You are on page 1of 302

Plant Evolution in the Mediterranean

This page intentionally left blank

Plant Evolution in
the Mediterranean
John D. Thompson
UMR 5175 Centre d’Ecologie Fonctionnelle et Evolutive,
CNRS, Montpellier

Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
São Paulo Shanghai Taipei Tokyo Toronto
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Oxford University Press 2005
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2005
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer
A catalogue record for this title is available from the British Library
Library of Congress Cataloging-in-Publication Data
Thompson, John D., 1959 Jan. 8–
Plant evolution in the mediterranean / John D. Thompson.
p. cm.
Includes bibliographical references and index.
ISBN 0–19–851533–2 (alk. paper) — ISBN 0–19–851534–0 (alk. paper)
1. Plants—Evolution—Mediterranean Region. 2. Plant ecophysiology—Mediterranean Region.
3. Acclimatization (Plants)—Mediterranean Region. I. Title.
QK314. 5. T49 2005
581. 3’8’091822—dc22
ISBN 0 19 851533 2 (Hbk)
ISBN 0 19 851534 0 (Pbk)
10 9 8 7 6 5 4 3 2 1
Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India
Printed in Great Britain
on acid-free paper by Antony Rowe Ltd., Chippenham, Wiltshire

My fascination for the Mediterranean, and its plants, Escarré, Rosylene Lumaret, and Max Debussche)
began when I arrived in Montpellier for a one-year provided me with informative and critical discus-
post-doctoral position in 1989. Since then I have sion of different aspects of functional and evolu-
become more and more interested in the ecological, tionary ecology in a Mediterranean context. Here,
genetic, and historical causes of diversity in the flora. the encouragement and advice of Jacques Blondel
My research has been focused on the ecology and was very important as my project to write a
evolution of just a few groups of plants in the flora, book unfolded. Next, the PhD and post-doctoral
which have allowed me to pursue my main interests students with whom I have interacted, hosted,
concerning how plant species respond to, and cope or supervised in one way or another (François
with, spatial variation in their environment. This Bretagnolle, Christophe Thébaud, Laurence Affre,
work naturally made me curious about the history of Michèle Tarayre, Christophe Petit, Thierry Pailler,
the region, its climate and its flora, including domes- Laurence Humeau, Anne Charpentier, Perrine
ticated and invasive species. In this book I have Gauthier, Angélique Quilichini, Sebastien Lavergne,
thus treated several subjects which are outside of Bodil Ehlers, Adeline Césaro, Justin Amiot, Isabelle
my primary research themes. I trust that I have dis- Litrico, and Emilie Andrieu). They have kept me on
cussed them to a level which satisfies others more my toes and allowed my research to progress. Then,
competent and knowledgeable than I in these fields. all the people who have helped with the plants them-
I did not fully realize what I needed to write this selves, particularly Christian Collin, Marie Maistre,
book until it was almost finished. First, I of course Annabelle Dos Santos, and Alain Renaux, the
needed a good story with a solid scientific basis. administrative support staff at CEFE and Geneviève
I trust I have supplied a text which is convincing. Debussche who kept my computer working. Finally
Second, you need endless motivation. My interest a particular thanks to Max Debussche, who first took
for the subject held me strong here. The third and me to see a wild population of Cyclamen. He has
final ingredient is the encouragement, help, and shared with me his interest in, and knowledge of,
support of my colleagues. That I managed to write the ecology and natural history of Mediterranean
this book attests to the excellence of the help and plants and has never been too shy to provide an alter-
encouragement I have received over the four years native explanation for something I had written. His
since I began this project. eye for detail and extensive knowledge have greatly
It is to the people in Montpellier to whom I intend contributed to improve my own understanding of
a special thanks. Their help has come in a variety plant evolution in the Mediterranean.
of ways. Bertrand Dommée, Isabelle Olivieri, and During the writing phase I received much posi-
Denis Couvet for their warm welcome in the late tive criticism and guidance. In this respect, I thank
1980s and their stimulating company as I settled into Gideon Rosenbaum who put me straight on the
a new chapter of my scientific life. Other colleagues geological history of the Mediterranean, Richard
at the Centre d’Ecologie Fonctionnelle et Evolutive Abbott for his knowledge of the phylogeography
(CNRS) laboratory in Montpellier (particularly José of Mediterranean taxa, Yan Linhart for his advice

vi P R E FA C E

on differentiation patterns, Spencer Barrett for his which interest those whose research bears on the
enthusiastic and constructive criticism of my dis- conservation of plant diversity in the region and
cussion of plant reproduction, and once again Max botanists who study and classify Mediterranean
Debussche for his attentive remarks. In addition, plants. Finally, I trust that this book will find an
many thanks to Geneviève Debussche who drew audience in the large community of naturalists and
several of the figures. I am also greatly indebted to all botanical society members who, always keen to see
the people who have provided me with figures, pho- a rarity or something new, have greatly improved
tos, and unpublished and published manuscripts, our knowledge of Mediterranean plants and their
and who have discussed with me and advised me distributions. I hope that the ecological and evolu-
about the plants they study. My apologies to those tionary themes I develop will thus stimulate people
whose work is not cited, I had to make a choice with a wide range of backgrounds as they pursue
in several places, and could not include every- their diverse interests.
thing. Finally, my thanks to Ian Sherman at Oxford I have cited a small number of general refer-
University Press for his encouragement, advice, and ences throughout the book to set the context for my
his endless patience. exploration of plant evolution in the Mediterranean
When I came to Montpellier I received a warm region and to lead the reader to recent key papers
welcome and entered into a stimulating atmosphere from which they can base a literature survey. I have
in which to work. Since then I have continually bene- also broadened my discussion where possible to
fited from the experience and advice of several close include examples from other Mediterranean-climate
colleagues who have shared their wide-ranging regions. To avoid littering the book with too many
knowledge of the Mediterranean environment and names, I have kept family names and species author-
its plants. There was just one thing missing: a book ities to the species list at the end of the book. I have
which introduced me to the evolutionary ecology of often included reference to my own unpublished
plants in the region. The absence of such a text was data and observations. My aim has been to cover
the primary reason motivating me to write this book where possible the extensive literature on Mediter-
on plant evolution in the Mediterranean. I hope that ranean plants and to extract and discuss in some
newcomers to the region and its flora will find in this detail a smaller number of case studies that I feel are
book a broad-based introduction to the evolution of particularly pertinent to the themes of the book.
plants in the Mediterranean Basin. I also hope that When sat in front of my computer screen, more
those researchers who know the region well will find than once my thoughts drifted back in time to my
something new and interesting. For those who have high school days when I first became interested in
never been, take this book as an invitation to come. plants through my mother’s passion for the plants
I have written this book with many people in in her garden and my Aunty Lynne who brought
mind. Therein lies my major problem: satisfying me the latest in biology textbooks. Thanks to their
the curiosity of different groups with very differ- encouragement, the fascination of a schoolboy for
ent backgrounds. I hope that there is something the natural world became a passion and a career.
new for molecular phylogeographers who study the Since starting to write this book it has been almost
evolution of distribution patterns in relation to cli- every day that Marie-Andrée has given me that
mate change and for plant population geneticists so-much-needed encouragement to keep on and
and ecologists interested in adaptation, plant repro- finish it.
duction, and the processes and consequences of John D. Thompson
landscape change. I also hope to develop themes Montpellier, June 2004

Introduction: Themes, structure, and objectives 1

1 The historical context of differentiation and diversity 10

1.1 Geology, climate, and human activities: the mould and sculptors of plant diversity . . . . . . . 10
1.2 A meeting of continents: a complex geological history . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3 Two seasons: the history of the climate and the vegetation . . . . . . . . . . . . . . . . . . . . . 18
1.4 Diversity and unity in the Mediterranean flora . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.5 Centres of diversity: concordance with history . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2 The biogeography and ecology of endemism 38

2.1 Narrow endemism: the cornerstone of Mediterranean plant diversity . . . . . . . . . . . . . . . 38
2.2 Endemism in the Mediterranean: patterns and classification . . . . . . . . . . . . . . . . . . . . 40
2.3 Endemism in the Mediterranean: community composition and biogeography . . . . . . . . . . 43
2.4 The biology and ecology of endemic plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3 The evolution of endemism: from population differentiation to species divergence 67

3.1 Endemism and evolution: the processes and scale of differentiation . . . . . . . . . . . . . . . . 67
3.2 Population variation in endemic plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.3 Climatic rhythms and differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4 Divergence in peripheral and marginal populations: isolation, inbreeding, and ecology . . . . 77
3.5 Hybridization and chromosome evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4 Trait variation, adaptation, and dispersal in the Mediterranean mosaic 109

4.1 Ecological constraints and adaptation in the Mediterranean . . . . . . . . . . . . . . . . . . . . . 109
4.2 Summer drought and nutrient stress: functional traits and their variability . . . . . . . . . . . . 111
4.3 The phenology of flowering and fruiting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.4 Dispersal and establishment: the template of local differentiation . . . . . . . . . . . . . . . . . 130
4.5 Variation and adaptation in aromatic plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

5 Variation and evolution of reproductive traits in the Mediterranean mosaic 167

5.1 Reproductive trait variation: the meeting of ecology and genetics . . . . . . . . . . . . . . . . . 167
5.2 Specialization and generalization in a mosaic pollination environment . . . . . . . . . . . . . . 168


5.3 Attracting pollinators . . . but avoiding herbivores . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

5.4 Mating system and gender variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.5 Pollination ecology and the evolution of style-length polymorphisms . . . . . . . . . . . . . . . 194
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

6 Ecology and evolution of domesticated and invasive species 207

6.1 Migration with man . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.2 The evolutionary history of domesticated plants . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.3 Invasive species in a Mediterranean environment . . . . . . . . . . . . . . . . . . . . . . . . . . 223
6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238

Conclusions: Endemism, adaptation, and conservation 240

Species list 246

Bibliography 253

Index 291
Introduction: Themes, structure, and

… areas of mediterranean climate afford not only repositories for relict plant families but great
natural laboratories for students of evolution …
P.H. Raven (1971: 132)

The primary goal of this book is to blend informa- Mediterranean Sea is reduced to shallow sills which
tion from diverse domains into a synthetic account further belie the history of land connections around
of evolutionary ecology in which the central theme the region. Almost all the way around its shores are
is differentiation, both among and within plant the mountains. This remarkable geology has been
species. To do so, I provide illustrations of principal instrumental in shaping patterns of plant species
evolutionary processes and emphasize the relative distribution.
roles of spatial isolation and ecological variation. The fundamental element of the Mediterranean
The central theme is developed by highlighting region is its highly seasonal climate. The essential
how population-level processes not only provide the and defining characteristic of this seasonality is that
template for differentiation but also the stimulus for the warmest season is associated with an effective
species evolution and by firmly setting trait vari- drought which limits plant growth. Although the
ation and evolution in the context of spatial habitat length and intensity of summer drought vary
variation. spatially, and its onset is fairly recent, the occurrence
The subject material of this book is the contem- of this climatic regime has had fundamental implica-
porary flora of the Mediterranean Basin. This flora tions for the ecology and evolution of plants in the
inhabits a region with a complex history and a highly region. Since the initial onset of the Mediterranean
heterogeneous landscape. The evolution of plant climate, many parts of the region have acted as
diversity in this flora has been greatly influenced by a refuge during periods of Quaternary glaciation.
its geological history, the oscillations of the climate, Climatic oscillations caused plant species ranges
and the impact of human activities. to contract and then to expand again as the cli-
On a map of the world, one can see the Medi- mate warmed. These oscillations opened the way
terranean, not just as an inland sea, but more as a for hybridization and evolution in new environ-
region where continents meet. The complex geo- ments and have been fundamental for patterns of
logical history of this meeting has decorated the differentiation and diversification in many groups
Mediterranean Sea with islands, which vary from of plants.
tiny fragments of previous land-bridge connections The Mediterranean is also the home of many
which barely keep their heads above water, to human civilizations. Human activities have been
the big islands with their massive mountains and modifying natural habitats and the spatial dis-
violent volcanoes. Plunging to vast depths in the tribution of species for thousands of years and
centre of its diverse basins, in many places the have thus played a key role in shaping recent


and contemporary evolutionary pressures in natural This unity has both a spatial and temporal context.
populations. The impact of human activities stems To delimit a biogeographic region, and thus its
from their effects on both the ecological condi- flora, one has to have reliable boundary lines. The
tions within habitats, which shape natural selection iso-climatic area proposed by Daget (1977a, b) is
pressures and adaptive variation, and the spatial fairly well accepted but actually extends away from
configuration of habitats in the landscape, which the Mediterranean Basin to the Canary islands,
determines gene flow and seed dispersal. By modi- south into sub-Saharan Africa, south-east into
fying the action of selection and gene flow, human Arabia, and north-east into other parts of western
activities have become a key element of the process Asia. At the other end of the extreme, classifications
of population differentiation. based on the distribution of particular species,
Mayr (1982) recognized that the interests of evolu- such as olives, or the distribution of sclerophyllous
tionary biologists range from those whose primary vegetation, all fall short of a true estimation of the
interest lies in the study of diversity (speciation in spatial extent of the Mediterranean region. In accor-
fact) and those for whom ‘adaptation … holds first dance with previous studies (Quézel and Barbero
place in their interest’ (p. 358). These two themes, 1982; Médail and Quézel 1997; Quézel and Médail
diversity and adaptation, provide the frame- 2003), I use a delimitation of the Mediterranean
work for my discussion of plant evolution in the region which falls between these different extremes
Mediterranean. My purpose is to firmly place the and which essentially covers the region where an
evolutionary processes which shape plant evolution effective drought occurs in the warmest part of the
into the context of the three main historical influ- year (Fig. I.1).
ences on vegetation in the Mediterranean region, The critical defining characteristic of the Medi-
that is, geology, climate, and human activities. I thus terranean region is thus that summer is the driest
attempt to write a story of plant evolution in the season, and that this dry season involves a period of
context of regional history. drought, that is, is biologically dry (Emberger 1930c;
To write about the evolution of plants that inhabit Quézel 1985). The high mountains that fringe the
the lands around the Mediterranean Sea requires shores of the Mediterranean and dominate many of
the conception of a certain unity which holds its islands as well as some of the steppe formations
together the immense diversity present in the flora. that spread across the Anatolian peninsula and large

500 km

Figure I.1 The delimitation of the Mediterranean region (redrawn from Quézel and Médail 2003).

parts of north-west Africa clearly have their place component to their territory, a large fraction of
here (di Castri 1973; Quézel and Médail 2003). On all their plant species occur in the Mediterranean
the summits of the Moroccan Atlas or the Taurus part of their territory. Even countries with a small
mountains of Greece, or in the Sierra Nevada of percentage of their territory in the Mediterranean
southern Spain, summer temperatures are not par- region have a high percentage of their total species
ticularly high, however, there is a prolonged dry complement which is present in the Mediterranean
period at this time of the year. Hence, the flora of region (Fig. I.2). For example, although only ∼10%
such areas can be considered to be ‘unequivocally of of the territory of continental France occurs in
a Mediterranean type’ (Quézel and Médial 2003: 25, the Mediterranean region, 66% of all species that
my translation). Some of the examples which I use occur in France occur in the Mediterranean zone.
occur on the margins of this geographic delimitation, One single administrative ‘département’ of southern
but in ecological settings which closely resemble France (l’Hérault) contains 2,400 species of vascu-
those within the strict confines of the Mediterranean- lar plants, that is, 55% of the flora of continental
climate region. It is my belief that species which France in just 1.1% of the total surface of the coun-
inhabit such peripheral Mediterranean areas, in par- try. Even in Mediterranean forests, where endemism
ticular those whose distribution crosses the border is low, woody species richness is twice that of
into more temperate, continental, or desert cli- temperate European forests (Quézel and Médail
mates provide ideal situations for the study of plant 2003). In addition, ∼60% of the native species in
evolution in the Mediterranean. the Mediterranean flora are endemic to the region
A melting pot of geological activity, climatic (Quézel 1985; Greuter 1991) making it one of the
evolution, and human civilizations, the Medi- world’s ‘hot spots’ of species diversity (Myers et al.
terranean Basin is a hot spot of plant biodiver- 2000).
sity. The flora of the Mediterranean Basin contains In Chapter 1, I build a framework for our present
∼24,000 plant species in a surface area of about 2.3 understanding of the biogeographic origins and
million km2 (Greuter 1991), that is, 10% of known diversity of the Mediterranean flora. To do so,
plant species in really what is only a small part of the I reconstruct the geological history of connections
world. In contrast, non-Mediterranean Europe cov- and isolation among different land masses and
ers about 9 million km2 but only has around 6,000 islands, the development of the Mediterranean-type
plant species. In 17 countries with a Mediterranean climate that currently reigns, and the history of
human activities in the last few millennia. These
three regional features will be used to structure
100 my discussion of plant evolution throughout
the book.
80 The Mediterranean flora shows extremely high
rates of narrow endemism in many regions,
% species

particularly in the mountains and on islands

(Greuter 1991; Médail and Quézel 1997). This nar-
row endemism is a key ingredient of plant bio-
diversity in the Mediterranean flora, and also the
other Mediterranean-climate regions where ecolo-
20 gical specialization and geographic isolation have
0 20 40 60 80 100
% territory
been primary determining factors (Cowling et al.
1996). A primary question motivating Chapters 2
Figure I.2 Percentage of land area and species richness in the
and 3 concerns the role of regional history and
Mediterranean part of the territory of 17 countries on the shores of the spatial environmental variation in the evolution
Mediterranean Sea (drawn from data in Médail and Quézel 1997). of endemism. In the Mediterranean, narrow

endemism often involves disjunct distributions effects, due to constraints on their action in associ-
among closely related taxa, creating an ideal set- ation with environmental variation. For example,
ting to link the study of population differentiation zones with deeper soils would have been the first to
with that of species divergence (Thompson 1999). In be cultivated as domesticated plants were dispersed
Chapter 2, I describe and explore the biogeography across the Mediterranean region. As a result, species
of endemism in the Mediterranean flora and assess more prevalent in open rocky habitats, and in and
whether ecological characteristics and biological around cliffs, may have been more persistent as
traits are associated with narrow endemism. In human activities developed and spread.
Chapter 3, I illustrate the diversity of evolutionary I will thus insist repeatedly on the importance of
processes acting on variation at the population and spatial heterogeneity. In a quantitative classification
species level in relation to the history of the region of the Mediterranean mosaic, Dufour-Dror (2002)
described in Chapter 1. To explore and assess the recognized 55 different vegetation types in Israel
history, ecology, and evolution of species divergence of which 30 were present in the Mt Carmel region.
and endemism I insist on the need to link population In Turkey, steppe vegetation occurs on an immense
differentiation to species divergence. diversity of substrate types (limestone, gneiss, ultra-
In Chapters 4 and 5, I switch to the theme basic rocks, and schists) creating a myriad of selec-
of trait variation and adaptation in a spatially tion pressures on plant populations within this
heterogeneous environment. The focus here is on type of vegetation. As Quézel and Médail (2003)
the ecological and historical factors which deter- point out, in such situations, the nature of the
mine trait variation and evolution and the ecological substrate plays a primary role in the composition
basis of adaptation in the highly heterogeneous and dynamics of local plant communities. Substrate
Mediterranean mosaic environment. may also contribute to patterns of endemism and
The Mediterranean region is where landscapes the immense diversity of the flora. Even on single
vary dramatically, often over short distances, with islands (Fig. I.3(a) and (b)) the combination of geo-
perhaps the most original and fascinating aspect of logical variation and altitude, along with strong
this spatial variation being the mosaic-like aspect of climatic variation among different slopes can create
the vegetation in the landscape. Anybody who has marked heterogeneity in the ecological forces acting
hiked across the Peloponnese peninsula in southern on the evolution of plant diversity. Such variation
Greece, the Sierras of southern Spain, or one of the can also occur on a highly localized spatial scale in
big islands such as Crete or Corsica will appreciate continental regions, such as the landscape depicted
this point. Such landscapes and islands contain a in Box I.1. Here the geology varies dramatically
diversity of ecosystems which harbour rich floras over a few kilometres, creating a mosaic of substrate
with striking local variation in community structure types. Local variation in soils (which are deeper and
and the presence of individual species: mountain more humid in the sedimentary basin), climate (the
pine forests, pungent arid garrigues and phrygana, sedimentary basin is a frost hollow in winter), and
humid canyons, deep and dense oak forests, human activities further accentuate spatial variation
savannahs where nothing moves out of the shade in ecological conditions across this landscape, where
of isolated trees in mid-summer, … the list goes on. genetic differentiation among populations of com-
Sharp cliffs, deep gorges, vast sedimentary basins, mon species has been reported (Chapter 4). This
and meandering rivers all multiply the effects of is but one example of localized mosaic vegetation,
substrate diversity and climatic stress. Human activ- which is common to many Mediterranean land-
ities have further added to this spatial complexity, scapes (Fig. I.4). The point here is that tectonic
reinforcing environmental variation in a landscape activity and edaphic variation have created a tem-
already structured by spatial heterogeneity of geo- plate for plant evolution, which has been further
logy, soils, and climate in many areas. In some zones modulated by climate and then more recently by
human activities have varied dramatically in their human activities.


1. Devonian: schist and sandstone

2. Lower Trias: silaceous
sandstone, clays, and conglomerates
3. Upper Trias: calcareous and
dolomitic limestone
4. Jurassic: limestone
1 5
5. Cretaceous: limestone
2 6
6. Miocene: limestone and
3 7
conglomerates 4
7. Dunes
10 km

1. Quarternary alluvium
2. Miocene limestone
3. Conglomerates and sandstone
(Jurassic and Pliocene)
4. Schist, serpentine, and
volcanic rocks
2 5. Granite

3 1


40 km

Figure I.3 The mosaic of substrate types on individual islands as illustrated by the simplified geology of the islands of (a) Minorca and (b) Corsica
(redrawn from de Bolòs and Molinier 1970 and Gamisans 1999, respectively).

Box I.1 Localized geological variation around the Pic St Loup in southern France
(based in part on an original figure, modified with permission, in Bousquet 1997)

Early Jurassic Late Jurassic Sedimentary Basin Late Cretaceous

(Mortiès depression) (Pic St Loup) (St Martin-de-Londres Basin) (Hortus cliffs)

50 Ma Late
135 Ma
180 Ma Early
Jurassic North
205 Ma Late Trias 2 km

The history and variety of geological formations than the surrounding higher elevation lands. The
and orogenic movements in this area of southern deeper and better water retention capacity of soils
France have been such that the Late Jurassic in the basin have promoted agricultural
limestone of the Pic St Loup (left) stands ‘face to exploitation of many areas, modifying the spatial
face’ with the Late Cretaceous cliffs of the Hortus configuration and size of semi-natural habitats in
(right). Subsidence and erosion has revealed the the landscape.
black marls of the Early Jurassic to the south of the The mosaic landscape is thus shaped by the
Pic St Loup, in a basin with a small area of Early interplay of geology, climate, and human activities.
Jurassic limestone. To the north and east a In fact, as Lepart and Debussche (1992) illustrate,
sedimentary basin (the St Martin-de-Londres Basin abiotic spatial heterogeneity has been a major
discussed in Chapter 4) rich in marine fossils covers determinant of the nature and impact of human
large expanses. This sedimentary basin is in a frost activities on natural habitats in the Mediterranean
hollow where the winter climate is much colder region.

The evolution of plants in such a spatially het- by modulating pollen and seed dispersal among
erogeneous landscape requires an understanding patches of favourable habitat. Spatial heterogene-
of regional history and the ecological differences ity is thus at the heart of my discussion of plant
which occur at a variety of spatial scales. The habitat evolution in Chapters 4 and 5 of the book. This
mosaic is associated withsharp and local vari- discussion is strongly motivated by my convic-
ation in selection pressures and regulates gene flow tion, that to understand plant evolution requires an




Figure I.4 Mosaic habitat variation in the Mediterranean: (a) shrub–woodland interface in the Sierra de Cazorla in southern Spain (photo kindly
supplied by C. Herrera), (b) Pinus brutia forest pocket in the mountains of Crete, (c) oak woodland, limestone cliffs, scree slopes, and open
garrigue vegetation adjacent to cultivated fields and abandoned cultivated areas in southern France.

understanding of how spatial variation in ecological vascular plant species (Cowling et al. 1996). They
processes regulates dispersal, thus creating a tem- also contain a large number of endemic species
plate for differentiation, and how trait variation and show strong patterns of localized or regional
influences establishment and reproduction, and differentiation. The similarities and differences of
thus affects long-term population dynamics and vegetation in these Mediterranean-climate regions
evolution. (Dallman 1998) have given rise to much interest in
The central theme of Chapters 4 and 5 con- the possible convergent evolution of vegetation and
cerns variation and adaptation within and among traits in these different regions. Rather than write
local populations in relation to the environmen- a single chapter on the comparative ecology and
tal constraints and selection pressures that natural evolution of floras in the different Mediterranean
populations encounter in the Mediterranean mosaic regions of the world, I have repeatedly broadened
landscape. The emphasis is thus on intraspecific my discussion to compare patterns with those in the
variation. In Mediterranean-climate regions, plant other Mediterranean-climate regions. In Chapter 2,
form and function has traditionally been interpreted I extend my exploration of the biology and eco-
as a response to climatic and edaphic constraints. logy of endemic plants in the Mediterranean to
I thus explore and evaluate evidence of adaptive trait encompass patterns observed in South Africa and
evolution in this context. A key issue I develop is Australia. In Chapter 3, the importance of cli-
that local populations occur in habitats that are part mate change in different Mediterranean regions
of a highly heterogeneous mosaic of environmental for the evolution of endemism is discussed. In
variation in the landscape. Population differentia- Chapter 4, I discuss variation in the occurrence of
tion is a balance between the local ecological and traits associated with sclerophylly and resprouting
population processes acting on the genetic varia- ability in different Mediterranean-climate regions.
tion in a habitat patch and the regional processes In Chapter 6, I compare patterns and processes of
which determine gene flow and migration among invasion in different Mediterranean regions. My
patches. I address this issue in Chapter 4 where purpose has not been to provide a comprehensive
I discuss functional trait variation, adaptation, and comparative examination of convergence in differ-
dispersal patterns, and in Chapter 5 where I dis- ent Mediterranean regions but to illustrate those fea-
cuss the ecology, spatial dynamics, and evolution tures of plant evolution in the Mediterranean flora
of reproductive strategies. which are common to other Mediterranean-climate
Then, in Chapter 6, I discuss the ecology and regions and those which are more closely tied
evolution of species whose distributions have been to specific aspects of the regional history of the
modified as a result of human-induced dispersal. Mediterranean Basin. I have thus selected specific
The focus of this final chapter is evolution under examples from other Mediterranean-climate regions
domestication and cultivation and the population to illustrate general patterns. In addition, I have
ecology of invasive species. By moving plants included sections which gives each chapter a broad-
around the Mediterranean Basin, and into and out based conceptual framework.
of the region, humans have not only created exas- Finally, the richness of endemic plants in the
peratingly complex problems for the conservation Mediterranean has lured botanists into the region
of differentiation diversity but have also set up for centuries. In more recent years, the popula-
experimental populations ready for the study of tion ecology and genetics of Mediterranean plants
plant evolution in a new environment. have received growing attention, with much inter-
The Mediterranean Basin, along with parts of est directed towards understanding the ecology
south-western Australia, the south-western Cape of and evolution of natural plant populations. There
South Africa, western California, and central Chile now exists a large body of information concerning
is one of five Mediterranean-climate regions of the various aspects of the biology of Mediterranean
world. These five regions of the world only occupy plants. In this book my aim is to draw together such
∼5% of the land surface but harbour 20% of known information in a general synthesis of evolutionary

ecology in which evolutionary processes are dis- context of the conservation of endemic plants in the
cussed in relation to regional history. To conclude, region. I argue that more emphasis should be placed
I recapitulate some of the main themes and issues on conservation strategies which explicitly integrate
of plant evolution in the Mediterranean within the ecological processes and evolutionary potential.

The historical context of

differentiation and diversity

The mountains and basins of the Mediterranean have been called the Enigma Variations of tectonic
geology. Certainly it is a symphony of the earth that is not easy to understand.
J.M. Houston (1964: 51)

northern and southern shores. Trapped in a collision

1.1 Geology, climate, and human
zone between the African and Eurasian plates, the
activities: the mould and sculptors of
Mediterranean Sea has only one narrow natural out-
plant diversity
let, via the Straits of Gibraltar, which provides an
To provide a framework for my discussion of plant exchange with the oceans outside. This setting rep-
evolution, I have outlined what I consider to be resents one of the most geologically complex areas of
the three dominant factors which have been instru- the world and a unique example of a sea surrounded
mental in shaping the evolutionary forces acting on by different continents. It is in this context that
plant variation and diversity in the Mediterranean plants have diversified. The geological complexity
region in a historical triptych (Box 1.1). Geolo- has untold ramifications for our understanding of
gical and climatic histories have greatly impacted the origins of the flora in the Mediterranean Basin
on species distributions, isolating individual popu- and provides a fascinating setting for the study of
lations or localized groups of populations and bring- plant evolution. As Oleg Polunin (1980: 1) pointed
ing into reproductive contact previously isolated but out in the opening sentence of his book on the flora of
closely related taxa. Human activities have modified the Balkans, ‘The geological history of the Balkans is
selection pressures and the potential for pollen and perhaps the most important single factor contribut-
seed dispersal across the landscape. Although these ing to the diversity of the present-day flora’. I will
three factors are presented in separate panels, in real- illustrate how similar statements could be made for
ity there are no sharp boundaries. Indeed, I will other regions around the Mediterranean Sea.
emphasize throughout this book that plant evolu- Second, the Mediterranean region has a charac-
tion has been greatly influenced by the interaction teristic climate with two main seasons. The essen-
among the three different elements of this triptych. tial characteristic of this climate is the occurrence
First, the Mediterranean region has a complex of hot and dry summers which impose an effect-
geological history. The Mediterranean is the largest ive drought on the plants. There is also a cool
inland sea in the world. From Gibraltar in the west, or cold moist season in which unpredictable and
the Mediterranean Sea stretches eastwards for just often intense rainfall events occur from autumn
over 3,500 km. Its width is highly variable: whereas through spring. Close to sea, this season is mild
∼750 km separate the south of France from Algeria, compared with inland, where freezing temperatures
in some places, such as across the Straits of commonly occur in winter. As I discuss later in this
Gibraltar or where Italy sleeks down via Sicily chapter, the relative length of the summer drought
towards Tunisia, only a few kilometres separate the and the amount and timing of rainfall in the moist

H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 11

Box 1.1 Historical triptych

What are considered in this book to be the three the Mediterranean are depicted here in the panels
key factors which have modulated the action of of a historical triptych. Each panel illustrates the
the main processes impacting on plant evolution in timing of some important events.

Human activities Geological history Climate

Last 150 years Pliocene (∼1.5–2 Ma) Last 100 years

Abandonment of traditional Alpine orogeny Gradual warming
rural land-use and (uplift and folding) Holocene
reforestation on northern shores Miocene (5.3–6 Ma) Increased aridity
Holocene The Messinian crisis 12,000 BP
Major forest clearance <25 Ma Late glacial warming
Prior to 8,000 bp Migration of Cyrno-Sardinian Quaternary
Early harvesting and microplate The major Pleistocene
cultivation of the wild relatives Oligocene (25–30 Ma) glaciations
of domesticated cereals, Alpine orogeny Pliocene (∼2–3 Ma)
legumes, and fruit trees Jurassic/Cretaceous Onset of the Mediterranean
in the Near East Apulian plate contacts Europe. climate
Rotation and north-east Middle Miocene (∼15 Ma)
migration of African plate Mild seasonal climatic
contrasts begin to develop
Early Miocene and
beyond . . .
Subtropical conditions

season show much spatial variation around the on the landscape has been dramatic in terms of the
Mediterranean Basin. It is the alternation of these spatial configuration and size of natural habitats.
two seasons which unifies the region, its land- Such impacts have created new opportunities for
scape and its flora. The climate we now experi- colonization in some species and caused others to
ence is a relatively recent phenomenon and has retract into isolated patches. As a result, human
oscillated repeatedly. Its evolution and oscillations activity should be viewed as an integral ecological
have, within the constraints of land connections and feature of the Mediterranean scene, modifying not
dispersal limitation, caused species’ range sizes only the spatial configuration of habitats in the land-
to contract and expand at repeated intervals. As scape, with consequent effects on gene flow and the
some species disappeared from the landscape, potential for differentiation, but also the local selec-
others expanded their range. Plant diversity tion pressures and constraints that determine plant
in the Mediterranean tells this tale of climate establishment, persistence, and evolution.
change. This chapter traces the history of the Medi-
Third, nowhere else in Europe has there been terranean flora in relation to the three factors pre-
such a long history of human presence and activity. sented in Box 1.1. My objective is to lay the
Harvesting, cultivation, and domestication began foundation for my subsequent exploration and
early, particularly in the eastern Mediterranean. evaluation of patterns of differentiation and diver-
Since the Neolithic, the impact of human activities gence in Chapters 2 and 3.

1.2 A meeting of continents: a complex (a) ~180 Ma

geological history EU

The Mediterranean has been fashioned by the meet- Subduction

ing of Eurasia and Africa. The precise geological
history of the Mediterranean is far from being com- IB
pletely understood, and the account I give in this AP
chapter attempts to synthesize (largely from the axis
geological literature) the extent of current knowl-
edge, some of which remains hypothetical since
some interpretations require further confirmation.
My impression is that our understanding of the geo-
(b) ~150 Ma
logical history of the Mediterranean is at a stage
similar to that of an evolutionary biologist staring at Subduction
a molecular phylogeny of a large genus based on one zone
or a small number of gene(s). Although the frame-
work of the tree is no doubt close to its true form, Accretion
IB axis
several species may not be in their correct clades. AP
This analogy should be kept in mind as one reads
through my interpretation of geological history.

1.2.1 From ancient Tethys to a series of basins

(c) ~70 Ma Subduction zone
The Mediterranean Sea has an ancestor named
Tethys, whose history is complex. Most evidence EU
points to the existence of an equatorial ocean, or
Paleotethys, between the northern and southern AR
continents of Pangea during the Triassic Accretion
IB axis
(Maldonado 1985). This ocean was wedge-shaped,
open to the east and closed to the west, where a
Hercynian continent linked what is now Africa to AF
north-western Europe (Fig. 1.1). Paleotethys closed
in the early Mesozoic due to the overall northward
Figure 1.1 The ancient Tethys and the historical positions and
motion of continental blocks that rifted away from movements of microplates during the development of the
Gondwana and collided with Eurasia (Sengör Mediterranean. AF: African plate, AP: Apulian microplate, EU: European
1979). This produced Neotethys, a Permian to plate, AR: Arabian plate, IB: Iberian microplate. Arrows represent plate
Jurassic ocean that is widely known from remnants movements (redrawn from figures in Biju-Duval et al. 1976).
of oceanic crust (ophiolitic structures) now found
in the Alpine Mediterranean belt. The whole of Atlantic Ocean in the Early and Middle Jurassic,
the eastern wedge of Tethys began to disappear as that is, at 165 Ma (used to signify ‘Mega Annum’
a result of subduction during the early Mesozoic as in the geological literature, this abbreviation gives
the ‘Eurasian’ continent spread. us a timescale in millions of years), Eurasia and
The configuration of the ancestral Mediterranean Africa began convergence motion which was to
Sea, a series of closed basins in the Oligocene and shape the early formation of the Alps and the
Miocene, was thus closely related to the struc- Mediterranean Basin (for details of what follows see:
tural relationships between the major tectonic belts Biju-Duval et al. 1976; Dewey et al. 1989; Rosenbaum
of Africa and Eurasia. With the opening of the et al. 2002b). During the Late Jurassic and Early
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 13

Cretaceous (170–120 Ma), the two plates showed associated with Alpine orogeny which occurred
left-lateral strike-slip motion and ∼200 km of dis- in two main periods. The first occurred in the
placement (Fig. 1.1(b)). Then, in the Cretaceous Cretaceous and Early Tertiary when compression
(120–80 Ma), plate convergence brought Africa and and mountain building produced the initial socle of
Europe closer together (Fig. 1.1(c)) and Alpine oro- mountains in many areas. The second followed later
genesis began. Collision may have commenced in in the Pliocene and Pleistocene and involved verti-
the Early Tertiary (∼65 Ma) although this remains cal uplift and fracturing. Quaternary processes were
unsure (Rosenbaum et al. 2002b). After a period thus important elements in the fashioning of the cur-
of relative quiescence, more convergence occurred rent day landscape both in the western (Houston
during the Eocene and Early Oligocene (55–45 Ma) 1964) and eastern (Zohary 1973) Mediterranean.
as Africa rotated by more than 50◦ relative to Europe,
swinging from a north-east/south-west tilt to its
1.2.2 Micro-plate configuration: dispersal and
present position, face to face with Europe. The
African plate is still moving.
The approach of the European and African Since at least the Tertiary, and perhaps during the
plates produced two characteristic features of the earlier stages of Alpine orogenesis (G. Rosenbaum,
Mediterranean landscape. University of Mainz, personal communication),
First, the Mediterranean Sea contains a series microplate individualization and dispersal have
of deep basins bordered by relatively shallow sills played a major role in the tectonic evolution of the
(Fig. 1.1; Box 1.2). The different basins are small Mediterranean Basin (Alvarez et al. 1974; Biju-Duval
in terms of their surface area, and the continental et al. 1976; Rosenbaum et al. 2002a, 2004). The
margins (i.e. the transition zone between continen- three most well studied are the Iberian microplate,
tal and oceanic crust) cover more than 100 km in Adria (or Apulian microplate comprising Italy, the
many areas and less than 10% of the surface of the Balkans, and Greece), and the Cyrno-Sardinian
different basins lies more than 100 km from the con- microplate.
tinental plateau. The different continental margins The Iberian microplate occupied a key position in
thus touch each other, adding to the unity which the geological evolution of the Mediterranean Basin
makes up the contemporary configuration of the due to its position at the western extremity of the
Mediterranean region. The formation of the different contact zone between the African and European
basins has been closely associated with the config- plates (Fig. 1.1). The geological evolution of this
uration of adjoining land masses and, along with region exhibits a complicated interplay of orogenic
its almost lack of any tidal regime (except in some processes and plate movements (Rosenbaum et al.
restricted areas) and its high salinity, is one of the 2002b). The Iberian microplate was initially attached
principal characteristics of the Mediterranean sea. to Europe, albeit further west than at present. The
Second, the Mediterranean region has many movement of the African plate pushed it north-
mountains. For example, the Atlas Mountains eastwards from the Late Jurassic to the Late Cre-
of North Africa (geologically speaking: the Rif taceous (∼70 Ma), causing the uplift of various
and Maghrebides), the Sierra Nevada (or Betic mountain ranges, notably the Pyrenees.
Cordillera), the Pyrenees, Appenines, Dinarides, The Apulian plate or Adria represents the con-
Taurus, and Anatolian chains and Mt Liban all tinental crust bridging the continental masses of
form an imposing backdrop. In some areas these Africa and Eurasia across the central Mediterranean
mountains drop directly into the sea, while in other where it separated the eastern and western basins
parts of the Mediterranean Basin the transition is (Rosenbaum et al. 2004; Fig. 1.1). This plate was
more gradual through low hills and a coastal plain. centred on what is now the Adriatic Sea. Connected
Centres of diversification, many of the mountain- to the African plate, perhaps as a promontory rather
ous areas represent hot spots of endemism (see than a detached fragment, this microplate came into
later in this chapter). Their formation was closely contact with the southern part of the European plate.

Box 1.2 Geological history of the basins under the Mediterranean Sea

The Mediterranean Sea is subdivided into islands are poised on these sills. In the east, the
individual basins (black) separated by shallow sills, Gibraltar sill maintains a degree of isolation from
some of which represent ancient arcs the Atlantic, with implications for sea currents and
of mountains now under the sea (dark grey). Several the climate of the Mediterranean region.



The different basins contain the different extensional rifting and subduction (Cherchi and
localized seas within the Mediterranean, that is, Montadert 1982; Mascle and Rehault 1991;
the Tyrrhenian, Alboran, Ionian, Adriatic, and Robertson and Grasso 1995).
Aegean Seas. The Aegean Sea developed in the Pliocene and
The Alboran Sea (AS) probably opened during Quaternary (Maldonado 1985; Robertson and
the north-westward movement of the Apulian Grasso 1995), primarily as a result of northward
microplate (Maldonado 1985). Collision of Africa subduction, continued back arc extension and
and Europe led to the formation of a volcanism. This development, like that of the
south-western Mediterranean microplate (Araña western Mediterranean, involved a reconfiguration
and Vegas 1974). At this time the Alboran Sea of ancient rocks.
was, with the Betic Cordillera, above sea level. The Ionian Basin (IB), which plunges to 5000 m,
Continued westward movement of the microplate and the Levant Basin (LB) are the only areas in the
caused the formation of the Gibraltar arc as the Mediterranean region where remnants of ancient
plate was over-thrusted on collision with the Neotethys oceanic crust underlie sea floor
Atlantic continental margin. The Calabrian and sediments (Rosenbaum et al. 2002a). The shallow
Hellenic arc formations probably formed in this submarine sill linking Calabria, Sicily, and Tunisia
way (Biju-Duval et al. 1976; Maldonado 1985). (at depths of <600–700 m) which subsided at the
The Tyrrhenian ‘back arc basin’ (TB) was more end of the Tertiary, represents an important
recently created (in the Late Miocene) by north–south historical connection.

Analysis of paleomagnetic, geophysical, and geo- Adria and Africa involved the opening of the Ionian
logical data point to a relatively coherent motion of Sea perhaps as early as the Permian. Movement of
Adria and the African plate since the Jurassic, albeit the African plate and the collision of its Arabian
with some independent rotation. The detachment of margin with Europe caused the Mediterranean Sea
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 15

to become closed in the east, and the rotation of the opening to the north appeared and then after the
Iberian microplate allowed for only a small outlet in Messinian (Fig. 1.2(d),(e)) an opening to the south
the west. developed. Corsica, Sardinia, and the other frag-
The history of the Cyrno-Sardinian microplate (see ments of the initial microplate have thus been iso-
Rosenbaum et al. 2002a) is critical to our under- lated from the Balearic islands and southern France
standing of endemism in the western Mediterranean since the Miocene. The Balearic islands have how-
(as we shall see in Chapter 2). In the Late ever had repeated connections among each other
Oligocene (35–30 Ma), a Hercynian massif con- (Minorca and Majorca had their latest connection in
nected the Pyrenees to the outer crystalline massifs the Pleistocene) and with the Iberian peninsula.
of the Maures—Esterel, and ultimately the Alps As Corsica and Sardinia rotated south-eastwards
via what are now the cliffs on the north-east tip during the Miocene, the Kabylies broke away
of Minorca, Corsica, and Sardinia (Fig. 1.2(a)). The from the Balearic islands in a southerly direction
latter two islands were then part of a continental (Fig. 1.2(b),(c)) and collided with the African margin
environment, with Corsica 30◦ and Sardinia a lit- (Fig. 1.2(c)). During this period, the Betic Cordillera
tle over 60◦ north-west of their present position and became separated from the eastward migrating
orientation (Hsu 1971; Westphal et al. 1976; Cohen fragments and accreted (∼10 Ma) to the south-
1980; Cherchi and Montadert 1982). Corsica and the eastern tip of Spain (Fig. 1.2(b)–(d)). Calabria and
Esterel (now part of continental France) were con- north-east Sicily continued their rotation till the
tiguous (Cohen 1980). Based on their geological sim- Pliocene when they arrived in their present position
ilarity, Alvarez (1976) postulated that north-eastern (Fig. 1.2(d),(e)).
Corsica, Calabria, the Kabylies (in North Africa),
and the Betic Cordillera were also linked to one
another in an Alpine Belt that extended around the
1.2.3 When the Mediterranean salted up
southern edge of the Hercynian massif (Fig. 1.2(a)).
In the Late Oligocene (Fig. 1.2(b)), from an initial In 1961, a newly developed type of echo-sounding
closed position against southern France and north- used by oceanographers produced what was then
west Italy, this microplate began to rotate south- a startling finding: the Mediterranean Sea floor
eastwards (Alvarez 1974; Rosenbaum et al. 2002a). is underlain by an array of pillar-like structures.
The dispersal and fragmentation of the Tyrrhenian Some of these exceed several kilometres in diameter,
islands and Calabria on a single microplate prob- reach 1,500 m in height, and protrude as knolls on
ably started due to the rifting-off of the European the sea floor. Today, some are exposed on land,
continental margin to produce the Cyrno-Sardinian the best examples being on Sicily (Robertson and
microplate (Cherchi and Montadert 1982; Robertson Grasso 1995). The resemblance of these structures
and Grasso 1995). The Balearic Basin opened behind to salt-domes put geologists onto the idea that vast
the rotating microplate. Once Corsica collided with salt deposits are currently hidden beneath the floor
the crust of the northern Appenines (∼20 Ma) it of the Mediterranean. This was the first hint that
could no longer rotate (Fig. 1.2(c)). As a result, the in the Messinian stage of the Late Miocene, the
depression where the Straits of Bonifacio now occur Mediterranean dried up (Hsü 1972; Hsü et al. 1973,
opened as Corsica became separated from Sardinia 1977; Cita 1982; Duggen et al. 2003). This ‘Messinian
and Calabria, which continued to rotate towards the salinity crisis’ is now known to have begun at
south-east. According to Alvarez (1974) the rotat- 5.96 Ma (Krijsman et al. 1999).
ing plate collided with the Tunisian margin of North In the Late Miocene (∼8 Ma), marine passages
Africa at ∼14 Ma, a collision which stopped the rota- in southern Spain and northern Morocco linked the
tion of Sardinia. In the Middle Miocene, Sardinia Mediterranean Sea to the Atlantic Ocean. Although
became separated from Calabria and north-east a global drop in sea level occurred at about this time,
Sicily. The opening of the Tyrrhenian Sea occurred this marine gateway from the Mediterranean to the
in two stages. From 9 to 5 Ma (Fig. 1.2(c),(d)) an Atlantic was probably closed as a result of uplifting

(a) (b)
Oligocene (30 Ma) Late Oligocene (25 Ma)

500 km 500 km

(c) (d)
Middle Miocene (15 Ma) Tortonian (10 Ma)

500 km 500 km

(e) (f)
Messinian (6 Ma) Late Pliocene (2 Ma)

500 km 500 km

Figure 1.2 Historical reconstruction of the history of the different islands and continents in the western Mediterranean since the Oligocene.
(reproduced with permission from Rosenbaum et al. 2002a).

along the African and Iberian plate margins in asso- a disjunct mosaic of large lakes in which thick and
ciation with mantle processes (Duggen et al. 2003). extensive evaporites precipitated, particularly in the
Marked regional aridity led to high levels of evap- deepest parts of the basins. The presence of fos-
oration from the closed Mediterranean Sea, which silized remains of light-demanding cyanobacteria,
led to a basin-wide lowering of sea level as the other which usually develop in shallow water, indicate
sea levels lowered. The Mediterranean Sea became that although the precipitation occurred within
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 17

the confines of deep basins, it occurred in rela- present configuration, with an extension of coastal
tively shallow water (Hsü et al. 1973). As these areas as sea levels declined during the different
authors discuss, the depth of the evaporite in Quaternary glaciations. For example, sea level was
some areas could not have occurred from a sin- ∼150 m lower than the present sea level, during
gle desiccation since there is simply not enough the last glacial maximum (Kaiser 1969), allowing for
salt in sea water to produce the immense salt many land-bridge connections in various parts of
deposits currently under the Mediterranean. In fact, the Mediterranean (Corsica with Sardinia, Majorca
cycles of desiccation–inundation probably repeated with Minorca, and among different Aegean islands).
themselves ∼8–10 times during the Messinian.
The Messinian salt crisis was, to quote Duggen
1.2.4 Recent geological history
et al. (2003: 602) ‘one of the most dramatic events
on Earth during the Cenozoic era’. During this Geological activity has remained a major feature of
period, the Mediterranean became a desert (Hsü the Mediterranean region in recent history and of
1973). There were land-bridge connections between, course continues. The volcanoes under and around
for example, Corsica, Sardinia and the north of the Mediterranean Sea have different magmatic ori-
Italy; connections linking Sicily with southern Italy gins (Rosenbaum and Lister 2004), being related
and perhaps North Africa, and connections from to either crust extension (those in the centre of the
continental Greece (a) to the south-east across the basin), convergence and subduction (the chain of
small Aegean islands (perhaps to Rhodes) and (b) to volcanoes in the Aeolian islands), or intra-plate mag-
Crete via the Peloponnese. Several authors (Bocquet matism. Subduction continues where plates meet
et al. 1978; Cardona and Contandriopoulos 1979) in many areas such as under the island of Cyprus
have thus discussed whether such land-bridge con- (Robertson and Grasso 1995) and in the Calabrian
nections permitted plant migration, indeed there is Arc. Volcanism has remodelled the Greek island
good evidence for the migration of animals during of Santorini and repeatedly caused extinction and
this period (Alcover et al. 1999). I do not doubt the re-colonization of plant communities on Vesuvius
occurrence of land connections during this period. and Etna (where more than 100 eruptions have been
What remains questionable, however, is how suit- signalled in the last 2,500 years). Volcanic activity
able such connections may have been for plant life, has never ceased, and continues to shake southern
and thus the migration of sedentary organisms. Italy, north Africa, and Turkey.
A problem here is that the climate is not thought to Another important recent event concerned the
have been markedly different during the Messinian contemporary Mediterranean coastline. Four main
compared to the preceding part of the Late Miocene types of erosion have been important here:
and the subsequent Early Pliocene (see below). In (a) mechanical action in the high mountains,
the absence of a change in climate it is difficult to (b) linear erosion by rivers, (c) lateral erosion in
envisage how vegetation could have dropped in alti- semi-arid areas, and (d) aeolian erosion in more arid
tude in order to allow for migration among newly regions. Marine terraces are an important feature of
connected areas (Suc 1989; Quézel 1995). this coastline (Houston 1964; Biju-Duval et al. 1976),
The end of the Messinian occurred suddenly and are particularly prominent along the Ionian
at 5.33 Ma (Krijgsman et al. 1999). The distinct coast of Italy, Corsica, Tunisia, and parts of southern
separation between Messinian evaporites and France and Spain.
Pliocene marine sediments confirms this abrupt I have detailed this complex history of land
end, which coincided with the opening of the masses in the Mediterranean region to illustrate
Gibraltar Straits and the establishment of a per- how geological history has no doubt been closely
manent connection between the Atlantic and the associated with the limitation of species distribu-
Mediterranean. Mantle-related causes may have tions and the creation of phylogeographic divi-
created a new marine connection to the Atlantic sions within the Mediterranean flora. Since the end
(Duggen et al. 2003). Since then the precise location of the Miocene, two other historical factors have
of the Mediterranean coastline has developed its become decisive elements in the shaping of plant

species distributions and endemism around the intensity of rainfall events. Third, Mediterranean
Mediterranean: the onset of a summer drought and, depressions rarely have clear-cut fronts and bare
more recently, the development of human activities. little resemblance to the Atlantic depressions that
regularly swing across north-west Europe. Fourth,
the majority of depressions originate within the
1.3 Two seasons: the history of the Mediterranean Basin itself.
climate and the vegetation An important feature of rainfall in the Medi-
terranean is the marked regional variation in
1.3.1 The contemporary climate
annual levels and timing of peak rainfall events.
The Mediterranean region has a climatic regime Rainfall may be concentrated in the autumn, win-
which is characterized by two main seasons. The ter, or spring, depending on the region. In the
first, a hot dry summer, is the essence of the Medi- western Mediterranean, the peak rainfall occurs
terranean climate region, whose definition relies on in the autumn. A smaller peak occurs in spring
the regular occurrence of an effective drought at the around the coastal areas of the northern rim of the
hottest time of the year (Quézel 1985). During the Mediterranean, in winter from the southern tip of
Mediterranean summer, the main weather centre of Spain (Andalousia) across north Africa and Sicily to
influence is the Atlantic anticyclone of the Azores southern Italy (Calabria), and in spring in the cen-
which imposes fairly uniform sunshine and a quasi- tre of the Iberian peninsula, the Moroccan Atlas and
absence or scarcity of rainfall. Although the onset the high plateau of Algeria. Overall, rainfall declines
of summer can be fairly gradual, its end is usually from west to east. More important, seasonal aridity
abrupt and accompanied by intense rainfall events. becomes longer and more severe in the south than
Thus begins the second or ‘wet’ season whose moist in the north, and in the east compared with the west.
and cool climate assures plant development. This These trends occur on three spatial scales: basin-
cooler period lasts from 5–10 months, depending on wide, on individual pieces of continents such as
the region. During this period, intense cold may the Balkans and Greece (where annual rainfall more
occur in some regions and limit plant development. than doubles as one moves from the eastern fringes
Such winter stress may limit species distribution, of north-west Greece to the western Balkans), and on
as noted for some sclerophyllous species (Mitrakos individual islands such as Crete (where once again
1982) and thus represents a second constraint on annual rainfall in the west is twice that of the eastern
the phenology and growth of Mediterranean plants tip of the island). Detailed graphs of sunshine and
(Chapter 4). To sum up, the Mediterranean climate precipitation are well documented elsewhere (e.g.
imposes a double constraint on plant growth, lack Houston 1964; Zohary 1973; Blondel and Aronson
of moisture in summer and cold temperatures in 1999; Grove and Rackham 2001; Quézel and Médail
winter. 2003), where they illustrate clearly the spatial het-
During the cool moist period the Mediterranean erogeneity in climatic regime in terms of the length
region lies between a cold anticyclone present in of the summer drought and the timing and amounts
Asia and the Atlantic anticyclone, that is, in a of rainfall.
zone of low pressure with minimum values over The contemporary Mediterranean climate is
the main sea basins, which are the centres of fre- a recent phenomenon and three main chapters of
quent cyclogenesis. The advection of air streams climatic history provide the backdrop to the consti-
from these sources dominate weather conditions tution of the contemporary vegetation.
and create important events and periods of rainfall.
There are four important characteristics of rain- 1. Prior to the Middle Pliocene (i.e. >3 Ma).
fall in the Mediterranean (Houston 1964). First, 2. Recent alternation of summer drought and cold
annual rainfall is concentrated into a small num- temperatures associated with glaciation.
ber of events. Second, in a given region, there is 3. Climate change in the presence of human activ-
enormous interannual variation in the timing and ities, that is, since the last glaciation.
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 19

Box 1.3 Pollen analyses and vegetation history

The study of pollen composition in cores of 2. It is difficult to distinguish congeners—which

sediment is a major tool in the study of vegetation may have very different ecological requirements. In
history. Nonetheless, important caveats in the use mixed oak forests, deciduous oak pollen may be
and interpretation of pollen data for assessing more abundant than evergreen oak pollen in soil
community composition require appreciation (see samples because they grow on deeper soils.
Bazille-Robert et al. 1980; Pons and Suc 1980; 3. Only particular conditions (such as in peat bogs)
Reille et al. 1996; Grove and Rackham allow preservation. Mediterranean ecosystems
2001): away from the coast are, for the most part, far
from being peat bogs. Such conditions are
1. A bias due to the marked variation among taxa uncommon and only occur in isolated spots.
in their preservation and their likely input to Mediterranean taxa may thus be underestimated
sediments due to differences in pollen production in historical reconstruction.
and release. It is highly unlikely that insect 4. Some areas of the Mediterranean have been
pollinated species will input as much pollen to a intensively studied (Spain, southern Italy, and the
sediment as wind pollinated species, even if the South of France) while others have received almost
two were at equal abundance in the landscape. no attention.

The historical study of Mediterranean vegetation the European plate (Mai 1989). In North Africa,
has a long tradition (de Saporta 1863). Much of what temperate rainforest occurred in the Saharan zone
is known is based on analysis of pollen frequencies and subtropical woodland savanna (with a species
in cores, which, for diverse regions, should be inter- composition that suggests the occurrence of a dry
preted with some caution (Box 1.3) with additional season) existed in the area where present day
information coming from fossilized plant parts and Mediterranean vegetation occurs, that is, the three
charcoal remnants. countries of the Maghreb and coastal areas of
Libya and Egypt (Quézel 1978). The decline of this
flora occurred during the cyclic periods of cool-
1.3.2 The onset of the Mediterranean climate
ing from the Late Miocene onwards. Relicts of the
The Early Tertiary Tertiary flora include the following: laurel forest
During the Early Tertiary the accuracy of vegetation vegetation (on the Macaronesian islands) in a few
analysis based on pollen remains is poor since only ancient forests in relatively humid areas (e.g. Laurus,
few tree species are identifiable. It is, however, gen- Prunus, Persea, Daphne, and Ocotea), Rhododendron in
erally thought that south of Tethys, the vegetation oceanic parts of the Iberian peninsula, the presence
was essentially tropical (forest and savanna) and dif- of Liquidambar, Parrotia, and Pterocarya in eastern
ferent to that to the north (Quézel 1995), where scle- Turkey, Zelkova abelicea and Phoenix theophrastii on
rophyllous vegetation, akin to that in western North Crete, and Nerium in stream beds on Corsica. The
America, was present: the so called Madro-tertiary findings of fossil leaves of several species have
Geoflora (Axelrod 1958) or ‘Madre-Tethyan’ been particularly instructive for our knowledge of
sclerophyll vegetation (Axelrod 1975). the overall vegetation types of this and subsequent
periods of climate history (Vernet 1997).
The Late Tertiary Otherwise, only small amounts of fossil mater-
In the Late Tertiary (Oligocene and Miocene, ial are available to help elucidate the origins
33–5 Ma) evergreen rainforest and laurel forests of the Mediterranean sclerophyll forests. Some
were the two most important vegetation types on species (Nerium, Olea, Cupressus, Punica, Pyracantha,

Jasminum) could have had their origins in the lau- and Rhamnus, and species of Prosopis, Vitis, and
rophyll floras of the Eocene (>33 Ma), while others Cistus, were present along the northern shores of
(Ceratonia, Cercis, Phillyrea, Pistacia, and the ever- the Mediterranean (Pons and Suc 1980; Bessedik
green oaks) may have belonged to the mixed meso- et al. 1984; Bessedik 1985). Elements of contempo-
phytic forest flora of the Oligocene (33–23 Ma). But rary Mediterranean-type vegetation were present,
these taxa were, towards the end of the Oligocene but only in complex vegetation associations that no
(∼25 Ma), only sporadic elements of a very dif- longer exist. Semi-arid elements were present as a
ferent vegetation to that in which they now occur secondary component of an ancient Mediterranean
(Medus and Pons 1980). The Mediterranean meso- landscape dominated by tropical and warm tem-
xerophytic sclerophyll forest in its contemporary perate, evergreen and deciduous elements, with a
composition and structure is a relatively recent phe- mangrove coastal vegetation.
nomenon, and could only have become established
after the disappearance of the laurophyll vegetation. The Middle Miocene
The analysis of fossil macrofloras (Palamarev In the Middle Miocene (16–14 Ma) seasonal contrasts
1989) attests to the appearance of a ligneous scle- in the temperature regime developed, perhaps as a
rophyll vegetation in the Late Eocene. These scle- consequence of glaciation in northerly latitudes and
rophyllous species were secondary elements of the the loss of water connections to the Indian Ocean.
widespread subtropical woody vegetation. Their Tropical elements began to disappear from pollen
abundance increased during the Oligocene when diagrams during this period (Pons et al. 1995) and
they formed more complete xerothermic communi- floristic richness declined due to the loss of whole
ties. The best-represented species were in the genera taxonomic groups from many regions. As a result,
Quercus, Arbutus, Pistacia, Ceratonia, Acer, Periploca, the flora began to resemble contemporary vegeta-
Smilax, and Pinus. tion. Pollen analyses from cores in southern France
show that this was the case for Bombax (Bomba-
The Early Miocene caceae), Alchornea (Euphorbiaceae), two genera of
In the Early Miocene (∼23 Ma) pollen spectra Icacinaceae, Simaroubaceae, Avicennia (Avicenni-
indicate that the flora of the northern sector of aceae), Rhodoleia and Eustigma (Hamamelidaceae),
the western Mediterranean was rich in subtrop- and Gunnera (Gunneraceae). For example, Avicennia
ical species and families (Bessedik et al. 1984; disappeared from the Languedoc in southern France
Bessedik 1985). In short, the climate was trop- (14 Ma), from Sicily at 5 Ma, and is now extant
ical, with little seasonal change in temperature on the Red Sea coastline (Suc et al. 1992). Extinc-
and fairly high levels of summer rainfall. Major tions were thus primarily in taxonomic groups
elements of this subtropical vegetation included with high temperature and humidity requirements.
representatives of the Taxodiaceae, Bombacaceae, Several of the groups that became extinct from the
Hamamelidaceae, Juglandaceae, Melicaceae, western Mediterranean, currently occur in tropical
Melastomataceae, Menispermaceae, Oleaceae, and subtropical regions of south-east Asia, Africa,
Restionaceae, Sapindaceae, Sapotaceae, and central America and the Neotropics (Bessedik
Simaroubaceae. In addition, in many coastal 1985).
areas of the western Mediterranean, mangrove
swamps dominated by the genus Avicennia were The Late Miocene
present. By the Late Miocene (10–6 Ma), prior to the
The overall vegetation was highly heterogeneous Messinian salinity crisis, important concentrations
and some parts of the western Mediterranean, in of Palaeo-Mediterranean species began to develop
particular the low plains (<500 m elevation), hosted as more tropical elements were lost. Bocquet et al.
a semi-arid open vegetation (Bessedik 1985). Some (1978) proposed that the drop in sea level during
of the elements of the current day Mediterranean the Messinian was associated with a drier climate
vegetation, for example, Olea, Pistacia, Nerium, during the Messinian salinity crisis, and a greater
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 21

possibility for plant migration as a result of land components (Pons 1984):

connections. However, it is probable that the climate
• a temperate vegetation similar in composition to
was not particularly dry during this period relative
that present in non-Mediterranean Europe;
to the preceding period (Hsü 1973), and vegetation
• groups of taxa now present either in North
does not show major changes during this period
America or the far east;
(Suc and Bessais 1990; Suc et al. 1992, 1995; Bertini
• an ancestral Mediterranean element with Abies,
1994; Fauquette et al. 1998). Mangroves (Avicennia)
Cedrus, Nerium, Parrotia, and Quercus.
continued to disappear, although increased salinity,
rather than aridity, may have been the cause. Com- During this period, vegetation was spatially hetero-
munities of a Mediterranean sclerophyllous-type, geneous. For example, open xeric assemblages were
during this period, contained evergreen oaks and probably more common in Catalonia (north-east
pines, along with Arbutus, Ceratonia, Olea, Phillyrea, Spain) compared to the Languedoc of southern
Pistacia, and Pyracantha (Palamarev 1989). In the France (Suc et al. 1992). South of Barcelona, xeric
mountains of North Africa, sclerophyllous ever- herbs were increasingly important with Ceratonia
green forests occurred (Quézel 1978). Various pollen and Palmae, while Taxodiaceae, Quercus, and other
evidence suggests the presence of Quercus pollen more mesophilous taxa show decreased abundance.
in the Mediterranean region in this period (Barbero So moving south, the mixed forest was replaced,
et al. 1992). by open, xeric, Mediterranean-like communities
(Bessais and Cravatte 1988).
The decline and disappearance of the Taxodiaceae
The Early and Middle Pliocene was not synchronous across western Europe. In the
In the Early Pliocene (∼4 Ma), the climate was prob- Mediterranean region, the decline set in during
ably a few degrees warmer and slightly more humid the Middle Pliocene whereas in the rest of Europe
than at the present time (Fauquette et al. 1998). Pollen this decline did not occur till the Late Pliocene
diagrams show that tropical elements had disap- and Early Pleistocene (Michaux et al. 1979). Indeed
peared from the flora of the northern shores of the by the Middle Pliocene, taxa in the Hamamel-
Mediterranean but were still present to the south. idaceae and Juglandaceae diminished and the
In the north-western Mediterranean, pollen analysis Taxodiaceae were completely lost from the coastal
suggest that coastal vegetation was dominated by areas. The loss of the Taxodiaceae and depletion
Taxodiaceae (with Myrica, Symplocos, and Nyssa) and of other groups was only part of a more general
Lauraceae, while drier inland areas had many Jung- increase in rates of extinction during the Mid-
landaceae and Hamamelidaceae (e.g. Liquidambar). dle Pliocene (Bessedik et al. 1984). In addition
A type of evergreen broad-leaved forest prevailed to the loss of Taxodiaceae, this period witnessed
at low altitudes, as did rainy summers. Pignatti the disappearance from the western Mediterranean
(1978) hypothesized that at the end of the Pliocene of several genera in the Sapotaceae, Restoni-
altitudinal transitions occurred on Mediterranean aceae, Agavaceae, Leea (Leeaceae), Embolanthera and
mountain slopes, from Laurophyllous forests at Hamamelis (Hamamelidaceae), Rhioptelea (Rhiopte-
low altitude, through Ilex-Taxus forests (with Buxus, leaceae), Symplopus (Symplocaceae), Microtropis
Ruscus, and Daphne) and mountain conifers (Picea, (Celastraceae), and Nyssa (Nyssaceae). These extinc-
Abies, Cedrus) to spiny shrubs (Astragalus, Genista) tions were not simultaneous in different regions;
at high altitude. The latter two belts can be observed Taxodiaceae, Symplocus, Nyssa, and a few others
on many southern massifs of the Mediterranean, remained for longer periods in Catalonia than they
while the Laurophyllous forests have disappeared did in the Languedoc. The progressive appear-
and the Ilex-Taxus belt now only occurs in relictual ance of the Mediterranean climatic regime, and
formations. in particular the precipitation regime was the key
During this period vegetation in the northern element in these extinctions. While setting the
sector of the western Mediterranean had three main scene for the evolution of one of the world’s most

diverse contemporary floras, the evolution of the taxa such as Cistus and Phlomis in some sites indi-
Mediterranean climate thus also caused high levels cates that the temperature had not cooled to a great
of extinction in the pre-existing flora. degree, but had become much drier. There was
It was thus in the Pliocene (∼3 Ma) that a gradual also much variation among sites in the composition
but profound climatic change in the Mediterranean of pollen spectra: well-developed Mediterranean
region began (Suc 1984). This change did not occur in communities appear to have already occurred in
a parallel fashion elsewhere in temperate Eurasia or southern Italy and Languedoc at this time whereas
in the tropics of Africa. At this time, the temperature a more steppe-like vegetation occurred in north-
regimes began to drop significantly, introducing a west Spain (Bazile-Robert et al. 1980; Pons et al.
marked seasonality, not just in temperature, but also 1995). Geographic variation in the development of
in the establishment of a marked and prolonged dry Mediterranean vegetation thus probably occurred.
season, which became more and more severe as its The fluctuations of these two types of pollen spectra
association with the newly established warm season suggest rapid climatic cycling of short duration but
developed. intense amplitude.
As coastal forests thinned out and disappeared Altitudinal zonation of the vegetation was present
and more xeric vegetation developed, a complex in the Late Pliocene (Suc 1984). For example, pollen
mosaic of vegetation types established in the land- spectra from Calabria contain pollen of different
scape of the Mediterranean (Suc 1984; Combourieu- vegetation associations, some of which no longer
Nebout 1993; Pons et al. 1995). The summer drought coexist in the Mediterranean and others which
stabilized at ∼2.8 Ma and by ∼2.3 Ma some of the are now absent from the region (Combourieu-
oldest signs of extensive Mediterranean vegetation Nebout 1993). These spectra suggest a succession of
can be detected in pollen cores (Suc 1984). The ori- deciduous forest (with abundant Quercus associated
gin of the Mediterranean climatic regime is thus very with Acer, Carpinus, Celtis, and others), subtropi-
recent and its onset had, as its main element, a fluc- cal humid forest (Taxodiaceae and Cathaya which
tuation in the rhythm of rainfall, rather than a strong now occur in western Asia), high-altitude conif-
contrast in temperature (Suc 1984). erous forest (Tsuga, Cedrus, Abies, and Picea) and
The onset of the highly seasonal climate showed open steppe vegetation (composition as above).
marked spatial variation. For example, the first signs This sequence most likely represents the vegetation
of a summer drought appear from the Late Miocene response to climatic change from warm and fairly
(10–6 Ma) in North Africa (Bachiri Taoufiq 2000), humid interglacial periods to colder, drier glacial
the Early Pliocene (∼4 Ma) in Calabria and Sicily periods (Combourieu-Nebout 1993). Fairly rapid
(Bertoldi et al. 1989), and ∼3.5 Ma in the north-west climatic oscillations, mostly due to changes in rain-
of the Mediterranean Basin (Suc 1984). fall, and less the result of temperature variations,
probably caused the shifts from subtropical forest
to herbaceous open vegetation. The presence of
The Late Pliocene strong altitudinal gradients around the shores of the
In the Late Pliocene (2.5–2.1 Ma) pollen spectra Mediterranean may have allowed different associa-
indicate strong fluctuations between two main tions to locally persist and thus rapidly track climate
vegetation types: (a) forest communities rich change. In the Late Pliocene, floristic differences
in deciduous species no longer present in the between the western and eastern Mediterranean
region (e.g. Carya, Pterocarya, and Parrotia) and regions were already apparent (Palmarev 1989).
(b) steppe communities dominated by Artemisia As the climate changed in the Pliocene a summer
accompanied by the genus Ephedra and a range of drought climatic zone was formed between 37◦ N
Amaranthaceae and Chenopodiaceae. The presence and 45◦ N where prominent sclerophyllous woody
of steppe vegetation attests to drier (and slightly ecosystems developed. The important presence in
cooler) climatic conditions than in the Middle the pollen spectra of this period of Cupressaceae,
Pliocene (Suc and Cravatte 1982). The presence of Pinus, Quercus, Olea, Phillyrea, Cistus, Helianthemum,
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 23

Rhus, and Rhamnus suggests that these initial During such cool periods, strong ecological dif-
Mediterranean plant communities developed on ferences developed between the eastern and western
low hillsides with a dry calcareous soil. Their direct Mediterranean regions. In parts of the eastern Medi-
descendants constitute much of woodland vegeta- terranean such as Israel (Horovitz 1979), pollen
tion of the contemporary Mediterranean flora. analyses suggest the occurrence of oaks and olives in
a steppe vegetation lacking Artemisia and more rem-
iniscent of western Mediterranean garrigues and
1.3.3 Climatic oscillations associated with maquis (Pistacia, Cupressus, Rosaceae, Poaceae). The
glaciation coniferous forests (with fir and scots pine and some
Late Pliocene to Early Pleistocene Corylus, Acer, Carpinus, Buxus, Tilia, and Fagus)
During this period, the gradual cooling and dry- present in southern Europe during the Quaternary
ing of the climate occasioned the extinction of many also showed an east–west variation in composition,
species from various regions. In some cases, whole with Betula and Hippophae in the west and deciduous
genera, sometimes the sole representatives of par- oaks in the east.
ticular families, became extinct in particular regions The warmest periods saw the localized develop-
of the Mediterranean. In southern France these ment of Mediterranean forest vegetation, containing
losses included genera such as Carya, Pterocarya, and Quercus, various Oleaceae, Pistacia, Cistus, Ostrya,
Juglans (Juglandaceae), Eucommia (Eucommiaceae), Vitis, Juglans, and Pinus. In the Late Quaternary,
Elaeagenus (Elaeagenaceae), Zelkova (Ulmaceae), and the climate dried and the vegetation took on a
genera such as Parrotia, Parrotiopsis, and Liquidambar more Mediterranean aspect with evergreen oaks and
in the Hamamelidaceae (Bazile-Robert et al. 1980; Cistus becoming more abundant (Bazile-Robert et al.
Bessedik et al. 1984). The loss of such groups was 1980; Pons and Suc 1980; Brenac 1984).
again variable between regions, occurring gradually
later towards the east (Bessedik et al. 1984). Aridity, The last glacial maximum
associated with reduced temperature, was thus a The last glacial maximum in southern Europe
key factor causing what was to be a third wave of occurred around 20,000 bp. This was a dra-
enhanced extinction rates. matic moment for Mediterranean vegetation which
declined to what Pons (1984) termed ‘état zéro’. Tem-
The Pleistocene
perature depression in southern Europe is thought
In the Pleistocene (1.8 Ma–15,000 bp) forest and
to have been something of the order of 5–7◦ C;
steppe continued to alternate, at a rhythm of
markedly less than the 15–16◦ C depression at the
∼100,000 years (Pons et al. 1995). Steppe vegetation
southern limits of permafrost further north (Kaiser
covered large expanses of the landscape as glaciers
1969). The seasonality of rainfall was also more
moved south across the northern parts of Europe.
marked than at the present time (Prentice et al. 1992).
In contrast to some of the previous colder periods,
Mediterranean plants persisted through this period
steppe vegetation comprised species indicative of
in isolated glacial refugia which were to be the
cooler temperatures. In the western Mediterranean,
sources for future re-colonization (Box 1.4).
the most common pollen types were Artemisia and
various Chenopodiaceae (Bazile-Robert et al. 1980).
Trees were not absent from the landscape, pines Start of the late glacial
were abundant, albeit with much spatio-temporal The start of the late glacial occurred from ∼15,000 bp
variation, Juniperus are recorded in the early parts in southern Europe. The pattern of late glacial veg-
of this period, as are, Betula and Hippophae dur- etation development in southern Europe showed
ing the coldest periods. Steppe vegetation occurred much spatial heterogeneity in relation to local cli-
in southern Spain (Pons and Reille 1988) and a matic conditions and soil type and moisture (Turner
savannah-like vegetation may have persisted in and Hannon 1988; Reille et al. 1996; Carrión 2001).
sheltered areas. The spread of vegetation in association with this

Box 1.4 Survival during ‘état zero’: types of glacial refugia and where they occurred

In Mediterranean Europe, major regions of refuge in north Africa). The existence of such pockets of
occurred in the southern Iberian peninsula, vegetation on south-facing slopes above the plains
Greece, and the Balkans. Other refugia no doubt helps explain why the reforestation of
occurred in the Middle East and North Africa. In Mediterranean mountains occurred so quickly in
addition to the major zones of refuge other the first few thousand years (particularly between
smaller refugia may have occurred across the 13,000–11,000 BP) after the glaciers began to beat
southern extremes of Europe. Refugia would have a retreat (Peñalba 1994; Reille et al. 1996, 1997).
been located in a landscape whose vegetation was 2. In the western Mediterranean isolated trees
otherwise dominated by grasses and Artemisia, and shrubs persisted in the lower parts of ravines
then widespread over southern Europe. This was and river gorges—many of these sites have
essentially a steppe vegetation associated with an probably since been submerged.
arid climate where lack of rainfall, perhaps as 3. Pockets of species-poor deciduous oak forest
much as low temperature, put a severe restriction on the southern shores of the Mediterranean, and
on tree growth (Kaiser 1969; Van Campo 1984). probably in southern tips of the continent on the
Exactly where Mediterranean vegetation northern shores.
persisted during the glacial periods and whether 4. Open deciduous oak forest with pines and a
such persistence occurred in more than just a few few other Mediterranean taxa near the sea in the
isolated and small pockets in protected rocky areas eastern Mediterranean.
around the coast is not completely known, 5. Highly isolated refugia probably dotted the
although in some precise locations long-term landscape in sheltered valleys and near the coast.
persistence of tree cover has been clearly
demonstrated (Tzedakis 1993). In the western Cliffs are a conspicuous feature of the
Mediterranean, glaciation may have had more Mediterranean landscape and probably played an
severe effects on plant distribution than further important role as a refuge, particularly maritime
east, and evergreen oak forests probably only cliffs and those with a southerly exposure. In cliffs,
persisted in the southern tips of Spain and Italy, open vegetation typical of contemporary
being otherwise displaced into large areas of north garrigues, phrygana, and maquis vegetation may
Africa and the south-east margin of Europe. Mixed have persisted alongside strict chasmophytes
deciduous forests would have been fairly extensive during the glacial maxima (Davis 1951; Snogerup
across the southern half of the Iberian peninsula, 1971). The fact that not all the species that occur
down the east and west coasts of Italy, and on limestone cliffs are chasmophytes supports the
perhaps in a small coastal band around parts of idea that at least a few elements of the
southern France. Otherwise many species probably Mediterranean vegetation found a refuge in cliffs
persisted in warmer and more humid localized during the Quaternary glaciations. For example, in
pockets of the landscape (Pons and Suc 1980; the Aegean, some species (e.g. Anthyllis
Pons 1984; Hermenger et al. 1996). hermanniae) occur as a chasmophyte in parts of
Pons (1984) provides several elements of their range and in phrygana vegetation elsewhere
response to the question of precisely what types of and several forest species (e.g. Quercus ilex,
habitat acted as refugia in a bleak and barren Pistacia terebinthus, and Rhamnus alaternus) can
Mediterranean landscape. be observed in cliffs (Snogerup 1971). Likewise in
limestone cliffs of the Iberian peninsula, many
1. In the western Mediterranean on south-facing common species are generalist chamaephytes (e.g.
slopes above the arid plains at 400–800 m Saxifraga monocayensis and Silene saxifraga)
elevation. During this period the treeline was which represent pioneer colonists, or
situated at around 800–1000 m on the northern nanophanerophytes (e.g. Rhamnus alpinus and
shores of the Mediterranean (but reached 1,500 m Lonicera pyrenaica).
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 25

warming varied in relation to latitude and the exist- Tilia, all of which now occur more commonly at
ence of local refugia which would have acted as higher latitudes and in cooler and wetter parts of
additional sources for expansion. Whereas oaks the Mediterranean landscape where they represent
began to spread form their probably quite vast and relicts of a once more widespread distribution. High
patchy network of refugia in southern Spain from frequencies of Pistacia in southern areas and Corylus
13,200 to 12,000 bp, their expansion is only recorded elsewhere point to the existence of a fairly open for-
from 10,000 bp in northern Spain (Peñalba 1994), est vegetation in some areas. This forest diversified
where refugia would have been few and far between. and in some areas became more dense and closed by
In north-west Syria, vegetation of a Mediterranean about 8,000 bp (Pons et al. 1995).
type is thought to have been an important com- There was marked geographic variation in the
ponent of the landscape by about 11,000 bp with dominant species present in these forests: ever-
assemblages of Quercus, Pistacia, and Olea present green oaks in drier areas of southern Spain, decidu-
in the plains and Cedrus, Carpinus, Ostrya, and Quer- ous oaks in southern France and in Italy, firs in
cus on the mountain slopes (Niklewski and van Zeist the northern Appenines, pines in the Maritime
1970). Alps and the eastern Pyrenees, and Pinus nigra
The initial warming of the climate was disrupted subsp. laricio forests on Corsica (Reille et al. 1996).
by short but intense cold periods, such as in the Pollen cores in southern Spain attest the presence
recent Dryas (11,000–10,000 bp). During these cold of a Mediterranean-type vegetation from around
snaps, forests retreated again and steppe vegetation 10,000 bp in some sites, for example, Pistacia pollen
spread (Pons and Reille 1988), Betula pollen showed has been recorded from 9,500 bp and cork oak
a marked decline in some sites (Turner and Hannon (Quercus suber) from 6,800 bp in the Sierra Nevada
1988). In the drier regions, such as in southern Spain, (Peñalba 1994; Grove and Rackham 2001). However
Quercus ilex showed a marked decline during this there is a great deal of heterogeneity among pollen
period (Reille et al. 1996). On Corsica, a marked sequences from different sites in Spain, where even
increase in Artemisia pollen and the absence of Pinus in the semi-arid south-east, pollen records show
nigra subsp. laricio pollen suggests that this charac- much variation in the timing of different vegetation
teristic tree in upland forests was not then present stages, perhaps as a result of local variation in topog-
on the island (Reille et al. 1997). However, even raphy and microclimate or time lags in vegetation
during these cold periods, when Artemisia pollen development linked to the vegetation present at a
dominated with various Chenopodiacae, Poaceae, site prior to any climate modification (Carrión 2001).
Apiaceae, Asteraceae, and Ephedra, a xerophytic Since ∼10,000 bp trees were present in the large
vegetation similar to that which is now present, per- majority of the landscape, forest was less abundant
sisted in sheltered lowland sites on Corsica (Reille in the eastern and southern parts of Mediterranean
et al. 1997). In the last period of climatic oscillations, Europe than in the north-west part of the basin. In
that is, 18,000 to 10,000 bp, Prunus pollen became southern Spain, the return of forest vegetation was
more abundant in some regions, as did Juniperus and rapid, probably because of the numerous localized
Cistus. Bushes and small trees thus appeared to pre- refugia that may have dotted the landscape during
dominate in the landscape, suggesting a fairly cool the ‘état zero’ of glacial maxima (Box 1.4).

1.3.4 Ever since glaciation: climate change and

From 10,000 bp onwards
human activities
From 10,000 bp onwards a more definitive warm-
ing began. At this time, deciduous oak forests In the Holocene (<10,000 bp), the natural ecology of
covered large areas on the slopes of many of all circum-Mediterranean regions came under the
the Mediterranean mountains (Pons et al. 1995; influence of a new ecological factor, namely human
Grove and Rackham 2001). Present in these forests activities (Triat-Laval 1979; Pons 1984; Barbero et al.
were Corylus, Alnus, Fraxinus, Betula, Ulmus, and 1990; Pons et al. 1995; Quézel and Médail 2003).

Humans have been present in the Mediterranean for development and the evolution of environmental
longer than anywhere else in Europe, as the skele- perception and conservation. Evidence for human-
ton dated at ∼400,000 years discovered in a cave near induced decline in forest vegetation up till the last
the village of Tautavel in the Roussillon of southern 200 years or so has been presented for a diverse array
France illustrates all too well. As the cave paintings of situations, most of which involve changes in the
of the western Mediterranean and its periphery also distribution and composition of forest communities,
illustrate, Cro-magnon humans were present in the and their reversion to shrubland over large areas, the
southern parts of the Europe for long periods during spread of evergreen oaks at the expense of decidu-
the Quaternary. ous oak and the spread of pines (Box 1.5). Human
Since the last glacial maximum, the composi- activities have clearly had a major impact on the
tion and spatial relationships of Mediterranean eco- Mediterranenan landscape (Lepart and Debussche
systems have been greatly modified by more extens- 1992).
ive human activities. However, in the same time One way in which human activities have had a
spell, the climate has also warmed and become drier. dramatic influence on vegetation in the Mediterran-
There has thus been some debate as to whether ean region has been via the use of fire. Fires are
major modifications to Mediterranean plant com- thought to have occurred naturally since at least
munities in the last 6–7,000 years are more the result the Miocene (Dubar et al. 1995). With the onset of a
of human activities than they are of climate change. highly seasonal association of high summer temper-
Since the intensity and timing of human activities atures, drought and often strong winds, fires prob-
were different in different places there is also much ably became more frequent and may have been an
spatial heterogeneity in the anthropogenic element important feature of the ecology of natural vegeta-
of pollen sequences (Carrión 2001). Pons and Quézel tion in the primeval Mediterranean forests. This
(1985) and Quézel and Médail (2003) argue that natural selection pressure has been greatly modi-
although climate change probably drove vegetation fied as human activities learnt its use and started
change up till ∼10,000 bp, human activities have to create pastures and enrich soils for cultivation.
since then become the determining factor influenc- In the current-day Mediterranean landscape, forest
ing Mediterranean forest cover, beginning in the east fires are almost exclusively linked to human activ-
and moving west. ities, hence their consideration as an anthropogenic
Around 10,000 bp, human-induced forest clear- disturbance in the landscape (Moreno and Oechel
ance was dotted around the landscape in the form of 1994).
small, temporary clearings in an otherwise forested In historical times, burning was a constant feature
landscape (Pons and Thinon 1987). Since then, the of land clearance and settlement in the Mediterran-
impact of human activities has increased dramat- ean, where the use of fire began much earlier than
ically (Pons 1984). One can identify the following elsewhere in Europe. Some of the oldest evidence
phases: (a) the development and diffusion of agri- for a human presence in the Mediterranean, such as
culture in the Neolithic with cereal cultivation on the cave settlements near Tautavel (southern France)
plains and low-altitude plateaux, (b) more gener- which date to ∼400,000 bp, attest to the, albeit prob-
alized forest clearance, a little after 3,200 BP in ably limited, use of fire. Some time in the Neolithic
north-west Greece, 2,800 bp in Provence, 2,700 bp on (∼8,000 bp), fires became more frequently used over
the Dalmatian coast, and 2,500 bp on Corsica (i.e. a period of about 2,000 years, and since ∼4,500 bp
during Greek Antiquity and the Roman Empire), they have become general practice and widely used,
(c) political events and socio-economic changes as the Early Holocene abundance of cork oak pollen
(e.g. medieval changes in land ownership, wars, suggests (Pons and Thinon 1987; Pons and Reille
and population movements and other demographic 1988; Grove and Rackham 2001). According to
changes) up to the end of the nineteenth century, and Pons and Thinon (1987: 10) the importance of fires
(d) twentieth-century rural depopulation, coastal associated with human activities became such that
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 27

Box 1.5 The diverse types of human impact on forest vegetation in the Mediterranean

Several lines of evidence for human impact on the Q. ilex, although present prior to human presence,
Mediterranean landscape, and in particular the has only become a dominant component of forests
destruction of forest and changes in tree/shrub as human activities (e.g. colonization by the
species composition have been Romans and the Republic of Genova) developed in
proposed. the last few thousand years.
4. Expansion of woody shrub and herbaceous
1. A reduction in cover and thinning out of vegetation and in particular the replacement of
deciduous oak forests and their replacement by evergreen oak woodland by maquis (Pons et al.
evergreen oaks, for example, in the Moroccan Rif 1995).
where Quercus canariensis and Q. toza have been 5. Contemporary extension of pine forest, for
replaced by Q. ilex and Q. suber (Reille et al. 1996). example, Pinus pinaster in the High Atlas and
In Provence, Q. ilex increased naturally in P. halepensis in the Languedoc of southern France
abundance, from a scattered tree in open (Quézel and Médail 2003).
communities dominated by Juniperus, to a more 6. The extension of Kermes oak garrigues in
open woodland as climate warmed between association with human fires and intense grazing
15–10,000 BP (Triat-Laval 1979). This expansion (Triat-Laval 1979).
occurred as Juniperus declined, not at the expense 7. The stopping of Fagus (and perhaps also
of deciduous oaks, and prior to significant human Carpinus) expansion in a westerly direction in
impacts in the region. It was only after the cutting northern Spain (Peñalba 1994).
of deciduous oak forests from 7,000 bp onwards 8. Species introductions, such as on Corsica where
that evergreen oak spread into what was two species which are now important vegetation
previously deciduous forest in this region. From elements stem from human introduction:
then on, the two types of oak show negative P. halepensis introduced as late as the nineteenth
correlations in abundance, suggesting the century (Reille 1992) and Castanea sativa, totally
replacement of deciduous oaks by evergreen oaks absent from the pollen spectra prior to the
as human activities expanded. In southern France sub-Boreal (Reille et al. 1997).
this may have been due to the occurrence of 9. Pollen analyses on Corsica (Pons and Reille
deciduous oaks on deeper soils which were the 1988; Reille and Pons 1992) show that the marked
most valuable for cultivation (Lepart and rise in abundance of Q. ilex and Q. suber only
Debussche 1992). happened as development of human activities
2. The regression of mountain forests, for proceeded, that is, after 6,000 bp. Prior to this,
example, Cedrus atlantica, C. libani, and Juniperus deciduous oaks were the dominant tree species.
thurifera in the Rif and Atlas (Pons 1984; Pons Although Q. ilex has been naturally present in
et al. 1995). many areas, it was probably not the dominant tree
3. Invasion of other forest types by Q. ilex and species, except in areas with a semi-arid climatic
pines: On Corsica, Reille (1992) showed that regime and/or on shallow soils.

‘the changes induced by . . . [fire were] . . . so great halepensis in southern France is more due to the aban-
that an analysis of the present relationships between donment of cultivation and pastoralism than the
ecosystems and environmental factors can be signif- direct result of fires (Barbero et al. 1987b).
icant only if it takes into prior consideration all the Although recent fire statistics are often difficult
anthropogenic past of the ecosystems’. This is never- to interpret (see Grove and Rackham 2001), the fire
theless a hefty statement, given the strong influence regime in the Mediterranean region has changed. In
of drought and nutrient stress and the impact of a general survey of Mediterranean forest fires over
land-use changes. For example, the spread of Pinus a 30 year period, Le Houerou (1987), documented

the gradual increase in the spatial extent of burnt The prevailing paradigm of landscape destruc-
areas from an average 200,000 ha/year (1960–71) tion as a result of human activities has been chal-
to 470,000 ha/year (1975–80), and 660,000 ha/year lenged by a number of authors, who argue that
(1981–85). The number of fires has increased human activities are not the major cause of land-
in parallel fashion. Since the 1980s, data for scape degradation in the Mediterranean region.
Mediterranean France indicate that the surface burnt In their recent book on the ecological history of
by fire has stabilized (albeit with much annual varia- the Mediterranean landscape, Grove and Rackham
tion) while the number of fires continues to increase (2001) argue at length against the idea that the con-
(Quézel and Médail 2003). Analysis of fire regimes temporary Mediterranean landscape is a ‘degraded’
in Catalonia, where detailed inventories exist for the landscape. Indeed, in many areas open landscapes
medieval period (1370–1462) and the late twentieth are not the result of human activities. Here are some
century (1966–96) indicates that although there has examples.
not been an increase in fire frequency between the
1. Many endemic species do not occur in forest
two periods, the surface burnt by individual fires has
habitats (Chapter 2), suggesting that the latter has
greatly increased and the annual number of summer
not been a ubiquitous landscape feature in the
fires has increased (Lloret and Marí 2001). As these
authors illustrate, the occurrence of a small number
2. The well-developed semi-arid floras of south-east
of very large fires has become, in the last 50 years,
Spain and south-east Crete attest to the historical
an integral part of the fire regime in Mediterranean
existence (throughout the Holocene and preceding
forests. In the light of the recent massive and numer-
interglacial periods) of areas too dry for forest, in
ous summer fires of 2003, the conclusion made
particular deciduous oak forest. On the Cyclades,
by Le Houerou (1987: 22) that ‘the ever increas-
tree species are few and well-developed garrigues
ing build up of fuel in Mediterranean forests and
or maquis are rare due to the poor soils and arid
shrublands as a result of rural depopulation and the
climate. In these areas, climate may have played a
abandonment of marginal lands . . . will sooner or
more important role in the lack of forest cover than
later make new legislations necessary as well as the
human-induced impacts.
adoption of new methods of prevention’, remains
3. On Corsica, maquis vegetation now omnipresent
in many areas on the island (e.g. in the Agriates and
Diminished human activity in forests, for
the Cap Corse) was a well-established feature of the
example, glass-making, tanning, and charcoal burn-
landscape prior to the onset of human activities, as
ing have become (almost) obsolete, is an essential
pollen spectra for Erica arborea attest (Reille 1992).
element of the fuel build-up and the occurrence
The abundance of E. arborea was probably due to the
of a small number of very large fires, or con-
chance absence of potential dominant tree species
flagrations. In what were once actively exploited
on this island and the nature of the soils formed on
forests there has been an increase in tree height
very compact acid rocks. This landscape is thus not
and density and thus a dramatic increase in woody
a degraded landscape linked to the removal of the
biomass (e.g. Debussche et al. 1999). In the absence
forest, although the spread of Q. ilex is probably due
of fire, the understorey vegetation of pine forests
to human activities and opening of the deciduous
in the Mediterranean changes gradually as woody
oak forest (Reille 1992).
shrubs such as Phillyrea, Viburnum, and Rhamnus
increase in abundance before the eventual appear- Grove and Rackham (2001) enumerate numerous
ance of oaks (Barbero et al. 1987a). This increase other examples to back up their doubt about ‘any
in woody biomass has no doubt contributed to the theory that the normal state of wildwood was
increased risk of fire in many Mediterranean forests trees upon trees upon trees . . .’ (p 153). For these
(Le Houerou 1987; Moreno et al. 1998), a trend which authors the critical changes in vegetation change
climate change may exacerbate in the future (Piñol during recent history involved two main processes:
et al. 1998). the advance of agriculture and pasturage in the
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 29

Bronze Age followed by a gradual drying of the During the twentieth century, no one can deny
climate, whose effects on the vegetation were com- that changes in human activities and land-use pat-
plete 2,000–3,000 bp. Evidence for Holocene vegeta- terns closely tied to socioeconomic developments
tion changes synchronous with climatic changes and political changes have had major impacts
support this idea (Beug 1967, 1975). There is no on the vegetation and ecology of many regions
doubt an element of truth in their argument. It is around the Mediterranean. Since the late nine-
difficult to precisely identify the structure of a pris- teenth century, human activities have had variable
tine landscape prior to human impacts because of consequences depending on whether one observes
the spatial and temporal variability of the abiotic vegetation on the southern or northern shores of
environment already present prior to the arrival of the Mediterranean and depending on whether one
humans. Landscape patterns were already natural observes littoral of hinterland vegetation. On the
mosaics prior to human activities, and, as Lepart and northern shores of the Mediterranean, forests are
Debussche (1992: 79) point out ‘in the Mediterranean spreading in the back country (e.g. Barbero and
region, it is a myth to think that homogeneous Quézel 1990; Debussche et al. 1999; Arianoutsou
forest was the essential element of the landscape 2001; Chapter 4) whereas littoral vegetation goes
before the arrival of humans’. But since then it is under concrete due to peri-urban and holiday
clear that in many areas humans have had a crit- resort development and sprawl. To the south of
ical effect on the flora and patterns of vegetation the Mediterranean Sea, the search for arable land
(Box 1.5). is ongoing, reducing natural forests to isolated
Perevolotsky and Seligman (1998) argue that trees and contributing to continued degradation of
although traditional grazing practices may well natural ecological systems.
have caused forest destruction in many Medi- The low-lying hills and upland plateaux around
terranean areas, such activities may be an efficient the northern shores of the Mediterranean have
and ecologically sound form of ecosystem manage- incurred much rural depopulation, in some places
ment, one that in no way implies degradation. For local populations have declined to less than one-
these authors (pp. 1007–1008), ‘even heavy grazing fourth of their numbers at the end of the nineteenth
by domestic ruminants on Mediterranean range- century. Punctuated collapses of the market for
lands is a relatively benign factor in ecosystem some important crops due to silk worm diseases
function and seldom in itself irreversibly destruc- and Phylloxera on vines at the end of nineteenth
tive to the soil or the vegetation’. The long history century, for example, and other more general socio-
of grazing in the Mediterranean region (not the case economic causes have led to the abandonment of
in other Mediterranean climates where the recent agricultural practices (abandon of intensive terrace
introduction of mammalian grazers has had dra- cultivation and extensive sheep and goat grazing
matic impacts on the ecology of natural systems) and the reduction of wood cutting) around the
means that such areas may not be fragile to such northern shores of the Mediterranean. In associa-
grazing. In fact, where grazing and agricultural tion with these changes there has, over the last 100
practices have been abandoned, Mediterranean years, been a marked change in perception of forest
woodlands are rapidly spreading, with important cover and open vegetation (Lepart et al. 2000). All
consequences for the maintenance of a traditional these changes are recent, occurring in the twentieth
Mediterranean mosaic landscape of open vegeta- century and most dramatically since the end of the
tion and woodland (Debussche et al. 1999; Lepart Second World War.
et al. 2001). Their capacity to recover, even after hun- Clearly human activities have greatly impacted
dreds of years of heavy grazing, is high, although on Mediterranean vegetation, and continue to do
understorey herbs with limited dispersal and cur- so. I will leave others to argue about the general-
rent distributions restricted to small isolated pockets ity of this phenomenon and its relative importance
of forest, may take a long time to follow the spread of compared to climatic variation. What is of con-
woodlands. cern in this book is where and how changes in

land use, species introductions, and other activ- delimitation of six bioclimatic types:
ities have modified both ecological processes acting
(1) per-arid: P < 100 mm, 11–12 dry months;
within natural plant populations and their spatial
(2) arid: P = 100–400 mm, 7–10 dry months;
distribution in the landscape.
(3) semi-arid: P = 400–600 mm, 5–7 dry months;
(4) subhumid: P = 600–800 mm, 3–5 dry months;
1.4 Diversity and unity in (5) humid: P = 800–1,200 mm, 1–3 dry months;
the Mediterranean flora (6) per-humid: P > 1, 200 mm, <1 dry month (and
thus barely ‘Mediterranean’).
The contemporary Mediterranean climate, in which
annual precipitation varies from 100 mm in the most Most recent work on the Mediterranean excludes
arid parts of the region, to ∼3,000 mm on some of the areas which fall into the first of these types, which
mountains (which nevertheless have a dry summer), have a climate and a vegetation which is more typ-
and in which temperature varies greatly within and ical of a desert (Médail and Quézel 1997; Joffre and
among regions, has stimulated various classifica- Rambal 2002; Quézel and Médail 2003). These six
tions of climatic diversity in the Mediterranean. types can be further subdivided in relation to winter
Two main classifications can be identified. The first temperatures.
involves a series of bioclimatic types in relation to A second classification, developed by phyto-
a rainfall–temperature coefficient (Q2 ) developed sociologists, describes Mediterranean vegetation as
by Emberger (1930a, b, c). This bioclimatic coeffi- a series of ‘étages’ in relation to thermal differences
cient relates mean annual precipitation (P ) to the along altitudinal gradients (Table 1.1).
mean minimum temperature in the coldest month So wherein lies the unity? A first step in defin-
(m) and the mean maximum temperature in the ing the unity of the Mediterranean flora can
hottest month (M) by the following equation: Q2 = be taken with reference to the work of Pierre
2, 000P /(M 2 − m2 ). This coefficient allows for the Quézel, the French botanist and ecologist who has

Table 1.1 Vegetation classification in relation to altitude in the Mediterranean region

Etage m (◦ C) T (◦ C) Principal vegetation and locations

Infra-Mediterranean >7 Arid communities: Argania spinosa and Acacia. Only in western Morocco.
Thermo-Mediterranean >3 >17 Sclerophyllous communities: with Olea europaea, Ceratonia siliqua, Pistacia lentiscus,
Pinus halepensis, Pinus brutia and Tetraclinis articulata. Circum-Mediterranean.
Often as a narrow band near the sea and in valleys with Nerium, but can reach 800 m
in North Africa.
Meso-Mediterranean 0–3 13–17 Sclerophyllous forests of Quercus ilex (western and central) or Quercus calliprinos
(eastern) and pines. Littoral to 400 m to the north of the Mediterranean Sea,
∼400 to ∼1,000 m to the south.
Supra-Mediterranean −3–0 8–13 Deciduous oak forests dominant in the humid bioclimate (with Ostrya and Carpinus),
sclerophyllous oaks in zones with low rainfall. 400–900 m to the north of the
Mediterranean Sea, up to 1,500 m on the south side.
Mountain-Mediterranean −7 to −3 4–8 Upland coniferous forests with Pinus nigra and Mediterranean firs and cedars.
900–1,400 m to the north of the Mediterranean Sea, 1,400–2,000 m to the south.
Oro-Mediterranean <−7 <4 Open vegetation with xerophytic shrubs, Juniperus and sometimes open pine forest.
This belt is not always composed of Mediterranean taxa. Mostly above 2,000 m
(Atlas, Taurus).
Alti-Mediterranean Dwarf chamaephytes—Atlas and Taurus mountains above 2,200 m.

Notes: m: mean minimum temperature of the coldest month; T : mean monthly annual temperature.
Source: Ozenda (1975), Quézel (1985), Médail and Quézel (2003).
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 31

probably done more than anyone else to shape Taxa with an autochtonous origin are, not surpris-
ideas on the biogeographic origins and distribu- ingly, the main constituent of the Mediterranean
tion limits of Mediterranean vegetation. In a review flora. The strictly Mediterranean elements of the
of this subject, Quézel (1985: 18) proposed that flora developed and differentiated during the
the Mediterranean flora be viewed as ‘a hetero- Tertiary in association with the existence of iso-
geneous entity associated with a region that is lated microplates and climatic change during this
largely defined by climatic criteria’. Many elements period (Zohary 1973). Some genera are limited
of contemporary Mediterranean vegetation existed to contemporary areas associated with ancient
prior to development of the current-day climate, plates and some have greatly diversified within the
others became important components of the flora limits of the zone (Section 1.5). For example, the
since the onset of the Mediterranean climate in the Iberian peninsula has 16 palaeo-endemic genera
Pliocene. These different climatic origins introduce and is also centre of diversification for many genera
much diversity into the flora (Quézel 1985; Quézel (e.g. Genista, Narcissus, Linaria, Thymus, Teucrium,
and Médail 2003), which has several biogeographic and several Cistaceae). In the Balkans (set on the
elements. Apulian plate), diversification of genera such as
Taxa with subtropical affinities make up a sizeable Silene and Stachys, to cite but two examples, has
proportion of the Mediterranean flora (Raven 1973; been rampant, while in this region other genera, for
Quézel et al. 1980; Quézel 1985). Some of these example, Jankaea, Petromarula, and Haberlea, contain
taxa have probably evolved from ancestral stocks palaeo-endemic species.
present prior to the opening of the North Atlantic Two other entities which evolved on the eastern
and separation of the southern continental plates, and southern borders of the Mediterranean can
for example, Borderea and Dioscorea (allied to taxa be added. From the east came the Irano-Turanian
in South Africa or South America), Tetraclinis and group which developed during the dry and cold
Aphyllanthes (whose affinities lie with Australian glacial periods of the Pliocene and Pleistocene—the
taxa), and Cneorum which has related species in east- most significant taxa being Artemisia, Ephedra, and
ern North America. Alternatively, others had ances- Salsola, and trees such as the Judas tree (Cercis
tors present in the circum-Mediterranean flora of the siliquastrum), the storax tree (Styrax officinalis), and
Oligocene and Miocene (e.g. Ceratonia, Chamaerops, some oaks. Most of the species in the Irano-Turanian
Jasminum, Olea, Phillyrea, and Nerium). Many of element have centres of diversity in the semi-arid
these taxa represent palaeo-tropical relicts that have steppes of central Asia, that is, a continental
persisted and evolved in situ as the climate became climate. The different species in this element prob-
Mediterranean. There are also clear links with ably penetrated the Mediterranean region during
the more semi-arid and arid flora of East Africa and episodes of climate change and geological activity
the Cape Province of South Africa. The existence since the Tertiary (Zohary 1973). To the south, the
of generic pairs between the Mediterranean and Saharo-Arab element differentiated from a xerophytic
Cape floras (e.g. Thymelaea with Passerina, Echium and heterogeneous ancestral stock, and as a result
with Echiostachys, and Iris with Moraea) support the several North African endemics have affinities with
idea of an ancient origin for many disjunct distribu- Saharan and Arabian taxa.
tions (Goldblatt 1978; Quézel 1985). Likewise, sev- The final group concerns Holarctic or Eurasiatic
eral genera (e.g. Pistacia, Anemone, Ceratonia, Coris, elements of the flora. One part of this group con-
Cyclamen, and Globularia) whose centres of diversity cerns taxa from the Laurasian flora present prior
occur around the Mediterranean also have isolated to the Miocene, which are now localized in parts
species in more arid parts of Africa (Quézel 1995), of the eastern Mediterranean (e.g. Aesculus hip-
indicative of historically more widespread distribu- pocastanum, Forsythia europaea, Liquidambar orientalis)
tions and/or migration between the ancestral or on islands (e.g. Zelkova abelicea on Crete and
Mediterranean region and parts of Africa, including Z. sicula on Sicily) that were little affected by
the east and as far south as the Cape Province. periods of glaciation. A second component includes

Figure 1.3 The areas of the Mediterranean Basin (shaded areas) recognized as ‘hot spots of biodiversity’ by Médail and Quézel (1997),
where rates of endemism exceed 10% of the local flora (reproduced with permission).

the many temperate Eurasiatic taxa that inhabit the of biodiversity within the Mediterranean region
Mediterranean mountains, where they found refuge (Fig. 1.3). These hot spots are areas where species
during periods of glaciation. Another part of this diversity and endemism are high. The proportion of
group, the Pontic entity has taxa which evolved endemic taxa in these hot spots varies from 7 to 10%
under a continental climate with a dry period, for on some of the islands to ∼20% in Andalucia and
example, Stipa, which no doubt migrated into the the Rif of North Africa, up to 27% in the moun-
Mediterranean during periods of glaciation. Finally, tains of Greece, and up to 31% in Turkey. These
a small number of arctic–alpine taxa, rarely with any centres of diversity occur where geological and cli-
endemic species in the Mediterranean, can be found matic histories have played crucial roles in shaping
on a few high mountain tops (e.g. in the Atlas and plant diversity. Based on data from Médail and
Taurus and on Corsica). Quézel (1997) and as a foretaste of things to come
To recapitulate, with regard to the diverse (Chapters 2 and 3), I now describe the main fea-
origin of the contemporary elements of the tures of plant diversity in relation to history and
Mediterranean flora, I am tempted to cite that setting in three major poles of diversity: (a) the
old adage of Shakespearian literature (used by Iberian Peninsula, (b) the Balkans and the Aegean,
other authors in their evaluation of plants and and (c) Anatolia and Cyprus.
their diverse strategies) . . . while some species were
born Mediterranean, others became Mediterranean 1.5.1 The Iberian peninsula
as they colonized the region, while others had
A complex geology, a beautiful and diverse scenery,
Mediterranean life thrust upon them! Many, but
and a spatially heterogeneous climate character-
not all, of the latter have now gone. Together
ize the Iberian peninsula. Other than the northern
these diverse elements make up the contemporary
fringe and the north-west corner of the penin-
Mediterranean flora.
sula, the climate is Mediterranean. Spain and
Portugal sit on a hard crystalline core of very
1.5 Centres of diversity: concordance
ancient and now partly metamorphosed rocks.
with history
Similar formations also occur in Galicia and the
In a rousing testament to the immense wealth Massif Central of south-central France. Around this
of the Mediterranean flora, Médail and Quézel ancient core there have been successive periods
(1997) proposed the delimitation of ten ‘hot spots’ of uplift, subsidence, folding, and faulting. These
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 33

processes have moulded the current day geology 6. The coastal plains, adjacent to and sometimes dis-
of the Iberian peninsula and the surface processes sected by coastal mountains. The Iberian peninsula
which have fashioned the current day landscape has more than 2,500 km of coastline, however
(see Polunin and Smythies 1973). In the Late this natural habitat has been decimated by tourist
Miocene, the Iberian plate was roughly in its developments.
present position and the main mountain ranges 7. The physical diversity of the peninsula is further
were in place. The Balearic islands, a continua- enhanced by the highly localized occurrence of rare
tion of the Betic Cordillera, were then connected substrate types such as serpentines and gypsum.
to the mainland extending the peninsula to the
The north-west tip and a narrow band of the
Iberian peninsula along the north coast do not
The contemporary Iberian Peninsula thus has
have a Mediterranean-type climate. In the extreme
a number of recognizable features.
south-west, facing the Atlantic, the Mediterranean
climate prevails with more than two months of
effective drought. In the centre of the peninsula,
1. The crystalline massifs in the north-west of the
extreme heat in summer and freezing in winter are
peninsula (particularly in Galicia), most of which
locally enhanced where mountains and high eleva-
do not have a Mediterranean climate and have
tion allow for extreme cold and create rain-shadows
relatively low rates of endemism.
(<300 mm rain may fall in a given year and drought
2. The central plains (mesetas) where more recent
may extend for up to seven months in some parts
Tertiary sediments overlie the ancient core, form-
of the central plateau and the south-east rim of the
ing vast plateaux comprising limestone, sandstone,
clays, and marls. This area is home to the Spanish
The diverse geology, substrates, and climate of the
salt-steppes and their unique vegetation, whose
Iberian peninsula are paralleled by its botanical rich-
species often have their closest relatives as far away
ness: 1,258 species or subspecies are endemic to this
as the salt-steppes in Turkey, the Caspian Sea, or in
region (Gómez-Campo and Malato-Beliz 1985). Half
North Africa.
of these belong to only 27 genera, while the other
3. The old-fold mountains (Iberian mountains,
half are distributed across no less than 286 genera.
Central Sierras, and Sierra Morena) formed directly
This illustrates the phylogenetic overrepresentation
from the ancient granite core. They are home to
of certain genera (in this case Hieracium, Centaurea,
many mountain plants.
Linaria, Saxifraga, Armeria, Sideritis, and Thymus all
4. The new-fold mountains. Once part of the ancient
have >20 endemic species in the region). To these
Hercynian massif, these represent the highest sum-
should be added the ∼500 Iberian endemics whose
mits of the peninsula (almost 3,500 m in the Betic
distribution also extends into the Rif in northern
Cordillera). These are young rocks, mostly sedimen-
Morocco, where the flora has close similarities to
tary in origin, formed by up-thrusting 50–100 Ma
that of Andalousia.
around the margins of the ancient core. Such places
often contain alpine plants in refugia where they
have either persisted since separation from their
1.5.2 Mountain configuration and island
closest relatives in the Massif Central or the Alps
isolation in the Balkans and the Aegean
or diverged more recently.
5. The depressions where subsidence has occurred The foundation of the Balkan and Aegean area is an
against the hardcore of the ancient rocks during ancient crystalline massif probably uplifted in the
Alpine orogeny. They contain the main river valleys Carboniferous period. This massif has two zones
(Guadalquivir, Tagus, Ebro, Guadiana, and Duero) (Polunin 1980). First the high mountains of the
whose watersheds drain the peninsula. Flooded Rhodope massif (culminating above 2,000 m) which
with Tertiary deposits these now form low-lying stretch across southern Bulgaria and Macedonia
plains. with two southerly extensions, one down through

eastern Greece (including Mt Olympus which occurs on the small islands south of Naxos. Granite
reaches 2,911 m) into the Euboea peninsula and the outcrops occur sporadically, for example, on Naxos,
other through Thrace to the islands of Thasos and Mikonos, and Tinos.
Samothrace in the north-east corner of the Aegean. In the Early Pliocene, the Aegean area resembled a
The second area, formed of a crystalline block, lies large continent extending from the Greek mainland
in the south-central part of the Aegean, where the across to Turkey, via the Cyclades and Rhodes, and
remains of the summits and crests of the ancient an arc of land connected the Peloponnese peninsula,
mountains are now the islands of the Cyclades and Crete, Karpathos, and Rhodes. The Cyclades and
some of the continental peninsula. Crete were separated by a sea, and have probably
During Alpine orogeny, new fold mountains been so ever since, and a large lake covered most
composed of Secondary and Tertiary sedimentary parts of the present North Aegean Sea. During the
limestone, sands and conglomerates were thrown Pliocene, Greece and the Cyclades became isolated
up to create several more recent mountain ranges. from Turkey, Rhodes, and other eastern islands by a
These stretch in a north-east to south-west direction sea extending northwards. What are now islands off
from the Dinaric Alps down the east coast of the the east coast of Turkey (such as the large islands of
Adriatic, through the Pindhos of central Greece and Kos, Patmos, Ikaria and Chios, and the many smaller
the Taiyetos and Parnon ranges in the Peloponnese islands in between) were probably connected to one
peninsula, forming an arc across the island of Crete another and to the mainland. At various times dur-
and up across the islands of Karpathos and Rhodes ing this period Crete was a series of 3–5 isolated
to the Taurus Mountains in Turkey. During the same mountain ranges rather than a single island.
period the Rhodope massif to the north uplifted to The origin of the islands of Crete and Rhodes,
form the Balkan Mountains of Bulgaria and Serbia. on the Tertiary arc of fold mountains that link the
Two main groups of rocks occur in these different Peloponnese peninsula to the mountains of south-
chains: west Turkey, around the Aegean margins, is more
recent than that of the islands of Corsica and Sardinia
1. Non-calcareous shales, sandstones, and marls
(Biju-Duval et al. 1976; Mascle and Rehault 1991).
which originated under shallow water and have
The multitudinous islands in the south and east
in some areas formed slates, quartzites, and con-
Aegean have had even more recent connections
glomerates and which contain many rivers and
among each other. During the first cold snap of the
early Pleistocene (∼2–1 Ma) (and perhaps later dur-
2. Calcareous sediments of deeper clearer water that
ing successive glacial maxima) further regression of
have given rise to the limestone and marble that
the sea (a drop in sea-level 150–200 m below present
characterize many areas in this region, notably the
levels occurred during the Riss glacial period) made
old Mesozoic limestone that forms the distinctive
land-bridge connections possible from Greece across
Karst scenery in Dalmatia and the many cliff habitats
the Cyclades, although the big islands probably
in the islands of the Aegean.
remained isolated. Again there would have been
In the Late Tertiary and Early Quaternary, the connections between many of the eastern islands
Tertiary fold mountains underwent much faulting off the west coast of Turkey with each other and
and shattering and further uplift contributed to the the Turkish mainland. During the successive glacial
formation of many of the escarpments and cliff maxima of the Pleistocene, connections between
formations that can be seen in the region. Although islands where the sea is <100 m deep, that is, many
some of the Aegean islands are comprised of mainly of the Cyclades, no doubt occurred.
volcanic rock (e.g. Milos and adjacent islands), So when one looks at a map of the Balkans and
most are crystalline schist of ancient origin. Hard the Aegean in relation to Turkey, it is not complete
crystalline limestone occurs on many islands where fantasy to envisage three main land-bridge con-
they create the cliffs and gorges that are famous in nections across the area: one across the Bosphorus
the region. Softer marly limestone is rare and only to the north of Turkey, one through the Cyclades
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 35

in the centre (perhaps with a link as far as Rhodes) to produce earthquakes and intense faulting and
and one around the arc of Tertiary Mountains from uplift have created some of the highest moun-
the Peloponnese to Crete, Karpathos, Rhodes, and tains of the Mediterranean region. These mountains
south-west Turkey. Deep basins north and south of are comprised of long narrow chains mostly ori-
the Cyclades separated these connections. As for ented either east–west (inland from the Aegean
the Iberian peninsula, the Balkans, including Greece coast) or north–south (to the south-east) and mostly
and its islands, have an amazing climatic diver- composed of sedimentary strata deposited in the
sity. Annual rainfall varies from <500 mm in the ancient geo-synclinal Tethys Sea. The mountain
south-east to over 2,000 mm at high elevation in the chains are separated by deep often broad valleys
north-west mountain ranges. or high plateaux (1,500–2,000 m elevation across
The richness of the flora of this region is one of vast expanses). Central Anatolia is a rolling plateau
the most diverse, for a given area, of the whole which stretches across vast expanses at 900–1,200 m
Mediterranean region, containing 6,500–7,000 elevation and is made up of Mesozoic hard lime-
native species, of which ∼1,500 are endemic to stone and Tertiary marls and soft chalks. The com-
the region (Polunin 1980). The arc of Tertiary fold plex coastline of the eastern Aegean Sea, comprising
mountains that delimit the southern part of this many islands and estuaries was formed by the
region contain particularly high rates of endemism foundering of the central part of the Aegean and
(Strid and Papanicolaou 1985). On Crete, which complex faulting in the Turkish mountains.
runs 245 km in an east–west direction along this The elevation gradient that exists in the region
arc, 70% of species are of Mediterranean origin. means that annual precipitation varies more than
This cortège of Mediterranean taxa includes many threefold over small distances (300–1,000 mm in
woodland and maquis species from the western southern Turkey). Within the region there is a dis-
Mediterranean, for example, Quercus coccifera (one tinct decline in rainfall from west to east and from
can see real trees of this species on Crete!!), Q. ilex, north to south and the Mediterranean vegetation
Pistacia terebinthus, Arbutus unedo, Erica arborea, and at low altitude is more xerophytic than that of
Euphorbia dendroides. Another 20% have their origin the western Mediterranean. Although most authors
in the Irano-Turanian floristic province. Compared include most of Turkey in the Mediterranean region,
with Cyprus (see below), the importance of the once one gets over the coastal mountains and moves
Irano-Turanian element in the island flora is of inland, the climate rapidly resembles a more conti-
minor importance. Crete has 150–170 endemic taxa nental regime. There is thus high spatial variation in
(Zohary 1973; Médail and Quézel 1997), which the climatic regime in this region, further details of
represent around 12% of the flora. A roughly equal which can be gleaned from Zohary (1973).
number of these endemic taxa have affinities, that is, The main centres of endemism in Turkey are
purported sister taxa, with either the Greek main- the mountains, particularly the Taurus, Armenian,
land or Turkey (∼25 species in each case). About and Kurdo-Zargrosian ranges and also the vast
40 of these endemic taxa are related to wide-ranging expanses of steppe vegetation on the central
species across the eastern Mediterranean including plateaux areas (Zohary 1973). The vegetation con-
Greece. Rates of endemism on Crete are almost tains many Mediterranean elements and even the
identical to those on the Peloponnese peninsula, transition zone with the Irano-Turanian floristic
where 12.5% of the flora is endemic. province includes many species that one can observe
in the strictly Mediterranean part of the territory.
As a result, in Anatolia the distinction between
1.5.3 The Anatolian peninsula and Cyprus
the Mediterranean territory and the eastern steppes
A major centre of diversity and endemism in the that spread across Asia is perhaps less marked
eastern Mediterranean concerns the mountains and than the distinction between the Mediterranean
coasts of southern Turkey. In this region, volcanic and Euro-Siberian and Atlantic provinces in the
activity is still in progress, subduction continues western Mediterranean. The precise delimitation of

a Mediterranean region is particularly difficult in spite of repeated extinctions and glacial impover-
some areas of western Turkey where the mountain ishment, the current-day flora is immensely rich,
ranges run east–west and where broad valleys are largely because of its heterogeneous origins, notably
open to the migration and persistence of steppe from the ancient floras of the tropical and temper-
elements from the Irano-Turanian floristic province. ate regions between which it has formed a sort of
The Mediterranean flora of Turkey contains ‘ecotone’ since at least the Cretaceous, and the dif-
around 5,000 species of which ∼30% are endemic to ferentiation of taxa in situ. These diverse floristic
the region, including 12 monotypic, endemic gen- origins have come together to create a diverse and
era (Zohary 1973; Demiraz and Baytop 1985). There spatially heterogeneous flora, unique on earth.
are two important features of this diversity. First, The geological history illustrates all too clearly
the presence of a large number of genera which that the probabilities of dispersal and gene flow
are very rich in species (sometimes several tens of among populations may, at certain times, have been
species), such as Astragalus, Trifolium, Centaurea, radically different to what they are at the present
Salvia, Stachys, Alyssum, Silene, and Dianthus to name time. Major geographical and geological barriers
but a few. In the genus Ebenus, all 14 species are have limited dispersal and created phylogeographic
endemic to Turkey. As a result of this radiation of divisions in many plant groups at various moments
a relatively small number of genera, endemism in during the history of the region. In particular,
the Anatolian peninsula reaches one of the highest the dispersal and fragmentation of microplates dur-
rates of endemism of any region on earth. Some- ing the Oligocene and Miocene initially provided a
where around half of the endemic taxa have evolved template for the evolution of many endemic species
from Irano-Turanian elements that have colonized in the Mediterranean.
the eastern Mediterranean. Second, the occurrence The development of contemporary
of a relatively large number of tree species (more Mediterranean-type vegetation occurred in the
than 300). Late Pliocene as the summer-drought regime set in.
The region has one main island, Cyprus. The geo- Prior to this time, Mediterranean sclerophyllous
logical foundation of this island is an extension of vegetation existed as isolated taxa adjacent to
the Taurus/Amanus folded system. Cyprus is flor- other vegetation types, but not as the character-
istically related to Syria, with affinities to southern istic vegetation. Since the Pliocene, the onset and
Turkey. On the island, 1,620 species occur of which development of the highly seasonal Mediterranean
roughly 10% are endemic. Unlike Crete, the Irano- climate has been more closely related to the
Turanian element is the dominant element of this precipitation regime than to an alternation of
flora. temperatures. Even during glacial maxima, aridity
probably had a major effect on species distributions.
Future climate change may thus have significant
1.6 Conclusions
effects on Mediterranean vegetation principally
The Mediterranean flora has had a complex and as a result of modified precipitation regimes, for
eventful history. The evidence is overwhelming: example, if the summer drought regime extends for
Mediterranean plant communities have continu- longer periods during the year via the modification
ously undergone profound compositional changes, of spring rainfall. Strong regional contrasts in
some gradual, some more abrupt, some episodic, precipitation have occurred over a long period of
and some repeated. The vegetation history of the time, even before the onset of a Mediterranean
Mediterranean has been strongly marked by succes- climate, adding to geological spatial heterogeneity.
sive extinctions of whole taxonomic groups during My synthesis of historical events points to the occur-
the Middle Miocene, the Middle/Late Pliocene, and rence of a sort of palaeo-mosaic in Mediterranean
the Early/Middle Pleistocene. A point to note here habitats. A spatial environmental mosaic has indeed
is that extinctions will have involved not just plants probably been present in many landscapes through-
but also their pollination and dispersal agents. In out the history of the Mediterranean Basin due to
H I S TO R I C A L C O N T E X T O F D I V E R S I T Y 37

local geological and edaphic variation and spatial combined with the evolution and oscillations of
variation in climate (Lepart and Debussche 1992). the climate, which forced plant migration, caused
Such historical spatial heterogeneity may thus have successive contractions and expansions of species’
had lasting effects on population differentiation distributions and imposed strong selection pres-
and species divergence in the Mediterranean Basin. sures on plant traits, are the natural foundations of
More recently, human activities have created even plant variation and evolution in the Mediterranean
more variation in a landscape already structured region. On to these can be added the powerful
by spatial heterogeneity of environmental factors. and longstanding influence of human activities.
In some areas, the spatial mosaic of environmental Ever since Neanderthal spread westwards from Asia
factors has determined the extent and impact of Minor, pre-empting the subsequent direction of
human activities, in others, human activities have spread of domesticated plants, the Mediterranean
reinforced the natural variation in the landscape. landscape has been deeply altered and re-modelled.
The three panels of my historical triptych (Box 1.1) Human activities have created massive contempor-
form the essential framework of this chapter. In ary changes in the spatial configuration of habitats
the rest of this book I will explore how these three and the ecological conditions within them, but have
components of regional history have moulded and also been constrained in their impact by the pres-
sculpted evolutionary processes acting on distribu- ence of spatial environmental variation imposed
tion patterns, differentiation, and divergence in the by geology, soils, and climate. The long history of
Mediterranean flora. Other authors have outlined human impacts should not be ignored as we try to
the roles of geology, climate, and human activities understand plant evolution in theMediterranean.
in the shaping of species diversity and distribu- Despite the dividing lines I have placed between
tions in the Mediterranean region (Zohary 1973; the three panels of Box 1.1, the boundaries between
Heywood 1995; Quézel and Médail 2003). In this them are fluid. Examination of the nature of the pro-
book, I will use these themes as a framework to dis- cesses involved in each of the three panels, and the
cuss the process of plant evolution, both within and interactions among them, allows us to understand
among species. A complex geological history, which the development of narrow endemism, which as
introduced diverse biogeographic origins to the I will explore in Chapter 2, is one of the most char-
flora and set physical limits on species distribution, acteristic features of Mediterranean plant diversity.

The biogeography and ecology of


Correlations between the palaeogeography of the Mediterranean and the flora of the island of
Corsica . . . [illustrate] . . . not only the diversity of biogeographic origins but also the remarkable
specificity of this flora.
J. Contandriopoulos (1990: 414–415, my translation)

The first step towards understanding these issues

2.1 Narrow endemism: the cornerstone
is to recognize that endemism has multiple causes
of Mediterranean plant diversity
and that a diversity of factors can influence vari-
The closely interrelated questions of what factors ation in range size. These factors include geological
limit range size, why species differ in relative abund- barriers to dispersal, modification of distributions in
ance within and among sites and which factors association with climate change, genetic factors such
regulate species diversity within local communit- as hybridization, polyploidy, and mating system
ies are central to the discipline of ecology. Over evolution which can greatly affect the capacity
the years, research on these issues has developed for adaptation to new ecological conditions, inter-
a solid understanding and a rich literature. In this actions with and/or absence of pollinators and
chapter, I will primarily be concerned with the first dispersal agents, and species traits which restrict
of these questions: what causes some species to dispersal. In this chapter, I recast some of the above
have restricted range size while others, even very questions in a Mediterranean setting in order to
closely related species, have more widespread distri- explore and evaluate the historical, biogeographic,
butions? This question has stimulated much interest and ecological features of endemic plant distribution
(Stebbins 1942, 1980; Kruckeberg and Rabinowitz patterns in the flora. I will take up the issue of how
1985; Gentry 1986; Major 1988), and a number evolutionary processes have acted to create patterns
of questions which relate to the causes of range of differentiation that are the template for endemism
size variation and endemism can be identified. in Chapter 3.
Are endemic distribution patterns simply a result The Mediterranean region is an ideal place to
of chance events during the evolutionary history study plant endemism. What is perhaps the major
of a species or group of related species? Have characteristic of the Mediterranean flora is the fact
endemic taxa evolved via ecological specialization that its well-known high overall floristic richness
at the range limits of widespread species or are (24–25,000 plant species) is to a large part due to
widespread species successful derivatives of their the high incidence of species turnover (β-diversity)
endemic congeners? Do endemic taxa have reduced and regional endemism. Rates of endemism in
genetic diversity, perhaps associated with differ- particular regions within the Mediterranean are fre-
entiation in an isolated situation? Are endemic quently above 10% and can exceed 20% of a local
species the relicts of previously widespread species flora (Greuter 1991; Médail and Quézel 1997). It
or new taxa that have not yet had time to spread? is in the different mountain chains, for example,


the Betic-Rifan complex on either side of the Straits just a single region within the Mediterranean. So
of Gibraltar, the Maritime Alps of south-east France, more than one-third of the native flora (∼37%) have
the mountains of southern Greece and Turkey, and restricted distribution patterns, they are narrow
on several of the islands which straddle previously endemic species. In addition, only ∼8% of species
connected mountain chains (notably Sicily, Crete endemic to the Mediterranean have a distribution
and Rhodes, Cyprus, Corsica and Sardinia, and pattern that encompasses more than five of the
the Balearic islands) that rates of endemism (Médail Med-checklist regions (for details concerning the
and Quézel 1997) and species diversity (Lobo et al. delimitation of these regions see Greuter 1991). In
2001) are at their highest. These zones of high contrast, 37% of non-endemic species have dis-
endemism correspond to zones of high tectonic tributions that extend to more than five regions.
activity and/or microplate fragmentation and isola- Narrow endemism is thus the integral component
tion and are areas which may have been less affected of endemism in the Mediterranean.
by more contemporary human activities. In this chapter, I will focus on narrow endemic
Somewhere close to 60% of all native taxa in plants whose distribution is restricted to a small
the Mediterranean region only occur in the Medi- region, most often one or a few adjacent islands or a
terranean, that is, are endemic to the region as a small and often distinct region of a continental area.
whole (Greuter 1991). An important trend within My objectives are as follows:
those species endemic to the Mediterranean is that
close to 60% are narrow endemic species, that • First, I discuss the diversity of endemic distribu-
is, species whose distribution is restricted to a tion patterns in the Mediterranean flora.
single well-defined area within a small part of the • Second, I explore the historical and biogeographic
Mediterranean region (Fig. 2.1). In contrast, only associations of endemic plants and set endemism in
28% of non-endemic species (i.e. species which also the context of the ecology and dynamics of the plant
occur outside the Mediterranean region) occur in communities in which they occur.

Number of endemic taxa in each range size category





1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
Range size (number of regions where a species is present)

Figure 2.1 The prevalence of narrow endemism in the Mediterranean region (drawn from data in Greuter 1991).

• Third, I evaluate the ecological and biological chapter, I use this definition in order to examine the
correlates of endemism. In particular, I examine biogeographical, ecological, and biological features
whether endemic species differ from widespread of narrow endemic plants within the Mediterranean
species in traits linked to ecological function and region.
habitat occupation. Endemic plants can be relicts or newly formed.
These two categories of endemic taxa are commonly
referred to as palaeo-endemic and neo-endemic
taxa, respectively (Favarger and Contandriopoulos
2.2 Endemism in the Mediterranean:
1961; Stebbins and Major 1965), or as Zohary (1973)
patterns and classification
preferred, primary (active) and secondary (relict)
A convenient place to start the treatment of a subject endemism. I will keep to the former terminology
is to define it. Endemism however is one of those which is more commonly used. Palaeo-endemic taxa
enigmatic issues where definitions are difficult to are ancient or relict elements of a given taxonomic
apply. The problem here relates to the relative group, often systematically isolated from other taxa.
and subjective nature of the phenomenon—plants Neo-endemic taxa are more recently evolved and
have distribution areas of all sizes and degrees have extant sister taxa. This duality of endemism,
of disjunction imaginable (Raven 1972), hence the which Stebbins and Major (1965) clearly highlighted
degree of endemism may increase as the size of in their analysis of endemic plants in California, has
an area increases (Major 1988). In consequence, been combined with karyotype variation to produce
defining endemism becomes highly problematic a fourfold classification of endemic taxa (Favarger
since ‘the scale of the investigation creates the phe- and Contandriopoulos 1961).
nomenon’ (Favarger and Contandriopoulos 1961:
384, my translation). Endemism is also a relative Palaeo-endemics. These are systematically isolated
phenomenon, since taxa may be endemic to a single taxa (isolated species in large genera or mono-
area (such as an island) or several now disjunct areas typic genera), which are clearly ancient and usually
that have a common history (e.g. an archipelago of show little variability. They are probably relict taxa
islands which once had land-bridge connections). that have persisted through long periods of time.
But let us not stop there, the relative nature of Although they do not necessarily occupy the region
endemism, far from being a major obstacle, actually in which they originated, they provide illustrations
renders its study all the more interesting because of of ancient lineages present in the Mediterranean
the novel insights it can provide for our understand- region, some since the Late Tertiary. Examples
ing of biogeography and speciation. include the monospecific genera of Naufraga balearica
Endemism has received several definitions. (endemic to Corsica and Minorca), Soleirolia soleirolii
Several authors have used the term endemism to (endemic to Corsica, Sardinia, Tavolara, and
describe taxa whose distribution is markedly more Majorca), and Nananthea perpusilla (Corsica and
restricted than the average distribution of taxa of Sardinia), and species with isolated positions
the same taxonomic rank. For Kruckeberg and within their genus, for example, Mercurialis corsica
Rabinowitz (1985: 448) ‘narrow endemic taxa are and Ruta corsica, both endemic to Corsica and
those that occur in one or a few small populations’. Sardinia. Palaeo-endemics may be of any ploidy
However, further on in their review, Kruckeberg level.
and Rabinowitz (1985: 451) recognize that ‘the term Patro-endemics. These are diploid endemic taxa
endemism, in its classical biogeographical usage, which represent the progenitors of now more
does not necessarily imply rarity or even small widespread polyploid entities. Examples of patro-
range’. Major (1988: 117) proposed that ‘a taxon is endemics on Corsica include Pinguicula corsica,
endemic if confined to a particular area through his- Chrysanthemum tomentosum, and Arrhenatherum
torical, ecological or physiological reasons’. In this elatius subsp. sardoum. In several patro-endemic

species which have given rise to autotetraploids, species’. There are multiple examples of this class
the diploids and tetraploids are not treated as of endemism in the Mediterranean flora (Fig. 2.2)
distinct species, primarily because of their mor- which illustrate the diverse patterns of isolation
phological similarity. For example, in both Dactylis and connections among different regions of the
glomerata on the Balearic islands and A. elatius Mediterranean.
on Corsica unambiguous distinction of diploids
from tetraploids requires cytological investigation. This classification illustrates a crucial point.
Strong reproductive isolation between the two Endemic species are not a homogeneous group,
ploidy levels can nevertheless occur due to the other than the fact that they can all be classi-
formation of sterile triploids when crosses are made fied as endemics because their distribution is lim-
between cytotypes. Given that the diploids are ited to a particular area. The different types of
ancestral, one should be careful not to treat them endemic taxa often co-occur in local floras, hence
as simple varieties or subspecies of an otherwise several authors have drawn attention to the impor-
polyploid taxon (Favarger and Contandriopoulos tance of the Mediterranean region for both (a) the
1961; Stebbins and Major 1965). conservation of ancient taxa whose distribution is
Apo-endemics. These are endemic polyploids relictual and (b) recent and ongoing diversifica-
whose distribution is a small portion of the range tion. In the Mediterranean flora, somewhere close
or a disjunct isolate of a more widespread ancestral to 75% of endemic taxa are neo-endemics, illustrat-
diploid. Apo-endemics are thus the reverse case ing the dynamic nature of speciation in the region, a
of patro-endemics. Sticking to Corsica, one can feature which conservation biologists should bear
cite a number of examples, including Cerastium in mind. The classification is however far from
soleirolii, Viola corsica, Genista corsica, and Cymbalaria being perfect and a number of points should be
hepaticaefolia. In some cases the diploid ancestor recognized when analysing patterns of endemism
is unknown. In addition, polyploid evolution and and the evolutionary processes which promote
endemism may be recurrent in particular diploids endemism.
in different areas. For example, multiple origins First, a given taxon may be classified into more
of endemic polyploids from a single widespread than one of the above classes. The example of poly-
diploid have been documented in Plantago subulata ploid complexes which contain different diploid
which has endemic polyploid variants in the Atlas subspecies with a common ancestor that has differ-
Mountains and the mountains of Corsica, and entiated in the different parts of its range could
in D. glomerata where different endemic diploid be classified as schizo-endemic taxa since they
subspecies have repeatedly given rise to endemic have the same ploidy level. Most of them are also
polyploids in the western Mediterranean (Lumaret the progenitors of new polyploids in each of the
1988). different regions and thus could also be classi-
Schizo-endemics. This class of endemics have fied as patro-endemic taxa. Indeed, the dichotomy
undergone differentiation due to the fragmenta- between old (relict) and recent (neo-) endemism is
tion of the range of a widespread ancestral taxon often presented without conclusive evidence con-
producing endemic taxa in different parts of the cerning whether a taxon has a relict distribution
original distribution. The resulting pattern is one of or is of recent origin (Stebbins and Major 1965).
disjunct distributions of closely related species with For example, Cyclamen balearicum, endemic to the
the same chromosome number. Schizo-endemic Balearic islands and southern France was considered
species may be of any age and degree of divergence as a palaeo-endemic species by Contandriopoulos
from the parental stock, which is often identifiable and Cardona (1984). However, its sister species are
on the basis of morphological and molecular tech- well known (Fig. 2.2; Chapter 3) and have the same
niques. As Stebbins and Major (1965: 5) quip … . ploidy level, indicative that this is a schizo-endemic
‘The older the divergence, the “better” are the species, albeit perhaps quite old.



5 6 3

500 km



500 km


2 3

500 km


1 2


500 km

Figure 2.2 Examples of schizo-endemic distribution patterns. (a) Mediterranean Scabiosa: (1) and (2) S. cretica, (3) S. hymetia,
(4) S. minoana, (5) S. albocincta, (6) S. variifolia. (b) Cephalaria squamiflora subspecies: (1) subsp. balearica, (2) as yet an unrecognized
variant which the authors call subsp. mediterranea (3) subsp. squamiflora. (c) Pinus nigra subspecies: (1) subsp. mauretanica, (2) subsp.
salzmanii, (3) subsp. laricio, (4) subsp. nigra, (5) subsp. pallasiana, (6) subsp. dalmatica. (d) Cyclamen subgenus Psilanthum:
(1) C. balearicum, (2)–(5) the different subspecies of C. repandum, with (2) subsp. repandum, (3) subsp. peloponnesiacum (which has two
almost allopatric varieties in the Peloponnese peninsula), (4) subsp. creticum, (5) subsp. rhodense. (a) & (b) Drawn from Contandriopoulos and
Cardona (1984), (c) Redrawn from Quézel and Médail (2003), and (d) Redrawn from Debussche and Thompson (2002).

The problem here is that the above classifica- endemic species have large populations and/or
tion says little about the processes underlying many populations in the region in which they
species differentiation and the evolution of endemic occur.
distributions. Some palaeo-endemic species will Another problem is that in many schizo-endemic
have originated as a result of similar processes groups of species, although ploidy level may
as more recent schizo-endemic or apo-endemic be constant, different endemic species may vary
species have done. Some apo-endemic species in chromosome number, structure, or types
whose diploid ancestors are extinct could also of rearrangements (Favarger and Siljak-Yakovlev
be considered as palaeo-endemic. Favarger and 1986). This variation in karyotype may be quite
Contandriopoulos (1961) suggest that the rates common in Mediterranean endemic species (see
of differentiation may vary among classes of Chapter 3). Hence, the category of schizo-endemics
endemism, being gradual in schizo-endemics or may encompass a rather diverse and heterogeneous
abrupt in patro- and apo-endemics, a suggestion group of endemics.
that is unlikely to be so simple and straightfor- A final point worth noting here is that in a given
ward. The conventional distinction between palaeo- region the diversity of endemic species may not only
endemics which are relict taxa and neo-endemics be high, but also may contain a range of endemic
which are recently evolved is of little impor- species with very different histories, some originat-
tance. What should be emphasized is the distinc- ing in the Late Tertiary, others evolving as the climate
tion between neo-endemics, where the genetic and oscillated during the Quaternary. The description of
phylogenetic relationships among sister taxa and endemism in the Maritime Alps (on the border of
their contemporary distributions can be established, southern France and Italy) by Barbero (1967) illus-
and palaeo-endemics which are systematically iso- trates all too well the localized co-occurrence of
lated, probably due to the extinction of ancestral and ancient and more recently derived endemic taxa and
sister taxa, and which also have a contemporary dis- the potential role of geology and climate during the
tribution that may not correspond to their site of different episodes of Mediterranean history. Similar
divergence. Neo-endemic taxa thus provide model analyses elsewhere in the Mediterranean would no
systems for the study of differentiation in relation to doubt produce comparable results. In a given place,
recent history and contemporary selection whereas different endemic species (even within a single
palaeo-endemic species provide the chance to study genus) reflect different episodes of Mediterranean
ancient divergence and persistence in relation to history.
geological history.
In addition, endemic distributions may arise in a
variety of ways (Stebbins 1950). These include long 2.3 Endemism in the Mediterranean:
distance dispersal, localized dispersal following community composition and
range restriction in almost contiguous areas, migra- biogeography
tion over historical land-connections, migration of
2.3.1 Endemism and community composition:
an ancestral stock into two different areas (with
islands and mountains
possible extinction in different parts of the range),
and of course straightforward isolation of two for- Understanding why some species have widespread
merly connected areas. The above classification in distributions and others are endemic to localized
some ways oversimplifies this diversity of causes. It regions requires appraisal of why species differ
should not be forgotten that endemism is a feature in relative abundance within local communities.
of distribution which says little about local popula- The abundance of species within samples often
tion characteristics in terms of abundance across the show regular patterns of distribution, and com-
regional landscape and numbers of individuals in munities regularly contain many species at low
local populations. An endemic species is not neces- abundance. This feature of community diversity
sarily rare in terms of local abundance, indeed some has long attracted the attention of ecologists and

continues to stimulate the development of testable Canary islands, or the Mascarene islands which
theories (Hubbell 2001). Species which are rare have never been connected to continental land
within sites, simply by virtue of their low abund- masses. The flora of the Mediterranean islands and
ance, may be less common at the regional and their adjacent continents were thus very similar
geographic scale, and thus have endemic distribu- prior to isolation, hence, the distance of islands from
tions. Within patches of habitat, common species each other and from continental areas, which may
are more common and rare species more rare than have a strong effect on species number on oceanic
one would expect because the latter are both more islands, is poorly correlated with species richness
extinction prone and less likely to colonize new for Mediterranean islands (e.g. Médail and Vidal
sites or areas where they have gone extinct than 1998).
common species (Hubbell 2001). Such patterns may The rate at which species increase in number with
occur because of random events associated with increasing area is one of the fundamental issues in
dispersal, colonization, and extinction, because community ecology and biogeography (McArthur
rare species have particular ecological requirements and Wilson 1967; Hubbell 2001). Species diversity
and thus only occur in certain types of habitat in any locality is a balance between local abiotic
which are themselves rare in the landscape, or factors and biotic interactions, which act to reduce
because rarity is associated with traits that reduce diversity, and immigration from other habitats and
the probability of movement across the landscape communities, which tends to increase local diversity
and thus the potential to expand range size. These and maintain diversity at equivalent levels across
issues are the subject matter for the rest of this the landscape (Schluter and Ricklefs 1993). In a
chapter. given region of relatively uniform climate, there
As outlined in Chapter 1, endemism in the commonly exists a close relationship between the
Mediterranean is highest in mountain ranges and size of the area and the number of taxa which
on islands. A feature of geological activity in the are present (Preston 1962a,b; McArthur and Wilson
Mediterranean is the discontinuity of land masses 1967). A common pattern is that species number (S)
and the heterogeneity of the landscape that have and area (A) are linked by a strong and recurrent
resulted from isolation by water, successive phases relationship with two fitted constants (c and z):
of mountain formation, and a variety of chemical S = cAz . A log–log plot of species number against
and physical substrate compositions. These pro- area produces a linear relationship of slope ‘z’. This
cesses have created strong geographic barriers to species–area curve is not just an inevitable rela-
dispersal. tionship, it provides a first step to understanding
Islands and archipelago systems have long fascin- how diversity and endemism are area-dependent
ated biologists, primarily because of their unique and closely related to the regional processes
and somewhat unusual faunas and floras (Darwin that act on community composition and species
1859; Wallace 1880; MacArthur and Wilson 1967; distribution.
Carlquist 1974). Although island floras may actu- The flora of local regions in the Mediterranean
ally have low levels of biodiversity in terms of provides some clear illustrations of the species–
species number, the proportion of endemic species area relationship across different spatial scales and
that occur on islands is unrivalled by continental in different biogeographic contexts. A significant
areas, illustrating the importance of isolation for relationship between species number and area can
the presence of endemic taxa. In the Mediterranean, be seen in the species composition of 48 small
islands are for the most part fragments of land islands in the Provence archipelago off the south-
that have become isolated due to their separa- east coast of France (Fig. 2.3(a)), for 7 large islands,
tion from continental areas (exceptions include the namely Corsica, Sardinia, Crete, Cyprus, Malta,
Aeolian islands off the coast of Sicily and south- Sicily, and the Balearic islands (Fig. 2.3(b)), surface
ern Italy). Most Mediterranean islands are thus very area in Andalusia in southern Spain (Fig. 2.3(c)),
different from oceanic islands such as Hawaii, the and in the Mediterranean regions of 17 continental

areas in different countries (Fig. 2.3(d)). The study is closely correlated not only with area but also
of the Provence archipelago by Médail and Vidal with maximum elevation. Of key importance is thus
(1998) (Fig. 2.3(a)) showed that the cornerstone the presence of mountains and in particular the
of this relationship is habitat diversity, which mid-elevation ecotone zones on the slopes of these
increases on larger islands, which have more alti- mountains, as previously suggested (Stebbins and
tudinal variation. As a result, taxonomic diversity Major 1965).

(a) 7
ln number of taxa

–4 –2 0 2 4 6 8
ln area (ha)

(b) 8 (c) 10
ln species number
ln number of taxa

6.5 2

6 0
5 6 7 8 9 10 11 –4 –2 0 2 4 6 8 10 12 14 16 18
ln area (ha) ln area (m2)

(d) 9 (e) 9
ln number of taxa

ln number of taxa

8.5 8.5

8 8

7.5 7.5

7 7
8 9 10 11 12 13 14 8 10 12 14 16
ln area (km2) ln area (km2)

Figure 2.3 Species–area relationships for islands and continental Mediterranean regions: (a) 48 small islands off the coast of Provence, (b) 7 large
islands, (c) southern Spain, (d) the Mediterranean sector of different countries, (e) whole countries with part of their land in the Mediterranean.
Data are natural logarithms (drawn from data in (a) Médail and Vidal (1998); (b), (d), and (e) Médail and Quézel (1997); (c) Ojeda et al. (2000a)).

There are several examples which illustrate that as area increases. If an area is greater than the size of a
the form of the species–area relationship varies local hotspot of diversity the slope of the relationship
across different spatial scales (MacArthur and will decline.
Wilson 1967; Shmida and Wilson 1985). The slope Second, dispersal limitation may control variation
of the curve is high on local spatial scales, relat- in the slope of the relationship, particularly for
ively low on intermediate (landscape or regional) island systems. On Mediterranean islands the rate
spatial scales, and is again high on very large (inter- of increase in taxonomic diversity with area is scale
region or intercontinental) spatial scales. Hubbell dependent, being greater for the 48 small islands
(2001) explores how at very local scales the species in the Provence archipelago (slope = 0.51) than for
accumulation curve is particularly sensitive to the the 17 larger Mediterranean regions on the contin-
abundance of species in local communities, as first ent (slope = 0.21), in accordance with the pattern
common and then rare species are encountered. On described above. Greuter (1991) also reported that
this local scale, the sampling size is smaller than within the Mediterranean the slope of the species–
most range sizes and dispersal limitation has a neg- area curve is greater than that for a similar range
ative effect on the rate of species accumulation. As a of areas in temperate Europe. In the Mediterranean
result, the relationship has a high slope. On regional the high incidence of narrow endemism means that
scales, the rate of encounter of a species depends small distribution ranges will allow for a greater
less on relative abundance and more on rates of dis- slope in the species–area curve for a given range of
persal and extinction and thus on species range size, areas.
which becomes smaller than the range of area sam- Insularity may strongly affect such patterns.
pled for more and more species. Dispersal limitation MacArthur and Wilson (1967) first pointed out that
thus has a positive effect as newly encountered the insularity prevents common species from being
(more dispersal-limited species) are encountered. present on a large number of islands since dis-
The overall result is that the rate of increase in the persal among islands is less probable than dispersal
slope of the species–area curve will decline. On among contiguous patches of mainland. High dis-
intercontinental scales, species accumulate as one persal rates will thus distribute common species
crosses major barriers to dispersal and enters distinct more widely on continental areas, where overall
biogeographical zones, with separate evolutionary species richness will be reduced due to a higher
histories, causing a sharp up-turn in the slope of rate of extinction of rare species across the land-
the S–A relation. The Mediterranean examples of scape (Hubbell 2001). On islands, overall species
the species–area curve also illustrate that the form richness may remain high since common species
of the relationship depends on spatial scale and will not displace rare species everywhere. However,
biogeographic context. as islands get smaller, the probability of extinction
First, the relationship observed for the Medi- increases, particularly for rare species which are
terranean regions of 17 countries (Fig. 2.3(d)) does likely to be lost from small islands more rapidly
not occur when diversity of whole countries is used than common species by virtue of their low ‘global’
(Fig. 2.3(e)). In the large North African countries abundance and reduced potential to re-colonize
(Algeria, Libya, and Egypt are the three points to sites. The fate of common and rare species is
the right in Fig. 2.3(e)) species richness of the south- thus context dependent. Three observations from
ern arid zones which occupy a large proportion of the Mediterranean flora are pertinent to these
their territory is low due to climatic constraints. So ideas.
although these countries have a large area, species
number does not increase consistently with area. First, palaeo-endemic species may be more preval-
Hence the importance of having a uniform climate ent in the endemic flora of islands compared
across the range of areas studied. A lower slope to continental areas (Verlaque et al. 1997). This
in continental regions may thus arise because new observation suggests that geographic isolation
habitats with low species diversity are encountered may be important not just because it promotes

genetic differentiation and the evolution of to a single ‘island’. They suggest that this result
endemism (Chapter 3) but also because it reduces is due to a higher competitive ability and lower
immigration of common species and thereby rate of extinction of polyploids. In addition, it
favours the persistence of endemic species on is possible that polyploids have a greater colon-
islands. ization capacity and thus a more widespread
Second, close inspection of the dataset for the 48 distribution.
Provence islands obtained by Médail and Vidal
(1998) shows that variation in taxonomic divers- Another important issue here is that endemic
ity among islands tends to decrease as islands plants may occur in a biological community
increase in size. If one compares the 24 largest which itself has a fairly unique and novel com-
islands with the 24 smallest islands, one finds position (Kruckeberg and Rabinowitz 1985). The
that among island variation in the number of communities of chasmophyte species on and around
taxa present on the small islands is twice that the limestone cliff faces in Mediterranean landscape
among large islands (F = 2.1; P < 0.05). This provide a clear illustration of this. To quote Polunin
may reflect greater amounts of random variation (1980: 44) ‘each gorge or cliff will have its own collec-
in presence/absence of taxa on small islands tion of species and many rarities are only to be found
(which have the least variation in ecological in such habitats’. Mediterranean cliffs often have
conditions). a high diversity of species, for example, limestone
Third, in a study of species composition in habitat cliffs in northern Spain have higher species divers-
fragments (‘islands’) in a landscape of cultiv- ity than equivalent areas on different substrata in the
ated fields in southern France, Lumaret et al. surrounding region (Escudero 1996).
(1997) reported that polyploid perennial herbs The islands of the Aegean are the place to see
occur on a larger number of such ‘islands’ than spectacular cliff formations (Fig. 2.4), Karpathos and
diploids, the latter being more often restricted south-west Crete being the most grandiose and well

100 km

Figure 2.4 The distribution of major cliff formations (shaded areas) in the Aegean archipelago (redrawn from Runemark 1971).

known. Chasmophytes are thus a notable element and Campanula subsect. Quinqueloculares, which
in the flora of the southern and central Aegean, are dominant components of cliff communities in
where the distribution of the habitat and the species the surrounding geographical regions have only
shows some intriguing patterns. More than 60 oblig- restricted distributions in the Aegean. These dis-
ate angiosperm chasmophytes are to be seen in these tribution patterns, the restriction of many species
cliffs, ∼50 of which occur primarily in hard lime- to limestone cliffs, and the taxonomic isolation of
stone cliffs. Most species occur in the maritime cliffs many of these chasmophytes suggest that the cliff
not far from the sea, some are endemic to the central flora is all that is left of a flora that once prospered in
Aegean (e.g. Helichrysum amorginum and Campanula larger and more connected areas of coastline in the
calaminthifolia), most are highly lignified and sev- Pliocene.
eral represent systematically isolated species (e.g. The fact that several chasmophyte species do
Eryngium amorginum and Linum arboreum). Several not occur on adjacent small islands, but cross
chasmophyte species have closely related species or phytogeographical divisions which span the region
occur as a single highly variable species with char- (Box 2.1), for example, with Crete where large areas
acteristic races in different parts of the Aegean. On of suitable habitat occur, suggests that community
some of the islands, for example, Ikaria, species composition and distribution may have a random
with affinities to the south, east, and west can component. Runemark (1969) argues that because
be observed together, indicative of a relict flora populations of chasmophyte species are characteris-
on an island once connected to different continen- tically small, new individuals probably arrive very
tal areas. The cliff communities on such islands rarely and the colonization of new sites is likely to
probably served as refugia for chasmophytes and occur by only a very small number of propagules.
other non-chasmophytes during the Quaternary Many species will thus have high local extinction
glaciations. risks because of their small numbers. So, irrespec-
Runemark (1969) illustrates how across the tive of the ecological preferences and competitive
Aegean Sea, the presence/absence of different chas- abilities of the species, the dynamic equilibrium
mophytes on different islands does not fit any between colonization and local extinction due to
particular geographic or ecological pattern. Even climate change and dispersal limitation may create
islands less than 10–20 km apart show marked dif- irregular patterns of distribution and abundance.
ferences in chasmophyte species presence. Many Runemark (1969) also points out that these
adjacent Aegean islands are separated by very short seemingly random patterns of distribution in chas-
distances (5–20 km in many cases), and the dis- mophytes are paralleled by similar patterns in the
tance between two adjacent central Aegean islands phrygana and sublittoral vegetation which occur
never exceeds 40 km. Despite repeated connec- on the Aegean islands. Many of the islands have
tions, dispersal limitation thus appears to have a normally developed phrygana, while others have
played a most effective role in shaping distribu- little trace of this vegetation type and a sub-
tion patterns in this areas. Small distances have littoral vegetation with a unique composition. For
been effective barriers to dispersal. Very few species Runemark (1969) it is not just random extinction fol-
reach the Greek mainland, yet many occur across lowing fragmentation that is important, low rates of
more ancient seas and occur on Crete or less often re-colonization due to ‘reproductive drift’ and fail-
on the eastern Aegean islands. None of the oblig- ure of most colonization events to establish a new
ate chasmophytes in the archipelago have an even population also play a crucial role in community
distribution pattern across the area and some of composition and distribution patterns. His expla-
the most widespread species in the archipelago nation is appealing and provides an instructive link
(e.g. Dianthus fruticosus and Scrophularia hetero- between the theory outlined above and observations
phylla) are hardly to be found elsewhere. Runemark of distribution patterns and community composi-
(1971) also points out that taxa such as Inula candida tion on Mediterranean islands.

Box 2.1 Biogeography of the Aegean islands

You do not have to pore over a map for long to isolation (Chapter 1), have created a patchwork
realize that the islands and continental margins of system well-adapted to the study of the limitations
the Aegean Sea provide a fascinating situation to on species distribution and the biogeography of
study the biogeography and community ecology of endemism.
endemic plants. The archipelago of multitudinous The islands of the Aegean Sea can be grouped
islands (∼200) of different size (∼100 have an area into four main sectors which reflect historical
<1 km2 ) within a sea surrounded on three sides connections and migration routes across
by continents with different histories and land-bridge connections and subsequent repeated
biogeographic connections, and across which periods of isolation (redrawn from Runemark
there has been a history of connections and island 1969).

100 km

1. The western small islands with very close Rhodes) where rates of endemism
affinities to the Greek mainland. are high.
2. The eastern islands with affinities to Turkey and 4. The Cyclades archipelago which occur on the
Thrace (due to the presence of a deep channel and eroded summits and crests of the ancient
more permanent sea in a north–south direction in mountains that once dominated this region. On
the eastern Aegean. these islands elevations reach 1,000 m on Naxos
3. The islands of the southern arc of Tertiary fold and a little less on Andros (944 m) but elsewhere
mountains (including Crete, Karpathos, and do not exceed 600 m.

In the Mediterranean it is not islands but regions 100

with high mountains which have the highest rates

% of all endemic taxa

of endemism (Médail and Quézel 1997; Verlaque 80
et al. 1997). For example, endemic taxa on lime-
stone in Mediterranean shrublands in southern
Spain are more common above 1,000 m elevation 40
than on similar bedrock at lower altitude in the
mountains of southern Spain (Arroyo and Marañón 20
1990). Where mountains occur as islands in the
sea then endemism is even more prevalent, CO CO + SA
for example, on Corsica, where the percent- Endemic distribution
age of endemism reaches 35% in the mountain
flora. The comparison of the affinities of taxa Figure 2.5 Percentage of taxa endemic only to Corsica (CO) or
endemic to Corsica and those endemic to both to both Corsica and Sardinia (SA) which have affinities to the
Corsica and Sardinia further illustrates this point Mediterranean (shaded part of bar) or non-Mediterranean (open bars)
(Contandriopoulos 1990). Whereas two-thirds of element of the flora (drawn from data in Gamisans 1999).
taxa whose endemism is limited to Corsica are
probably of non-Mediterranean origin (most of 1,600
these taxa have Eurasiatic affinities) and only 1,400
Number of taxa

one-third are of Mediterranean origin, for taxa

endemic to both Corsica and Sardinia, the pattern 1,000
is reversed, the majority have Mediterranean affin- 800
ities (Fig. 2.5). If Sardinia had mountains similar in 600
elevation to those of Corsica, the affinities between 400
the flora of the two islands would certainly be 200
tighter. 0
Associated with the increased rates of endemism LI ME SM MO OR SA AL
in the mountain flora of Corsica is a decline in [9.2] [6.7] [13.3] [21.4] [51] [32.5] [45.3]
species diversity (Fig. 2.6). On Corsica, endemic Vegetation belt [% endemism in belt]
taxa have diversified in a relatively impoverished
mountain flora. Similar patterns occur on Crete Figure 2.6 Variation in the percentage of endemism in different
(Greuter 1972) and in the mountains of southern vegetation belts on the island of Corsica. Each histogram represents
the number of endemic (open portion) and non-endemic (closed
Spain, where most endemic species occur in species
portion) taxa in each vegetation belt. LI: coastal vegetation,
poor, open heathlands on mountain ridges (Ojeda ME: thermo- and meso-Mediterranean, SM: supra-Mediterranean,
et al. 1995). It is uncertain if Corsica has had MO: mountain-Mediterranean, OR: mountain, SA: sub-alpine,
any continental connections (suitable for plant AL: alpine (drawn from data in Gamisans 1999).
migration) since the Miocene when its rotation
with the other parts of the microplate was stopped
by collision with the Apulian plate. On Corsica, as Crete (Greuter 1972). Impoverishment in
most of the endemic mountain taxa have affin- the mountain floras of such islands is thus most
ities to alpine–arctic species, which probably date likely due to a long history of extreme isolation
to periods of climate change in the Late Tertiary during which extinction rates probably exceeded
(Contandriopoulos 1962). The migration of new immigration rates, the flora being in a sort of non-
taxa to Corsica, particularly the mountains, has equilibrium. Indeed in some places in the mountains
thus been very low for a very long period of time. of Corsica one can observe taxa that one would
The same may be true of other large islands such not expect to be present in a given community.

A certain disharmony may thus reign in the struc- (a) 40.0

ture of some communities on Mediterranean islands,
as proposed for oceanic islands (Carlquist 1974),
despite the fundamental differences in how the 30.0

% endemism
communities came to be on these two types of
islands. A potential consequence of low divers- 20.0
ity is reduced competition in some habitats which
may allow species to colonize new ecological condi-
tions. Gamisans (1991) has suggested that reduced 10.0
competition has allowed some endemic species
(e.g. Stachys corsica, Robertia taraxacoides, Cerastium
soleirolii, and Galium corsicum) to have a wider 0 1,000 2,000 3,000 4,000 5,000
ecological amplitude than that of related taxa else- Number of taxa
where. Such a wider range ecological amplitude
(b) 12.0
on islands has been reported for C. balearicum
on the Balearic islands compared to continental 10.0
France (Debussche and Thompson 2003), and is pre-
dicted in many texts on the biogeography of island 8.0
% endemism

On a geographic scale, centres of endemism tend 6.0
to be regions with high species richness (Fig. 2.7;
Chapter 1). Hence, there is a complex relationship
between diversity and endemism which depends on 2.0
the balance between (a) local abiotic and biotic inter-
actions which act to reduce diversity and (b) immig- 0.0
ration from other habitats and communities which 0 1,000 2,000 3,000
may increase local diversity. Resource diversity and Number of taxa
habitat productivity can have a strong effect on
species diversity since although theory predicts a Figure 2.7 Relationship between proportion of endemic species and
species diversity in the flora of (a) the Mediterranean area of 17
positive relationship between resource diversity and
countries, and (b) 7 large islands in the Mediterranean (drawn from
the diversity of coexisting species which compete data in Médail and Quézel 1997).
for these resources, decreased availability of nutri-
ents may prevent the most competitive species from
excluding others, and thus maintain diversity at
higher levels than in highly productive habitats a situation which provides insights into the histori-
(Tilman 1994). I discuss this issue in more detail later cal component of range size variation.
in the chapter. Endemic plants on the island of Corsica, studied
by researchers based primarily in Marseille and
Geneva over the second half of the twentieth century
illustrate this theme well; details of what follows
2.3.2 Endemic species with disjunct
can be found in Contandriopoulos (1962, 1990),
Gamisans (1999), and Gamisans and Jeanmonod
The discussion of endemism is often confined to (1995). The island of Corsica has a little over 2,100
species unique to a single piece of land, be it a moun- indigenous vascular plant species plus 400 differ-
tain range or an island. However, many endemic ent subspecies/varieties. Almost 130 of these are
species have distribution patterns which encom- endemic to the island, that is, 5% of the total
pass a small number of disjunct land fragments, flora. For species on Corsica, endemism extends

beyond the boundaries of this island to the other 50

islands and fragments of continent which have his-
torical land connections with Corsica (Chapter 1).

Number of taxa
In fact, the number of endemic species on Corsica 30
whose distribution does not extend beyond this
island is <50% of endemic species on the island 20

(Box 2.2). Two-thirds of the taxa which show 10

endemic distributions that extend beyond Corsica
also occur on Sardinia. The majority of these 0
disjunct endemic distribution patterns encompass Taxa endemic to the Taxa endemic to the
the different parts of the microplate which frag- Gymnesian islands Pityusic islands
mented and dispersed in the Oligocene and Miocene
(Chapter 1). A smaller number of species also Figure 2.8 The distribution patterns of endemic taxa in the Balearic
occur on the continental parts of the Hercynian islands. Not included in this graph are taxa whose endemic
massif, with which they have a longer history of distribution encompasses the Gymnesian and Pityusic islands
(18 taxa). S: single island endemics, G: taxa endemic to two or more
geographic isolation. An important point here is
Gymnesian islands, P: taxa endemic to two Pityusic islands, T: taxa
that where disjunct endemism involves the Balearic whose distribution extends to the Tyrrhenian islands and/or
and Tyrrhenian islands ancient endemism is the surrounding continents, I: taxa whose distribution extends to the
prevailing pattern, for example, palaeo-endemic eastern Iberian peninsula (drawn using the distribution patterns of
species such as S. soleirolii, N. balearica, Cym- 124 endemic taxa described by Alomar et al. 1997).
balaria aequitriloba, Delphinium pictum). In con-
trast, no palaeo-endemic species on Corsica have tend to be endemic to this subset of islands or
endemic distributions which extend to Sicily and/or have distributions which encompass the Tyrrhenian
Calabria. islands and surrounding continents, endemic plants
A similar pattern of disjunct endemism exists on the Pityusic islands are either endemic to these
for the endemic plants of the Balearic islands islands or have distributions which extend to the
(Contandriopoulos and Cardona 1984; Alomar et al. eastern Iberian peninsula (Fig. 2.8). There is a
1997), which are made up of two main groups marked break across the Balearic islands, a sort
of islands, the Gymnesian islands of Majorca, of Wallace-line type separation between the Gym-
Minorca, and Cabrera (plus some very small adja- nesian and Pityusic islands, that is, through the
cent islands) and the Pityusic islands of Ibiza and middle of the archipelago (Box 2.2). The stronger
Formentera (plus adjacent small islands). Endemic historical affinities of the Gymnesian islands to
taxa in this archipelago can be restricted to a the ancient microplate means that palaeo-endemic
single island (e.g. Daphne rodriguezii on Minorca, taxa occur primarily on the islands of Majorca and
Globularia cambessedesii on Majorca, Viola stolonifera Minorca and only extremely rarely on the Pityusic
on Minorca, and Allium grosii on Ibiza), a subset of islands.
islands, usually to the different Gymnesian islands
2.3.3 The east–west divide
of Majorca and Minorca and sometimes Cabr-
era (e.g. Thymelaea velutina and Rhamnus ludovici- East–west vicariance is common in the Medi-
salvatoris), to the Gymnesian islands plus parts terranean and involves some characteristic
of the original microplate (Corsica and Sardinia), Mediterranean species (Dallman 1998; Quézel
or to the Pityusic islands and the eastern Iberian and Médail 2003). Despite the separation of
peninsula (e.g. Silene cambessedesii). Although there North Africa from Europe by the Mediterranean
are some exceptions, these distribution patterns Sea, there is a weaker floristic relationship between
illustrate nicely the different biogeographic affini- the eastern and western Mediterranean Europe
ties of taxa on the Gymnesian and Pityusic islands. than there is between northern and southern
Whereas endemic taxa on the Gymnesian islands shores. This is particularly apparent when one

Box 2.2 Patterns of single-island and disjunct endemism in the western Mediterranean
as illustrated by species endemic to Corsica and associated regions

CO: Corsica
SA: Sardinia
SI: Sicily
PY IT TU: Tuscan archipelago
IT: Italian peninsula
BA: Balearic islands
Gymnesian PY: Pyrenees
islands AL: Alps
PR: Provence

The histogram, drawn from data in Gamisans endemic distribution (e.g. CO + BA is the
(1999), shows the percentage of all endemic taxa percentage of all endemic species on Corsica
which occur on Corsica as a function of their whose distribution extends to the Balearic islands).

% of endemic taxa on Corsica






[126] [69] [18] [17] [13] [13] [6] [7] [9]

There is a clear concordance between endemic following areas:

distribution patterns and the geological history of • Sicily (e.g. Berberis aetnensis, Carduus
the region since ∼30% of endemic taxa on cephalanthus, and Cardamine chelidonia);
Corsica have distributions which encompass either • the Balearic Islands (e.g. Naufraga balearica,
the different parts of the microplate, which Soleirolia soleirolii, and Arenaria balearica);
fragmented and migrated in the Miocene, or • the Tuscan or Hyères islands (e.g. Teucrium
Continental France and Spain. The most common marum and Delphinium pictum);
pattern of ‘disjunct endemism’ involves taxa which • Italy (e.g. Alnus cordata and Pinus nigra subsp.
occur on Corsica and Sardinia, and one of the laricio).

compares the flora and dominant vegetation of the the retreat of a large number of species into dif-
western (Iberian peninsula–Morocco) and eastern ferent geographically isolated refugia during the
(Turkey–Israel) limits of the Mediterranean. First, Quaternary glaciations. Second, historical and
in the different altitudinal vegetation belts there current day differences in climate, in particular
is a closer resemblance of dominant tree species rainfall and seasonal aridity, are more extreme
between the south-east and north-east and between from east to west than they are in a north–
the south-west and north-west regions, than there south direction. Finally, the eastern Mediterranean
is from east to west (Quézel and Médail 2003). flora has been strongly influenced by immigra-
Second, the flora of the southern part of the Iberian tion of Irano-Turanian floristic elements, much
peninsula and the Moroccan Rif have 75% of more than the western Mediterranean (Zohary
plant species in common (Valdés 1991). Messinian 1973). For example, the flora of Crete has a pre-
land-bridge connections (∼5–6 Ma) between the dominantly Mediterranean origin whereas that of
Iberian peninsula and Morocco, a similar climate Cyprus has a dominant Irano-Turanian element
and ecology on either side of the Mediterranean (Chapter 1).
and the migration of bird dispersers in the autumn The eastern Mediterranean, particularly in and
have undoubtedly contributed to the homogeniza- around the Aegean Sea, has undergone major tec-
tion of the flora of the two regions which form tonic changes and configuration of land masses
the basis of the Ibero-Mauritanian floristic zone, and island formation in more recent history than
which has more than 500 endemic species (Quézel Corsica and the Balearic islands (Chapter 1). It
1978). is thus not surprising to find that the percent-
East–west differentiation is particularly appar- age of all endemic species in the four classes
ent when one examines the distribution of closely of endemism described in Section 2.1 becomes
related species and patterns of differentiation within more biased towards neo-endemics as one moves
individual species in the Mediterranean. Patterns eastwards. In the Peloponnese peninsula and on
of schizo-endemism frequently involve disjunction Crete 83–85% of endemic species are schizo- or
and divergence in an east–west direction. I have apo-endemic taxa and less than 17% are ancient
shown some examples in Fig. 2.2, and many more endemics. In contrast, on Corsica, in the Balearic
can be seen in other works on the Mediterranean islands, and in south-east France there is a more
flora, such as those depicting the distribution of mixed blend of ancient (28–32%) and neo-endemics
tree species in the Mediterranean (Quézel and (68–72%).
Médail 2003), for example, the east–west disjunct Another feature which distinguishes groups
distributions of Pinus brutia and Pinus halepensis of islands in an east–west direction across the
and Quercus coccifera subsp. calliprinos and subsp. Mediterranean concerns the relative proportion of
coccifera on either side of the Aegean, and Quer- single island and disjunct endemic species. Whereas
cus ilex subsp. rotundifolia (central and southern the relative proportion of these two types of
Spain) and Q. ilex subsp. ilex (from Provence to the endemics is roughly equal (48–52%) on Corsica,
shores of the Aegean) whose distributions are con- Sardinia, and the Balearic islands (and Malta), sin-
nected by the presence of an intermediate ‘morpho- gle island endemics are far more common (75–85%
type’ present across southern France and north-east of all endemics) on the islands of Sicily, Cyprus, and
Spain. Crete (Médail and Quézel 1997). The large number
Three major factors (all discussed in Chapter 1) of disjunct endemics present in Corsica, Sardinia,
have no doubt greatly contributed to east–west and the Balearic islands are related to historical
floristic and species differentiation and thus the connections (Box 2.2). The higher rates of strict
patterns of endemism depicted above. First, endemism on Sicily, Cyprus, and Crete suggest that
the climatic history of the northern shores of the either ancestral connections were lost earlier (not so
Mediterranean has probably been a major ele- for Crete), that once connected areas are now sub-
ment in east–west patterns of divergence due to merged, or, as is probably the case for Sicily, that

parts of these islands were not part of an ances- identify 26 localities around the Mediterranean
tral connection. The high degree of single island where relict vegetation elements occur, notably in
endemism and widespread taxa at their distribu- the mountains and coastal areas of Morocco, south-
tion limits on Crete (e.g. P. brutia, Q. coccifera subsp. ern Spain, Sicily, Calabria, Crete, the mountains of
calliprinos, etc.) highlights the unique position and Greece, south-west and south-east Turkey, and the
botanical significance of this island in the southern Amanus and Liban mountains. Scattered trees in
Aegean. these relictual fragments are all that is left of the
Tertiary palaeoflora (Table 2.1). Several of these areas
illustrate where Tertiary vegetation found refuge as
the Mediterranean climate set in and they all may
have given refuge to species from a combination of
2.3.4 Endemic relicts from the Tertiary
reduced temperature and greater aridity during the
glacial maxima of the Quaternary (see Chapter 1).
This discussion of endemism would be incom- Endemic Tertiary relict species can also be observed
plete without mention of Mediterranean endemics in the herbaceous flora of cliffs and crevices, for
which represent the last remaining evidence that example, the five species of Gesneraiceae, of which
a few million years ago a luxuriant broad-leaved Ramonda myconi occurs in the Pyrenees and four
forest clothed the landscape around the shores others in the Balkans, that is, Ramonda nathaliae and
of the Mediterranean. Quézel and Médail (2003) Ramonda serbica, Haberlea rhodopensis, and Jankaea

Table 2.1 Examples of relictual populations of ‘Tertiary endemics’ in the Mediterranean region

Species and family Contemporary distribution in the Past distribution

Mediterranean (see also Chapter 1)

Liquidambar Humid bioclimate (riparian habitat) of the An ecologically important genus in the Miocene with
orientalis thermo-Mediterranean in south-west Turkey and on many fossil remainsb . Known from the Late Pleistocene
Rhodesa in Italyc
Laurus azorica Atlas Mountainsd Neogene ancestors occurred as far north as south-central
Zelkova sicula Single locality on Sicily (Monte Lauro)e Widespread genus during the Tertiary, up till the Early
Zelkova abelicea Three mountain ranges on Cretef Pleistocene (30,000 BP) in the western and central
Phoenix theophrasti Coastal vegetation in riparian habitat and gorges More widely distributed in Greece and perhaps Turkey and
(<350 m elevation) in the Peloponnese, Crete and the Aegean islands
south-west Turkeyg
Rhododendron Disjunct populations of subspecies in the southern Widely distributed throughout southern Europe during the
ponticum Iberian Peninsula and east of the Aegeanh Tertiaryi . Now an invasive species in the British Isles
Frangula alnus A relict subspecies has been described in riparian Other subspecies still widely distributed in central Europe
forest vegetation in southern Spainj and northern
Tetraclinis articulata Semi-arid bioclimate of north-west Africa, southern Ancestral taxon present in Europe during the Eocenel
Spain and Maltag

a Akman et al. (1993). b Roiron (1992). c Combourieu-Nebout (1993). d Barbero et al. (1981).
e Di Pasquale et al. (1992). f Barbero and Quézel (1980). g Quézel and Médail (2003). h Mejías et al. (2002).
i Mai (1989). j Hampe and Arroyo (2002). k Fennane and Ibn Tatou (1998). l Palmarev (1989).

heldreichii which is endemic to Mt Olympus (Polunin 2.4 The biology and ecology of
and Smythies 1973; Polunin 1980; Vokou et al. 1990). endemic plants
These relict populations provide us with a contem-
porary image of the past flora of the Mediterranean 2.4.1 The comparison of endemic and
region. widespread species
Such relict distributions concern not only indi- The study of range size limitation is a central
vidual species but in some cases a whole ‘cortège’ of theme in evolutionary ecology. To pinpoint traits
plants. In a study of the flora of south-west Morocco, associated with endemism is no easy task, and for
Médail and Quézel (1999) identified this region as an some authors a fruitless endeavour. Comparisons of
important refuge for Mediterranean elements of the rare and common plant species nevertheless indicate
flora, many of which have dispersed to the Canary that there may exist general differences in the bio-
islands (e.g. succulent Euphorbia, Aeonium, Dracaena, logy of endemic and widespread species (Kunin
and Sonchus), and various elements of the Tertiary and Gaston 1993). In plants, endemism may be
palaeoflora. The occurrence of summer fogs, com- particularly prevalent in some taxonomic groups,
mon in other Mediterranean-type ecosystems which for example, the Ericaceae and Proteaceae of South
flank oceans rather than a closed sea, maintain air Africa (Cowling and Holmes 1992). The relative
humidity at high levels in this region. In addition, frequency of endemic plants may vary among life-
the contrasting relief may have permitted species form classes (Major 1988) and nutritionally unusual
to survive the various periods of major climatic and rare substrates (e.g. serpentine soils) often har-
change since the Pliocene. It is noteworthy that sev- bour high rates of endemic taxa (Kruckeberg and
eral Tertiary endemic taxa occur in riparian and/or Rabinowitz 1985). Finally, as discussed above, dis-
coastal areas where the summer drought may be persal limitation may be critical in the determination
mitigated. of endemism; after all one of the primary reasons
Some relictual populations, for example, Frangula why a species is not present in a given area may be
alnus subsp. baetica (Hampe et al. 2003; Chapter 3), that it has not yet reached the site in question.
contain high levels of genetic diversity which is not A number of important questions concerning the
encountered elsewhere in the species range, and biology of endemic plants can be asked.
thus represent an important part of the evolutionary
potential of the species. They are not just genetically 1. Is the restricted distribution of narrow endemic
depauperate relict populations. They are of signifi- taxa in the Mediterranean linked to their ecological
cance for our understanding of plant evolution and characteristics? Are endemic species more ecolo-
also for the conservation of a representative sample gically specialized than widespread species?
of genetic diversity. However, in the last century or Do endemic species have traits which reflect a
so, many of the species in Table 2.1 have seen their lack of competitive ability or a more stress-tolerant
populations further depleted and fragmented as a ecological strategy?
result of human activities and the conservation of 2. Has human impact been greater on the habitats
such species has become a pressing problem, whose of endemic taxa?
resolution is problematic. For example, as a result 3. Do endemic taxa have reproductive or dispersal
of forest destruction less than 25% of the habitat traits which limit dispersal ability?
of Liquidambar orientalis known to have existed in 4. Are endemic plants a taxonomically hetero-
the early twentieth century now remains (Quézel geneous assemblage or do a small number of
and Medail 2003). In Zelkova sicula, only ∼200 particular groups represent the majority of endemic
plants now exist in a single population on Sicily, taxa in a region?
where overgazing, a precarious ecological situation
and poor pollen viability in this andromonoecious In attempts to analyse these questions one must
tree (Nakagawa et al. 1998) all threaten the demo- be cautious. First, although the traits of endemic
graphic viability of this species (Di Pasquale et al. species may indeed be different from widespread
1992). species, such differences may be a consequence of

restricted distribution, and not the cause. Second, in a small number of genera/families is observed in
despite observations that geographic range often other Mediterranean-climate regions (Section 2.4.4).
closely reflects local ecological breadth (Hodgson Recent studies using phylogenetically controlled
1986; Kunin and Shmida 1997) and that there may approaches indicate how the biology and ecology
be the associations between local or regional abund- of Mediterranean endemic taxa may differ from
ance and geographic range (Hubbell 2001), species widespread species.
with widespread distribution are not always locally
abundant, nor do they consistently have a wider
2.4.2 Reproduction and rarity in crucifers
ecological breadth. Hence, if a relationship is found
between particular traits and estimates of abund- In a study of 52 Mediterranean annual crucifers in
ance on a local scale, similar relationships may not Israel, Kunin and Shmida (1997) applied the PICs
occur when the traits are examined in relation to procedure to test for relationships between repro-
range size. The interpretation of differences between ductive traits and rarity on three scales: (a) local
widespread and endemic species may thus depend density, (b) regional abundance, and (c) geographic
on the scale of investigation (Kunin and Shmida range size in Israel. The reproductive traits they
1997). examined involved floral traits (e.g. flower depth,
Another difficulty concerns the choice of study petal length, and floral longevity) and an estimate of
species. Many comparative studies of widespread whether species are self-compatible. These authors
and endemic species have involved a single rare found that species whose populations contain low
species and a single closely related widespread densities are more likely to be self-compatible (and
species. This approach provides valuable informa- in most cases capable of spontaneous selfing without
tion for the conservation of particular species any pollinators) than are species whose populations
but provides only low power for generalization. tend to contain dense clumps of plants. This result no
Comparative evaluation requires large numbers of doubt reflects the sensitivity of reproductive success
species, some of which may share traits due to a to local density in self-incompatible annual plants.
common ancestral condition. This so-called phylo- Kunin and Shmida (1997) also found that rare
genetic constraint can be accounted for in two ways. species show more extreme trait values than
widespread species. Within self-incompatible taxa,
1. The use of phylogenetically independent con-
rare species have larger flowers than common
trasts (PICs) in which contrasts of trait values among
species. In direct contrast, in self-compatible taxa,
related species are made at each node of a phylo-
rare species have smaller flowers than common
genetic tree, rather than by using the trait values
species. Likewise, in self-incompatible taxa, rare
of the different species (Harvey and Pagel 1991).
species have longer floral longevities than com-
This method reduces the sample size to account
mon species, probably due to both their larger
for non-independence associated with phylogenetic
flowers and infrequent or rare pollinator visitation.
The annual life-history of the study species means
2. Multiple contrasts of species pairs in different
that all species require some pollination in the year
phylogenetic backgrounds, for example, pairs of
they flower for them to contribute to subsequent
widespread and endemic species in different genera.
generations. Self-compatible taxa show no require-
The importance of such a phylogenetically con- ment to extend floral lifetime since most species
trolled approach is illustrated by the fact that in were capable of seed set in the absence of pollinators.
different parts of the Mediterranean, endemism is
over-represented in certain genera and families, for
2.4.3 A comparison of congeneric endemic
example, the Iberian peninsula where half of the
and widespread species
∼1,250 endemic taxa occur in only 27 genera and
the other half are distributed across 286 genera In a comparative analysis of endemic and
(Chapter 1). A similar concentration of endemic taxa widespread species in 20 genera, Lavergne

et al. (2004b) studied whether endemic taxa in the the persistence of endemic species is that such
western Mediterranean differ from widespread habitats are relatively stable, both in relation to
congeners in terms of their habitat, community char- vegetation succession and human activities. In such
acteristics, traits linked to resource acquisition and habitats, pioneer species which colonize rocky cliffs
allocation, reproduction and dispersal, amounts may facilitate the establishment of dwarf shrubs.
of herbivory, and maternal fertility (Box 2.3). For example, Escudero (1996) suggests that the
Surprisingly, levels of herbivory and ecophysiolog- establishment of nano-phanerophytes in cliffs in
ical traits (specific leaf area, leaf dry matter content, the Iberian peninsula (e.g. Rhamnus alpinus and
leaf nitrogen concentration, and rates of photo- Lonicera pyrenaica) depends on a process of facilita-
synthesis) showed no overall difference between tion whereby colonist chamaephytes (e.g. Saxifraga
related endemic and widespread taxa. Endemic and moncayensis and Silene saxifraga) cause an accumu-
widespread species did however differ for a number lation of soil in crevices and hollows. The nano-
of ecological characteristics and biological traits, phanerophytes can grow to much greater size and
which emphasize three important features of the their root development probably contributes to a
population ecology of endemic plants: their habitat shattering of the rock and erosion, allowing for
in terms of abiotic and community characteristics, re-colonization by the pioneer chamaephytes. The
floral traits and the size and maternal fertility of cliff faces which are home to many endemic species
individual plants. The discussion of these findings thus appear as a complex mosaic of discrete micro-
is based on Lavergne et al. (2004b) and illustrated in communities which succeed one another in a col-
Box 2.3. onization, succession, extinction cycle. However,
severe environmental constraints on vegetation
establishment may halt the successional develop-
Distinct ecology ment of a forest cover. Their inaccessibility and
Endemic species have a distinct ecology compared unsuitability for cultivation may have allowed such
to widespread congeners. The principal component habitats to serve as a refuge for endemic taxa during
of this difference concerns the occurrence of endemic periods of intense human-induced landscape modi-
species in rocky and steep slopes in low, open fication. Polunin (1980) pointed out that due to the
vegetation with low species richness. This indic- reduced impact of disturbance by humans and graz-
ates that endemic species grow at sites not only for ing animals, the vegetation of Mediterranean gorges
historical reasons which have isolated their distri- and cliffs is often rich in endemic and rare species,
butions and limited their dispersal, but also because which find a refuge there from the competition
of a ‘fit’ between their ecology and prevailing site of aggressive species in the surrounding habitats.
conditions. This trend for endemic species to occur Zohary (1973: 315) also commented on how ‘the
in distinct ecological conditions compared to their lack of competitive vigour is probably one of the
widespread congeners, in particular the associa- major causes for the stenochory of so many endemics
tion with open habitats rather than forest vegeta- which flourish when artificially grown in an envi-
tion has been commented on elsewhere (Grove and ronment free from competition’. It is thus probable
Rackham 2001; Quézel and Médail 2003). Indeed, that the persistence of endemics may have been
a survey of different flora illustrates the frequent favoured by their capacity to grow in rocky habitats
occurrence of endemic species in rupicolous hab- where competitive interactions, which may have
itats, often in and around cliffs, for example, in the fundamental effects on diversity and the persist-
Balearic islands (Alomar et al. 1997), Greece, and ence of endemic plants (Box 2.4), may also be
the Balkans (Polunin 1980; Strid and Papanicolaou limited.
1985). The question thus arises as to whether the ecolo-
A characteristic of open rocky habitats on steep gical strategies of endemic species, in terms of
slopes (sometimes on scree slopes, sometimes in functional trait variation, differ from those of
and around cliff faces) which may be crucial for widespread species. Does trait variation reflect

reduced competitive ability, increased tolerance of in association with increased persistence of endemic
stress, and/or reduced dispersal ability? Lavergne taxa if there is a negative correlation between annual
et al. (2004b) detected a clear trend for endemic investment in seeds and vegetative persistence. In a
species to be smaller than widespread congeners. few of the genera studied by Lavergne et al. (2004b)
However, congeners showed no differences in leaf individuals of the endemic species live longer than
traits or photosynthetic capacity and thus their those of the widespread species (even though they
capacity to acquire resources. There was thus no both show a perennial life history).
evidence for a stress-tolerant ecological strategy in
endemic taxa, as suggested by a previous study
based on an analysis of life-history strategies in Habitat, fertility and population ecology
endemic species (Médail and Verlaque 1997). Of The differences in ecology and fertility suggest
course, other unstudied factors may be involved, marked differences in the population ecology
for example, allocation to root growth, which could of endemic and widespread species. Populations of
contribute to the greater affinity of Mediterranean endemic species have traits and recent history that
endemics for rocky, steep sloping habitats in the suggest high local persistence and low population
Mediterranean and cause reduced allocation to turnover. A study of population persistence in the
above-ground stature. Hérault département of southern France from 1886
(based on a detailed local flora published at the end
Floral traits of the nineteenth century) to 2001 (from a database
Endemic species have floral traits which suggest created by the Conservatoire Botanique National
that their populations are more inbred than those of Méditerranéen de Porquerolles) by Lavergne et al.
widespread congeners, that is, smaller flowers with (2004c) illustrates that individual populations
a lower pollen : ovule ratio and less herkogamy (i.e. of endemic species at a given site have a greater
stigmas closer to the anthers) than flowers of their temporal stability than those of widespread species.
widespread congeners. However, I am inclined to In contrast, the latter show stability in numbers of
interpret this difference as being a consequence of populations in the same area but in different sites.
their restricted range size and small population size Widespread species may thus have populations
rather than a cause of limited distribution. Many which are more closely connected to one another by
invasive and colonizing species are inbred, hence virtue of higher rates of colonization and extinction.
it is hard to see why inbreeding should limit the Widespread species may thus move around the
distribution of endemic species. landscape to a greater extent and function more as
typical meta-population systems have been sug-
Lower maternal fertility gested to do. Having said this, the marked spatial
Endemic species have a significantly lower mater- and temporal variation in the dynamics of six small
nal fertility than their widespread congeners. In populations of Centaurea corymbosa (i.e. all known
addition to causes linked to pollen and resource populations of this species which is endemic to an
limitation, the male function of hermaphrodite area of ∼3 km2 on a single massif in southern France)
flowers and inbreeding depression, perennial over a period of eight years detected by Fréville
plants may reduce annual reproduction in order to et al. (2004) suggest that such meta-population
optimize lifetime fitness. It is thus typical to observe dynamics may also occur, but on a much more
low maternal fertility in herbaceous perennial localized scale, in narrow endemic species (B. Colas
plants in the Mediterranean (J. Herrera 1988, 1991a; et al., unpublished manuscript). Differences in
C.M. Herrera 1993; Thompson and Dommée 1993; the dynamics of other narrow endemic species in
Baker et al. 2000a; Méndez and Traveset 2003). Such relation to microhabitat variation (Albert et al. 2001;
low maternal fertility in endemic taxa could con- Chapter 4) further illustrate this mosaic pattern of
tribute to reduced dispersal across the landscape but population dynamics. I am tempted to suggest here
at the same time may be a strategy that has evolved that endemic and widespread species may in fact

Box 2.3 The comparison of the biology and ecology of endemic and widespread species
in the flora of Languedoc-Roussillon region of southern France (redrawn from
Lavergne et al. 2004b)

Based on one narrow endemic species (upper traits. Nineteen genera contain herbaceous
species in each genus) and one widespread species perennials, one genus involved annuals. The pairs
(lower species) in 20 genera in which the paired of species were sampled in geographically close
species have the same pollination and dispersal sites to minimize climatic variation (38 of the
modes and growth form, and do not show 40 species were sampled in an arc extending
consistent differences in ploidy level, Lavergne from the Rhone Valley to the Pyrenees). Of the
et al. (2004) performed multiple comparisons of 20 endemic species, 11 are protected
ecological, ecophysiological, and reproductive by law.

Scrophulariaceae Odontites jaubertiana

Odontites lutea
Labiaceae Thymus nitens
Thymus pulegioides
Caprifoliaceae Lonicera pyrenaica
Lonicera xylosteum
Asteraceae Centaurea corymbosa
Centaurea maculosa
Phyteuma charmelii
Phyteuma orbiculare
Cyclamen balearicum
Primulaceae Cyclamen repandum
Lysimachia ephemerum
Lysimachia vulgaris
Genista villarsii
Fabaceae Genista pilosa
Lathyrus cirrhosus
Lathyrus latifolius
Cistaceae Cistus pouzolzii
Cistus monspeliensis
Brassicaceae Hormatophylla pyrenaica
Hormatophylla spinosa
Resedaceae Reseda jacquinii
Reseda phyteuma
Polygonaceae Polygonum romanum
Polygonum maritimum
Aquilegia viscosa
Renunculaceae Aquilegia vulgaris
Thalictrum tuberosum
Thalictrum minus
Iridaceae Iris xiphium
Iris spuria
Amaryllidaceae Narcissus dubius
Narcissus tazetta
Cyperaceae Carex olbiensis
Carex glauca
Gramineae Melica amethystina
Melica ciliata
Liliaceae Lilium pyrenaicum
Lilium martagon

Source: Phylogeny from Soltis et al. (2000).


(b) Independent, pair-wise comparisons of means that the value for the widespread species is
ecological variables and biological traits were greater than that of the endemic species in a given
made for each genera. A point above the bisector genus.

% rock cover Slope pH

80 1,000 9
P < 0.001 P < 0.001
60 100
40 10
20 1
P > 0.1
0 0.1 4
0 20 40 60 80 0.1 1 10 100 1,000 4 5 6 7 8 9

% cover by trees and shrubs Number of co-occurring species % damaged flowers

45 100

80 35
Trait value in the widespread species

P < 0.05 P < 0.01 P > 0.1
0 5 0
0 20 40 60 80 100 5 15 25 35 45 0 1 10 100

Specific leaf area Photosynthetic rate Plant size

40 1,000
9 100
10 10
P > 0.1 P > 0.1 P < 0.05
0 7.5 1
0 10 20 30 40 7.5 8 8.5 9 9.5 1 10 100 1,000

Flower size Pollen ovule ratio Stigma–anther separation

100 100,000 100

10,000 10
1,000 1

P < 0.05 P < 0.01 P < 0.05

1 100 0.1
1 10 100 100 1,000 10,000 100,000
0.1 1 10 100

Flower number Seed production Seed mass

1,000 1,000 10,000

100 1,000
10 100
1 10
0.1 P > 0.1
P < 0.05 P < 0.05
0.01 1
0.01 0.1 1 10 100 1,000 1 10 100 1,000 10,000
0.1 1 10 100 1,000

Trait value in the endemic species


Box 2.4 Species diversity, endemism, and competitive interactions

Tilman and Pacala (1993) discussed how conditions for diversity are also the most common
competitive interactions, in relation to the habitat type in a landscape or those where
productivity and resource heterogeneity of the disturbance is also at intermediate levels (and thus
local environment, may contribute to patterns of acting to promote diversity), then species diversity
variation in community diversity. These authors may be further increased.
suggest that productivity gradients affect species In Mediterranean grasslands (Puerto et al. 1990)
diversity via spatial variation in nutrient supply and and South African Fynbos (Bond 1983) high
light availability along productivity gradients, that diversity occurs at intermediate levels of
is, from habitats with low rates of supply of soil productivity. Similarly, comparison of localities in
resources and high light penetration to those with southern Spain by Ojeda et al. (1995) has shown
high rates of supply of soil resources and low light that heathlands on nutrient poor soils and Quercus
penetration. In nutrient-poor habitats, all species canariensis woodlands in more productive habitats
are limited by soil resource supply, and the best have higher rates of endemism but lower species
competitor will dominate. At the other end of the diversity than cork oak (Quercus suber)
gradient the same process, this time in relation to woodlands, where intermediate levels of fertility
light availability, should cause domination by the occur. Indeed, species richness (measured as the
best competitor for light. In this manner, low number of woody species) of shrubland and
nutrient availability or the large size of organisms heathland communities in southern Spain is
lead to marked declines in resource heterogeneity negatively correlated with soil fertility (Arroyo and
at the two extremes of the productivity gradient. Marañón 1990). In the latter case, heathland
In addition, in the central part of the gradient, communities on nutrient-poor acid soils have
different factors may promote coexistence. As a higher diversity than shrublands on nutrient-rich
result, species diversity should show a unimodal basic soils. There may thus be complex
relationship with productivity (Tilman and Pacala relationships between species richness and rates of
1993). If the habitats with the necessary endemism.

show little or no difference in their capacity for long- Mediterranean Basin) have attracted a great deal of
distance colonization (which is a very rare event for interest concerning the extent to which the overall
both types of species). What may differ however is similarities one can see in terms of community struc-
the spatial scale of meta-population function, which ture and form, represent functional convergence
in turn affects geographic range. The probability of in similar ecological conditions (Chapter 4). The
natural extinction of any given population may in four other Mediterranean type ecosystems also
fact be higher in widespread species than in endemic contain floras which combine immense diversity
species. (Dallman 1998) and high rates of local endemism,
for example, almost 30% in the Cape Penin-
sula (Simmons and Cowling 1996) and the Agul-
has Plain (Cowling and Holmes 1992) in South
2.4.4 Consistent patterns in different
Africa and the Barrens of south-west Australia
Mediterranean-climate regions of the world?
(Cowling et al. 1994). Hence the question: are simi-
The five different regions of the world with lar ecological and biological features associated with
a Mediterranean climate (i.e. California, Chile, endemism in the different Mediterranean-climate
South Africa, and Australia in addition to the ecosystems?

In what has become a classic study of endemism present in a small area: acidic and rather infertile
and speciation, Stebbins and Major (1965) described soils on sandstones and quartzites which make up
how areas with the highest degrees of endemism most of the low mountains of the area, more fer-
in the Californian flora are those which have the tile acidic soils on the shales which separate the
greatest variety of habitats and ecological conditions main mountains, shallow infertile siliceous soils
and which have escaped glaciation during the in pockets across the landscape, well-drained cal-
Pleistocene. In general, low-lying mountains are careous sands on limestone bedrock in the coastal
centres of endemism. In striking similarity to zone, calcareous aeolian quartz and dunes along
the Mediterranean examples discussed above, the the coast and valley, flood plains with alluvial or
endemic species in California have very diverse colluvial topsoils on a clay-base soil. Strong rainfall
origins. Many endemics have relict distributions gradients also exist in this region. Hence, despite
while others are recently evolved and, as these a relative lack of physical separation of different
authors suggest, may not yet have achieved their land masses, ecological differentiation can occur
maximum possible area of distribution. The for- in what is an intricate mosaic of spatially hetero-
mer tend to have persisted in regions where the geneous edaphic environments. Despite the short
climate has been fairly stable (some of the fairly distances between habitats on these different sub-
mesic zones and in the highly arid interior regions) strates, Cowling and Holmes (1992), found that
while the latter have probably evolved in regions more than two-thirds (69%) of regional endemics
where spatial heterogeneity has promoted active and 85% of local endemics in the flora in the Agul-
differentiation. has Plain only occur on a single type of substrate.
In the Cape Floristic Region, well known for its Of the total endemic flora, more than one-third
diversity of endemic plants in a small surface area, (37%) only occur on limestone. Edaphic specializa-
studies of the taxonomic, ecological, and biologi- tion is thus rife in endemic plants of this region,
cal characteristics have provided critical and novel suggesting that adaptive differentiation has played
insights into the biology and ecology of endemism an important role in the high rates of endemism in
in this region. In the Agulhas Plain, a rolling coastal this area.
plain which occupies 1,600 km2 at the southern tip Cowling and Holmes (1992) also show that
of South Africa, rates of endemism are ‘extraordinar- endemism is not common to all the plant groups
ily high for a lowland continental region’ (Cowling present and is particularly present in species with a
and Holmes 1992: 376). The plant communities of certain biological profile. Some families have statist-
this area are the characteristic (and endemic) Fynbos ically more regional endemic species than expected
vegetation in which ∼30% of the flora is endemic by chance, witness the amazing diversity of cer-
(Cowling and Richardson 1995). This area has no tain genera in the Ericaceae (45% of speceis in
long history of physical isolation with surround- the genus Erica are regional endemics), Proteaceae
ing land and has been submerged more than once (47% of the genus Leucadendron), Polygalaceae (46%
in recent geological history, that is, in the Miocene of species in the genus Muraltia), and Rutaceae
(∼15 Ma) and Pliocene (∼4 Ma). Diversification (38% of the genus Agathosma). Others are under-
typical of physically isolated islands and moun- represented in endemic taxa (e.g. Orchidaceae,
tain areas has thus occurred recently and rapidly in Scrophulariaceae, and Poaceae). Second, endemic
a continental setting. This diversity appears to be species fit a biological profile of non-sprouting dwarf
related to ecological specialization by a small num- shrubs with short-distance (often ant or passive)
ber of taxonomic groups with a similar biological seed dispersal with (in some families) a symbiotic
profile. relationship with soil microorganisms. Cowling and
The geology of the region with a Mediterranean Holmes (1992) argue that lineages of such species
climate in South Africa is complex (Cowling and may be particularly vulnerable to population reduc-
Holmes 1992), with a diverse number of substrates tion and local extinction and at the same time more

prone to rapid edaphic specialization. A similar have a fairly original flora (Arroyo and Marañón
biological profile has been identified in mountain 1990; Ojeda et al. 2000a). In this region, heath-
Fynbos (McDonald and Cowling 1995). However, land communities are more diverse than those in
multivariate logistic regression analysis of the rela- the non-Mediterranean Europe (Ojeda et al. 1998)
tionship between such traits and endemism by and show higher rates of endemism than nearby
McDonald et al. (1995) showed that the primary communities in cork oak woodland (Ojeda et al.
factor determining high rates of endemism is short- 1995 Box 2.4).
distance dispersal, and that the interaction between Depending on their substrate preference, endemic
growth form and regeneration strategy plays only a taxa in southern Spain show contrasting distri-
secondary role. bution patterns. Whereas on limestone, endemic
In a comparison of patterns of endemism in the species occur only in the Betic cordillera of south-
Cape Floristic Region with matched sites (in terms east Spain, the endemic taxa of heathlands on acid
of historical and contemporary climate, landforms, soils also occur in North Africa and the western part
and soils) in the Barrens in south-west Australia, of the Iberian peninsula where similar habitats on
where species richness and endemism are very high Oligocene siliceous sandstone also exist. In addition,
(Crisp et al. 2001), Cowling et al. (1994) found similar acid soils, on which many narrow endemic species
patterns of endemism in the two regions. These pat- occur in the southern Spanish heathlands, are not
terns are a high degree of edaphic specialization for common around the Mediterranean. Hence many
nutrient-poor acid soils, a broadly similar biological widespread species which occur on more common
profile of endemic taxa (non-sprouting dwarf shrubs and widespread calcareous soils may be absent from
with short-distance dispersal), although serotinous the acidic and sandy soils which harbour heathland
species are also important elements of endemism species. So both heathland communities on acid soils
in south-west Australia (Beard et al. 2000), and and limestone shrublands tend to occur as edaphic
an over-representation of endemic taxa in cer- or geographically isolated ‘islands’ in the landscape,
tain taxonomic groups (Proteaceae, Epacridaceae, hence the importance of narrow endemism in such
and Myrtaceae). The association between local communities. Once again specialization on nutrient-
endemism and the non-spouting growth form has poor soils and the spatial occurrence of suitable hab-
also been detected in genera such as Arctostaphy- itat are critical for the geographic range of endemic
los and Ceanothus in the California chaparral (Wells species.
The consistency of the association between
2.5 Conclusions
endemism and nutrient-poor acid soils and shrubby
growth form has support from studies in the Endemic taxa represent ‘something old and some-
Mediterranean region. In the heathlands of the thing new’. Ancient endemic species provide clues
Sierras de Algeciras in the Aljibe mountains near to historical connections while neo-endemic species
the Straits of Gibraltar in southern Spain, endemism provide models for the study of more recent and
is associated with a shrubby growth form and often rapid differentiation in relation to climate and
is negatively correlated with substrate fertility ecology. What is clear from this chapter is that we
(Ojeda et al. 2001). This region comprises large have gained much ground in our appreciation of
areas of acidic soils (pH 4–5) on siliceous sandstone the two sets of conditions which favour high rates
(dating to the Oligocene–Miocene) with low nutri- of endemism: the factors that favour persistence of
ent content and high levels of assimilable aluminium endemic plants over long periods and the condi-
and other heavy metals harmful to normal plant tions which promote divergence. Endemic species
development (Arroyo 1997). The Aljibe mountains are not all relatively recent in origin and do not
probably represented an important glacial refuge for all occur at the tips of phylogenetic trees. Nor do
many species during the Pleistocene and currently they have evolutionary lifetimes shorter than more

widespread species. Many Mediterranean endemics small island systems indicate that they are the rem-
are either very old (i.e. palaeo-endemic taxa con- nants of a larger flora in which random variation
served from the Palaeogene flora of the initial has marked community composition. The process
microplates) or, as is the case for patro-endemics, the is akin to that of random differentiation in gene
progenitors of more widespread species. Endemic frequencies on small island systems which, as I
distribution patterns cover the whole range of spa- illustrate in the next chapter, may produce striking
tial scale from taxa endemic to single mountains patterns of differentiation. Indeed, community and
or islands, groups of disjunct islands or contin- population patterns may often show parallel trends,
ental areas, to those endemic to whole regions such which may have similarity in their underlying
as the western or eastern Mediterranean. These causes.
diverse distribution patterns provide insights into A clear conclusion concerns the association
the roles of history and ecology in the determina- between endemism and ecological conditions.
tion of species’ ranges. Indeed, when associated Endemic plants tend to occur on steep rocky slopes
with the tools of population genetics and sys- and cliffs, and soils which are nutrient poor or
tematic and phylogenetic analysis, these patterns have intermediate nutrient levels. Home to many
can be used to analyse the evolutionary process endemic plants, such habitats illustrate both the ran-
(Chapter 3). dom elements of variation in species composition
The occurrence of high rates of endemism in due to dispersal limitation and the importance of a
areas with strong geographic barriers to dispersal suitable habitat for the presence of endemic species.
(islands and mountains) point to the important role Cliffs and steep rocky areas are stable habitats; their
of history in shaping species distributions around inaccessibility puts a halt to human cultivation and
the Mediterranean. Indeed, for some authors, ‘the even goat and sheep grazing and, because of con-
multitude of discontinuities created by geological straints on plant development, succession towards
processes is perhaps the ultimate cause of local a climax forest is prevented. This stability has prob-
rarity and narrow endemism’ (Kruckeberg and ably favoured the persistence of endemic plants. The
Rabinowitz (1985: 465)). In the Mediterranean, to importance of local persistence may have moulded
quote Zohary (1973: 320), ‘the historical reasons for trait evolution in endemic plants, which may have
the outburst of endemism are certainly the most a greater longevity associated with decreased repro-
weighty ones’. This author outlined three major ductive effort in a given year and whose lower
historical events which may explain the high rates abundance across the landscape reduces the prob-
of local endemism in the eastern Mediterranean: ability of dispersal and colonization of new sites.
(a) the recession of the Tethys and the expansion What remains to be more fully explored is whether
of the Irano-Turanian flora across vast areas of new Mediterranean endemic taxa have a narrower range
ground; (b) the upheaval of the Tertiary fold moun- of ecological tolerance than their widespread con-
tains which created insular habitats of diverse eco- geners. Although there is some evidence that nar-
logy and produced sharp climatic discontinuities; row endemics have a more specialized ecology
(c) the onset of summer drought and repeated cold (Debussche and Thompson 2003), confirmation of
cycles associated with glaciation, which may have the generality of this issue will require detailed
created strong selection pressures for adaptation and comparative field studies across the range of
and caused many distribution patterns to contract the distribution of endemic and widespread con-
into isolated refugia. geners, accompanied by manipulative transplant
There are several pieces of evidence which sug- experiments.
gest that the organization and structure of plant Endemic species in the Mediterranean flora pro-
communities on small islands, and thus potential vide a fascinating material for the study of the factors
endemism is linked to random historical processes regulating plant distribution and long-term persist-
associated with isolation. Patterns of distribution in ence. I have discussed how endemism and diversity

in the Mediterranean landscape depend on the bal- confines of their habitat and, to colonize new hab-
ance between regional processes of immigration and itats, evolutionary change may be required. The
gene flow (and thus the spatial configuration of hab- evolutionary processes involved in the diversifica-
itats) and local ecological interactions within habitat tion of endemic species is the subject of the following
patches. Species originate and evolve within the chapter.

The evolution of endemism: from

population differentiation to
species divergence

. . . the processes involved in ‘descent with modification’, to use Darwin’s classic phrase, can be
shown clearly to apply to differentiation within species, as well as to the further divergence of
species . . . once they have become separated from each other.
G.L. Stebbins (1950: 190)

3.1 Endemism and evolution: the species divergence and narrow endemism in the
processes and scale of differentiation Mediterranean flora.
Species are the basic units of plant taxonomy
To determine how species evolve requires informa- and classification, they allow us to characterize and
tion concerning both the relatedness among study the dynamics of biodiversity and to go out
different taxa and the microevolutionary forces in the field and discover new taxa or a species in
acting on local populations of individual taxa. a new area. It is difficult however to provide a
The pertinence of such microevolutionary forces single definition of the term species that would be
for our understanding of plant species diver- acceptable to all botanists. I thus do not attempt to
gence was first brought to the fore by one of the provide a species definition in this book. For the
great plant evolutionary biologists of the twenti- range of definitions that can be applied to plant
eth century, George Ledyard Stebbins. In three species I refer the reader to Rieseberg and Brouillet
books, Stebbins (1950, 1971, 1974) showed all too (1994) and Levin (2000). My aim in this chapter is to
clearly how the fundamental evolutionary pro- explore how the evolutionary process of divergence
cesses of selection and drift, which act to pro- has been modulated by the geological and climatic
mote genetic differentiation among populations, history of the Mediterranean region. I attach par-
and migration of pollen and seeds, which coun- ticular importance to a discussion of the scale on
teract differentiation and act to homogenize gene which species have diverged and endemic patterns
frequencies across the landscape, have shaped of distribution developed, and the role of ecologi-
not only the contours of differentiation within cal differentiation in the process of speciation. My
species but also divergence among species. Onto premise here is that to understand plant evolution
this foundation he incorporated the natural setting in the Mediterranean requires a knowledge of the
of spatial habitat heterogeneity, the potential for role of microevolutionary processes in relation to
hybridization and reproductive isolation, chromo- the historical framework developed in the two pre-
some evolution, and stochastic factors associated vious chapters. Three questions form the basis for
with dispersal requirements. In this chapter, I will this exploration.
make the jump from the ecology of endemism
discussed in the previous chapter to explore • Are endemic species genetically depauperate and
the evolutionary processes which have caused do they differ from widespread congeners in the


spatial structure of genetic diversity across the genetic variation may be a less important cause of
landscape? endemism and rarity than ecological preferences.
• What are the evolutionary processes and spatial In the last 30 years it has become possible to pre-
scales of differentiation and divergence? cisely quantify genetic variation using a number of
• How different are closely related species with techniques based on protein analysis with allozyme
disjunct distribution patterns? electrophoresis or more direct quantification of vari-
ation in DNA profiles using the tools of molecular
biology (Table 3.1). Such methods make it possible
3.2 Population variation in endemic
to quantify and compare genetic variability based
on allele number per locus, the percentage of loci
Endemic species are rare in the sense that they which show a genetic polymorphism (i.e. more
have a restricted geographic distribution. Their than one allele) and heterozygosity (presence of
restricted geographic range has long attracted atten- more than one allele at individual loci in individual
tion from ecologists seeking to determine whether plants).
endemism is associated with ecological specializa- Comparisons of 34 pairs of rare and widespread
tion (Chapter 2). In addition, from a genetic or congeners by Gitzendanner and Soltis (2000) have
evolutionary perspective, a long-standing issue con- shown that although endemic species have globally
cerning endemic plants has been whether they show significantly lower levels of genetic diversity than
reduced genetic variation relative to widespread their more widespread congeners, in a number of
species. Two ideas underlie the issue of reduced genera the differences are slight, and occasionally
genetic variability in endemic taxa. First, if endemic reversed. The differences are primarily due to high
taxa are specialized to local ecological conditions levels of diversity in a small number of widespread
then natural selection will whittle down varia- species. In fact in 24–29% of the genera, rare species
tion as non-adapted alleles are eliminated. Second, were at least as variable if not more variable than
small population sizes associated with founder their widespread congeners and diversity levels
events and/or genetic bottlenecks may induce rapid in rare species covered a similar range to those
rises in the amounts of inbreeding. As a result, of widespread species. Despite the overall trend,
genetic variability will decline rapidly within popu- it appears that rare species are not a homogen-
lations which become more and more different from ous group with low levels of genetic diversity. In
one another. To study the evolution of endemism addition, Gitzendanner and Soltis (2000) detected a
thus requires an appreciation of levels of popula- significant correlation in levels of diversity across
tion diversity and differentiation in endemic taxa, genera, a finding which underscores the need for
relative to that in common and widespread taxa. multiple comparisons among closely related rare
Stebbins (1942) first addressed the issue of and widespread species.
whether rare species show reduced genetic variation In the Mediterranean flora although there are
compared to more widespread congeners. Indeed some clear examples of reduced genetic variabil-
he was one of the first plant evolutionary biologists ity in endemic species relative to widespread
to ask the question: ‘why are some plant species congeners, there is no clear-cut and consistent
widespread and common, while others are rare trend. In four Mediterranean oaks, genetic divers-
and local’ (p. 241). This question remains all too ity in populations of the narrow endemic oak
pertinent today in our search to better understand Quercus alnifolia (endemic to ultrabasic soils on
plant evolution, be it in the Mediterranean or else- Cyprus) is less than that in more widespread
where. The examples he cites lead to the conclusion species such as Quercus coccifera (Toumi and Lumaret
that rarity is associated with reduced genetic vari- 2001). Likewise, Cyclamen balearicum, which is
ation. However, in a somewhat revised and more endemic to the Balearic islands and southern France,
balanced evaluation of this question some years shows markedly less within-population variation
later, Stebbins (1980) concluded that low levels of than its widespread congener Cyclamen repandum

Table 3.1 A glossary of terms used in this chapter concerning the analysis of genetic variation and population differentiation

Term Brief description

FST , GST , or θ Estimation of the proportion of genetic variability in a total sample that is due to genetic differentiation among
populations (based on allele frequency variation).
Allozyme electrophoresis Based on differences in the charge of proteins associated with different enzymes and their relative migration in
an electric field it is possible to extract proteins from plant tissues and stain gels with patterns that are
interpretable in terms of Mendelian genotypes at particular loci. Allele and heterozygote frequencies can thus
be estimated.
RFLP Restriction fragment-length polymorphism (RFLP) is identified by cutting DNA with enzymes which produce
different fragments that can be sorted and visualized on a gel according to their molecular weight. RFLP is a
technique in which population differentiation can be assessed based on the length of the fragments produced
(which will differ when the DNA is digested with a restriction enzyme). The similarity of the patterns generated
can be used to differentiate species and populations.
RAPD Randomly amplified polymorphic DNA (RAPD) is based on polymerase chain reaction (PCR) which is used to
amplify a sequence of DNA. This amplification can produce evidence of polymorphism due to presence of
different bands after gel staining. However the target DNA sequence is unknown and RAPDs are dominant in
the sense that the presence of an RAPD band does not allow distinction between heterozygous and
homozygous states.
AFLP Amplified fragment-length polymorphism (AFLP) is a PCR-based fingerprinting technology. In its most basic
form, AFLP involves the restriction of genomic DNA, followed by ligation of adaptors complimentary to the
restriction sites and selective PCR amplification of a subset of the adapted restriction fragments. These
fragments are visualized on denaturing polyacrylamide gels. The availability of many different restriction
enzymes and corresponding primer combinations provides a great deal of flexibility. A main advantage of this
method is the possibility of polymorphism detection at the total-genome level.
cpDNA haplotype Chloroplast DNA (cpDNA) is transmitted primarily in maternal lineages (gymnosperms are a well-known
exception). Genotypes for cpDNA represent non-recombining characters transmitted by female parents
through their seeds. They are referred to as haplotypes.

(Affre and Thompson 1997a; Affre et al. 1997). This geographically isolated and marginal populations,
trend is not however consistent, Cyclamen creticum endemism may be a predictable evolutionary out-
endemic to the island of Crete does not show come (Fréville et al. 1998). However, the com-
less genetic variation than widespread C. repan- parative study by Gitzendanner and Soltis (2000)
dum (Affre and Thompson 1997b). In this group and examination of population differentiation in
it is differences in mating system rather than geo- Mediterranean plants (Thompson 1999; Table 3.2)
graphic distribution that occasion the differences in show that there is no general difference between
genetic variability (see below). Other examples of rare and widespread species in terms of the spatial
high levels of genetic variability in narrow endemic organization of genetic variation. This is not surpris-
species can be seen in the genus Antirrhinum in the ing since similar processes act on the populations
Iberian peninsula (Mateu-Andrés 1999). The conclu- of both endemic and widespread species. All plants
sion is clear, endemic species and their populations are sedentary, and it is only by gene movement that
do not in general harbour less variation than their differentiation can be countered. So one expects to
widespread congeners. observe some level of genetic differentiation in a
What about levels of population differentia- species, even in rare or endemic species. Rare and
tion? This question is of particular pertinence here endemic species differ (as much as do widespread
because differentiation is the template on which species) in terms of their life histories, breeding sys-
endemic species divergence occurs. Indeed, where tems, and modes of dispersal, hence they will also
widespread species show marked differentiation in show variable amounts of differentiation. Historical

Table 3.2 Examples of genetic differentiation among populations in widespread and endemic Mediterranean plants. Based in part on Thompson

Species Location and scale of study Estimate of differentiation Reference

Widespread species
Bromus intermedius Algeria GST = 0.24 Ainouche et al. (1995)
Bromus squarrosus GST = 0.23
Bromus lanceolatus GST = 0.25
Bromus hordeaceus GST = 0.06
Centaurea maculosa Southern France FST = 0.26∗∗∗ Fréville et al. (1998)
Cyclamen hederifolium Corsica FST = 0.13∗∗∗ Affre and Thompson (1997a)
Cyclamen repandum Corsica FST = 0.42∗∗∗ Affre and Thompson (1997a)
Ecballium elaterium Across Spain FST = 0.23 (subsp. elaterium) Costich and Meagher (1992)
FST = 0.96 (subsp. dioica)
Fagus sylvatica Balkans FST = 0.01 Gömöry et al. (1999)
Medicago sativa Spain FST = 0.05∗∗∗ Jenczewski et al. (1998)
Medicago truncatula Corsica and southern France, FST = 0.51∗∗∗ Bonnin et al. (1996)
Subpopulations ∼50 m apart FST = 0.32∗∗∗
Subpopulations ∼10 m apart FST = 0.15∗∗∗
Quercus suber Across the Mediterranean Basin GST = 0.08 Toumi and Lumaret (2001)
Quercus ilex GST = 0.14
Quercus coccifera GST = 0.19
Quercus ilex Western Mediterranean GST = 0.10 Michaud et al. (1995)
Senecio gallicus Iberian peninsula and France θ = 0.56∗ (cpDNA) Comes and Abbott (1998)
θ = 0.15∗ (allozymes)
Senecio glaucus Eastern Mediterranean θ = 0.43∗ (cpDNA) Comes and Abbott (1999b)
θ = 0.12∗ (allozymes)
Senecio rupestris Central Mediterranean θ = 0.37∗ Abbott et al. (2002)
Senecio vernalis Eastern Mediterranean θ = 0.05 (cpDNA) Comes and Abbott (1999b)
θ = 0.04∗ (allozymes)
Thymus vulgaris Southern France FST = 0.04∗∗∗ (allozymes) Tarayre and Thompson (1997)
FST = 0.24∗∗∗ (cpDNA) Tarayre et al. (1997)
Endemic species
Antirrhinum mollissimum Southern Spain GST = 0.11 Mateu-Andrés (1999)
Antirrhinum microphyllum Southern Spain GST = 0.04 Mateu-Andrés (1999)
Antirrhinum valentinum Southern Spain GST = 0.48 Mateu-Andrés and
Segarra-Moragues (2000)
Argania spinosa Throughout Morocco GST = 0.60 (cpDNA) El Mousadik and Petit
GST = 0.25 (allozymes) (1996a,b)
Brassica insularis Corsica GST = 0.11∗∗∗ Hurtrez-Boussès (1996)
Centaurea corymbosa Subpopulations 250 m–2.5 km FST = 0.34∗∗∗ Colas et al. (1997)
Cytisus aeolicus Eolian Islands GST = 0.01 Conte et al. (1998)
Cyclamen balearicum Among regions in France FST = 0.42∗∗∗ Affre et al. (1997)
Among the Balearic islands FST = 0.11∗∗∗
Among populations (Cévennes) FST = 0.26∗∗∗
Among populations (Majorca) FST = 0.16∗∗∗
Cyclamen creticum Crete FST = 0.17∗∗∗ Affre and Thompson (1997b)

Table 3.2 (Continued)

Species Location and scale of study Estimate of differentiation Reference

Fagus moesiaca Balkans (allozymes) FST = 0.02 Gömöry et al. (1999)

Silene diclinis Spain GST = 0.06 Prentice and Anderson (1997)
Narcissus longispathus Spain θ = 0.15∗ Barrett et al. (2004a)
Quercus alnifolia Cyprus GST = 0.06 Toumi and Lumaret (1991)
Thymus loscosii North-east Spain GST = 0.03 López-Pujol et al. (2004)
Triticum dicoccoides 100 m transect GST = 0.26 Nevo et al. (1988a)
subpopulations GST = 0.41 Golenberg (1987)
∗ ∗∗∗
P < 0.05, P < 0.001. For some studies I employed reported CI values to test for significant differentiation at P < 0.05.

effects will also influence genetic variation in a diversification in Oenothera (Klein 1970) in asso-
similar fashion for both endemic and widespread ciation with chromosome rearrangement and the
species. Finally, endemic species are not neces- colonization of contrasting and often extreme eco-
sarily rare in terms of their population sizes or logical conditions since the start of the Pleistocene
even in terms of the number of populations in a in the Mediterranean-climatic region and deserts of
given area, they can show considerable variation in California also illustrates this pattern.
the size, number, and spatial organization of their Within the constraints of geographical barriers to
populations. For all these reasons, levels of differ- dispersal, climatic change at the end of the Tertiary
entiation will vary among endemic species, which and during the Quaternary had important effects on
as a group, are unlikely to differ from widespread species’ distributions in the Mediterranean region
species. (Chapter 1). Climatic variation caused local extinc-
tions and genetic isolation among populations and
thus the potential for species divergence. Indeed,
3.3 Climatic rhythms and the impact of Pleistocene glaciation on population
differentiation genetic structure has been documented in several
Taxonomic diversity in the different Mediterranean European forest tree species (e.g. R.J. Petit et al. 1997).
floras of the world has frequently been inter- In this section, I present evidence for the idea that
preted as resulting from major climatic changes. Quaternary climatic rhythms have contributed to
For example, a large proportion of the taxonomic species divergence and population differentiation in
diversity of the Cape Flora in South Africa may different glacial refugia, and thus some of the dis-
have been ‘generated during massive bursts of junct distribution patterns that characterize groups
speciation associated with the range extensions of related plants in the Mediterranean.
of the sclerophylous vegetation that survived the
drastic climatic changes at the Tertiary-Pleistocene
3.3.1 Species divergence in relation to
boundary’ (Linder et al. 1992: 132). Climatic oscil-
climate change
lations associated with glaciations may have fur-
ther promoted speciation in this region during the The isolation of populations in association with cli-
Pleistocene (Midgley et al. 2001). Molecular analy- mate change may have contributed to two common
sis of gene sequence variation in the genus Phylica distribution patterns in the Mediterranean flora (see
(Rhamnaceae) from the Cape Floristic Region illus- Chapter 2): (a) the restricted distribution of (schizo-)
trates how extensive diversification beginning at endemic species that have diverged following frag-
∼7–8 Ma coincided with marked aridification of the mentation and isolation of parts of the distribution of
climate in association with changes in ocean currents more widespread ancestral species and (b) east–west
during that period (Richardson et al. 2001). Rapid vicariance.

A genus which has provided key information a distinct lack of differentiation among taxa in terms
on differentiation and divergence in relation to cli- of their cpDNA haplotype profiles and nuclear gene
mate change in the Mediterranean is the genus sequence variation. Several species in the section
Senecio. In a combined study of genetic variation Senecio share the same range of cpDNA haplotypes
for cytoplasmic and nuclear genes in 18 species, and show a phylogenetic tree (based on nuclear
Comes and Abbott (1999a, 2001) illustrate a num- gene sequence variation) with very short terminal
ber of interesting features associated with the branches and a lack of resolution among species
diversification of Mediterranean taxa. They report (Fig. 3.1). The lack of distinct differentiation among


leucanthemifolius [Morocco] A, B, C, F, J
squalidus [C ltaly] A, B
glaucus [Israel] A, B, C, F J

Glaucus group
1 72 3
flavus ssp. breviflorus [Israel] F
glaucus [Morocco] A, C, F
4 4 A, B, C, F
gallicus [Spain]
1 B
chrysanthemifolius [Sicily]
3 1 A, B
aethnensis [Sicily]
5 B,C,F, J
vernalis [Israel]
Vernalis group

6 6 A,B,C,E
83 vulgaris [Israel]
1 1 0
squalidus [Greece (Mistras)] C
89 0 vernalis [Germany] C

Glandular tetraploids

1 viscosus [Romania] D
82 2 nebrodensis [Spain (Cantabrian Mts.)] M, D

5 85 0 nebrodensis [Spain (S. Nevada)] D

74 1
lividus [Spain] M
95 2 M
sylvaticus [Spain]
flavus ssp. flavus [Sinal] N
2 4
aegyptius [Egypt] O

malacitanus P

Figure 3.1 Molecular phylogeny of 19 accessions (14 species) of Mediterranean Senecio section Senecio based on ITS sequence data. CpDNA
haplotypes detected in each species/accession are provided for comparison of nuclear and cytoplasmic differentiation (reproduced with permission
from Comes and Abbott 1999a). Numbers of nucleotide substitutions are shown above the branches, and bootstrap percentages (100 replicates)
are shown below.

taxa is particularly apparent for the widespread shores of the Mediterranean became too cold under
diploid species which inhabit the western and the influence of glaciation for the persistence of the
central part of the Mediterranean region and whose former taxon.
cpDNA diversity includes the range of variation in
related endemic diploid taxa (Comes and Abbott
3.3.2 Population differentiation in
1999a). For Comes and Abbott (2001: 1953) this lack
Mediterranean glacial refugia
of phylogenetic resolution ‘reflects a real historical
phenomenon of a near simultaneous and relatively The analysis of population differentiation within
recent diversification’. The timescale for differenti- species provides further evidence for the role of
ation in the Mediterranean Senecio species, which climate change in species divergence. Isolation in
probably began at ∼1 Ma, is indicative of diver- glacial refugia will have created the context for
sification as a result of rapid and repeated climate genetic differentiation to occur, and in some cases
change. may have pushed populations into new evolution-
Molecular phylogenetic analysis of the silver ary trajectories. The degree of such geographic isola-
saxifrages (Saxifraga Section Lingulatae) by Conti tion probably varied greatly for different species.
et al. (1999) illustrates the evolution of schizo- Many species probably had highly contracted dis-
endemic patterns of distribution in relation to tributions as they retreated into one or a small
climate change. In this group, narrow endemic number of geographically distinct refugia in south-
species with disjunct distributions, for example, ern Europe (Iberian peninsula, Italy, Balkans).
Saxifraga cochlearis which is endemic to limestone Patterns of genetic differentiation in several tem-
rocks in a small area of the Maritime Alps and on perate forest tree species across Mediterranean
the Ligurian coast and Saxifraga crustata which is and temperate regions of Europe (e.g. R.J. Petit
endemic to limestone and dolomite in the south- et al. 1997; Gömöry et al. 1999) confirm histor-
east Alps and the Balkans, appear to have evolved ical palynological data (Huntley and Birks 1983)
from populations of widespread Saxifraga paniculata which suggest the persistence of tree species in geo-
during periods of isolation in distinct glacial refuge graphically separate Mediterranean refugia located
populations. A similar process probably occurred in in the southern Balkans, Italy, and the Iberian
mountain populations of Abies and other tree gen- peninsula. Strong patterns of east–west genetic dif-
era in the Mediterranean region (Quézel and Médail ferentiation for molecular markers detected in two
2003). emblematic Mediterranean trees, the cork oak Quer-
Mediterranean oaks provide a well-known cus suber (Toumi and Lumaret 1998) and wild olives
example of east–west vicariance (Chapter 2). In the Olea europaea subsp. europaea var. sylvestris (Besnard
western Mediterranean, two closely related ever- et al. 2002; see Chapter 6) are consistent with these
green taxa have been described: Quercus ilex with propositions. The molecular divergence of Turkish
elongated leaves bearing 8–9 nerves and Q. rotun- and Iberian populations of Rhododendron ponticum
difolia with rounder leaves with 6–8 nerves. Q. ilex (Milne and Abbott 2000) is also thought to have
occurs in temperate, subhumid and humid biocli- occurred during episodes of Pleistocene glaciation
mates whereas Q. rotundifolia occurs more often in since there is good fossil evidence that this species
drier habitats. In southern France intermediate habi- was present in the Alps in the Mindel-Riss inter-
tats contain both types and a range of variation glacial (∼0.25 Ma) (Jessen et al. 1959). Other species
in morphology. The causal relationship and adap- may have persisted in smaller fragments of suitable
tive significance of this variation remains unknown, vegetation in sheltered sites dotted across the land-
and it is probable that the two taxa represent the scape on the northern shores of the Mediterranean.
extremes of morphological variation in a single vari- Finally, for some species, changes in distribution
able complex (Barbero et al. 1992). It is thought that may have been less dramatic, species distributions
the divergence of Q. rotundifolia and Q. ilex arose becoming fragmented as they persisted in small
during periods of refuge when the most northern ‘peri-glacial’ refugia.

During the Pleistocene, many species will have of this pattern has been revealed by Hampe et al.
experienced several periods of expansion and (2003) in a study of genetic variability in the differ-
retreat in response to climatic fluctuations. It is ent parts of the range of alder buckthorn, Frangula
thus possible that not all refugia remained inde- alnus, a small tree widespread across temperate
pendent over the course of the different glacial Europe, whose seeds are dispersed by birds. This
maxima, since some of the smaller refugia may species has relict populations in southern Spain,
have been part of colonization routes from major northern Morocco, and in Anatolia. Populations
refugia. This scenario has been proposed for Fagus in these disjunct regions are highly differentiated
sylvatica and related beech tree species whose from one another (different subspecies have been
re-colonization of Europe may have occurred from described) and the populations in southern Spain
either non-differentiated or perhaps only one refugia have maintained high levels of variation. The latter
(Demesure et al. 1996; Gömöry et al. 1999). Beech do not appear to have contributed to post-glacial
populations in Calabria are genetically more sim- re-colonization, which has been exclusively from
ilar to Balkan beech than to beech distributed across refugia in the Balkans. An important result is thus
southern and temperate Europe (Gömöry et al. 1999). that relict populations in Spain contain a unique
Populations from the central part of the Dinarian store of genetic variation absent from the rest of
region are intermediate between these two southern the species’ range. The conservation value of these
extremes, indicating historical connections between refuge populations is thus clear.
Calabria and the Balkans, perhaps during the The geographic genetic structure of populations
Pliocene. Further evidence for this connection can of herbaceous plants also provide evidence for dif-
be seen in the close genetic link among populations ferentiation in glacial refugia. Molecular analysis of
of Senecio rupestris in central Italy and the south- Senecio gallicus across its distribution in the west-
ern Balkans (Abbott et al. 2002). The latter study ern Mediterranean nicely illustrates how current
also illustrates how populations in refugia are more day patterns of differentiation can provide revealing
genetically variable than those in formerly glaciated evidence on the historical process of re-colonization
areas. In maritime pine, Pinus pinaster, patterns after glaciation (Comes and Abbott 1998, 2000). This
of population differentiation have also been inter- widespread outcrossing diploid, which like many
preted in relation to climatic history. Three distinct Senecio species has a marked capacity for relatively
groups of populations (Atlantic, North African, and long-distance dispersal, occurs in open ruderal hab-
Ibero-Tyrrhenian that includes populations from itats, on dunes, river banks, and sandy soils in fields
Andalousia to south-east France, Liguria, and the with more natural vegetation, in coastal and inland
islands of Corsica and Sardinia) have been identified regions of the western Mediterranean from southern
on the basis of molecular genetic and biochemical France across the Iberian peninsula. Comparative
markers (Petit et al. 1995; Salvador et al. 2000). analysis of cpDNA and nuclear allozyme variabil-
Migration out of glacial refugia in Andalousia and ity across this distribution has shown that allozyme
the central Spanish mountains is thought to have variation has a relatively random spatial structure
created these distinct groups in different areas. with low variation, and cpDNA haplotypes show
Much work has been done with the aim being to a significant increase in the frequency of a derived
identify the source populations, and better under- haplotype and a decrease in the number of haplo-
stand the process, of post-glacial re-colonization types from coastal to inland sites. Different coastal
of temperate Europe. Such work has often shown sites showed significant differences among each
that some refuge populations have not contributed other in the frequency of cpDNA haplotypes, a
to the genetic diversity of re-colonizing populations. pattern not shown by inland populations where
The contemporary populations present in these the frequency of cpDNA haplotypes showed little
refuge areas may thus retain a unique store of genetic variation (Fig. 3.2). Variation in quantitative mor-
variation, not present elsewhere in the widespread phological traits and RAPD markers (Table 3.1)
distribution of the species. An illustrative example also showed differentiation between coastal and

inland populations indicating that the phylogeo- still possible that directional selection has favoured
graphic divide for cpDNA also exists for other mark- cpDNA haplotype ‘G’ (Fig. 3.2) in inland habitats,
ers, indicating that the geographical uniformity of either directly, due to some form of ecological asso-
allozymes is more a result of low rates of evolution ciation, or indirectly, via linkage to other genes
or low genome representation than high contempo- involved in adaptation to inland habitats.
rary pollen flow. Recent range expansion associated The population genetic structure of endemic
with a more rapid rate of cpDNA lineage sorting C. balearicum, a forest dwelling geophyte on rocky
may be the cause of the above patterns. north-facing slopes in the western Mediterranean
Founder events during colonization and sub- also provides evidence that glaciation, by virtue
sequent genetic drift in isolated populations can of its effects on species distribution and popula-
create sharp patterns of genetic differentiation in tion sizes, may have had a strong impact on levels
plant populations. During glacial maxima S. gallicus of population differentiation. This species illus-
would have been absent from the north-west and trates the pattern of disjunct endemism discussed in
central parts of the Iberian peninsula and its alti- Chapter 2: its distribution encompasses two distinct
tudinal range would have been more restricted. regions where the scale of isolation is similar, but
The genetic structure of coastal populations sug- the history of isolation is different (Fig. 3.3(a)). First,
gests that S. gallicus persisted in distinct coastal it occurs on several of the Balearic islands where
refugia, out of which it has recently spread to cover populations on different islands have been isolated
its present range. Its recent advance may have from one another since at least the last glaciation
contributed to the discordance between cpDNA period and perhaps longer, particularly those on the
and nuclear allozyme variation. Nevertheless, it is island of Ibiza compared to those on the Gymnesian







0 200 400 km

cpDNA haplotypes

Figure 3.2 Geographic distribution of cpDNA haplotypes in S. gallicus in the Iberian peninsula and southern France. Reproduced with
permission from Comes and Abbott (1998).

(a) This would have created the template for differenti-

ation to occur. Human activities (forest clearance
and grazing) in combination with its ecological hab-
itat requirements (Debussche and Thompson 2003)
and reliance on ant dispersal (Affre et al. 1995) would
have restricted the subsequent spread of this species,
maintaining isolation among ‘habitat islands’ in
southern France. Following severe bottlenecks on
population size, the highly inbreeding nature of this
species, which is rarely if ever visited by pollin-
ators and which can self in the absence of pollinators
(Affre et al. 1995; Affre and Thompson 1999), will
have promoted gene fixation via random drift. In
(b) 0.5 fact, populations in southern France show marked
fixation of alleles in each population, whereas popu-
lations on the Balearic islands tend to contain more
than one allele at each studied locus per population
FST (± SE)

0.25 (Affre et al. 1997). In contrast, populations on the

Balearic islands may have been larger and more con-
nected due to lower sea levels in the Quaternary. As
an epilogue, this is a story to be followed, in the wake
0 of forest advance on abandoned terraces in southern
France Spain France Spain France, the species may be currently re-expanding
Among islands Among populations its local distribution in some areas (Chapter 4).
Some Mediterranean mountain species, may have
Figure 3.3 The (a) distribution and (b) genetic differentiation persisted during the Quaternary in peri-glacial
(mean FST values) among populations of Cyclamen balearicum on
refugia, with little change in range size in response to
the Balearic islands and populations in habitat fragments in southern
France (redrawn from Affre et al. 1997). glaciation. An illustration of such distributional sta-
sis has been reported in Anthyllis montana, a species
islands of Majorca, Menorca, and Cabrera. Second, which occurs in low mountains around the western
C. balearicum occurs naturally in five geographic- Mediterranean (Kropf et al. 2002). Phylogenetic ana-
ally isolated zones in southern France. Affre et al. lyses across the distribution of A. montana revealed
(1997) reported that genetic differentiation among a major genetic subdivision between eastern popu-
populations in different habitat islands in southern lations (Greece, Balkans, Italy, Maritime Alps) and
France was more than twice that among popula- western populations in France (French Alps and
tions on the different Balearic islands (Fig. 3.3(b)). the Cévennes), and Spain (Pyrenees, Cantabrican
By comparing the amount of differentiation among Mountains, and Sierra Nevada). The western and
populations on the largest of the Balearic islands eastern lineages correspond to two different sub-
(Majorca) with that among different populations species which are separated by the Alps where the
in one region in southern France (the Cévennes), species is uncommon. The absence of this species
these authors also found that population differenti- from what one would predict to be suitable habitats
ation was greater among populations within habitat in the Alps and the distinct pattern of geographic
islands compared to on true islands. The most plaus- differentiation on either side of the Maritime Alps
ible explanation for these patterns is that glaciation probably reflect a long-standing separation at the
hammered population size and number in southern population level and thus stasis in distribution pat-
France, reducing the distribution of this species to a terns. The extant distribution of some populations
small number of isolated pockets in sheltered cliffs. in the south-eastern margins of the Alps coincides

with an area suggested to represent sites of peri- (e.g. Arum pictum, Cyclamen balearicum, Orchis
glacial refugia for different alpine taxa (e.g. Stehilik insularis, Soleirolia soleirolii, and Arenaria balearica)
2000). So in some microenvironments a number show a lack of distinct geographic variation in
of Mediterranean mountain taxa may have found morphology among the geographically disjunct
refuge during glacial maxima in a fairly wide part parts of their range.
of their current distribution. Other taxa with disjunct and endemic distri-
butions do however show evidence of distinct
morphological differentiation among closely related
3.4 Divergence in peripheral and species (e.g. (a) Erodium corsicum (endemic to Cor-
marginal populations: isolation, sica and Sardinia) and Erodium reichardii (endemic
inbreeding, and ecology to Majorca and Minorca) and (b) Pastinaca latifolia
(endemic to Corsica) and Pastinaca lucida (endemic
3.4.1 Morphological differentiation in
to Majorca and Minorca)) and at the subspecies
disjunct populations
level (e.g. (a) Helleborus lividus subsp. corsicus
In Chapter 2, I outlined how closely related wide- (endemic to Corsica and Sardinia) and subsp. lividus
spread and endemic taxa can show significant (endemic to Majorca and Cabrera), (b) Sesleria insu-
levels of variation in biological traits and ecological laris subsp. cordata (endemic to Corsica) and subsp.
requirements. Precise quantitative investigation of insularis (endemic to Sardinia and Majorca), (c) Alnus
the degree of morphological differentiation among viridis subsp. viridis (Alps) and subsp. suaveolens
disjunct but closely related taxa or among disjunct (Corsica), and (d) Herniaria latifolia subsp. lat-
populations of a single taxon have however rarely ifolia (Pyrenees and Spain) and subsp. litardierei
been performed. (endemic to Corsica and Sardinia)). Several species
Several studies attest to a lack of population with disjunct distributions on the Pityusic islands
differentiation in morphology among highly (Ibiza and Formentera) and the Iberian penin-
disjunct populations. Despite a long history of sula also show differentiation in morphological
isolation on small fragments of what were once traits suggesting recent geographical differentiation
larger microplates or connections of land, many (Contandriopoulos and Cardona 1984).
species with disjunct endemic distributions show Precise analysis of morphological variation within
little or no morphological differentiation on the and among closely related endemic taxa may also
now disjunct fragments or islands. For instance, in blur taxonomic distinctions. In a study of quant-
the southern Aegean, 23 species have distributions itative morphological variation in a group of three
that encompass western Crete and Andikithira closely related Cyclamen species (Plate 1), Debussche
(between Crete and the Peloponnese) or eastern and Thompson (2002) showed that C. creticum
Crete and Karpathos, that is, across what are (endemic to Crete) should be considered as a geo-
thought to be important phytogeographical barriers graphic subspecies of the widespread C. repandum.
(Greuter 1972). In 21 of these cases, no significant A conclusion which concords with a phylogeo-
morphological differentiation was observed, that is, graphic study of this group (see below). This
no varieties or subspecies on the different islands, raises the question of how different closely related
and just two groups showed differentiation on endemic species really are and illustrates a central
Crete and nearby islands: Campanula saxatilis on theme of this chapter: what differs among species
western Crete has subsp. cytherea on Andikithira often varies considerably within species.
(and Kithira) and Silene ammophila on eastern
Crete has subsp. carpathae on Karpathos. Such
3.4.2 The geographic and local scales of
morphological stasis is surprising given the length
of time that such islands have been isolated. Like-
wise in the western Mediterranean, several endemic In this section, I will focus on the evolution of
species with geographically disjunct populations schizo-endemic distribution patterns, that is, where

endemic species diverge in the different parts of Indeed, local genetic differentiation is common in
the range of an ancestral species, to create a con- plants (Table 3.2; Linhart and Grant 1996) and may
temporary pattern of disjunct distributions among be an important step towards speciation (Levin
taxa with the same chromosome number. This 2000). However, it is well known that advantageous
evolution is assumed to be gradual in the different genes may spread rapidly, and thus even small
parts of the range of an ancestral taxon (Favarger amounts of gene flow may allow for geographic
and Contadriopoulos 1961). The fact that many speciation to occur (see Morjan and Rieseberg
closely related species have disjunct distributions 2004). Second, in plants there is enormous potential
around the shores of the Mediterranean is no doubt for sympatric or peripatric speciation via rapid
the source and inspiration for this idea. Favarger genetic change linked to hybridization, polyploidy,
and Contadriopoulos recognize that such endemism inbreeding, and local adaptation on fine spatial
may evolve as a result of isolation followed by scales. Finally, geographic speciation requires fairly
differentiation and divergence in allopatry or as a uniform selection pressures across the area in
result of initial differentiation in a variable taxon in which divergence occurs, an assumption which
the different parts of its range. In the latter case, is difficult to uphold for plant populations in the
‘geographic isolation of the different parts of the Mediterranean landscape (see Chapter 4).
ancestral distribution follows the initial differenti- Local speciation relies on the operation of the
ation, producing schizo-endemic taxa’ (Favarger two microevolutionary processes which promote
and Contadriopoulos 1961: 398, my translation). population differentiation. First, random genetic
There is in both cases an implicit assumption that drift in gene frequency and the fixation of new gene
speciation is geographic, although in the latter combinations due to founder events and genetic
case the authors imply that the initial impetus for bottlenecks will favour genetic drift and random
differentiation was not reproductive isolation in differentiation and thus provide a template for
allopatry. They were unable to identify the causes of local speciation, in the absence of geographic isola-
this process of initial divergence, despite reference tion (Levin 1993). Second, novel variants may be
to some cases of ecological differentiation among favoured by selection if they have an adaptive
Mediterranean and Alpine taxa. advantage. This may occur for selfing variants if
The reliance on a role for gradual accumulation they provide reproductive assurance in the absence
of genetic differences in allopatry reflects a tradi- of pollinators, favour seed set in novel ecological
tionally accepted view of the importance of spatial conditions or provide reproductive isolation from
isolation in species divergence (Levin 2000). This ancestral species. In other words, new gene com-
is no doubt an important process in long-lived binations (that appear by random fixation) may
species with efficient mechanisms of gene flow and in fact have an adaptive value for colonization
high effective population sizes. However, if one and persistence in novel ecological conditions.
analyses speciation at the scale of the population- However, theoretical work illustrates that the dis-
level processes that promote differentiation or ruption of gene flow in peripheral populations may
homogenize variation, several features of plant allow for selection to cause speciation, even in the
population ecology and evolution suggest that this absence of contributions from random genetic drift
mode of speciation may be less important than in (García-Ramos and Kirkpatrick 1997).
animals. First, gene flow is often spatially limited This twofold process may be particularly import-
in plants, hence the spread of novel genes over ant where species colonize islands or in peripheral
large population systems may require extremely populations of a widespread species. Populations
long periods of time. Recent theoretical work at the geographic margins of species’ ranges will
attests to the possible evolution of phylogeographic be particularly prone to the evolution of random
breaks without geographical barriers to gene flow and adaptive differentiation due to their small size,
in species with low average individual dispersal potential isolation, reduced gene flow, and faster
distances and small population sizes (Irwin 2002). population turnover (extinction and colonization)

than in the central part of the range (Antonovics In C. balearicum, population differentiation in
1976). Genes with no major impact on fitness in southern France is thought to have evolved as
large populations in the central part of the range a result of gene fixation via random drift in a
may confer a selective advantage in the novel small population system as a result of climatically
conditions experienced in peripheral isolates on the induced habitat fragmentation (see above). Com-
edge of the species range, where the breakdown parison of the genetic distance among populations
of developmental canalization due to rapid rise of C. balearicum and its closely related congeners
in levels of inbreeding and strong selection will C. repandum on Corsica and C. creticum from Crete
allow previously masked variation to be expressed indicates that random fixation has caused allele
(Levin 2000). The significance of an abrupt shift to frequencies in one population of C. repandum to
intense inbreeding following dramatic reductions resemble more those of populations of other species
in population size and differentiation in response than other conspecific populations (Fig. 3.4(a)). The
to novel ecological conditions has in fact long single population of C. repandum that is genetically
been stressed in the literature (Lewis 1962, 1966; distinct from the other Corsican populations is one
Raven 1964). This form of speciation may thus with a very high level of homozygosity. In this group
often be associated with a shift from outcrossing to of species it is probable that inbreeding causes pop-
selfing, that is, one of the major transitions in the ulations to diverge and show gene frequencies more
reproductive system of plants (Barrett 2002a). typical of other closely related species. A similar
In the local speciation model, a new species will pattern has been reported for populations of three
begin with a limited distribution. From then on a vari- Centaurea taxa in Mediterranean France (Fig. 3.4(b);
ety of processes may occur. Some new species may Fréville et al. 1998).
rapidly go extinct, others may establish large pop- In both Cyclamen and Centaurea, localized
ulations but only persist in a limited area, and still endemism is frequently correlated with the appear-
others may expand well beyond the site of origin (in ance of albino floral variants. Such flower-colour
allopatry or within the range of the progenitor). Spe- variation may be a common feature of random fixa-
ciation in geographically peripheral and ecologically tion and species divergence in isolated populations
marginal isolates followed by range expansion in and closely related taxa. For example, in Satureja
allopatry (and the subsequent development of geo- cordata which has two subspecies in the Balearic
graphic barriers to dispersal) could thus produce islands, one is an erect plant with purple flow-
a contemporary distribution pattern similar to that ers (subsp. rodriguezii) and one is a prostrate plant
which would occur following geographic speciation. with white flowers (subsp. filiformis). Paeonia and
Several plant groups in the Mediterranean illus- probably many other groups show similar patterns
trate how random loss and fixation of alleles in of flower colour variation in association with the
disjunct parts of a species range can create sharp isolation of taxa on Mediterranean islands.
patterns of differentiation among populations of Random changes in gene frequency due to genetic
a single species, with important implications for drift in small and isolated populations may be the
understanding the process of species divergence. cause of marked differentiation among populations
In Pinus maritima, Petit et al. (1995) documented of the chasmophyte Brassica insularis on Corsica
significant geographic differentiation (GST = 0.17) (Hurtrez-Boussès 1996) and the patchy distribution
for isozyme loci among the fragmented parts of of rare and infrequent alleles among subpopulations
the geographic range of this species in the western and populations in the rare Silene diclinis, endemic
Mediterranean (i.e. Corsica, Sardinia, Italy, Spain, to a small area of south-east Spain (Prentice 1984;
Portugal, and south-western France). In contrast, Prentice and Andersson 1997). The results of these
on a more regional scale, from northern Morocco authors (including a strong relation between pop-
to northern Spain, González-Martínez et al. (2002) ulation size and allele number) are consistent with
reported much lower genetic differentiation at the idea that the loss of gene diversity will be slower
isozyme loci among populations (GST = 0.05). than the loss of alleles in a population that has

(a) Species Location


0.5 0.4 0.3 0.2 0.1 0 F
C. corymbosa
C. corymbosa
C. corymbosa
C. maculosa subsp. albida
C. maculosa subsp. maculosa
C. maculosa subsp. maculosa
C. corymbosa
C. maculosa subsp. maculosa
C. corymbosa
C. corymbosa
C. maculosa subsp. maculosa
C. maculosa subsp. maculosa

Figure 3.4 Genetic distance relationships among populations of closely related taxa. (a) Three closely related Cyclamen species
(BA—C. balearicum, CR—C. creticum, RE—C. repandum) from Crete (CR), Corsica (CO), southern France (FR), and the Balearic islands (BA).
Based on data in Affre et al. (1997) and Affre and Thompson (1997a,b). (b) Three closely related taxa of Centaurea in southern France (redrawn
from Fréville et al. 1998).

nean Sea



e At

Agadir t iA
0 200 400 km

Figure 3.5 The distribution of Argania spinosa in North Africa (reproduced with permission from El Mousadik and Petit (1996b)).

undergone a rapid reduction in size. Work on isol- invites study of the processes causing species diver-
ated populations of the argan tree (Argania spinosa) gence at range limits (Fig. 3.6(a)). Previous work has
endemic to south-west Morocco has also shown that generally recognized three distinct species in this
rare alleles have a more scattered distribution than subgenus, C. repandum, C. balearicum, and C. creticum
common alleles (El Mousadik and Petit 1996a,b). The (Grey-Wilson 1997). In recent studies, the classifica-
distribution of these rare alleles and their concentra- tion shown in Chapter 2, with C. creticum reduced
tion in the more isolated populations of this species to a subspecies of C. repandum, has been suggested
(Fig. 3.5) cause the latter to contribute more to total (Debussche and Thompson 2002). The different taxa
diversity, despite their reduced allelic richness, than in this subgenus form a closely related and distinct
populations in the central portion of the range. unit in the genus on the basis of molecular and
Random population differentiation in isolated morphological data (Anderberg 1993; Anderberg
populations, and in the peripheral parts of the range et al. 2000), have the same diploid chromosome
of a species can thus be commonly observed and number, and all flower in the spring. These charm-
invites the question: is there a link to local speciation ing little plants can produce unmistakable carpets
in geographically peripheral populations? of small single-stemmed flowers with characterist-
ically reflexed petals in many woodlands of the
western Mediterranean in the spring. In contrast,
their fruits are almost invisible and very difficult to
3.4.3 Species divergence in western find. Due to a coiling of the pedicel after fertiliza-
Mediterranean Cyclamen tion, the fruits are brought to ground level where
The different taxa of Cyclamen subgenus Psilan- they mature in leaf litter at soil level and open to lib-
thum (Plate 1) have a distribution pattern which erate the seeds, and then are dispersed primarily by




400 km C.c.





400 km cre.

62 rhodense
cpDNA type repandum

96 repandum
cpDNA type balearicum

creticum cpDNA
vividum type peloponnesiacum

vividum cpDNA type vividum
creticum cpDNA type creticum

Figure 3.6 (a) Geographic distribution of the various taxa in Cyclamen subgenus Psilanthum. Abbreviations are as follows: C.b.—Cyclamen
balearicum,—Cyclamen repandum subsp. repandum, C.r.rh.—Cyclamen repandum subsp. rhodense, C.r.p.p.—Cyclamen
repandum subsp. peloponnesiacum var. peloponnesiacum, C.r.p.v.—Cyclamen repandum subsp. peloponnesiacum var. vividum,
and C.c.—Cyclamen creticum (now thought to be a subspecies of C. repandum. (b) Tentative geographic distribution of the five cpDNA types
in the phylogenetic tree shown in (c). CpDNA types are: bal.—balearicum, rep.—repandum, pel.—peloponnesiacum, viv.—vividum, and
cre.—creticum. (c) The single most parsimonious tree for cpDNA haplotypes of C. repandum and its two allopatric congeners, C. creticum and
C. balearicum. Numbers adjacent to branches represent bootstrap values obtained from 1,000 replications. Reproduced with permission from
Gielly et al. (2001).

ants (Affre et al. 1995). Disjunct distribution patterns other narrow endemic species in the Mediterranean,
thus represent the consequences of allopatric frag- such as Saxifraga (Conti et al. 1999; Vargas et al. 1999).
mentation and vicariance, rather than long-distance Whereas C. creticum flowers resemble C. repandum
colonization, in these species. in overall size, stigma–anther separation, and
Using cpDNA trnL (UAA) intron sequence ana- pollen–ovule ratio, they more closely resemble
lyses on samples from across the entire distribu- C. balearicum in their colour (Affre and Thompson
tion of the different taxa in this group, Gielly 1998; Debussche and Thompson 2002). Pair-wise
et al. (2001) obtained a phylogenetic tree composed comparisons of the three species show that the flow-
of five haplotypes regrouped in two main clades ers of C. repandum and C. balearicum are very different
(Fig. 3.6(b),(c)). The two clades suggest divergence from each other (Fig. 3.7). This floral variability is
in the disjunct parts of the range of C. repandum. typical of differences regularly observed between
outcrossing (C. repandum) and selfing (C. balearicum)
One clade contains samples of C. repandum subsp.
species. Indeed, controlled experiments in an insect-
repandum from Croatia, Italy, southern France,
free glasshouse by Affre and Thompson (1999) have
Corsica, Sardinia, and Sicily, C. repandum subsp.
shown that C. balearicum is capable of autonomous
rhodense from Rhodes and Kos and C. balearicum
selfing (in the absence of pollinators), while C. repan-
from the Balearic islands and southern France.
dum and C. creticum require an external vector to
In this clade, the divergent position of different
assure high levels of seed production (Fig. 3.7(e)).
samples of C. repandum subsp. repandum sug-
Field observations of pollinators have revealed
gests that this subspecies has diverged in the
that insect visitation to C. balearicum is almost
different geographic limits of its range to produce
non-existent, whereas bumble-bees are frequently
C. repandum subsp. rhodense and C. balearicum.
observed visiting flowers of C. repandum on Corsica.
The second clade contains the samples of C. repan-
C. repandum and C. balearicum also show a distinct
dum subsp. peloponnesiacum (Peloponnese
difference in their habitat conditions and overall
peninsula) and C. creticum (Crete). This clade
ecology (Debussche and Thompson 2003). Whereas
suggests an important phylogeographic separa-
C. repandum on Corsica is primarily found in either
tion of taxa in the Peloponnese peninsula and
coniferous or deciduous woodlands on a range
Crete from those elsewhere in the distribution of
of bedrocks and with an important litter cover,
this subgenus (including Rhodes and Kos).
C. balearicum (throughout its entire distribution)
Despite the fact that in many groups a major floristic occurs almost exclusively on rocky limestone sub-
division splits the Cyclades from the eastern islands strates, evergreen shrublands, and open woodlands
close to Turkey (Chapter 2), C. repandum subsp. (Fig. 3.8). To sum up, divergence of C. balearicum
rhodense is more closely related to C. repandum subsp. at the margins of the distribution of C. repandum has
repandum. Hence, although C. creticum appears to probably been closely linked to increased inbreeding
have diverged in allopatry from C. repandum subsp. and ecological specialization.
peloponnesiacum, C. repandum subsp. rhodense does The link between species disjunction, reproduc-
not appear to have evolved at the end of a chain tion, and ecology and the possibility of local specia-
from the Peloponnese peninsula across Crete and tion in this group has become a distinct possibility in
Karpathos to Rhodes. C. repandum subsp. rhodense the light of a recent discovery of what appears to be
appears to have evolved as a result of geographic a very old and disjunct population of C. balearicum.
isolation following the loss of land-bridge connec- As Stebbins (1942: 77) commented, ‘every field
tions across the Cyclades, perhaps in the Pliocene botanist can recall the thrill of excitement that comes
(Chapter 1). with the discovery of . . . a well known species far
The nested position of C. balearicum is typical outside of its normal range of distribution’. Such
of a phylogenetic pattern expected by local speci- ‘peripheral isolates’ are of key importance to the
ation (Rieseberg and Brouillet 1994). Such nested study of plant evolution in a biogeographic setting.
phylogenetic patterns have been shown to occur for It was thus with much excitement that my colleague

(a) Flowers of the three taxa

C. balearicum C. repandum C. creticum

(b) Mean petal length (mm) (d) Mean stigma/anther distance (mm)
25 2.5

20 2
15 1.5
10 1
5 0.5
0 0
(c) Mean pollen/ovule ratio (log) (e) % fruit set
2 100
1.5 80
0 0
C. balearicum C. repandum C. creticum C. balearicum C. repandum C. creticum

Figure 3.7 (a)–(d) Floral trait variation and (e) capacity for autonomous self-pollination in three taxa of Cyclamen subgenus Psilanthum
(Primulaceae) (drawn from data in Affre and Thompson 1998, 1999). In (e) open bars are autonomous selfing, hatched bars are manual selfing and
black bars are outcrossing.

F2 Max Debussche and I recently reported that

+1 C. balearicum may also occur outside of its previously
C. balearicum documented range on Corsica, on a limestone massif
(Balearic islands)
C. balearicum near the village of St Florent, very close to popu-
C. repandum lations of C. repandum (Debussche and Thompson
4 (Corsica) 2000). The cpDNA haplotypes characteristic of
1 2 C. balearicum and C. repandum subsp. repandum both
–1 +1 F1 occur in this site (Fig. 3.6(b)). The size, pollen–
3 ovule ratios, and stigma–anther separation of white-
flowered plants in the St Florent populations, where
white-flowered plants are at a high frequency, is
more similar to C. balearicum than (white-flowered)
C. repandum (Box 3.1). The ecology of this site is typ-
ical of C. balearicum habitats in the Balearic islands
and on the French mainland, and is thus quite dis-
Figure 3.8 Multivariate analysis of ecological habitat differentiation tinct from other C. repandum populations on Corsica
of C. repandum and C. balearicum. The small area of habitat
(Fig. 3.8). This site contains an immense divers-
space occupied by C. repandum with a dashed line corresponds to
three populations on a limestone massif near St Florent in Corsica ity of floral forms, in terms of the combination
where many plants have the morphology of C. balearicum. of flower colour, size, and stigma–anther separa-
Reproduced with permission from Debussche and Thompson (2003). tion. Plants can have flowers which resemble either

Box 3.1 Cyclamen on Corsica

(a) (b)



The occurrence of a high percentage of stigma–anther separation, and pollen–ovule ratio)

white-flowered plants in three populations (L1–L3) more similar to C. balearicum than to
of Cyclamen repandum on limestone in northern pink-flowered C. repandum (filled bars).
Corsica (b) compared to other sites on Corsica
(d) 100
which occur on granite or schist (c). In (a), each
% of polylorphic loci

circle is a population and the frequency of 80

white-flowered plants is the open sector of the
circle. The high frequency of white-flowered plants 60
in the ecologically marginal sites is accompanied 40
by the presence of plants with the attractive silvery
grey leaf markings typical of the dark green leaves 20
of Cyclamen balearicum, which contrast to the
lighter green and yellow patches on the leaves of B R BIC BSE BSI RSE
C. repandum. A population of C. balearicum may
thus have been historically present on this island. (e) 3
Genetic (d) and morphological (e) data strongly 2.5
separation (mm)

support this idea. (d) The percentage of 2

polymorphic AFLP loci in plants on limestone (BIC,
bicoloured flowers; BSI, C. balearicum-like flowers;
BSE, white flowers with exerted style; RSE, 1
C. repandum-like flowers) compared to allopatric 0.5
populations of C. balearicum (B) and C. repandum 0
(R) (M. Gaudeul unpublished data). (e) In the three 1 2 3 4 5 6 7 8
populations on limestone (6–8), white flowers Population
(open bars) are morphologically (size, colour,

typical C. repandum subsp. repandum, C. balearicum, Finally, the fact that C. balearicum has persisted
flowers produced by artificial hybridization among on this limestone outcrop located on an otherwise
C. repandum and C. balearicum and flowers of other ‘granite island’ (Fig. I.3) implies that initial diver-
C. repandum subspecies (Plate 1). Analysis of genetic gence occurred in geographically peripheral and
diversity using the technique of AFLP (Table 3.1) ecologically marginal populations on an unusual
has shown that all plants on the limestone massif soil type. Populations in this zone of Corsica are
(whether they resemble C. repandum, C. balearicum, very small, the number of plants barely exceed
or hybrids) combine the diversity of C. repandum and several hundred and occur in a small area com-
C. balearicum (Box 3.1). This combination of genetic pared to populations of C. repandum elsewhere on
markers, specific to one or other of the parental Corsica, where populations are usually much larger
species, tips the balance towards the idea that in terms of number and spatial extent. Differenti-
C. balearicum has been historically present on Corsica ation may have been facilitated by developmental
and that the populations on limestone represent modifications since the newly opened flowers of
secondary contact between a relict population of C. repandum are similar in size and floral design
C. balearicum and local C. repandum subsp. repandum. to C. balearicum (Fig. 3.9). An abrupt developmental
Hybridization between these species in a single area change or some form of instability in develop-
of Corsica appears to have reproduced a remarkable ment could thus have triggered the evolution of the
array of morphological recombinant types whose floral phenotype of C. balearicum. In contact with
variation encompasses the floral variability of the the parental species, plants which have persisted
entire subgenus. on limestone after the isolation of Corsica have,
Another result of the genetic analyses is that all of despite genetic introgression, maintained the floral
the plants in the hybrid site studied show evidence phenotypes of C. balearicum. In contrast, molecular
of genetic introgression between the two species and divergence has evolved in allopatry as C. balearicum
none have a genetic constituency that falls within the has increased its distribution. Although geographic
range of variation present in allopatric C. balearicum. isolation may thus have contributed to genetic
This illustrates the swamping role of gene flow in differentiation in this progenitor-derivative species
peripheral and marginal populations (Antonovics pair, which otherwise forms a hybrid swarm in
1968; García-Ramos and Kirkpatrick 1997). All of the contact zones, I would argue that this Cyclamen story
C. balearicum present in these populations are thus provides evidence for the local speciation model dis-
introgressed forms. However, many plants have cussed at the beginning of this section in which
a floral morphology which is identical to that of speciation is initiated in local populations. If this
C. balearicum, others are like C. repandum, while some interpretation was true then initial divergence must
others have a combination of traits that resemble one have been ancient, pre-dating the tectonic activity
or other of the two parental species. The mainten- which isolated Corsica from the Balearic islands and
ance of a floral morphology akin to C. balearicum, in southern France (Chapter 1).
the face of strong genetic introgression from C. repan- Several other examples of ecological special-
dum, strongly suggests a role for natural selection. ization in narrow endemic species relative to
Without wishing to be too speculative, I would widespread congeners are known in Mediterranean
suggest that temporal variability in the pollination plants (Box 3.2). The occurrence of edaphic adapta-
environment, perhaps due to the absence of polli- tion in geographically peripheral and ecologically
nators or years with a early summer drought and a marginal populations may thus have been an
need for rapid flowering and fruiting, may favour important feature of the evolutionary process that
the persistence of a highly selfing strategy in these has given rise to the plethora of narrow endemic
sites. Without some selective advantage, it is hard species in the Mediterranean. However, in none
to understand what maintains the floral phenotype of these cases is there evidence for adaptive
of C. balearicum in these sites given the levels of differentiation, that is, that ecological differentiation
introgression observed. contributes to fitness. To understand more fully the

Figure 3.9 From left to right, 3-day old flower of C. repandum subsp. repandum, newly opened flower of C. repandum subsp. repandum,
and a 3-day old flower of C. balearicum.

role of ecological differentiation in local speciation for long-distance dispersal, isolation can hardly be
will require that adaptive variation be identified. countered by gene flow.
3. High seed germination rates and lack of dorm-
3.4.4 Random differentiation in an ancy prevent the establishment of seed banks that
archipelago system would normally buffer the effects of genetic drift.

In the Aegean Sea the multitudinous islands of dif- The different taxa of Erysimum also show marked
ferent size and ecology and the history of island differences in their ecological distribution, Erysimum
isolation have created a patchwork system in which corinthium occurs primarily in maritime cliffs,
many species groups may have diversified. In this Erysimum senoneri subsp. senoneri occurs inland at
part of the Mediterranean, small population sys- low altitude whereas E. senoneri subsp. amorginum is
tems, with marked among-population isolation, are found only in cliffs above 500 m elevation, Erysimum
a characteristic feature of chasmophytic species that naxense and Erysimum rhodium occur at intermediate
inhabit cliffs and crevices, particularly on hard lime- elevations (between 500–800 m and 200–600 m on
stone (Chapter 2). Studies of two genera in this Naxos and Rhodes, respectively). So selection for
region, Erysimum and Nigella, illustrate random ecological specialization may have played a role in
differentiation in such small population systems. population differentiation in the different parts of
The different diploid taxa of Erysimum sect the range of this group.
Cheiranthus which occur in the sheltered parts of The different taxa most probably evolved follow-
cliffs in the Aegean region show a classic schizo- ing an initial break up of continental areas and
endemic distribution pattern (Fig. 3.10). In this island isolation during the Pliocene. The amount of
group a range of population characteristics interact morphological differentiation that occurs among the
with biological traits to facilitate the occurrence of different taxa closely parallels the history of isolation
genetic drift (Snogerup 1967). among their current distributions. For example,
1. 50% of populations have less than 50 individuals E. naxense (endemic to Naxos) is morphologically
and taxa are self-compatible. more similar to the different subspecies of E. senoneri
2. The cliff habitat isolates populations across the than to other species. Successive land-bridge con-
landscape. Since seeds are large with no adaptations nections across this area during glacial maxima in

Box 3.2 Examples of ecological differentiation among closely related species in the

Comparative studies of endemic and widespread nutrient-poor acidic sandy soils in southern
species in the Mediterranean (Chapter 2), the Spain and Morocco show significant differences
analysis of endemism in Mediterranean South in soil chemistry and shading (Ojeda et al.
Africa and patterns of habitat variation in 2000b).
Cyclamen mentioned in this chapter have revealed 4. Antirrhinum lopesianum only known from one
a consistent pattern of ecological differentiation population in Spain and Portugal occurs on
among closely related species. This pattern serpentine soils whereas related Antirrhinum
suggests a possible role of ecological mollissimum and Antirrhinum microphyllum have a
differentiation in marginal populations for species more typical rupicolous habitat (Mateu-Andrés
divergence. There are several other examples of 1999).
ecological differentiation which can be used to 5. Several genera contain closely related calcifuge
strengthen this claim. and calcicole species which may have diverged in
relation to substrate, e.g. Pinguicula
1. In western Mediterranean Senecio (Fig. 3.1) (Contandriopoulos 1962) and two sclerophyllous
two species, Senecio gallicus and Senecio petraeus oaks: Quercus calliprinos on calcareous soils and
show a clear progenitor–derivative relationship the endemic Quercus alnifolia on ultrabasic rocks
(Comes and Abbott 2001). S. petraeus is endemic on Cyprus (Barbero et al. 1992).
to a small calcareous mountain range in southern 6. The six subspecies of Pinus nigra, which have
Spain whereas S. gallicus has a distribution which an almost completely vicariant distribution
covers the Iberian peninsula and stretches into (Fig. 2.1; Barbero et al. 1998), all occur in the
southern France. humid and/or subhumid bioclimatic zones in the
2. The narrow endemic Saxifraga cochlearis shows supra- or montane-Mediterranean belts. The
ecological specialization relative to its widespread different subspecies also differ in the range of
progenitor Saxifraga paniculata (Conti et al. 1999). substrates they occupy (Quézel and Médail 2003).
3. Despite broadly similar ecological requirements, The subspecies with the most generalist substrate
three heathland Erica species which occur on preferences are the most widespread.

the Pleistocene may have delayed the differentiation The confinement to mostly maritime limestone
of this taxon on Naxos, where E. senoneri does not cliffs, the taxonomic isolation of many of the species,
occur. This looks suspiciously like another example and the pronounced local differentiation all concord
of the splitting off of an endemic species with a with the idea that the contemporary cliff flora is a
restricted distribution within the range of a more relictual version of a more extensive flora that inhab-
widespread species, as a result of random drift in ited the coasts of Crete and the Aegean islands in the
small populations. Several populations show evid- Pliocene. The geographic setting and the historical
ence of variation in morphology and a reduced framework for the patterns of isolation and differen-
fertility which could result from genetic instability tiation in this group of plants has made Snogerup’s
in small inbred populations. In the more widespread (1967) work well known, but we still do not know
Erysimum candicum on Crete and E. corinthium in how much of the variation is truly genetically based.
Greece populations are more uniform, suggesting Quantitative genetic studies and a molecular phylo-
that historical and contemporary gene flow on larger geographic study of Erysimum in and around the
areas of land have prevented the differentiation of Aegean would be most useful here.
populations in these taxa and their diversification In his studies of the genus Nigella, and in par-
into new taxa. ticular the Nigella arvensis complex, Strid (1969,


2A 4


100 km

Figure 3.10 Distribution of the different taxa in Erysimum sect. Cheiranthus in the Aegean region. (1) E. candicum: (A) subsp. candicum
and (B) subsp. carpathum. (2) E. senoneri: (A) subsp. senoneri, (B) subsp. icarium, and (C) subsp. amorginum. (3) E. corinthium,
(4) E. naxense, and (5) E. rhodium (redrawn from Snogerup 1967).

1970, 1972) provides a parallel illustration of dif- populations are responsible for the pattern of differ-
ferentiation in a small population system in the entiation. Nuclear DNA sequences suggest a recent
Aegean islands, which he argued to be the result of diversification at 1.7–2 Ma. For cpDNA variation,
allopatric differentiation as a result of genetic drift. they have found that >80% of variation is due to
He provided a comprehensive account of distinct differences among island populations, indicating
morphological variation among the different taxa that cytoplasmic gene flow is too low to prevent
which make up the N. arvensis complex (Fig. 3.11(a)). differentiation due to drift in small isolated frag-
On different islands in the Aegean Sea, morpho- ments. Diversity was markedly higher among island
logical variation is highly discontinuous among the populations than among continental populations
different taxa which differ strikingly in growth habit and there was no evidence for isolation due to
and a range of morphological characters. Across distance. In fact all the analyses made by these
this archipelago there is thus a mosaic of non- workers concord with the hypothesis proposed by
overlapping distributions of morphologically dis- Arne Strid: allopatric fragmentation during periods
tinct species and subspecies on different islands of climate change and isolation of small populations
(Fig. 3.11(a)). These sharp morphological discon- have set the scene for random differentiation due to
tinuities among island populations stand in stark genetic drift. Patterns of morphological differenti-
contrast to more clinal morphological variation ation in Ranunculus species on the Aegean islands,
among continental populations of two subspecies where ‘almost every population could be regarded
of N. arvensis (Fig. 3.11(b)). In current work on the as a separate taxon’ (Dahlgren and Svensson 1994:
nuclear and cpDNA sequence variation in 60 popu- 268) further illustrate the extent of morphological
lations of the N. arvensis complex in the same variation that can be observed in this archipelago.
region, H. Bitkau and H.P. Comes (University Studies of the distribution patterns of other species
of Mainz, unpublished data) have analysed in in this region provide further evidence that patterns
detail the hypothesis that fragmentation during of differentiation result from allopatric fragmenta-
the Pleistocene and genetic drift in small isolated tion and that contemporary patterns are relicts of


2C 1B
1A 2B



100 km

(b) 100 km

1A 1B

Figure 3.11 (a) The distribution of different taxa in the N. arvensis complex in the Aegean. (1) N. arvensis: (A) subsp. aristata, (B) subsp.
glauca, and (C) subsp. brevifolia. (2) N. degenii: (A) subsp. degenii, (B) subsp. jenny, (C) subsp. barbro, and (D) subsp. minor.
(3) N. icarica, and (4) N. carpatha. (b) Clinal variation in morphology among continental populations of two subspecies of N. arvensis in the
northern Aegean region (redrawn from Strid 1972).

the once continuous distribution patterns (Greuter the Cyclades, Crete, and Andikithira. Whereas most
1979). Nigella species have traits which suggest predom-
One member of the N. arvensis complex, Nigella inant outcrossing (although most are thought to be
degenii, shows striking morphological variation self-compatible), N. doerfleri is one of only two
among populations on the different islands where species in the Aegean region with characteristics typ-
it occurs, to the extent that different subspecies ical of a selfing species (Table 3.3). In addition, this
have been recognized in the different parts of its species bears remnants of floral structures present
range. This is in marked contrast to Nigella doer- in the outcrossing members of the genus (e.g. nec-
fleri which is morphologically more uniform across tar producing petals). The morphological reduction,

Table 3.3 Biological traits of two Nigella species in the Aegean forms in the absence of geographic barriers to gene
(from Strid 1969) flow (another example is shown in Box 3.3).
Nigella degenii Nigella doerfleri

Flowering time May–July Mid-April–May

3.5 Hybridization and chromosome
Habitat Large, mesic islands Also on low-lying
arid islands Common in plants, rare in animals, hybridization
Pollen production High Low and polyploidy are key elements in plant evolution
Seed fertility Variable High
(Stebbins 1950; Levin 1983; Abbott 1992; Thompson
Flowers Large and highly Smaller and less
and Lumaret 1992; Arnold 1997; Rieseberg et al.
coloured coloured
Morphology Highly variable Relatively uniform
2003). Hybridization and polyploidy can occur inde-
among islands among islands pendently or in concert. The incorporation of two
diploid genomes into a hybrid species which has
a complete set of chromosomes from each parental
rapid development, and arid habitat of N. doerfleri species, the process of allopolyploidy, is a pri-
fit the general pattern for the evolution of self- mary force in plant evolution. Hybridization in
ing. Based on the offspring of controlled crosses, the absence of chromosome doubling, can have
Andersson (1997) suggested that pleiotropic rela- manifold effects on plant performance and evo-
tionships between floral and leaf traits may have lution hence its recognition in the titles of some
facilitated a reduction in floral morphology of isol- of the classic papers as being an ‘evolutionary
ated populations in arid habitats in association with stimulus’ (Anderson and Stebbins 1954) or a ‘source
the evolution of inbreeding. The more uniform mor- of variation for adaptation to new environments’
phology of N. doerfleri across its range is indicative (Lewontin and Birch 1966) which more recently has
that original isolation and evolution was a highly been confirmed in the paper by Rieseberg et al.
localized phenomenon and that this species has sub- (2003), whose title reveals that ‘major ecological
sequently expanded its range across the Aegean transitions . . . [are] . . . facilitated by hybridization’.
(either along ancient land-bridge connections or by Finally, polyploidy can occur within a single species
dispersal across water barriers in recent history). (autopolyploidy) in the absence of hybridiza-
A reliance on selfing may have allowed for rapid tion via the production of unreduced gametes
reproduction on small arid islands and provided (Bretagnolle and Thompson 1995). The role of
reproductive isolation on the larger mesic islands these two processes in plant evolution took some
where contact with other taxa is possible. time to be accepted in the evolutionary litera-
Finally, crosses among N. arvensis subsp. aristata ture. The development and application of molecu-
and subsp. glauca are sterile, indicating that geo- lar tools in the last two decades of the twentieth
graphic isolation is associated with the evolution century has however allowed evolutionary biolo-
of sterility barriers (Strid 1970). Since this species gists to detect more and more clear examples of
has a fairly continuous distribution with areas of autopolyploidy and also to illustrate the hybrid
mixed populations between the subspecies, sterility origin of new species in the absence of chromosome
of hybrid types may have facilitated the evolution doubling (homoploid diploid speciation), mostly
of different forms in the extreme parts of the species based on comparative analyses of phylogenetic
range. In contrast, isolated subspecies on differ- datasets based on nuclear and cytoplasmic genes
ent islands produce fertile hybrids when crossed in (Rieseberg 1997). In this section, I provide illustra-
cultivation (as do geographically isolated Cyclamen tions of the role of hybridization and polyploidy in
and Centaurea discussed above). In continental areas, the evolution of diversity in Mediterranean plants.
reproductive isolation via hybrid sterility may thus I tackle hybridization first, using a series of examples
be a key component of taxonomic divergence by to illustrate its role in geographic differentiation,
virtue of its role in the maintenance of different ecological adaptation, and speciation.

Box 3.3 Reproductive isolation and divergence in Campanula dichotoma

(redrawn from Nyman 1991)

Campanula dichotoma contains a range of geographic barriers to dispersal, providing an

geographic morphological variation, shown here illustration of the historical connections
as different symbols, which some authors variously from the south-western Balearic islands,
classify as endemic species or subspecies (Nyman southern Spain, Morocco across
1991). The distribution patterns of the different North Africa to Sicily and
morphological variants cross contemporary southern Italy.


The distribution pattern of the four variants are hybridization, which may facilitate divergence. The
almost non-overlapping in areas where there are variants differ in floral traits and their ability to
no major geographic discontinuities, indicative of self-pollinate suggests that selfing may have
reproductive isolation. Although crosses among facilitated differentiation, reproductive isolation,
them produce seeds, the latter have poor viability. and their ability to colonize new
There are thus post-zygotic barriers to areas.

The complex and dynamic nature of the history of have been reported in the Iberian peninsula
the Mediterranean region means that fragmentation, where a total of 35 species occur (Morales 2002).
contraction, and expansion of species ranges have Of the six possible hybrid combinations among
repeatedly occurred. In this context, hybridization four species of Rhododendron which have sym-
and polyploidy play prominent roles in plant evolu- patric localities in Turkey, five showed evidence
tion. A series of recent talks at the IOPB meeting on of hybrid formation in the wild (Milne et al. 1999)
‘Plant Evolution in Mediterranean Climate Zones’ These authors illustrate that ecological differentia-
in Valencia illustrate the repeated occurrence of tion favours reproductive isolation. This introduces
hybridization in many genera in the Mediterranean, an issue which I will emphasize in this section,
for example, Limonium, Helichrysum, Phlomis, genetic changes associated with hybridization and
Armeria, Orchis, etc. In some groups, hybridization polyploidy provide the stimulus and ability for
may in fact be rife. In the genus Thymus, 60 hybrids ecological adaptation.

3.5.1 Hybridization and speciation in A second example from the genus Senecio illus-
Senecio trates how hybridization can produce variant types
which may produce a new species where reproduct-
Molecular studies of the genus Senecio illustrate
ive isolation from the parents occurs. The case in
the role of hybridization in species divergence and
point involves Senecio squalidus, a ruderal species in
the ecological success of derivative species.
Great Britain which has colonized Britain since its
Senecio flavus and S. glaucus have parapatric dis-
introduction to the Oxford Botanical Gardens from
tributions in North Africa and the south-east corner
seed sources collected on Mt Etna (Sicily). There is
of the Mediterranean region. A study of cpDNA and
now molecular evidence that the seed source used
nuclear DNA diversity in section Senecio (in which
in the introduction originated from hybridization
both species occur) illustrates reticulation due to
between two Sicilian endemic species, Senecio aeth-
hybridization (Comes and Abbott 1999a,b; 2001). In a
nensis and Senecio chrysanthemifolius (Abbott et al.
nutshell, although RAPD nuclear markers produce a
2000, 2002). Studies of morphological and isozyme
phylogenetic tree which concords with the morpho-
variation in cultivation provide two lines of evidence
logical classification of the section, cpDNA haplo-
for this claim. First, S. squalidus closely resemb-
types and sequence variation in the ITS nuclear gene
les plants derived from hybrid swarms of the
of S. flavus are more akin to those of S. glaucus than to
two endemic species present on Mt Etna. Second,
the other species that are thought to be its closest rel-
S. squalidus also contains genes present in pure
atives (Fig. 3.1). The latter result is strong evidence
populations of the two endemic ‘parental’ species.
for hybridization and subsequent introgression of
The evidence thus points to a hybrid origin of a
genes from S. glaucus into S. flavus. The combined
diploid species (all three species have the same
use of different markers shows that the introgres-
diploid chromosome number of 2n = 20), and thus
sion of cpDNA haplotypes and sequence variation
a new example of homoploid hybrid speciation (see
in the ITS nuclear region are not accompanied by
Rieseberg 1997 for the few known examples of this
introgression of the rest of the genome. Hybridiza-
phenomenon). In this case, speciation in the nat-
tion between these two species is also thought to
ural range of the different taxa has not produced
have occurred early in the Pleistocene when climatic
a new species probably because of a lack of repro-
changes may have caused distribution changes
ductive isolation. In contrast, introduction of hybrid
which in turn brought the two species into repro-
material to the Oxford Botanical Gardens in the
ductive contact. Hybridization, accompanied by
eighteenth century has allowed for stabilization of
chromosome doubling, is thought to have pro-
hybrid material and speciation to occur in allopatry.
duced a third (polyploid) species Senecio mohavensis
So although the first step in the process of specia-
sometime in the Late Pliocene or Early Pleistocene
tion occurred in the wild, via hybridization, the next
(Coleman et al. 2001). The latter now occurs as
and necessary step of stabilization and reproductive
subsp. breviflorus in the eastern Mediterranean and
isolation only occurred in a geographically isolated
the arid areas of south-west Asia and as subsp.
site. The fact that neither of the two parental species
mohavensis in south-west North America. During
have spread following their introduction to Britain
the Pleistocene, S. mohavensis is thought to have
suggests that the hybridization has conferred an
dispersed to North America by long distance dis-
ability to colonize a new environment, very different
persal, perhaps in association with bird migration
to that in which the parents occur on Mt Etna.
rather than wind dispersal (Coleman et al. 2003). The
historical occurrence of hybridization and diversifi-
cation has thus been modulated by climatic changes
3.5.2 Geographic differentiation via
which have enabled contact and then isolated taxa
hybridization in Armeria
in disjunct areas. In addition, recent hybridiza-
tion may have provided the genetic material that The genus Armeria has a holarctic distribution, with
has enabled S. flavus to diversify and expand its a centre of diversity in the Iberian peninsula where
range. ∼60% of the ∼120 mostly diploid species occur

(Fuertes Aguilar et al. 1999). Phylogenetic analyses plausible explanation for the haplotype sharing
of ITS nuclear gene sequence data for 55 samples is that there have been repeated cycles of distri-
in the Iberian peninsula (35 species and several butional changes associated with climatic warm-
samples of geographically disjunct subspecies of ing and cooling during the Quaternary glaciations
the morphologically variable Armeria villosa) is which initially allowed the three taxa (perhaps inde-
indicative of a significant role for reticulation in tree pendently) to colonize the region. Climatic oscilla-
structure (Fuertes Aguilar et al. 1999). First, the dif- tions probably caused range extensions and contrac-
ferent subspecies of A. villosa do not always group tions; allowing contact and subsequently isolating
together within a single clade. Second, the composi- introgressed populations. Once again, the action of
tion of the five main clades detected in the analyses hybridization in plant evolution has been modu-
is more congruent with geographic distribution than lated by historical changes in climate which caused
their predicted systematic relationships. This pat- distribution changes and the possibility of reproduc-
tern could be produced by lineage sorting in a highly tive contact among previously isolated but closely
polymorphic ancestral species. This is unlikely in related taxa.
Armeria, since it would require an extremely vari-
able ancestral species. Lineage sorting would also
require a loss of polymorphism (or lack of sampling)
3.5.3 Hybridization at the polyploid level
within accessions in the different clades. These two
conditions make lineage sorting an unlikely cause The combination of two complete but different
of the separation of different subspecies of A. villosa. genomes within a single individual via hybridiza-
A more plausible interpretation for the polyphyly tion and chromosome doubling is the process
in the gene tree, and thus the apparition of the of allopolyploidy. Unlike many autopolyploids,
morphologically different and geographically sep- allopolyploids are almost immediately recognized
arated subspecies of A. villosa, is reticulation due to as new species due to their frequent morpho-
hybridization between a widespread ancestral sub- logical dissimilarity and their almost immediate
species of A. villosa and geographically contiguous reproductive isolation from parental diploids. The
populations of other species. In Armeria many taxa genetic diversity that this process introduces at
are cross-compatible, hence the formation and estab- the individual level, is thought to be critical for
lishment of hybrids is a distinct possibility where the spread of allopolyploid plants (Thompson and
contact occurs between different species. What is Lumaret 1992). In addition, allopolyploids, with
more, the morphology, ecology, and geographic dis- independent origins, can hybridize among each
tribution of the different subspecies in relation to other. Such hybridization may reshuffle available
the sympatric species with which there may have genetic variation and further contribute to evolution
been hybridization, support the claim that the dif- and speciation. Two examples of this phenomenon,
ferent subspecies have originated via hybridization whose domesticated relatives have become very
(Fuertes Aguilar et al. 1999). These authors inter- well-known cultivated or horticultural plants, are
pret their data by suggesting that A. villosa has the wild ancestors of cultivated wheats and garden
‘captured’ genetic polymorphism from sympatric peonies. These two groups illustrate how hybridiza-
taxa by introgression. In the case of A. villosa, failure tion at the tetraploid level has played a decisive role
to develop reproductive barriers with other species in the establishment, persistence, and evolution of
may have allowed hybridization with local sym- polyploid plants.
patric species to promote geographic differentiation Wild wheats occur in two genera, Aegilops and
within the widespread species and in so doing Triticum, which occur wild in the eastern Medi-
stimulate adaptation to local ecological conditions. terranean and the Irano-Turanian floristic province
In the Sierra Nevada, three Armeria taxa which (Harlan and Zohary 1966). These genera contain
share cpDNA haplotypes, A. villosa, A. splendens, six groups of diploids and three clusters of poly-
and A. filicaulis, currently occur in distinct altitu- ploids based on their genomic group. In the poly-
dinal belts (Gutiérrez Larena et al. 2002). The most ploids, the main ‘pivotal’ genome is identical to one

chromosome set at the diploid level. Wild species during domestication (see Chapter 6). It is in this
within a cluster differ as a result of the additional modified part of the genome that new chromosomal
genome they contain, which is not identical to one combinations, that is, ‘the material for variations
of the diploid ancestors. Polyploids thus show a on the three basic themes’ (Zohary 1965: 417), has
rather peculiar situation in that they contain one evolved. The recent rise and spread of the wild
‘unaltered’ genome, common to a group of species, polyploid wheats may thus have been favoured by
and one modified genome. This situation may have the juxtaposition of two genome complements.
been critical for their evolutionary success which The Mediterranean is an important centre of dis-
appears to have been determined by hybridization tribution and diversity of wild peonies. In addi-
among different polyploids (Zohary and Feldman tion to a group of narrow patro-endemic diploids
1962; Zohary 1965). often restricted to islands, for example, Paeonia
Diploid wheats represent species with a sound rhodia (Rhodes), Paeonia clusii (Crete and Karpathos),
taxonomic base. Each species shows limited vari- and Paeonia cambessedesii (Balearic islands) and
ation in vegetative traits and characteristic spike continental diploids with more widespread dis-
morphology and seed dispersal strategy and can tributions, for example, Paeonia broteroi (fairly
readily be classified into one of six genome groups. widespread across the Iberian peninsula), a num-
Species in different groups are isolated by steril- ber of tetraploids also occur in the Mediterranean
ity barriers. Diploids also have fairly specialized region, for example, Paeonia coriacea (southern Spain,
ecological requirements. At the diploid level, one Corsica, Sardinia, and North Africa), Paeonia offici-
thus observes distinctive morphology and habitat nalis (south and central Europe), and Paeonia mascula
specialization. A very different situation occurs subsp. arietina (south-east Europe to Turkey). Most
in the polyploids, which show marked variation of the polyploids have more widespread dis-
in morphology within species and overlap among tribution than the diploids, indicative that the
species, such that a series of morphological inter- genetic variability inherent in the polyploids may
mediates make species delimitation difficult. Some have favoured post-glacial re-colonization in the
polyploids exhibit combinations of traits observed Pleistocene. The polyploids show much evidence
in different diploid groups. The differences in dis- of reticulation in phylogenetic analyses and dis-
persal strategies observed at the diploid level do cordance between nuclear and cpDNA gene trees
not occur among polyploids, and polyploid species (Sang et al. 1997; Ferguson and Sang 2001), two
in different clusters are not fully intersterile. As a results that provide strong evidence that hybridiza-
result, the genome clusters resemble aggregates or tion has played an important role in the colonization
species complexes. potential of polyploid peonies.
In wild wheats, polyploids tend to be weedy with An intriguing and novel feature of this hybridiza-
large overlaps in geographical and ecological dis- tion is that it has involved hybrid speciation at
tributions. Colonization of disturbed open habitats the polyploid level. For example, P. officinalis has
would appear to be closely related to hybridiza- evolved after hybridization between two allopoly-
tion among different polyploids, perhaps as a result ploids, one in the group containing allopolyploid
of range changes in response to climatic variations Paeonia arietina and one related to Paeonia peregrina
and habitat availability during the Quaternary. The (Ferguson and Sang 2001). The genetic config-
common genome in each cluster may have served uration of this species in terms of similar copies of
as a buffer during the hybridization process and a particular genomes (Ferguson and Sang 2001) may
source of the pivotal genomes, which Zohary (1965: have facilitated meiotic chromosome pairing and
417) describes as the ‘main evolutionary themes for thus the initial establishment of hybrid populations.
the polyploid clusters’. The presence of a second As Ferguson and Sang (2001: 3918) point out, ‘the
genome has allowed for the introgression and dif- integration of genes from multiple genomes is
ferential modification that has evolved during the likely to have provided the variability necessary
spread of the polyploids, a process also apparent for the colonization of new, more wide-ranging

habitats’. Such integration of multiple genomes has the successful spread of the now cosmopolitan
occurred in the absence of constraints associated tetraploids as a result of the incorporation of
with further chromosome doubling. This example locally adapted genes from the endemic taxa (as in
also illustrates how current distributions can bear A. villosa discussed above). Likewise, distribution
little resemblance to historical distributions, species patterns in Cruciata species and in Knautia dipsacifo-
now present only in Asia must have coexisted lia also illustrate, to cite Ehrendorfer (1980: 49), how
at some time (probably in the late Tertiary) with ‘neopolyploids which combine genomes from dif-
European Mediterranean species for the proposed ferent diploids (or . . . polyploids) often surpass their
hybridization to have occurred. ancestors in variation, adaptability and invasion
Increased colonization capacity may thus be potential in new habitats and virgin areas’.
strongly linked to hybridization among diploids
to produce allopolyploids or among independently
3.5.4 Reproductive isolation and
produced tetraploids (an idea championed by
the maintenance of genetic integrity in
Stebbins 1985). The polyploid complex Dactylis
parental stocks
glomerata illustrates this theme (Lumaret 1988).
In this species nine diploid subspecies have dis- The above examples illustrate that hybridization
junct endemic distributions within different parts can be frequent and have important effects on
of the Mediterranean region: for example, subsp. evolutionary potential. This raises the question of
ibizensis on the Balearic islands, subsp. juncinella how species maintain their genetic integrity where
at high altitude in the Sierra Nevada and subsp. their distribution overlaps with other closely related
reichenbachii in northern Italy. This diversifica- species. Reproductive isolation can be favoured
tion at the diploid level has probably occurred by a number of traits which either by preventing
since the end of the Tertiary. Diversification of pollen exchange among species (e.g. ecological
diploids has been followed by independent pro- differences in flowering phenology, pollinators,
duction of different polyploids in the different and habitat or if one species is highly selfing) or
parts of the species range. Although the deriva- by preventing the formation of viable offspring
tive polyploids have roughly the same distribu- (e.g. cross incompatibility and hybrid inviability).
tion pattern as the diploids they are not schizo- Depending on the precise situation, different factors
endemic taxa, since they have independent ori- are likely to be involved, ecology and pollination
gins from already differentiated diploid subspecies being particularly important.
(Chapter 2). D. glomerata also contains three main
tetraploid types that have cosmopolitan distribu- Pre- and post-zygotic isolation in Mediterranean
tions across temperate or Mediterranean Europe orchids
and on the Macaronesian islands. These taxa Orchid flowers are often thought to have highly
have spread across an open landscape since the specialized interactions with particular pollinators,
last glaciation, often in association with human which may contribute to reproductive isolation
activities which have spread their seeds actively among coexisting species. In Mediterranean Ophrys,
for the creation of pastures for grazing. The different species can have distinctly shaped and
three widespread tetraploids show distinct pat- coloured flowers which produce a different bou-
terns of morphological variants in the different quet of monoterpenes in their floral fragrance,
parts of their range indicative of hybridization with traits which may differentially influence pollina-
local tetraploid subspecies (Lumaret and Barrien- tor attraction and which should limit pollinator
tos 1990). Crossing studies have shown that fer- visitation to congeners. Different species often dif-
tile hybrids can be obtained from crosses among ferentially attract male Hymenoptera, often specific
several different diploid subspecies and among dif- to particular species, hence the possibility that
ferent polyploid subspecies (Lumaret 1988). Such odour, in combination with the shape and colour
hybridization may have been a key element in of the flowers, contributes to limit hybridization

and maintain the integrity of different species in species (Comes and Abbott 1999b). The cpDNA
sympatry (Paulus and Gack 1990). Hybridization introgression may thus be sporadic and historical
has nevertheless been detected among sympatric rather than a common feature of contemporary pop-
species, for example, between Ophrys lutea and ulations, whose ecological isolation may limit such
O. fusca in North Africa (Stebbins and Ferlan 1956). introgression and maintain species integrity.
In the group of closely related species that make
up the O. sphegodes group, different species often
3.5.5 The evolutionary significance of
flower simultaneously in sympatry in southern
France and Italy. Soliva and Widmer (2003) have
shown that although genetic differentiation among Polyploidy, even in the absence of hybridization
geographically distant populations of the same among taxa, is a major process of plant evo-
species was lower than differentiation among lution which allows the different stages of spe-
sympatric populations of different species, the ciation, that is, genetic origins and population
strength of genetic differentiation among species establishment and persistence, to be observed and
was lower than that usually observed for different studied experimentally (Thompson and Lumaret
orchid species, probably as a result of some gene 1992). The occurrence of closely related taxa of
flow among species in sympatry. So, although in different ploidy level is also an integral compo-
this group the sexual deceit pollination system may nent of Mediterranean endemism (Verlaque and
be less specific than thought, perhaps as a result Contandriopoulos 1990). Finally, polyploidy pro-
of occasional mistakes by pollinators, the fairly vides fascinating model systems for the study of
specific nature of the Ophrys-pollinator interaction how taxa maintain their genetic integrity in areas
may be a key element of pre-zygotic reproductive where polyploids come into secondary contact with
isolation among taxa and thus a potentially import- their related diploids. In plants, differentiation and
ant feature of orchid speciation (Paulus and Gack diversification in association with polyploidy can
1990). As I mention later in this chapter, post-zygotic occur following hybridization among closely related
factors such as chromosome divergence may also species (see above) or by chromosome doubling
be important here. in a single taxon (autopolyploidy), usually as a
result of fusion between gametes with an unreduced
Ecological isolation in Senecio chromosome number (Bretagnolle and Thompson
Population-level analyses of genetic structure in 1995).
two morphologically distinct, highly outcrossing A key question here concerns how newly formed
Senecio species whose flowering times and geo- polyploids establish and spread. A requirement for
graphic distributions slightly overlap to form a zone the establishment of any new derivative popula-
of secondary contact in the eastern Mediterranean tion is that it attains some form of reproductive
provides instructive information concerning the isolation from its progenitors. A difference in chro-
maintenance of species differences in the face of mosome number is, of course, very important here
potential hybridization and introgression. Whereas since hybrids may be immediately checked as a
S. glaucus is typically found in sandy dunes and arid result of chromosomal imbalance. In addition, a
regions of the steppic Irano-Turanian floristic zone doubling of the chromosome number, even in the
or the Saharo-Arabian desert zone of the Near East, absence of hybridization, can have manifold con-
S. vernalis is very common in anthropogenic and sequences since the increased number of gene copies
disturbed habitats of the mesic-Mediterranean zone at individual loci allows for new genetic combina-
(Zohary 1973). Zones of potential hybridization tions which provide a template for future evolu-
occur in the eastern Mediterranean where there tion (Levin 1983). Such genetic changes may allow
is clear evidence for introgression of cytoplas- for polyploids to tolerate and colonize a wider
mic genes, although this has not impacted on range of ecological conditions by two mechanisms.
the phenotypic morphological integrity of the two First, individual plants may have increased capacity

to modify their phenotype (i.e. show phenotypic occurs because a rare cytotype, for example, a new
plasticity) in different environments. Alternatively, polyploid in an otherwise diploid population, will
the higher genetic variability may open new avenues primarily be pollinated by, and its pollen lost on
of adaptive evolution in different environments. To the stigmas of, the majority cytotype. Such inter-
understand these issues requires comparative study cytotype mating may come at a significant cost
of the distribution of diploids and polyploids in due to hybrid inviability associated with triploid
order to quantify the processes which (a) favour formation (e.g. see van Dijk et al. 1992).
the coexistence of diploids and derivative poly- Triploid inviability is common to many poly-
ploids (both in primary and secondary contact) and ploid complexes, and may lead to the selection
(b) enable the spread of polyploids. The success- of reproductive isolation where the two cytotypes
ful establishment and persistence of new polyploid are in reproductive contact. In the polyploid com-
variants may nevertheless require subsequent muta- plex D. glomerata crosses among diploids and poly-
tion, regularization of chromosome pairing, inac- ploids are for the most part incompatible (Lumaret
tivation of duplicate loci, or hybridization among 1988). The offspring that are produced are either
related but independently produced polyploids to triploid or tetraploid. In the multiple contact zones
further facilitate persistence (Stebbins 1980). which occur among diploids and polyploids of this
In the Mediterranean, polyploidy, with or with- species around the Mediterranean very few triploid
out hybridization has been of prime importance hybrids have been found (Borrill and Lindner
in plant evolution. The fact that previous authors 1971; Lumaret 1988; Lumaret and Barrientos 1990).
have been able to classify endemic species on the The absence of triploids in most areas is due
basis of ploidy variation (see Chapter 2) indicates to a combination of pre-zygotic isolation caused
the importance of ploidy variation in the flora and by differences in flowering time (Lumaret and
its endemic complement of species. Two impor- Barreintos 1990; Bretagnolle and Thompson 1996)
tant trends should be distinguished here (Verlaque and habitat differentiation (Lumaret et al. 1987)
and Contandriopoulos 1990). First, polyploidy has and post-zygotic isolation caused by the abortion
played a major role in the diversification of new of embryos or lack of successful fertilization as a
(apo-) endemic Mediterranean taxa. Second, the result of the unbalanced chromosome numbers in
flora is home to a large number of (patro-) endemic a diploid × tetraploid cross. The frequent absence
diploid species which represent the progenitor taxa of any triploids in D. glomerata contact zones is
of polyploids that have spread to occupy more wide- indicative that direct formation of tetraploids and
ranging distributions. It is thus not surprising that crosses among diploids and tetraploids occur via
diploids and polyploids do not show consistent dif- the production of unreduced gametes in the diploids
ferences in geographical distribution (Ehrendorfer (Borill and Linder 1971; Bretagnolle and Thompson
1980; Stebbins and Dawe 1987). 1995). An exception to this trend concerns the
occurrence of relatively large numbers of triploids
in a mixed diploid–tetraploid population in Israel
3.5.6 Polyploidy and ecological
(Zohary and Nur 1959). Such triploids produced
tetraploid offspring, indicating that triploids may
A common feature of distribution patterns in poly- serve as a bridge for gene flow from the diploid to
ploid groups concerns the occurrence of segregated tetraploid level. This ‘triploid bridge’ facilitates the
distributions of different ploidy levels. What this incorporation of genetic variability into the poly-
means is that polyploid complexes can be subdiv- ploid and thus enhances their probability of per-
ided into ‘single cytotype areas’ over large parts of sistence and spread. There is also strong evidence
their geographic range. These single cytotype areas of gene flow among locally present ploidy levels
are thought to be maintained by the exclusion of via the ‘triploid bridge’ in Ornithogalum umbella-
the minority cytotype which incurs a reproduct- tum (Box 3.4). Such gene flow can limit morpho-
ive disadvantage when it is rare (Levin 1975). This logical differentiation and thus maintain genetic

Box 3.4 Variation in vegetative propagation and gene flow among the different ploidy levels
of Ornithogalum umbellatum (from Moret 1991; Moret and Favereau 1991;
Moret et al. 1991; Moret and Galland 1992)

Ornithogalum umbellatum contains diploids

(2x = 18), triploids (2x = 27), tetraploids
(2x = 36), pentaploids (2x = 45), and
hexaploids (2x = 54). Diploids occur mostly in
natural mountain areas close to the coast in the
western Mediterranean, whereas higher ploidy
levels occur to the north and in
non-Mediterranean temperate Europe where they
are common in disturbed habitats.

Vegetative propagule number per plant

Polyploids above the triploid level rely on clonality
and asexual reproduction, in which a combination
of two strategies—vegetative propagation and 40
apomixis—assure population persistence. Diploids
are more dependent on normal sexual
reproduction. In diploid populations which do
practice vegetative propagation (production of 20
sister bulbs), the precise modality is different to
that of tetraploids (which propagate by bulbils).
2x 3x 4x 5x 6x


The surprisingly high pollen and seed fertility of 80

triploids, which occur in a number of disjunct
% Pollen viability

populations, illustrates that triploidy can provide

an effective means of gene transfer among ploidy 60
levels. In fact, diploid and triploid populations in
the Mediterranean region are morphologically 40
more similar to local polyploids than to other
disjunct populations of the same ploidy level.

2x 3x 4x 5x 6x
Ploidy level

There is thus strong evidence that gene flow across ploidy levels can thus be considered as a single
the ‘triploid bridge’ maintains genetic cohesion species complex in which spatial isolation is more
among ploidy levels and introduces important important than ploidy level in the evolution of
genetic variation into the polyploids. The different morphological differentiation.

cohesion in a polyploid complex as well as permit- wider range of environments. Several other studies
ting the transfer of genetic variability among ploidy support this idea.
levels. In their classic work on endemism and speciation
Unless a polyploid is reproductively isolated from in California, Stebbins and Major (1965) showed that
its parental diploids, minority cytotype disadvan- patro- and apo-endemic species differ in their eco-
tage may cause the elimination of one or other of logical requirements. Whereas patro-endemics are
the different cytotypes. An important question thus most frequent in the Central Coast Region, par-
concerns whether polyploids differ from diploids ticularly where summer fogs maintain humidity,
in terms of their ecological amplitude. A response and at mid-elevation where an old stable giant
to this question is a first step towards under- redwood forest flora has persisted, apo-endemics
standing the conditions which regulate polyploid are more frequent in areas with abrupt climatic
establishment and persistence and whether coloni- gradients between two distinct areas (where ances-
zation ability in polyploids is determined by a tral diploids may have occurred). The frequency of
greater capacity to buffer environmental variation recently derived endemics is thus closely correlated
(phenotypic plasticity) or more the result of adaptive with climatic and ecological heterogeneity, which
differentiation to spatial heterogeneity of selection may favour the establishment of hybrid derivatives
pressures. Two pieces of evidence suggest that in and the ecological segregation of autopolypoids.
a given geographical region, polyploids do not Ehrendorfer (1980) suggested that in many groups
generally have a wider ecological amplitude than diploids are more common in stable habitats of
diploids. permanent climax communities whereas polyploids
The first line of evidence comes from a study of are often found in more open, disturbed succes-
the ecological distribution of endemic species on sional communities, which often involve habitats
Corsica by Contandriopoulos (1962) who estimated that became available for colonization after the
the altitudinal distribution (in five altitudinal belts: Pleistocene glaciations. This increased coloniza-
littoral, lowlands <600 m, uplands 600–1,200 m, tion capacity may be strongly linked to hybridiza-
sub-alpine 1,200–1,800 m and alpine >1,800 m) tion among diploids to produce allopolyploids
of more than 100 diploid and tetraploid endemic (see examples above). In the genus Pinguicula,
species. If one compares the distribution of diploid the widespread species in Europe are those with
and tetraploid species in her list (it is possible to a high ploidy level, whereas diploids tend to
do so for 29 polyploids and 65 diploids) one finds have restricted ranges, for example, the patro-
that both diploids and tetraploids occur in a surpris- endemic Pinguicula corsica is endemic to the island
ingly similar number of altitudinal belts (mean = from which it gets its name. In this genus, the
2.08 for diploids and 2.07 for polyploids). Second, different species and ploidy levels differ clearly
in a study of the ecological amplitude of species in their occupation of limestone or acidic sub-
in the flora of the Pyrenees, Petit and Thompson strates (Contandriopoulos 1962). Asimilar pattern of
(1999) found no significant variation in the eco- European-wide distribution of polyploids and frag-
logical amplitude of diploids and polyploids. This mented and smaller distribution ranges of diploids
is not to say however that different ploidy levels in Mediterranean Europe can be seen in many other
occur in the same types of habitat. In fact, Petit and plant groups, for example, Plantago media (van Dijk
Thompson (1999) showed that the presence of poly- et al. 1992), Rumex acetosella (den Nijs 1983; den Nijs
ploidy in a taxonomic group is closely associated et al. 1985), Cardamine pratensis (Lihová et al. 2003),
with taxonomic diversity, which itself is closely cor- D. glomerata (Lumaret 1988), and Arrhenatherum
related with the ecological range of the taxonomic elatius (Petit and Thompson 1997).
group. This result suggests that successful establish- To quantify the capacity of diploids and tetra-
ment of polyploids, and thus taxonomic diversity of ploids to buffer environmental habitat variation
a given group, is facilitated by ecological divergence and/or adapt to variation in irradiance levels,
from progenitor diploids and adaptive differentia- Petit and Thompson (1997) studied the response of
tion in polyploids, and not the capacity to colonize a diploid and tetraploid populations of A. elatius to

artificial shading. They found that tetraploids have reproduction, is a reproductive strategy that may
greater vegetative size and flower production than have contributed to the reproductive isolation of
diploids, as reported in other species, for example, polyploids and their colonization capacity in new
D. glomerata (Bretagnolle and Thompson 2001) and areas (Thompson and Lumaret 1992). Different
Anemone palmata (Médail et al. 2002). Tetraploids groups in the Mediterranean illustrate these trends.
did not however have a greater ability to buffer
1. In Senecio sect. Senecio, diploids primarily have
environmental variation than diploids. The plas-
long well-developed outer ray florets which are
ticity of morphological traits, flowering phenology,
female in addition to the central hermaphrodite disc
and fertility were similar in the two cytotypes.
florets. All examined diploids appear to be self-
In contrast, tetraploid populations showed greater
incompatible (R.J. Abbott, University of St Andrews,
differentiation among woodland and open habitat
personal communication). In contrast, in tetraploids
populations than diploid populations, indicative
and hexaploids the outer female ray florets are
of local adaptation. In subsequent work (Petit and
absent or reduced in size (Alexander 1979) and they
Thompson 1998) the direction of selection in these
are self-compatible.
two types of habitat was reported to concord with
2. An improved capacity for vegetative reproduc-
the patterns of morphological variation, that is, taller
tion may also have allowed many polyploids to
plants favoured in open habitats. In this complex
establish in new, often disturbed, environments not
it would thus appear that adaptive differentiation
exploited by their diploid progenitors. Examples
among tetraploid populations has been an integral
of this increased reliance on vegetative propaga-
component of the spread of the polyploid cytotype
tion can be seen in Tulipa oculus-solis in the eastern
into different environments.
Mediterranean (Horovitz et al. 1972) and in differ-
A final example which illustrates the role of eco-
ent Ornithogalum species in the eastern (Kushnir
logical differentiation in polyploid evolution comes
et al. 1977) and western (Moret and Favereau 1991)
from the classic work on annual Mercurialis annua
Mediterranean (Box 3.4).
in the western Mediterranean by Durand (1963).
In this highly diverse species the three elements of In Mercurialis annua the trend towards increased
my introductory triptych, geological history, climate selfing is also apparent, but with no direct
change, and human activities, have jointly shaped link to colonization capacity. Diploids in tem-
differentiation and distribution in a single species in perate Europe are dioecious whereas the diverse
relation to ecological and genetic processes acting on Mediterranean polyploids are primarily mono-
plant diversification (Box 3.5). ecious and self-compatible and thus subject to
The argument I thus propose is that the divergence some degree of inbreeding (Durand 1963). The
and evolutionary success of polyploids is closely variation in ploidy and habitat in this com-
related to their ability to adapt to new ecological plex is thus accompanied by the evolution of
conditions. The new gene combinations present in a more inbreeding mating system. This may
polyploids, and the reshuffling of such variation explain why polyploid populations show greater
by natural selection, are likely to be key ingredi- amounts of population differentiation for morpho-
ents here. The extreme rarity of mixed cytotype logical traits than do diploid populations in this
populations, even in overlapping regions, suggests species (where morphological variation is clinal in
the strong selective importance of local ecological nature).
conditions in this process.
3.5.8 The combination of auto and
3.5.7 Polyploidy and the evolution of allopolyploidy
uniparental reproduction
Western Mediterranean Ornithogalum illustrate how
The potential for uniparental reproduction, either the combination of autopolyploidy and allopoly-
by a greater reliance on selfing or vegetative ploidy contributes to diversification within a group

Box 3.5 The distribution limits of the different cytotypes of Mercurialis annua
(redrawn from Durand 1963)

In M. annua distribution patterns of different extension through Italy and Sicily to northern
ploidy levels coincide with bioclimatic regions. Tunisia. This points to a Mediterranean refugia for
First, diploids (2x) have a mostly temperate this taxon in Calabria, Sicily, and Tunisia, during
distribution (note their presence across northern the Quaternary glaciation cycles. In Tunisia,
Spain which does not have a Mediterranean diploids are limited to humid and subhumid
climate), with a meridional Mediterranean Mediterranean belts in the north.

In contrast, polyploids only occur in the 200 km

Mediterranean. 2x

In the Iberian peninsula, hexaploids occur around

the Mediterranean coastal zones in the 2x
thermo-Mediterranean belt. Cold winters probably
restrict the distribution of hexaploids to these
coastal areas. In Tunisia hexaploids occur in the
arid bioclimatic regions and octoploids in semi-arid 6x
conditions. 4x
In Morocco only tetraploids and hexaploids are
present. The latter have a more widespread
distribution than in Spain (in a range of bioclimatic
zones), while tetraploids are confined to semi-arid
Atlantic coastal zone.

As Durand (1963: 616) points out (my translation), Italy and Sicily to Tunisia illustrates. Finally, all the
‘the adaptation of the different cytotypes to different cytotypes grow in disturbed habitats
different bioclimatic zones is not the only factor (e.g. cultivated fields, gardens, road-sides, and
contributing to their contemporary patterns of waste-grounds). Hence the role of human activities
distribution’. In addition to contemporary climate, in their dispersal, and their limits to dispersal,
distribution patterns have been strongly moulded cannot be ignored. In this group, human activities
and constrained by the geological history of land may have favoured the expansion of diploids
connections, as the historical refugia zone through rather than polyploids.

of closely related species (Moret and Galland 1992). primary and secondary habitats over a wider and
Polyploids of Ornithogalum which occur in the often more stressful range of conditions than diploid
Iberian peninsula and North Africa have a very dif- O. umbellatum in the same region. The occurrences
ferent morphology from those further north, which of distinct geographic variants on either side of the
show almost exclusive reliance on sexual repro- Straits of Gibraltar indicate that the origins of this
duction and are allopolyploid (2x = 52). These allopolyploid probably date to the Late Tertiary.
polyploids are thought to represent a distinct species This combination of autopolyploidy and allopoly-
which originated from hybridization among two ploidy also occurs in several other Mediterranean
ancestral diploids, one of which would have been plant groups, for example, Thymus (Morales
in the O. umbellatum complex (see Box 3.4). In the 2002), Mercurialis (Durand 1963), Campanula (Geslot
mountains of Spain and North Africa, allopoly- 1983; Contandriopoulos 1984), Scilla (Parker et al.
ploidy has produced a lineage which occurs in both 1991; Vaughan et al. 1997), and Triticum (Zohary

and Feldman 1962). The different ways of being and thus a recourse to diverse methods in botany,
polyploid can thus greatly contribute to the diver- ecology, and genetics, with a strong flavour of
sification of large Mediterranean genera. natural history. It was this wise combination of
The conclusion to be drawn is that the different but approaches that allowed G.L. Stebbins to do so much
closely interconnected processes of my introductory for our understanding of plant evolutionary biology.
triptych, in relation with the genetic changes asso-
ciated with polyploidy and hybridization, particu-
larly introgression among different polyploids, have 3.5.9 Karyotype differentiation and
been crucial elements in the evolutionary dynamics evolution
of many species complexes in the Mediterranean
In addition to polyploidy, there are two other major
flora. The genetic changes which accompany poly-
types of chromosome variation and evolution which
ploidy and hybridization stimulate and allow for
can influence population differentiation and species
adaptive evolution in relation to local ecological con-
divergence: quantitative variation in chromosome
ditions. Cytotype differentiation can represent a key
number and changes in structure and organiza-
element of within species differentiation in many
tion associated with breakage and reunion of chro-
taxa and thus may play a central role in the evolution
mosomes (Jones 1970; Stebbins 1971; Jones 1995;
of widespread distributions. An intriguing feature
Rieseberg 2001).
of this diversification is that it frequently occurs
without major changes in morphological traits and Quantitative variation in chromosome number. This
in several of the above-cited studies one cannot is frequent among closely related species and even
be 100% sure of the ploidy level of plants in the within species. Many examples of this phenomenon
field without recourse to chromosome counts or have been reported in Mediterranean plants, some
flow cytometry analysis of DNA content. In addi- of which involve quantitative variation without
tion, in many groups, polyploids have had multiple any ploidy level variation while others involve
origins in different parts of the range of the pro- chromosome loss and/or addition subsequent to
genitor diploid. Adaptation to novel conditions, polyploidization (Table 3.4). In the latter case, an
geographic barriers to dispersal, and the spread of amazing range of chromosome numbers can be
different taxa in association with human-induced observed within a given species or group of closely
habitat changes have all contributed to patterns of related species. One way in which such chromosome
diversification. Understanding plant diversity in increments occur is associated with the presence
such groups requires study of the joint roles of of ‘B’ chromosomes (Jones 1995), in other cases
such processes in creating and maintaining diversity chromosome number may decline via aneuploidy.

Table 3.4 Some examples of chromosome number variation within and between closely related
Mediterranean plants

Species Chromosome series (2n) Reference

Cyclamen 20, 22, 30, 34 (68), 48, 84, and 96 Grey Wilson (1997)
Nonea 14, 16, 18, 20, 28, 30, 32, 40, 44, 60 Luque (1995)
Lithodora 16, 26, 28, 32, 35, 40 Luque and Valdés (1984)
Lupinus 32, 36, 38, 40, 42, 50, and 52 Plitmann (1981)
Leucojum 14, 16, 18, 22 Contandriopoulos (1962)
Phlomis 20, 22, 24 Azizian and Cutler (1982)
Reichardia 14, 16, 18 Siljak-Yakovlev (1996)
Cardamine pratensis 16, 24, 20, 32, 40, 46, 48, 55 Lihová et al. (2003)
Erysimum grandiflorum 14, 26, 34, 38, 40, 48 Kupfer (1981)

Variation in karyotype structure. This stems mostly basis of morphological traits. Two geophytes in
from chromosome breakage and reunion which can the Liliaceae provide pertinent examples of pop-
cause deficiencies by elimination of segments, dupli- ulation differentiation in chromosomal structure
cation via the integration of segments from homolo- within species that have widespread geographic
gous chromosomes, inversion within chromosomes, distribution and similar morphology around the
and translocation of segments among chromosomes. Mediterranean.
As a result of such changes, karyotype variation
among populations or closely related taxa may occur 1. In Muscari comosum different inversion heterozy-
in the absence of any variation in chromosome num- gosities have been detected on different islands in
ber, providing a new source of genetic variation the Aegean (Bentzer and Ellmer 1975) and among
for evolution and genetic isolation. In some cases populations in continental Spain (Ruiz-Rejon and
such rearrangements may accompany a progressive Oliver 1981). Bentzer and Ellmer (1975) interpret the
reduction in chromosome number, for example, pattern of fixation of different chromosomal vari-
in species of Reichardia (Siljak-Yakovlev 1996). The ants on different islands as being the result of genetic
Mediterranean flora provides numerous examples drift. In contrast, the finding of similar patterns and
of how such variation may occur within and among types of variation in Spain, and their comparison
closely related taxa, some of which provide evidence with allozyme variability do not concord with this
of the potential importance of the fixation of chro- hypothesis, leading Ruiz-Rejon and Oliver (1981:
mosomal variants for population differentiation. In 407) to the conclusion that ‘although genetic drift
some cases chromosomal rearrangements are not cannot be rejected, we think that the interaction
associated with any observable morphological dif- between adaptive and historical factors, related to
ferences, in others they occur among taxa that also the colonization process, have played the major role
differ in morphology. in determining the geographical distribution of the
two coupled stable chromosome polymorphisms in
Two examples illustrate how variation in karyo- Muscari comosum’.
type can be accompanied by observable morpholog- 2. In Scilla autumnalis, a widespread species around
ical differentiation (Verlaque et al. 1991). the Mediterranean, studies of variation in chro-
1. Scabiosa cretica shows variation in karyotype mosome complement by Parker et al. (1991) and
organization but no variation in chromosome Vaughan et al. (1997) have detected 10 distinct chro-
number, among populations in the different parts mosome races in the different parts of the range
of its range. In this species, variation in karyo- of this species (Fig. 3.12). The variation includes
type organization among populations on different diploids (2n = 14), tetraploids (2n = 28), and
islands is accompanied by significant differences in hexaploids (2n = 42) plus changes in DNA content
overall plant morphology. and chromosomal rearrangements. There are five
2. Delphinium requienii on the Hyères islands off the haploid genome sets: A, B7 , B5 , B∗ , D, which have
south-east coast of France shows marked differentia- given rise to a range of diploids and polyploids. The
tion in the symmetry of its chromosome complement diploids occur in different parts of the distribution of
compared to its disjunct ‘sister species’ Delphinium this species around the Mediterranean, four on the
pictum, endemic to Corsica, Sardinia, and Majorca. island of Crete (Fig. 3.12). Three of the four variants
These two species are however very similar. on Crete were not detected elsewhere. Some vari-
ants are widespread, others endemic. Once again, it
Within species variability in chromosome struc- is the polyploid variants which have spread north
ture and organization may occur in the absence to colonize temperate Europe while diploids have
of any observable morphological differences, as remained endemic to localized parts of the range
in many autopolyploid complexes where diploids in the Mediterranean. Northward migration would
and polyploids are often indistinguishable on the appear to have occurred from a Balkan refugia in

B7B7B7B7B7B7 B7B7B7B7B7B7

B7B7B7B7 B7B7B7B7
B7B7 B7 B7



Figure 3.12 Geographical distribution of the chromosome variants of Scilla autumnalis. The B7 B7 B7 B7 karyotype also occurs in Britain as does
an additional type (AAB7 B7 B7 B7 ) (redrawn from Vaughan et al. 1997).

this species. The morphology of the different ‘races’ mountains (Bou Dagher-Kharrat et al. 2001). In the
is very similar and they are able to cross fertilize latter, differences in chromosome size are associated
in the glasshouse (Vaughan et al. 1997), indicating with variation in life-history among closely related
that only small parts of the genome have diverged species, as Stebbins (1950) first proposed. Differ-
in allopatry. ent species vary in habitat occupation in relation to
degree of aridity, suggesting a role for chromosome
Closely related endemic species may also show rearrangements in plant adaptation and evolution,
subtle differences in chromosome structure. In as mentioned earlier in this chapter (see Klein 1970).
three endemic Lilium species in the mountains of Fixation of chromosomal variants in the different
the Mediterranean, yellow-flowered Lilium pyre- parts of the range of a widespread species or in dis-
naicum from the Pyrenees and Cantabrian moun- junct populations may thus be a key process in the
tains, red-flowered Lilium poponium endemic to the divergence of endemic species. Variation in chro-
Maritime Alps, and yellow/orange-flowered Lilium mosome number and structure can be of critical
carniolicum which is endemic to the Balkans and the importance for the delimitation of new taxa, even
south-east Alps, Siljak-Yakovlev et al. (2003) have in the absence of morphological variation, and may
shown marked differences in the number and posi- thus be a key ingredient of plant evolution in the
tion of secondary constrictions and particular gene Mediterranean.
loci, and overall genome size (Fig. 3.13). Similar Differences in karyotype also may play a role in
types of variation in karyotype have been docu- the evolution of crossing barriers and reproduct-
mented in different groups with different life his- ive isolation. An elegant demonstration of this
tories and in different parts of the Mediterranean, has been provided for sympatric Mediterranean
for example, Vicia in the Eastern Mediterranean Orchis by Cozzolino et al. (2004). As mentioned
(Zohary and Plitmann 1979), Lactuca in the Iberian earlier in this chapter, selection for pre-zygotic isola-
peninsula (Mejías 1993), Asphodelus in the western tion and species-specific pollination may strongly
Mediterranean (Díaz Lifante 1996), and the dis- contribute to reproductive isolation and diversifica-
junct endemic Cedrus in different Mediterranean tion of orchids. However, some Mediterranean

(a) orchids share pollinators, but maintain integrity in

sympatry. The above authors show that pairs of
species in sympatry which also share pollinators
have significantly more divergence in karyotype
structure than species pairs which have specific pol-
linators. Their interpretation is that post-zygotic
reproductive barriers which result from karyotype
divergence may be an important causal factor in
orchid diversification in the Mediterranean. There
is some evidence that such karyotype divergence
may occur in peripheral isolates of widespread taxa
(b) (D’Emerico et al. 2002), another possible case of local
population differentiation that may ultimately lead
to speciation.
The importance of chromosomal rearrangements
in divergence and speciation has received a fair
amount of attention since early work which
proposed that speciation can occur when popu-
lations become fixed for chromosomal rearrange-
ments that reduce fitness when heterozygous (Lewis
1966; Stebbins 1971). In a recent review, Rieseberg
(2001) discusses the various arguments developed
in the literature which cast doubt on this idea.
(c) For this author, chromosomal rearrangements may
reduce gene flow more often through their effects
on recombination rates than through any affect
on fitness. The potential for reduced recomb-
ination offered by chromosomal rearrangements
increases the possibility of speciation where gene
flow can occur, for example, in cases of sym-
patric speciation or as a newly derived peripheral
‘neospecies’ expands its range and encounters popu-
lations a progenitor species (see also Levin 2000).
Rieseberg (2001), argues the need for additional
1 2 3 4 5 6 7 8 9 10 11 12 work, now possible with the development of
10 µm powerful molecular techniques and their applica-
tion to plant genetics, to improve our understanding
of the role of chromosomal variation in plant speci-
Figure 3.13 Schematic illustration of differences in the position of
secondary constrictions (cut in the chromosome), banding patterns for ation. Some of the above examples would provide
chromomycin A3 (), and the position of different ribosomal DNA loci, ideal model systems for such work.
that is, 18S () and 5S () on each of the 12 chromosomes in the
karyotype of three endemic Lilium species: (a) L. pyrenaicum,
(b) L. poponium, and (c) L. carniolicum. Reproduced with
permission from Siljak-Yakovlev et al. (2003). 3.6 Conclusions
In this chapter, I argue that by integrating species
and population-level approaches a more precise

understanding of the role of different microevo- natural selection in the wild, it is impossible to
lutionary processes, ecological factors, and histor- fully identify any adaptive function of such varia-
ical effects in the evolution of new species can be tion. So, despite the fact that Lewis (1966), nearly
reached. Evolutionary diversification of species-rich 40 years ago, and more recently Rieseberg (2001)
Mediterranean groups is a complicated problem due have argued for an important role of chromo-
to the complex history of the region and the diversity some rearrangements in plant speciation, this role
of the evolutionary processes involved. The work has yet to be confirmed within the Mediterranean
I have presented in this chapter illustrates repeat- flora.
edly how genetic discontinuities, within and among Several of the examples discussed in this chapter
species, have arisen on the template laid out by the illustrate the potential evolutionary significance
geological history of the Mediterranean Basin and of local speciation in geographically peripheral or
the contractions and expansions of species during ecologically marginal populations of widespread
climatic change and oscillation. A similar tale can be taxa. Instructive for our understanding of evolution,
told for other Mediterranean-climate regions such these examples also illustrate the immense value
as California (Calsbeek et al. 2003), where climatic of marginal populations of widespread species
aridity became marked at ∼3 Ma causing sharp eco- for the conservation of evolutionary potential.
logical gradients across the region, and South Africa The existence of such populations should thus
(Richardson et al. 2001). Geological history has be recognized in the elaboration of priorities for
thus left its mark not only on species distributions conservation. Somewhere at the limits to the distri-
(previous chapter) but also the differentiation and bution of widespread species, new taxa are evolving
evolution of new species. The patterns of diversifica- that will become the rare endemics of future con-
tion have however varied in time and thus differ for servation efforts. The case of one of the prettiest
different groups of species (Vargas 2003). A char- Cyclamen species I have seen, C. repandum and its
acteristic feature of Mediterranean plant endemism four allopatric subspecies, and its highly selfing
is that many genera show distinct centres of species derivative C. balearicum, is particularly enlightening
diversity in either the Iberian peninsula or in the here. Where some botanists may clamour for a range
Balkans, Greece, and Anatolia (Chapter 1). In addi- of disjunct species others may limit the distinction
tion to having been a refuge during periods of to two species, one with several subspecies. Under-
glacial maxima, such areas have also been impor- standing the precise nature of variation among and
tant sites for the diversification of many genera. within such taxa is central to our delimitation of taxa
Future climatic change will likely cause more shifts in such groups and to our understanding of their
in distribution and create more cases of endemism evolution. The example of Cyclamen is one which
for us to delight upon around the shores of repeats itself in many groups of Mediterranean
Mare Nostrum. plants. Yet there have been few studies which com-
An integral component of endemism and specia- pare levels of quantitative variation in order to assess
tion is chromosome evolution and hybridization. which traits may have been important in divergence,
That the genetic changes which accompany poly- and where. In the case of C. repandum, it would
ploidy and hybridization stimulate and provide appear that divergence of subspecies is ongoing,
the genetic capacity for ecological differentia- and that the variation which occurs within sub-
tion and adaptation is now clearly demonstrated species is that which ultimately causes differences
(Rieseberg et al. 2003) and the examples I discuss among taxa.
lend support to this idea. Other forms of chro- Finally, I have discussed examples which provide
mosome variation may also be common in the evidence that ecological factors have been critical
Mediterranean flora and thus play an important for differentiation and divergence and the main-
role in diversification. However, in the absence tenance of different taxa in geographic proximity.
of work linking chromosome rearrangements to Although differentiation in response to ecological

conditions is greatly facilitated by geographic isola- and seed movements) are critical to such diver-
tion, in some cases divergence may proceed in local gence. They are also fundamental to the develop-
populations, particularly small populations in geo- ment of patterns of within-species differentiation in
graphically peripheral and ecologically marginal relation to the spatial configuration of habitats in
sites. The interplay of factors which promote diver- Mediterranean mosaic landscapes, where ecological
gence (selection and drift) and those which homog- factors vary dramatically over short distances. This
enize gene frequencies in the landscape (pollen is the subject of Chapters 4 and 5.

Trait variation, adaptation, and

dispersal in the Mediterranean mosaic

Adaptation, insofar as such a concept existed, could no longer be considered a static condition, a
product of a creative past, and became instead a continuing dynamic process. Organisms are
doomed to extinction unless they change continuously in order to keep step with the constantly
changing physical and biotic environment. Such changes are ubiquitous, since climates change,
competitiors invade the area, predators become extinct, food sources fluctuate; indeed hardly any
component of the environment remains constant. When this was finally realized, adaptation became
a scientific problem.
E. Mayr (1982: 483–484)

4.1 Ecological constraints and The Mediterranean climate is characterized by

adaptation in the Mediterranean strong seasonality which involves (a) the associa-
tion of a drought period when temperatures are
Evolutionary theory takes as one of its simplest and at their hottest and (b) a cool (and cold in many
most fundamental premises that natural selection areas) moist period. The summer drought can limit
shapes patterns of trait variation in a way which growth, flowering, and fruiting, and is a major
maximizes the fitness of individuals in different cause of seedling mortality, as illustrated by studies
environments. Fitness can be maximized in different on Nerium oleander (J. Herrera 1991a), Helianthemum
environments in two ways. squamatum (Escudero et al. 1999), Lavandula latifo-
lia (C.M. Herrera 2000a), Olea europaea (Rey and
• Traits may show genetically based adaptive dif-
ferentiation, that is, different genotypes in distinct Alcántara 2000), Frangula alnus (Hampe and Arroyo
environments. 2002), Thymus vulgaris (my own data), and Rham-
nus ludovici-salvatoris (Traveset et al. 2003). Summer
• Phenotypic plasticity may enable individual
genotypes to alter their phenotype and thus adopt drought may thus impose a severe demographic
an adaptive phenotype in different environments. constraint. Cold temperatures in winter may also
limit growth and cause mortality in many regions
Distinguishing among these solutions to environ- of the Mediterranean. The most favourable periods
mental variation is a central theme in evolutionary for growth are thus the autumn and spring. But
ecology. In the Mediterranean region, microclimate even then, the unpredictability of rainfall and its
(particularly rainfall and drought stress), geology, occurrence in rapid bursts of intense rainfall can
soils and a long history of human activities have impose constraints on germination, establishment,
fashioned a mosaic of ecological conditions, often on and growth. Rainfall during the growing period
highly localized scales. Such localized spatial hetero- may vary enormously from one year to another,
geneity of ecological factors is ideal for the study of and this variation can have marked effects on
plasticity and adaptive differentiation in relation to species composition (e.g. Figueroa and Davy 1991),
levels of gene flow across the landscape. growth, and flowering. Variability in the timing


and amounts of rainfall can thus greatly impact may be of primary importance for the occurrence
population dynamics. of vegetation patterns on such slopes (Pigott and
Initial work on how plant traits represent an adap- Pigott 1993), with an additional limitation imposed
tive response to the Mediterranean climate was by the drying effect of some of the strong winds
mostly oriented towards the study of the adaptive which impose themselves in the Mediterranean
significance of sclerophyllous vegetation in the dif- region (e.g. see Barbero and Quézel 1975). Else-
ferent Mediterranean parts of the world (Box 4.1). where it has been reported that even long-lived tree
There are in fact two main strategies of coping with species on north- and south-facing slopes can show
drought stress: tolerance and avoidance. To toler- marked genetic differentiation (Linhart and Grant
ate drought requires functional attributes that allow 1996), hence the distribution patterns on ‘hubac’
for the maintenance of cell functions and plant per- and ‘adret’ expositions are not without consequence
sistence in the absence of the necessary moisture for for plant evolution. Several other examples of distri-
normal growth and development. To avoid drought, bution patterns of individual species (Thuiller et al.
plants may adopt a strategy of summer dormancy. 2003) and patterns of vegetation structure (Zavala
As I will discuss in this chapter, both the tolerance et al. 2000) which are determined at least in part by
and avoidance strategy are well illustrated in the aridity are known in the Mediterranean landscape.
Mediterranean flora. This co-occurrence of different In addition, human activities have greatly mod-
strategies is well known in the other ‘Mediterranean’ ified the selection pressures acting in local popula-
parts of the world. Take, for example, the plant com- tions and the spatial organization of natural habitats
munities of Southern California, where drought- in the Mediterranean landscape. A major feature
deciduous coastal sage communities and evergreen of such variation involves not just the fragmenta-
chaparral ‘occur intimately associated in an intri- tion and disturbance of natural habitats but also,
cate mosaic pattern throughout large areas of the following the abandonment of human activities,
landscape’ (Harrison et al. 1971: 868). the colonization of open areas (pastures, cultivated
An important point to make from the outset is fields, old mine sites) and the successional develop-
that moisture limitation can allow variation in other ment of forest vegetation. The variable stages of
environmental factors, such as diversity in topo- such succession are often all locally present, even
graphy, aspect, and nutrient availability, to exert an over very short distances. Population differentiation
important controlling effect on plant distribution in such landscapes will ultimately depend on rates
and population ecology. Indeed, mosaic vegetation and distances of dispersal across the landscape, any
in the Mediterranean is often linked to vari- spatial heterogeneity in dispersal patterns, whether
ation in soils, which are generally poor in nutrients selection pressures are strong enough to override
and highly variable in space. Two integral ingredi- the homogenizing effects on gene flow, and the
ents of this variation are depth and moisture reten- capacity of plants to establish populations and per-
tion capacity. Topography can also contribute to this sist in a heterogeneous landscape. Let us not forget,
mosaic, as the differential distribution of species if selection is strong enough, even in the presence of
and vegetation belts on south-facing and north- gene flow, adaptive variation may rapidly evolve.
facing slopes illustrates. For a well documented and My investigation of variation and adaptation of
illustrated description of such vegetation patterns, functional traits in Mediterranean plants in this
I refer the reader to Gamisan’s floristic inventory chapter is structured around three main themes.
of the island of Corsica (1991). The striking nature
of such vegetation differences on north- and south- 1. I explore variation in functional traits and pheno-
facing slopes in southern France is such that local logy in relation to the classical themes of aridity
languages evolved words (now used in the botan- and nutrient stress in Mediterranean communities.
ical and ecological literature) to describe the baked The contemporary climate is a recent phenomenon.
south-facing slopes (‘adret’) and cooler, shady north- The association of trait combinations in many
facing slopes (‘hubac’). Differences in water supply species no doubt evolved in the region prior to
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 111

the onset of a Mediterranean-climate regime, as the understanding of how contemporary trait variation
presence of a large group of Pre-Pliocene sclerophyl- and species distributions have been shaped by
lous, unisexual, fleshy-fruited species illustrates. historical climatic variation.
I would argue that the occurrence of sclerophylly The dramatic changes in vegetation across the
in the Mediterranean is an issue which has largely xeric Mediterranean–desert transition on the eastern
overshadowed other important aspects of plant vari- fringes of the Mediterranean illustrate the import-
ation and adaptation, which I will bring more to ance of constraints on plant growth associated with
the fore in this chapter as I examine evidence for increasing aridity (Kutiel et al. 1995). Since the begin-
variation in precise functional attributes. My stand- ning of the twentieth century, ecologists have been
point is that to observe plant evolution in relation fascinated by the similarities of vegetation structure
to climate requires analysis of trait variation within and function in the different Mediterranean-type
and among closely related species in relation to ecosystems of the world and the possibility that
ecological stress and environmental variation. such patterns represent independent convergence
2. I examine the constraints and selection pressures of traits as species adapt to the Mediterranean sum-
acting in plant populations during the coloniza- mer drought. Much of this interest has been directed
tion of open habitats following the abandonment towards discerning the adaptive nature of the ever-
of human activities and the successional establish- green and sclerophyllous habit of Mediterranean
ment of forest vegetation. I stress the roles of dis- shrubs and trees (Box 4.1).
persal limitation and spatial variation in dispersal In the Mediterranean, soil moisture becomes
patterns and regeneration capacity in relation to decreasingly available during the drought period,
habitat heterogeneity for the evolution of population even for deep-rooted sclerophyllous species, which
differentiation. do not show a physiologically latent phase during
3. Finally, I take up the theme of why there are the summer drought. Their vegetative activity is
so many aromatic plants in the Mediterranean fairly constant throughout the year with seasonal
flora. My focus here is on variation among closely peaks in photosynthetic activity associated with
related species and among populations of individ- bud-opening and shoot growth (Gratani et al. 1992).
ual species in their dominant secondary metabolites As the progressive drought sets in, stomatal control
and whether such variation is adaptive. is probably the most effective means of regulating
water loss due to transpiration (Mooney and Dunn
Compared to previous chapters, the focus has
1970a; Joffre et al. 1999). During late summer, water
now shifted from problems of diversity to problems
stress can cause cell turgor and water potential
of adaptation.
values to drop to extremely low values which may
cause major physiological dysfunction. Stomatal
4.2 Summer drought and nutrient closure may provide an effective control on such
stress: functional traits and their stress by allowing plants to increase water use
variability efficiency during summer drought (Cowan and
Farquhar 1977), although this may be less than pre-
4.2.1 Constraints and functional attributes
dicted if transpired water affects air humidity in and
Vegetation–climate relationships are one of the around the canopy (Tenhunen et al. 1990). Seasonal
foundations of plant biogeography and provide the changes in leaf physiological efficiency may also be
basis for the classification of Mediterranean forests of prime importance in controlling water loss during
and other vegetation of the region (Quézel and the summer drought (Tenhunen et al. 1990; Gratani
Médail 2003; Chapter 1). As our climate continues et al. 1992). Summer is also characterized by a high
to evolve, enormous efforts are being channelled air temperature, a vapour pressure deficit, and high
into attempts to predict how future climate change solar irradiance. Stomatal closure in response to
will affect the distributions of species and com- water stress will also cause reduced gas exchange
munities. To make realistic predictions requires an and thus reduce photosynthesis. Last but not least,

Box 4.1 Sclerophyllous vegetation in the different Mediterranean regions of the world

Sclerophyllous trees and shrubs are a characteristic drought stress, and are excluded from extremely
feature of Mediterranean vegetation. There has arid areas. Sclerophylly in Mediterranean plants
thus been much interest in the adaptive may thus be effective only if drought stress does
significance of sclerophylly and the possible not cause a reduction in transpiration which
evolutionary convergence of form and function in provokes irrevocable damage to xylem water
the five geographically disjunct conductance as a result of cavitation. If drought
Mediterranean-type climate regions of the stress is too intense or too long it may prevent
world. recovery of cavitated xylem, and there will be little
In the Mediterranean, the (winter) deciduous advantage to being sclerophyllous. In a
strategy has traditionally been thought to be comparison of drought resistance in two
disadvantageous due to a short sclerophyllous (Viburnum tinus and Ilex aquifolium)
photosynthentically active period which coincides and non-sclerophyllous (Hedera helix and
with summer drought. As Mooney and Dunn Sambucus nigra) species, Salleo et al. (1997)
(1970b) argued, the cost of maintaining evergreen showed that sclerophyllous species do not recover
leaves that can withstand drought stress and deter from water loss and stress more rapidly but do
herbivory is less than that of producing a recover more completely from xylem cavitation. In
completely new foliage each year. The combination their study of the mosaic distribution of coastal
of leaf traits associated with sclerophylly (i.e. a sage and chaparral vegetation in California,
relatively low photosynthetic capacity, a high Harrison et al. (1971) showed that although
proportion of stored carbon, a low leaf-nitrogen evergreen chaparral plants are less sensitive to
concentration, a low surface to volume ratio, thick long periods of summer drought by virtue of their
cell walls, and a thick rigid cuticle) may facilitate low rates of transpiration, their low photosynthetic
tolerance of negative turgor pressure under water activity during the favourable season puts them at
stress. Sclerophyllous species also show a much a disadvantage. Hence in extremely arid areas, a
higher degree of stomatal regulation than several summer-deciduous drought avoiding strategy, as
deciduous species in the Mediterranean (Duhme adopted in the coastal sage community, may be a
and Hinckley 1992). Finally, the nutrient acquisition more beneficial strategy by virtue of the higher
and allocation strategy of sclerophyllous species rates of assimilation of the species in this
may permit growth in a nutrient-poor soil and community during the favourable season for
their chemical defence may assure long leaf growth. In contrast, in areas which benefit from a
lifetime. Many authors have considered such traits, slight increase in soil moisture availability during
and the sclerophyllous evergreen habit, to summer, the xeromorphic morphology and greater
represent an adaptation to life in a Mediterranean water economy of the chaparral vegetation can
climate, as a couple of citations illustrate. allow them to maintain photosynthetic activity
during the dry period. The mosaic of chaparral and
• ‘Thus, the environmental conditions are such coastal sage communities in California, with the
that evergreenness is an appropriate strategy in
latter dominating the most xeric sites fits this idea
Mediterranean-type climates’ (Mooney and Dunn
and that of Kummerow (1973: 166) for whom
1970b: 298).
sclerophyllous species represent ‘moderate
• ‘The evergreen shrub seems to be the xeromorphic plants’.
morphological life-form best adapted to
In fact, despite broad similarities, and probable
Mediterranean stress conditions with seasonally
convergence of traits linked to overall form and
restricted growth’ (Seufert et al. 1995: 352).
function, there are important differences in
Drought resistance of sclerophyllous species may patterns of divergence in form and function of
arise from an ability to recover from damage to sclerophyllous species in different
xylem conductance rather than a faster recovery Mediterranean-climate regions (Naveh 1967;
from higher rates of water loss. Sclerophyllous Specht 1969a,b; Mooney and Dunn 1970b; Raven
species cannot however tolerate overly severe 1971; Parsons 1976; Cody and Mooney 1978;
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 113

Cowling and Campbell 1980; Shmida 1981; balance of evidence suggests that low nutritive
Milewski 1983; Cowling and Witkowski 1994). quality, high concentrations of secondary
Historical and contemporary differences in local metabolism, and tough, sclerophyllous foliage
climate and the importance of different fire interact to provide formidable barriers to herbivory
regimes and soil nutrient status are likely to be the in resource poor environments’.
most important causes of such differences (Joffre Finally, many sclerophyllous species (which have
and Rambal 2002). Indeed, a recent review by been the object of the majority of ecophysiological
Rambal (2001) illustrates that leaf photosythetic studies) evolved from a tropical and temperate
performance of Mediterranean woody species sclerophyllous vegetation present during the
does not generally differ from that of species in Tertiary, that is, prior to the onset of the seasonal
other biomes and that drought tolerance and Mediterranean climate. Such species did not
stomatal sensitivity are not consistently related to evolve sclerophylly in response to the onset of a
rooting depth, resource availability, or the degree Mediterranean-type climate. In a detailed analysis
of seasonal water stress. In fact, the maintenance of the character syndromes of the woody
of a canopy of drought tolerant sclerophyllous vegetation of southern Spain, C.M. Herrera
leaves through the summer may be more closely (1992a) revealed how character associations in
associated with nutrient stress rather than drought contemporary Mediterranean woody plants in
stress per se (Loveless 1961, 1962; Rundel 1988, Spain (notably the occurrence of sclerophylly in
1995; Specht and Rundel 1990; Joffre et al. 1999). association with unisexual, uncoloured flowers and
Overall, growth rates will be slow, except during large vertebrate-dispersed seeds and fruits) are not
marked seasonal flushes during periods of water observed in a dataset limited to those genera that
availability, and thus allocation to maintenance have only been present since the onset of a
costs and a longer leaf longevity could be Mediterranean-climatic regime. Verdú et al. (2003)
favoured. Indeed, levels of annual dry matter produced similar results for woody plants in the
production of evergreen and deciduous species Mediterranean parts of the Iberian peninsula,
can be equivalent in nutrient-poor sites, whereas California, and Chile. The similarities of vegetation
in nutrient-rich sites, deciduous species have a in different Mediterranean-climate ecosystems are
greater productivity, hence, some authors go so far thus strongly linked to the long-term persistence
as to consider sclerophylly as a simple of Tertiary lineages which evolved under a
epiphenomenon of phosphorous and other subtropical climate. The general conclusion is thus
nutrient deficiency (Loveless 1961, that although sclerophylly may provide certain
1962; Rundel 1988). advantages in a Mediterranean setting, it pre-dates
Having long-lived leaves with a low the Mediterranean climate and, at least in terms of
photosynthetic activity and high carbon its basic structural features, is not an evolved
concentrations may come at a cost if herbivory is adaptation to the highly seasonal Mediterranean
frequent. The low photosynthetic capacity means climate (Seddon 1974; Axelrod 1975; C.M. Herrera
that the replenishment of stored carbon reserves, 1992a; Box 1997; Joffre and Rambal 2002). The
lost as a result of herbivory, will be slow. Leaves similar traits one observes in the different regions
must therefore be defended against herbivory and have previously been interpreted in relation to
have a long lifetime in order for their construction convergence in association with the independent
costs to be recovered. The leathery and thick onset of Mediterranean climates in different parts
cuticle of sclerophyllous leaves, which are not of the world (Cody and Mooney 1978). Traits such
consistently more expensive to produce than as sclerophylly were present prior to the onset of
leaves of deciduous species (Merino et al. 1982), the Mediterranean-climate regime in the different
and which often contain various monoterpenes Mediterranean regions and thus do not represent
and tanins as significant proportions of fixed convergent evolution (Verdú et al. 2003). This is a
carbon, illustrate that traits associated with clear-cut illustration of how ecological patterns can
sclerophylly introduce benefits other than coping result from historical processes associated with the
with a summer drought. Indeed, Herms and dynamics of regional taxonomic assemblages and
Mattson (1992: 305–306) conclude that ‘the the persistence of historical lineages.

a reduction in transpiration can cause a modification of different physiological mechanisms may permit
of the energy balance of the leaf, which may result in the dissipation of such excess absorbed energy, for
overheating, a problem that can be exacerbated by example, decreased chlorophyll content (Kyparis-
high solar irradiance and temperature. If excessive, sis et al. 1995) and the production of volatile oils
such overheating can cause photo-inhibitory dam- (Section 4.5).
age. Several constraints can thus simultaneously A range of additional functional attributes
limit plant growth and development. may enable deciduous species to persist in a
Mediterranean plants show a diverse array of Mediterranean setting, alongside their evergreen
morphological and physiological attributes which relatives. Deciduous trees, neglected by many
help accommodate drought stress. Joffre et al. (1999) early evaluations of plant–water relations in the
have classified these according to the timescale of Mediterranean flora, are an important component of
drought stress. Mediterranean woodlands. Indeed, one of the major
climax oak species in the western Mediterranean,
1. Seasonal variation in water availability may be
Quercus pubescens, is deciduous. This species does
buffered by reduced leaf size. Low leaf surface
not avoid summer drought by early leaf senescence
area is common in Mediterranean shrubs, as the
as do some deciduous oaks in California (Griffin
plethora of Mediterranean Lamiaceae (e.g. Satureja,
1973). In Q. pubescens leaves appear in spring and
Thymus, and Rosmarinus) illustrate. At xeric sites in
begin to senesce just prior to the autumn rains.
California, species with small leaves contribute to
Individual trees thus renew their entire foliar tissue
∼80% of vegetation cover. In some Lamiaceae, indi-
and must maintain sufficient resources to survive
vidual plants may vary leaf size among seasons—
the drought period and then renew their leaves.
producing small leaves in summer and large leaves
Photosynthetic assimilation prior to the physio-
in winter (Margaris 1976). During a season, as the
logical stress triggered by drought stress is thus
water supply diminishes, plants may favour the
growth of deeply penetrating roots, and soil below
Measures of water potential made on Q. pubescens
∼2 m contributes an increasing fraction of total plant
during the hot and dry summer of 1994 in southern
water uptake. For example, in a study of Quercus coc-
France (when March–August rainfall was only one-
cifera Rambal (1984) showed that this contribution
third of the mean rainfall for that period in the
may be as much as ∼25% of total uptake in severe
study site), showed that water losses lead to a grad-
drought periods.
ual and severe decrease (similar to that shown by
2. Diurnal variation in water availability can be
the evergreen Quercus ilex at the same site) in pre-
countered by the regulation of stomatal activity.
dawn water potential (Damesin and Rambal 1995).
For instance, stomatal closure can allow for rapid
The similarity of low water potential values in the
reductions in rates of transpiration and thus a more
deciduous and evergreen species and the fact that
conservative water-use. Such stomatal closure may
they do not decrease below a certain level sug-
be progressive, for example, in Quercus species
gests that even the deciduous species regulates its
(Acherar and Rambal 1992) and N. oleander (Gollan
physiological activity in order to tolerate extreme
et al. 1985) or, as in Pinus halepensis, more dependent
aridity. In this species, conservative water use is
on a threshold water stress (Joffre et al. 1999).
promoted by progressive stomatal closure as water
Plants which do not escape the summer drought potential drops and by daytime reductions in net gas
via dormancy must also tolerate high solar irradi- exchange (measured as the stomatal conductance to
ance and thus cope with excess intercepted solar water vapour and net CO2 assimilation rate), pre-
radiation at the time when carbon assimilation is venting irreversible cell dehydration. In addition,
limited by stomatal closure and photosynthesis by the above authors found no evidence of photode-
water stress. Light energy may thus exceed that struction; inhibition of the photosynthetic apparatus
required for carbon fixation and cause damage or due to high solar irradiance and temperature was
dysfunction in the photosynthetic system. A range only temporary. Deciduous species can thus cope
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 115

with drought stress and have a lower area-based leaf In a study of leaf morphology and structure of
construction cost. Although Q. pubescens does not Arbutus unedo and Pistacia lentiscus at two sites on
have a higher photosynthetic capacity than Q. ilex Sardinia, Gravano et al. (2000) reported marked
(to offset a shorter period of photosynthetic activity), variation in structural leaf traits in relation to local
the proportion of nitrogen in fallen leaves relative climate. Although these authors did not detect
to mature leaves on the tree is only 44%, compared differences in overall morphology (leaf area, dry
to 78% in fallen leaves of Q. ilex at the same site weight, and weight/area ratio) they did find an
(Damesin et al. 1998). Hence the deciduous species increased tannin content in A. unedo and a thicker
may more efficiently remove nitrogen from senesc- leaf cuticle in P. lentiscus in an arid site compared to a
ing leaves. These studies thus illustrate the import- more mesic habitat. Both these traits could occasion
ance of studying a range of trait functions that may a greater resistance to water stress in the arid site.
be involved in coping with drought stress; no single In Dactylis glomerata, plants from populations in the
factor allows individuals to withstand the extreme Mediterranean region (relative to populations from
drought conditions that occur in some years. temperate Europe) show better turgor maintenance
The precise nature of plant adaptation to drought under low water potential which facilitates survival
stress may also vary among scleropyllous species. of the summer drought (Roy 1981) and an enhanced
For example, sclerophyllous Mediterranean plants ability to accumulate water-soluble carbohydrate
show marked variation in the thickness of the cuti- which permits rapid regrowth in the autumn
cle and the structure of their outer cell walls and their (Volaire 1995). Likewise, the leaf anatomy of Satureja
mechanisms of drought resistance (Kummerow horvatii populations suggests a xerophytic ecotype
1973). Comparisons of the physiological strat- in lowland Mediterranean habitats (280 m) when
egy of drought resistance in three species with compared with plants from the oro-Mediterranean
roughly equal leaf dry weight/surface area ratios mountain conditions (1540 m elevation) in the
by Lo Gullo and Salleo (1988) showed differences in Balkans (Todorovíc and Stevanovíc 1994).
the physiology of leaf–water relations. Olea europaea It is not only leaf morphology which can show
shows a ‘drought tolerant’ strategy (with a large variation among populations in relation to con-
diurnal osmotic stress in the warmest hours of trasting ecological conditions associated with
the day late in summer), Ceratonia siliqua contin- variation in climatic conditions. For example,
ues extracting enough water such that turgor and Narcissus triandrus, a widespread and locally
transpiration are little affected by drought (‘water- abundant geophyte in the central, north-western,
spending’ strategy), and Laurus nobilis had inter- and northern parts of the Iberian peninsula, shows
mediate values of leaf conductance and high leaf continuous geographic variation in stature and
relative water content (a more ‘drought avoidance’ floral morphology across its range (Barrett et al.
strategy) due to low losses of water and rapid 2004). In central Spain, plants bear 1–2 pale-lemon
recovery of leaf water content. flowers per inflorescence, have small narrow leaves,
and are generally small in stature. In contrast, in the
north-western part of the Iberian peninsula, that
4.2.2 Drought and nutrients: adaptation in
is, outside of the Mediterranean-climate region,
the mosaic
plants are wide-leaved, taller, and bear larger
Whether Mediterranean plants show intraspecific inflorescences with up to 10 creamy white flowers,
genetic variability in their capacity to withstand what Blanchard (1990: 112) calls ‘an impressive
and/or respond to the summer drought, particu- plant’. Flowers are significantly larger in size
larly in relation to the functional attributes discussed in the non-Mediterranean parts of the range of
above, has unfortunately been a neglected issue. this species. The overall change in phenotype
There are nevertheless some examples of population thus accompanies an ecological gradient from
variation in relation to climatic and soil variation in Mediterranean to non-Mediterranean conditions.
the Mediterranean region. It would be fascinating to test by manipulative

transplant experiments the extent to which such levels of differentiation in the deserts which fringe
patterns of variation, that is, the small-leaved and the Mediterranean probably stems from the fact
small-flowered plants of central Spain, represent that Mediterranean habitats are themselves highly
adaptive variation in Mediterranean conditions. heterogeneous.
In L. latifolia, a common herb in lowland Reproductive effort may vary in relation to
and mid-elevation open habitats in the western aridity. For example, several studies illustrate
Mediterranean, maternal plants differ both in their increased allocation of resources to reproduction
inherent capacity to produce seeds, which produce in more arid habitats. This trend has been reported
seedlings when sown into natural sites in southern in desert populations of Erucaria hispanica (Bras-
Spain (C.M. Herrera 2000a,b). This variation is pri- sicaceae) and two grasses Bromus fasciculatus and
marily related to a differential ability to tolerate Brachypodium distachyon relative to Mediterranean
and survive the summer drought. In addition there populations in Israel (Aronson et al. 1993; Boaz
exists an inverse relationship between seed produc- et al. 1994) and in Bromus erectus populations
tion and seed and seedling viability, hence, adap- from sites with dry, stony, fersialitic soils in sites
tion to drought will be strongly influenced by a with mild winters compared to populations on
trade-off between fecundity and viability. Increased deeper moister soils that are less stony in sites with
fitness associated with increased flower number colder winter temperatures (Ehlers and Thompson
will be compromised by lower viability of the off- 2004b). The higher proportional investment in
spring from plants with many flowers, probably due reproduction in more arid sites may result from a
to among-flower self-pollination and inbreeding trade-off with allocation to vegetative biomass and
depression (reduced viability of offspring produced competitive ability or a competition–colonization
by selfing compared to outcrossing). The precise trade-off if seedling survival is low in arid sites.
analysis of the different components of variation Whatever the cause, evidence from several species
in maternal fitness represents an essential element suggests that increasing aridity is associated with
in our understanding of fitness variation in natural the adoption of a life-history strategy which
plant populations. may be adaptive in unpredictable environments
To examine the idea that aridity may cause greater (see Box 4.2).
levels of population differentiation, as a result The soil environment may also be an important
of greater isolation among populations (Stebbins component of trait variation among plant popu-
1952), Comes and Abbott (1999b) compared the lations. In his classic work on the genetic dif-
spatial organization of genetic variation in popu- ferentiation of Anthoxanthum odoratum populations
lations of two closely related annual, outcrossing grown on the different soils of the Park Grass
Senecio species, Senecio vernalis and S. glaucus. In Experiment, Snaydon (1970) revealed a mosaic of
the eastern Mediterranean S. vernalis is common in A. odoratum populations locally adapted to soil
mesic habitats in the Mediterranean-climate zone heterogeneity. Differentiation among populations
of Israel whereas S. glaucus is primarily a species had evolved rapidly (in plots less than 40 years old)
of maritime sands and semi-desert environments. and over very short distances (plots were some-
Based on comparisons of genetic differentiation times <30 m apart). In the Mediterranean mosaic,
for nuclear and cytoplasmic genes, these authors substrate and soil conditions (Box 4.3) can vary
found no evidence for greater genetic differentia- markedly among and within sites, creating a mosaic
tion in arid habitats. Volis et al. (2002a,b) found of variable selection pressures over very local-
a similar lack of difference in the organization of ized scales. Such variation may have a strong
genetic variation at allozyme loci and for a range of effect on spatial and temporal patterns of seedling
phenotypic traits among desert and Mediterranean recruitment and thus the dynamics of natural plant
populations of wild barley (Hordeum spontaneum) populations.
in Israel. Hence, the hypothesis of greater levels A striking example of localized variation in
of population differentiation in arid environments female reproductive success has been documented
does not apply to these species. The lack of higher by Albert et al. (2001) for a narrow endemic species
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 117

Box 4.2 Temporary marshes: diversity and adaptation in relation to spatial and
temporary variation in ecological conditions

The alternation of a cool moist period and summer this trait variation fits the idea that spatio-temporal
drought is pushed to the extreme in wetland areas environmental gradients are strong enough to
which dry out in summer. Such temporary marshes counter local gene flow and produce striking
(called vernal pools in California) occur in localized patterns of within-population adaptive
depressions with no natural outlet, show striking variation.
variation in the relative length of flooding (which is Second, temporary pools are well known for
nevertheless long enough to allow the their species diversity, e.g. in California (Holland
development of hygromorphic (water saturated) and Jain 1981), Chile (Bliss et al. 1998), and the
soils and aquatic vegetation), and show strong Mediterranean Basin (Barbero et al. 1982; Grillas
gradients in environmental conditions both in and Roché 1997; Quézel 1998). Like the vernal
space (among marshes and from the centre to the pools of California (Zedler 1990; Bliss and Zedler
peripheral parts of a given pool) and time (among 1998), temporary marshes in the Mediterranean
years and during a given year in relation to the Basin are characterized by a large number of
alternation of flooding and drying). In the annual plants with a rapid life-cycle (Barbero et al.
Mediterranean Basin, temporary marshes also 1982) and a highly sensitive germination strategy
show immense variation in size, (Grillas and Roché which reduces the probability of germination at an
1997). Temporary marshes are not exclusive to unfavourable time and maintains a large seed
Mediterranean-climate regions and provide ideal bank (Bonis et al. 1995). This bet-hedging strategy
systems for the study of ecology and evolution in is typical of plants in a temporally variable
relation to spatio-temporal habitat variation. Three environment. These authors argue that this seed
main features of these habitats are relevant to the bank (which can reach tens to hundreds of
study of plant evolution. thousands of seeds per square metre in the
First, temporary marshes show striking spatial topsoil) has a ‘storage effect’ which magnifies the
variation in ecological conditions, both among and effect of favourable years relative to poor years. In
within marshes. In the temporary marshes of the conjunction with variation in reproductive
Mediterranean Basin, species composition varies strategies, this seed bank may contribute to the
spatially as a result of the interaction between persistence of a species-rich community. The
life-history traits and strong salinity gradients, both interaction between species traits and
in space and time, which are closely related to the environmental gradients in space and time can
date of flooding, its duration, and the depth of thus act to maintain species diversity in these
water (Grillas et al. 1991; Grillas and Roché 1997). communities, in a way similar to that proposed by
Position in a marsh is of particular importance lottery models (Chesson and Warner 1981).
here. Towards the periphery of the pool, drying is Finally, temporary marshes in several areas of
more frequent and severe and the environment the Mediterranean Basin are currently threatened
more variable. Hence plants may often germinate and in many areas in danger of being lost from the
and grow in shallower or drier conditions, and may landscape. Close to the coast, some are prime sites
experience a more variable environment and in for drainage to provide a rich soil for farming or
some cases less intraspecific but higher yet more space for tourism. Others are threatened
interspecific competition in the peripheral parts of by invasion of shrub species and their sustainability
a pool relative to the centre of a pool. In addition, will require the implementation of management
the granulometry of the sediment, temperature, techniques to counter such encroachment (Médail
irradiance, and salinity may all vary along a et al. 1998; Rhazi et al. 2004). A sound scientific
gradient from the centre to the periphery of a understanding of these marshes is thus important,
marsh. This spatial variation can create divergent not just for those of us interested in plant
selection pressures on plants of a given species at evolution, but also for the establishment of
different places in a pool (Linhart 1988). Most of conservation plans (Grillas and Roché 1997).

Box 4.3 Major types of soil in the Mediterranean region

Mediterranean soils are primarily of two main in arid and semi-arid bioclimates, rendizina with a
types (Lossaint 1973; Zohary 1973; Bottner 1982): dark surface horizon that sharply contrasts with
the parent rock (usually softer limestones, chalks,
1. Brown forest soils (fairly shallow clay-loams)
and marly limestones) and alluvial soils in the river
which develop under various woody vegetation
plains and serpentine soils in some sites.
around the Mediterranean and which differ from
Soils in the Mediterranean region are often thin
those in temperate regions by their frequent high
and stony with strong local affinities to the parent
concentrations of Ca2+ , Mg2+ , and Na+ ions in
rock, which undergoes relatively slow chemical
mineral–nutrient cycles, the persistence of
weathering (Paskoff 1973). They tend to occur in
carbonates in the soil and in which the seasonal
one of three nutritional states: (a) moderately
drought has a strong effect on the development of
leached with a well balanced (albeit fairly low)
humus and litter decomposition.
nutrient status (particularly nitrogen and
2. Iron-rich terra rossa which is derived from hard
phosphorous), (b) strongly leached and nutrient
or dolomitic limestone and is characteristic of
poor, and (c) calcium rich (high pH) with
garrigue and maquis vegetation particularly in
micronutrients in relatively high concentrations
subhumid and humid Mediterranean bioclimates
that can render phosphorous insoluble (Specht and
(see Chapter 1).
Moll 1983). All in all, soils are often oxidized and
In addition, one can observe carbonate-rich soils nutrient poor, a key feature being the paucity of
with a superficial crust that develop on limestone available nitrogen and phosphorous.

of Erodium which occurs in a single locality in the In a long-term field experiment on patterns
Lozoya Valley in central Spain. At this locality, of seedling recruitment in L. latifolia in the Sierra
plants occur on either rocky or lithosol microhab- de Cazorla in southern Spain, C.M. Herrera (2002)
itats on an isolated dolomitic outcrop in a land- showed how spatial patterns of the soil environment
scape dominated by extensive siliceous formations. affect both the probability of seedling emergence
The rock micro-habitat is characterized by chas- and the survival of seedlings over the six years of
mophytes and the lithosol micro-habitat by species his study. Although all the soils at the site originate
more typical of meadows and pastures. Plants are from calcareous or dolomitic limestone, they show
fairly evenly distributed in these two microhabi- marked spatial differences in texture, organic matter
tats which are themselves equally represented in the content, and nutrient concentrations over small
habitat island. In this mosaic of different communi- distances. Emergence was closely related to textural
ties, plant reproductive success is highly variable: features of the soil (associated with drainage), while
seed production varies among years in the lithosol survival depended more closely on soil fertility.
microhabitat (where seed production is very high The patchiness of seedling establishment (and
in favourable years) and is low but stable in rocky adult plant distribution) in this species may thus
microhabitats. As a result, in favourable years high be closely related to how the regeneration niche is
seed production can occur (in the lithosol habitat) differentially affected by spatial heterogeneity of
while in less favourable conditions persistence will the soil environment. Later stages of the life cycle,
be favoured (rock habitat). This species thus occurs for example, flowering probability and reproduc-
in a mosaic of microhabitats with different dynam- tive success, may also have different ecological
ics, another example of highly localized meta- correlates.
population function, which in this case is closely All these studies illustrate patterns of morpho-
related to local ecological conditions (see Chapter 2). logical and functional trait variation in relation
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 119

to contrasting ecological conditions associated selection in different sites, among which gene flow
with climatic and/or edaphic variation. More is restricted. An evaluation of the ecological factors
examples will come later in this chapter as I dis- creating such divergent selection was not, however,
cuss other features of spatial habitat variation in examined.
the Mediterranean mosaic. It should however be In studies of genetic variation and differentiation
realized that most of the examples concern patterns among natural populations of the wild relatives of
of variation which will require experimental val- two cultivated crops, wild barley (H. spontaneum)
idation. Manipulative studies using transplanted and wild emmer wheat (Triticum dicoccoides), Nevo
material among sites which quantify performance et al. (1983, 1988a) reported significant variation
variation in relation to environmental conditions in allozyme frequencies in relation to soil type. In
would be extremely valuable here. this area, brown basaltic soils produced during the
Another informative way of detecting patterns of Pleistocene and terra rossa soils on Middle Eocene
adaptive trait variation is to compare estimates of hard limestone vary over small distances (transects
population variation based on molecular markers were 100 m long). Similar patterns of population
(which quantify variation which may have evolved differentiation were detected in relation to highly
as a result of both random and/or adaptive causes) localized and geographic variation in the climate
with variation in quantitative traits (morphology, (Nevo et al. 1979, 1982, 1988b). Selection gradi-
phenology, and physiology) related to plant fitness, ents in the Mediterranean mosaic could thus be
that is, trait variation that is likely to result from acting on the allozyme markers. In particular, het-
the action of natural selection. If molecular mark- erozygosity was often positively correlated with
ers vary little compared to traits related to fitness increasing aridity and unpredictable rainfall, indica-
then one can suspect variation in selection gradi- tive that aridity stress represents a major selective
ents among sites to be an important cause of trait cause of genetic differentiation. Morphological vari-
variation among populations. A brief review of the ation among populations supports this idea (Nevo
literature shows that different species show very et al. 1984a,b). Although Volis et al. (2002a) found low
different patterns. levels of genetic differentiation between desert and
By comparing variation at polymorphic allozyme Mediterranean populations of wild barley (H. spon-
loci with variation in trunk morphology, growth and taneum), they detected significant differentiation in
survival for 19 native populations of the maritime phenotypic traits such as awn morphology, plant
pine, Pinus pinaster, mostly in the Iberian peninsula, size (but surprisingly not leaf size which was more
González-Martínez et al. (2002) found that quan- variable among populations within the two regions),
titative traits showed higher differentiation than and spikelet biomass. Reciprocal transplanting of
allozymes. Variation in the three quantitative traits seeds and seedlings among these two types of habi-
was closely associated with local patterns of pre- tat produced strong evidence for local adaptation
cipitation, temperature, and soil type, indicative of to the two types of environment (Volis et al. 2002b).
local adaptation to variation in these features of the As mentioned in other species above, plants from
environment. In particular the authors detected a Mediterranean habitats have a higher competitive
latitudinal cline in survival rates which could reflect ability (vegetative vigour) and lower reproductive
climatic effects on survival. Despite the long-lived effort than those from desert habitats.
nature of this tree, populations of maritime pine Within-site variation in climatic and soil condi-
show evidence for adaptation to highly diverse eco- tions can develop as a result of modifications to the
logical conditions, adaptation that has evolved since local environment by dominant species. Such effects
the last glaciation. In the annual legume, Medicago may alter the balance of competitive interactions
truncatula, Bonnin et al. (1996) reported that popu- among species and even facilitate the establishment
lations are more differentiated for morphological of later-successional species (Connell and Slatyer
traits than marker loci, suggesting that quantita- 1977; Bertness and Callaway 1994). For instance,
tive characters have undergone strong divergent individual trees are well known to modify the local

soil environment on which they and other species factors is not constant among species. Whereas in
may grow (e.g. Boettcher and Kalisz 1990), shad- some species temperature plays a critical role in
ing may impede the reproduction of gap species the onset of dormancy, for example, in Anemone
but provide a more favourable microclimate for coronaria (Ben-Hod et al. 1988), daylength, which
their regeneration, and roots may compete with may be further stimulated by high temperatures,
other species for nutrients and water. Concomi- is the primary cue in other species such as Poa
tantly, increased litter decomposition may create a bulbosa (Ofir and Kigel 1999). High concentra-
nutrient-rich environment, favourable to the growth tions of abscisic acid in leaf tissue accompany this
of associated species. This is important because in photoperiodic induction (Ofir and Kigel 1998).
the Mediterranean litter decomposition may be slow In the Mediterranean, autumn is an important
due to the coriaceous leaf structure of many species, period of photosynthetic activity and may thus be an
their high concentration of secondary compounds, appropriate moment to flower. Autumn flowering
and low moisture content of soils (Arianoutsou enables rapid deployment of flowers after the sum-
and Radea 2000). Two examples from the western mer drought when few other species are in flower
Mediterranean illustrate these processes and how and competition for pollinators may be reduced
species not only respond to environmental variation (Shmida and Dafni 1989) and there is probably less
but also create spatial heterogeneity for other species chance of pollen loss or stigma clogging due to
(Box 4.4). heterospecific pollen transfer. There are however
several constraints on flowering in the autumn. First,
flowering occurs in the rainy season. Contact with
4.3 The phenology of flowering and rain may inhibit pollen germination and reduce per-
fruiting formance, physically damage the flower, and dilute
nectar. Shmida and Dafni (1989) and Dafni (1996)
4.3.1 Summer dormancy
discuss how the several floral traits, such as timing
Many species survive the summer drought by avoid- of flower opening, pendulous flowers, pollen ger-
ing it. Such avoidance can be achieved by an annual mination inhibitors, and hydrophobic floral parts,
life history, very prevalent in the Mediterranean may be adaptive in such conditions. Second, flow-
flora and in the Mediterranean-climate region of ering may have to be rapid once rains begin. Finally,
California (Shmida 1981) or, in perennial plants, by pollinators are less abundant than in spring (Shmida
summer dormancy. This dormancy involves the die- and Dafni 1989).
back of above-ground parts and a resting period of Mediterranean geophytes provide an example
highly reduced physiological activity. of how to exploit the autumnal pollination niche.
A well-known form of summer dormancy in The subterranean storage organ not only provides
perennial plants is the geophyte life form, in which an effective dormant period during the summer
all the above-ground parts of the plant are shed drought, but also provides a supply of nutrients for
and the plant remains dormant with a perennating rapid growth at the beginning of the rainy season.
bud on a subterranean storage organ (tuber, bulb, Rapid development after the summer drought is
corm, etc.). Perennial geophytes are an emblematic important because the exact timing of the period
component of the Mediterranean flora (Shmida favourable for photosynthetic activity and growth
1981) not only in monocotyledons (Narcissus, is unpredictable and the length of this period vari-
Asphodelus, Iris, Muscari, Crocus, Tulipa, Ophrys, able, and quite short in some years. Once the
etc.) but also in some dicotyledons (Cyclamen is a geophyte bulb attains a given size, resource stock-
well-known example). age may exceed that required for flowering. This
Summer dormancy may be triggered by one or a accumulation of stored nutrients is likely to be very
combination of several environmental factors which important for flowering in autumnal flowering geo-
co-vary as summer develops: daylength, drought, phytes (see below). Although most Mediterranean
and temperature. The relative importance of these geophytes flower in spring, there is a secondary
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 121

Box 4.4 Local modification of ecological conditions by dominant species

The savannah-like dehesa landscape (photo kindly and/or Q. suber) managed for acorn,
supplied by R. Joffre) which covers ∼55,000 km2 cork, and cereal production (Joffre
of the southern Iberian peninsula has a low density et al. 1999).
(40/50 per hectare) of oak trees (mostly Q. ilex

Species number and vegetation structure differ in plants has a significantly higher organic matter
the open areas and under the canopy (Marañon content compared to soils <5 m distant but in
1986; Joffre et al. 1988): for example, the mean patches of grass, and associated grass species
number of species under the oak canopy show higher biomass on soil collected from under
(17 species/4 m2 ) is significantly less than that on thyme plants relative to soil from elsewhere in a
the canopy edge (32 species/4 m2 ) and in open given site (Ehlers and Thompson 2004b).
grassland >5 m distant (27 species/4 m2 ). These
floristic differences may arise because soils under
the canopy have a greater water-retention capacity
(Joffre and Rambal 1993) and are richer in
nutrients and organic matter due to litter
decomposition (factors which are likely to promote
the dominance of a small number of species)
and/or the shelter trees provide for domestic
livestock (Escudero 1985). In this
ancient agroforestry system, natural
selection pressures may vary sharply and
repeatedly over small distances; a fascinating
mosaic for the study of local adaptation.

The woody shrub Thymus vulgaris is associated

with a significant modification of soil properties in
garrigue habitats in southern France where this
species is a common and abundant feature of local
communities. Under thyme plants, soils have large
amounts of leaf litter due to the shedding of
leaves in late summer. Soil from under thyme

peak in autumn. In fact, perennial geophytes can Mediterranean species such as C. repandum, there is
comprise 80% of the autumn-flowering species in always a Cyclamen to be seen in flower during the
a given Mediterranean community (Shmida and cooler moister season of the Mediterranean climate.
Dafni 1989). Two phenological strategies are closely To see a species in flower during the summer
associated with this bimodal flowering phenology drought, you have to leave the Mediterranean and
(Dafni et al. 1981a,b). head to the Alps, where C. purpurascens completes
the annual cycle of flowering in the summer. This
1. In species with ‘synanthous’ leaves peak flow-
variation in flowering phenology among species,
ering occurs when the leaves are fully developed.
combined with the thrill of seeing a host of pink
The dormant period is followed by a period of
cyclamens in flower in the litter of a Mediterranean
leaf growth and resource acquisition and storage
forest, is enough to make any biologist want to study
followed by flowering and fruit production. Nearly
the evolution of floral traits in this genus.
all spring-flowering geophytes have this strategy.
The genus Cyclamen has ∼20 species, mostly
2. In species with ‘hysteranthous’ leaves flowering
distributed around the Mediterranean Sea and
is uncoupled from leafing and occurs primarily in
the Black Sea, plus one species in central Europe
the autumn or in winter. The first event after dor-
(C. purpurascens) and one in Somalia (C. somalense).
mancy is the appearance of flowers, and only once
Once a Cyclamen flower is fertilized, the single-
flowering passes its peak, do leaves appear.
flowered stem coils or bends downwards to place
Hysteranthy may have evolved from synanthy the capsules at soil level where they mature and
in autumnal flowering species in response to open for seeds to be dispersed by ants (Hildebrand
the unpredictability of rainfall and the need to 1898; Affre et al. 1995). Whereas flowers are easy to
flower rapidly after the onset of the first autumn see, sometimes in spectacular carpets on the forest
rains (Dafni et al. 1981a,b). This strategy may floor, fruits are extremely difficult to find in the wild.
be more prevalent in arid conditions, indicat- Hierarchial analyses of variation of flowering time
ing that flowering is in response to temper- show that flowering time in this genus is primarily
ature and/or daylength and not the onset of determined by subgenus membership (Box 4.5).
autumn rains. Although in most spring-flowering This result indicates that major phenological shifts
Mediterranean geophytes, flowering time occurs (between the two main peaks of autumn and spring
later with increasing altitude (e.g. Arroyo 1990a,b), flowering) occurred prior to the diversification of
the opposite is true for autumnal-flowering species. species in each subgenus. The divergence of ances-
For example, on Mt Hermon in Israel increasing tral taxa thus represented an important stage in
elevation is correlated with later flowering in spring- the diversification of phenological strategies in this
flowering Hyacinthus orientalis and earlier flowering genus. Since then, phenology has been relatively
in autumn-flowering Crocus olbanus (Dafni et al. conserved as species have diverged within different
1981b). On Corsica, two congeneric Cyclamen species lineages, except in the subgenus Gyrophoebe in the
(spring-flowering Cyclamen repandum and autumn- eastern Mediterranean. This phylogenetic compo-
flowering C. hederifolium) which co-occur over an nent to flowering phenology has been detected in
altitudinal gradient of ∼1,000 m, show a similar several other studies (e.g. Johnson 1992; Ollerton
pattern. and Lack 1992). In Cyclamen, the combination of
Cyclamen illustrates the different aspects of flower, leafing and flowering phenology shows a gradient
leaf, and fruiting phenology in Mediterranean geo- of variation from strictly hysteranthous species,
phytes. The first point to make here is that, whatever for example, C. africanum, C. hederifolium, and
the time of year there is always at least one species C. rohlfsianum, to strictly synanthous species, for
in flower. From the autumnal drifts of C. hederifolium example, C. persicum, C. coum, and C. repandum,
flowers in the low mountain forests across the while in some species leaves appear before the
northern shores of the Mediterranean, through the end of flowering (Debussche et al. 2004). After
winter appearance of C. coum and C. persicum in flowering, the long period of leaf production in
the east to the spring appearance of the most western autumnal flowering species may allow for a greater
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 123

accumulation of stored reserves which in turn 3. Mediterranean populations of the annual

may enable rapid autumn flowering prior to leaf composite Crepis sancta flower three weeks earlier
production (autumn-flowering species tend to have than populations from Northern France when
a much large maximum corm size). grown in a common environment (Imbert et al.
Hysteranthous species also occur in the geophytes 1999a). This difference is despite the fact that
of other Mediterranean-climate regions (Ornduff Mediterranean populations germinate later, after
1969; Dafni et al. 1981b; Johnson 1992). Hence, there the onset of autumnal rains.
is evidence for convergent evolution of this strategy 4. The innate dormancy of Senecio vulgaris seeds
in different Mediterranean-climate regions. Geo- from Mediterranean populations (an inhibitor in the
phytes which are active throughout the year tend to seed coat may prevent the synthesis or liberation of
occur in the tropics or (as in C. purpurascens) in non- gibberellins required for germination) over a wide
Mediterranean temperate climates which do not range of temperatures relative to seeds from British
experience a summer drought. This suggests that populations, which germinate (∼80%) immediately
the abundance of geophytes in a Mediterranean- at 20˚C, allows for a winter annual life cycle in a
type climate has been favoured by the onset of Mediterranean climate (Ren and Abbott 1991).
this climate since the Pliocene. As aridity becomes
more important, hysteranthous leaf phenology In the coastal shrublands of southern Spain, species
may have been selected for (Dafni et al. 1981a,b). which flower outside of the spring peak tend to
Unfortunately the genetic basis and degree of occur in more mesic sites (J. Herrera 1986, 1991a).
plasticity of this trait is unknown. Arroyo (1990a,b) reported a similar but less marked
pattern. The similarity of species phenology and
their patterns of interannual variability suggest
4.3.2 General patterns of phenology in the that climatic constraints are more important than
Mediterranean phylogenetic relationships in determining flow-
ering time (Petanidou et al. 1995). This may be
Asymmetric bimodality of flowering phenology
particularly important for high-mountain species,
with a sharp peak in spring and a secondary
for example, Hormathophylla spinosa in the Sierra
peak in the autumn is a conspicuous feature of
Nevada where the extremely short and dry growing
Mediterranean plant communities (Box 4.6; Dafni
period imposes a severe constraint on flowering
and O’Toole 1994). The summer drought thus
phenology (Gómez 1993).
imposes a severe constraint on phenology, and
As in Cyclamen, the constraint can also be of a phy-
flowering time, as illustrated by several studies of
logenetic nature. In various Tertiary endemics (see
intraspecific variation in the timing of germination,
Chapter 2) flowering phenology no doubt evolved
growth, and reproduction.
prior to the onset of the Mediterranean climate. For
1. In Quercus ithaburensis in the eastern example, the early summer flowering of N. oleander
Mediterranean, one can observe variation in leaf (J. Herrera 1991a) and the long flowering period
phenology; some trees are deciduous and others (November to June) of Cneorum tricoccon (Traveset
almost evergreen due to a very short duration of 1995) suggest a historical constraint on flowering
leaflessness (Ne’eman 1993). Both phenotypic plas- phenology which is a highly conserved trait in such
ticity and genetic variation appear to influence this species. The reproductive failure of Ruscus aculeatus
variation in phenology, which may be in a transition has also been attributed to a historical constraint,
stage from a deciduous to an evergreen strategy. this species persisting in a pollination environment
2. The phenology of annual grass species popula- very different from its original habitat in the Late
tions from California and the Mediterranean Basin Tertiary (Martínez-Pallé and Aronne 2000). In con-
have a phenology which corresponds with a longer trast, the flowering of the relict populations of
and earlier summer drought in California (Jackson F. alnus in southern Spain is significantly earlier than
and Roy 1986). in populations in temperate Europe (Hampe 2004).

Box 4.5 The phenology of Cyclamen species

Morphological and molecular phylogenetic Mediterranean-climatic regime (column numbers

analyses suggest that the genus Cyclamen can be refer to months). H—hysteranthous species
be subdivided into four subgenera. Debussche (flowering prior to leaf emergence), S—synanthous
et al. (2004) quantified the phenology of leafing, species (flowers and leaves at the same time).
flowering (grey bars), and fruit maturity (F) in 17 Figure drawn from data in Debussche et al. (2004),
species in controlled conditions under a photos kindly supplied by Max Debussche.

Subgenus/ species 7 8 9 10 11 12 1 2 3 4 5 6
Cyclamen F
purpurascens H F
africanum H F
hederifolium H

graecum H F
rohlfsianum H F
persicum S F

Gyrophoebe Hysteranthous
mirabile H F C. hederifolium
cilicium H
intaminatum S
cyprium S F
coum S F
libanoticum S F
trochopterantum S F
pseudibericum S F

balearicum S F
repandum S F Synanthous
creticum S F C. pseudibericum

All studied species are summer dormant except for subgenera, there is a striking lack of variation in
the non-Mediterranean Cyclamen purpurascens the timing of seed release. Even species such as
which occurs under a continental climate and the spring-flowering synanthous Cyclamen
continues growth in the summer. Caught between repandum and the autumn-flowering
the constraints of cold winters and summer hysteranthous Cyclamen hederifolium which grow
droughts, species in the the spring-flowering together in the wild show almost identical periods
subgenus Psilanthum have a very short period of of seed release. The length of seed maturation
photosynthetic activity, and thus a high capacity for thus varies from 2–3 months in C. repandum to
resource acquisition (i.e. high values of specific leaf 10 months in C. purpurascens and the seed release
area—a trait correlated with photosynthetic activity. period is independent of the flowering period and
Despite the strong seasonal differences in closely linked to the end of vegetation activity prior
flowering among species in different Cyclamen to the onset of summer drought.
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 125

Box 4.6 Flowering phenology in Mediterranean plant communities

(a) Total number of species which flower in a particular period, with the mean number per month in

Community and region Spring Summer Autumn Winter Reference

Greek phrygana 27 2 5 0 Petanidou and

(9) (1) (1.7) (0) Vokou (1990)
Mediterranean coastal 19 3 5 3 J. Herrera
scrub community (∼6) (1.5) (1.7) (<1) (1986, 1988)
Mediterranean 14 2 1 0 Bosch et al.
grassland (∼5) (1) (<1) (0) (1997)

(b) Number of species in flower during each month 80

Number of species in flower
over three years in the Greek phrygana. Drawn from 70
data for 133 species in Petanidou et al. (1995). 60
0 1 2 3 4 5 6 7 8 9 10 11 12

In southern Spain, the flowering period is very maritima in coastal scrubland in north-east Spain
short and flowers that develop in late spring lose (Picó 2000; Picó and Retana 2001).
a large proportion of ovules to desiccation. Contem- The spring peak of flowering produces a large
porary climate thus imposes a strong constraint on supply of flowers which in some situations may
reproductive phenology and fruiting in this taxon. create severe competition for local pollinators. The
In a study of the summer flowering shrub result is a large investment in reward (nectar
L. latifolia which flowers in mid-summer, that is, and pollen) and attractiveness (large floral dis-
when fewer other species are in flower, C.M. Herrera plays, colourful flowers) in spring-flowering species
(1992c) reported that although early flowering (Cohen and Shmida 1993). In contrast, outside of
plants had higher seed production than later the spring peak, there are fewer species in flower,
flowering plants, this result was not due to vari- but also fewer insects in flight, and it has been pre-
ation in water availability. This suggests that sum- dicted that one should observe lower investment in
mer flowering is associated with some form of reward (Shmida and Dafni 1989). A trend from high-
inherent constraint that may indirectly cause the rewarding spring flowers to less-rewarding summer
pattern of temporal variation in seed produc- flowers is seen in the Lamiaceae in Israel where
tion. In other species, extended flowering may there is a decline in flower size (Shmida and Dukas
allow reproductive failure in some years to be 1990) and a parallel reduction in nectar reward
compensated by reproduction during an extended (Petanidou et al. 2000) during the spring and early
period of flowering, for example, in Lobularia summer. Larger flowers and nectar reward in spring

may enhance competitiveness in terms of pollinator Hence the question: is phenology and morphology
attraction. Flowers may nevertheless remain showy adaptive in relation to vertebrate dispersers?
outside of the spring to attract the few pollinators Several studies have attempted to evaluate
that are present at this time, what has been termed whether variation in fruiting phenology and fruit
‘discovery advertisement’ (Shmida and Dafni 1989; morphology represent a response to selection by
Cohen and Shmida 1993). Long-distance attraction avian dispersers in Mediterranean fleshy-fruited
of potential pollinators may thus be at a premium plants. In a 12-year study of fruit dispersal in
in autumn-flowering species when pollinators and 12 species (9 genera) in the Sierra de Cazorla in
flowers are rare. Experimental analysis of these ideas south-east Spain, C.M. Herrera (1998) revealed a
and trends would be useful. striking independence of fruit production and bird
abundance over several years. This was despite a
more than 10-fold annual variation in fruit produc-
4.3.3 Variation in fleshy-fruit phenology and
tion by all the studied species and marked variation
in the composition of the fruit spectrum available
Seed dispersal by animals brings a mutual benefit: to dispersers. Except for a positive relationship
animal dispersers obtain food and plants dis- between Phillyrea latifolia and Sylvia atricapilla, which
perse their offspring, hopefully towards a suitable was primarily due to a greater abundance of birds
site for germination and establishment. In the in the two mast-fruiting years of this shrub, there
contemporary Mediterranean flora, vertebrate- was no correlated variation between fruit avail-
dispersed plants are abundant (C.M. Herrera 1995a) ability and either fruits in the diet of avian dis-
and, in contrast to temperate plant communities, persers or disperser abundance. Bird abundance
local groups tend to be taxonomically diverse was more closely related to November temperatures
at the family level. This is primarily because of than fruit availability. Variation in fruit produc-
the persistence of 1–2 species in several genera tion most probably arises as a result of resource
and families in the tropical element of the con- availability exceeding a sufficient level for their
temporary Mediterranean flora, for example, allocation to reproduction and high levels of fruit
Jasminum, Phillyrea, Olea, and Osyris to name but production. The study by C.M. Herrera (1998) and
a few (Fig. 4.1). previous work by Jordano (1987) indicate that the
The abundance and nutritional characteristics primary limiting factor here is likely to be water
of fleshy-fruited species makes them an import- availability. The absence of a consistent relationship
ant element of biodiversity functional relation- between fruit traits and frugivory and any signifi-
ships in the Mediterranean (C.M. Herrera 1984a,b; cant variation among populations in the interaction
Debussche 1988; Rey 1995). Many fleshy-fruited between frugivory, plant traits, and reproductive
species in the Mediterranean have peak fruit success has also been reported for Juniperus commu-
maturity when avian dispersers are most abundant nis in the Mediterranean mountains of south-east
(C.M. Herrera 1984a,b; Debussche and Isenmann Spain (García et al. 2001). In contrast, in ant-
1992); but whether this is due to selection by dispersed species it has been reported that habitat
dispersal agents and not just a consequence of type, by its effect on disperser communities, may
resource limitation imposed by the dry summer has cause plant–disperser interactions to adopt distinct
been questioned (Herrera 1985). Although rates of evolutionary trajectories in different populations
fruit removal are often very high, they may vary (Garrido et al. 2002).
among and within species (Jordano 1987; Thébaud Another limitation on adaptive variation in
and Debussche 1992; C.M. Herrera 1995c) and fruit traits stems from the finding that, with just a
many species show variation in size and shape few exceptions (C.M. Herrera 1984b), fleshy-fruited
of fruits which may be differentially consumed plant species in the Mediterranean tend to have their
by frugivores, for example, Prunus mahaleb seeds dispersed by either a diverse array of bird
(Jordano 1995) and O. europaea (Rey et al. 1997). species (C.M. Herrera 1984a,b; Jordano 1987, 1989;
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 127

(a) (b)

(c) (d)

(e) (f)

Figure 4.1 Examples of fleshy-fruited species present (or with ancestral species present) in the Mediterranean region prior to the onset of the
Mediterranean climate: (a) Phillyrea latifolia, (b) Olea europaea, (c) Osyris quadripartita, (d) Jasminum fruticans, (e) Rhamnus
alaternus, and (f) Pistacia lentiscus. Photos kindly supplied by C.M. Herrera ((a)–(c)) and M. Debussche and B.O. Krüsi ((e)–(f)).

Debussche and Isenmann 1989; Guitián Fuentes and Other factors may further contribute to reduce
Sanchez 1992) or combinations of birds with either the selective impact of dispersal agents on fruit
rodents, for example, Juniperis communis (García traits (Box 4.7). First, although variation in traits
et al. 2001), or ants, for example, Rhamnus alaternus may influence rates of dispersal, the effect on final
(Aronne and Wilcock 1994). Spatio-temporal hetero- reproductive output, and fitness, may be minimal,
geneity in the composition and relative abundance as a result of the complexity of interactions that may
of dispersal agents across the distribution of a given occur and the overwhelming importance of total
plant species may limit the opportunity for adap- crop size. Second, fruit number may be significantly
tive evolution in fruit traits (C.M. Herrera 1995c). reduced by predispersal abortion and predation of

fruits (Jordano 1989) which reduce the contribution not be adaptive with respect to contemporary dis-
of variation in crop size to reproductive success. persal agents and the Mediterranean climate. Never-
In the hemiparasitic dioecious woody shrub theless, since the Pliocene, fleshy-fruited genera
Osyris quadripartita (Fig. 4.1), common in warm have gone extinct at a lower rate than non-fleshy-
shrublands in the western Mediterranean, a rather fruited genera, suggesting that their high frequency
surprising temporal pattern of fruit production can in the flora is not just due to phylogenetic inertia.
be observed (C.M. Herrera 1984c, 1988a). In popula- The tropical element also has a majority of species
tions of this species in southern Spain, females pro- with small often unisexual flowers and fleshy fruits
duce ripe fruits throughout the year, despite the fact (C.M. Herrera 1982a; Aronne and Wilcock 1994;
that individual plants only flower once a year. Due Quézel and Médail 2003), a syndrome which is cer-
to the gradual ripening of fruits from a single flow- tainly not limited to the Mediterranean. Although
ering bout, fruits resulting from the previous year’s it has been predicted that such dioecious species
flowering ripen as new fruits develop. The only sea- should show greater absolute maternal investment
sonality is in the fraction of plants ripening fruits in seed dispersal compared to hermaphrodites
which peaks in late-autumn/early winter. In Osyris, (Givnish 1980), C.M. Herrera (1982a) found no evid-
the persistence of a continuous phenological strat- ence of this in a study of dispersal-related maternal
egy through the summer drought, may have been reproductive investment in 73 species in southern
facilitated by the ‘generalist’ hemiparasitic nature of Spain. Three points are important here: (a) selection
this species, which permits additional nutrient and pressures due to dispersal agents may be complex
water uptake (C.M. Herrera 1984c, 1988b). and highly variable in space and time; (b) the sig-
This long period of growth, flowering, and fruit- nificance of crop size for fitness is limited since
ing, with extensive temporal overlap in such activi- many seeds are dispersed into unfavourable habi-
ties, is typical of species in a tropical climate and is tats; (c) many dioecious shrubs in the Mediterranean
not related to potential selection pressures imposed have probably shown little modification over a
by contemporary dispersal agents. As I discussed long period of time. This association of charac-
in Chapter 1, many fleshy-fruited woody species ters is thus primarily due to phylogenetic con-
(e.g. species in the Oleaceae, Anacardiaceae, Santa- straints and sorting processes, having evolved
laceae, Rhamnaceae, Myrtaceae, Cneoraceae) in the long before the Mediterranean-climatic regime set
Mediterranean flora, originated prior to the onset of in (Box 4.1).
a Mediterranean seasonal climate and form a fairly In the tropical relict, Jasminum fruticans (Fig. 4.1;
distinct Pre-Pliocene group. Species in these groups Plate 2), fleshy fruits are also ripened gradually such
were thus present when conditions were subtropical that many are mature only in late winter (Puech
in this region. Of woody shrub and tree genera now 1986). Late ripening in such species is associated
present in southern Spain that were present prior with a lower pulp biomass and smaller fruits (Puech
to the Pliocene 94–95% have fleshy fruits, whereas 1986; C.M. Herrera 1988a). Species which have a
only 5–6% of genera that have immigrated since the ripening period which occurs in late autumn and
onset of a Mediterranean-climate regime have fleshy early winter also have different fruit characteris-
fruits (C.M. Herrera 1992a; 1995a). Although, many tics (noticeably a decrease in water and an increase
species disappeared from the region as the climate in lipid content in late-ripening fruits) compared
evolved, several have persisted. In the Late Tertiary, to those which ripen their fruits in summer and
vegetation extinctions would have been paralleled early autumn (C.M. Herrera 1982b; Debussche et al.
by the extinction of animals, particularly large birds 1987). Such variation is unrelated to variation in
capable of dispersing fruits which were present prior avian dispersers. The co-occurrence of peak fruit
to the Pliocene onset of the Mediterranean climate availability and the abundance of avian dispersers
(Mourer-Chauviré 1989). Hence many plant species is more the result of climatic constraints, pest pres-
may have persisted following the loss of their seed sure, and opportunistic behaviour of frugivores
dispersal agents, hence their dispersal mode may (Debussche and Isenmann 1992; C.M. Herrera 1995c)
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 129

than the result of selective pressures exerted by suggest that a similar conclusion can be drawn for
dispersal agents. Hence, it would appear that dis- variation in the size and shape of fleshy fruits in
persal agents impose only weak selective pressures the Mediterranean (Box 4.7). There is thus a fairly
on the fruiting phenology of fleshy-fruited shrubs large body of evidence which attests to a lack of
in the Mediterranean region, due to the complex- direct interaction and little evidence of any adaptive
ity of the interaction between single plant species variation in fruiting phenology and morphology in
and their multiple dispersal agents. Several studies relation to avian dispersal agents. However, as I

Box 4.7 Fleshy fruits: morphology and vertebrate dispersal

At the present time there is a paucity of evidence relationships). Instead, allometry and genus- and
of adaptive variation in fruit morphology in species-specific variation in fruit shape are more
relation to avian dispersal agents in Mediterranean important than dispersal agent in explaining the
fleshy-fruited species. patterns of variation. Nested analyses of variance
First, C.M. Herrera (1988a) reported significant showed a significant phylogenetic component to
variation among individuals of Osyris quadripartita variation in fruit size and shape. Diversification in
in fruit size and production over four years of fruit morphology occurred at a level above the
study. However, the patterns of variation were species level and closely related species share
highly variable over a single year and among years similar traits. The combined effects of allometry
due to differences in rainfall events in different and taxonomic relationships rule out adaptive
years (once again the unpredictability of rainfall explanations of variation in such traits.
events is important). Although fruit characteristics Third, any potential direct effect of fugivorous
influence dispersal rate, they only have a minor birds on plant traits that enhance fruit removal
contribution (0.2%) to realized reproductive may be masked or buffered due to the impact of
output in a population, due to the overriding other ecological parameters. One of the most
contribution of total fruit production. For these important of these is herbivory and pre-dispersal
two reasons, variation in fruit morphology does seed predation. In a study of the effects of such
not translate into fitness differences and thus may factors on fruit production and removal in Olea
have little selective value. This primary importance europaea in southern Spain, Jordano (1987)
of absolute numbers of fruits produced has been reported that although individual fruit yield was
documented in other Mediterranean herbs and positively correlated with fruit removal by
shrubs, for example, Lavandula latifolia dispersers, plants with high yield were massively
(C.M. Herrera 1991) and Prunus mahaleb (Jordano infested by Dacus oleae (a Tephritide fly), which
and Schupp 2000). More fecund plants may thus reduced the contribution of trees to total fruits
disperse more propagules despite variation in fruit removed in a population.
size or relative rates of seed set. Nevertheless, in species whose distribution
Second, in comparative analyses (correcting for extends beyond the bounds of the Mediterranean
statistical non-independence associated with fruit traits have been reported to vary between the
phylogenetic relationships) of 117 fleshy-fruited Mediterranean and temperate regions, perhaps in
species (from 35 families) in the Iberian peninsula, relation to the local climate and variation in
C.M. Herrera (1992b) found that fruit shape disperser presence and activity (Hampe and
variation in Mediterranean vertebrate-dispersed Bairlein 2000; Hampe 2003). Hence, it is possible
plants does not support the prediction that groups that colonization and/or persistence in the
of plants with dispersal agents of contrasting sizes Mediterranean has to some degree impacted on
and fruit handling capabilities have differently particular dispersal-related traits in some
shaped fruits (in terms of length–width species.

will discuss later, dispersal patterns show marked in many areas been quite simply abandoned. The
spatial patterns in the Mediterranean mosaic, with result was, and continues to be, rural depopulation.
potential consequences for the evolution of genetic For example, in the Languedoc-Roussillon region
differentiation among populations. of southern France the number of inhabitants in
rural areas decreased 2–3 fold over the period
1936–90 (Cheylan 1990). Towards the coast, human
4.4 Dispersal and establishment: the populations remained stable during this period,
template of local differentiation and since the 1960s have begun to grow, at rates
faster than anywhere else in France. A similar story
4.4.1 Landscape change: process and
occurs in many areas of the Mediterranean. On
Crete, major landscape changes in the second half of
Landscapes are changing around the Mediterra- the twentieth century involve tourist development
nean. Although vegetation patterns and species along the north coast (where human populations
distributions have frequently changed in relation have remained stable) and forest spread and closure
to climatic variations, even since the initial onset of previously open woodland in inland areas, where
of a Mediterranean-type climate in the Pliocene the mountain village populations have declined by
(see Chapter 1), contemporary rates of change more than one-third (Arianoutsou 2001). Hand in
are probably unprecedented in Mediterranean hand with these changes in human populations,
history. Instead of gradual climatic oscillations, the landscape has changed as woody plants have
Mediterranean vegetation is now faced with the spread to form woodland and forest on previously
rapid, extensive, and often brutal effects of human open landscapes. This forest spread is extremely
activities. Near the sea, wetlands and semi- rapid in some areas (Fig. 4.3).
natural coastal and lowland habitats are rapidly On the northern shores of the Mediterranean
being destroyed and fragmented by urbanization there is a distinct difference in the degree of human
and tourist resort development, causing major impacts on natural ecosystems which depends on
loss of biodiversity and creating severe problems location: whereas human impacts are increasing
for the viability of natural populations. Inland, near the coast, further inland, particularly in low
a different story can be told. Throughout the mountain areas, they have actually diminished.
European backcountry of the Mediterranean region, In this section, I will focus on the ecological pro-
forests are spreading due to agricultural decline cesses and consequences of natural forest spread.
and rural depopulation (Fig. 4.2). Some famous This discussion is motivated by the question: what
Mediterranean landscapes are changing, as woody are the implications of such landscape change for
species colonize the rolling hills of Tuscany, the lime- plant population ecology and differentiation?
stone grasslands near Roquefort, and the mountains The principal land-use changes associated with
of Crete. The contours of the Mediterranean habitat modified human impact and rural depopulation
mosaic are thus changing rapidly. include the abandonment and dramatic decline of
The reforestation of many inland areas around the terrace agriculture on hillsides, a shift towards
northern rim of the Mediterranean is a natural eco- intensive smaller-scale farming with a decline in
logical process, but one that has been set in motion extensive grazing practices over large expanses of
by important socioeconomic changes that began in lowland and upland plains and reduced exploitation
the late nineteenth and early twentieth centuries of woodlands for glass-making, charcoal, and even
(Lepart and Debussche 1992). These changes caused firewood collection (Lepart and Debussche 1992).
traditional agricultural practices, pastoralism, and Striking examples of the rapidity of forest spread
forest activities to become economically unattrac- associated with these changes can be seen in the
tive and in many areas unprofitable (e.g. Marty Mediterranean backcountry (Figs. 4.2 and 4.3),
et al. 2003). Too much labour and not enough profit, for example, in the Cévennes (Debussche et al.
traditional farming has modified its structure and 1999) and Maritime Alps (Barbero et al. 1990) in
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 131




Figure 4.2 Some examples of landscapes currently experiencing natural spread of woody vegetation and forest closure in the Mediterranean:
(a) the spread and establishment of Pinus brutia woodland on Crete, (b) the spread of deciduous oak (Quercus pubescens) on the Causse du
Larzac (southern France), (c) the spread of Buxus sempervirens on the Causse du Larzac in southern France (photo kindly supplied by P. Marty).



Figure 4.3 The rapidity of natural spread of woody vegetation (mostly evergreen oaks) and forest closure in the Basses Cévennes area of southern
France: (a) a postcard view in 1929 and (b) the same view in 1992 (reproduced with permission from Lepart et al. 1996).

southern France and in mainland Greece and Crete which covered ∼36,000 ha of Mediterranean France
(Arianoutsou 2001; Grove and Rackham 2001). In at the end of the nineteenth century, now occupies
the Languedoc-Roussillon region of southern France more than 200,000 ha of this region (Barbero and
mentioned above, forest cover has more than dou- Quézel 1990). The Mediterranean mosaic is thus sub-
bled since the beginning of the twentieth century, ject to important changes: woodland interspersed
increasing from 400,000 ha to ∼975,000 ha (Agence with patchy open habitats and localized inten-
Méditerranéenne de l’Environnement 2000). An sive cultivation is becoming a typical landscape in
important element of this reforestation involves the many areas.
return of evergreen oak woodland and the spread Forest spread may have significant ecological and
of pioneer pine trees. For example, P. halepensis, evolutionary consequences. First, local ecological
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 133

conditions are modified, creating new constraints of factors affecting dispersal and establishment dur-
and selection pressures for plants which arrive at a ing the colonization process can create sharp spatial
site. Second, the spatial configuration of habitats in patterns in natural populations.
the landscape has been dramatically modified. This
means that dispersal of seeds and pollen, and thus
4.4.3 Spatial patterns of dispersal across
gene flow, among sites are subject to marked mod-
open habitats
ifications. Populations of woodland plants isolated
from one another for several centuries may now be The rate and spatial extent of establishment of
increasing in size and may come back into close con- vertebrate-dispersed woody plants in open land-
tact with possibilities for crosses among previously scapes in the Mediterranean, like elsewhere, are
isolated populations. In contrast, populations of closely dependent on the spatial distribution of
species that occur in open habitats or gaps in the seed sources. Studies of the several fleshy-fruited
forest may decrease in size and become more iso- and wind-dispersed woody plants shows that most
lated from another if they cannot survive shading dispersal occurs close to maternal trees, that is,
and forest spread. The implications for genetic over a few metres or tens of metres, for example,
differentiation in these two groups of species are Buxus sempervirens and Fraxinus angustifolia in
thus very different. Mediterranean old-fields (Debussche and Lepart
1992), mast-fruiting vertebrate-dispersed P. latifolia
4.4.2 Dispersal and establishment: colonization in mid-elevation scrubland and forest in southern
processes and population differentiation Spain (C.M. Herrera et al. 1994), wind-dispersed
P. halepensis on the lower slopes of Mt Carmel in
Colonization and population establishment depend Israel (Nathan et al. 2000), several fleshy-fruited
on rates of arrival at a site (and thus dispersal-related pioneer plants in early-succession old-fields
traits and rare episodes of long-distance dispersal) (Debussche et al. 1985; Debussche and Isenmann
and the establishment of a local population from the 1994), and O. europaea subsp. europaea var. sylvestris
initial colonists (i.e. the regeneration of seedlings in scrubland and grazed sites in southern Spain
in a novel environment). Because plant mortality (Alcántara et al. 2000). In several of these studies
tends to be high at the seedling stage in many plant the distribution pattern of seeds in the seed rain
populations (see references above) there should be was very similar (particularly in open scrubland) to
strong selection on dispersal-related traits that con- the spatial pattern of seedling establishment. The
sistently favour the chances of putting a seed into a spatial aggregation of seedfall around maternal
site favourable for seed germination, seedling estab- plants and the coupled pattern of seed rain and
lishment, and seedling survival. Recognized as a established seedlings provide strong evidence that
key element in the maintenance of diversity in plant dispersal limitation will create spatial pattern,
communities (Grubb 1977), the regeneration niche with a potentially ‘lasting impact on … population
is a key element in the population ecology of indi- dynamics" (C.M. Herrera et al. 1994: 315).
vidual species. In addition, in recent years there An important component of this spatial pattern
has been increased recognition of the important role involves nucleation around initial colonists. For
played by dispersal limitation in the patterns of many Mediterranean fleshy-fruited species, the
abundance and distribution of individual species presence of other tree and shrub species in a more
and community structure (Hubbell 2001). Evidence or less open landscape promote nucleation as birds
from studies of different Mediterranean systems use conspecific trees and other species as perching
illustrate the combined role of spatial patterns of dis- points (Debussche et al. 1985; Debussche and Lepart
persal and its restricted nature and ecological factors 1992; Debussche and Isenmann 1994; C.M. Herrera
influencing regeneration for the dynamics of natural et al. 1994; Verdú and Garciá-Fayos 1996; Alcántara
populations in the Mediterranean mosaic landscape. et al. 1997, 2000; Garciá et al. 2000; Traveset et al.
The point that is important here is that the interplay 2003). For example, Debussche et al. (1985) showed

that although 15% of the seeds of 14 fleshy-fruited vegetation structure can favour dispersal over
species in Mediterranean old-fields are dropped longer distances: the establishment of seedlings
within 20 m of a seed bearer, noticeable peaks in >100 m from any potential source illustrating this.
the seed shadow occur around perching places In a study of bird-dispersed P. mahaleb in southern
(isolated trees). In wind- and animal-dispersed Spain, Jordano and Schupp (2000) detected a highly
species, dispersal limitation is thus not completely heterogeneous spatial pattern of seed dispersal.
random in the landscape. A pattern of nucle- In this species the seed shadow was determined
ation can also be observed in the colonization by a complex interaction between the foraging
of open grassland by wind-dispersed species, patterns of a suite of frugivores that differ in their
for example, Pinus sylvestris which is actively habitat preferences and the precise distribution of
colonizing peri-Mediterranean grasslands (Box 4.8). different habitat patches in the landscape. Covered
Dispersal patterns can be closely related to micro- patches received seed dispersed by 7–11 bird
habitat variation in the Mediterranean mosaic, species whereas seed dispersal into open habitats
particularly in vertebrate-dispersed plants. The was by only 1–2 species. As Jordano and Schupp
first point to note here is that the mechanical and (2000) suggest, one would predict greater genetic
chemical action of the dispersal agent has little variability among P. mahaleb propagules within the
importance for overall germination capacity in covered habitat patches. In addition, reduced and
many Mediterranean species (Debussche 1985, spatially aggregated seedfall patterns in more open
1988; Izhaki and Safriel 1990). Frugivory may habitats may favour differentiation among patches
however promote dispersal into favourable sites (each from potentially different seed sources but
for germination. Alcántara et al. (2000) integrated with little variation within patches).
the patterns of seed fall both in the seed shadow
(under trees) and the seed rain (population level) to
examine the relative contribution of spatial, micro- 4.4.4 Spatial heterogeneity and regeneration
habitat, and tree characteristics for seed dispersal
Contemporary reforestation of many areas is an eco-
from wild olive trees (O. europaea subsp. europaea
logical change involving primarily native species
var. sylvestris) at two study sites in the Sierra Sur de
and is thus a natural example of secondary
Jaén in southern Spain. They found that differences
succession. Vegetation succession has fascinated
among frugivores in sizes of seeds dispersed and
ecologists ever since the science of ecology shifted
foraging movements created a pattern of seed
its emphasis from describing patterns to under-
dispersal in which different microhabitats received
standing their causes. In their classic paper which
different distributions of seed sizes. The spatial
focused attention on the mechanisms which allow
distribution of seeds was found to depend on a
and favour the transition from pioneer communities
complex interaction between the fruit traits which
to climax forest, Connell and Slatyer (1977) outlined
attract frugivores, distance to the source tree, and
three models of vegetation succession:
microhabitat. The latter was associated with high
seed densities in patches of shrub species with • Facilitation: pioneer species modify the environ-
dense foliage (Phillyrea latifolia, Pistacia lentiscus, ment in a way which makes ecological conditions
and Q. coccifera), low seed densities under species suitable for the establishment of later successional
with open foliage (Pistacia terebinthus and Rhamnus species. Without such modifications the latter can-
lycioides) and low seed dispersal into open spaces. not establish at a site
A low rate of seed dispersal by frugivores into • Tolerance: late successional species establish
open spaces is commonly observed in fleshy-fruited independently of pioneer species
species in Mediterranean habitats (C.M. Herrera • Inhibition: establishment of late successional
and Jordano 1981; Izhaki et al. 1991; Debussche species only occurs as a result of disturbance
and Isenmann 1994; C.M. Herrera et al. 1994). or mortality of pioneer and early succession
Debussche et al. (1985) showed that a complex species.
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 135

Box 4.8 Dispersal, nucleation, and colonization of limestone grasslands by

Pinus sylvestris in southern France

An example of the role of nucleation for rapidly increase the speed of colonization in this
the colonization of open areas by wind-dispersed landscape due to the creation of new focal points
species can be seen in pine trees colonizing upland (or ‘nascent foci’) of colonization, distant from
limestone grasslands in southern France. In this the primary source populations, which become
area, the spread of P. sylvestris since the beginning secondary source populations for subsequent
of the twentieth century (when cereal cultivation spread (see also Moody and Mack 1988; Hengeveld
began to decline and milk production became 1989; Cain et al. 2000). The expansion of species
based on intensive, specialized methods) illustrates such as the pine trees colonizing the Causse
dispersal mediated patterns of spatial aggregation Méjean, is also favoured by their reproductive and
(Debain 2003; Debain et al. 2003a,b). Dispersal seed biology, that is, early juvenile reproduction,
across the plateau from the western forested part a short interval between large seed crops
of the plateau eastwards onto the open ‘Causse’ and the production of many small winged seeds:
has occurred by occasional episodes of long that is, a suite of traits which favours dispersal
distance dispersal and establishment of isolated and rapid population establishment (Barbero
trees, followed by local colonization and nucleation et al. 1990; Rejmánek and Richardson 1996).
around these pioneer trees. Isolated trees produce This expansionist strategy (sensu Barbero et al.
up to four times as many cones as dominant trees 1990) can also be observed in Pinus halepensis
in groups and more than ten times the number colonizing abandoned agricultural fields
of cones produced by subordinate trees in dense in the plains of southern France, Spain, Croatia,
stands. Long-distance dispersal and nucleation and north-west Italy (Quézel and Médail 2003).

These models, although quite simple when a shifting balance between positive and negative
compared to the complex reality of the process, stim- interactions in relation to the life-cycle stage of
ulated ecologists to identify mechanisms of interac- individual plants, local environment (in particu-
tion and their role in what is a central process in lar abiotic stress), and successional stage (Pugnaire
plant population ecology. Since the publication of and Luque 2001). Studies of plant–plant interactions
the above paper it has become clear that to under- in successional old-fields, semi-arid grasslands,
stand the process of succession requires explicit and woody plant colonization of upland pas-
consideration of how interspecific interactions occur tures provide some illustrations of this dynamic
in relation to abiotic environmental conditions, par- balance between competition and facilitation in
ticularly soil water and nutrient levels and light, Mediterranean communities.
and the role of diverse trade-offs in colonization, First, in old-field populations that have devel-
community structure, competitive ability, and allo- oped following the abandonment of agricultural
cation to different traits (Tilman 1990; Garnier activities near Montpellier in southern France,
et al. 2004). there is a well-documented sequence of vegetation
A range of different modes and mechanisms succession over time (Escarré et al. 1983). Following
of succession no doubt occur in the succession a period of initial dominance by annuals, perennial
of vegetation after agricultural abandon in the Lamiaceae become important components of the
Mediterranean (Escarré et al. 1983; Lepart and vegetation within ∼10 years. This is followed by the
Escarré 1983; Debussche et al. 1985, 1996; Debussche establishment of a community containing dominant
and Lepart 1992; Zavala et al. 2000). In fact, the rel- perennial grasses (20–50 years), in which establish
ative importance of different processes may show several woody species. Forest closure occurs after

50 years or more. In the early parts of this succession, and Juniperus, hence the spatial heterogeneity of
Sans et al. (1998) reported that facilitation is import- the environment in which oak seeds germinate,
ant for regeneration of the short-lived herbaceous that is, the precise distribution and identity of other
species Picris hieracioides whereas nutrient competi- species, may have an important effect on the path-
tion is an important determinant of persistence. So way and speed of woody plant colonization and
although neighbouring plants positively promote succession.
seedling recruitment (perhaps via an improvement The works of Rousset and Lepart (1999, 2000)
of microclimatic conditions), they have a strong illustrate two important points. First, facilitation is
negative affect on later growth and reproduction. not limited to early successional stages where stress-
Second, field experiments in two semi-arid grass- ful conditions commonly occur. Even the establish-
land communities in south-east Spain by Maestre ment of late succession woody species, albeit in
et al. (2003) revealed how the bunch grass Stipa a hostile environment where drought and grazing
tenacissima consistently facilitates the establishment limit seedling recruitment, also depends on positive
of the shrub P. lentiscus due to a positive effect interactions. Indeed, a positive balance in species
on soil moisture, fertility, and microclimate despite interactions, due to the mediation of improved
strong below-ground competition between these soil conditions (particularly reduced nutrient and
two species. The magnitude of this facilitation water stress), may be a primary force regulating
effect varied significantly in relation to small-scale regeneration in semi-arid communities (Pugnaire
environmental variation, being greatest in stressful et al. 2004). Second, the extensive spread of woody
conditions. species across the Causses limestone plateaux of
Third, experimental work on the Causse du southern France is more due to the abandonment
Larzac, an upland limestone plateau in southern of cultivation than to reduced grazing pressure.
France, where Buxus sempervirens, Juniperus Although seedling recruitment is strongly limited
communis, and Q. pubescens are rapidly spreading by sheep grazing, which is responsible for 2/3 of
across the landscape (Fig. 4.2), illustrates how the seedling mortality of B. sempervirens, grazing did
establishment and persistence of woody species not have a strong effect on plant growth. Grazing
depends on the balance of positive and negative controls the spread of woody species but does not
interactions (Rousset and Lepart 2000). In this completely prevent it. Of course, a combination of
area Buxus has spread rapidly due to it no longer intense herbivory and drought stress (in addition
being cut by farmers (it was formerly used in to limited seed dispersal) can restrict establishment
stables and sheep pens and for soil enrichment). and survival of herbaceous species in open areas. For
Quercus seedlings are 50 times more abundant example, in disturbed sites in south-west Spain, the
under the canopy of Buxus and Juniperus (par- winter annual Geranium purpureum has a localized
ticularly on north-facing side of such shrubs) distribution that is almost invariably spatially aggre-
than in open grassland. Experimentally planted gated under dominant Juniperus trees (J. Herrera
acorns of Q. pubescens under pioneer Buxus and 1991b).
Juniperus showed significantly higher germination Microhabitat specific patterns of seed
and reduced seedling mortality than in open areas, germination, seedling survival and growth,
or under shrubs where the canopy had been cut. and reproduction are also known to occur in fleshy-
Microclimate and reduced grazing caused these fruited species whose seed rain is non-random
results. Although seedling growth was less under (C.M. Herrera et al. 1994; Jordano and Herrera
Buxus than under Juniperus and in open areas, 1995). Indeed, in some species there may be a strong
once the canopy was overtopped such effects spatial discordance between seed rain and seedling
disappeared. Facilitation due to reduced summer establishment, indicating the importance of eco-
drought stress and grazing can thus offset the logical constraints on regeneration. In Rhamnus
negative effects of competition. Important here is ludovici-salvatoris (endemic to the Balearic islands),
that different results were obtained under Buxus Traveset et al. (2003) found that whereas most seeds
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 137

are dispersed to sites under conspecifics, both in model of succession while the frequent installa-
open and woodland habitats, patterns of seedling tion of fleshy-fruited species under pioneer shrubs
distribution are not the same in the two habitats. and trees is indicative of a model of facilitation
In a Quercus ilex forest, most established seedlings via the presence of perching posts which facilitate
were found under the oak trees (despite the fact both arrival at a site and seedling establishment
that most seeds were dispersed to sites under due to improved local environmental conditions and
conspecific trees) whereas in an open field, juve- reduced herbivory. The relative importance of differ-
niles are more evenly distributed across different ent processes may vary in relation to the life-cycle
microhabitats. In this endemic species, population stage of individual plants, the local environment (in
dynamics are more limited by seed dispersal than particular abiotic stress), and the successional stage
the availability of microsites in both habitats. In of the vegetation. When conspecifics are scattered
addition, the ecological processes acting on seedling in an open landscape, spatial aggregation may be
recruitment and survival vary markedly between intense. However, a recruitment deficit may actually
microhabitats. Spatial patterns of recruitment thus occur close to seed bearers due to competition, either
vary markedly in space and cannot simply be pre- with the maternal plant or conspecific seedlings
dicted from seed dispersal patterns in this species. when at high density. Indeed in several species, high
In species with a secondary means of dispersal, seedling density under conspecifics has been found
such as by water in species that occur in riparian to be associated with higher rates of seedling mor-
habitats (e.g. Thébaud and Debussche 1991; Hampe tality (Debussche and Lepart 1992; C.M. Herrera
2004), a spatial decoupling of wind or vertebrate- et al. 1994; Traveset et al. 2003), probably because
dispersed seed rain and seedling distribution may of intraspecific competition, although localized pat-
be frequent. terns of herbivory and parasitism could also con-
In other cases, nucleation due to dispersal pat- tribute. In contrast, a more continuous shrub cover
terns can be further enhanced by a positive (nurse) may extend the tail of the seed shadow and poten-
effect of adults on seedling establishment due to tially reduce intraspecific competition and thus
reduced herbivory, improved microclimate due increase local recruitment. The dynamics of such
to the buffering effect of litter on evapotranspi- populations will thus be closely linked to the spatial
ration and temperature extremes, improved soil structure of the vegetation within and among popu-
nutrient and organic matter content, less compact lations, which may set the stage for differentiation to
soil structure and reduced aridity. In P. latifolia in evolve.
the Sierra de Cazorla of southern Spain seedling
survival varies significantly among microhabitats
4.4.5 Consequences of reforestation for
in scrubland but not in forest due to the role of
landscape pattern and biodiversity
facilitation played by shrubs in the former habitat
type (C.M. Herrera et al. 1994). The role of stress Contemporary reforestation is a natural process
avoidance in facilitation is well known (Bertness and which shows distinct spatial heterogeneity of dom-
Callaway 1994) hence facilitation, and consequent inant species and speed among regions (Lepart and
nucleation, may be key processes in the spatial orga- Debussche 1992). For example, in southern France,
nization of plant populations in Mediterranean habi- old-fields are primarily colonized by vertebrate-
tats where abiotic stress is common (Koechlin et al. dispersed fleshy-fruited species (e.g. in Pistacia,
1986). Rhamnus, Juniperus, and Phillyrea), many areas of
This section illustrates the diversity of ecolog- lowland garrigues have become dominated by
ical factors influencing dispersal and regenera- P. halepensis, in riparian areas a multispecies forest
tion, and thus the mechanisms of succession, in with Populus and Alnus has developed and upland
the Mediterranean mosaic landscape. The close limestone plateaux are being actively colonized
fit between seed rain and seedling distribution of by B. sempervirens, J. communis, Q. pubescens and
woody species in old-fields supports the tolerance P. sylvestris in many areas, or exotic Pinus nigra

close to early twentieth century plantations. This during periods of extensive traditional agricultural
regional specificity of forest spread has developed processes. In addition, species in understorey herba-
in response to three principal causes (Debussche ceous vegetation are not renowned for their rapid
et al. 1999): (a) differences in the proximity of an colonization capacity. A study of temporal changes
existing pool of tree species for reforestation and in the vegetation of Q. pubescens woodland after
herbaceous understorey species, (b) local ecological the abandonment of coppicing and grazing illus-
conditions and topography, and (c) the spatial trates that although the vegetation of previously
organization of habitats in the landscape. exploited and grazed forest has become similar to
The importance of proximity of source popula- that of undisturbed forest, the floristic diversity may
tions is illustrated by the spectacular disappearance never regain levels similar to those of undisturbed
of terrace agriculture, where the small size and diffi- woodlands (Debussche et al. 2001). The rarity of
cult access of terraces rapidly made them unsuitable such undisturbed woodland in the landscape and
and unprofitable for mechanized agriculture (Lepart the lack of mechanisms favouring long-distance dis-
and Debussche 1992; Grove and Rackham 2001). persal in many forest plant species are probably the
Terraces by definition occur on hill sides, that is, causes. So even though a natural forest cover may
close to remnant areas of forest in less accessible be re-establishing itself in many areas, its biodiver-
and less hospitable areas (on crests, ridges and steep sity, in terms of herbaceous species richness, may be
slopes, and closer to summits). This proximity has low, relative to the species diversity in open habitats
allowed for the rapid arrival of tree species onto (e.g. the species rich open grassland and semi-
abandoned terraces, whose deep soils and open flat rupicolous communities of southern France) and
surface promotes the rapid reconstitution of forest. the original primeval forests that probably occurred
In many areas, oaks have directly colonized such in these areas. Plantations of exotic species clearly
areas, without a temporary stage of colonization have significantly negative effects on species diver-
by pioneer species (Debussche et al. 1999). In the sity, notably a decline in the presence of endemic
absence of disturbance, succession leads primarily species, as reported for heathlands in southern Spain
to deciduous oak dominance, whereas repeated dis- (Andrés and Ojeda 2002).
turbance leads to the development of more open Second, although natural reforestation and other
sclerophyllous vegetation (Tatoni et al. 1994). The landscape changes associated with alterations in
colonization of burnt areas by P. halepensis closely human land-use patterns may have little effect
depends on the proximity of seed bearing trees near on natural populations of many protected species,
the burnt site (Abbas et al. 1984). Dispersal limitation due to their rupicolous and chasmophytic nature
may thus slow re-colonization after a large fire. The Chapter 2, some protected species, in addition to
spontaneous spread of species from plantations of those in coastal habitats, are clearly threatened
human origin also illustrates the importance of the by contemporary landscape change. Species which
proximity of source populations. Examples of this require gaps and open habitats for flowering are
phenomenon (see Chapter 6) include P. pinaster in a prime candidate. An example in point concerns
the Cévennes and the Maures massif in Provence native peonies such as Paeonia mascula and Paeo-
(Mazurek and Romane 1986), Fraxinus ornus in ripar- nia officinalis which give a bright splash of colour
ian vegetation along the Hérault river (Thébaud to forest gaps during their flowering period. Pop-
and Debussche 1991) and P. nigra on the Causse ulations of both species (in Israel and southern
Méjean (Lepart et al. 2001). In addition to the spread France, respectively) are threatened by rapid habi-
of some exotic species, there are other ecological tat change and forest closure. Both species show
consequences of rapid reforestation. significantly reduced flowering in dense shade com-
First, in some areas, the development of fairly pared to open clearings and forest gaps (Andrieu
natural and diverse understorey vegetation may be 2002; Ne’eman 2003), a pattern also reported for
rapid, but in other areas very slow due to their isola- peonies in the Californian chaparral (Keeley 1991).
tion from areas where a forest cover was maintained For such long-lived perennial species, the speed of
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 139

forest closure in some areas is so fast that plants now for local adaptation in life-history traits related
in the forest include many which established them- to survival, growth, and reproduction. In the
selves in an open habitat. Population persistence Mediterranean mosaic of neighbouring populations
will thus depend on how different components of which differ greatly in age and ecology, populations
regeneration (germination, seedling establishment, may frequently colonize ecological conditions
and survival) and reproduction (flowering, fruiting, different from their source populations, which, in
and seed dispersal) are affected by forest spread. conjunction with the temporary nature of pioneer
Finally, a decrease in the grassland–shrubland populations may limit the evolution of adaptive
mosaic landscape as a result of reduced grazing variation. In addition, within-habitat spatial hetero-
pressure and cultivation may in some areas be asso- geneity of local conditions may mask any selective
ciated with a loss of biodiversity. For example, in differences among sites. The achene heteromor-
a comparison of evergreen oak woodland, grass- phism in this species is likely to be important here.
land, shrubland, and grass-land–shrubland mosaic Peripheral achenes have a poor dispersal capacity
vegetation, Verdú et al. (2000) reported that species in space but produce more competitive offspring
richness is maximum in the grassland–shrubland that the smaller more easily dispersed central ach-
mosaic landscape. A return to the forest may thus enes (Imbert et al. 1997). Hence, within individual
have a negative consequence for plant biodiversity. variation in seed traits, an excellent expression of
phenotypic plasticity, may provide a highly flexible
adaptive strategy in a mosaic of successional old-
4.4.6 Trait variation in relation to succession
fields where persistence of annual plants depends
and reforestation
on both high dispersal rates and good competitive
Populations of a single species which occur in differ- ability.
ent successional stages provide model systems for The role of functional trait variation in the suc-
the study of trait variation as environmental con- cessional dynamics of Mediterranean old-fields has
ditions change. Yet, there have been surprisingly been analysed by Garnier et al. (2004). These authors
few direct studies of trait variation along succes- reported that early succession habitats are domin-
sional gradients and the few studies that have been ated by fast-growing plant species with high specific
done provide little evidence for adaptive variation, leaf area (the ratio of water-saturated leaf area to leaf
population persistence along such gradients being dry mass which is positively correlated with photo-
greatly favoured by phenotypic plasticity (Gray synthetic activity) and leaf nitrogen content but low
1993; Thompson et al. 1993). The Mediterranean leaf dry matter content (the ratio of leaf dry mass to
mosaic of old-fields of variable age and contrast- water-saturated fresh mass) and tend to have high
ing open areas and woodland provides an ideal rates of resource processing per unit biomass. In late
landscape context for such work. succession habitats the reverse is the case. These dif-
Two studies based on reciprocal transplant ferences in traits among successional stages point
experiments have provided consistent support to the replacement of fast-growing species which
for the idea that trait variation along successional acquire external resources rapidly in early stages
gradients is primarily a result of phenotypic plas- by more slowly-growing species which efficiently
ticity. Using a pioneer (3 year old) population and conserve resources as succession proceeds. The fact
a 40 yr old population of the biennial P. hieracioides, that these relationships are maintained when only
Sans et al. (1998) showed that different populations the traits of the two dominant species are consid-
in the succession show a similar response to the ered strongly suggests that modifications to com-
selection pressures in the two habitats. Likewise, munity dynamics and structure depend more on
experiments with the colonist annual Crepis sancta species traits than on species number. These results
in three Mediterranean old-fields that differ in age reflect what Herms and Mattson (1992) term a bal-
since abandonment (6, 10, and 15 years) by Imbert ance between growth and differentiation in which
et al. (1999b) showed only very limited evidence growth-dominated plants invest a high proportion

of their resources into resource acquisition and high habitats than in open habitats where plants cap-
relative growth rates, while differentiation dom- ture significantly fewer insects (Zamora 1995). The
inated plants primarily invest resources into the decline in performance was manifest even when
non-growth processes and structures which facili- plants received additional prey, hence carnivory did
tate efficient use of resources, and consequent low not compensate for diminished photosynthesis. In
relative growth rate. this species, plants in open habitats bear curled
Rapid reforestation in the Mediterranean means leaves (probably to reduce water loss) with higher
that plants have to cope with deep shade in addi- levels of mucilage secretion than plants in deep
tion to summer drought. Sack and Grubb (2002) shade which bear more flattened leaves to maximize
subjected a range of species, including two species photon capture. The latter may only capture smaller
common in Mediterranean forests, Viburnum tinus insects but in such habitats insect prey is more plen-
and Hedera helix, to this combination of constraints. tiful than in open areas (Zamora 1995; Zamora et al.
For these species they did not find, as trade-off the- 1998). These patterns illustrate once again the com-
ory would predict, a greater impact of drought stress plexity of factors acting on plant response to nutrient
when plants were also subject to deep shade. This is uptake (and in this case prey availability and cap-
surprising since it is difficult for a plant to specialize ture), water loss, and irradiance in a Mediterranean
both in shoot traits that favour irradiance capture context. The favoured habitat of this species is tran-
and root function to avoid and/or tolerate drought. sitional habitats between open areas (where there is
The result suggests that some traits must allow a surplus of light but little prey and potential water
for both reduced demand for irradiance and water stress) and deep shaded areas (where water and food
simultaneously; for example, reduced size and low are available, but additional prey cannot compen-
respiration rates, that is, the presence of long-lived sate for lack of light). Hence, along the environmen-
leaves with low specific leaf area. In the polyploid tal gradient on which this species occurs, the limiting
complex Arrhenatherum elatius (see Chapter 3), Petit factor changes. The message is clear, to study plant
and Thompson (1997, 1998) reported marked differ- response and adaptation requires appreciation of
entiation in morphological traits among woodland responses to diverse environmental variables.
and open habitat populations of the tetraploid cyto-
type which concords with the direction of selection
4.4.7 Forest fires: successional dynamics and
in such habitats (taller plants favoured in open
trait adaptation
habitats). These authors suggest that adaptive dif-
ferentiation among tetraploid populations in open Fires were an integral component of the
and woodland habitats in the Mediterranean mosaic Mediterranean landscape long before humans
has been an integral component of the spread and learnt their use (Chapter 1) and may thus have
persistence of the polyploid cytotype in a wider influenced trait evolution. Mediterranean fires may
range of environments than the diploid progenitor. impact on plant population ecology and evolution
A fascinating example of plant response to in two main ways. As a disturbance, fires open
variation in irradiance occurs in the carnivorous the environment, creating spatio-temporal hetero-
endemic Pinguicula vallisneriifolia. In this species geneity of the environment and new opportunities
variation in light availability influences investment for colonization and succession. As Barbero et al.
in carnivory in natural situations. This species (1987a: 48) commented ‘the fire-shaped landscape
inhabits damp rocky habitats over a range of light perfectly corresponds to the definition of the mosaic
environments in south-east Spain (Zamora et al. community’. Second, as a selective agent, fire may
1996). In an experimental study of transplanted indi- shape the evolution of traits that allow species to
viduals, Zamora et al. (1998) reported a decreasing adapt to fires.
gradient in performance from sunny environments The dynamics of vegetation after fire has been
to deep shade over a distance of a few metres. well documented in Mediterranean communities
However prey availability is greater in damp shady (Trabaud and Lepart 1980; Trabaud 1987, 1992).
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 141

Immediately following the passage of a fire, species threatened (Moreno et al. 1998), hence the frequency
diversity remains low, reaching a peak within of fires may have a major effect on the mosaic pat-
three years after the passage of a fire. After about tern of vegetation and species diversity. This effect of
five years, species diversity declines to a level equiv- fire frequency on mosaic structure of vegetation has
alent to that prior to the fire. A key point here is also been discussed in other Mediterranean-climate
that the composition of many plant communities ecosystems (e.g. Zedler et al. 1983).
returns to its pre-fire state fairly quickly (Casal 1987; Although it is difficult to assign the status of a
Trabaud 1987). For example, garrigue communities true pyrophyte (i.e. a species whose propagation or
in southern France studied by Trabaud and Lepart reproduction is stimulated by fire or which resist
(1980) recovered more than 75% of their original veg- fires) there are a number of traits which illustrate
etation within 10 years after a fire. After fire, one thus adaptation to fire. These have been classified in dif-
observes the return of previously present species, ferent ways by different authors (Kunholtz-Lordat
rather than a process of species turnover which is 1958; Naveh 1975; Grove and Rackham 2001). Here,
more characteristic of secondary succession. I identify five main strategies:
Fire frequency can affect pattern and diversity in
Mediterranean plant communities. Very high fire Resistance. Traits such as the thick insulating and
frequencies create a patchwork vegetation struc- protective bark of trees such as Quercus suber (Pausas
ture with a low species diversity dominated by 1997), Q. ithaburensis (Naveh 1975), or Pinus bru-
resprouting species which rapidly propagate into a tia (Eron 1987) provide a means of surviving a fire
dense cover, for example, Q. coccifera in the west- and thus regeneration. The flammability of many
ern Mediterranean. In north-east Spain, sites with pines, sclerophyllous shrubs and trees, and aromatic
a high fire recurrence have recently shifted species species may also contribute to this resistance by has-
composition from domination by P. halepensis to tening the passage of a fire. In P. halepensis the loose
domination by evergreen oaks because the fire fre- fallen needles act as fuel which allows ground fires
quency is shorter than the time (15–20 years) for to spread rapidly.
P. halepensis individuals to produce seeds and accu- Resprouting. Many plants may have their above-
mulate a seed storage capacity for self-replacement ground parts burnt to ground level without major
(Pausas 2001). In contrast, as shown by Trabaud and damage to the roots, allowing them to resprout after
Galtié (1996) working in the Massif des Aspres in fire: e.g. Q. coccifera and Q. lusitanica in the western
southern France, low fire frequency causes greater Mediterranean, and Q. ithaburensis in the eastern
spatial heterogeneity in the landscape and greater Mediterranean, and species of Phillyrea, Rhamnus,
diversity in plant communities, with the estab- Arbutus, and Erica (Naveh 1975; Trabaud and
lishment from seed of numerous pioneer woody Lepart 1980; Barbero et al. 1987a; Kummerow et al.
species such as Thymus, Lavandula, and Cistus, which 1990; Ojeda et al. 1996b). Critical to this resprouting
develop rapidly. Such seeder species may however strategy is the presence of burls, root crowns, and
be eliminated at very low fire frequency in the rootstocks, particularly swollen lignotubers rich in
absence of resprouting species if serotinous species starch at the base of the stem. As shown for Erica
are present. In the absence of any disturbance such species of the Cape floristic region, plants with a
as fire, evergreen oaks such as Q. ilex will replace resprouting strategy build-up starch reserves in
P. halepensis by virtue of its enhanced shade toler- the roots, even in young seedlings which provide
ance (Zavala et al. 2000). These patterns illustrate a stock of rapidly usable reserves for growth after
that the predicted dominance of seeders at high a disturbance such as fire (Bell and Ojeda 1999;
levels of disturbance (in this case high fire fre- Verdaguer and Ojeda 2002). The resprouting strat-
quency) does not apply to fire-prone ecosystems egy is favoured by high fire frequency (Barbero et al.
of the Mediterranean which contain species with 1987a) and severe summer drought (Ojeda 1998).
traits that facilitate resprouting. In fact, in such sys- Seed and seedling biology. In situations where
tems, if fires are too frequent, biodiversity may be fire frequency is reduced, a high reproductive

output and a suite of seed traits may be favoured, trees (Tapias et al. 2001). This age-related plasticity
for example, small seeds, with hard coats and a may be adaptive. However, critical analysis of this
fairly long persistence in the soil seed bank, or phenomenon has shown that a large amount of seed
the fire-stimulated germination and rapid seedling release (∼60%) may occur in the absence of fire, sug-
establishment observed in Cistus in the western gesting that this species is only partially serotinous
Mediterranean (Thanos and Georghiou 1988; and that this trait may have evolved in response
Trabaud and Oustric 1989; Roy and Sonié 1992). to other environmental factors (Nathan et al. 1999).
In Cistus, some species produce two types of seeds, Serotiny is more common in other Mediterranean-
some which germinate without fire and others, with type ecosystems than in the Mediterranean flora.
a hard seed coat that germinate after treatment at
high temperatures (Thanos and Georghiou 1988). Interpretation of the adaptive significance
Seedlings of species with a ‘seeder’ response to fire of such traits requires a certain prudence. The
also show a better drought resistance compared to different strategies which favour resistance to, or
seedlings in species which resprout and produce few re-colonization after, fire, also favour tolerance
seedlings after fire (Ojeda 1998). Another seed trait of summer drought and other environmental
which may favour resistance to fire is trypanocarpy, stress (Naveh 1975). Hence, it would be unwise to
the capacity of seeds to bury themselves in the soil. conclude, on the basis of present knowledge, on
Such seed movements are often associated with the adaptive nature of such traits solely in terms
the presence of a hygroscopic awn, for example, in of a response to selection pressures imposed by
Stipa tortilis in fire-prone steppe and arid vegetation fire. As Pausas (2001) argues, it is necessary to
of the eastern Mediterranean (Naveh 1975). In other consider a range of traits (age at maturity, serotiny,
Mediterranean-climate regions, the adaptive nature seed dormancy and germination, and survival in
of germination strategies (e.g. in response to heat shaded conditions) and fire regimes in order to
and charred wood, or following the release of nitro- make biologically realistic predictions of the relative
gen and smoke) is firmly established (e.g. Keeley fitness of different strategies.
et al. 1985; Thanos and Rundel 1995). The persis- Different combinations of traits may evolve in
tence of such seeder species depends on their ability systems with different fire histories. Such vari-
to produce seeds between two fires, seed survival ation could help explain the relative lack of seroti-
during a fire, and enhancement of regeneration and nous species in the Mediterranean flora relative
recruitment as a result of a fire passage. to other Mediterranean-climate areas. Pausas et al.
Dispersal and re-colonization. Traits which favour (2004) developed a fourfold classification of species
dispersal and re-colonization may allow species in relation to the basic fire-response strategies
which lack resistance to rapidly re-colonize burnt (i.e. resprouting ability and propagule persistence):
areas. In P. halepensis the light, wind-dispersed resprouters (R+), non-resprouters (R−), propagule
seeds favour dispersal to open areas and have persisters (P+) or propagule non-persisters (P−).
favoured the spread of this species in burnt areas They then examined how combinations of supposed
(Achérar et al. 1984; Barbero et al. 1987b). fire-response traits vary in different regions and the
Serotiny. This is the capacity to maintain old extent to which different strategies co-occur with
unopened cones for several years in a ‘canopy- other traits. Their study showed marked variation
stored’ seed bank and release seeds only after the in the relative abundance of each of the four types in
passage of a fire. It is present in just a few species in different Mediterranean ecosystems. For example,
the Mediterranean flora, for example, P. halepensis most resprouters in the Mediterranean flora do not
and P. brutia, Cupressus sempervirens, and Tetraclinis persist as propagules, that is, they are R+ P−, while
articulata (Quézel and Médail 2003). In P. brutia in California, resprouters are evenly distributed
serotiny may favour re-colonization of burnt sites among propagule persisters (R+ P+) and propag-
(Eron 1987) and in P. halepensis young trees require ule non-persisters (R+ P−). The two basic response
a higher temperature for seed release than do older traits were found to be closely correlated with other
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 143

traits, for example, resprouters are longer-lived and also Verdú 2000). These results are consistent with
allocate more resources to storage and basal buds. the idea that allocation trade-offs cause the negative
However, not all resprouters may respond in a sim- relationship between the capacity for resprouting
ilar way to fire because of different combinations of and propagule persistence. The fact that resprout-
associated traits in different regions, that is, differ- ing precedes propagule persistence strongly points
ences in stored seed banks and modes of dispersal. to an origin for the former trait in the older lin-
Hence attempts to predict population-level changes eages present in the Tertiary flora, that is, prior
and the relative fitness of seeder and resprouting to the onset of the Mediterranean-climate regime.
species require integration of the various combina- Resprouting capacity thus appears to have been part
tions of traits and fire regimes in different regions. of the Pre-Pliocene suite of traits that has not evolved
One should thus be cautious in any interpretation in response to environmental changes associated
of the adaptive significance of the above traits in rela- with the onset of a Mediterranean climate.
tion to fire. Several of these traits may have evolved
in association with other selection pressures, per-
4.4.8 Colonization and adaptation on
haps prior to the onset of a Mediterranean-climate
abandoned mines
regime. They may thus represent an example of
pre-adaptation in Mediterranean fire-prone com- Since the early 1960s, plant colonization on mine
munities. Two studies back up this claim. First, spoils has provided a clear illustration of how strong
resprouting is a widespread trait in plants of shrub- natural selection can be and how quickly plant
land in the mountains of Mexico (Lloret et al. 1999). populations can adapt to local conditions (McNeilly
The marked morphological and floristic similar- and Antonovics 1968; MacNair 1987). These stud-
ity of this ‘mexical’ shrubland to Mediterranean- ies illustrated that plants from soils polluted with
type vegetation is strong evidence that it represents heavy metals such as zinc, lead, cadmium, copper,
a relict of the Tertiary flora which has persisted and nickel can show a greater capacity for growth
either under (in California) or in the absence of and survival in the presence of such elements at
(Mexico) a Mediterranean-type climate. The preva- high doses than plants of the same species from
lence of resprouters characterized by the presence adjacent unpolluted sites. What fascinated biolo-
of a lignotuber or burl in the mexical shrub- gists at the time was that even though gene flow
land community strongly suggests that resprouting was occurring from unpolluted sites into the pol-
evolved to cope with disturbance other than fire, luted population, this was not enough to counter
that is, prior to the onset of the Mediterranean genetic differentiation between populations on the
climate in California. Second, in a phylogeneti- different soils. The clear message was that selection
cally controlled analysis of the correlated and inde- can be strong enough to produce adaptation, even
pendent occurrence of resprouting capacity and when gene flow occurs at significant rates. To reduce
propagule-persistence ability in plants of the Iberian the ‘harmful’ effects of breeding with ‘unadapted’
peninsula, Pausas and Verdú (2004) detected a individuals from different populations, populations
significant negative association between the two on polluted sites were also found to have a greater
response strategies, that is, R+ P− and R− P+ were chance of self-fertilization than those on unpolluted
more frequent than expected by chance. In addition, soils. In the absence of heavy metals however, plants
their results do not suggest a correlated evolution from polluted sites were less vigorous than those
of these two traits in R+ P+ species as a result of from unpolluted sites, they have a cost associated
selection on one trait. In their dataset, resprouting with their resistance in terms of a fitness reduction in
capacity (R+) preceded propagule persistence (P+) the absence of the metals. As a result, the frequency
in species which combine these traits. Their study of resistance genes should decline away from pol-
also confirmed the co-occurrence of longevity with luted sites. More recently, there has been much
resprouting capacity and the trend for juveniles of interest in the precise physiological and genetic
resprouters to grow more slowly than seeders (see mechanisms which allow plants to (a) adapt to what

are highly toxic molecules (MacNair 1987, 1993) and of physiological traits that allow plants to colonize
(b) hyper-accumulate heavy metals in their aerial severe environments. Based on controlled cross-
parts (Escarré et al. 2000) . ing experiments, Frérot et al. (2003) found that
Abandoned mines litter the back country of sev- zinc accumulation was significantly higher in the
eral Mediterranean regions where they provide offspring of crosses among non-metallicolous par-
replicated sites for experimental analysis of the traits ents than in the offspring of both crosses among
which facilitate colonization and adaptation of new metallicolous parents and crosses between the two
environments. Rates of gene flow among popula- ecotypes (the latter having intermediate capacities
tions can now be assessed with powerful genetic for zinc uptake). There is thus heritable variation
markers that were not available to evolutionary bio- for zinc uptake and accumulation capacity in the
logists in North Wales in the early 1960s. How plants studied populations, which may involve a simple
cope with several compounds at high doses can also monogenic control with two alleles. A homozygote
be evaluated now. This is important because toler- recessive for the locus involved would be non-
ance of one metal does not go hand in hand with tol- metallicolous, a single dominant gene producing
erance of others, the genes for resistance to different intermediate zinc uptake and a dominant homozy-
compounds can be different (MacNair 1987, 1993). gote blocking uptake and accumulation capacity.
Adaptation to one metal may thus be compromised Such a simple genetic control may facilitate the
by genetic adaptation to another metal. Finally, in evolution of hyperaccumulation capacity, both in
the case of mines in the Mediterranean region let us the wild and in breeding programmes designed for
not forget that the toxic environment is made even phytoremediation and restoration purposes.
more stressful by summer drought, hence tolerance
may be even more difficult to achieve.
Work on Thlaspi caerulescens, well studied for its 4.5 Variation and adaptation in
metal tolerance and ability to colonize polluted soils aromatic plants
in temperate Europe illustrate some of these issues,
4.5.1 Abundance and diversity of aromatic
as well as being of particular interest to restora-
plants in the Mediterranean
tion ecologists and landscape project managers by
virtue of its capacity to accumulate high concentra- Where else in the world do plants smell so
tions of heavy metals in its tissues (Box 4.9). Despite good? Although most plants produce secondary
ecotypic differentiation between the two types of compounds, their presence is enhanced in the
population, the genes which confer tolerance are Mediterranean where one of the most striking
not limited to contaminated sites and plants from features of many plant communities is their
contaminated and uncontaminated sites show no fragrance. One of the most characteristic and
clear pattern of isozyme differentiation (Dubois et al. abundant types of volatile compounds produced
2003). Either the two ecotypes have evolved recently by Mediterranean plants are the monoterpenes
or there is ongoing gene flow among the two types produced in essential oils. Monoterpenes have a
of population. 10-carbon, 16-hydrogen structure which may
Plants of T. caerulescens also have the ability to take the form of acyclic and cyclic non-aromatic
accumulate high concentrations of the heavy metals molecules as well as truly aromatic types. They
in their aerial parts, although this accumulation does are ubiquitous in higher plants, and accumulate
occur in the same way for different metals. Whereas to significant levels in ∼50 angiosperm families
plants from metallicolous populations accumulate (Seufert et al. 1995). To put a number on essential
cadmium, those from non-metallicolous popula- oil presence in the Mediterranean flora, Ross and
tions accumulate zinc (Escarré et al. 2000). This Sombrero (1991) recompiled data in Guenther’s
ability to hyperaccumulate toxic compounds is (1948–52) classic work on the presence of essential
particularly interesting for the restoration of such oils in the angiosperms. In a total of 153 genera (from
ecosystems and also for studies of the genetic basis 49 families) that have species which accumulate
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 145

Box 4.9 Re-colonization and adaptation to heavy metals on abandoned mines in

southern France (photos and graph kindly supplied by H. Frérot)

In southern France, Thalaspi caerulescens and that is, present (albeit at lower levels) in plants
Anthyllis vulneraria occur in ‘metallicolous’ (M) in non-metallicolous populations. In fact, in
populations on a range of abandoned mine sites T. caerulescens most plants can survive on
where soils are contaminated with zinc, lead, and contaminated soils regardless of their origin,
cadmium and in ‘non-metallicolous’ (NM) unlike in other species such as A. vulneraria
populations in the surrounding landscape. where populations from polluted sites have
Metallicolous populations show greater tolerance a greater tolerance. In T. caerulescens,
of polluted soils (based on the ratio of aerial however, growth and reproduction on
biomass on contaminated/uncontaminated soils) contaminated soil is severely reduced
than non-metallicolous populations (Escarré et al. in plants from non-metallicolous
2000). This tolerance is however ‘constitutive’, populations.


Thlaspi M
Survival (proportion)

0.8 ns
Thlaspi NM
0.6 Anthyllis M
Anthyllis NM


0 1,000 2,000 3,000 4,000
Zinc concentration (µm)

essential oils, 58 occur only in Mediterranean- Table 4.1 Storage organs and their volatile oils in some common
type ecosystems, 32 in Mediterranean and other Mediterranean genera
regions, 42 are tropical, 17 temperate, and 4 are
Storage organ Family Genus Compounds
cosmopolitan. Hence, 90 of the 153 genera occur
in a Mediterranean-type ecosystem. Although, this Resin canals Coniferae Pinus Oleoresins
exploration lacks a rigorous comparative analysis Picea
of numbers of available genera for study in each Juniperus
region and the necessary correction for phyloge- Anacardiaceae Pistacia
netic relatedness and effects of sorting processes
(the large number of genera may be linked to high External Cistaceae Cistus Aromatic resins
rates of diversification in a few families such as the glandular hairs Helianthemum
Lamiaceae, Apiaceae, and Asteraceae which are Glandular Lamiaceae Salvia Monoterpenes
rich in secondary compounds), the abundance of trichomes Thymus
aromatic plants is clear. Cannabaceae Humulus
Essential oils and resins which have a volatile Cannabis
oil component are typical of perennial, evergreen, Myricaceae Myrica
xeromorphic shrubs, and are, not surprisingly, less Oil cavities Rutaceae Citrus
abundant in fast growing and drought avoiding Myrtaceae Eucalyptus
species, particularly annuals. In many species the Schizogenous Umbelliferae Coriandrum
biosynthesis of volatile oils represents a sizeable oil ducts Asteraceae Carthamus
fraction of resources in terms of the amount of
fixed carbon used up in their production (Ross and
Sombrero 1991). For example, in Q. ilex, monoter- in production among closely related species and
penes represent a significant feature of carbon among populations of individual species. Although
cycling (Staudt et al. 2001). the material of this section differs somewhat from
Volatile oils may contain toxic byproducts of pri- that of previous sections, the questions are sim-
mary metabolite synthesis and thus require storage ple variants on the common theme of the chapter:
in specialized tissues where they do not disrupt nor- what is the function of such compounds and is
mal cell function (Table 4.1). Glandular hairs and intraspecific variation adaptive in a Mediterranean
trichomes on the leaf surface are a common form of context?
such storage. Glandular hairs can be of two main To understand why volatile essential oils become
types: peltate and capitate, which differ in structure accumulated in specialized organs in many
and function (Werker et al. 1985). Peltate hairs consist Mediterranean plants requires recognition of why
of a basal cell, a short, wide stalk cell, and an almost they are produced in the first place. Two view
circular head of secretory cells. Capitate hairs differ points can be read on this issue. First, volatile oils
by having a uni or bicellular head. In most species represent secondary compounds. In other words
(for an unknown reason Rosmarinus officinalis is an they represent the byproducts of phytohormone,
exception) the hairs can be seen on both leaf surfaces, phytosterol, and carotenoid production (Banthorpe
on calyces and on green stems. et al. 1972), although their accumulation may then
Monoterpenes are ideal compounds for the study be of adaptive significance in relation to several
of how ecological interactions are mediated by sec- features of the abiotic and biotic environment.
ondary compounds for a number of reasons (see also Alternatively, some authors (e.g. Seigler and Price
Lerdeau et al. 1994): their biosynthetic pathways are 1976; Croteau 1987) have proposed that such
fairly well known, they can play multiple ecologi- compounds are not simple waste products because
cal roles due to their effects on biological processes they are often catabolized in a highly specific and
in the surrounding environment, they represent ordered fashion and may thus play a primary role
an important investment of fixed carbon and may in physiological processes. Although I recognize the
thus be costly to produce, and they show variation possibility of this primary function, I will often refer
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 147

to volatile essential oils, simply for convenience, that potential herbivores may respond to variation
as secondary compounds. They nevertheless have in monoterpene composition in Mediterranean
several important ecological functions. species. Several Mediterranean species have high
concentrations of secondary compounds in their
fruits (e.g. Milesi et al. 2001), suggesting a role in
4.5.2 Ecological functions of secondary
seed protection. But what is lacking, as I write this
compounds: biotic interactions
book, is direct evidence from a natural situation of
A principal function of chemical compounds in monoterpene-mediated defence in a Mediterranean
plants is their role in the mediation of biotic inter- plant. Evidence of a role for chemical defence is
actions, in particular (and there is an immense variable, in some studies increased phenolic content
literature on these subjects) defence against para- of leaves across species has been shown to be asso-
sites, predators, and herbivores, allelopathic effects ciated with reduced browsing (Massei et al. 2000)
on associated species which reduce competition and while in others variation in the phenolic content of
pollinator attraction. the leaves was not associated with differences in
Following the claim by Fraenkel (1959) that insect attack (Glyphis and Puttick 1989).
defence against herbivores and parasites represents The presence of secondary compounds may medi-
the veritable ‘raison d’être’ of secondary compounds, ate or influence community structure where they
Ehrlich and Raven (1964) argued for the idea that affect the germination, growth, and survival of
novel chemical compounds allow plants to escape associated plant species. The role of such ‘allelopa-
such attacks. They discussed how the coevolution- thy’ as an ecological process acting to modulate
ary response by predators and parasites, which in competitive interactions and structure plant com-
turn evolve to counter new defence mechanisms, munities has been a hotly debated subject. In the
push defensive systems to continually evolve if they early 1960s observations of bare zones around some
are to be effective. The result of this is radiation, both aromatic shrubs in Californian chaparral, combined
for the plants and their enemies. There is now a solid with experiments in controlled conditions, led to
body of literature on the importance of secondary the claim that allelopathic effects of dominant shrub
compounds in defence against herbivores (Herms species such as Adenostoma fasciculatum suppress
and Mattson 1992). the germination and growth of associated species
There are many precise examples from and inhibit the germination of conspecific seeds
Mediterranean ecosystems of how herbivory (e.g. McPherson and Muller 1969). Such observa-
and parasitism can have a lasting and sometimes tions were used to develop the idea that allelo-
severe impact on population dynamics (Thébaud pathic interactions are a key feature of plant–plant
et al. 1996; Escarré et al. 1999) and community interactions and community organization in such
ecology (Noy-Meir et al. 1989; Sternberg et al. 2000), ecosystems. At about the same time, Whittaker and
how variation in pest infestation may be linked Feeney (1971) emphasized the role of chemical com-
to particular plant traits (e.g. Alonso and Herrera pounds in the mediation of interspecific interactions.
1996), and how variation in pest populations However convincing experiments in natural condi-
may closely parallel patterns of the host (Burban tions have rarely if ever been performed and as a
et al. 1999). However, there is surprisingly little result the experimental results and their applicabil-
published information on the role of monoterpenes ity to natural systems are far from being clear cut
in defence against herbivory. The role of conifer (Rice 1979; Inderjit and Dakshini 1995). In the case
resins and monoterpene alcohols in defence against of A. fasciculatum it was later found that the concen-
insect and small mammal attack due to their tration of phenolics in the soil under adult plants
impact on digestive processes is well known from was insufficient to prevent germination and growth
studies elsewhere (Seufert et al. 1995) and studies of associated species, soil toxicity being more the
on T. vulgaris by Linhart and Thompson (1995, result of phytotoxins produced by the soil microbes
1999) in controlled conditions provide evidence under the dominant shrubs (Kaminsky 1981).

In the Mediterranean, an allelopathic role of essen- concentrations in the topsoil are ecologically sig-
tial oils has often been proposed. For example, nificant remains to be precisely determined. The
Barbero et al. (1987a: 45) stated that ‘Mediterranean addition of essential oils obtained from aromatic
communities characterized by Thymus vulgaris, plants (S. thymbra and Thymus (Coridothymus)
Dorycnium suffruticosum, Staehelina dubia, and Lavan- capitatus) can stimulate soil respiration and cause
dula latifolia are not very diversified … because of an increased organic matter content of soils and at
allelopathic reactions’. Although direct proof that the same time diminish fungal spore germination
this is true in the wild remains to be furnished, in controlled conditions (Vokou and Margaris
there have been some encouraging studies of several 1984, 1988). Increased respiration is the result from
Lamiaceae and the Cistaceae. increased numbers and activity of the bacterial
community in the soil which may use the oils as
1. In the eastern Mediterranean, individuals of
a substrate for growth. Recent work by Ehlers and
the aromatic shrub Coridothymus capitatus are often
Thompson (2004b) and Y.B. Linhart (University of
encircled by a ring in which annual plants are absent
Colorado, unpublished data), who grew seedlings
(Katz et al. 1987). These authors showed that germi-
of a variety of species which occur in Mediterranean
nation of two annuals (Plantago psyllium and Erucaria
garrigue communities on soils collected in the field,
hispanica) is suppressed both by leachates and the
suggest that monoterpenes leached from the
volatile oil extracted from shoots in laboratory con-
aromatic plant T. vulgaris are indeed ecologically
ditions and when seeds are hand planted into bare
significant, with beneficial and negative effects on
rings around plants of C. capitatus in the field. They
seedling growth, depending on the species and the
also note that plants vary in their capacity to inhibit
monoterpene. The search for the ecological signifi-
the establishment of seedlings of the two annuals.
cance of allelopathy is thus an important issue in the
2. The essential oil and leaf litter of T. vulgaris
Mediterranean, particularly in open garrigue-type
inhibits the germination of an associated species
vegetation where community dynamics may be
Brachypodium phoenicoides, a potentially important
strongly influenced by aridity and nutrient stress
competitor species in garrigue communities where
and competition among low woody and herba-
wild thyme occurs in southern France (Tarayre et al.
ceous vegetation. The task ahead will be to identify
exactly the role of allelopathy in the functioning of
3. Vokou and Margaris (1982) showed a strong
plant–plant interactions in natural situations.
inhibitive effect of high concentrations of the oils
The release of monoterpenes during litter decom-
of Thymus (Coridothymus) capitatus, Satureja thymbra,
position, or their presence in calyces and fruits
Teucrium polium, and Rosmarinus officinalis on two
which enclose the seed, may also influence the
cultivated species but only a weak effect on three
germination of conspecific seeds. In his seminal
wild annual species (Astragalus hamosus, Hymeno-
paper on plant regeneration, Grubb (1977) dis-
carpus circinatus, and Medicago minima) which grow
cussed the idea, and the lack of critical evidence
in the same communities as the species used as a
for it, that regeneration of seedlings beneath their
chemical source.
parents is less than under a comparable degree
4. Vokou et al. (2003) have shown potential strong
of shade due to leaching of chemical compounds.
effects of the monoterpenes present in many
McPherson and Muller (1969) were among the first
Mediterranean Lamiaceae on seedling growth of
to claim a role of allelochemicals in the dormancy
cultivated plants.
of A. fasciculatum seeds in the California chaparral
5. Finally, Robles et al. (1999) have reported poten-
until fires consume the adult plant. More recently,
tial allelopathic and auto-allelopathic effects of
however, Keeley (1991) suggested that the direct
Cistus albidus.
effects of fire (e.g. charred wood and smoke) and
It is known that monoterpenes are soluble and other ecological constraints may be more impor-
that they can be transported to the soil surface by tant determinants of germination. The leaching of
precipitation (e.g. Hyder et al. 2002) but whether volatile oils which inhibit early germination could in
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 149

fact represent an adaptive response to such irregular In some species vegetative tissues may emit
germination cues (Augspurger 1979; Keeley 1991; volatile compounds that attract pollinators. In the
Fox 1995). In a Mediterranean ecosystem the first Mediterranean dwarf palm, Chamaerops humilis
bout of heavy rain in late summer is often followed (Dufaÿ et al. 2003) the leaves can attract a specific pol-
by a subsequent period of drought stress. Germina- linating weevil due to fragrances emitted by fairly
tion may thus be more successful if it is timed to large structures at the sinuses of the palmate leaf.
occur when more consistent autumnal rains occur. Of course the leaves themselves provide no reward
There is in fact diverse evidence for climatically to the insect, which in the process of pollination
adapted germination strategies in Mediterranean also lays its eggs in the rachis of male plants. In
plants? Majorana syriaca (Beker et al. 1989) honey bees react
differently to and discriminate between leaves and
• Mediterranean populations of D. glomerata set inflorescences that vary in their ratio of carvacrol
seed early in the spring but only drop their seeds
and thymol. It is the vegetative parts which emit the
in the fall when rainfall becomes more reliable,
volatile oils which facilitate long-distance attraction
whereas populations in Atlantic climate areas of
to the plant. The close range signal, once a bee is
France set seed later in spring and release their seeds
close to the plant, is then provided by the flower.
in the summer (Knight 1973).
Floral fragrance is easy to appreciate in Narcissus
• Margaris and Vokou (1982) illustrate how, for which are well known for their sweet smell. The
a number of aromatic species in the phrygana of
volatile floral fragrance of nine Narcissus species in
Greece which produce seeds in late spring/early
southern Spain, which cover seven sections of the
summer, seeds remain in the soil and because
genus, includes no less than 84 volatile compounds
they are (p. 456) ‘equipped with after-ripening dor-
(Dobson et al. 1997). Some of these (such as trans-β-
mancy’ seeds are ‘protected from untimely germi-
ocimene) are well known in cultivated forms, others
represent a whole new range of natural product
• In C. capitatus, germination is significantly slower diversity. These authors recognized three main ‘fra-
in the presence of calyces (which are the unit of
grance types’: fatty acid derivatives, isoprenoids,
dispersal and which contain the essential oil) than
and benzenoids (without trans-β-ocimene) or iso-
when seeds are germinated alone, an effect relieved
prenoids and benzenoids (with trans-β-ocimene).
by leaching of the essential oil over time (Thanos
The basic organization of biosynthetic pathways is
et al. 1995). In this species the strong inhibitory
common to all species, hence, fragrance differences
effect of pure oils is lost once seeds are trans-
are most probably due to differences in the regula-
ferred and washed and leaf litter has no effect on
tion of enzyme activity rather than to the presence
the germination of conspecific seeds (Vokou and
or absence of specific enzymes. Hypecoum species in
Margaris 1986).
the eastern Mediterranean show marked variability
• In T. vulgaris the presence of litter or pure oils in their floral emissions (Dahl et al. 1990).
causes a ∼50% inhibition of early seed germination,
In Ophrys orchids, individual plants have a floral
and such effects may gradually wear off (Tarayre
fragrance containing many compounds and sev-
et al. 1995).
eral species show evidence of intra and interspecific
Volatile oils may also play a role in pollinator attrac- variation in fragrance composition. For example,
tion (Williams et al. 1983; Ackerman et al. 1997). For Ophrys lutea has a fragrance with farnesol and geran-
volatile compounds to have an attractive potential iol or a mixture of the two whereas Ophrys fusca
they must be associated with a modification of the is a mixture of farnesolic, cironellalic, and other
reward, otherwise the insect will remain indifferent alcohols (Kullenbeg 1976). In Ophrys orchids pol-
to any fragrance and its variation. Several examples linated by male wasps, Bergstrom (1978) showed
from the Mediterranean flora provide insights how their fragrance is (a) similar to that of female
into the role of fragrance variation in pollination insects in terms of the presence of a certain num-
biology. ber of compounds and (b) associated with the vital

needs of the pollinating bees in terms of their dissipate the excess of absorbed energy without
requirements for mating, feeding, and nesting. In any damage to cell function and it has been sug-
Mediterranean Ophrys the low frequency of pollina- gested that the production and accumulation of
tor visitation to many species is likely to sharpen essential oils could serve this function (Ross and
the interaction between attractive traits and pol- Sombrero 1991; Joffre et al. 1999). During the sum-
linators since traits which favour the chance of mer drought, photosynthates may be shunted into
visitation and also the specificity of the interaction secondary metabolic pathways, minimizing their
should be under strong selection. Indeed, Ayasse accumulation and potential negative effects on the
et al. (2000) have shown that in Ophrys sphegodes photosynthetic system. The fact that different aro-
(a common species in Mediterranean garrigue) bio- matic Lamiaceae in Greece (Vokou and Margaris
logically active components of the floral fragrance 1986) and Israel (Basker and Putievsky 1978) have
show less variation among plants than do non- a significantly higher yield of essential oils in sum-
active compounds, a result which suggests polli- mer than in spring and autumn supports this idea. In
nator mediated selection on the composition of the Ruta graveolens, a pungent smelling herb in dry rocky
bouquet. areas in lowland garrigue-type habitats, the yield
Volatile oils may thus be critical elements of mutu- of furanocoumarins is strikingly higher (10-fold) in
alistic plant–pollinator interactions, both in highly the fruits than in stems, leaves, or roots, indicative
specialized interactions such as those between of a seasonal increase in production during summer
orchids and their pollinators and dwarf palms and (Milesi et al. 2001). This could also represent vari-
their specialist pollinator (see Chapter 5) and in ation which enables enhanced protection of fruits
species where the interaction is less species-specific, and seeds from predators and pests. However, such
such as in Narcissus and various Lamiaceae. studies do not precisely control for seasonal varia-
tion in other traits, such as leaf surface area, which
may decrease in summer (Ross and Sombrero 1991),
4.5.3 Ecological functions of secondary
hence, more controlled investigation will be neces-
compounds: abiotic interactions
sary for any adaptive significance to be unravelled.
The diversity and abundance of aromatic plants Second, glandular trichomes and hairs con-
has stimulated much discussion of the ecological taining volatile oils on the leaf surface may
function of volatile oils in relation to abiotic envi- enhance tolerance of high leaf surface tempera-
ronmental constraints in the Mediterranean region tures and reduce excessive water loss by trap-
(Margaris and Vokou 1982; Ross and Sombrero 1991; ping air which helps buffer high temperature and
Seufert et al. 1995; Thompson et al. 1998; Joffre increases resistance of the laminar boundary, allow-
et al. 1999). This literature illustrates the diversity ing for a reduction in transpirational water loss
of abiotic roles that volatile oil accumulation may (Audus and Cheetham 1940). Volatile oils may also
play in Mediterranean plants. have a cooling effect and initiate stomatal closure
First, the presence and accumulation of essen- and the emission of isoprene and monoterpenes
tial oils may improve tolerance of water constraints may provide thermal protection for leaves subject to
and high solar radiation. In a Mediterranean- high temperature and potential damage of cell and
climate, plants receive high levels of intercepted membrane functions (Sharkey and Singsaas 1995).
solar radiation at times when photosynthetic capac- Third, secondary compounds may be reconverted
ity and growth are limited and carbon assimilation and re-utilized after their release. In Mentha piperata
is restricted by water stress and high temperatures. in California neo-menthyl glucoside produced by
As discussed earlier in this chapter, during the sum- leaves is transported to roots and rhizomes where
mer drought, the absorption of light energy may it is converted into other lipid-like metabolites
be in excess of that required for carbon fixation, (Croteau 1987). According to this author (p. 951),
with potential damage to the photosystem. Differ- several other species provide … ‘compelling
ent physiological regulatory mechanisms may help evidence that monoterpenes can be degraded to
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 151

metabolites that are reutilized in lipid biosynthesis hypothesis relies on the assumption that moderate
in the developing root or rhizome or can conceivably nutrient deficiency limits growth more than photo-
be further oxidized in energy production’. synthesis, leading to an increase in C/N ratio of
Finally, essential oils may play a role in enzyme the plant. Excess carbon is then allocated to the
maintenance during summer when metabolism and production of secondary metabolites. Evidence
growth are depressed. It has been hypothesized for this idea comes from a study of four woody
(Banthorpe et al. 1972) that the biosynthesis of essen- species in different genera in southern France by
tial oils could maintain the appropriate enzyme Glyphis and Puttick (1989). These authors detected
systems in a state which could allow the rapid reacti- a negative relationship between foliar nitrogen
vation of the metabolic system once favourable con- concentration and leaf-phenolics among sites. In
ditions (the onset of which is often unpredictable) sites where plants had a lower phenolic content
for rapid regrowth occur. they also had a lower leaf surface area and lower
Precise experimental tests of these hypotheses concentrations of phosphorus.
remain to be done in order to fully evaluate their The growth-differentiation (G/D) balance hypothesis.
pertinence in a Mediterranean setting. This hypothesis is similar, but slightly more com-
prehensive than the previous hypothesis. It takes as
its basic premise the idea that there is a physiolog-
4.5.4 Secondary compound production: are ical trade-off between growth and differentiation
costs reduced in a Mediterranean context? (e.g. the production of structures necessary for
secondary compound storage). Unlike the C/N
To fully understand the function and adaptive
balance hypothesis, the G/D balance hypothesis
nature of secondary compound production requires
integrates the role of developmental constraints and
a formal understanding of the costs of manufac-
external agents and thus provides a more complete
ture, the benefits of their production, and their
picture of the environmental constraints influencing
multiple roles. The production of secondary com-
the production and defence function of secondary
pounds requires the use of fixed carbon and the
compounds. This hypothesis predicts that more
differentiation of structures for their storage requires
photosynthates are available for the production of
that resources be diverted from growth, into other
carbon-based secondary compounds in nutrient-
functions. This resource allocation may represent
poor environments. However, theoretical models
an important use of carbon and other resources,
which incorporate a feedback effect of plants on
hence the accumulation of secondary compounds
resource supply indicate that the simple prediction
may come at a cost in terms of reduced growth.
(based on resource availability) of high investment
The ‘trade-off’ this implies may have fundamental
in antiherbivore defence in resource-poor envi-
ecological and evolutionary consequences. Indeed,
ronments may not always be valid (Loreau and
as Herms and Mattson (1992) describe, the produc-
Mazancourt 1999).
tion of secondary metabolites in high concentrations
can represent something of a dilemma to plants, and
Although there is evidence in some species for
to grow or to defend is a question that is often posed.
a cost to secondary compound production, trade-
In their in-depth review of the theory of resource
offs in terms of reduced growth due to alloca-
allocation to defence, the above authors outline two
tion to carbon-based defence compounds may not
main hypotheses.
always occur. This may be particularly the case
The carbon/nutrient (C/N) balance hypothesis. This in the Mediterranean where water shortage and
hypothesis predicts that the C/N ratio of a plant nutrient deficiency limit growth. The opportunity
should be positively correlated with the concen- cost (in terms of not being able to grow because
tration of carbon-based secondary compounds resources have been allocated to defensive com-
(e.g. monoterpenes) and inversely correlated with pounds) of secondary metabolism will thus be low
concentrations of nitrogen-based compounds. This (Coley et al. 1985). In Mediterranean habitats plant

growth is often limited by nutrients, hence growth here is that a combination of multiple roles and
may decline more rapidly than photosynthetic activ- an environment where aridity and nutrient stress
ity, creating an excess carbon above the requirements reduce the effective (or opportunity) cost of produc-
for growth. In such conditions woody plants may ing monoterpenes provide a possible explanation for
accumulate carbon-based compounds which lack the abundance of species with volatile oils in the
nitrogen and phosphorous. Such carbon input to sec- Mediterranean flora.
ondary compounds may thus be at little resource
cost because even if more resources were avail-
4.5.5 Variation in chemical composition
able, growth cannot be increased. Such a reduced
among closely related species
resource-based cost to growth (in terms of the use of
fixed carbon for secondary compound production) Closely related plant species in the Mediterranean
is a distinct possibility for Mediterranean plants flora often show striking differences in the blend of
which may continue photosynthesis in the summer monoterpenes they produce. Species of the genus
when environmental constraints prevent growth. Lavandula, aromatic species ‘par excellence’, provide
In addition, some species may shift allocation a convenient place to start. The species cultivated
from one type of defence to another depending on in the high plains of Provence since the end of the
attack rates and the resource cost of accumulat- nineteenth century, Lavandula angustifolia (or true
ing essential oils may be buffered by their multiple lavender), is probably the most well known of any
roles. As outlined by Herms and Mattson (1992), of the different species and commercial varieties. Its
in resource-limited environments, the production rich oil, coveted by the perfume industry for over
and storage of essential oils which act as a deter- 100 years, and its vivid purple bouquet have made
rent to herbivores and parasites may also contribute it an unmistakable emblem of the Provence land-
to the competitive ability of a plant or its toler- scape. This species grows wild above ∼500 m in the
ance of environmental constraints and thus further western Mediterranean, and produces an essential
enhance fitness. In other words, defence may not be oil rich in linalool and linalyle acetate. Below 600 m
as costly as some models for the evolution of chem- elevation, its congener L. latifolia is common on lime-
ical defence, assume. A dual role of reduced her- stone in natural garrigue formations. Its oil is quite
bivory and reduced competition could thus improve different, being composed mostly of 1,8-cineole, bor-
plant fitness in a resource poor Mediterranean envi- neone, camphor, and linalool. At similar elevations
ronment, without incurring major costs. In some on acidic schists and granite, Lavandula stoechas has a
(perhaps many) species, storage organs such as different oil, which is rich in borneol. The ecological
glandular hairs are already well developed and and geographic segregation of these species, as well
abundant on very young seedlings (references in as some hybridization (to produce the high-yielding
Herms and Mattson 1992). Indeed in T. vulgaris it sterile lavindin hybrid), and their chemical diversity,
suffices to touch the first leaves as they open to real- make them a promising model for the study of plant
ize that the glandular hairs are rapidly filled with evolution, a work which would surely interest the
monoterpenes. Such precocious differentiation of flourishing lavender industry.
storage sites suggests the importance of essential Less well known, but by no means less signifi-
oils in the defence of very young seedlings and illus- cant, is the variation among Mediterranean oak
trates that growth and differentiation can occur both species in the blend of terpenes they emit; some
rapidly and simultaneously. species are predominantly isoprene emitters, others
Finally, to my knowledge, the exact nature of the predominantly monoterpene emitters (Loreto et al.
relation among resource availability, cell differenti- 1998; Cisky and Seufert 1999). Instead of being
ation related to secondary compound production, stored in specialized secretory structures, as is the
and plant growth, and thus the true nature of the case for most aromatic oils, or induced when plants
cost of such compounds, has not been precisely iden- are damaged, the volatile organics emitted by oak
tified in any Mediterranean plant. What I suggest trees are constitutively produced but not stored.
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 153

Synthesized from recently fixed carbon, they are contain either high amounts of thymol and small
almost immediately emitted at rates dependent concentrations of carvacrol or vice versa (Ravid and
on light and temperature conditions. Although Putievski 1983).
Loreto et al. (1998) suggest that the variation among
species may be of little ecological and adaptive Quantitative variation in the monoterpene com-
significance, it being an ancestral character that position of essential oils may be common in many
may have served multiple functions, further work Mediterranean aromatic species, as the following
would be worthwhile on this issue. examples illustrate.
Finally, the composition of the essential oil may
be a distinguishing feature of endemic species.
For example, Ruta corsica, endemic to Corsica 1. In Q. ilex, a sample of 146 trees from natural
and Sardinia has an essential oil characterized sources in southern France revealed the existence
by a predominance of alkyl acetates whereas of three main chemical profiles based on five main
congeneric and more widespread species in the monoterpenes; α-pinene, β-pinene, sabinene,
western Mediterranean have high concentrations limonene, and myrcene (Staudt et al. 2001).
of 2-ketones (Bertrand et al. 2003). All plants emitted at least a small amount of each of
these molecules, but in distinctly different propor-
tions. Three groups can be identified: a pinene type,
4.5.6 Variation in chemical composition dominated by two types of pinene (with sabine and
within species myrcene as secondary components and limonene as
a minor component) occurred in 71% of sampled
Variation in secondary metabolite formation can trees; 31% of the trees were of the limonene type;
also occur within species, either as qualitative varia- and 8% emitted a monoterpene blend dominated
tion, which allows the recognition of two or more by myrcene. Lack of environmental or seasonal
forms (or ‘chemotypes’) or as quantitative vari- variation suggests that such variation is genetically
ation when individuals produce a different blend based in Q. ilex.
of similar monoterpenes. 2. In P. lentiscus on Corsica, limonene and pinene
The presence of qualitative intraspecific vari- types have been described with similar levels of
ation in chemical composition, which suggests relative abundance in sampled trees (Castola et al.
the occurrence of distinct chemotypes in natural 2000).
populations, has been documented in several 3. In the widespread P. halepensis, resin composi-
groups of Mediterranean plants. tion is highly variable among geographic localities
(Schiller and Grunwald 1987).
• Different species of Thymus (Stahl-Biskup 2002;
Table 4.2), notably the six chemotypes of T. vulgaris 4. In R. officinalis, the major monoterpene is variable,
in southern France (Thompson 2002) which I discuss with plants having an oil dominated by one of four
in detail below. main molecules—camphor, eucalyptol, verbenone,
or α-pinene (Granger et al. 1973).
• The common mint, Mentha spicata which has four
distinct chemotypes in Greece (Kokkini and Vokou 5. The two different varieties of Cistus ladanifer in
1989). southern France, var albiflorus and var. maculata,
show significant variation in the blend of monoter-
• Withania somnifera which has been reported to
have three chemotypes in Israel (Abraham et al. penes in their essential oil (Robles et al. 2003).
• Origanum vulgare subsp. hirtum which has plants The first step towards understanding the evolution-
whose oil is dominated by either thymol or carvacrol ary significance of intraspecific variation is to docu-
in Greece (Vokou et al. 1993). ment its genetic basis. In T. vulgaris six well-defined
• Individual plants of Coridothymus capitatus chemotypes have been identified in southern France
and Majorana syriaca in the eastern Mediterranean (Granger and Passet 1973; Vernet et al. 1977a,b, 1986).

Table 4.2 Dominant monoterpenes and potential chemotype differentiation (based on a review by Stahl-Biskup
2002) in different species of the genus Thymus in the western Mediterranean

Section Species G A B U 18C CA L C T

Mastichina T. albicans ∗ ∗

T. mastichina ∗ ∗

Micantes T. caespititius ∗ ∗ ∗

Piperella T. piperella ∗ ∗

Pseudothymbra T. funkii ∗

T. moroderi ∗ ∗

T. membranaceus ∗ ∗

T. longiflorus ∗ ∗ ∗

T. villosus ∗ ∗ ∗

T. antoninae ∗

Thymus T. camphoratus ∗ ∗

T. carnosus ∗ ∗ ∗

T. hyemalis ∗ ∗ ∗

T. orospedanus ∗ ∗ ∗

T. baeticus ∗ ∗ ∗ ∗ ∗ ∗ ∗

T. vulgaris ∗ ∗ ∗ ∗ ∗ ∗ ∗

T. loscosii ∗

T. zygis
subsp. gracilis ∗ ∗ ∗ ∗

subsp. sylvestris ∗ ∗ ∗ ∗ ∗ ∗

subsp. zygis ∗ ∗ ∗

T. serpylloides ∗ ∗ ∗ ∗ ∗ ∗

T. hirtus ∗

Hyphodromi T. bracteatus ∗

T. leptophyllus ∗

Serpyllum T. wilkomii ∗ ∗ ∗

T. pulegioides ∗

T. nitens ∗ ∗ ∗ ∗

T. herba-barona ∗ ∗

Notes: G: geraniol, A: α-terpineol, B: borneol, U: thuyanol, 18C: 1,8-cineole, CA: camphor, L: linalool, C: carvacrol, T:

These chemotypes can be distinguished on the basis will become apparent later. A large series of crossing
of their dominant monoterpene (present in glandu- experiments and progeny analysis first identified
lar trichomes on the leaves, stems, and calyces), that the underlying genetic control involves 5 (or
that is, geraniol (G), α-terpineol (A), thuyanol-4 6) loci (Vernet et al. 1986). A plant with a dominant
with terpinen-4-ol (U), linalool (L), carvacrol (C), G allele at one or other of two loci will have the
and thymol (T). The six monoterpenes are all pro- geraniol phenotype, regardless of whether it has
duced from geranyl pyrophosphate in a series of dominant or recessive alleles at the other loci. If
changes in configuration and hydroxylation and the plant is homozygous recessive at the G loci and
thus have fairly similar molecular structures, but has at least one dominant A allele then it will have
with a major distinction between phenolic and non- the α-terpineol phenotype. The sequence contin-
phenolic chemotypes (Box 4.10) whose importance ues in the order (G, A, U, L, C) until all loci are
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 155

Box 4.10 Biosynthesis, dominant molecule and genotype of the six chemotypes of
Thymus vulgaris in southern France

pyrophosphate G./../../../..



Terpinyl-8 gg/A./../.../.


Neryl- gg/aa/Uu/../..
pyrophosphate Terpinyl-4


λ-Terpinene OH




Plants with a phenolic chemotype have the monoterpene, except the thuyanol chemotype
characteristic thyme odour which makes them whose oil is dominated by a combination of
readily distinguishable from the four non-phenolic primarily thuyanol-4 and terpinene-4-ol and also
types. Another chemotype easy to distinguish in contains non-negligable amounts of myrcenol-8
the field is geraniol, which has a lemon fragrance. and linalool (Granger and Passet 1973; Thompson
All these chemotypes have a single dominant et al. 2003b).

homozygous recessive, which produces the thymol

% dominant monoterpene in oil

H—Home sites
A—Away sites
This mode of inheritance may be slightly more 80
complicated than just a series of epistatic inter-
actions (Thompson et al. 2003b). These authors 60

obtained samples of each chemotype from two dif-

ferent population contexts: (a) where a chemotype
is the most abundant chemotype (‘home’ sites) and
(b) populations where the same chemotype is rare H A H A H A H A H A
Geraniol -terpineol Linalool Carvacrol Thymol
or occurs as part of mixed-chemotype populations
(‘away’ sites). For five chemotypes, plants sampled
Figure 4.4 Mean percentage (±SE) of the dominant monoterpene
at away sites had a significantly lower proportion of
in the essential oil of five T. vulgaris chemotypes in southern France
their dominant monoterpene in the oil than plants at ‘home’ sites where that chemotype is the majority type (closed bars)
at a home site (Fig. 4.4). For nearly all plants at and ‘away’ sites where that chemotype is rare or in a
‘away’ sites, the decline in proportion of the charac- mixed-chemotype population (open bars). (Reproduced with
teristic dominant monoterpene is correlated with a permission from Thompson et al. 2003b).
significant increase in a secondary component of the
oil, which is the monoterpene present in the locally chemotype. In fact, rather than being due to epista-
abundant chemotype. So, when linalool plants were sis, the genetic control of monoterpene production
sampled in populations dominated by the thuyanol may be a simple consequence of gene dosage effects
chemotype, the decrease in linalool in their oil was along a chain of loci. Under this hypothesis, one
accompanied by an increase in thuyanol, whereas if would expect an increases in the production of
carvacrol was predominant then the linalool plants linalool across a range of genotypes of the geran-
had a high proportion of carvacrol. A similar pat- iol chemotype: from GG/ll (no linalool), to GG/Ll,
tern was observed for geraniol plants, which had Gg/Ll, and Gg/LL. Modifier genes could also allow
a high proportion of linalool in their oil at sites for the production of secondary monoterpenes.
where the linalool chemotype was locally abun- Distinguishing the precise cause will necessitate
dant, for α-terpineol plants in a population dom- more detailed genetic investigations of secondary
inated by the thuyanol and linalool chemotypes, compound production in this species.
and carvacrol plants in a population dominated In M. spicata in Greece (Kokkini and Vokou
by the thymol chemotype. The thymol chemotype 1989) four chemotypes occur; one dominated by
showed in almost all cases a decline in thymol asso- linalool (65–75% of the essential oil), one with
ciated with an increase in the composition of its variable amounts of carvone and di-hydrocarvone,
precursors, and thus not the presence of one of one with either piperitone oxide or piperiteonene
the other characteristic monoterpenes. The latter oxide depending on the site where plants are
result is not surprising since thymol is at the end of sampled, and one with a blend of menthone
the chain. (18–45%), isomenthone (3–15%), and which may
The decline in the percentage composition of the or may not have some pulegone (up to 31%).
dominant monoterpene was thus associated with a These types fit nicely onto the probably pathway
significant increase in the presence of a monoterpene of monoterpene synthesis, which would produce a
produced by a dominant allele at a locus that would first branch to linalool, a second branch to carvone
normally not be expressed in the chemotype (e.g. (and di-hydrocarvone), and a subsequent series
an L allele in a geraniol plant). The presence of a through piperitonene (and piperiteonene), pule-
‘secondary’ dominant gene may thus not be com- gone and finally menthone (Murray and Lincoln
pletely switched off by the dominant gene prior 1970). The chemotypic segregation of interlinked
in the chain, allowing for a limited production compounds in a chain of synthesis with branches
of a secondary monoterpene by plants of a given strongly suggests that dominant genes regulate
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 157

(a) (b)

100 km 100 km

Figure 4.5 Geographic variation in the percentage of carvacrol (black sector) and thymol (dotted sector) in the essential oil of populations (each
circle represents a population) of (a) O. vulgare subsp. hirtum and (b) O. onites (drawn from data in Vokou et al. 1988, 1993). The lightly
shaded and open sectors of each circle represent precursors and their compounds in the oil.

enzyme activity at points in the chain which sup- site, and thus favoured in warmer climates. In
press activity further down the chain. In addition, addition the southern and eastern populations,
several species show spatial variation in chemotype that is, those with the highest thermal efficiency
abundance, inviting investigation of the adaptive (milder winters and drier hot summers) tend to
significance of variation in secondary compound produce an oil composed of a relatively higher total
production. content of phenolic monoterpenes and tend to be
The genus Origanum provides an indication pure carvacrol (Fig. 4.5). The potential influence of
that local climate is closely related to population climate on this pattern is supported by work on
variation in monoterpene production. In a study of other closely related taxa:
23 populations of Origanum vulgare subsp. hirtum
across Greece and onto the Aegean islands, Vokou 1. In O. vulgare subsp. vulgare and subsp. viridulum,
et al. (1993) found that the local climate, in terms which have more northerly distributions, essential
of thermal efficiency, had a significant effect on oil yield is lower than in subsp. hirtum (Kokkini et al.
essential oil yield, which is higher at low altitude 1991).
and decreases in cooler sites and at high elevation. 2. In Origanum onites, which has a more southerly
Although there is no consistent effect on compo- and easterly distribution, all plants have carvacrol
sition, the sum of the two phenolic monoterpenes, as the dominant component of their oil (Vokou et al.
and even the sum of the four compounds in the 1988; Fig. 4.5).
phenolic pathway (thymol, carvacrol, and their 3. In Portugal, the phenolic content of the oil of
two precursors para-cymene and λ-terpenene) O. vulgare increases in a southerly direction (Carmo
were positively related to thermal efficiency at a et al. 1989).

Camphor Eucalyptol




Figure 4.6 Geographic variation in the principal chemotype of R. officinalis in the western Mediterranean. Redrawn from Granger
et al. (1973).

A study of 509 populations of 109 different aro- Spain, where they produce carpets of pink flowers in
matic labiates taxa in the flora of Greece by Kokkini April and May and shrivel to produce greyish small
et al. (1989) further supports this pattern. These bushes in late August. A true Mediterranean plant,
authors identified three categories of essential oil if ever there was one. Several species in different
production: oil-rich taxa (>2 ml/100 g dry weight) sections of the genus show evidence of chemotype
restricted to low elevation (<300 m); low-yielding variation (Table 4.2). For example in T. zygis three
taxa (<0.5%) at all elevations (sea level to 2,000 m); allopatric subspecies show chemical variation in
and intermediate taxa which show a gradual decline relation to latitude: the most southern subsp. gra-
in frequency of populations with altitude. Within cilis has thymol, carvacrol, and linalool chemotypes;
the high-yielding taxa there is a gradual decline in the more widespread subsp. sylvestris in central
yield with altitude. There is thus a dominant trend Spain has thymol, carvacrol, linalool, α-terpineol,
for oil-rich plants, particularly those with pheno- 1,8-cineole, and geraniol chemotypes; and the
lic monoterpenes, to occur in the hotter and drier most northern subsp. zygis has carvacrol, linalool,
Mediterranean communities. α-terpineol, and geraniol chemotypes. The trend is
In Rosmarinus officinalis the relative propor- thus towards non-phenolic oils in more temperate
tion of one of 4–5 major monoterpenes—all of sites with colder winters. A similar spatial pattern
which occur in most plants—varies geographically can be observed in T. vulgaris in southern France,
(Granger et al. 1973). In Spain and southern France, where phenolic chemotypes (carvacrol and thymol)
the oil is rich in camphor, in the eastern and southern dominate populations in hot dry sites close to the
(i.e. hotter and drier) parts of the range, eucalyptol Mediterranean Sea and non-phenolic chemotypes
is the dominant component of the oil, in the mid- (geraniol, α-terpineol, thuyanol-4 and linalool) are
dle of the western Mediterranean and on Corsica, abundant further inland, particularly above 400 m
the oil is almost exclusively composed of verbenone elevation, that is, in wetter, cooler climates (Granger
(Fig. 4.6). Unfortunately, I know of no study which and Passet 1973; Thompson 2002).
has explored the ecological and genetic basis to this In T. vulgaris a highly localized pattern of spatial
pattern of differentiation. differentiation occurs in and around the St Martin-
Chemotype differentiation is probably a general de-Londres Basin in southern France (Plate 2). Inside
feature of adaptive genetic diversity in western the basin a temperature inversion during winter
Mediterranean Thymus. Species of thyme are one of causes the accumulation of cold moist air against
the most characteristic plants of the open garrigue the imposing north face of the Pic St Loup. In win-
vegetation of southern France and the tomillares of ter, temperatures are often several degrees lower
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 159

(extreme freezing temperatures may reach −20˚C in (a) 180

Number of populations
the basin) than in the hills that surround the basin 150
where soils are more stony, shallower and drier, and 120
different in terms of carbon and nitrogen mineral- 90
ization when compared with those inside the basin 60
(Billès et al. 1971, 1975; Gouyon et al. 1986). Cold win- 30
ter temperatures probably prevent the colonization 0
1 2 3 4 5 6
of the basin by P. halepensis, and differences in soil Number of chemotypes/population
create a mosaic for the distribution of deciduous
(b) 80
oaks on deeper soils and evergreen oaks in more
rocky areas (see photograph in Box 4.11). This cli-

Number of populations
with two chemotypes
matic and soil gradient, which occurs on a localized 60

mosaic of geological variation (see Introduction) has 50

been found to be correlated with differences in the 40
distribution of different chemotypes over a very 30
short distance. Whereas phenolic chemotypes pre- 20
dominate over large areas around the rims of the 10
basin above 250 m elevation, below 200 m elevation 0
inside the basin, where populations are often frag- 2 Non- 1 Phenolic + 2 Phenolic
phenolic 1 non-phenolic
mented by agricultural land-use, there is a mosaic
of smaller thyme populations where non-phenolic
Figure 4.7 Chemotype diversity in natural populations of
chemotypes are most abundant (Plate 2). T. vulgaris in southern France: (a) the number of populations with
Since geraniol is produced by the dominant allele different numbers of chemotypes in an overall sample of ∼330
of the locus at the start of the chain, and is thus populations; (b) the number of populations with two chemotypes that
the only gene that cannot ‘hide’ behind other genes have either two phenolic chemotypes, two non-phenolic chemotypes,
which cause a particular phenotype to be expressed, or one phenolic and one non-phenolic chemotype. Reproduced with
permission from Thompson (2002) and based on original data
the dominant G allele may be more prone to loss collected by Vernet et al. (1997a,b) and Gouyon et al. (1986).
during episodes of colonization and extinction. This
chemotype is the rarest chemotype in the region.
Hence, some of the variation in frequency among
populations could have a stochastic cause due to and non-phenolic chemotypes (i.e. a combination
random loss of genes and fixation of others. This is of those with 2, 3, 4, or 5 chemotypes) are rarer
however, hardly likely to cause the striking spatial than expected (Fig. 4.7). Most populations which
segregation of phenolic and non-phenolic chemo- contain appreciable percentages of both phenolic
types in different ecological conditions. Indeed, and non-phenolic chemotypes occur in the eleva-
there is growing evidence for the role of selection tion transition between 200 m and 250 m (Plate 2).
in the maintenance of spatial chemotype differenti- The sharp spatial separation of phenolic and non-
ation in this region. phenolic chemotypes, despite the large number of
First, in 336 populations analysed by Gouyon mixed chemotype populations, is strong evidence
et al. (1986) ∼20% contained only a single chemotype that natural selection eliminates one or the other,
and 50% contained a mixture of two chemotypes, depending on the spatial location of the site.
another 20% had three chemotypes, and few popu- Second, the comparison of isozyme variation with
lations had more than three chemotypes (Fig. 4.7). chemotype variation indicates that allele frequencies
Although populations with two chemotypes are based on isozymes show little genetic differentia-
the most common, those with mixtures of phe- tion relative to those coding for the monoterpenes
nolic and non-phenolic chemotypes are rare and, (Box 4.11). Most pollination of thyme is by honey
in general, populations containing both phenolic bees which fly mostly between adjacent plants

Box 4.11 The frequency of chemotypes, and isozyme variation at two loci (PGM-2 and GOT-1)
along a transect in southern France across the St Martin-de-Londres Basin

In the basin, left of dotted line, non-phenolic chemotypes predominate, whereas over the shoulder of the
Pic-St Loup and onto its south-facing slope only phenolic chemotypes occur. Isozyme data from Tarayre
and Thompson (1997), chemotype data from Thompson et al. (2003b). Each circle represents the
frequency of alleles or chemotypes in a given population.

Non-Phenolic populations Phenolic populations




500 m

Pic St Loup

200 m
150 m

(Brabant et al., 1980) but can transport pollen over Third, transplant experiments I have done with
longer distances and even between populations several colleagues have given us a closer insight
(B. Vaissière, INRA Avignon, personal communi- into the nature of local chemotype adaptation in this
cation). In addition several butterfly species which region.
travel long distances between plants create a poten-
tial for pollen flow among populations. Tarayre 1. In seedlings transplanted among eight pairs of
et al. (1997) show that gene flow for pollen is much phenolic and non-phenolic populations recipro-
greater than that via seed, attesting to at least occa- cally transplanted in and around the St Martin-de-
sional among-population pollinator movements. In Londres Basin (Plate 2), seedlings from non-phenolic
the face of such pollen movement, selection on sites show a marked reduction in survival compared
chemotype genes must be strong. to seedlings of phenolic parents in sites where the
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 161

latter naturally occur. This reduced survival appears (a) 800 (b) 100
to be due to poor resistance of soil conditions. In the 600 90

% Survival
Plant size
non-phenolic sites, there is a tendency for seedlings
400 80
from the carvacrol population to have reduced sur-
vival compared to non-phenolic seedlings. 200 70
2. Transplantation of replicated cuttings from 15 0 60
clones of each of the six chemotypes into an N1 N2 P1 P2 N1 N2 P1 P2

experimental garden in Montpellier (where pheno- (c) 900

lic chemotypes—primarily carvacrol—would nat-
urally be the most dominant form) and near the
village of Mévouillon at 800 m elevation in the 700

Plant size
hills north of the Mt Ventoux in Provence (where 600
only non-phenolic chemotypes are present in natu- 500
ral populations) show that the survival and biomass
of phenolic and non-phenolic chemotypes is maxi-
mized in their original environment. 300
N1 N2 P1 P2
3. In a study of the second generation offspring from
two phenolic and two non-phenolic populations
Figure 4.8 Performance of Thymus vulgaris offspring grown in a
grown in the experimental garden in Montpellier, common garden at a site where phenolic, primarily carvacrol,
Thompson et al. (2004) have shown (Fig. 4.8) that chemotypes would normally be present: (a) plant size at reproduction,
size and survival of offspring from maternal plants (b) four-year survival rates in maternal offspring from two non-phenolic
from phenolic populations, particularly those from (N1 and N2, hatched bars) and two phenolic (P1 and P2, filled bars)
populations, (c) size at reproduction in maternal offspring (as above)
the nearby carvacrol population, are significantly
with pollen donors from phenolic (filled bars) or non-phenolic
greater than those from maternal plants sampled (hatched bars) sites (redrawn from Thompson et al. 2004).
in non-phenolic populations. An interesting twist to
this study is the finding that vigour depended not
only on the maternal origin of the offspring but also
on the origin of the pollen donor: offspring size being chemotypes. Likewise, Pomente (1987) reported
significantly greater if the pollen donor originated that seedlings of phenolic (thymol and carvacrol)
from a local phenolic population (Fig. 4.8). plants had a better tolerance of drought stress than
those of non-phenolic plants. Phenolic types may
Although these studies provide convincing evid- thus be better adapted to drier soils. Investigations
ence for monoterpene-mediated local adaptation are currently underway comparing the growth and
and that natural selection contributes to the pat- survival of different chemotypes on the different
tern of spatial chemotype differentiation in thyme, soils collected from different sites where they occur.
they do not identify the precise ecological cause(s) Phenolic chemotypes, particularly carvacrol, are
of population differentiation. Based on the distribu- absent from sites which experience extreme sub-
tion pattern one would predict that soil and localized freezing temperatures. Although Varinard (1983)
climatic variation are two main factors that either found no effect of freezing on the survival of
directly or indirectly (perhaps because of correlated seedlings of the carvacrol, thymol, and thuyanol
variation in herbivores and parasites) contribute to chemotypes in controlled conditions, Amiot et al.
spatial differentiation. There is some evidence that (unpublished manuscript) have found that carvacrol
this may be the case. and thymol plants are more sensitive to freez-
In controlled conditions Couvet (1982) found that ing below −10◦ C than non-phenolic plants, par-
the non-phenolic α-termineol chemotype is signif- ticularly early in winter. Transplanted out of the
icantly less resistant to drought and hot tempera- Mediterranean into very cold winter climates, the
ture stress than the carvacrol, thymol, and linalool germination and survival of offspring from phenolic

plants is significantly less than those from non- Observations of the occurrence of galls induced
phenolic chemotypes (Y.B. Linhart, University of by oviposition and larval development in the area
Colorado, unpublished data). Phenolic chemotypes around the St Martin-de-Londres Basin have shown
of thyme may be less resistant to freezing, perhaps that rates of infestation show spatial aggregation in
due to a greater toxicity of the phenolic molecules relation to chemotype (Box 4.12). Gall-formation is
which, following freezing and the rupture of cell rare on plants in populations dominated by the phe-
membranes, cause mortality during harsh winters. nolic chemotypes and higher in non-phenolic popu-
Why plants with a carvacrol phenotype may be less lations, with a striking peak in populations where
resistant to colder sites than those with a thymol the geraniol chemotype is abundant. In a mixed
chemotype is not known. The ecological segregation population of geraniol and linalool, infestation rates
of carvacrol and thymol may in fact be related to a were significantly higher on geraniol plants. It could
better tolerance of heat and/or drought in the latter, thus be that the rarity of the geraniol chemotype in
since I have found that survival of transplants of the the region where this work has been done (Plate 2) is
carvacrol chemotype has exceeded not only that of at least in part due to its susceptibility to such para-
non-phenolic transplants but also those of the thy- sitism. Where galls develop, they incorporate a large
mol chemotype in very hot and dry sites during the proportion of leaf biomass on a stem and thus may
prolonged and very hot summer of 2003. Climate have a fitness cost in terms of resource acquisition
effects appear to exert a strong selective pressure on and reproduction. This is currently being inves-
genetic variation in this species. tigated alongside the issue of whether adults can
Spatial differentiation of thyme chemotypes may choose oviposition sites (chemotypes) or whether
also be closely related to herbivory and parasite chemotypes differ in their toxicity for larval devel-
attack, as illustrated by several studies in con- opment. Examples of genetic variation in resistance
trolled conditions (Gouyon et al. 1983; Linhart and to gall-forming insects has been documented
Thompson 1995, 1999). Linhart and Thompson elsewhere, for example, on Salix (Fritz 1990) and
(1995) showed that snails (Helix aspersa) have a Solidago (Horner and Abrahamson 1992). The case
preference for non-phenolics, particularly the plants of thyme and its specialized fly may well provide
with a linalool chemotype, and a marked distaste another intriguing model system for this theme.
for the two phenolics. What is more, snails fed on Studies of a range of different potential herbivores
a diet of exclusively carvacrol plants lose weight. and different microorganisms show that variation
An interesting twist to this tale is the possibility that in chemical defence among chemotypes depends on
when plants with a genotype that should produce a the external agent and no one chemotype provides
linalool phenotype are at the young seedling stage, the best defence across the spectrum of potential
their leaves may not have a linalool phenotype but a herbivores and pathogens, although in general the
phenolic phenotype, only developing their ‘correct’ phenolic chemotypes do tend to be more deter-
phenotype after the young seedling stage. Linhart rent than non-phenolic chemotypes (Linhart and
and Thompson (1995) suggest that such develop- Thompson 1999). In addition to their resistance to
mental plasticity could allow the genotype which is snail herbivory and parasite attack phenolic chemo-
the least repellent to snails (linalool) to ‘hide’ behind types of T. vulgaris have a stronger antibacterial
a less palatable phenotype (thymol or carvacrol) effect than non-phenolic chemotypes (Simeon de
during early seedling development. Bouchberg et al. 1976), a pattern which can be seen in
Attempts to observe fitness effects of such comparisons of different species which have either
snail herbivory in the field have however been phenolic or non-phenolic chemotypes (Tantaoui-
inconclusive, even on small transplants, defying Elaraki et al. 1993). However due to variation in
further interpretation of the above studies. In deterrence of the spectrum of potential antagonis-
contrast, the occurrence of a specialized parasite, tic organisms, any spatio-temporal variation in the
the gall-forming Dipteran fly Janetiella thymicola, abundance of different herbivores, parasites, and
can be frequently observed in the wild (Box 4.12). pathogens could contribute to spatial variation in
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 163

Box 4.12 Presence and abundance of galls on Thymus vulgaris induced by oviposition of
Janetiella thymicola (data kindly provided by J. Amiot)

(a) & (b) Percentage of infected plants per

population and a mixed G/L population
(c) & (d) Mean number of galls per plant (only
infected plants) per population and in a mixed G/L


Dominant chemotypes in each population are: G,

geraniol; A, α-terpineol; L, linalool; U, thuyanol; C,
carvacrol; T, thymol.

(a) 25
% of infected plants

20 (b) 20
15 10


(c) 5 (d) 6
Number of galls per plant

4 4
3 2
2 0


the relative abundance of chemotypes and the main- In controlled conditions in an experimental garden
tenance of the chemotype polymorphism (Linhart Pomente (1987) reported that the thuyanol chemo-
and Thompson 1999). type is particularly favoured in humid conditions,
These different facets of the biotic environment and that the presence of an associated grass was
may interact with spatial variation in the abiotic correlated with a decrease in the survival of thyme
environment to influence chemotype abundance. plants in conditions of drought stress. This effect of

grass presence was not an effect of competition, but Leaf litter may thus represent not only a critical com-
rather because the grass maintained a more humid ponent of mineral nutrient cycling, which affects
environment in which slugs sheltered and subse- the composition and function of local communi-
quently caused a greater mortality of thyme plants. ties and their successional development, but also
The thuyanol chemotype grew best in humid con- a component of spatial heterogeneity of the local
ditions and is also the chemotype most deterrent to environment, and thus a cause of divergent selection
slugs (Gouyon et al. 1983). The response to differ- pressures.
ent ecological factors may thus go hand in hand, Spatial heterorgeneity in the soil environment
complicating our efforts to understand causal rela- may thus develop as a result of genetic differ-
tionships in the maintenance of this complex genetic entiation among populations of common species.
polymorphism. To understand how species interactions shape the
Chemotype variation in thyme may also influ- balance between facilitation and competition in
ence interactions with co-occurring plant species. Mediterranean garrigue communities may thus
Experimental work in controlled conditions shows require an appreciation of both the processes which
how the litter and soils under different chemotypes shape adaptive variation at the population level
can affect the regeneration and growth of associ- alongside those which regulate species interactions.
ated species. Phenolic leaf litter or soils from under To understand the dynamics of biodiversity in such
phenolic plants of T. vulgaris and other species sig- systems requires a close link between population
nificantly reduce germination and growth (Vokou and community ecology.
and Margaris 1982; Tarayre et al. 1995; Karamanoli
et al. 2000; Ehlers and Thompson 2004b) relative
4.5.7 More than just essential oils
to those from non-phenolic plants. An interesting
question raised by some of the work by Despina I have focused this section on a small subset of the
Vokou and her colleagues working in the phry- arsenal of chemical compounds that plants contain
gana of Greece is whether associated species may and employ for diverse purposes. Such roles are
show tolerance to the presence of the compounds in of course far from being limited to essential oils.
the soil. Ehlers and Thompson (2004b) have shown Closely related Mediterranean species can also dif-
(a) that Bromus erectus seedlings and adult plants fer significantly in secondary compound profile for
have greater biomass on soils collected from thyme other types of compound. To finish this chapter, let
patches dominated by non-phenolic chemotyes than us consider two examples.
on soil collected from patches of phenolic chemo- First, western Mediterranean species of Ruta
types and (b) that seedling and adult biomass of which show variation in essential oil composi-
B. erectus plants from non-phenolic thyme sites are tion (see above) also show marked differences in
on average greater on soil from thyme patches in the the relative composition of four furanocoumarins
original non-phenolic site than brome from a phe- (Milesi et al. 2001). Whereas psoralen (46%) and
nolic site. The reciprocal is also true on soil from xanthotoxin (43%) are predominant in Ruta angus-
phenolic sites. This pattern of ‘home site advantage’ tifolia, bergapten (67%) is the main compound in
is detected on soil from patches of thyme but not Ruta chalepensis and xanthotoxin is the single most
on soil from the same sites but in patches with no important furanocoumarin in Ruta montana. In Ruta
thyme plants, indicative that the growth response of graveolens, the species with the highest total yield
B. erectus to topsoil heterogeneity is closely related of furanocoumarins, a blend of three main fura-
to modifications associated with the local thyme nocoumarins was detected—bergapten (40%), pso-
chemotype. Species in communities where aromatic ralen (33%), and xanthotoxin (25%). The high yield
plants are abundant may thus adapt to high con- and blend of different compounds in R. grave-
centrations of secondary metabolites in the soil or olens are no doubt associated with the use of this
other modifications of the soil environment linked to species in aroma-therapy, to clean bee hives (when
the decomposition of litter rich in such compounds. used with thyme and rosemary) and in traditional
T R A I T VA R I AT I O N , A D A P TAT I O N A N D D I S P E R S A L 165

medicine (San Miguel 2003). Aqueous extracts from appreciation of how ecological variation creates dif-
R. graveolens have a significant fungistatic activ- ferences in selection pressures in different habitats
ity (Oliva et al. 1999) and can inhibit seed ger- and how the dramatic contemporary changes in
mination of cultivated and weedy species (Aliotta the spatial configuration of these habitats associated
et al. 1994). The elegant work by Berenbaum et al. with human activities modifies the spatial pattern
(1991) illustrating the adaptive significance of fura- of seed dispersal and thus the potential for genetic
nocoumarin variation, suggests that the natural differentiation.
substances in Mediterranean Ruta may play a sig- That plants have not had time to adapt to the
nificant ecological role and may have contributed Mediterranean climatic regime is to my mind not a
to the diversification of different (morphologically generally acceptable idea. When selection is strong,
similar) species in different parts of the range of this plants evolve rapidly and there are many examples
group. The functional and evolutionary significance of highly localized genetic differentiation within
of secondary compound variation in this genus thus plant populations, even long-lived trees (Linhart
awaits attention. and Grant 1996). Of course, many phenological
Second, Ferula communis shows chemotype dif- and growth strategies such as sclerophylly repre-
ferentiation in terms of its hydroxycoumarin com- sent nothing more than the ‘ghost of selection past’,
position, with one chemotype being highly toxic or ‘ecological phantoms’ since they clearly evolved
to livestock in Morocco and on Sardinia. On the prior to the onset of the Mediterranean climate
island of Sardinia, Marchi et al. (2003) have reported (Herrera 1992a). In this context, Grove and Rackham
that genetic differentiation for enzyme loci is sig- (2001: 45) argue that …
nificantly greater among populations of the two
chemotypes than among populations of geograph- European ecology is dominated by environmental change
ically distant populations of the same chemotype. rather than evolution. Most environments have not existed
This pattern of gene flow suggests reproductive long in evolutionary terms. Plants, except perhaps for
isolation among chemotypes which may facilitate annuals which breed every year, have had to make the
chemotype differentiation on the island. best of environments into which accidents of climatic and
geological history have thrust them, rather than becoming
adapted to some specific environmental niche. … Plants
4.6 Conclusions have lived with the present climate for barely longer than
they have lived with major human activity. They have yet
The Mediterranean mosaic is changing, and so are
to become fully adapted to it.
the plant populations it contains. Of prime import-
ance here is the way human activities currently
modify and structure the Mediterranean mosaic. In this chapter I have attempted to balance this
Such effects are ongoing and require consideration argument with illustrations of how diverse func-
and integration, not just as an external agent, but tional attributes and phenological strategies vary in
as a key ecological factor influencing patterns of relation to the complexity of the climatic constraints
dispersal and colonization and the functioning of and selection pressures acting in Mediterranean
natural plant populations. The evidence abounds habitats. Most of the evidence for adaptive vari-
for a strong impact of spatial habitat heterogene- ation concerns geographic and regional patterns
ity on patterns of dispersal and regeneration, two of trait variation and variability among closely
critical events for the development of spatial dif- related sclerophyllous species in the precise mech-
ferentiation among populations. More work which anisms they use to resist water stress. Although
integrates both demographic, dispersal and genetic limited in extent, this evidence can be used as
data within the context of ecological habitat hetero- a basis for manipulative experimental investiga-
geneity would be particularly worthwhile here. The tion of intraspecific variability in traits related to
message here is that to understand plant evolution aridity, nutrients, and other features of the land-
in the contemporary Mediterranean requires a full scape mosaic. Less emphasis should be given to

the occurrence of trait associations and syndromes, of aromatic plants in the Mediterranean flora make
such as sclerophylly per se and more experimental the region an important source of such emissions.
research on intraspecific variation is necessary for Since CO2 levels likely influence volatile oil pro-
us to conclude on whether species in the contem- duction it is clear that research on the dynamics
porary flora show evidence for climatic adaptation. of landscape change at the level of within-species
Such experimental work would also be of value variation in plant performance combined with close
for the development of realistic model predictions analysis of genetic variation in volatile oil produc-
for biodiversity changes as the climate continues to tion represent important themes to help broaden the
change in the Mediterranean. Species which occur debate on the ecological consequences of climate
on the fringes of the Mediterranean or which cover change. The productivity of Mediterranean woody
the complete climatic gradient from Mediterranean species can be greatly influenced by levels of CO2
to non-Mediterranean-climatic regimes would be (Hättenschwiler et al. 1997; Rathgeber et al. 1999),
particularly useful model systems here (see also hence such selection pressures may also change as
Chapter 6). My viewpoint is that more precise anal- the climate alters.
ysis, such as the transplant experiments described in Finally, landscapes are rapidly changing in the
the section on aromatic plants, could well show that Mediterranean. First, fire regimes are far from being
localized adaptation to climatic and other ecological constant, and as the climate alters, landscapes may
features of the Mediterranean scene is common in vary dramatically in the future as a result of massive
many plants, and not just annuals. fires and future changes in the fire regime. Second,
In a resource-limited environment such as the loss of natural coastal habitat on the northern
the Mediterranean, several factors may favour shores and woodland in North Africa, and the pro-
the investment of large amounts of fixed carbon gression of woodlands in the back country of the
in the metabolism of secondary compounds. The European Mediterranean countries is also causing
most important of these concerns selection pressures the spatial structure of the Mediterranean mosaic
associated with herbivory, parasitism, and compe- to change dramatically. Selection pressures on plant
tition and the potential contribution of secondary populations and opportunities for gene flow are thus
metabolites to multiple functions. The take home also being rapidly altered. An important compo-
message here is that the multiplicity of ecological nent of this landscape change concerns the processes
roles that secondary compounds play may furnish involved in plant–plant and plant–animal interac-
them with a low opportunity cost, and thus favour tions which may set the scene for population differ-
their presence in the Mediterranean flora. In addi- entiation during dispersal and regeneration. As we
tion, the production of many secondary compounds have seen, facilitation may promote the establish-
shows genetically based variation, which may thus ment of woody species due to the stressful effects
be a key feature of diversification. Such intraspecific of soil aridity in Mediterranean communities and
variation in secondary metabolite concentrations species may respond to genetic variation in associ-
occurs in other Mediterranean floras (Mabry and ated species, in a way which may alter the outcome
Difeo 1973) and is thus of general significance. of interspecific interactions. To further understand
The emission of volatile organic compounds may these ecological processes, and the evolutionary sig-
be of global significance since they may rival the nificance of species interactions, will require close
emission of methane and other compounds of contact between the disciplines of population and
anthropogenic origin. The abundance and diversity community ecology.

Variation and evolution of

reproductive traits in the
Mediterranean mosaic

Through the pioneering efforts of Darwin . . . the naturalists of the last quarter of the nineteenth
century were given good reason to believe that cross-pollination was beneficial to most flowering
plants and that there were, as a result, distinct advantages to the plants in the attraction of potential
pollinators. As a result of this stimulus, the naturalists searched for, and satisfied themselves that
they had found, adaptive advantages in every morphological and behavioural feature of flowers
and inflorescences.
H.G. Baker (1961: 64)

The subject matter of this chapter is the popu-

5.1 Reproductive trait variation:
lation ecology and evolution of reproductive traits
the meeting of ecology and genetics
in the spatially heterogeneous Mediterranean envir-
To understand plant evolution, one must know how onment. For plant species which interact with
a plant reproduces. The study of plant reproduc- animals for their pollination, reproduction and
tion includes two central themes in evolutionary fitness depend on an efficient interaction. The
ecology: the balance between sexual and asexual Darwinian approach to evolution views floral
reproduction and the evolution of the mating system traits in animal-pollinated plants as adaptations to
(i.e. who mates with who, and how often). The pollinators. However, the flowers of many species
evolution of the mating system has long attracted are visited by large numbers of insect species. Any
the attention of evolutionary biologists because it spatial and temporal variation in the abundance and
determines the transmission of genes and thus has composition of pollinator assemblages, when cou-
a crucial effect on the levels of genetic variability pled with differences in pollinator efficiency or in
in a population. Many plants interact with animals their response to floral trait variation, may create
which serve as vectors for pollination and seed dis- variation in selection pressures which may condition
persal, others depend on wind or water. Pollen the adaptive responses of flowers to their pollina-
transfer and seed dispersal are the means by which tors. In addition, attracting pollinators may come
genes migrate among populations, hence, under- at a cost if herbivores, predators, and parasites are
standing the ecological processes of pollination and also attracted. This cost may further limit pollinator-
seed dispersal is critical for our understanding mediated selection and adaptation. In plants, many
of plant population differentiation and evolution. species are sexually monomorphic but show fairly
What is more, diversification in floral morphology continuous quantitative variation in floral traits
can contribute to reproductive isolation and spe- and/or gender. Others show polymorphic variation
ciation. The study of plant reproduction is where in their reproductive system, that is, the occur-
diversity and adaptation meet. rence of two or more mating types (or morphs)


in a population. Understanding the evolution and seasonal Mediterranean climate. They are species
maintenance of such polymorphisms has been a that originated from the Tertiary flora present in
central theme in evolutionary biology ever since the Mediterranean region prior to the onset of the
Darwin. A fundamental feature of many species Mediterranean climate, the mutualisms have thus
with sexual polymorphism is that the relative fre- persisted despite the important climatic oscillations
quency of the different morphs may vary across and distributional shifts the plants and their pollina-
the landscape due to a complex interaction between tors have experienced over the last 2 million years.
the ecological and genetic factors acting on sex The genus Ficus is well known for its tight mutual-
expression, pollination, and the population dynam- istic interaction with particular pollinators (Janzen
ics of the species. In many cases, the stability of 1979; Anstett et al. 1997). A typically tropical genus,
such polymorphisms is closely linked to both the figs find their way into this book by virtue of a sin-
precise functional interactions between plants and gle species, the dioecious common fig Ficus carica.
their pollinators and the population ecology of the The sole pollinator of this species is the agaonid
plant. fig wasp Blastophaga psenes. The inflorescence of the
In this chapter, I discuss three of these issues with fig occurs inside an almost-closed urn-shaped syco-
reference to studies of reproductive trait variation nium, which has a small entrance called an ostiole
and evolution in Mediterranean mosaic habitats. which is covered by several bracts. The inner wall of
the syconium is lined with uni-ovulate female and
• Spatial and temporal pollinator diversity and its in some cases male flowers. The mutualistic inter-
effects on floral trait variation and evolution.
action involves a cycle of events (Valdeyron and
• The interactive effects of pollination and her- Lloyd 1979; Kjellberg et al. 1987). The wasp enters a
bivory on plant fitness.
syconium when female flowers are receptive (prior
• Sexual polymorphisms and their variation among to male flower maturity) and lays its eggs in the
ovules. Some flowers (with a long style) produce
seeds (if pollinated) others (usually with a shorter
5.2 Specialization and generalization in style relative to the length of the wasp oviposi-
a mosaic pollination environment tor) produce a new generation of fig wasps which
are adult when male flowers produce pollen. The
5.2.1 Specialized interactions in
wasps actually mate within the syconium and it is
a seasonal climate
the female wasps, usually bearing pollen, which
Plant species that rely on animals for pollen trans- emerge and then search for a receptive fig for the next
fer may evolve specialized floral traits that facilitate round of oviposition. F. carica is functionally dioe-
their interaction with particular classes of pollina- cious because on some trees all the female flowers
tors (Darwin 1877; Grant and Grant 1965; Stebbins in a syconium have short enough styles for the fig
1970). The evolution of specialized floral morpho- wasps to lay eggs, and they are thus functionally
logy in relation to interactions between floral traits male. In contrast, other trees have only flowers with
and pollination efficiency (a theme I will pick up on longer styles where oviposition is not possible and
later in this chapter in my discussion of style–length only seeds are produced, that is, the tree is female.
polymorphisms) can be illustrated by reference Clearly, a fig wasp which enters a syconium on a
to the stability of species-specific plant–pollinator female tree will have no reproductive success, it is
mutualisms, in which the plant benefits from pol- as if it enters a trap, a deadly one.
lination by a single species of insect which itself One would thus predict strong selection on the
is specific to the plant, often because it lays its wasp to avoid female figs. However, the sex-specific
eggs in plant tissues and thus assures its repro- seasonal phenology of F. carica means that male
duction. Two Mediterranean plants, the common and female syconia are almost never simultaneously
fig and the dwarf palm, illustrate several aspects available to the fig wasp. Thus asynchrony of attract-
of the stability of such mutualisms in the highly ivity among sexes prevents wasps from having
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 169

a choice (Valdeyron and Lloyd 1979; Kjellberg et al. complete discrimination by the pollinator. Females
1987). When the majority of wasps emerge from thus benefit from pollination but do not provide a
male syconia (thus bearing pollen), the only recept- site for larval development, they thus ‘cheat’ their
ive syconia elsewhere (with no wasps inside them) pollinator which may have reduced reproductive
are female. Wasps only live for ∼2 days so they success as a result of visiting females (Dufaÿ and
have no choice, enter a female synconia or per- Anstett 2004).
ish. In fact, in experimentally manipulated arrays, In both these ‘nursery pollination mutualisms’
the pollinators of F. carica show an ability to dis- where the plant hosts larval development, an
criminate between male and female inflorescences important feature of the plant species is that they
(Anstett et al. 1998). The blend of volatile compounds are dioecious: female plants prevent egg laying or
emitted by receptive male and female syconia con- larval development of the pollinator, and thus avoid
tains a very similar array of compounds (Gibernau the major potential cost of such a mutualism. The
et al. 1997; Grison-Pigé et al. 2001), however, the costs of such larval development thus differ between
proportions of the different compounds in the blend males and females, primarily since the phenology
differ among males and females (Grison-Pigé et al. of larval development is asynchronous with male
2001). Hence wasps can choose, if they are given a function. Both species are single representatives of
choice. Pollination of females is thus by deceit and Pre-Pliocene palaeo-tropical ancestral lineages in
the asynchronous phenology of males and females, the Mediterranean and thus pre-dated the evolution
in a highly seasonal environment, causes selection of a highly seasonal climate (Chapter 1). Although
on strict sexual mimicry in this system to be weak, the tight mutualistic interaction pre-dates their
thereby promoting the stability of the mutualism. existence under a Mediterranean-climate regime,
On the northern shores of the western Medi- these two species have nevertheless maintained
terranean, the dioecious dwarf palm, Chamaerops a highly specific interaction with their pollinator.
humilis is pollinated by a single species of weevil During the repeated oscillations of the climate
(Derelomus chamaeropsis: Curculonidae), whose eggs since the Pliocene the pollinators have thus tracked
develop and pupate in the rachises of male inflo- distribution changes in the plant and the mutualism
rescences and adults transfer pollen from males to has persisted.
females (Anstett 1999). So, in exchange for pollen
dispersal, dwarf palms provide the weevils with
5.2.2 Generalization in Mediterranean
shelter, egg-laying sites, and food, and the two
plant communities
partners have highly co-adapted life cycles and
reproductive biology. Although weevils lay eggs For a plant to evolve floral adaptations to a particu-
in females as they pollinate, processes associated lar class of pollinators, it is necessary that different
with fruit development prevent larval development potential pollinators vary in the strength of their
(Dufaÿ and Anstett 2004), hence males assure the interactions with the plant (Schemske and Horvitz
next generation of potential pollinators. Despite 1984). Such variation will be governed by two key
their lack of reproductive success, weevils continue elements of the interaction: the relative ‘quantity’
to visit females, albeit perhaps less often. An import- of visits by different insects and the ‘quality’ of the
ant facet of this interaction concerns the chemical interaction in terms of rates and distances of pollen
attraction of the weevils by volatile chemicals emit- dispersal within and among plants (C.M. Herrera
ted by vegetative parts of the plant (Dufaÿ et al. 1987, 1988c, 1989).
2003; Chapter 4). Although the blend composition of If a plant species relies on more than one species
the volatile attractive fragrance varies among plants, for pollination, and if each pollinator visits more
and males may produce more scent, there is no sig- than one plant species, generalized pollination
nificant variation in the composition of the fragrance systems may develop. Several factors favour the
emitted by females and males (Dufaÿ et al. 2004). maintenance of generalist pollination systems and
Such similarity in attractive potential may prevent a network of plant–pollinator interactions in local

communities (Waser et al. 1996; Olesen and Jordano a degree of asymmetry in the plant–pollinator rela-
2002): (a) spatial or temporal variation in pollina- tions of such systems. Although individual plant
tor abundance which could cause pollen limitation species are visited by many insect species, individ-
of seed set or even plant extinction in a particular ual species have a smaller number of plant targets.
region if a preferred pollinator is absent; (b) simi- In the coastal scrubland of southern Spain, J. Herrera
larities in floral rewards among co-occurring plant (1988) reported that whereas a single insect visited
species; (c) if pollinators suffer energetic constraints on average just two plant species, a given plant
on pollinator flight distances and a long lifetime species was, on average, visited by 16 different
of insects relative to the flowers in a community; insect species. The large number of rare insect vis-
(d) lack of variation in pollination efficiency, that is, itors and the tendency to focus studies on fairly
a sort of functional equivalence. Waser et al. (1996) common plant species no doubt contribute to this
and Olesen and Jordano (2002) argue that these sit- asymmetry. Additionally, plants flowering at the
uations commonly exist in nature, to quote Waser same time tend to be visited by the same insect taxa,
et al. (1996: 1053) they may represent ‘the rule rather more so than plants with similar floral morpholo-
than the exception’. However, spatial variability in gies but which have different dates of maximum
pollinator abundance, if coupled with differences flowering (Petanidou and Vokou 1993).
among pollinators in their pollination efficiency, At the plant community level the pollinator fauna
could also create a spatial mosaic of more or less can also change dramatically during the flower-
specialized interactions (Thompson 1994). In this ing season (J. Herrera 1988; Petanidou and Vokou
case, different evolutionary trajectories in plant– 1990). In the Lamiaceae of the Greek phrygana,
animal interactions and different floral trait adap- Petanidou and Voukou (1993) reported that plants
tations may evolve in a single species in different which flower in late winter or early spring have a
environments. small number of flowers and are visited by few insect
Several Mediterranean plant communities have taxa while those which flower later in spring bear
generalized pollination systems, in which plant many flowers and are visited by a large number of
species are visited by a large number of insects, different insects.
themselves unspecialized on a single plant species, The major type of reward in Mediterranean
for example, in coastal scrub community, at the flowering plants is pollen (Kugler 1977). Many
Reserva Biológica Doñana in south-west Spain Mediterranean plant species, in fact most animal-
(J. Herrera 1988), open garrigues of southern France pollinated species are pollinated by wild bees. This
(my own data), herbaceous grassland in northern predominance of bees as pollinators has been doc-
Spain (Bosch et al. 1997) and the phrygana ecosystem umented in Spain (J. Herrera 1987, 1988), Greece
in Greece (Petanidou and Vokou 1990; Petanidou (Petanidou and Vokou 1990; Petanidou and Ellis
and Ellis 1993). The occurrence of generalized pol- 1993), and Israel (Potts et al. 2001). In southern Spain
lination systems does not mean that there is no only 35% of a total of 122 plant species were con-
observable pattern in communities, nor does it rule sidered nectariferous by J. Herrera (1985). In the
out the existence of a ‘certain structure in the plant phrygana ecosystem of Greece, Lepidoptera repres-
pollinator interactions’ (Bosch et al. 1997: 590). The ent a small fraction of insect visitation which is
latter authors identified four functional groups of primarily by pollen collecting bees (Petanidou and
plants based on their main types of insect visi- Vokou 1990). These authors interpret this pattern
tors, with plants with similar reward composition in relation to drought which can have a negative
(in terms of relative dependence on pollen and effect on nectar production (see also Potts et al. 2001).
nectar) having similar visitors. So, for example, Hence, to quote J. Herrera (1985: 51) ‘species that
Fabaceae tend to be visited primarily by bees, rely on a copious and constant production of pollen
Cistaceae by Diptera and pollen collecting bees, may have a higher or more constant . . . reproductive
and Lonicera by hawkmoths. Such patterns remain, success’. This conclusion awaits experimental
however, coarse (J. Herrera 1988). There is also confirmation.
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 171

An extreme functional extension of this lack of and are thus pollinated by deceit (Aronne et al.
nectar reward can be observed in the formation of 1993). Such pollination by deceit requires that polli-
completely rewardless flowers that rely on decep- nation be to some extent unspecialized. In nectarless
tion of insects for pollination (like in figs above). Cyclamen, several species possess a buzz-pollination
For example, the orchid flora of the Mediterranean strategy where bumblebees actively vibrate flow-
region, in common with that in southern Australia, ers to extract pollen. In such species, the functional
shows an important element of mimicry and non- aspects of the plant–pollinator interaction are likely
rewarding flowers (Dafni and Bernhardt 1990). In to be specialized.
Orchis caspia in the eastern Mediterranean, this Many Mediterranean plant species have diverse
deception involves more than one insect species pollinator assemblages (Table 5.1). This includes
and several other plant species (Dafni 1983). In species with papillonaceous flowers and those with
different sites, O. caspia attracts pollinators from fairly long and narrow tubular corollas which could
different rewarding species according to the abun- in theory restrict visitation to particular types of
dance of the latter. In this case the deception insect (J. Herrera 1988; C.M. Herrera 1996; Bosch
relies on the occurrence of a generalized pollination et al. 1997; Thompson 2001). In addition, some
system in which pollinators show little discrim- species may be pollinated by a combination of
ination among plant species. In Nerium oleander, very different groups, for example, insects and ants
the lack of nectar is associated with a strategy of in Lobularia maritima in south-west Spain (Gómez
pollination by deceit since even the pollen is dif- 2000) and lizards and insects in Euphorbia dendroides
ficult to access due to the concealed position of on the Balearic islands (Traveset and Sáez 1997).
the anthers in the flowers (J. Herrera 1991a). In the In addition, some species may combine both wind
dioecious hemiparasitic shrub Osyris alba, females and animal pollination, for example, Hormatho-
neither produce pollen, nor have any nectar reward, phylla spinosa in the Sierra Nevada (Gómez and

Table 5.1 Examples of insect-pollinated plant species with diverse pollinator assemblages in the Mediterranean region

Species Family Region Pollinator assemblage Pollinator variation

Cistus libanotisa Cistaceae Eastern Mediterranean Diverse Hymenoptera, Among sites

Coleoptera, and Diptera
Three Cistus speciesb Cistaceae North-east Spain Diverse Hymenoptera, Not studied
Coleoptera, and Diptera
Paeonia broteroi c Paeoniaceae Iberian peninsula >25 insect species, mostly One species of bee is the
Hymenoptera principal pollinator
Asphodelus albusd Liliaceae Northern Spain 13 species of Hymenoptera, Among years
Diptera, and Lepidoptera
Satureja thymbrae Lamiaceae Greece At least 34 bee species Not studied
Lavandula latifoliaf Lamiaceae Sierra de Cazorla, ∼70 species of Hymenoptera, Among and within sites, and
south-east Spain Diptera, and Lepidoptera among and within years
Jasminum fruticansg Oleaceae Southern France and >28 species of Hymenoptera, Among sites and
north-east Spain Diptera, and Lepidoptera among years
Lobularia maritimah Cruciferae North-east Spain >50 species of Hymenoptera, Not studied
Diptera, and Coleoptera
Hormathophylla spinosai Cruciferae Altitudinal gradient in ∼70 species (19 families in 5 orders) Among 3 populations
the Sierra Nevada

a Talavera et al. (2001). b Bosch (1992). c Sánchez-Lafuente et al. (1999).

d Obeso (1992). e Potts et al. (2001). f C.M. Herrera (1988c).
g Thompson (2001). h Gómez (2000). i Gómez and Zamora (1999).

Zamora 1996) and Urginea maritima in coastal areas dominant component of the pollinator assemblage
of the eastern Mediterranean (Dafni and Dukas in four years, Lepidoptera numerically dominated
1986). Despite this diversity, in insect-pollinated the assemblage in one year, and a more balanced
species which have received detailed investigation, assemblage occurred in the remaining year. Even
for example, Lavandula latifolia, (C.M. Herrera the identity of species in the group of abundant
1988c), Jasminum fruticans (Thompson 2001), and H. visitors varied among years. Intensive observa-
spinosa (Gómez and Zamora 1999), a small number tions revealed that the cortège of visitors showed
of abundant pollinators dominate the pollinator strong seasonal dynamics during a single year and
assemblage which otherwise contains a large that pollinators do not all visit at the same time
number of fairly rare floral visitors. For example, of the day. For example, Anthophora bees exhib-
the 10 most abundant taxa represent ∼80% of ited two peaks, one in mid-morning the other in
all visits to both L. latifolia and J. fruticans. These mid-afternoon, a pattern observed in other species
diverse pollinator assemblages thus show a highly with different flowering ecologies, for example,
skewed (log-normal) frequency distribution of Petrocoptis grandiflora (Navarro et al. 1993) and
species abundance and rates of visitation, indicative Narcissus assoanus.
that a small number of strong interactions may 2. Pollinator abundance and composition varied
exist within the buzzing array of visitor diversity. significantly in space both among populations, in
Evidence for specialized interactions in species with relation to proximity to water, and within-sites, due
diverse pollinators remains however elusive, as the to a mosaic of sun-shade patches. Although for some
two examples in the next section illustrate. pollinators foraging behaviour varied among sun
and shade patches, others principally visited plants
in full sunshine. Hence, variation in the pollinator
5.2.3 The spatio-temporal pollination mosaic: assemblage was highly localized in relation to micro-
consequences for floral trait variation and habitat conditions. Although this pollination mosaic
evolution can be related to some environmental features and
the population cycles of some insects, it also has a
Detailed analyses of pollinator diversity in a six-year
strong stochastic component.
study of L. latifolia (Plate 3) in south-east Spain by
C.M. Herrera (1987, 1988c, 1989, 1990a, 1995a) illus-
For such variation to create a spatio-temporal mosaic
trates very clearly the spatial and temporal variation
in selection regimes on floral traits pollinators must
that may occur in the cortège of pollinators visit-
vary in their activity and pollination efficiency in a
ing a single plant species. Flowering at a time of
way which has consequences for plant fitness. There
the year when very few species are in flower (peak
should thus be some association between abund-
flowering is in July and August) L. latifolia is a tar-
ance, foraging efficiency, and pollination effective-
get for ∼70 insect species searching for pollen and
ness (rates and distance of pollen transfer). In
nectar. Although most interactions involve a small
L. latifolia the quality and quantity of pollination
number of species, there is no consistent relationship
do not go hand in hand (C.M. Herrera 1987). Pol-
between abundance and visitation rates, since no
linators differ in (a) the number of pollen grains
pollinator was both abundant and an efficient for-
they deposit on the stigma during flower visita-
ager, and the few species with the most visits occur
tion, (b) their visitation rates to male-phase and
in diverse taxonomic groups. What is more, pollina-
female-phase flowers (L. latifolia flowers are protan-
tor species abundance and the composition of the
drous), and (c) patterns of flight distance among
pollinator assemblage showed significant temporal
plants. For example, pollination by Lepidoptera
and spatial variation (Fig. 5.1).
may occasion greater rates of outcrossing as a
1. Variation in time occurred among years, at dif- result of their longer flight distances among plants.
ferent moments of the flowering period and at Hence, subtle differences in the composition of
different times of the day. Hymenoptera were the the pollinator assemblage in the landscape could
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 173

(a) 60
Number of visits per observation period

1982 1983 1984 1985 1986 1987

(b) 80




Late July Early August Late August Early September Late September Early October

(c) 80 (d)
60 12
40 8
20 4
0 0
Population Sector within a population

Figure 5.1 Spatio-temporal variation in the pollinator assemblage visiting L. latifolia at sites in south-east Spain. Pollinator composition can be
seen in the relative portions of each bar and is based on the mean number of individuals per observation period by Hymenoptera (hatched portion
of bar), Diptera (closed portion), and Leipdoptera (open portion). Graphs show (a) temporal variation among years, (b) temporal variation within
a year in a single population, (c) spatial variation among populations, and (d) spatial variation among locations within a population. Data are
number of visits per observation period (redrawn from C.M. Herrera 1988c).

have profound consequences for plant population for plant reproduction and floral evolution concerns
dynamics (C.M. Herrera 2000a,b). Finally, manip- J. fruticans (Table 5.1). Whereas L. latifolia occurs in
ulation of the corolla lobes of L. latifolia (one, both, the Lamiaceae, which is part of the Mediterranean
or neither of the two corolla lips were modified), element of the flora that has shown explosive radia-
showed that the modification of one trait was with- tion during the recent history of the Mediterranean,
out consequence for pollen removal and pollen J. fruticans is a tropical relict in the Oleaceae (from
deposition (C. Herrera 2001). Corolla integration the ancient Tertiary flora), and is the only native
is thus not closely related to selection pressures member of the genus in the western Mediterranean,
imposed by pollinators, and may stem more from where it occurs at the northern distribution limits
genetic correlation among traits. of the genus. J. fruticans is a clonal shrub (Plate 3)
To sum up, the spatio-temporal pollination common on scree slopes and around fields and
mosaic described for L. latifolia provides no evid- vineyards in the western Mediterranean. It bears
ence of a mosaic of specialized interactions, rather tubular yellow flowers in May when the garrigues of
a diversification which may reduce the impact of southern France is alive with potential pollinators.
pollinator-mediated selection on plant traits. Its fleshy fruits are dispersed by birds and hence its
A second example of the spatio-temporal mosaic frequent occurrence under other trees and shrubs
in the pollinator assemblage and its consequences or along old stone walls used as perching posts.

At least 28 insect species (Hymenoptera, Diptera, proportion of flowers. Insect groups also differ in
and Lepidoptera) in six main groups visit J. fruticans their behavioural response to variation in floral
flowers, where they actively feed on nectar or collect design (e.g. flower size) and floral display (numbers
pollen (Thompson 2001): of open flowers). Depending on the insect, visitation
rate was positively related to the number of flowers
(1) short-tongued bees (mostly Apis mellifera:
on a stem (short-tongued bees), the number of open
flowers in a patch (bee flies and butterflies), corolla
(2) long-tongued bees (Anthophora, Eucera, and
tube length (hawk moths), or corolla lobe length
Xylocopa: Hymenoptera);
(butterflies). These different foraging responses
(3) bumblebees (Bombus: Hymenoptera);
reflect differences in the biology of the insects.
(4) bombylid flies (Bombylius: Diptera);
The divergent selection pressures that variation
(5) hawkmoths (Macroglossum stellatarum and
in pollinator visitation may create are all the more
Hiemaris fuciformis: Lepidoptera);
intriguing in J. fruticans because of its floral bio-
(6) several butterfly species (Lepidoptera).
logy. This species is distylous (Box 5.1), a floral
As in L. latifolia, the composition of the pollina- polymorphism in which the morphs differ in the
tor assemblage varies significantly in space (among sequence of heights at which stigmas and anthers
populations in southern France and northern are positioned within the flowers (see Section 5.5).
Spain) and time (among years in a single popula- The reciprocal positioning of stigmas and anthers
tion) (Fig. 5.2), and insect groups vary markedly in the different floral morphs promotes cross-
in their foraging efficiency (Fig. 5.3). For example, pollination among morphs due to precise pollen
whereas hawkmoths and bombylid flies rapidly positioning on pollinators and efficient pollen trans-
visit a large proportion of open flowers on a given fer among sex organs at each level in the flower
flowering stem, butterflies slowly visit a small
(a) 160
(a) 14
Visitation rate

Number of visits per observation period

8 80
4 40
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Population Population
(b) 14 (b)
Proportion visited

8 1.2
6 0.8
0 0.0
1994 1995 1996 1 2 3 4 5 6
Figure 5.2 Spatio-temporal variation in the number of visits of
different pollinators to J. fruticans (a) in six populations and (b) over Figure 5.3 Variation in the foraging behaviour of different
three years in a single population. Data are number of visits per pollinators visiting J. fruticans (a) visitation rates to flowers in a
observation period. From the bottom of each bar: hatched given time and (b) proportion of flowers visited per stem during a visit
portion—short-tongued bees, grey portion—long-tongued bees, (drawn from data in Thompson 2001). From the bottom of each bar:
black portion—bumblebees, dotted portion—bombylid flies, open hatched portion—short-tongued bees, grey portion—long-tongued
portion—hawkmoths, diagonal portion—butterflies (drawn based on bees, black portion—bumblebees, dotted portion—bombylid flies,
data in Thompson 2001). open portion—hawkmoths, diagonal portion—butterflies.
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 175

(see later in this chapter). One would thus predict divergent selection on style length in the L-morph.
strong selection on stigma and anther position. For the S-morph, however, it is unclear whether
However, in J. fruticans, in contrast to many dis- there is any cost or benefit associated with vari-
tylous species (see Thompson and Dommée 2000), ation in style length. Quantitative variation in the
there is marked quantitative variation in these floral S-morph may thus be neutral and maintained by
traits (Box 5.1). Two aspects of this variation are a simple genetic correlation between style length
particularly intriguing. in related individuals of the two morphs. In disty-
First, style length shows continuous variation, lous species such as J. fruticans where each morph is
and in some short-styled plants (S-morph) the only fertilized by the other morph, and because the
stigma is at the same level as the anthers (at the genetic control of the polymorphism has a simple
mouth of the corolla tube). Plants with this floral genetic basis (the L-morph is usually homozygous
morphology have pollen grains that are (a) similar recessive ‘ss’ and the S-morph heterozygous ‘Ss’)
in size to those produced by plants of the S-morph, each morph produces equal numbers of the two
that is, significantly larger than pollen grains of long- morphs in its offspring. This means that if style-
styled plants (L-morph) and (b) incompatible on length variation within a morph is due to the
other plants of the S-morph. As in many distylous expression of quantitative genetic effects or modifier
species, J. fruticans is self-incompatible and crosses genes, then offspring of both L-morph and S-morph
among plants of a given morph are also incom- maternal parents will show both qualitative and
patible. This occurrence of S-morph plants which quantitative style-length variability. In J. fruticans,
lack any stigma–anther separation is an unusual fea- Thompson and Dommée (2000) reported correlated
ture for a distylous species, where the occurrence variation in the mean style-length of the two morphs
of such ‘homostyly’ is usually associated with the among populations.
occurrence of a novel variant morphology and a In summary, the patterns of floral variation sug-
breakdown in the heterostylous system (Section 5.5). gest that the spatio-temporal mosaic of variation
This is not so in jasmine where homostylous flowers in the pollinator cortège, coupled with variation in
are part of the S-morph variability. pollination efficiency of different groups, may cre-
Second, many plants of the L-morph have a curled ate divergent selection on floral traits such as style
style, and the longer the style, the greater the num- length and stigma position, even within a floral
ber of plants with a curled style in a population morph. In the L-morph, having a longer than aver-
(Box 5.1). Any advantage to having a longer than age style may provide an advantage (production
average style may be counterbalanced by costs asso- of more vigorous offspring), however it is neces-
ciated with placing the stigma too far out of the sary to keep the stigma close to the mouth of the
flower. Indeed, as stigma exertion of the L-morph corolla in order to receive pollen. A long, but curled,
increases, I have observed a marked decline in the style satisfies these conflicting selection pressures
probability that a stigma will receive S-morph pollen and may represent an ‘adaptive compromise’ to
and variation among insect groups in whether they divergent selection pressures on floral design (see
cause pollination. There is also a significant corre- Armbruster 1996). In addition, a genetic correla-
lation between maternal style length and seed and tion may maintain variation in both morphs. One
cotyledon size of young seedlings in the offspring of last point here, in J. fruticans style-length variation
the L-morph (but not the S-morph). Such enhanced occurs within the confines of a complex genetic
seedling vigour may result from some advantage polymorphism, which itself no doubt evolved in
to increased style length in the L-morph, perhaps this genus (the tropical and cultivated flowers of
via gametophytic selection (i.e. where more vigor- this genus I have seen have a floral morphology
ous pollen outcompetes pollen of poorer viability which suggests distyly) prior to the Mediterranean-
for fertilization in flowers with longer than average climate regime and thus in a very different polli-
styles) or correlated variation of other traits. nation environment. Quantitative floral variation
Why then do both morphs show variation in may thus have evolved in association with a fairly
style length? We can see above that there may be recent diversification of pollinators. Indeed, studies

Box 5.1 Floral biology of Jasminum fruticans.



Jasminium fruticans is a distylous species (Guitián et al. 1998; Thompson and Dommée 2000), that is, with
two floral morphs

(a) 16 (b) 100

Anther height (mm)

% of plants with
a curled style

10 60
8 40
2 0
2 4 6 8 10 12 14 16 11 12 13 14
Stigma height (mm) Mean style length

(a) Some short-styled plants (open squares) have stigmas adjacent to the anthers, while in long-styled plants
all plants (filled squares) all plants have a marked stigma–anther separation and (b) many have a curled style.
The degree of curling in 12 populations in southern France and Spain is positively related to the mean style
length of the L-morph in a population.
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 177

of endemic Viola cazorlensis (C.M. Herrera 1990b,c; evaluation of traits in relation to variation in
1993; Plate 3) and Arum italicum (Méndez and Diaz diverse ecological parameters that create selection
2001) in Spain produced no evidence that floral or and constraints in natural situations. In perennial
inflorescence traits influence fitness. plants, establishing the precise role of traits in
Abiotic factors may also affect the function- fitness variation and assessing the true contribu-
ing of plant–pollinator interactions (as the patch- tion of single-year estimates of fertility is difficult
dependent pollination of L. latifolia would suggest) due to demographic costs of reproduction and the
and cause pollen limitation on fruit and seed pro- complexity of untangling the role of different eco-
duction to occur and vary in space and time. An logical factors (e.g. C.M. Herrera 1991). In addi-
important factor here is climate unpredictability tion, floral trait variation may be closely related
which can have a strong impact on pollinator activ- to differential male fitness via an effect on pollen
ity. In Narcissus longispathus, which flowers early removal and transfer. So evidence for adaptation,
in spring when low temperatures may hinder bee and its absence, must be cautiously and carefully
activity and successful pollination, a favourable interpreted.
microclimate inside the large tubular flowers, a
long floral lifetime (14-day-old flowers are still cap-
5.3 Attracting pollinators . . . but
able of high seed set), and the thermal biology
avoiding herbivores
of the principal bee pollinator (Andrena bicolour
which raises thoracic temperature by basking in In animal-pollinated plants, reproduction requires
and on the corolla) are critical for pollen limita- that pollinators be attracted to a flower. At the
tion to occur in this species (C.M. Herrera 1995b). same time, however, plants must deter poten-
In the Lamiaceae, nectar production is maximized tial herbivores and seed predators, which may
at high temperature (Petanidou and Smets 1996) negate the benefits provided by mutualistic pol-
and species which flower late in spring have a linators. Pollination and herbivory may rarely
higher nectar sugar concentration than earlier flow- operate independently and their ecological effects
ering species (Petanidou et al. 2000). In N. assoanus and evolutionary significance may be closely inter-
in southern France, local climate (late snow and connected, particularly when they affect the same
heavy rain during peak flowering) reduces seed pro- traits (Strauss and Armbruster 1997). Indeed, attrac-
duction to almost zero in upland populations on tive and defensive traits may be linked by function or
limestone plateau areas (>700 m) which contain tens because of phylogenetic constraints and some traits
of thousands of plants in some years. Indeed, sev- involved in defence may have subsequently evolved
eral Mediterranean species show spatio-temporal to attract pollinators (Armbruster et al. 1997). If
heterogeneity in pollen limitation on seed produc- herbivores affect traits which influence pollination,
tion, for example, N. assoanus in France (Baker et al. potential selection exerted by pollinators may be
2000a) and dioecious Rhamnus ludovici-salvatoris on altered, masked, or rendered insignificant. Her-
the Balearic Islands (Traveset et al. 2003). In Helle- bivory may thus alter the nature and strength of
borus foetidus, pollinator visitation may be highly pollinator-mediated selection on floral traits and
variable and infrequent because of early flowering may exert quite different, even conflicting, selective
in relation to pollinator abundance and/or climatic pressures on reproductive traits. To fully under-
limitations (C.M. Herrera et al. 2001). stand how reproductive traits evolve thus requires
Several of the above examples suggest that that any non-additive (i.e. interaction) effects of her-
pollinator-mediated selection may be rare. How- bivory (including predation of seeds and parasitism)
ever, natural selection is notoriously difficult to and pollination be evaluated.
detect. So in some ways the examples I discuss In the Mediterranean, insect diversity is high,
here illustrate more that pollinator-mediated selec- hence the need to fully appreciate the com-
tion may be very complex and that to study floral bined effects of pollination and herbivory on
evolution requires precise, detailed and long-term reproductive success. In several Mediterranean

species, pre-dispersal flower, fruit, and seed preda- 25

tion has been reported to have significant impact
on fitness, and thus potentially constrain selection 20

Maternal fertility
related to pollinators. In Lavandula stoechas, pre-
dispersal seed predation by insects can account for
a 31% reduction in seed set (J. Herrera 1991c). In 10
N. assoanus, flower herbivory and fruit predation
may cause up to 60% of plants in a given site in 5
a given year to have no fruit production. In relict
populations of Frangula alnus in southern Spain, Excluded Present
floral herbivory strongly limits reproductive output Pollinators
(Hampe and Arroyo 2002; Hampe 2004). Finally,
in five Onopordum species in Greece, 34 out of 46 Figure 5.4 Schematic representation of the interactive effects of
populations have reduced seed production due to pollination and herbivory on maternal fitness (seed production,
seedling emergence, and seedling recruitment) in P. broteroi and
infestation by the weevil Larinus latus (Briese 2000).
H. foetidus populations in Spain (based on C.M. Herrera 2000 and
In endemic Ebenus on the Mediterranean coast of C.M. Herrera et al. 2002, respectively). Open symbols: herbivores
Egypt and Libya, ∼90% of seeds are consumed or present, closed sysmbols: herbivores excluded.
damaged by a bruchid beetle (Hegazy and Eesa
1991). Post-dispersal seed predation is also very 1,000 to 2,000 m above sea level in several moun-
important in several Mediterranean woody species tain ranges in southern and eastern Spain. In the
(Hulme 1997; Quézel and Médail 2003). Sierra Nevada, plants grow as perennial rosettes
The balance of positive affects related to pol- for up to ∼4 years, after which they produce many
lination and the negative affects of herbivory have reproductive stems bearing up to several hundred
been studied in a number of Mediterranean plants. bright-yellow hermaphroditic flowers. In the Sierra
Studies of the interactive effects of herbivory and Nevada, E. mediohispanicum is primarily pollinated
pollination on seed production in a population of by the pollen collecting beetle Meligethes maurus
Paeonia broteroi (Plate 3; C.M. Herrera 2000c) and (Nitulidae), has its flowers and fruits browsed by
the recruitment of seedlings by maternal plants of the Spanish Ibex (Capra pyrenaica), its sap sucked
H. foetidus in two different regions (C.M. Herrera by several bugs (Corimeris denticulate and Eurydema
et al. 2002) have revealed such non-additive fitness subsp.), and its stems and fruits bored by wee-
consequences of exposure to herbivory and pollina- vils. Gómez (2003) reported two results indicative
tion with two important trends (depicted schemat- of potential pollinator mediated selection in the
ically in Fig. 5.4). First, pollination has a positive absence of ungulate herbivores (the major herbi-
effect on seed production, seedling emergence, and vores on this species). First, lifetime maternal repro-
recruitment only if herbivores are excluded. Second, ductive success (easily measured since although
the exclusion of pollinators renders the negative plants are perennial they die after one bout of
effect of herbivores on seedling recruitment non- flowering) was positively correlated with pollina-
significant. In other studies, the effects have been tor visitation rates. Second, several traits influenced
found to be additive in nature since herbivores exert pollinator visitation rates, indicative of directional
a negative effect on performance even in the absence selection on these traits. For instance, plants with
of pollinators, due to the capacity of the studied more flowers had higher absolute (total seed num-
species to autonomously self-pollinate (e.g. Box 5.2). ber per plant) and relative (seed number per fruit)
In the montane herb Erysimum mediohispan- seed production. Positive directional selection was
icum (Plate 3), Gómez (2003) studied whether also detected for flowering stalk height and two
pollinator-mediated selection on floral traits is mod- floral traits (petal length and inner diameter). In the
ified by the effects of insect and ungulate her- presence of ungulate herbivory, selection on flower
bivory. This monocarpic perennial herb occurs from number was markedly reduced and selection on
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 179

Box 5.2 A comparative study of herbivory and pollination in natural populations of two
congeneric Aquilegia in southern France (Plate 3) (graphs drawn from data in
Lavergne et al. (2004a)

Aquilegia viscosa is endemic to the Languedoc to floral buds and prevent flowering, and various
Roussillon region of southern France and Diptera and Lepidoptera) and pollen limitation
north-east Spain and its congener A. vulgaris is (pollination is essentially by bumblebees) thus
widespread across western Europe. In these two combine to limit the maternal fertility of the
species floral predation (primarily Curculionideae endemic species but not the congeneric
larvae, which cause major damage widespread species in the same region.

Widespread A. vulgaris Endemic A. viscosa

140 1. Absence of
140 predators and pollinators
Number of seeds per fruit

Number of seeds per fruit

100 2. Absence of predators
100 with outcross pollination
3. Predator exclusion
60 and open to pollination
40 40
4. Open to predation
20 20 and pollination

0 0
1 2 3 4 3 4 1 2 3 4 3 4
Glasshouse Pop 1 Pop 2 Glasshouse Pop 1 Pop 2

Herbivory significantly limits reproductive success rare), it would appear that pollen limitation also
(by up to 56%) in the endemic species but not in constrains female fertility in the endemic species.
the widespread species. In unpredated flowers, The ability to attract pollinators, but avoid floral
seed set in the endemic species remains herbivores, both contribute to female fertility and
significantly less than that of the widespread perhaps influence the differences in distribution of
species. Given that in the absence of both the two species. In endemic A. viscosa, herbivory
pollinators and herbivory (i.e. in the glasshouse) imposes a strong constraint on maternal fertility
the two species show equivalent levels of even in the absence of active pollination due
autonomous self-pollination compared to seed set to autonomous self-pollination which provides
on outcrossing and because pollinator visitation some reproductive assurance in the absence of
rates are significantly higher in populations of the pollinators. Herbivory may thus continue to limit
widespread species than in populations of the maternal fertility even where pollinators
endemic species (where pollinators are extremely are rare.

floral traits became non-existent, that is, as in the are present. These two opposing selective forces may
above studies, the detrimental effect of herbivory limit the evolution of adaptive variation in flower
on maternal fecundity was so strong that it cancelled number.
out any pollinator mediated selection on particular In the high mountain zone of the Sierra Nevada,
traits. In fact, whereas tall plants with many flowers the mass-flowering crucifer Hormatophylla spinosa
are strongly favoured in the absences of herbivores, (Plate 3) shows significant variation in traits related
they are counter-selected when ungulate herbivores to how it interacts simultaneously with a number

of antagonistic herbivores and seed predators and This cost may be of a genetic nature if inbreed-
mutualistic pollinators (Gómez and Zamora 2000; ing occurs and progeny suffer reduced viability
Box 5.3). These authors found that only ungulates (inbreeding depression), or more ecological in
(mainly Spanish Ibex) and pollinators significantly nature if in self-incompatible species self-pollination
affect plant fitness and the relative effects of these causes reduced mating opportunities (i.e. self-
interacting organisms varied greatly among study interference).
sites—indicative of functional specialization within Despite the prevalence of hermaphroditism, in
populations. In two populations, ungulate pressure flowering plants there is an immense diversity of
was most intense and significantly reduced per- mating strategies. Between the extremes of ‘perfect’
formance. In the population where this effect was hermaphrodite flowers and unisexuality (dioecy)
most severe, plants had a significantly higher num- lie a wide range of intermediate types, the most
ber of thorns. In contrast, where ungulates were prominent of which include:
less abundant, and thus caused less damage to
• monoecy: male and female flowers on the same
plants, maternal fitness was closely correlated with
pollinator visitation. As above, floral traits only sig-
• andromonoecy: male and hermaphrodite flowers
nificantly contributed to performance variation (via
on the same plant;
pollinator attraction) in sites where herbivore pres-
• gynomonoecy: female and hermaphrodite
sures were non-significant. So in this species, only a
flowers on the same plant;
highly restricted sample of all the potential interact-
• gynodioecy: the coexistence of female and
ing organisms is likely to exert a selection pressure
hermaphrodite plants;
on plant traits in a given population.
• androdioecy: the coexistence of male and
Ultimately, plants with traits that enhance pol-
hermaphrodite plants.
lination and allow some form of defence against or
escape from herbivory should be selected. However, These different sexual systems combine variation
the frequent occurrence of herbivory on reproduct- among flowers/inflorescences on a single indi-
ive structures of herbaceous Mediterranean plants vidual and among plants in a population.
suggests that such interactive effects may commonly Even within hermaphrodites there can be much
affect plant reproduction in this region and perhaps variation in the mating system. First, some species
elsewhere (C.M. Herrera 2000c; C.M. Herrera et al. may be almost completely outcrossing, while others
2002). Once again, the message is clear, to fully quan- may rely on selfing to assure seed production,
tify and understand the role of pollinator-mediated and others may have a mixed mating system
selection requires appreciation of the range of eco- with intermediate, and often variable, rates of
logical factors influencing reproductive success in selfing and outcrossing (Barrett 2002a). Second,
natural environments. hermaphrodites may vary in their relative allocation
of resources to male and female function as a result
of environmental and ontogenetic causes (Lloyd
5.4 Mating system and and Bawa 1984).
gender variation The occurrence of variation in plant sexual sys-
tems entered early into the literature on plant evolu-
5.4.1 Mating system variation and evolution
tion. There thus exists a long tradition of population
Most plant species are hermaphroditic (co-sexual), genetic studies of the evolution of the mating system
with flowers that contain both male (stamens) and and the relative effects of selfing and outcrossing on
female (pistils) sex organs. Co-sexuality is favoured genetic variation which has been paralleled by the
in species with high resource allocation to attrac- growth of pollination ecology. In the last 15 years
tive structures that are shared by male and female these two fields have gradually merged to produce
function, particularly if they suffer pollen limitation. a new synthesis in plant evolutionary biology whose
Hermaphrodism nevertheless comes with a cost. central theme is the ecology and evolution of plant
VA R I AT I O N A N D E V O L U T I O N O F R E P R O D U C T I V E T R A I T S 181

Box 5.3 The selction mosaic of interactions in Hormathophylla spinosa positive impacts
are represented by solid lines; negative impacts by broken lines (redrawn from
Gómez and Zamora 2000), the bolder the line, the stronger the interaction

Pollinators: nectarivorous ants, pollen collecting

flies (Gomez and Zamora 1992, 1996, 1999)
Trait variation
and fitness
Herbivores: ungulates—mostly Spanish Ibex,
birds, flower herbivores, and seed predators

Trait variation among three populations

Flower density Flower size Flower number Thorn number

35 8 16,000 60
30 7 50
25 6 12,000
20 5 8,000 30
15 4
10 4,000 20
5 3 10
0 2 0 0

Strength and direction of interactions in three populations

Flower density

Flower size Pollinators

A Plant relative
Flower number
Thorn density Herbivores

Flower density

Flower size Pollinators

Flower number Plant relative
Thorn density Herbivores

Flower density

Flower size Pollinators

C Plant relative
Flower number
Thorn density Herbivores

reproduction. The evolution of plant reproduction The possibility that inbreeding depression may
is thus not only where adaptation and diversity vary among populations is particularly pertinent in
meet, but also where an integrative approach based the Mediterranean habitat mosaic where ecological
on ecology and genetics is necessary. In addition, conditions vary dramatically in a localized man-
stochastic factors associated with population history ner (Chapter 4). Studies of Crepis sancta, a ruderal
and colonization dynamics can closely interact with annual common in old-fields, vineyards, and along
natural selection (due to environmental differences roadsides, where it adds a vivid splash of yellow
in resource status and other ecological parameters) to the landscape in spring, illustrate how diverse
to shape patterns of variation and the evolution of ecological conditions can affect the evolution of the
genetic polymorphism. mating system. In a study of several populations
from Mediterranean old-fields of different succes-
sional stage, both drought stress (Cheptou et al.
5.4.2 From outcrossing to selfing
2000a) and increased competition (Cheptou et al.
‘It is well established that, in certain circum- 2000b) have been reported to cause the magnitude of
stances, outbreeding and the consequent promotion inbreeding depression to increase, creating a strong
of genetical variation are adaptively advantageous selection pressure against the evolution of inbreed-
while, in other circumstances, the genetical fix- ing. This may underlie why, in this species, selfing is
ity and reliable seed formation made possible by significant in young successional populations where
inbreeding fit the needs of a population more appro- competition is weak but not in older populations,
priately’ (Baker 1966: 349). Many species have thus which are almost exclusively outcrossing (Cheptou
‘abandoned’ outcrossing for a highly autogamous et al. 2002), and where competition from peren-
mating strategy and there has been a great deal of nial grasses is important feature of the environment.
interest in the causes and biological consequences Spatial heterogeneity of abiotic and biotic environ-
of the evolutionary shift from outcrossing to selfing, mental conditions in the mosaic of Mediterranean
all the more so because of the frequent occurrence of old-fields may thus create the conditions for d