You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263966116

Survival of Resident Neotropical Birds: Considerations for Sampling and Analysis


Based on 20 Years of Bird-Banding Efforts in Mexico

Article in The Auk · July 2012


DOI: 10.1525/auk.2012.11171

CITATIONS READS
42 393

7 authors, including:

Viviana Ruiz-Gutierrez Eduardo Santana


Cornell University Univ of Guadalajara & Univ WisconsinMadison
61 PUBLICATIONS 1,964 CITATIONS 138 PUBLICATIONS 965 CITATIONS

SEE PROFILE SEE PROFILE

Sarahy Contreras-Martínez Jorge E. Schondube


University of Guadalajara Universidad Nacional Autónoma de México
50 PUBLICATIONS 191 CITATIONS 114 PUBLICATIONS 2,636 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The idea that ecosystems have monetary value now has glibal support-and creates a route to protecting Earth´s endangered regions. View project

Vocal communication of the San Blas Jay (Cyanocorax sanblasianus) View project

All content following this page was uploaded by Sarahy Contreras-Martínez on 07 September 2016.

The user has requested enhancement of the downloaded file.


The Auk 129(3):500−509, 2012
 The American Ornithologists’ Union, 2012.
Printed in USA.

SURVIVAL OF RESIDENT NEOTROPICAL BIRDS: CONSIDERATIONS FOR


SAMPLING AND ANALYSIS BASED ON 20 YEARS OF BIRD-BANDING
EFFORTS IN MEXICO
VIVIANA RUIZ-GUTIÉRREZ,1,5,6 PAUL F. DOHERTY, JR.,1 EDUARDO SANTANA C.,2,3
SARAHY CONTRERAS MARTÍNEZ,2 JORGE SCHONDUBE,4 HERIBERTO VERDUGO MUNGUÍA,2
AND EDUARDO IÑIGO-ELIAS5
1
Department of Fish, Wildlife and Conservation Biology, Colorado State University, Fort Collins, Colorado, USA;
2
Instituto Manantlán de Ecología y Conservación de la Biodiversidad-DERN, Universidad de Guadalajara, Jalisco, Mexico;
3
Department of Forest and Wildlife Ecology, University of Wisconsin-Madison, Madison, Wisconsin, USA;
4
Centro de Investigaciones Ecosistemas, Universidad Nacional Autónoma de México, Morelia, Michoacán, Mexico; and
5
Cornell Laboratory of Ornithology, Cornell University, Ithaca, New York, USA

A.—Long-term bird-banding programs provide knowledge of the demographic rates of avian populations, but we currently
lack information on demographic rates of most bird populations in species-rich tropical ecosystems. Banding in tropical regions is typically
conducted with protracted or irregular sampling occasions that make the choice of the proper capture–mark–recapture (CMR) model
difficult. Here, we address common challenges related to collecting and analyzing data to estimate survival rates of resident Neotropical
birds using  years of banding efforts in Mexico as a case study. We applied Cormack-Jolly-Seber and Barker models to estimate apparent
survival and recapture probabilities of species with sufficient data for survival analyses. We were able to analyze  resident species of
 total species; apparent survival probabilities ranged from . to ., and recapture probabilities from . to .. For monitoring
programs with existing data collected at continuous, uneven, or irregular intervals, we recommend the application of the Barker model over
other models because it was more efficient in the use of available banding data and less often violated CMR assumptions. We recommend
that monitoring programs last > years and provide additional protocol suggestions for primary and secondary sampling occasions, as well
as the number of nets, potential net configurations, and the extent of the spatial scale. These baseline recommendations are likely to foster
an increase in our knowledge of avian survival rates in tropical ecosystems, which is imperative for managing tropical bird populations
under changing environmental conditions. Received  August , accepted  April .

Key words: Barker model, mark–recapture statistics, mist netting, population modeling, tropical birds.

Supervivencia de Aves Residentes Neotropicales: Consideraciones para el Muestreo y Análisis con base en un
Esfuerzo de  Años de Anillamiento de Aves en México

R.—Los programas de anillamiento de aves a largo plazo brindan conocimiento sobre las tasas demográficas de las poblaciones
de aves, pero todavía hace falta información sobre las tasas demográficas de la mayoría de poblaciones de aves en ecosistemas tropicales
áltamente diversos. El anillamiento en las regiones tropicales es típicamente conducido con muestreos prolongados o irregulares, que
hacen difícil la escogencia del modelo apropiado de captura-marcado-recaptura (CMR). En este trabajo abordamos las dificultades más
comunes relacionadas con la recolección y análisis de datos para estimar las tasas de supervivencia de aves neotropicales residentes, usando
 años de esfuerzos de anillamiento en México como un caso de estudio. Aplicamos los modelos de Cormack-Jolly-Seber y de Barker para
estimar la supervivencia aparente y las probabilidades de recaptura de especies con datos suficientes para los análisis de supervivencia.
Pudimos analizar seis especies residentes de un total de ; las probabilidades aparentes de supervivencia estuvieron entre . y ., y las
probabilidades de recaptura entre . y .. Para los programas de monitoreo con datos recolectados en intervalos continuos, desiguales
o irregulares, recomendamos la aplicación del modelo de Barker por encima de otros modelos, ya que éste fue más eficiente en el uso de los
datos de anillamiento disponibles, y viola menos frecuentemente los supuestos del CMR. Recomendamos que los programas de monitoreo
duren más de  años y presentamos sugerencias adicionales para el protocolo de ocasiones de anillamiento primario y secundario, así
como del número de redes, de la configuración potencial de las redes y de la extensión de la escala espacial. Estas recomendaciones base

6
E-mail: vruizg@colostate.edu

The Auk, Vol. , Number , pages −. ISSN -, electronic ISSN -.   by The American Ornithologists’ Union. All rights reserved. Please direct all
requests for permission to photocopy or reproduce article content through the University of California Press’s Rights and Permissions website, http://www.ucpressjournals.
com/reprintInfo.asp. DOI: ./auk..

— 500 —
JULY 2012 — PROTOCOL TO ESTIMATE SURVIVAL FOR RESIDENT TROPICAL BIRDS — 501

podrían promover el incremento de nuestro conocimiento de las tasas de supervivencia de las aves en ecosistemas tropicales, lo que es
imperativo para la administración de las poblaciones de aves tropicales bajo condiciones ambientales cambiantes.

E   management of avian populations these species are long lived (Robinson et al. ) and we need ad-
requires baseline information on the relationship between demo- ditional information about the relative importance of survival and
graphic rates and environmental factors. This knowledge is fun- fecundity across latitudes, life histories, and environments (Martin
damental to our understanding of how changing environmental , ; Ricklefs and Wikelski ) to target conservation ef-
conditions will influence population vital rates (e.g., survival and forts (Pfister , Clark and Martin ).
reproductive success) and is critical for predicting the viability and In the past two decades, advancements in capture–mark–
likelihood of persistence of threatened populations. Several long- recapture (CMR) sampling methods and statistical techniques
term banding programs have been established to monitor bird (Williams et al. , Sandercock ) have been used to esti-
populations in Europe (e.g., Baillie ) and North America (e.g., mate annual survival probabilities for ~% of resident Neotropi-
DeSante et al. ), generating information on demographic rates cal bird species (Table ), but these are restricted to a few sites and
of many resident and migratory bird populations. Despite increas- leave % of species unstudied. Survival information is lacking for
ing efforts, the geographic extent of most information generated by Neotropical bird populations because data do not exist for many
bird-banding monitoring programs remains limited to the Nearc- species and, when data exist, they are communicated through in-
tic and Palaearctic regions, which support only ~% of the world’s ternal project reports and not in peer-reviewed literature. Second,
avian biodiversity (Newton and Dale ). Roughly % of the the main objective of many banding projects is to gather data on
world’s bird species reside in tropical ecosystems whose forests (e.g., Neotropical migrants (DeSante et al. ), with less focus on res-
lowland, moist, dry, subtropical, and montane) contain almost half ident species. Third, data collection methods might present some
of the world’s threatened bird species (BirdLife International ). analysis and interpretation challenges. For example, current pro-
Thus, we are faced with the challenge of preserving tropical forest tocols for resident and migratory bird species suggest continuous,
birds without a clear understanding of the changes in demographic monthly sampling events (e.g., Ralph et al. , Latta et al. ).
rates that drive population declines of threatened species. More im- This may result in pooling of data across space and time, which
portantly, we lack information to develop and prioritize bird con- may violate the common assumptions in many CMR models of
servation objectives (e.g., minimize adult mortality and maximize demographic and geographic closure during sampling.
dispersal) and the means to determine whether management ac- Here, we address several of the challenges typically faced when
tions have succeeded in achieving the objectives themselves. collecting and analyzing data to estimate survival rates of resident
Most published research on bird population dynamics in tropi- Neotropical birds while adhering to CMR assumptions. We focus
cal ecosystems has been conducted in the Neotropics, where % of our analyses on the assumption of closure within a sampling pe-
the world’s bird species reside. These studies generally focus on es- riod and forgo discussing the details of “why” and “what” to moni-
timating survival rates, similar to monitoring programs established tor, given that these topics have been described at length elsewhere
in Europe and North America. Monitoring trends in survival is crit- (Yoccoz et al. ). Instead, we assume that “what to monitor” has
ical, especially for forest understory, resident Neotropical birds (i.e., been identified as survival and deal specifically with “how to moni-
understory forest species that do not seasonally migrate over long tor.” We use  years of banding efforts from the Sierra de Man-
distances, hereafter “resident Neotropical birds”), because some of antlán Bird Monitoring and Conservation Project, which is part of

TABLE 1. A compilation of banding studies that used a capture–mark–recapture framework to estimate survival probabilities of understory resident
bird species in tropical ecosystems in Latin America: publication, location and years of data collection, sampling intervals with number of days per
primary sampling event, months sampled, number of sites sampled, and number of species for which survival and recapture probabilities were re-
ported (NI = not included).

Sites Species
Reference Location Years Intervals Months (n) (n)

Karr et al. 1990 Panama a 1977–1986 3 days (primary)/month March, July 8 25


Brawn et al. 1995 Panama a 1977–1986 3 days (primary)/month March, July 8 25
Faaborg and Arendt 1995 Puerto Rico 1973–1990 3 days (primary)/year January or February 1 9
Johnston et al. 1997 Trinidad 1986–1995 NI June–August 3 17
Francis et al. 1999 Peru 1973–1982 3–4 days/year NI 6 14
Brawn et al. 1999 Panama a 1977–1997 3 days (primary)/month March, July 8–12 11
March–April,
Jullien and Clobert 2000 French Guiana 1991–1994 4 days (primary)/2 months September–October 7 17
Blake and Loiselle 2002 Costa Rica 1985–1994 2 days (primary)/5–6 weeks December–October 3 3
Parker et al. 2006 Ecuador 1998–2004 3 days (primary)/year December–February 2 5
Pearce-Higgins et al. 2007 Bolivia 1994–1999 2 weeks (primary) July–September 2 1
Blake and Loiselle 2008 Ecuador 2001–2006 1 day (primary) January, March 16 31
Ruiz-Gutierrez et al. 2008 Costa Rica 1993–2006 2–4 days (primary)/year January–March 8 1
Hau et al. 2010 Panama a 1979–2006 3 days (primary)/month March, July 8 NI
Present study Mexico 1993–2010 2–3 days (primary)/month January–December 1 9
a
The four published works for Panama are for the same study.
502 — RUIZ-GUTIÉRREZ ET AL.— AUK, VOL. 129

Mexico’s Long-term Ecological Research program, as a case study Loiselle ) or by defining relatively smaller intervals (e.g., days and
(Santana , Guallar et al. ). We use subsets of available months). But when different months are sampled inconsistently or
banding data to represent common variations in CMR analysis and sampling is not instantaneous (e.g., sampling occurs once per month,
evaluate the potential of the Barker model to estimate apparent an- as recommended by Ralph et al. []), correcting for potential bias
nual survival probabilities using data collected continuously or at can be difficult. The Barker () model can correct for potential
irregular intervals. Further, we use our results to provide sugges- bias under these sampling scenarios by allowing the consideration
tions for protocols that () incorporate sampling approaches found of a “primary sampling occasion” for which we can assume a closed
in the published literature, () meet CMR assumptions, () correct population, define intervals (e.g., yearly) between primary occasions,
for sampling bias, and () consider the ecology and life histories of but also include recaptures collected outside of this primary period
resident Neotropical birds. By providing a broader range of CMR as ancillary information. Although the Barker model was intended
models to analyze existing bird-banding data, as well as suggestions to mainly incorporate information from dead recoveries and resight-
concerning sampling protocols that might lead to more powerful ings outside of the primary sampling occasion, the model has been
data collection and analysis, we hope to improve the ability of inves- successfully applied to live-capture data by setting the dead-recov-
tigators to estimate survival of Neotropical and other birds. ery probability to zero (Collins and Doherty , LeDee et al. ).
For method , we generated capture histories from banding data col-
M ETHODS lected from  to , defining January–February as the primary
sampling occasion, and used recaptures from March–December as
Study site and sampling.—The study was conducted at the ,-ha ancillary information.
Las Joyas Research Station located within the Sierra de Manantlán For method , we generated capture histories from data col-
Biosphere Reserve and managed by the Manantlán Institute of Ecol- lected in January and February from  years of sampling (–
ogy and Conservation of Biodiversity. Our analyses were restricted ). After , sampling outside of the primary occasion
to the San Campus (SC) banding site, one of seven at the station, (January–February) was minimal or nonexistent, and there were
because SC has been actively monitored since  (Santana ). not enough recaptures outside of the primary occasion to yield
The habitat surrounding SC is early-successional montane pine robust estimates using the Barker model from –. We in-
(Pinus spp.) and oak (Quercus spp.) forest (Cuevas and Jardel ). clude this method because it probably represents our best survival
Sampling occurred in SC for  to  consecutive days. From  to estimates given the long period.
, sampling occurred once or twice a year; from  to , Statistical analysis for survival.—We fit a set of models that
sampling occurred nearly monthly; from  to , sampling varied in their assumptions of apparent survival (φ) and recapture
occurred from November through March. Sampling consisted of probabilities (p) in Program MARK (White and Burnham ).
constant-effort mist netting, with nets placed at the same location Apparent survival is the probability that an individual alive at time
in each site throughout the duration of the study. Birds were sam- t is also alive and has not emigrated from the study area at time
pled using  to  mist nets (-mm mesh) that were opened for  t + ; the product of true survival and fidelity (F), where F is defined
to  consecutive hours, starting at dawn. Nets were distributed over as the probability that an individual that was alive and present in
an area of approximately  to . ha. Nets were checked every – the study area in time t is also alive and has not emigrated from
min, and all individuals were identified to species and banded with the study area at time t + . We considered time-dependent (t) and
a uniquely numbered aluminum band. constant (.) models for both survival and recapture probabilities.
Methods for estimating survival.—To estimate survival, we We also used a time-since-marking (TSM) model to estimate ap-
used three methods (Table  and Table ). Methods  and  are parent survival (Pradel et al. ) to correct for the potential un-
Cormack-Jolly-Seber (CJS) models (Cormack , Jolly , Se- derestimation of these parameters as a result of transient and/
ber ), and method  is a model developed by Barker (). or young individuals (Johnston et al. ). The TSM model is a
Method  came from the realization that assuming both a closed version of the CJS model and allows the partitioning of apparent
population and constant site fidelity across all sampled months survival into survival during the first year after banding (φ) and
(e.g., January–December), spanning both breeding and non- survival from subsequent years (φ+), commonly referred to as sur-
breeding seasons, was not a likely biological scenario for resident vival of resident or adult birds. Time-since-marking models have
Neotropical birds, or for many bird species in general. To correctly been applied to estimate apparent survival and recapture prob-
estimate survival, the length of the sampling occasion should also abilities of tropical birds and are recommended for analyses of
only be a small fraction of the interval over which we want to make tropical bird data (Johnston et al. , Brawn et al. , Parker
inferences (Williams et al. ). To meet CMR assumptions, we
generated capture histories that collapsed data collected only in TABLE 2. The types of survival models, months, and years of data used in
January and February, from  to . Data were not collected methods 1–3 of the present study, and numbers of species (n) for which
in either of those months from  to , and so we had to dis- we were able to estimate survival and recapture probabilities.
card data from those years.
Method Survival model Months Years n
Method  examined the potential applicability of an alternate
approach to CJS developed by Barker (). In a CJS context, uneven 1 CJS a January–February 1994–1999 1
intervals and continuous sampling violate the closure assumption 2 Barker b January–February (primary) 1994–1999 4
and can bias estimates of apparent survival and recapture probabili- March–December (auxiliary) 1994–1999
ties. When sampling occurs “instantaneously” (e.g., – consecutive 3 CJS January–February 1994–2010 6
days) but at multiple times during a year, we can correct for potential a
CJS = Cormack-Jolly-Seber model (Cormack 1964, Jolly 1965, Seber 1965).
bias by defining each interval as a fraction of the year (e.g., Blake and b
Barker = the model developed by Barker (1997).
JULY 2012 — PROTOCOL TO ESTIMATE SURVIVAL FOR RESIDENT TROPICAL BIRDS — 503

TABLE 3. Notation and interpretation of the candidate model set tested using methods 1–3.

Model notation Interpretation

Methods 1 and 3 (Cormack-Jolly-Seber)


φ1(.)φ2+(.) p(.) Constant survival during the first year (φ1), survival during subsequent years (φ2+),
and recapture (p)
φ1(.)φ2+(t) p(t) Constant survival during the first year (φ1), time variation in recapture (p), and
time variation in survival for subsequent years (φ2+)
φ(.) p(.) Constant survival (φ) and recapture (p)
φ(t) p(.) Time variation in survival (φ) and constant recapture (p)
φ(t) p(t) Time variation in survival (φ) and recapture (p)
Method 2 (Barker)
All models have recovery rate (r) fixed to zero, with constant probabilities of be-
ing alive and having been recaptured (R), dying and having been recaptured alive
(R′), fidelity (F), and return probability of temporary emigrants (F′)
r = 0,S1(.)S2+(.)p(.)R(.)R’(.)F(.)F’(.) Constant survival after the first year (S1) and survival during subsequent years (S2+)
r = 0,S(.)p(.)R(.)R’(.)F(.)F’(.) Constant survival (S) and recapture probability (p)
r = 0,S(t)p(.)R(.)R’(.)F(.)F’(.) Time variation in survival (S) and constant recapture probability (p)
r = 0,S(t)p(t)R(.)R’(.)F(.)F’(.) Time variation in survival (S) and recapture probability (p)

et al. , Sandercock ). We had only enough data to model we used the quasi-likelihood modified Akaike’s information crite-
the first-year survival as constant (φ[.]), but considered both con- rion corrected for small sample sizes (QAICc) for model selection
stant (φ+[.]) and time dependence [φ+[t]) on survival and recap- and model averaging (Burnham and Anderson ). We present
ture probabilities of birds considered to be residents during the model-averaged estimates (Burnham and Anderson ).
sampling event, for methods  and  (Table ).
The Barker model (method ) in Program MARK has seven R ESULTS
estimable parameters: St = survival (probability that an individual
that was alive at time t is also alive at time t + ); pt = recapture The -year data set from the SC study site constitutes one of the
(probability that an individual that is alive in the study area at time largest known data sets in the Neotropics: , total captures,
t is recaptured); rt = recovery rate (probability that an individual of which , are recaptures of , individuals belonging to
dies between time t and t +  and was found dead and reported);  species. However, only  species had > total captures and
Rt = probability that an individual that survived from time t and > individuals, with most species averaging <. recaptures in-
t +  was recaptured and alive between t and t + ; Rt ’ = probability dividual– (Fig. ). Using the subset of data from  to  re-
that an individual that died between t and t +  was recaptured and sulted in model convergence on  of the  species. Of these, we
alive between t and t + ; Ft = fidelity probability (probability of re- had only enough marked individuals and recaptures for survival
turning to the study area given that the individual survives); and analyses for  migratory and  resident species using CMR meth-
Ft ’ = return of temporary emigrants (probability that an individual ods (e.g., SE range: –) for method . Of the six resident species,
that has emigrated temporarily from the study area and was not two belong to the family Trochilidae (Colibri thalassinus [COTH],
recaptured during time t is alive and returns to the study area at Hylocharis leucotis [HYLE]), two are in Turdidae (Myadestes
t + ). We set r = . for all models because the data did not contain occidentalis [MYOC] and Catharus occidentalis [CAOC]), one
band recoveries from dead birds (e.g., Collins and Doherty ). in Traupidae (Diglossa baritula [DIBA]), and one in Emberizidae
We cannot make any inferences on fidelity and other related pro- (Atlapetes pileatus [ATPI]). The models with the most support
cesses (F, F’, R, and R’) because all recaptures for the secondary oc- across species had recapture probabilities as constant, p(.), and ap-
casions were from the same area as the study site. Therefore, these parent survival as constant, φ(.) or S(.), or had a constant effect of
parameters were all set to be individually constant (.) in the candi- transient or juvenile individuals on apparent survival, φ(.)φ+(.) or
date model set (Table ). Because of our sampling scheme and the S(.)S+(.) (Table ).
inherent inability to estimate fidelity, we interpret what is typi- Using method , we had only enough data for DIBA to model
cally defined as true survival (S) in the Barker model as apparent apparent survival and recapture probabilities. Apparent survival
survival. Similarly, we interpret recapture probability as the prob- was φ+ = . ± . (SE), and recapture probability was p = .
ability that the individual (conditional on its being alive) remains ± . (Table ). Using the Barker model (method ), we were able
in the study area and is encountered. We modeled constancy and to generate estimates for four species: HYLE, COTH, DIBA, and
time dependence on survival and recapture probabilities, survival ATPI (φ+ = .–.; p = .–.; Table ). Using method ,
during the first year after banding (φ), and survival from subse- we obtained survival estimates for MYOC, CAOC, and the other
quent years (φ+) of the TSM model (Table ). species analyzed with methods  and : COTH, HYLE, DIBA, and
We constructed our models for each species independently. ATPI (φ+ = .–.; p = .–.; Table ).
We evaluated overdispersion by generating a variance inflation A subset of the data that included only captures from January
factor (ĉ) using the “median c-hat” estimation capabilities within and February better met the biological and closure model assump-
Program MARK for our most parameterized models. When ĉ > , tions (method ). However, this allowed us to obtain estimates for
504 — RUIZ-GUTIÉRREZ ET AL.— AUK, VOL. 129

from methods  and  (Table ). All estimates of apparent survival


and recapture probabilities using method  overlapped with es-
timates obtained through method , which represent  years of
monitoring efforts (Table ).

D ISCUSSION

Matching data collection with available analysis methods prior


to sampling is the most efficient approach to avoid potential pit-
falls when making statistical inferences in CMR studies (Lindberg
). In Latin America, published banding protocols for moni-
toring survival and other demographic rates for Neotropical res-
ident birds are often based on continuous sampling throughout
the year (Ralph et al. , Latta et al. ). Challenges common
to many banding programs, like logistical constraints and a lack
of consistency in funding across years, could also result in data
being collected over protracted or uneven sampling periods. The
result is that a large proportion of existing tropical bird-banding
data likely represent banding events spread over multiple months,
uneven spacing and sampling effort across years, or both. We fo-
cused on  years of banding effort in the Las Joyas Research Sta-
tion in the Sierra de Manantlán Biosphere Reserve, Mexico, as a
FIG. 1. Summary of banding data collected at the San Campus site, 1991–
case study to contrast different approaches for analyzing existing
2010. Histograms represent total captures, total number of individuals
bird-banding data to estimate survival probabilities of resident
captured, total recaptures, and the ratio of average number of recaptures
Neotropical birds.
per individual, estimated at the species level.
Our results suggest that the Barker model (method ) can
successfully be applied to such existing banding data for some
only  of  species (DIBA; Table ). The inclusion of captures species, while still adhering to the closure and interval-length
from March–December as supplementary information (method ) assumptions required by CMR analyses. The Barker model also
allowed us to estimate parameters for DIBA and three additional enables the combination of primary banding data, with auxiliary
species. The cost of adding the additional parameters of the Barker information from other sites that were not sampled as consis-
model was evident from estimates of recapture probabilities, tently. The proportion of species for which we can estimate sur-
which had very large standard errors compared with estimates vival using the Barker model remains minimal (%), even after

TABLE 4. Species code, scientific name, total number of individuals (n), method used (1–3), best-supported (top) models, model
averaged estimates of apparent survival of resident adult birds (S2+ or φ2+) and recapture probabilities (p), with their respective
unconditional standard error (SE) and lower (LCI) and upper (UCI) bounds of the confidence interval.

Code Species n Method Top models S2+ or φ2+ SE LCI UCI p SE LCI UCI

DIBA Diglossa baritula 291 1 φ(.) p(.) 0.49 0.23 0.14 0.85 0.40 0.23 0.09 0.81
φ1(.)φ2+(.)p(.)
103 2 S1(.)S2+(.)p(.) 0.77 0.13 0.43 0.94 0.44 0.19 0.15 0.78
S(.)p(.)
50 3 φ1(.)φ2+(.) p(.) 0.68 0.09 0.48 0.83 0.30 0.10 0.14 0.53
φ(.) p(.)
HYLE Hylocharis leucotis 74 2 S1(.)S2+(.)p(.) 0.52 0.22 0.16 0.86 0.24 0.21 0.03 0.75
193 3 φ1(.)φ2+(.) p(.) 0.52 0.12 0.29 0.74 0.36 0.16 0.13 0.68
MYOC Myadestes occidentalis 152 3 φ(.) p(.) 0.65 0.09 0.46 0.80 0.17 0.07 0.07 0.36
φ1(.)φ2+(.) p(.)
ATPI Atlapetes pileatus 58 2 S(.)p(.) 0.62 0.15 0.32 0.85 0.52 0.31 0.09 0.92
φ1(.)φ2+(.) p(.)
48 3 φ1(.)φ2+(.) p(.) 0.74 0.08 0.56 0.87 0.44 0.10 0.27 0.62
φ(.) p(.)
CAOC Catharus occidentalis 260 3 φ1(.)φ2+(.) p(.) 0.68 0.04 0.59 0.75 0.36 0.05 0.26 0.46
COTH Colibri thalassinus 71 2 S(.)p(.) 0.30 0.17 0.08 0.67 0.21 0.24 0.01 0.83
S(t) p(.)
121 3 φ(.) p(.) 0.43 0.17 0.16 0.75 0.11 0.13 0.01 0.61
φ1(.)φ2+(.) p(.)
JULY 2012 — PROTOCOL TO ESTIMATE SURVIVAL FOR RESIDENT TROPICAL BIRDS — 505

 years of continuous monitoring (– in our study). How- from methods developers, and user-friendly software (e.g., Pro-
ever, this rate was not unlike many published studies that gener- gram MARK) is available to conduct such discrete-time analyses.
ated data for < species, even after as much as  years of data Time span of monitoring.—The appropriate time frame over
collection (Table ). which to estimate survival is likely to differ between temperate
The Barker model remains preferable over truncating data to and tropical forest species, given that the latter are considered to
create a sampling interval closer to  year (e.g., method ), which be longer lived, on average (see review in Robinson et al. ).
greatly reduced the data and the number of species for which we We recommend against monitoring adult survival over short time
could estimate annual survival probabilities. Collapsing band- frames (e.g.,  month; Githiru and Lens ) in relation to the
ing data across several months to increase the number of species longevity of the species (> years), because true survival rates
for which we can estimate survival (e.g., Blake and Loiselle ) over short intervals would be ~. and would not be expected to
yields estimates that are difficult to interpret as annual survival differ from one interval to the next. Therefore, observed differ-
rates, even if effort is constant across years (e.g., McGregor et al. ences in apparent monthly survival estimates are likely a result
) because of the violation of the closure assumption within of movement patterns and do not reflect differences in true sur-
a sampling occasion. Small sample sizes also create the tempta- vival (e.g., Cruz-Angón et al. ). For long-term monitoring
tion to combine capture histories across species to provide a value programs, DeSante et al. () recommended using ≥ years of
of “average” survival for a group of interest (e.g., across guilds or data to model temporal and spatial variation in annual survival
habitat types). The biological justification for combining capture for bird species in North America. We suggest that a longer time
histories of diverse tropical bird species is tenuous because of the frame (more than – years) might be required to study popu-
implicit assumption that all individuals across species have the lation dynamics of most resident bird species in Neotropical sys-
same survival and recapture probabilities, including rates of fidel- tems, given the longevity and low recapture probabilities (p < .;
ity, and temporary and permanent immigration–emigration. The Sandercock and Beissinger ) that are likely. We also note that
resulting survival estimates are a weighted average, and the high- models that estimate population growth explicitly may be useful
est weight will be associated with species with the most individu- with such long-term (approximately – years) data sets (De-
als and recaptures. Sante et al. ).
Below, we build on the work of Ralph et al. () and Latta Frequency and timing of primary sampling events.—Ralph et
et al. () and present protocols that are specifically designed al. () suggest operating each site for  day every  days from
to model survival as a discrete process using a CMR framework June to August for temperate species and monitoring longer than
(Table ), explicitly linking data collection to analysis (Lindberg et  continuous months in the Neotropics. Latta et al. () pro-
al. ) using existing user-friendly software. Our suggested pro- posed similar recommendations specifically for monitoring de-
tocols also incorporate published methods used to estimate sur- mographic rates of tropical bird populations and suggested that
vival of Neotropical birds (Table ). Ralph et al. () developed sampling continuously throughout the entire year at regular in-
protocols to accomplish multiple objectives of monitoring of bird tervals (e.g., daily, weekly, monthly, quarterly) was more produc-
populations, but Latta et al. () focused specifically on mon- tive than sampling once a year. Sampling month to month, over
itoring Neotropical resident landbirds and wintering migrants. multiple years, is known to cause long-term net-avoidance behav-
The continuous sampling methods of Ralph et al. () and Latta ior in tropical lowland birds (J. D. Brawn pers. comm.) and prob-
et al. () have successfully yielded critical information on molt ably resulted in lower recapture probabilities for the entire length
patterns, food habits, and temporal changes in body condition of our study. To reduce this potential effect, we recommend moni-
(e.g., Santana , Schondube et al. , Guallar et al. ). Al- toring during discrete primary sampling events. The timing and
though analytical methods to model survival and recapture prob- number of the primary sampling events should be determined by
abilities as a continuous process exist, continuous sampling (e.g., the specific objectives of the monitoring project and would need to
monthly) is likely to result in net avoidance behaviors, leading to be optimized with simulations, but the beginning, middle, or end
low recapture rates (see below). Discrete-time survival models are of the peak of the breeding season for most species (e.g., March,
also most commonly used and have received the most attention May, and July in Central America) may be a logical starting point.
TABLE 5. Summary of suggestions for sampling Neotropical resident birds to estimate apparent an-
nual survival. Each column represents a sampling topic, sampling considerations under each topic,
and our suggestions for sampling to increase sample size and recapture probabilities in relation to
our study.

Sampling topic Considerations Suggestions

Time span Average survival ~ 10 years


Trends in survival ~ 10–30 years
Sampling approach Frequency and timing Robust design a
Primary and secondary occasions
Spatial extent Core netting area >25 ha
Fidelity and movements >1 km surrounding core
Spatial configuration Continuous sites Clusters of nets
Discrete sites Scaling of clusters by habitat types or sizes
One focal site, surrounding ancilliary data
a
Robust design stands for the model described by W. L. Kendall (1997).
506 — RUIZ-GUTIÉRREZ ET AL.— AUK, VOL. 129

Examples include Faaborg and Arendt (), whose sampling permanently emigrate or switch territories > year after initial
took place every January from  to the present; Brawn et al. capture (e.g., widowed individuals). Although differences in fi-
(), with sampling in March and July from  to the pres- delity between northern temperate and tropical bird species
ent; and Blake and Loiselle (), with sampling in January and have never been formally tested, the evidence suggests that the
March from  to the present (Table ). Reducing continuous common assumption that fidelity to a study site is ~ for tropi-
monitoring to one or a few primary sampling periods each year cal birds is also dependent on the spatial extent of sampling. Our
could reduce the cost and effort associated with current protocols application of the Barker model did not allow us to estimate fi-
and is likely to reduce the effects of net avoidance by individuals. delity and, hence, true survival because we did not have any re-
Secondary sampling occasions.—We recommend using a captures beyond our small study area that would have enabled us
robust-design (RD) sampling approach (Kendall et al. ), in to model permanent, large-scale movements. To estimate differ-
which sampling events are separated into primary and secondary ences in fidelity among tropical bird species and test for effects of
occasions. We have had success with primary sampling occasions permanent emigration on apparent survival rates, we recommend
that are made up of  consecutive days of mist netting or secondary increasing the spatial extent of the effective study area. One way
occasions (e.g., Fabborg and Arendt , Brawn et al. , Ruiz- in which this can be accomplished is by collecting ancillary move-
Gutiérrez et al. ). The main objective of the RD analysis ap- ment data, such as telemetry data (Powell et al. ) or resight-
proach is that it allows for the additional estimation of temporary ings of marked individuals (Cilimburg et al. ) at a reasonable
emigration parameters. The probability of emigrating temporar- distance outside the study area (> km).
ily is likely to be high for frugivorous birds (e.g., tanagers), insec- Spatial configuration of study sites.—Species-rich tropical
tivores (e.g., flycatchers), and nectarivores (e.g., hummingbirds), ecosystems generally contain low densities of individual terri-
as well as differing between sexes and between breeding and non- tories for most bird species (Karr et al. ). Although the area
breeding seasons. High or variable rates of temporary emigration used by individuals can expand and contract between breeding
will result in recapture probabilities that are biased low and possi- and non-breeding seasons, the number of individuals per species
bly underestimate survival rates (Sandercock and Beissinger , that were consistently recaptured in our relatively small netting
Parker et al. ). Although recapture rates will likely not be high sites was low. To increase sample sizes and recaptures, we recom-
enough to estimate temporary emigration parameters for most mend that several clusters or net lanes (“N” in Fig. ) be used to
species, a benefit of the RD sampling approach is that banding in- monitor continuous sites using the RD sampling approach. Moni-
formation can be collapsed across secondary sampling occasions toring can be consecutive, sampling one cluster per each second-
(e.g., to generate one value for the entire primary sampling event), ary sampling occasion (e.g., Jullien and Clobert ; Fig. A), or
without violating CMR closure assumptions (e.g., most work in a subset of available nets per cluster can be opened at each occa-
Table ). Although we used an RD sampling approach in our study, sion per site (e.g., Johnston et al. ; Fig. B). Another possibil-
we had consecutive primary sampling occasions (monthly) and ity is to have a core set of net clusters or lanes that are monitored
used very few mist nets (–) within a small area (< ha), which on a regular basis, and other clusters or lanes outside of the core
resulted in too few recaptures and marked individuals to apply RD
models. However, the RD sampling approach allowed us to col-
lapse our capture histories across secondary occasions and apply
CJS and Barker models to our data. The RD can also be extended
to use ancillary recaptures between primary sessions (see Lind-
berg ).
Spatial extent of sampling.—The small spatial extent of our
sampling area (< ha) also could have contributed to our low sam-
ple sizes and low recapture rates. We recommend that the spatial
extent of study areas be large enough to maximize the number of
potential territories per site (e.g., > ha). Otherwise, permanent
emigration rates can underestimate dispersal distances and esti-
mates of apparent survival for bird populations (Cilimburg et al.
, Zimmermann et al. , Cooper et al. ). Banded birds
are a combination of hatch-year birds, nonbreeding and nonterri-
torial floaters, delayed dispersers, and resident breeders. Through
the application of TSM models, we can account for lower survival
of hatch-year birds and for some of the nonbreeding floaters that
never returned to the study area after the initial capture. How- FIG. 2. Examples of potential spatial configurations for monitoring sur-
ever, Francis et al. () found that  of  species of tropical vival of resident Neotropical birds. For continuous sites, (A) net lanes
birds in Peru were recaptured > m away after the year of initial or clusters (“N”) are sampled consecutively (N1–N3) at each sampling
capture, with movement distances ranging from  m to > km. occasion or (B) a subset of each net cluster–lane (all N1) is sampled si-
This is likely the result of protracted breeding seasons (Schon- multaneously. For discrete sites or habitat types, (C) individuals can be
dube et al. ), delayed dispersal, and individuals that disperse, captured and marked in focal areas (“N”) and recaptures or resightings
float, and fail to breed (Tarwater and Brawn ). Time-since- (“R”) can take place outside the focal site. Sampling effort must also be
marking models do not account for recaptured individuals that scaled across (D) habitat types and sizes.
JULY 2012 — PROTOCOL TO ESTIMATE SURVIVAL FOR RESIDENT TROPICAL BIRDS — 507

(“R” in Fig. ) that are run less frequently but might generate valu- ACKNOWLEDGMENTS
able recaptures (e.g., Francis et al. ; Fig. C). To monitor mul-
tiple discrete sites (e.g., forest patches) or sites located in different Data analysis was funded by the Cornell Laboratory of Ornithol-
habitat types (e.g., shade vs. sun coffee), sampling designs should ogy (CLO); the MacArthur Foundation (E.I.E. and V.R.G.); the
take into consideration that, as with detection probabilities (Ruiz- Department of Fish, Wildlife and Conservation Biology (V.R.G.
Gutiérrez et al. ), recapture probabilities are likely to differ and P.F.D.); and the Colorado AGEP Postdoctoral Fellowship Pro-
between species in a group and could contrast sharply between gram (V.R.G.). The field work was funded by the Universidad de
habitat types (e.g., primary forest vs. pasture) and patch sizes. In Guadalajara, CONACYT, CONANP, Point Reyes Bird Observa-
this case, sampling effort (e.g., number of clusters) could be scaled tory, Rocky Mountain Bird Observatory, CLO, Catalán Institute
across habitat types and patch sizes (Fig. D), and the focus should of Ornithology, University of Wisconsin-Madison, University of
be on increasing recapture probabilities across all site categories. Arizona-Tucson, San Francisco State University, World Wildlife
Individuals could also be captured and marked in focal habitats Fund, National Fish and Wildlife Foundation, Audubon Society
or patches, and recaptures or resightings can take place outside of of Greater Denver, U.S. Fish and Wildlife Service, U.S.-A.I.D.,
the focal sites (e.g., surrounding matrix habitats; Fig. C) and po- U.S. Forest Service, National Institutes of Health MIRT program,
tentially incorporated in the analysis using Lindberg et al.’s () Grace J. Calder Trust, General Services Foundation, Colorado
and Barker’s models (Barker and White ). Wildlife Heritage Foundation, Vilas Scholarship, WWF-Interna-
Number of nets.—Our relatively small number of nets (– tional Prince Bernhard Fellowship, and the Paul and Bay Founda-
mist nets) could also have contributed to our low sample sizes and tions’ Biodiversity Leadership Award. Comments by J. M. Lepak,
recapture probabilities. The number of nets should be based on M. Lindberg, J. D. Brawn, and C. M. Francis on the manuscript
the number and configuration of net clusters or lanes, limited by were very much appreciated. The Sierra de Manantlán long-term
the total number of birds that can be handled and processed safely. banding project was possible thanks to > individuals who
Ralph et al. () recommend using  to  nets site–, but we ob- helped in multiple technical and logistical chores, too many to list;
served a range of – nets cluster– in the published literature we acknowledge and are grateful for their support.
(cited in Table ) and have also had success with clusters of ~
nets. The density of the nets will vary depending on the latitude, LITERATURE CITED
altitude, topography, and vegetation of the forest sites, but we rec-
ommend as even spacing as possible to reduce heterogeneity. B, S. R. . The contribution of ringing to the conservation
Overall, the estimation of annual survival through CMR and management of bird populations: A review. Ardea :–
sampling and analysis methods for most Neotropical bird species .
will require a substantial amount of long-term effort. We believe B, R. J. . Joint modeling of live-recapture, tag-resight, and
that our low sample sizes, even after  years of monitoring, likely tag-recovery data. Biometrics :–.
resulted from a combination of net avoidance behavior related to B, R. J.,  G. C. W. . Joint analysis of live and dead
our continuous sampling, the small number of nets we used, and encounters of marked animals. Pages – in Proceedings
the limited spatial extent of our sampling area. The sampling and of the Second International Wildlife Management Congress (R.
analysis protocols presented here are meant as suggestions for Field, R. J. Warren, H. Okarma, and P. R. Sievert, Eds.). Wildlife
long-term sampling and monitoring and are designed to increase Society, Bethesda, Maryland.
the potential number of species for which we can obtain survival BL I. . Threatened birds occur in all hab-
rates. To evaluate the performance of our suggestions, existing itats, but the majority are found in forest. [Online.] Available from
banding data should be frequently analyzed and, if needed, the www.birdlife.org/datazone/sowb/casestudy/.
protocols should be adjusted on the basis of preliminary findings B, J. G.,  B. A. L. . Manakins (Pipridae) in
or results from simulations. To this end, we need trained quan- second-growth and old-growth forests: Patterns of habitat use,
titative ecologists working for, and collaborating with, agencies movement, and survival. Auk :–.
and institutions in charge of monitoring resident Neotropical bird B, J. G.,  B. A. L. . Estimates of apparent sur-
populations. Much of our current knowledge on survival stems vival rates for forest birds in eastern Ecuador. Biotropica :–
from scientists who are trained in CMR analysis techniques, have .
access to such training, or collaborate with individuals who are B, J. D., J. R. K,  J. D. N. . Demography of
well versed in the subject (Table ). Our knowledge of survival birds in a Neotropical forest: Effects of allometry, taxonomy, and
rates of resident Neotropical birds will not likely increase un- ecology. Ecology :–.
less existing and future international bird-monitoring programs B, J. D., J. R. K, J. D. N,  W. D. R. .
shift from a blanket approach of teaching the mechanics of how Demography of forest birds in Panama: How do transients affect
to capture and band birds, toward need-based, local capacity- estimates of survival? Pages – in Proceedings nd Inter-
building efforts focused on teaching the data analysis and inter- national Ornithological Congress (N. J. Adams and R. H. Slotow,
pretation methodology that we have applied and discussed here. Eds.). Bird Life South Africa, Johannesburg.
All the above recommendations, along with other complementary B, K. P.,  D. R. A. . Model Selection
ones aimed at establishing long-term monitoring projects in Latin and Multimodel Inference: A Practical Information-Theoretic
America (Villaseñor and Santana ), will increase the amount Approach, nd ed. Springer-Verlag, New York.
of knowledge of survival rates, providing insight into the popula- C, A. B., M. S. L, J. J. T,  S. J.
tion ecology of resident Neotropical birds, as well as help guide H. . Effects of dispersal on survival probability of adult
future conservation and management actions. Yellow Warblers (Dendroica petechia). Auk :–.
508 — RUIZ-GUTIÉRREZ ET AL.— AUK, VOL. 129

C, M. E.,  T. E. M. . Modeling tradeoffs in avian K, J. R., J. D. N, M. K. K,  J. D. B.
life history traits and consequences for population growth. Eco- . Survival rates of birds of tropical and temperate forests:
logical Modelling :–. Will the dogma survive? American Naturalist :–.
C, C. T.,  P. F. D, J. . Survival estimates for K, W. L., J. D. N,  J. E. H. . Estimating
Royal Terns in southern California. Journal of Field Ornithology temporary emigration using capture–recapture data with Pol-
:–. lock’s robust design. Ecology :–.
C, C. B., S. J. D,  J. R. W. . Can we L, S. C., C. J. R,  G. R. G. . Strategies for
improve estimates of juvenile dispersal distance and survival? the conservation monitoring of resident landbirds and wintering
Ecology :–. Neotropical migrants in the Americas. Ornitologia Neotropical
C, R. M. . Estimates of survival from the sighting of :–.
marked animals. Biometrika :–. LD, O. E., T. W. A, E. A. R,  F. J. C.
C-A, A., T. S. S,  R. G. . An . Use of breeding and nonbreeding encounters to estimate
experimental study of habitat selection by birds in a coffee planta- survival and breeding-site fidelity of the Piping Plover at the Great
tion. Ecology :–. Lakes. Condor :–.
C, G. R.,  E. J. J. . Flora y Vegetación de la L, M. S. . Designs for capture–mark–recapture studies:
Estación Científica Las Joyas. Departamento de Ecología y Review, assessment, and emerging issues. Journal of Ornithology
Recursos Naturales-IMECBIO, Centro Universitario de la  (Supplement ):S–S.
Costa Sur, Universidad de Guadalajara, Jalisco, Mexico. L, M. S., W. L. K, J. E. H,  M. G. A-
DS, D. F., M. P. N,  D. R. O’G. . Identify- . . Combining band recovery data and Pollock’s robust
ing the proximate demographic cause(s) of population change by design to model temporary and permanent emigration. Biomet-
modeling spatial variation in productivity, survivorship, and pop- rics :–.
ulation trends. Ardea :–. M, T. E. . Life history evolution in tropical and south tem-
DS, D. F., T. S. S, R. B. S, J. F. S, C. A. perate birds: What do we really know? Journal of Avian Biology
R  V A, S. M, A. C, D. R. :–.
K, M. G,  B. M. . MoSI (Moni- M, T. E. . Avian life-history evolution has an eminent
toreo de Sobrevivencia Invernal): Assessing habitat-specific over- past: Does it have a bright future? Auk :–.
wintering survival of Neotropical migratory landbirds. Pages MG, R., M. J. W,  W. C. .
– in Bird Conservation Implementation and Integration Survival rates of tropical birds in Nigeria, West Africa. Ibis
in the Americas (C. J. Ralph and T. D. Rich, Eds.). U.S. Depart- :–.
ment of Agriculture, Forest Service General Technical Report N, I.,  L. D. . A comparative analysis of the avi-
PSW-GTR-. faunas of different zoogeographical regions. Journal of Zoology
F, J.,  W. J. A. . Survival rates of Puerto Rican (London) :–.
birds: Are islands really that different? Auk :–. P, T. H., C. D. B, B. K. S,  A. E.
F, C. M., J. S. T,  J. W. F. . A. . Apparent survival estimates for five species of
Survival rates of understory forest birds in Peru. Pages – tropical birds in an endangered forest habitat in western Ecuador.
in Proceedings nd International Ornithological Congress Biotropica :–.
(N. J. Adams and R. H. Slotow, Eds.). BirdLife South Africa, P-H, J. W., R. C. B,  J. H. .
Johannesburg. Survival of Band-tailed Manakins. Condor :–.
G, M.,  L. L. . Demography of an Afrotropical P, C. A. . Patterns of variance in stage-structured popula-
passerine in a highly fragmented landscape. Animal Conserva- tions: Evolutionary predictions and ecological implications. Pro-
tion :–. ceedings of the National Academy of Sciences USA :–.
G, S., E. S C., S. C, H. V,  P, L. A., M. J. C, J. E. H, J. D. N, 
A. G. . Paseriformes del occidente de México: Mor- D. G. K. . Simultaneous use of mark–recapture
fometría, datación y sexado. Monografía no. , Museu de Ciències and radio telemetry to estimate survival, movement, and capture
Naturals, Barcelona. rates. Journal of Wildlife Management :–.
H, M., R. E. R, M. W, K. A. L,  J. D. P, R., J. E. H, J.-D. L,  J. D. N. .
B. . Corticosterone, testosterone and life-history strat- Capture–recapture survival models taking account of transients.
egies of birds. Proceedings of the Royal Society of London, Series Biometrics :–.
B :–. R, C. J., G. R. G, P. P, T. E. M,  D. F.
J, J. P., W. J. P, R. D. G,  S. A. W. DS. . Handbook of field methods for monitoring land-
. Survival rates of tropical and temperate passerines: A Trini- birds. U.S. Department of Agriculture, Forest Service General
dadian perspective. American Naturalist :–. Technical Report PSW-GTR-.
J, G. M. . Explicit estimates from capture–recapture data R, R. E.,  M. W. . The physiology/life-
with both death and immigration-stochastic model. Biometrika history nexus. Trends in Ecology & Evolution :–.
:–. R, W. D., M. H, K. C. K, M. W, J. D. B,
J, M.,  J. C. . The survival value of flocking in S. H. A, C. E. T,  R. E. R. . Diversi-
Neotropical birds: Reality or fiction? Ecology :–. fication of life histories in New World birds. Auk :–.
JULY 2012 — PROTOCOL TO ESTIMATE SURVIVAL FOR RESIDENT TROPICAL BIRDS — 509

R-G, V., T. A. G,  A. A. D. . Habi- T, C. E.,  J. D. B. . Family living in a Neo-
tat fragmentation lowers survival of a tropical forest bird. Ecolog- tropical bird: Variation in timing of dispersal and higher survival
ical Applications :–. for delayed dispersers. Animal Behaviour :–.
R-G, V., E. F. Z,  A. A. D. . V, F.,  E. S C. . El monitoreo de poblacio-
Occupancy dynamics in a tropical bird community: Unexpect- nes: Herramienta necesaria para la conservación de aves en México.
edly high forest use by birds classified as non-forest species. Jour- Pages – in Conservación de Aves: Experiencias en México
nal of Applied Ecology :–. (H. Gómez de Silva and A. Oliveras de Ita, Eds.). CIPAMEX, Mexico.
S, B. K. . Estimation of demographic parameters W, G. C.,  K. P. B. . Program MARK: Sur-
from live-encounter data: A summary review. Journal of Wildlife vival estimation from populations of marked animals. Bird Study
Management :–.  (Supplement):S–S.
S, B. K.,  S. R. B. . Estimating rates W, B. K., J. D. N,  M. J. C. . Analy-
of population change for a Neotropical parrot with ratio, mark– sis and Management of Animal Populations. Academic Press, San
recapture and matrix methods. Journal of Applied Statistics Diego, California.
:–. Y, N. G., J. D. N,  T. B. . Monitor-
S [C.], E. . Dynamics of understory birds along a cloud ing of biological diversity in space and time. Trends in Ecology &
forest successional gradient. Ph.D. dissertation, University of Evolution :–.
Wisconsin, Madison. Z, G. S., R. J. G,  W. S. L. .
S, J., E. S C.,  I. R-T. . Bian- Finite study areas and vital rates: Sampling effects on estimates of
nual cycles of the Cinnamon-bellied Flowerpiercer. Biotropica Spotted Owl survival and population trends. Journal of Applied
:–. Ecology :–.
S, G. A. F. . A note on the multiple-recapture census.
Biometrika :–. Associate Editor: M. Lindberg

View publication stats

You might also like