Professional Documents
Culture Documents
Topografia, Umidade Do Solo e Composição Da Vegetação Allan
Topografia, Umidade Do Solo e Composição Da Vegetação Allan
Citation: Moeslund, J. E., L. Arge, P. K. Bøcher, T. Dalgaard, M. V. Odgaard, B. Nygaard, and J.-C. Svenning. 2013.
Topographically controlled soil moisture is the primary driver of local vegetation patterns across a lowland region.
Ecosphere 4(7):91. http://dx.doi.org/10.1890/ES13-00134.1
Key words: Ellenberg indicator values; Europe; NATURA 2000; plant community assembly; plant diversity; solar
radiation; spatial scale; species richness; topographic wetness index; wind.
Received 12 April 2013; revised 31 May 2013; accepted 3 June 2013; published 31 July 2013. Corresponding Editor: D. P.
C. Peters.
Copyright: Ó 2013 Moeslund et al. This is an open-access article distributed under the terms of the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original author and source are credited. http://creativecommons.org/licenses/by/3.0/
5
Present address: Wildlife Ecology and Biodiversity, Department of Bioscience, Aarhus University, Grenåvej 14, DK-8410
Rønde, Denmark.
E-mail: jesper.moeslund@biology.au.dk
the scale dependence of topographic variables (e.g., obvious identification errors). The final
and the underlying mechanisms by which they dataset contained data from 30544 plots distrib-
operate. Our specific study objectives were to uted among the 901 monitoring sites (Fig. 1A),
determine (1) the spatial scale at which relation- and included a total of 1262 vascular plant
ships between topography and local plant species, subspecies and varieties.
diversity and distributions operate, (2) the extent The DK-DEM (‘‘Danmarks højdemodel’’) ele-
to which topography controls local plant species vation model provided LiDAR based elevation
diversity and distribution patterns in natural and data of the study region at 1.6 m horizontal
semi-natural areas across the lowland study area, resolution and 0.10–0.15 m vertical precision
(3) the mechanisms through which topography (Knudsen et al. 2008). In addition to the first
influences vegetation patterns, and (4) site-level order horizontal resolution of 1.6 m, we aggre-
environmental factors that may affect the gated (mean) the model to horizontal resolution
strength and nature of local (within-site) vegeta- scales of 9.6, 49.6, 100.8 and 249.6 m. For
tion–topography relationships. simplicity, these resolutions are referred to
according to respective orders of magnitude as
METHODS the 2, 10, 50, 100 and 250 m resolutions in the text
that follows. Terrain slope, a heat index, potential
Study sites solar radiation, wind exposure and a topograph-
This study included 901 sites (0.16 km2 on ical wetness index (TWI) were calculated at each
average). A majority of the sites are located horizontal resolution scale in order to evaluate
within special areas of conservation protected the impact of both direct and indirect topograph-
under the European Union’s Habitats Directive ic factors. These were used as explanatory
(Council of the European Communities 1992). variables in the statistical analyses conducted in
Each site covered one or more of the natural and/ this study. The text below describes each variable
or semi-natural NATURA 2000 habitat types except the elevation variable (also see Appendix
(Commission of the European communities B).
2002) ranging from dry dunes and grassland to Slope, aspect and potential solar radiation
wet flushes and mires. The sites are distributed layers were computed using the ‘Slope’, ‘Aspect’
throughout Denmark (Fig. 1A) and their plant and ‘Area Solar Radiation’ tools in the ArcGIS 10
species compositions were recorded by the Spatial Analyst module (Environmental Systems
Danish National Monitoring and Assessment Research Institute, Redlands, California, USA).
Programme for the Aquatic and Terrestrial The default settings were used for each tool
Environment (NOVANA; Svendsen and Norup except in the case of the Area Solar Radiation. For
2005; data available via www.naturdata.dk). this tool we selected the time configuration
Table A1 (Appendix A) gives detailed descrip- ‘whole year (2009) with monthly interval’. As
tions of the habitat types included in this study. circular variables do not fit into the statistical
analyses conducted here and to represent heat
Vegetation and topographic data balance, the aspect variable was converted into a
The data used in this study consisted of heat index using the formula (1 cos(h 45))/2,
vegetation records from randomly selected sites where h represents the aspect in degrees
that included 20 to 60 circular sample plots (5-m (McCune and Keon 2002).
radius) georeferenced at 2–5 m precision. Vege- The ‘Hillshade’ tool (Spatial Analyst, ArcGIS
tation sampling procedures included recording 10, Environmental Systems Research Institute,
the type and abundance of all vascular plant Redlands, California, USA) was used to represent
species rooted within the 0.25 m2 quadrat at the topographically controlled wind exposure ac-
center of each plot (Svendsen and Norup 2005). cording to methods described by Mikita and
The vascular plant species data used in this study Klimánek (2010). See Appendix B for details.
was collected from 2004 to 2009 from 37422 plots. To represent topographically controlled effects
Following initial extraction of the data from the on soil moisture, we calculated a topographic
NOVANA database, we removed duplicates and wetness index (TWI) as a proportional measure
entries having incomplete or erroneous values of water retention for a given area (Hengl and
Fig. 1. The distribution of the 901 natural and semi-natural study sites located throughout Denmark underlain
by a digital elevation model used in this study. Panel (A) shows the general strength of the vegetation–
topography relationships (average Spearman r 2 values for all vegetation–topography correlations) at each site.
Panel (B) and (C) show the average Ellenberg indicator values (Ellenberg et al. 2001) for soil moisture and
temperature, respectively.
Reuter 2009). Appendix B explains the TWI 2001) were used to indicate plant species’ pre-
calculation procedures in detail. ferred position along key environmental niche
axes. Using these indicator values to identify
Data and analysis relationships between vegetation and environ-
Local within-site relationships: preparation of ment is a well-known and widely used method
data.—We selected 10 measures of plant diversity in vegetation ecology (Diekmann 2003). Here, we
and composition to assess topographic effects on calculated weighted average (by abundance)
local vegetation patterns among sample plots Ellenberg indicator values per plot and used these
within each study site (i.e., these measures were as measures of functional composition along
used as response variables). These measures individual key environmental niche axes. We
included species richness, a measure of species included values for light (EIVlight), temperature
composition, seven measures of functional com- (EIVtemp), continentality (EIVcont), soil moisture
position (i.e., species preferences along key (EIVmoist), soil reaction (EIVreac), soil nitrogen
environmental niche axes), and a synthetic content (EIVnitr) and soil salinity (EIVsalt). When-
measure of functional composition based on the ever possible, we used the EIV values from Hill et
previous measures of functional composition. al. (1999), since these offered a better fit for the
The species richness measure represented the Danish flora. Ellenberg et al. (2001) for example
number of taxa per plot counting all species, assign the grasses Leymus arenarius (L.) Hochst.
subspecies and varieties consistently identified and Festuca arundinacea Schreb. an EIVsalt value of
within the plot for all years of observation. The 1 (non-saline to periodically slightly saline)
far majority of the observations used were at the whereas Hill et al. (1999) assigned the same
level of species. Plot scores on the first axis of a species an EIVsalt value of 3 (coastal, slightly
Detrended Correspondence Analysis (DCA) of saline soils). These grass species are found almost
the species composition in the plots within a site exclusively within coastal areas in Denmark
provided a measure of the plots’ position on the where soil salinity is undoubtedly higher than
main local floristic gradient (cf. Ejrnæs and that suggested by an EIVsalt value of 1 (i.e., the
Bruun 2000). original values given in Ellenberg et al. 2001). The
Ellenberg indicator values (EIV; Ellenberg et al. wetland species Silene flos-cuculi (L.) Greuter &
Burdet serves as an additional example of the all topographic variables at the five different
higher specificity of the Hill et al. (1999) indicator horizontal resolution scales (45050 models total
values. Ellenberg et al. (2001) assigned this species for all sites). The 10-m horizontal resolution scale
an EIVmoist value of 5 (dampness indicator revealed the strongest links between topography
species) whereas Hill et al. (1999) assigned it a and vegetation (see Results). All further statistical
value of 7 (wet-site indicator species). The latter analyses were therefore conducted using this
value more accurately reflects this species’ prefer- scale of horizontal resolution.
ence for wetlands in Denmark (see Hill et al. 1999 We implemented both parametric and non-
for further details on indicator values). Values for parametric models to assess local relationships
taxa not found in Hill et al. (1999) were taken from between vegetation and topography. Parametric
Ellenberg et al. (2001), or if unavailable therein, tests included multiple OLS regressions of each
from Landolt et al. (2010). Values taken from response variable against the explanatory vari-
Landolt et al. (2010) were adjusted to match ables for each study site (9010 local models in
Ellenberg indicator value ranges (i.e., maximum total). Non-parametric tests included Spearman
Landolt indicator values correspond to maximum correlation statistics calculated between each pair
EIVs and the same for minimum and intermediate of response and explanatory variable (60 pairs,
values). For taxa not listed in the aforementioned 54060 correlation coefficients total). The results
sources, we estimated indicator values primarily identified elevation as one of the most important
using distribution information from Danish floris- topographic variables influencing plant diversity
tic handbooks (Appendix A: Table A3). A and distribution patterns (see Results). Due to the
correlation analysis showed that this approach influence of elevation on all the other topograph-
did not introduce bias to our data: the mean plot ic variables, we removed the elevation variable
EIVs used here correlates highly with mean plot and re-calculated local models to determine
EIVs based solely on data from the literature relationships between vegetation and the remain-
mentioned above (Pearson’s r . 0.99 for almost all ing, more mechanistic variables.
EIVs, Appendix A: Table A4). Subspecies and Cross-regional relationships (all plots).—In order
varieties were treated as unique taxa given that to examine regional vegetation–topography rela-
they typically exhibit divergent environmental tionships, we performed a multiple OLS regres-
preferences. If a subspecies or variety did not sion of each response variable against all
consistently appear each year in a given plot, explanatory variables and the primary habitat
records for these organisms were combined with type using the full set of plot data from all sites
those belonging to the taxonomic rank immedi- combined (10 regional models). For these models,
ately above. Together, these methods allowed us the first axis of a DCA and a PCA including all
to assemble the most accurate measures of plots from the regional dataset represented the
ecological preference (average indicator values) main regional floristic and functional composi-
for the taxa present in each plot. tion. To avoid bias in DCA results introduced by
Plot scores on the first axis of a Principal rare species entries, we omitted plots having less
Component Analysis (PCA) calculated on the than five species. The regional models were also
seven average Ellenberg indicator values for all run without habitat type included in order to
plots within a site represented the plots’ position evaluate the effect of habitat type.
on the main local functional gradient, and were Calculations of Moran’s I for OLS model
used as a synthetic measure of functional residuals revealed a small degree of spatial
composition. autocorrelation within regional models on rela-
Local within-site relationships: analysis proce- tively short distance scales. To address this issue,
dure.—We initially used ordinary least square we calculated a series of analogue Linear Mixed
(OLS) multiple regression modeling to explore Effects (LME) models (Zuur et al. 2009) treating
the relative effects of various horizontal resolu- site as a random effect. This procedure removed
tion scales exerted on vegetation–topography the spatial autocorrelation giving Moran’s I
relationships. This analysis included 50 different ,j60.1j for all model residuals except for the
models for each of the 901 study sites, namely EIVnitr and EIVreac models, for which the Moran’s
regressions of the ten response variables against I still exhibited low values of 0.3–0.45 in the first
distance class (0–20 km). For the regional models, cedures were performed using the R (ver. 2.13.2
we therefore report P-values from LME analyses for 64-bit operating systems) statistical program-
rather than from the OLS models noting that the ming environment (R Development Core Team
two approaches identified the same factors as 2011).
having the strongest influence on vegetation.
Site effects on local relationships.—We assembled RESULTS
a number of measures of the overall environment
at each site. These included the main habitat type The strength of the vegetation–topography
according to NATURA 2000 habitat codes (Table relationships varied with the scale of horizontal
A1 and A2, Appendix A), as well as site means spatial resolution, with the highest adjusted
and standard deviations for the seven average coefficients of determination (R 2adj) at the 10 3
Ellenberg indicator values and for all plot-level 10 m resolution and the lowest R 2adj at the 250 3
topographic explanatory variables. Fig. 1b and 1c 250 m resolution (Fig. 2).
show examples of the average conditions and Topography explained a significant proportion
variability in conditions at sites throughout the of variation in all measures of plant diversity and
study region. To determine whether site-level distribution within sites (local models) as well as
environmental factors affect the strength of local across Denmark (regional models). All the 60
(within-site) vegetation–topography relation- within-site correlation pairs between plant and
ships, we regressed the squared Spearman topography measures were statistically signifi-
correlation coefficients (r 2; following Healey cant in .5% of the sites (i.e., at frequencies above
2006) for five correlation pairs against all of the those expected to randomly occur, Table 1). The
site-level environmental descriptors using OLS average strength (r 2) of the strongest Spearman
multiple regression models. The five correlation correlation in each site was 38% (616% standard
pairs included: (1) the strongest correlation pair deviation; see Fig. C1, Appendix C for examples),
at each site, (2) the DCA first axis plot scores’ indicating a generally moderately strong topo-
correlation with elevation, (3) the PCA first axis graphic effect on the local vegetation. Seventy
plot scores’ correlation with elevation, (4) species percent of the vegetation–topography relation-
richness’ correlation with elevation, and (5) the ships were significantly weaker across the whole
EIVmoist-TWI correlation. The last four pairs study region (regional models) than within sites
cover the most important vegetation measures (local models, Fig. 3).
and topographic variables. Within sites, the relationship between soil
Prior to these analyses, all explanatory vari- moisture preference (EIVmoist) and topography
ables (except for habitat type) were checked for was always among the strongest (Table 1, Fig. 3).
multicollinearity. Those having a tolerance In addition, strong and consistent local relation-
(Quinn and Keough 2002) below 0.1 were not ships between topography and the most impor-
included in the analysis. Hence, the following tant local floristic and functional gradients (DCA
measures (where l indicates a mean and r and PCA first axis plot scores, respectively) were
indicates a standard deviation) were used as also evident (Table 1, Fig. 3). While all topo-
explanatory variables: EIVlight (l, r), EIVtemp (l, graphic variables had statistically significant
r), EIVcont (l, r), EIVmoist (l, r), EIVreac (r), relations to the plant measures, elevation and
EIVnitr (r), EIVsalt (r), potential solar radiation slope had the strongest effects (Table 1). Similar
(l), wind exposure (l), elevation (l), heat index to the local relationships, EIVmoist also exhibited
(l, r) and TWI (l ,r). Moran’s I calculations the strongest relationship with topography
using model residuals (Legendre and Legendre across the whole study region (Fig. 3). In
1998) demonstrated the absence of spatial auto- contrast, relationships between the main regional
correlation for all 60 models. floristic and functional gradients (DCA and PCA
Statistical issues and software.—We applied log- first axis plot score) and topography were much
or square-root transformations as needed to weaker than their local relationships (Fig. 3).
fulfill assumptions regarding the normal distri- The habitat-specific patterns in the local
bution of model residuals and the validity of vegetation–topography relationships were large-
associated P-values. All statistical analysis pro- ly consistent with the overall local patterns just
Fig. 2. A summary of the R 2adj values for the local (within-site) OLS multiple regression models linking each of
the ten response variables to all topographic variables at the five resolutions. Upper extreme, upper quartile,
median, lower quartile and lower extreme are represented for each box. Circles mark the means. Different letters
denote significantly different average R 2adj values (Student’s t-test, P , 0.05). The number of statistically
significant models (out of 9010) for each resolution scale is shown next to each bar.
Table 1. Average squared Spearman’s correlation coefficients (r 2) for topography and vegetation within each of
the 901 natural and semi-natural sites distributed throughout Denmark.
Fig. 3. Coefficients of determination (R 2adj) for each vegetation measure from the regional-scale models
(combining plot data excluding habitat type from all 901 study sites across Denmark) and from the within-site
(local) models (including elevation as explanatory variable). For the local models 1r error bars are shown.
Abbreviations next to each bar identify statistically significant (P , 0.05) explanatory variables. Ele, Elevation;
Heat, Heat Index; Slo, Slope; Rad, Potential Solar Radiation; TWI, Topographic Wetness Index; Wind, Wind
Index. Asterisks indicate that the local model mean R 2adj was significantly greater (Student’s t-test, P , 0.05) than
that of the regional model. See Table 1 for the remaining abbreviations and Fig. C2 and C3 (Appendix C) for
results for the regional models including habitat as explanatory variable and for the local models excluding
elevation as explanatory factor. The local-model-R 2adj shown here (i.e., including elevation as explanatory
variable) were in all cases significantly different from models not including elevation (Student’s t-test, P , 0.05).
Fig. 4. Habitat-specific average Spearman correlation (vegetation–topography) r 2 values for (A) the two
strongest topographic determinants of vegetation measures and (B) the two vegetation measures exhibiting the
strongest topographic relationships. Panel (C) shows habitat specific r 2 values for pairs of variables exhibiting (1)
the highest overall correlation and (2) the highest correlation between functional variables (i.e., excluding
elevation and slope, given their mainly indirect effects). The color codes from panel (A) and (B) are used in (C) to
denote the correlation in question (e.g., blue-blue ¼ TWI-EIVmoist). A plus denotes a positive relationship, a minus
denotes a negative one. See Table 1 for abbreviations.
described, but with some variability at least vegetation–topography relationships, with the
partially related to specific environmental condi- strongest relationships in wet habitats (Table 2).
tions in certain habitats, e.g., the uniquely strong Correspondingly, the site-means of TWI and
relationship between salt tolerance (EIVsalt) and secondarily EIVmoist had the clearest effects on
topography in salt meadows (Fig. 4). When the local relationships (Table 3).
considering only the more functional variables
(i.e., not elevation and slope), the strongest local DISCUSSION
relationships in all habitats always included at
least one of the hydrology-related variables, TWI The presented cross-scale and cross-habitat
or EIVmoist, and often both (Fig. 4C). analysis of vegetation–topography relationships
Environmental site characteristics such as based on .30500 plots and very high-resolution
habitat type affected the strength of the local topographic data provides the first geographi-
Table 2. An overview of the NATURA 2000 habitat slope and TWI having the strongest relation-
types as well as their frequencies observed from the ships. Hydrology-related variables exhibited the
901 natural and semi-natural study sites located strongest correlation coefficients among the more
throughout Denmark. functional topographic and plant-related vari-
ables across all habitat types. Topography,
No. sites
Habitat type included Average R 2adj furthermore, exerted the strongest influence on
Salt meadows 95 0.26A local vegetation–topography relations in wet
Xeric calcareous grasslands 6 0.20ABC habitats, although it was also important in dry
Humid dune slacks 41 0.20B habitats.
Acidic bogs 78 0.19B
Grasslands (unspecified) 31 0.19BC
Molinia meadows 58 0.17BC Scale effects on vegetation–topography
Calcareous fens 160 0.16BC
Wet heaths 39 0.16BC relationships
Calcareous grasslands 80 0.15BC Spatial scale affects ecological patterns and
Pioneer communities 11 0.15BCD
Forests 13 0.15BCD processes (Levin 1992). However, multiscale
Acidic grasslands 96 0.15C ecological studies are relatively rare (Dalgaard
Calcareous dunes 68 0.14C
Xeric heaths 89 0.10D
et al. 2003, Moser et al. 2007), probably due to the
Relatively dry acidic dunes 65 0.09D lack of high-resolution data covering large
Notes: See Table A1 and A2, Appendix A for further details. geographic areas. We found that the topographic
The habitat-specific average R 2adj values were calculated from influence on local plant diversity and distribution
OLS multiple regression models running each vegetation
measure against all the topography measures for local sites.
patterns became less evident at horizontal
The R 2adj values represent the general strength of the resolutions coarser than 10 m. Hence, our multi-
vegetation–topography relationships for a given habitat. scale modeling suggests that the mechanisms
Different letters indicate significantly different mean R 2adj
values (Tukey’s Honestly Significant Difference test, P , 0.05). underlying the local vegetation–topography re-
lationships presented here operate primarily at
the 10 m scale. The weaker vegetation–topogra-
cally comprehensive assessment of the impor- phy relationships of the 2 3 2 m models may,
tance of topography as a determinant of local however, reflect incomplete sampling: a 2 3 2 m
plant diversity and distribution patterns. Topo- area located in the center of the plot would not
graphic variables had the strongest explanatory necessarily capture the topographic variation
power in models scaled at 10 3 10 m horizontal affecting the vegetation in a larger vegetation
resolution and the weakest in models at 250 3 sample-plot like the ones used here. That local
250 m. All topographic variables were shown to (within-site) vegetation–topography relation-
influence vegetation patterns, with elevation, ships were generally stronger than cross-region
Table 3. Results from OLS multiple regression models linking the correlation strength of five selected correlation
pairs to site-level environmental variables for the 901 study sites located throughout Denmark.
Environmental variables
2
Response variable Full model R adj Mean TWI SD of EIVmoist Mean EIVtemp
Strongest correlation pair at each site 0.14 0.18 0.11 0.14
EIVmoist-TWI 0.10 0.06 0.31 0.10
DCA-elev 0.10 0.24 0.09 0.07
PCA-elev 0.09 0.14 0.07 0.07
SpRich-elev 0.06 0.24 0.07 0.15
Mean absolute regression coefficient 0.17 0.13 0.11
Notes: The five response variables are the strength (r 2) of (1) the strongest correlation pair (between a vegetation and a
topography variable) within each site, and (2–5) four within-site Spearman correlation pairs selected to cover the most
important vegetation measures and topographic variables. For each response variable, the R 2adj is given for the full (also
including environmental variables not shown here) multiple regression model along with standardized regression coefficients
for each of the three most important (see below) environmental variables. The mean absolute standardized regression
coefficient for the five response variables represents the general importance of each of the environmental variables in
determining the strength of the local (within-site) vegetation–topography relationships. TWI, topographic wetness index. See
Table 1 for the remaining abbreviations. For full results including all environmental variables used in this study, see Table C2,
Appendix C.
relationships, probably reflects that other factors thus reflect productivity- and/or species-pool-
such as climate become more important for related mechanisms (Zobel 1997, Mittelbach et al.
vegetation patterns at regional extents (Levin 2001).
1992).
Underlying mechanisms for vegetation–topography
Topographic control of vegetation patterns relationships
Generally topography was important for local In this study, elevation itself exerted the
plant diversity and distribution patterns. It strongest influence on local vegetation patterns,
consistently influenced the plant’s average pref- followed by slope, TWI, potential solar radiation,
erence for moisture (EIVmoist), resembling previ- wind exposure and heat index. The role of
ous studies, which have observed strong elevation in a low relief region (Denmark’s
relationships between local topography and the highest point: 171 m above sea level) is surprising
soil moisture affinity of different plant species since altitude is mostly considered to influence
(Vivian-Smith 1997, Silvertown et al. 1999, Øk- vegetation patterns through temperature effects
land et al. 2008, Raulings et al. 2010). Topography as seen in mountains (Grytnes 2003, but see
was also strongly and consistently linked to the Lenoir et al. 2013). Broadly, lowland areas
most important local floristic and functional experience only short temperature gradients
gradients (DCA and PCA), as expected from given standard temperature lapse rates (Calvert
niche-based species sorting (Leibold et al. 2004). 1990, Sanders and Rahbek 2012). The relative
Other factors such as dispersal, history, compe- importance of elevation evident in the current
tition, facilitation, and stochasticity have also study probably reflects: (1) its effect on niche
been suggested as important for local plant gradients important to plant species, (2) that
diversity (Wilson 2011). These factors may these gradients and their role for plant growth
explain some of the variation that topography may vary among sites and (3) that the nature of
could not explain. However, the aforementioned such gradients may be complex due to geo-
clear link between topography and the main graphic factors. Elevation represents both slope
floristic gradient suggests that topographically (Burbank 1992) and soil moisture (TWI; Wilson
controlled niche-based sorting generally is the and Gallant 2000). Furthermore, elevation corre-
predominant driver of local plant community lates with soil salinity in coastal settings (Moe-
assembly across habitat types in the lowland slund et al. 2011) and with pH in areas underlain
study region. by limestone, which is common in many areas of
Topography was also consistently, but less Denmark (Thomsen 1995).
strongly linked to local species richness patterns, Studies of plant diversity especially those
especially in salt meadows, unspecified grass- conducted locally have treated slope as a
lands, forests and xeric heaths. These results predictive variable (Gottfried et al. 1998, Jones
demonstrate that topography structures local et al. 2011). This factor typically explains a
plant species richness patterns not only in saline relatively high proportion of the observed vari-
(Moeslund et al. 2011) and freshwater wetland ation in vegetation patterns (Luoto et al. 2002,
plant communities (Silvertown et al. 1999, Øk- Bennie et al. 2006). The present research therefore
land et al. 2008, Raulings et al. 2010), but also in agrees with previous studies. Resembling eleva-
other habitat types. The present study is the first tion, slope is a non-functional variable (Guisan
to identify a link between local species richness and Zimmermann 2000). It may capture eleva-
and topography in dry habitats generally. Be- tion variation within an area and thus relate to
sides soil moisture, other topographically depen- the potential microhabitat variation occurring
dent features such as soil pH and fertility may here. Given that such variation and species
explain this finding: in the current study, richness is often tightly linked (Vivian-Smith
topography influenced average plant preferences 1997), the positive correlation between slope and
for soil pH and nitrogen in a manner similar to species richness in this study supports this
how it affects species richness in xeric heaths and interpretation. Slope also affects the amount of
unspecified grasslands (not shown). These topo- solar radiation received (Gottfried et al. 1998,
graphic effects on local richness patterns could Bennie et al. 2008), soil moisture (Wilson and
Gallant 2000) and parameters such as erosion influence on local vegetation patterns. Potential
rates, litter accumulation and biological dynam- solar radiation exerted a stronger influence than
ics (Troeh and Thompson 2005, Gomes de Freitas the heat index (aspect) whereas it remained less
et al. 2012). Together, these factors may explain than that of the slope variable, in spite of the fact
why slope was one of the most important factors that both aspect and slope is considered in the
for local vegetation patterns. calculation of potential solar radiation. Given the
TWI was the third strongest variable affecting contributions of these factors, potential solar
plant diversity and distribution patterns and the radiation should theoretically exert greater influ-
second strongest determinant of plant preferenc- ence on vegetation patterns than that exerted by
es for moist (EIVmoist). TWI has been shown to slope and aspect individually. However, our
provide fairly accurate local estimates of the results controvert these expectations and instead
average EIV for moisture (Kopecký and Čı́žkova comply with studies showing that simple topo-
2010) as well as of actual measurements of graphic parameters reflect local thermal climate
groundwater and soil moisture levels (Sørensen better than potential solar radiation in humid
et al. 2006). Numerous studies have demonstrat- temperate regions due to the effects of cloud
ed the role of soil moisture in determining plant cover in obscuring solar radiation estimates
species distributions and community structure (Reger et al. 2011).
within single habitats and/or small areas (e.g., We observed only weak associations between
Økland et al. 2008, Raulings et al. 2010). Silver- aspect-determined heat balance and vegetation
town et al. (1999) and Araya et al. (2011) for patterns indicating that the heat index exerted
example showed the influence of hydrologic very little influence relative to other topographic
factors in 2–18 meadow sites in England and 8 variables. Additionally, the index exhibited neg-
fynbo sites in South Africa. To our knowledge, no ative correlation with the average preferences for
previous study has investigated the role of temperature and light contrary to expectations.
topographically controlled soil moisture across Aspect-determined heat therefore did not appear
a wide range of habitats and sites in order to to be substantial in determining local vegetation–
determine its general importance for local vege- topography relationships within the study region
tation patterns. Our results indeed show that even though it does appear to influence these
topographically controlled soil moisture exerts a relations quite a bit in other studies (Boyko 1947).
strong and ubiquitous influence on these pat- The complexity of interactions between topog-
terns. TWI was the most important functional raphy and wind has prevented successful mod-
variable affecting local vegetation patterns across eling and rigorous study of the relationships
all habitats. This suggests that local topography– between topographically controlled wind re-
soil moisture dynamics serve as the primary gimes and local vegetation patterns (but see
mechanism underlying topographic control of Knapp 1985, Chen et al. 1997, Ennos 1997, Lortie
local vegetation patterns across a number of and Cushman 2007, Li et al. 2009). In the current
different habitat types, and over large areas. study, wind exposure exhibited only weak
Hence, the topographically affected niche axes correlation to these patterns. Nonetheless, the
generally important for local plant community correlations between wind exposure and average
assembly seems to be soil moisture related across soil moisture preference (EIVmoist) were among
habitat types in the study region. the strongest observed for forests as well as xeric
Results from this study indicate that potential and mesic calcareous grasslands in this study.
solar radiation does not strongly influence local Elevated levels of evapotranspiration and soil
plant diversity and distributions. Nevertheless, it evaporation caused by intensified wind exposure
did influence vegetation patterns in xeric calcar- (Knapp 1985) provide a reasonable explanation
eous and acidic grasslands, indicating that as a for these observations. Under such circumstanc-
community-assembly mechanism, it primarily es, we would expect a negative correlation
operates in relatively dry semi-natural areas between wind exposure and EIVmoist. The afore-
(Ejrnæs and Bruun 2000). Hence, under certain mentioned habitats affirmed this expectation in
circumstances, solar radiation may contribute showing strong negative correlations between
(along with hydrologic factors) to topography’s wind and the EIVmoist variable (in contrast to its
positive and weak relationship observed in other inventory allowed us to comprehensively assess
habitats). Wind-driven desiccation is recognized the general role played by topography in
as an important factor in determining local plant structuring plant communities across habitats
species distributions in forests (Ellenberg 1988). and scales in a large lowland region. Topograph-
ically controlled factors such as temperature are
Importance of topography well-known determinants of vegetation patterns
in different environments in mountainous regions. Surprisingly, our results
The results given here indicate that the degree showed that topography also consistently influ-
to which topography influences vegetation de- ences vegetation patterns at local scales in natural
pends on site-level environmental parameters and semi-natural lowland habitats. Topography’s
including habitat type and moisture conditions control of soil moisture patterns rather than
(as indicated by the site-mean TWI). Notably, the temperature emerged as the primary mechanism
strongest relationships between topographic fac- underlying the observed lowland vegetation–
tors and vegetation were observed in wet topography relationships, even within dry hab-
habitats, and particularly so in tidal salt mead- itat sites. Overall, our findings suggest that niche-
ows, while the weakest relationships occurred in based species sorting along topographically
dry, sandy habitats and forest. The strength of controlled gradients in soil moisture are among
species-environment relationships depends on the primary local plant community assembly
the length of the environmental gradients under mechanisms across large lowland regions.
investigation (Austin 1985). As the hydrologic
gradients in wetlands is longer than in other ACKNOWLEDGMENTS
habitats (Adam 1990), this explains the stronger
relationships there, i.e., when considering the We gratefully acknowledge funding from the Aar-
evidence that topography’s control of soil mois- hus University Research Foundation via the Center for
ture is the primary mechanism underlying the Interdisciplinary Geospatial Informatics Research (CI-
GIR), the Danish Strategic Research Council, Center for
local vegetation–topography relationships.
Massive Data Algorithmics, a Center of the Danish
Sandy soil types have a relatively low water National Research Foundation, and the Oticon Foun-
retention capacity (Troeh and Thompson 2005), dation (grant to JEM).
reducing the influence of topography on soil
moisture in ecosystems that depend on infiltra- LITERATURE CITED
tion from precipitation. This can explain the
relatively weak vegetation–topography relations Adam, P. 1990. Saltmarsh ecology. Cambridge Univer-
observed in dry, sandy habitats. We note how- sity Press, Cambridge, UK.
ever, that in groundwater-fed ecosystems, topo- Araya, Y. N., J. Silvertown, D. J. Gowing, K. J.
graphically controlled hydrology may play a McConway, H. Peter Linder, and G. Midgley.
major role for vegetation patterns even on sandy 2011. A fundamental, eco-hydrological basis for
niche segregation in plant communities. New
soils (Araya et al. 2011).
Phytologist 189:253–258.
Albeit the vegetation–topography relation- Augustine, D. 2003. Spatial heterogeneity in the
ships were weaker within dry sites generally, herbaceous layer of a semi-arid savanna ecosystem.
TWI was still the most important functional Plant Ecology 167:319–332.
variable in determining vegetation patterns even Austin, M. 1985. Continuum concept, ordination
in the relatively dry dunes and heathlands. methods, and niche theory. Annual Review of
Hence, our results suggest that topographically Ecology and Systematics 16:39–61.
controlled soil moisture is an important structur- Barbour, M. G., R. B. Craig, F. R. Drysdale, and M. T.
ing component of local vegetation patterns in Ghiselin. 1974. Coastal ecology: Bodega Head.
University of California Press, Berkeley, California,
both wet and dry habitats.
USA.
Bennie, J., M. O. Hill, R. Baxter, and B. Huntley. 2006.
CONCLUSIONS Influence of slope and aspect on long-term vege-
tation change in British chalk grasslands. Journal of
New high-resolution LiDAR-based digital ele- Ecology 94:355–368.
vation models and a massive vegetation plot Bennie, J., B. Huntley, A. Wiltshire, M. O. Hill, and R.
Baxter. 2008. Slope, aspect and climate: spatially Relationships between soil characteristics, topogra-
explicit and implicit models of topographic micro- phy and plant diversity in a heterogeneous
climate in chalk grassland. Ecological Modelling deciduous broad-leaved forest near Beijing, China.
216:47–59. Plant and Soil 261:47–54.
Boose, E. R., D. R. Foster, and M. Fluet. 1994. Gomes de Freitas, C., F. R. Costa, Capellotto, J.-C.
Hurricane impacts to tropical and temperate forest Svenning, and H. Balslev. 2012. Topographic
landscapes. Ecological Monographs 64:369–400. separation of two sympatric palms in the central
Boyko, H. 1947. On the role of plants as quantitative Amazon—Does dispersal play a role? Acta Oeco-
climate indicators and the geo-ecological law of logica 39:128–135.
distribution. Journal of Ecology 35:138–157. Gottfried, M., H. Pauli, and G. Grabherr. 1998.
Burbank, D. W. 1992. Characteristic size of relief. Prediction of vegetation patterns at the limits of
Nature 359:483–484. plant life: a new view of the alpine-nival ecotone.
Calvert, J. G. 1990. Glossary of atmospheric chemistry Arctic and Alpine Research 30:207–221.
terms. Pure and Applied Chemistry 62:2167–2219. Gould, W. A., and M. D. Walker. 1997. Landscape-scale
Cappelen, J., and B. Jørgensen. 1999. Observed wind patterns in plant species richness along an arctic
speed and direction in Denmark—with climatolog- river. Canadian Journal of Botany 75:1748–1765.
ical standard normals, 1961–90. Report no. 99-13. Grace, J. B. 1999. The factors controlling species density
Danish Meteorological Institute, Copenhagen, Den- in herbaceous plant communities: an assessment.
mark. Perspectives in Plant Ecology, Evolution and
Chen, Z. S., C. F. Hsieh, F. Y. Jiang, T. H. Hsieh, and I. F. Systematics 2:1–28.
Sun. 1997. Relations of soil properties to topogra- Grytnes, J. A. 2003. Species-richness patterns of
phy and vegetation in a subtropical rain forest in vascular plants along seven altitudinal transects
southern Taiwan. Plant Ecology 132:229–241. in Norway. Ecography 26:291–300.
Christiansen, M. S. and H. Anthon. 1970. Danmarks Guisan, A., and N. E. Zimmermann. 2000. Predictive
Vilde Planter. Branner og Korch, Copenhagen, habitat distribution models in ecology. Ecological
Denmark. Modelling 135:147–186.
Commission of the European Communities. 2002. Healey, J. F. 2006. The essentials of statistics: a tool for
Commission working document on NATURA social research. Wadsworth, Belmont, California,
2000. The Commission of the European Commu- USA.
nities, EU. Hengl, T., and H. I. Reuter. 2009. Geomorphometry:
Council of the European Communities. 1992. Council concepts, software, applications. Elsevier, Amster-
Directive 92/43/EEC of 21 May 1992 on the dam, The Netherlands.
conservation of natural habitats and of wild fauna Hill, M. O., J. O. Mountford, D. B. Roy, and R. G. H.
and flora. Report no. 206. The council of the Bunce. 1999. Ellenberg’s indicator values for British
European Communities, EU. plants. ECOFACT 2a Technical Annex. Centre for
Crawford, D. J. 1975. Systematic relationships in the Ecology and Hydrology, National Environmental
narrow-leaved species of Chenopodium of the Research Council, Swindon, UK.
western United States. Brittonia 27:279–288. Jones, M. M., B. Szyska, and M. Kessler. 2011.
Dalgaard, T., N. J. Hutchings, and J. R. Porter. 2003. Microhabitat partitioning promotes plant diversity
Agroecology, scaling and interdisciplinarity. Agri- in a tropical montane forest. Global Ecology and
culture, Ecosystems & Environment 100:39–51. Biogeography 20:558–569.
Diekmann, M. 2003. Species indicator values as an Knapp, A. K. 1985. Early season production and
important tool in applied plant ecology—a review. microclimate associated with topography in a C4
Basic and Applied Ecology 4:493–506. dominated grassland. Acta Oecologica 6:337–345.
Ejrnæs, R., and H. H. Bruun. 2000. Gradient analysis of Knudsen, T., R. C. Andersen, N. S. Dalå, S. L.
dry grassland vegetation in Denmark. Journal of Kokkendorff, B. P. Olsen, B. C. Rosenkranz, and
Vegetation Science 11:573–584. M. Wind. 2008. DK-DEM: one name—four prod-
Ellenberg, H. 1988. Near-natural woods and thickets. ucts. Page 4 in T. Knudsen and B. P. Olsen, editors.
Pages 43–283 in H. Ellenberg, editor. Vegetation Proceedings of the Second NKG workshop on
ecology of Central Europe. Cambridge University national DEMs, Copenhagen, November 11–13,
Press, Cambridge, UK. 2008. National Survey and Cadastre, Copenhagen,
Ellenberg, H., H. E. Weber, R. Düll, V. Wirth, and W. Denmark.
Werner. 2001. Zeigerwerte von planzen in Mitte- Kopecký, M., and Š. Čı́žková. 2010. Using topographic
leuropa. Erich Goltze, Göttingen, Germany. wetness index in vegetation ecology: does the
Ennos, A. R. 1997. Wind as an ecological factor. Trends algorithm matter? Applied Vegetation Science
in Ecology & Evolution 12:108–111. 13:450–459.
Fu, B. J., S. L. Liu, K. M. Ma, and Y. G. Zhu. 2004. Landolt, E. et al. 2010. Flora indicativa—ökologische
zeigerwerte und biologische kennzeichen zur flora Norway: the role of surface microtopography.
der Schweiz und der Alpen. Haupt Verlag AG, Journal of Vegetation Science 19:67–74.
Bern, Switzerland. Olivero, A. M., and D. M. Hix. 1998. Influence of aspect
Legendre, P., and L. Legendre. 1998. Numerical and stand age on ground flora of southeastern
ecology. Elsevier, New York, New York, USA. Ohio forest ecosystems. Plant Ecology 139:177–187.
Leibold, M. A. et al. 2004. The metacommunity Pausas, J. G., and J. Carreras. 1995. The effect of
concept: a framework for multi-scale community bedrock type, temperature and moisture on species
ecology. Ecology Letters 7:601–613. richness of Pyrenean scots pine (Pinus sylvestris L.)
Lenoir, J. et al. 2013. Local temperatures inferred from forests. Plant Ecology 116:85–92.
plant communities suggest strong spatial buffering Pedersen, H. Æ., and N. Faurholdt. 2010. Danmarks
of climate warming across Northern Europe. Vilde Orkideer. Gyldendal, Copenhagen, Denmark.
Global Change Biology 19:1470–1481. Quinn, G. P., and M. J. Keough. 2002. Multiple and
Levin, S. A. 1992. The problem of pattern and scale in complex regression. Pages 111–154 in G. P. Quinn
ecology: the Robert H. MacArthur Award lecture. and M. J. Keough, editors. Experimental design
Ecology 73:1943–1967. and data analysis for biologists. Cambridge Uni-
Li, F., L. Zhao, H. Zhang, J. Liu, H. Lu, and L. Kang. versity Press, Cambridge, UK.
2009. Habitat degradation, topography and rainfall R Development Core Team. 2011. R: A language and
variability interact to determine seed distribution environment for statistical computing. R Founda-
and recruitment in a sand dune grassland. Journal tion for Statistical Computing, Vienna, Austria.
of Vegetation Science 20:847–859. Raulings, E. J., K. Morris, M. C. Roache, and P. I. Boon.
Lortie, C. J., and J. H. Cushman. 2007. Effects of a 2010. The importance of water regimes operating at
directional abiotic gradient on plant community small spatial scales for the diversity and structure
dynamics and invasion in a coastal dune system. of wetland vegetation. Freshwater Biology 55:701–
Journal of Ecology 95:468–481. 715.
Luoto, M., T. Toivonen, and R. Heikkinen. 2002. Reger, B., C. Kölling, and J. Ewald. 2011. Modelling
Prediction of total and rare plant species richness effective thermal climate for mountain forests in
in agricultural landscapes from satellite images and the Bavarian Alps: Which is the best model?
topographic data. Landscape Ecology 17:195–217. Journal of Vegetation Science 22:677–687.
McCune, B., and D. Keon. 2002. Equations for potential Sanders, N. J., and C. Rahbek. 2012. The patterns and
annual direct incident radiation and heat load. causes of elevational diversity gradients. Ecogra-
Journal of Vegetation Science 13:603–606. phy 35:1–3.
Mikita, T., and M. Klimánek. 2010. Topographic Schou, J.-C. 2001. Danmarks Høgeurter (Hieracium,
exposure and its practical applications. Journal of Pilosella). Aarhus University Press, Aarhus, Den-
Landscape Ecology 3:42–51. mark.
Mittelbach, G. G., C. F. Steiner, S. M. Scheiner, K. L. Schou, J.-C. 2006. Danmarks Halvgræsser. BFN’s
Gross, H. L. Reynolds, R. B. Waide, M. R. Willig, forlag, Thisted, Denmark.
S. I. Dodson, and L. Gough. 2001. What is the Schou, J.-C., P. Wind, and S. Lægaard. 2009. Danmarks
observed relationship between species richness and Græsser. BFN’s forlag, Thisted, Denmark.
productivity? Ecology 82:2381–2396. Shmida, A., and M. V. Wilson. 1985. Biological
Moeslund, J. E., L. Arge, P. Bøcher, B. Nygaard, and J. determinants of species diversity. Journal of Bioge-
Svenning. 2011. Geographically comprehensive ography 12:1–20.
assessment of salt-meadow vegetation-elevation Silvertown, J., M. E. Dodd, D. J. G. Gowing, and J. O.
relations using LiDAR. Wetlands 31:471–482. Mountford. 1999. Hydrologically defined niches
Moser, K., C. Ahn, and G. Noe. 2007. Characterization reveal a basis for species richness in plant
of microtopography and its influence on vegetation communities. Nature 400:61–63.
patterns in created wetlands. Wetlands 27:1081– Sørensen, R., U. Zinko, and J. Seibert. 2006. On the
1097. calculation of the topographic wetness index:
Mossberg, B. and L. Stenberg. 2005. Den Nye Nordiske evaluation of different methods based on field
Flora. Gyldendal, Copenhagen, Denmark. observations. Hydrology and Earth System Scienc-
O’Callaghan, J. F., and D. M. Mark. 1984. The es 10:101–112.
extraction of drainage networks from digital Svendsen, L. M., and B. Norup. 2005. NOVANA.
elevation data. Computer Vision Graphics and Nationwide monitoring and assessment pro-
Image Processing 28:323–344. gramme for the aquatic and terrestrial environ-
Oke, T. R. 1987. Boundary layer climates. Second ments. Programme description—part 1. NERI
edition. Methuen, London, UK. Technical Report No. 532. National Environmental
Økland, R. H., K. Rydgren, and T. Økland. 2008. Research Institute, Ministry of the Environment,
Species richness in boreal swamp forests of SE Denmark.
Svenning, J., D. Harlev, M. Sørensen, and H. Balslev. ity and floristic diversity in experimental wetland
2009. Topographic and spatial controls of palm communities. Journal of Ecology 85:71–82.
species distributions in a montane rain forest, Wilson, J. P. and J. C. Gallant. 2000. Terrain analysis:
southern Ecuador. Biodiversity and Conservation principles and applications. Wiley, New York, New
18:219–228. York, USA.
Thomsen, E. 1995. Kalk og kridt i den danske under- Wilson, J. B. 2011. The twelve theories of co-existence
grund. Pages 31–67 in O. B. Nielsen, editor. in plant communities: the doubtful, the important
Danmarks geologi fra Kridt til i dag. Aarhus and the unexplored. Journal of Vegetation Science
University, Aarhus, Denmark. 22:184–195.
Troeh, F. R., and L. M. Thompson. 2005. Soils and soil Zobel, M. 1997. The relative role of species pools in
fertility. Blackwell, Oxford, UK. determining plant species richness: an alternative
Vierling, K. T., L. A. Vierling, W. A. G. Gould, S. explanation of species coexistence. Trends in
Martinuzzi, and R. M. Clawges. 2008. LiDAR: Ecology and Evolution 12:266–269.
shedding new light on habitat characterization and Zuur, A. F., E. N. Leno, N. J. Walker, A. A. Saveliev,
modeling. Frontiers in Ecology and the Environ- and G. M. Smith. 2009. Mixed effects models and
ment 6:90–98. extensions in ecology with R. Springer, New York,
Vivian-Smith, G. 1997. Microtopographic heterogene- New York, USA.
SUPPLEMENTAL MATERIAL
APPENDIX A
Table A1. An overview of the habitat types included in this study (a subset of Annex 1 of the EU Habitats
Directive; Council of the European Communities 1992).
Table A3. An overview of plants species that were assigned at least one Ellenberg indicator value (Ellenberg et al.
2001) by the authors of this study (bold numbers).
Indicator value r
EIVlight 0.999
EIVtemp 0.987
EIVcont 0.767
EIVmoist 0.998
EIVreac .0.999
EIVnitr 0.998
EIVsalt .0.999
Notes: The table is based on the plot mean indicator values
for the first 100 sites in the dataset embracing 3878 plots
equally covering all habitat types included in the study.
APPENDIX B
Detailed descriptions of the calculation of the wind polating (Inverse Distance Weighting, IDW) data
exposure and the topographic wetness index from 62 weather stations distributed throughout
Denmark (Cappelen and Jørgensen 1999). The
Wind exposure.—The ‘‘Hillshade’’ tool in the Spa- IDW calculations were performed using the
tial Analyst module in ArcGIS 10 creates a ‘‘IDW’’ tool in the ‘‘Geostatistical Analyst’’
shaded relief from a DEM by considering the module in ArcGIS 10 with default settings
azimuth (compass direction) and altitudinal (default power ¼ 2) except that we used an
angles of an artificial illumination source. By output cell size of 250 m since the general wind
using an altitude angle of 58 hillshade addition- speed and direction is unlikely to change notably
ally represents the potential wind exposure from over such short distances. These two maps were
a given direction (Boose et al. 1994, Mikita and used for weighting (equal weights) the potential
Klimánek 2010). Using this tool and wind speed impact by wind as represented by the corre-
and direction data for Denmark we created a sponding hillshade layers, allowing us to com-
map of potential wind exposure for all Denmark. pute a map of weighted-average potential wind
Initially, one hillshade raster was created for each exposure for all of Denmark.
of the following compass directions: 08, 308, 608,
908, 1208, 1508, 1808, 2108, 2408, 2708, 3008 and Topographic wetness index.—The TWI of a DEM
3308. In Denmark wind is not equally distributed Cell (i ) is given by ln(As/tan(h i )), where As is the
among the compass directions (Cappelen and specific catchment area associated with i (de-
Jørgensen 1999). To account for this we incorpo- tailed below) and hi is the slope angle of i
rated data on general wind speed and direction: (Wilson and Gallant 2000). As is defined as Au/Li,
for each compass direction a map of average where Au is the area of the upslope cells draining
wind speed and a map of frequency for that into i and Li is the dimension of i measured as
particular wind direction were created by inter- the width of the cell in the direction orthogonal
to the water flow direction. The calculations of ogy software package version 1.0.5 (SCALGO,
Au and Li are typically based on outputs from Aarhus, Denmark) was used for these opera-
flow accumulation and flow direction computer tions. Based on these the TWI was computed in
algorithms. Here, we used the output of a single-
flow direction algorithm (D8) (Wilson and the ArcGIS 10 environment using the above
Gallant 2000) for the subsequent run of a flow formula. For a flowchart showing all TWI
accumulation algorithm. The SCALGO Hydrol- calculation details, see Fig. B1.
Fig. B1. Flowchart of the procedure used for calculation of the topographic wetness index (TWI).
Below, the steps requiring further explanation are detailed. Steps initiated by the word SCALGO was
performed using the SCALGO Hydrology software package version 1.0.5 (SCALGO, Aarhus, Denmark). The
remaining steps were performed using ArcGIS 10 (Environmental Systems Research Institute, Redlands,
California, USA). Blue boxes are input data, yellow ovals represent calculation procedures, green boxes are
output data sets, and transparent boxes are output data sets not used here.
SCALGO SFD flow directions: This module models a water flow regime where the water is allowed to flow in
only one direction, namely to the lowest lying downslope neighbor cell relative to the initial cell (O’Callaghan
and Mark 1984).
SCALGO flow accumulation: This module constructs a flow accumulation raster where each cell is assigned a
value proportional to the amount of water that potentially reaches it from all cells draining into it according to the
given flow direction method (cf. above).
3 3 3 focal minimum: For each cell the lowest value of the 8 neighboring cells is allocated. This is taken to be the
elevation value of the cell to which the flow is directed (as calculated by SFD-flow). The focal statistics tool in
ArcGIS was used for this operation.
Calculate DZ: The elevation difference between each raster cell and the cell to which it flows (under an SFD
regime) is computed by subtracting the output of ‘‘3 3 3 focal minimum’’ from the DEM.
Calculate cell dimension (L): For the cardinal directions the value is equal to the length of the cell edge; for the
diagonal directions the value is equal to the diagonal length of the cell.
Calculate catchment size (number of cells): This is equal to the flow accumulation value plus 1 (the cell itself ).
Calculate catchment area (Au): The catchment area is equal to catchment size multiplied by the cell area.
Calculate Tan(h): Tan(h) is equal to ‘‘rise’’/’’run’’, i.e., DZ/L. h is the slope between the cell in question and the
lowest lying downslope neighbor cell.
Calculate specific catchment area (As): The As is equal to the Au divided by tan(h ).
Shift 0 with 0.000001: since tan(h ) is used as denominator in a division (and it may take the value zero), all zero
entrances in the raster is assigned the value 0.000001.
Calculate TWI: calculates ln(As/ tan(h i )), the topographic wetness index.
APPENDIX C
Implementation examples and detailed results
Fig. C1. Vegetation–topography relationships at two sites (both alkaline fens). Panel (A) illustrates one of the
strongest (r ¼0.68) correlations between the topographical wetness index (TWI) and the average preference for
moisture (EIVmoist). Site location: 55.978 N, 12.358 E. TWI is superimposed upon the elevation model for the area.
White means high TWI (22.5) and black low (2.3). Elevation ranges 9.2–21.4 meters above sea-level. Panel (B)
shows the same as (A), but for a strong elevation-DCA first axis correlation (r ¼ 0.94). Site location: 55.478 N, 9.798
E. Here, the white-black gradient refers to terrain elevation, 0.3–7.9 meters above sea-level.
Fig. C2. Coefficients of determination (R 2adj) for each vegetation measure from the regional-scale models
combining plot data (including habitat type) from all 901 study sites across Denmark. Abbreviations next to each
bar identify statistically significant (P , 0.05) explanatory variables. Hab, Habitat type; Ele, Elevation; Heat, Heat
Index; Slo, Slope; Rad, Potential Solar Radiation; TWI, Topographic Wetness Index; Wind, Wind Index. See Table
1 for abbreviations.
Fig. C3. Mean coefficients of determination (R 2adj) for each vegetation measure from the within-site (local)
models with elevation excluded as explanatory variable. See Table 1 for abbreviations.
Table C1. Average squared Spearman’s correlation coefficients (r 2) for topography and vegetation within each of
the 901 natural and semi-natural sites distributed throughout Denmark.
Table C2. Results from OLS multiple regression models linking the correlation strength of five selected correlation
pairs to 8 of the 18 site-level environmental variables for the 901 study sites located throughout Denmark.
Table C3. Results from OLS multiple regression models linking the correlation strength of five selected correlation
pairs to the remaining 10 of the 18 site-level environmental variables for the 901 study sites located throughout
Denmark.