Long-term demography of a reintroduced population of endangered falcons

Abstract

We examined long-term demography of an endangered subspecies, the Northern Aplomado Falcon (Falco femoralis septentrionalis), in South Texas, USA. The population has been managed and monitored since reintroductions began in 1993. Data spanning 1993–2018 enabled us to build an integrated population model (IPM) and a Cormack-Jolly-Seber model to estimate survival for three life stages (first-year, non-breeders, and breeders) and both sexes, abundance of males, fecundity, immigration, and emigration. Male falcons survived at lower rates than females during their first year; Hurricane Harvey caused a decline in survival rates of first-years and breeders; and fecundity increased after 2011 coinciding with changes in management focused on improving nest platforms and habitat quality. Both immigration of non-breeders and emigration were likely negligible for this population suggesting a potentially isolated population. The IPM likely overestimated immigration of breeders warranting further research. Population growth rates were greatest during years having more released captive-reared young and greater probabilities of breeder survival. Importantly, an apparent decrease in breeder survival of unknown cause occurred during 2006 to 2009 when breeder survival declined and remained low for several years. Our ability to identify the cause for reduced survival is now greatly hampered by the extended time that has passed, limiting the usefulness of our recent awareness of reduced survival for informing management and further highlighting the importance of real-time monitoring for proactive decision making processes. Our study greatly improves knowledge of demographics for a reintroduced, isolated, and intensively managed population of Aplomado Falcons. Applying this IPM to new data each year will enable adaptive management of the South Texas population by providing annual evaluations of vital rates along with revised assessments of monitoring and management.

Full text

Global Ecology and Conservation 38 (2022) e02226
Available online 19 July 2022
2351-9894/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Long-term demography of a reintroduced population of
endangered falcons
Brian W. Rolek
a
,
*
, Leah Dunn
a
, Benjamin Pauli
b
, Alberto Macias-Duarte
c
,
Brian Mutch
a
, Paul Juergens
a
, Tim Anderson
d
, Chris N. Parish
a
, Jeff A. Johnson
e
,
Brian Millsap
f
, Christopher J.W. McClure
a
a
The Peregrine Fund, Boise, ID, USA
b
Biology Department, Saint Mary’s University of Minnesota, Winona, MN, USA
c
Universidad Estatal de Sonora, Cuerpo Acad´
emico de Recursos Naturales, Hermosillo, Sonora, Mexico
d
U.S. Fish and Wildlife Service, Corpus Christi, TX, USA
e
Wolf Creek Operating Foundation, Wolf, WY, USA
f
U.S. Fish and Wildlife Service, National Raptor Program, Division of Migratory Bird Management, Albuquerque, NM, USA
ARTICLE INFO
Keywords:
Emigration
Hurricane
Immigration
Integrated population model
Population trend
Reintroduction
ABSTRACT
We examined long-term demography of an endangered subspecies, the Northern Aplomado Fal-
con (Falco femoralis septentrionalis), in South Texas, USA. The population has been managed and
monitored since reintroductions began in 1993. Data spanning 1993–2018 enabled us to build an
integrated population model (IPM) and a Cormack-Jolly-Seber model to estimate survival for
three life stages (rst-year, non-breeders, and breeders) and both sexes, abundance of males,
fecundity, immigration, and emigration. Male falcons survived at lower rates than females during
their rst year; Hurricane Harvey caused a decline in survival rates of rst-years and breeders;
and fecundity increased after 2011 coinciding with changes in management focused on improving
nest platforms and habitat quality. Both immigration of non-breeders and emigration were likely
negligible for this population suggesting a potentially isolated population. The IPM likely over-
estimated immigration of breeders warranting further research. Population growth rates were
greatest during years having more released captive-reared young and greater probabilities of
breeder survival. Importantly, an apparent decrease in breeder survival of unknown cause
occurred during 2006–2009 when breeder survival declined and remained low for several years.
Our ability to identify the cause for reduced survival is now greatly hampered by the extended
time that has passed, limiting the usefulness of this recent awareness for informing management
and further highlighting the importance of real-time monitoring for proactive decision making
processes. Our study greatly improves knowledge of demographics for a reintroduced, isolated,
and intensively managed population of Aplomado Falcons. Applying this IPM to new data each
year will enable adaptive management of the South Texas population by providing annual
evaluations of vital rates along with revised assessments of monitoring and management.
* Corresponding author.
E-mail address: rolek.brian@peregrinefund.org (B.W. Rolek).
Contents lists available at ScienceDirect
Global Ecology and Conservation
journal homepage: www.elsevier.com/locate/gecco
https://doi.org/10.1016/j.gecco.2022.e02226
Received 4 March 2022; Received in revised form 7 July 2022; Accepted 9 July 2022

Global Ecology and Conservation 38 (2022) e02226
2
1. Introduction
Long-term monitoring is important to reveal drivers of population processes, yet such programs are uncommon (Wiens, 1984;
Lindenmayer and Likens, 2010; Hughes et al., 2017). Monitoring during short time-scales could produce incomplete or misleading
results (Wiens, 1984) because researchers might observe fragments of a population cycle or fail to record catastrophes. Monitoring of
relevant demographic rates including survival, reproduction, emigration, and immigration is important (Katzner et al., 2007) for the
management of wildlife populations (Williams et al., 2002) especially for reintroduced or translocated populations (McClure et al.,
2017b). Demographic rates of post-translocated populations may provide insights into population viability at a site, and potential
threats (e.g. predation, management, and habitat; Bubac et al., 2019) that may prevent reestablishment.
Numerous processes can affect a population’s demographic rates, few of which are under human control. For example, catastrophes
such as hurricanes can substantially reduce wildlife populations because hurricane winds, rains, and storm surges can cause mortality
and other indirect effects (Wiley and Wunderle, 1993). Isolation of small populations further increases extinction risk (MacArthur and
Wilson, 1967). Conversely, immigration can rescue populations from extinction (Brown and Kodric-Brown, 1977) and might be
especially important for stability or growth of raptor populations (Brown and Collopy, 2013; Altwegg et al., 2014; Lieury et al., 2015;
McClure et al., 2021). Management actions can similarly increase population viability (McClure et al., 2017a; Crawford et al., 2018;
Saunders et al., 2018).
The Northern Aplomado Falcon (Falco femoralis septentrionalis; hereafter, Aplomado Falcon) is listed under the United States En-
dangered Species Act (United States Department of Interior, Fish and Wildlife Service 1986). Historically, this subspecies was a
year-round resident of the southwestern United States (Ligon, 1961; Hector, 1987) and northern Mexico south to Belize and
Guatemala, but was extirpated in the United States by the 1950s (Kiff and Peakall, 1980). Causes of population declines are uncertain
but are thought to have resulted from the conversion of open grasslands to agriculture and brushlands, leading to increased predator
populations and suppressed re regimes (Hunt et al., 2013). Releases of captive-reared individuals into coastal South Texas began in
1985 but failed to reestablish the population until additional releases occurred during 1993–2004. The population has appeared stable
Fig. 1. A map depicting three spatial aggregations of Northern Aplomado Falcon including Chihuahua in the northwest, Tropical Lowlands in the
south, and our focal population in South Texas, USA.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
3
at roughly 35 pairs since the bulk of releases ceased in 2004 (Jenny et al., 2004; Hunt et al., 2013; McClure et al., 2017a). The Peregrine
Fund has been intensively monitoring the population since 1993. Persistence of this population is now heavily reliant on enhanced
reproduction due to the provision of nesting platforms that reduce nestling predation prior to edging (McClure et al., 2017a).
Long-term demographic studies of reintroduced populations are rare. The reintroduced population of Aplomado Falcons presents
an opportunity to examine processes that affect small and reintroduced populations. These falcons are roughly 500 km from the nearest
known breeding population in the tropical lowlands of Mexico, and the demographic consequences of this geographic isolation have
not been quantied. After the breeding season of 2017, Hurricane Harvey made landfall in South Texas, providing the opportunity to
discern the effect of catastrophes on this small population. Importantly, the only study of survival in this population was a short-term
analysis examining data from 2002 to 2004 (Brown et al., 2006). This study has been inuential as the basis of population modeling
(McClure et al., 2017a) and management strategies, but whether this brief period is representative of survival across the span of the
entire program remains less clear. Additionally, the sex ratio of the population is female-biased for unknown reasons. Simulations
suggest that survival of juvenile males is less than that of juvenile females (McClure et al., 2017a), but this did not seem to be the case
during 2002–2004 (Brown et al., 2006).
Here, we use a class of hierarchical models called integrated population models (IPMs, Besbeas et al., 2002, Abadi et al., 2010a,
Schaub and Abadi, 2011) to examine 26 years of demographic data from the reintroduced population in South Texas. IPMs simul-
taneously model multiple data sets that can share information to improve demographic estimates (Schaub et al., 2007). Importantly,
IPMs can estimate latent (unobserved) rates that are challenging to measure in the eld, such as immigration, when explicit data are
unavailable (Besbeas et al., 2002; Schaub et al., 2007; Abadi et al., 2010b). We expected low rates of immigration due to geographic
isolation, lower survival in 2018 than other years due to Hurricane Harvey, and that survival of juvenile males is lower than that of
juvenile females.
Our objectives included (1) to create a modeling framework that will enable us to monitor vital rates each year; (2) to estimate vital
rates; (3) to evaluate the effects from previous management; (4) to identify effects from catastrophes; and (5) to identify drivers of
population growth rates.
2. Methods
2.1. Field sites
Our focal population included two spatial aggregations of the Aplomado Falcon in coastal South Texas centered near Laguna
Atascosa National Wildlife Refuge (26◦13′14′′ N, 97◦20′50′′ W) and Matagorda Island (28◦10′09 N, 96◦44′19 W) within Cameron and
Aransas counties, respectively (see Hunt et al., 2013 for details). These sites included both publicly owned (National Wildlife Refuge
and municipal) and privately-owned (ranches) lands. The South Texas population is one of three known spatial aggregations of
Aplomado Falcons including South Texas, Chihuahua, and Tropical Lowlands (Fig. 1).
Aplomado Falcon habitat included open prairie with widely scattered woody vegetation (Jenny et al., 2004). Habitat near Laguna
Atascosa National Wildlife Refuge included larger expanses of salt prairie that was seasonally inundated and vegetated by gulf
cordgrass (Spartina spartinae), sea oxeye daisy (Borrichia frutescens), saltat grass (Monanthochloe littoralis), and glasswort (Salicornia
sp.). Habitat at Matagorda Island included gulf cordgrass, marsh hay cordgrass (S. patens), gulf dune paspalum (Paspalum mono-
stachyum), and gulf bluestem (Schizachyrium maritimum). Aplomado Falcons reuse nests built by other large birds and establish nests in
wild vegetation including tree-yucca (principally Yucca treculeana), honey mesquite (Prosopis glandulosa), spiny hackberry (Celtis
ehrenbergiana), along with other species of vegetation (summarized by Hunt et al., 2013). The falcons often also nest in suitable
human-constructed articial nest structures (Hunt et al., 2013).
2.2. Management
The coastal South Texas population was reintroduced and augmented by releasing captive-reared birds during 1993–2004, and
again during 2012 and 2013. During releases, biologists housed Aplomado Falcons using human-constructed articial shelters, often
referred to as ‘hack boxes’. These hack boxes provided shelter and protection from predation for juvenile falcons while biologists
supplied resources necessary for survival, including food. Biologists gradually released juveniles into the wild over time by allowing
access outside of hack boxes, coinciding with their ability to edge, sustain ight, and obtain food on their own (Sherrod, 1982).
Hereafter, we refer to captive-reared and released individuals as ‘hacked’ and individuals reared by wild parents as ‘wild-reared’,
consistent with terminology used in other raptor studies.
Naturally-occurring nest substrates (e.g. Yucca treculeana) were scarce in South Texas during reintroductions, and nest predation
became increasingly widespread (Hunt et al., 2013); therefore, biologists installed human-constructed elevated articial nest struc-
tures to improve reproductive success of wild-breeding Aplomado Falcons. These articial nest structures were boxes or platforms
raised approximately 2–6 m in height except one box that was 21 m in height on a power pole. The articial nest structures had bars
spaced 7.62 cm along the sides with a solid top (Hunt et al., 2013). They were elevated to discourage ground predators (e.g. Raccoon,
Procyon lotor) and barred sides excluded large avian predators (e.g. Great-horned Owl, Bubo virginianus), while allowing adult Aplo-
mado Falcons to access nests.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
4
2.3. Data
We constructed an IPM using three primary data sources that included count data of individuals by life stage (c), mark-resight-dead
recovery data (Y), and productivity data (J). Count data included only males, mark-resight data included both males and females, and
productivity data included both male and female nestlings per pair. We chose to include counts of males as shared parameters within
the IPM rather than females because eld observations of unpaired females suggested that the fewer number of males could limit
populations, and previous analyses suggested lower survival of rst-year males (McClure et al., 2017a). We tested this assumption by
examining differences in survival of rst–year males and females (see Preliminary analyses).
We used ground-based survey methods (Hunt et al., 2013) to collect these data during 1993–2018. Adults and nestling Aplomado
Falcons were captured and outtted with colored-leg bands with unique color-combinations to allow individual identication. Each
year, we recorded the number of breeding adults (≥1 year old), non-breeding adults (≥1 year old), and nestlings including both
banded and unbanded birds. We identied banded individuals using spotting scopes, binoculars, or recaptures to record colored-leg
bands. Count surveys, banding, and resight efforts varied each year depending on resources, funding, and other logistical constraints.
We collected productivity data during May through July to count the number of wild-reared edglings (large young) per nesting
attempt by territorial pairs. We estimated productivity for all known territorial pairs during 1995–2006 and 2012–2018; however, we
monitored fewer pairs during 2007–2011, and we excluded productivity data during 2007 and 2009 because surveys were minimal.
We estimated survival, recruitment, and resight probabilities using the models described below (see Survival, recruitment, and
resight- Cormack-Jolly-Seber submodel), and we included explanatory variables for these parameters that included sex (male or fe-
male) and hacked status (wild or hacked) of individuals. Additionally, we included annual resight effort (low or high) as an
explanatory variable for resight probabilities. We assigned resight effort as “high” during years when we were able to monitor pro-
ductivity of all breeding pairs, and when we had resources to perform nearly-complete count surveys of Aplomado Falcon. Otherwise,
we assigned survey effort as “low”. We assigned sex to birds using morphometric measurements coupled with behavioral observations.
Hacked individuals were of captive origin and released into the wild (described in detail in “Management”).
Articial nest structures were updated in 2012 by changing the spacing of bars designed to exclude nest predators and by increasing
the number of articial nest structures (McClure et al., 2017a). Articial nest structures and improvements to these structures
increased daily nest survival rates (Brown and Collopy, 2008, 2012) and produced approximately three times the number of edglings
compared with nests on natural substrates (Hunt et al., 2013). Additionally, the United States Fish and Wildlife Service implemented
habitat management on approximately 3600 ha during 2012 to reduce shrub and woody encroachment that provided habitat for
predators near Aplomado Falcon breeding sites. This management included a combination of mechanical, chemical, and prescribed
re treatments to remove selected plants followed by prescribed burning to maintain falcon habitat (Verderber, 2015; Watson et al.,
2019). Combined, these management actions affect reproduction; therefore, we included a categorical covariate so fecundity and its
standard error could have different values before (<2012) and after (≥2012) management changes were implemented.
2.4. Life-cycle and IPM structure
We created a post-birth-pulse and stage-structured IPM for the Aplomado Falcon (Figs. 2 and 3). This IPM is ‘post-birth’ because we
consider surveys to have occurred after the hatching of young, and ‘stage structured’ because we grouped individuals by stages using
relevant life history traits.
Fig. 2. A graphical depiction of an integrated population model for Northern Aplomado Falcon. Demographic parameters are represented by blue
rectangles, observation parameters with purple circles, and data with orange diamonds. Dependences between nodes are depicted using arrows,
which are reduced in number for clarity. Submodels are depicted by large rectangles with dashed outlines. Data nodes included productivity (J),
counts (c), and mark-resight-dead recovery (Y). Demographic parameters included fecundity (F), number of individuals (N), immigration (
ω
),
emigration (δ), survival (ϕ), and recruitment (
ψ
). Parameters included observation error for count data (
σ
), resight probability (p), and dead re-
covery probability (r). Superscripts indicate life stages including rst year (aged zero years, 0), non-breeding adults (A), and breeder adults (B).
Figure adapted from Schaub and Abadi (2011).
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
5
We created a life-cycle model (Fig. 3) of the Aplomado Falcon to inform the IPM such that individual birds could be assigned to
three life stages: rst-years aged <1 year old (hereafter ‘rst-years’ and labeled in equations with superscript ‘0′because these birds are
zero years old), non-breeding adults (hereafter ‘non-breeders’ and labeled in equations with superscript ‘A′because these birds are
adult non-breeders), and breeding adults (hereafter ‘breeders’ and labeled in equations with superscript ‘B′). First-years were banded
approximately 17 days after hatching. Non-breeders were adults present during the breeding season that were not observed using
nests, maintaining nests, or provisioning young at a nest. Breeders were observed using or maintaining nests or provisioning young at a
nest.
Fig. 3. A diagram of the stage-structured life-cycle model used to develop an integrated population model for the Aplomado Falcon with population
counts during the post-breeding period. Circles depict abundance at each life stage and arrows depict potential recruitment between life stages.
Superscripts label parameters of rst-years aged zero to <1 year old (0), non-breeding adults ≥1 year old (A), and breeding adults ≥1 year old (B).
Potential recruitment is depicted as arrows and transitions are dependent on survival probability (ϕ), recruitment probability (
ψ
), and fecundity (F).
Fecundity is halved, because the life cycle graph is male-based and assumes an equal sex ratio at birth. Immigration and emigration are excluded
here for clarity. Bryce W. Robinson provided illustration of Aplomado Falcon.
Table 1
A glossary of symbols used in the integrated population model for Northern Aplomado Falcon.
Type Symbol Description
Labels 0 First-year (age zero)
A Adult non-breeder
B Adult breeder
H First-year hacked
0B, AB, BA, AA,
BB
Recruitment transitions are labeled as two life stages side-by-side having the rst label as the origin. For example, 0B indicates
rst-year to adult breeder, and BB indicates breeder remaining breeder.
Data c Count data of males
Y Mark-resight data of both males and females
J Annual productivity data
Indices i Marked individual
t Time step for each breeding season having 1993 as the rst year
Parameters N Abundance indices
ψ
Recruitment probability
ϕ Survival probability
p Resight probability
F Fecundity of males
ω
Immigration rate
δ Probability of emigration
r Probability of dead recovery
λ Population growth rate
σ
Standard error of normal distributions. Labeled using superscripts.
α
Regression intercepts and coefcients
ρ
Intercepts as probabilities
ε
Random effects
γ Productivity for each territory and year
μ
Average fecundity
η
,
υ
Negative binomial parameters for counts of breeders.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
6
2.5. Preliminary analyses
Preliminary analyses had two primary components: (1) an evaluation of the probabilities of survival, recruitment, and resight in
response to explanatory variables (i.e. hacked, sex, and effort) for each life stage; and (2) an evaluation of emigration and immigration
probabilities. These preliminary analyses are fully described in Supplementary materials (Appendix S1), and were used to retain
covariates and vital rates within the IPM when important to the South Texas population. Briey, we used mark-resight-dead recovery
data within a Cormack-Jolly-Seber (CJS) model to estimate survival, recruitment, resight probability, dead recovery probability, and
emigration. The explanatory variable ‘sex’ (either female or male) was included as a covariate of survival, recruitment, and resight
probabilities; ‘hacked status’ (wild-reared or hacked) was included as a covariate of survival, recruitment, and resight probabilities;
and ‘survey effort’ (low or high) was included as a covariate of resight probabilities in a given year as low or high effort. Next, we used
an IPM that included immigration as a latent vital rate. We excluded immigration in IPMs if they were negligible (i.e. median rate of ≤
1 individual per year) because inclusion of latent immigration rates in this circumstance can bias other demographic rates (Schaub and
Fletcher, 2015; Paquet et al., 2021). Further, we excluded these rates if they were poorly estimated having large 95 % highest density
intervals or median values appeared biologically implausible given eld observations.
We used preliminary tests to evaluate the goodness-of-t for Poisson, negative binomial, and zero-inated distributions for pro-
ductivity data and count data alone and statistics used for these tests are described in “Model implementation”. We chose the best
tting distribution for use within the IPM, and reassessed their t during the nal implementation.
2.6. Integrated population model (IPM)
2.6.1. Population submodel- integrating abundance index, survival, recruitment, and fecundity
We formulated a population submodel allowing integration of data sets that provided information about vital rates. This population
submodel species temporal dynamics for the population while considering survival, recruitment, fecundity, and the number of in-
dividuals in each life stage. We provide a table of symbols used hereafter for reference (Table 1).
The total number of breeder males during each year (NB
t) was estimated by summing the number of rst-year males (N0
t) that
survived (ϕ0
t) and recruited (
ψ
0B
t) to breeders; the number of non-breeder males (NA
t) that survived (ϕA
t) and recruited to breeders (
ψ
AB
t);
the number of breeder males (NB
t) that survived (ϕB
t) and remained breeders (i.e. did not recruit to non-breeders, 1 −
ψ
BA
t); the number
of immigrants to breeders N
ω
B
t; and then subtracting the number of emigrants (NδB
t) from breeders
NB
t=N0B
t+NAB
t+NBB
t+N
ω
B
t−NδB
t.
The probability of recruitment was denoted using two superscripts that indicated the initial state then the recruiting state of in-
dividuals. For example, the probability of recruiting from rst-years (0) to breeders (B) was denoted as
ψ
0B.
Next, we describe the specication of population dynamics of breeders in the IPM. We assigned binomial distributions to constrain
the number of recruited breeders from each life stage because the number recruited from each life stage must be less than or equal to
the number in each life stage during the previous year. The binomial distribution, often parameterized as y∼Binomial(p,n), is useful
here because it has two discrete parameters, where y is bound to integer values less than or equal to n, and p is the success probability
for each trial. For example, the number of rst-year males recruiting to breeders during the following year must be less than or equal to
the number of rst years (i.e. N0B
t+1 ≤N0
t).
N0B
t+1∼binomialϕ0
t
ψ
0B
t,N0
t
NAB
t+1∼binomialϕA
t
ψ
AB
t,NA
t
NBB
t+1∼binomialϕB
t1−
ψ
BA
t,NB
t
The number of immigrants during the following year was the product of the number of breeders during the previous year (NB
t) and
the breeder immigration rate (
ω
B
t) with a Poisson distribution
N
ω
B
t+1∼Poisson(
ω
B
tNB
t).
The number of emigrants the following year were the product of the number of breeders during the previous year and the
emigration rate (δt) with a Poisson distribution
NδB
t+1∼Poisson(δtNB
t)
This emigration rate was informed by the mark-resight-dead recovery data described below (See Survival, recruitment, and resight
section).
The number of rst-year males (N0) at the beginning of the following time steps (t+1) were estimated from the total number of
breeder males multiplied by fecundity halved, and by adding the number of captive-bred young males that were released. We esti-
mated the total number of breeder males by summing the following: the number of rst-year males that survived and recruited to
breeders (N0
tϕ0
t
ψ
0B
t) during the following year; the number of non-breeder males that survived and recruited to breeders (NA
tϕA
t
ψ
AB
t)
during the following year; and the number of breeder males (NB
t) that survived (ϕB
t) and remained breeders (i.e. did not recruit to
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
7
breeders, 1−
ψ
BA
t)during the following year. Next, we multiplied the number of breeding males by fecundity halved (Ft+1
2) because we
specied a male-based IPM and assumed an equal sex ratio of births (K´
ery and Schaub, 2012). We assigned a Poisson distribution to
constrain the number of wild-reared rst-years to discrete integers greater than or equal to zero. Abundance of wild-reared rst-years
was augmented with hacked (captive-reared) young during 1993–2004, 2012, and 2013. Abundance of hacked young (NH) was known
with certainty and these birds were deterministically added to the annual abundance of rst-years. Then we subtracted the number of
rst-year emigrants (NδO
t). Thus, the total number of rst-year males during the following year was estimated as
N0
t+1∼PoissonN0
tϕ0
t
ψ
0B
t+NA
tϕA
t
ψ
AB
t+NB
tϕB
t(1−
ψ
BA
t)Ft+1
2+NH
t+1−NδO
t
The number of rst-year emigrants (NδO) was estimated as NδO
t+1∼Poisson(δtNO
t).
We estimated the total number of non-breeder males (NA) during following years (t+1) by summing the number of rst-year males
(N0
t) that survived (ϕ0
t) and did not recruit to breeders (1 −
ψ
0B
t); the number of non-breeder males (NA
t) that survived (ϕA
t) and did not
recruit to breeders (1 −
ψ
AB
t); the number of breeder males NB
t that survived (ϕB
t) and recruited to non-breeders (
ψ
BA
t); the number of
immigrants to non-breeders (N
ω
A
t); then subtracting the number of emigrants to non-breeders (NδA
t).
NA
t=N0A
t+NAA
t+NBA
t+N
ω
A
t−NδA
t
We constrained the number recruited and remaining non-breeders using binomial distributions because the number recruiting from
each life stage during the following time step cannot exceed the number available. Thus, the number of non-breeder males during
following years was estimated as
N0A
t+1∼Binomial ϕ0
t1−
ψ
0B
t,N0
t
NAA
t+1∼BinomialϕA
t1−
ψ
AB
t,NA
t
NBA
t+1∼BinomialϕB
t
ψ
BA
t,NB
t
And the numbers of immigrants and emigrants to non-breeders were estimated as
N
ω
A
t+1∼Poisson(
ω
A
tNA
t)
NδA
t+1∼Poisson(δtNA
t)
Abundances of each life stage can be further subdivided as needed to include population segments that respond differently to
important covariates, and we included additional subdivisions in our implementation for Northern Aplomado Falcon when these
population segments had diverging vital rates (Appendix S1).
2.6.2. Count state-space submodel
We estimated an index of abundance within the IPM using count data (c) and a state-space formulation for each life stage with a
Poisson distribution for rst-years and non-breeders, while we assigned breeders a negative binomial distribution because goodness-of-
ts tests suggested this as an appropriate distribution.
cA
t∼Poisson(NA
t)
c0
t∼Poisson(N0
t)
cB
t∼negative binomial(
η
t,
υ
)
η
t=
υ
NB
t
υ
We used a negative binomial distribution for counts of breeders because goodness-of t tests suggested a lack of t using a Poisson
distribution, and a negative binomial distribution provided adequate t.
2.6.3. Survival, recruitment, and resight- multi-state Cormack-Jolly-Seber submodel
We created a stage-structured multi-state mark-recapture submodel (e.g. McClure et al., 2017b) using a Cormack-Jolly-Seber
(hereafter CJS) state-space formulation (Gimenez et al., 2007, Royle 2008). Observation data were a matrix of mark-resight obser-
vations (Yi,t) with four observed states designated with Yi,t=1 as rst-year, Yi,t=2 as non-breeder, Yi,t=3 as breeder, Yi,t =4 as
recovered dead, and Yi,t=5 as ‘not seen’. We related observed states to true states of individuals using the observation matrix:
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
8
.
This observation matrix was specied so that breeders and non-breeders that were ‘not seen’ could have been undetected. First-
years have a perfect (1.0) probability of resight because they must be captured and banded to enter the study, and cannot be
resighted as rst-years thereafter. We included dead recovery probability in the state-transition matrix to allow implementation in
JAGS. We included sex, hacked status, and resight effort as xed effects of resight probability when these covariates were determined
to be important in preliminary analyses (Appendix S1). We allowed resight probabilities and recruitment to vary by year by including
random factors (
ε
t). Resight probabilities for non-breeders to varied with effort and hacked status as logit(pA
i,t) =
α
pA
effort,hacked +
ε
pA
hacked,t
where
ε
pA
hacked,t∼normal(0,
σ
pA
hacked)and breeders as logit(pB
i,t) =
α
pB
effort +
ε
pB
t where
ε
pB
t∼normal(0,
σ
pB).
Next, we specied a state-transition matrix for the CJS submodel that governed state dynamics to estimate the probabilities of
apparent survival (ϕ) and life stage recruitment (
ψ
), dead recovery (r), and emigration (δ) from time step (t, rows) to the following time
step (t+1, columns) as.
.
First-years could not be resighted as the same state during the following year because they must recruit to breeders, non-breeders,
or die. We included sex and hacked status as xed effects of survival and recruitment when these covariates were determined to be
important (Appendix S1). We allowed survival and recruitment to vary by year by including random factors (
ε
t). We specied survival
of rst-years to vary with sex, hacked status, and year as logit(ϕ0
i,t) =
α
ϕ0
sex,hacked +
ε
ϕ0
sex,hacked,t where
ε
ϕ0
sex,hacked,t∼normal(0,
σ
ϕ0
sex,hacked). We
specied recruitment from rst-year to breeder to vary with sex and year as logit(
ψ
0B
i,t) =
αψ
0B
sex +
εψ
0B
sex,t where
εψ
0B
sex,t∼normal(0,
σψ
0B
sex )and
recruitment from non-breeder to breeder varied with sex and hacked status as logit(
ψ
AB
i,t) =
αψ
AB
sex,hacked +
εψ
AB
sex,hacked,t where
εψ
AB
sex,hacked,t∼
normal(0,
σψ
AB
sex,hacked). All other survival and recruitment parameters did not include effects of sex or hacked status but did include the
random factor of year. For example, survival of non-breeder males was specied as logit(ϕA
i,t) =
α
ϕA+
ε
ϕA
t and
ε
ϕA
t∼normal(0,
σ
ϕA).
This CJS submodel estimated true survival rather than apparent survival, which can be confounded with emigration (K´
ery and
Schaub, 2012). We included data from both males and females in the CJS submodel to obtain survival estimates for both sexes using a
single analysis. However, we integrated parameters from only males into the IPM when we observed potentially important differences
between sexes in survival, recruitment, or resight (see Preliminary analyses) consistent with a male-based IPM.
2.6.4. Fecundity- Poisson regression submodel
We used annual productivity data (J) of the total number of edglings from all monitored territories during each year (t) with a
Poisson distribution having a mean that was the product of fecundity (F) and the number of breeders, Jt∼Poisson(FtNB
t). This equation
links annual fecundity to the population submodel. We included before (<2012) and after (≥2012) management (McClure et al.,
2017b) as a covariate which included management of both habitat and nest boxes, as a xed effect of the intercept as log(Ft)
=log(
μ
F
management[t]) +
ε
F
t where
ε
F
t∼normal(0,
σ
F)was a random factor of time.
2.6.5. IPM implementation
We implemented models using Bayesian methods in Just Another Gibbs Sampler v4.3.0 (JAGS, Plummer, 2003) with R statistical
software v3.6.1 (R Core Team, 2015) and the package jagsUI v1.5.1 (Kellner, 2016) as interfaces. We implemented each model using
three chains with each having ≥1000 adaptation, 150,000 burn-in, and ≥50,000 posterior iterations. We thinned each chain by
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
9
retaining one of 50 iterations totaling 3000 iterations for each posterior distribution to avoid autocorrelation between posterior draws
and increase effective sample sizes from the posterior. We assessed convergence of posterior chains using traceplots, density plots, and
Gelman-Rubin diagnostic (
R, Gelman and Rubin, 1992), and we assigned adequate convergence when traceplots of parameters did not
visually appear to drift and 
R≥1.1. Scripts for implementation are archived online (Appendix S1). We considered explanatory var-
iables potentially important when 85 % HDIs of differences between groups did not intersect zero. We used vague priors for all pa-
rameters constrained to exclude biologically implausible values (Appendix S2).
We used posterior predictive checks to assess goodness-of-t of the count state-space submodels and this test can be considered as
an omnibus test for the entire IPM (Schaub and Kery, 2021). This procedure is fully described elsewhere (Gelman and Hill, 2007;
Schaub and Kery, 2021). We used mean absolute percentage error as a discrepancy measure (Besbeas and Morgan, 2014) to assesses
goodness-of-t. For the fecundity Poisson regression submodel, we evaluated goodness-of-t by examining the deviance and var-
iance/mean ratio, and we visually examined plots of discrepancies of observed data compared to the discrepancies of replicate data
simulated from the submodel using Bayesian P values (Schaub and Kery, 2021).
2.7. Effects of Hurricane Harvey on survival
We evaluated the effect from Hurricane Harvey on survival. Hurricane Harvey occurred during the fall of 2017 after post-breeding
surveys occurred; therefore, the observed effects from this hurricane would be within survival estimates for 2018. We compared the
average annual rates of survival among all population segments and overall averages among all population segments (e.g. averaged
over rst-years, non-breeders, and breeders) during three periods: 1993–2017, 2017, and 2018. For rst-year falcons, we only
examined estimates from wild-reared young because there were no hacked young within the population at the time. Estimates from
1993 to 2017 provided a comparison of long-term averages of annual survival to the hurricane year (2018), while 2017 was expected
to be more autocorrelated with the hurricane year because they are nearer in time and could provide a more direct “before and after”
comparison. We used derived parameters to calculate average rates of survival for population segments. We compared rates of survival
between population segments using probability of direction (Makowski et al., 2019).
2.8. Population growth rates
We evaluated inter-annual correlations between model estimates of population growth rates and vital rates because insights into
drivers of population growth rates can be attained by examining correlations among vital rates over time (Robinson et al., 2004;
Schaub and Abadi, 2011). We used Pearson’s correlation coefcients to estimate correlations between population growth rates and
some vital rates (i.e. survival and fecundity) from the IPM (K´
ery and Schaub, 2012). We assigned importance to correlation coefcients
when 90 % HDIs did not intersect zero, and the probability of the correlation coefcient being greater than zero was ≥0.90. This is
consistent with a one-tailed test because we expected survival and fecundity to be positively correlated with population growth rates.
3. Results
3.1. Data summaries
During 1993–2018, we banded a total of 1334 individuals. Observed states over time included 1318 rst-years, 253 non-breeders,
279 breeders, and 38 dead recoveries. Of these 1334 individuals, 441 were wild-reared while 893 were hacked, and 631 were female
while 703 were male. Productivity averaged 25.9 (SD =21.4, range =0–59) young per year and 1.5 (SD =1.25, range =0–4) young
per nest attempt.
3.2. IPM goodness-of-t
Goodness-of-t tests for state-space and productivity submodels suggested adequate model t to the data (Appendix S1
Figs. S9–S11).
3.3. Emigration and immigration
Preliminary analyses estimated emigration rates near zero (median =0.0009, 95 % HDIs =0.0000–0.0027) using mark-resight-
dead recovery data with a CJS model. Using an IPM, immigration rates of non-breeders were also near zero (median =0.012, 95
% HDIs =0.0000–0.0589). The IPM estimated breeders having greater immigration rates than non-breeders and had greater uncer-
tainty (median =0.11, 95 % HDIs =0.02–0.19, Appendix S1 Figs. S2–S7). We believed these immigration rates were biologically
implausible because (1) estimates were greater than expected given the geographic distance to other known breeding populations; (2)
we observed low numbers of unmarked birds each year excluding some years when several nestlings remained unbanded due to
logistical constraints; and (3) we observed almost no movement of banded birds between two subsites (Matagorda Island and Laguna
Atacosa NWR). Previous studies have documented the tendency of IPMs to overestimate immigration (Paquet et al., 2021). Hereafter,
we considered emigration and immigration to be negligible, and we present results from an IPM that set immigration rates to zero
because including immigration could potentially bias estimates of other vital rates (Paquet et al., 2021). Emigration estimates were low
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
10
enough that we did not expect this rate to bias estimates of vital rates, and was therefore retained in the model.
3.4. Index of abundance
Index of total abundance of male Aplomado Falcon peaked around 2001, declined until around 2011, increased again until 2017
when Hurricane Harvey impacted the study area and decreased in 2018, the last year of the study (Fig. 4).
3.5. Fecundity
Fecundity increased after 2011 (Fig. 5), as documented in a previous study (McClure et al., 2017a). The 1994–2011 median
fecundity rate was 0.58 (95 % HDIs =0.47–0.72), and the 2011–2018 rate was 1.74 (95 % HDIs =0.82–2.99). The 1994–2011 rate
excludes 1993 when the population consisted only of hacked rst-years and therefore lacked reproduction.
3.6. Survival, recruitment, and resight
Breeders had the greatest overall survival rate followed by non-breeders, rst-year wild-reared females, rst-year hacked females,
rst-year wild-reared males, and rst-year hacked males (Fig. 6). We observed important differences in survival between rst-year
males and females, and rst-year wild-reared and hacked (Appendix S1 Fig. S1). However, we did not observe differences in sur-
vival between sexes during other life stages, nor were there differences in survival between hacked and wild-reared falcons of non-
breeders and breeders (Fig. 7). Importantly, survival rates of breeders and rst-years were low during 2006 to roughly 2009 and
rebounded briey until Hurricane Harvey in 2017 (Fig. 8). Non-breeder males that were wild-reared had the greatest probability of
recruiting to breeders, followed by hacked non-breeder males, wild-reared non-breeder females, and hacked non-breeder females
(Figs. 6 and 7). Survey effort affected resight rates of both breeders and non-breeders (Fig. 6). Both breeders and non-breeders had the
greatest probability of being resighted during years with greater survey effort (Table 2, Fig. 7), and survey effort had more pronounced
effects on resight rates of breeders, consistent with our study design targeting breeding birds.
3.7. Effects of Hurricane Harvey on survival
Average survival of breeders and rst-year males was less in 2018 compared to the previous year, and compared to the average
from 1993 to 2017. Average survival across all population segments had a 0.96 probability of being less during 2018 compared to
2017, and survival had a 0.98 probability of being lower during 2018 compared to the period 1993–2017 (Table 3; Fig. 9).
3.8. Population growth rates
Inter-annual population growth rates increased with breeder survival and number of hacked individuals (Fig. 10). Breeder survival
had the greatest magnitude of correlation, while the number of hacked individuals had the greatest probability of being greater than
zero.
Fig. 4. Inter-annual index of abundance of male Aplomado Falcons in South Texas, USA estimated from an integrated population model. Black lines
depict model-estimated medians, and gray shading depicts 95 % highest density intervals. Dashed lines depict count data.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
11
Fig. 5. Inter-annual estimates of fecundity from 1993 to 2018 of Aplomado Falcons in South Texas from an integrated population model. The black
line depicts the median, and gray shading depicts 95 % highest density intervals. The vertical dashed line indicates the year (2012) when biologists
improved nest boxes and increased the number of articial nest structures. Fecundity is dened as edglings per occupied territory and excludes
captive-reared (i.e. hacked) edglings.
Fig. 6. Average probabilities of survival, recruitment, and resight for population segments of Aplomado Falcons in South Texas, USA during
1993–2018. Estimates from an integrated population model are depicted as medians (circles) ±85 % and 95 % highest density intervals (thick and
thin lines).
Fig. 7. Group differences of survival, recruitment, and recapture estimates of Northern Aplomado Falcon in South Texas, USA during 1993–2018.
Estimates are from an integrated population model that included data from population counts, mark-resight, and productivity. Differences are
presented for each life stage (i.e., rst-year, non-breeder, breeder) and differences were estimated between groups (i.e., hacked, sex, effort). Medians
(points), 85 % highest density intervals (thick vertical lines), and 95 % highest density intervals (thin vertical lines) are depicted. The horizontal line
at zero references no difference. Positive valued differences indicate greater values of wild hatched, female, and greater effort, while negative values
indicate greater values of hacked, males, and lower effort. The reduced survival model presented here includes covariates that did not intersect zero
from the global CJS model using 95 % highest density intervals.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
12
4. Discussion
Our results highlight the importance ongoing management of a population of Aplomado Falcons in South Texas. Management had a
large effect on the population growth rate, and the number of captive bred males released per year was correlated with increases in
population growth rates. Further, the rate of fecundity increased after 2011 as predicted, when nest boxes were updated and more nest
structures were installed (McClure et al., 2017a). The effects that we observed from intensive management on this reintroduced
population are consistent with past studies. For example, Perkins et al. (2008) showed that reintroductions could potentially increase
population viability of Florida Grasshopper Sparrows (Ammodramus savannarum oridanus), and the reintroduced Puerto Rican Parrot
(Amazona vittata) is reliant on articial nest sites for breeding (White et al., 2006), similar to the falcons in our study population.
Fig. 8. Inter-annual estimates of survival from an integrated population model during 1993–2018 of Aplomado Falcons in South Texas, USA. Black
lines depict medians, and gray shading depicts 95 % highest density intervals.
Table 2
Medians and 95 % highest density intervals (95 % LHDI and 95 % UHDI) of estimates from the Cormack-Jolly-Seber (CJS) submodel of an integrated
population model including estimates of survival, recruitment, and resight rates for population segments of Northern Aplomado Falcons in South
Texas. The CJS submodel included data from both sexes (males and females), and only males were linked to the integrated population model when
signicant differences were observed between sexes. ‘Hacked’ indicates birds that were bred in captivity and released via hacking, and ‘wild’ indicates
birds that reared by wild adults.
Parameter Life stage Sex, hacked, effort Median 95 % LHDI 95 % UHDI
Survival First-year female, wild 0.44 0.32 0.57
male, wild 0.12 0.02 0.25
female, hacked 0.22 0.14 0.32
male, hacked 0.11 0.06 0.18
Nonbreeder 0.78 0.71 0.86
Breeder 0.86 0.78 0.94
Recruitment First-year to breeder female 0.05 0.00 0.17
male 0.36 0.11 0.68
Nonbreeder to breeder female, wild 0.28 0.17 0.39
male, wild 0.85 0.71 0.96
female, hacked 0.16 0.02 0.33
male, hacked 0.59 0.30 0.88
Breeder to nonbreeder 0.03 0.00 0.07
Resight Nonbreeder female, wild 0.30 0.17 0.49
male, wild 0.42 0.29 0.55
female, hacked 0.18 0.07 0.33
male, hacked 0.45 0.22 0.72
Breeder female 0.04 0.01 0.08
male 0.88 0.71 0.99
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
13
Table 3
Probabilities of annual survival of population segments being different during certain time pe-
riods. Hurricane Harvey impacted this population in the fall of 2017; therefore, the survival rate
during 2018 represents the probability of an individual surviving between breeding seasons of
2017 and 2018, the time period impacted by the hurricane.
Life stage and sex 2018 vs 2017 2018 vs
1993–2017
First-year male 0.95 0.98
First-year female 0.74 0.81
Non-breeders 0.42 0.38
Breeders 0.90 0.85
Fig. 9. Half-eye plots representing posterior distributions of average annual survival rates across all population segments of Aplomado Falcons in
South Texas, USA. Medians (points), 85 % (thick vertical lines) and 95 % highest density intervals (thin vertical lines) are depicted. Hurricane
Harvey impacted this population in the fall of 2017. Therefore, the survival rate during 2018 represents the probability of an individual surviving
between breeding seasons of 2017 and 2018, the time period impacted by the hurricane.
Fig. 10. Inter-annual correlations between population growth rates and demographic rates for Aplomado Falcon in South Texas, USA during
1993–2018. Correlations were estimated using Pearson’s correlation coefcients, where r is the mode and 90 % highest density intervals are in
parentheses. The probability that correlation coefcients (r) were greater than zero are indicated by P(r >0). A stable population growth rate is
equal to one and is depicted by horizontal dashed lines. Error bars (solid lines) depict 95 % HDIs of point estimates.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
14
Breeder survival in our study population was weakly correlated with population growth rates. Adults often have greater survival
when animals have slow life histories (Sæther and Bakke, 2000; Clark and Martin, 2007), especially raptors (Newton, 1979). Similarly,
a recent study of American Kestrels across four disparate study sites used IPMs to demonstrate that adult survival and immigration
were most correlated with population growth (McClure et al., 2021).
We expected our study population to be fairly isolated, and a lack of immigration would explain the need for reintroduction efforts
to reestablish the population in South Texas. Estimates provided by the IPM appear to overestimate immigration by breeders having a
median of four (95 % HDIs=0–6.2) males immigrating each year into a population of breeders that ranged between 0 and 44 males. We
believe the IPM provided biologically implausible immigration estimates given the geographic distances between known breeding
populations and small number of unmarked birds observed each year, and IPMs have a tendency to overestimate immigration using
similar methods (Paquet et al., 2021). Despite the large geographic distance between known breeding populations, several Aplomado
Falcon have been documented in closer proximity to south Texas within Tamaulipas, Mexico (Rodríguez-Ruíz et al., 2012) suggesting a
few Aplomado Falcons could occasionally immigrate. The IPM without immigration t the data slightly better than the model
including immigration, and we consider the South Texas population of Aplomado Falcons to be likely isolated from other populations
but further exploration of this topic is warranted. Isolation could be concerning given that immigration can be an integral part of
maintaining stability or growth of raptor populations (Brown and Collopy, 2013; Altwegg et al., 2014; Lieury et al., 2015; McClure
et al., 2021). Isolation followed by a declining population size can lead to inbreeding depression and other potential tness related
problems caused by genetic factors (e.g. Bortoluzzi et al., 2020), although Johnson et al. (2021) found no evidence of inbreeding in this
population using samples collected between 2004 and 2016, or within a few generations following the release of captive reared in-
dividuals. Inbreeding might be a problem in the future and thus continued genetic monitoring is warranted (see also Johnson et al.,
2021).
Isolation can also hinder recovery from catastrophes such as hurricanes, when potential recruits are available only from repro-
duction but not from immigration. Hurricane Harvey in 2017 substantially reduced survival of breeders and rst-year males, thereby
lowering the number of breeding pairs in the population. With climate change predicted to increase the damage cause by hurricanes in
the United States (Dinan, 2017), resilience to hurricanes is a concern for this population. Also of concern, this population appears to
have also suffered decreased breeder survival during 2006–2009, unbeknownst to managers (Fig. 8). Although we observed population
declines during this period, we assumed that this decline was caused by the cessation of captive releases, not due to a decline in adult
breeder survival. This event seems to have similar effects on the population to those observed from Hurricane Harvey (Fig. 8) and has
implications for the frequency and severity of catastrophes that this population will likely experience.
Population modeling should be an integral part of any reintroduction program (Grifth et al., 1989; Wolf et al., 1998; Seddon et al.,
2007; Armstrong and Seddon, 2008; Armstrong and Reynolds, 2012). Given the importance of adult survival to population dynamics,
estimates should be representative of survival rates and their temporal variation when evaluating their inuence on populations
(Morris and Doak, 2002; McGowan et al., 2011). Prior to this study, the only estimates of survival available were those of Brown et al.
(2006) using data collected over a much shorter time period (2001–2004) than used in this study (1993–2018). Brown et al.’s (2006)
survival estimates were uncharacteristically high with low variance compared to our estimates because this period of study coincided
with some of the highest survival estimates observed (Fig. 8), and did not span temporal periods with any apparent stochastic events
that could negatively impact the population such as Hurricane Harvey in 2017 or the decrease in breeder survival during 2006–2009.
Therefore, population modeling conducted by McClure et al. (2017a) using previous survival estimates might have been overly
optimistic about the fate of this population in light of the lower survival estimates as obtained from this study. Our results highlight the
pitfalls of relying on short-term studies for obtaining accurate inference into population processes. Future work should incorporate
these new survival estimates and the potential effects from catastrophes into population assessments for this subspecies.
Long-term datasets can also be more useful for explaining sex-biased differences that result from more nuanced processes than
those that may occur over the short term. The South Texas population had more lone territorial females observed than males and more
pairs containing sub-adult males than sub-adult females suggesting a female-biased population (McClure et al., 2017a). However,
survival of males and females did not differ during 2002–2004 using intersecting data (Brown et al., 2006). The prediction of McClure
et al. (2017a) appears to have been correct that rst-year survival was lower for males than females over the longer duration of our
study. The resulting sex bias of this population likely has implications for population viability, which should be the subject of
simulation modeling.
Our results demonstrate the importance of long-term monitoring for inference into population processes. Our focal population has
been the subject of several short-term studies and substantial monitoring (Hunt et al., 2013). Our analyses using 26 years of data
provided greater inference about population dynamics than obtained previously, yet additional years of study could provide additional
insights. Adaptive monitoring (Lindenmayer and Likens, 2009) provides a tool to update knowledge regularly and can inform adaptive
management (Holling, 1986) so that managers can respond in real-time to improve conservation outcomes. For example, had we been
conducting adaptive monitoring during the decrease in breeder survival of 2006–2009, we could have recognized the downturn in
adult survival and could have increased effort to collect additional data and identify potential causes. Our current ability to discern and
compensate for causes of this decrease in breeder survival is severely hindered because we identied it more than a decade after it
occurred. Long-term and real-time adaptive monitoring is therefore important, especially for isolated and reintroduced small pop-
ulations such as the Aplomado Falcon in South Texas. By implementing analyses annually, we plan to create a framework for adaptive
management and monitoring of the South Texas population of Aplomado Falcon and our study provides a template for other con-
servation programs focused on single-species management.
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
15
Declaration of Competing Interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to
inuence the work reported in this paper.
Data availability
Links to code and data are included within the manuscript and are openly accessible.
Acknowledgements
Data and code are available online: https://doi.org/10.5281/zenodo.6656702 and an R Markdown workow is provided in the le
“docs/index.html”. This project was funded by a grant from the U.S. Fish and Wildlife Service Science Support Grant Number
F17AC01040. Bryce W. Robinson provided illustrations of Aplomado Falcon and additional artwork can be found at www.ornithologi.
com. BWR performed analyses and wrote and edited the manuscript. CJWM performed analyses, wrote the manuscript, and
conceptualized the study. BP, AM, TA, CP, and JJ edited the manuscript and conceptualized the study. BMu, and PJ edited the
manuscript, conceptualized the study, collected data, and managed eld sites. BMi provided guidance on analyses and edited the
manuscript. LD edited the manuscript and provided a gure. We are grateful for the thoughtful and thorough comments from two
anonymous reviewers that greatly improved this manuscript.
Data and code are available at the link below, which we state in the body of the manuscript and the acknowledgements.
“Scripts for implementation are archived online at https://doi.org/10.5281/zenodo.6656702 and an R Markdown workow is
provided in the le “docs/index.html”.”
Appendix A. Supporting information
Supplementary data associated with this article can be found in the online version at doi:10.1016/j.gecco.2022.e02226.
References
Abadi, F., Gimenez, O., Arlettaz, R., Schaub, M., 2010a. An assessment of integrated population models: bias, accuracy, and violation of the assumption of
independence. Ecology 91, 7–14.
Abadi, F., Gimenez, O., Ullrich, B., Arlettaz, R., Schaub, M., 2010b. Estimation of immigration rate using integrated population models. J. Appl. Ecol. 47, 393–400.
Altwegg, R., Jenkins, A., Abadi, F., 2014. Nestboxes and immigration drive the growth of an urban Peregrine Falcon Falco peregrinus population. Ibis 156, 107–115.
Armstrong, D.P., Reynolds, M.H., 2012. Modelling reintroduced populations: The state of the art and future directions. In: Ewen, J.G., Armstrong, D.P., Parker, K.A.,
Seddon, P.J. (Eds.), Reintroduction Biology: Integrating Science and Management. Blackwell Publishing Ltd, Oxford, UK, pp. 165–222.
Armstrong, D.P., Seddon, P.J., 2008. Directions in reintroduction biology. Trends Ecol. Evol. 23, 20–25.
Besbeas, P., Morgan, B.J.T., 2014. Goodness-of-t of integrated population models using calibrated simulation. Methods Ecol. Evol. 5, 1373–1382.
Besbeas, P., Freeman, S.N., Morgan, B.J.T., Catchpole, E.A., 2002. Integrating mark–recapture–recovery and census data to estimate animal abundance and
demographic parameters. Biometrics 58, 540–547.
Bortoluzzi, C., Bosse, M., Derks, M.F.L., Crooijmans, R.P.M.A., Groenen, M.A.M., Megens, H.J., 2020. The type of bottleneck matters: Insights into the deleterious
variation landscape of small managed populations. Evolut. Appl. 13, 330–341.
Brown, J.H., Kodric-Brown, A., 1977. Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58, 445–449.
Brown, J.L., Collopy, M.W., 2008. Nest-site characteristics affect daily nest-survival rates of Northern Aplomado Falcons (Falco femoralis septentrionalis). Auk 125,
105–112.
Brown, J.L., Collopy, M.W., 2012. Bayesian hierarchical model assessment of nest site and landscape effects on nest survival of Aplomado Falcons. J. Wildl. Manag.
76, 800–812.
Brown, J.L., Collopy, M.W., 2013. Immigration stabilizes a population of threatened cavity-nesting raptors despite possibility of nest box imprinting. J. Avian Biol. 44,
141–148.
Brown, J.L., Collopy, M.W., Gott, E.J., Juergens, P.W., Montoya, A.B., Hunt, W.G., 2006. Wild-reared Aplomado Falcons survive and recruit at higher rates than
hacked falcons in a common environment. Biol. Conserv. 131, 453–458.
Bubac, C.M., Johnson, A.C., Fox, J.A., Cullingham, C.I., 2019. Conservation translocations and post-release monitoring: identifying trends in failures, biases, and
challenges from around the world. Biol. Conserv. 238, 108239.
Clark, M.E., Martin, T.E., 2007. Modeling tradeoffs in avian life history traits and consequences for population growth. Ecol. Model. 209, 110–120.
Crawford, B.A., Moore, C.T., Norton, T.M., Maerz, J.C., 2018. Integrated analysis for population estimation, management impact evaluation, and decision-making for
a declining species. Biol. Conserv. 222, 33–43.
Dinan, T., 2017. Projected increases in hurricane damage in the United States: the role of climate change and coastal development. Ecol. Econ. 138, 186–198.
Gelman, A., Hill, J., 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, New York, NY, USA.
Gelman, A., Rubin, D.B., 1992. Inference from iterative simulation using multiple sequences. Stat. Sci. 7, 457–472.
Gimenez, O., Rossi, V., Choquet, R., Dehais, C., Doris, B., Varella, H., Vila, J., Pradel, R., 2007. State-space modelling of data on marked individuals. Ecol. Model. 206
(3–4), 431–438. https://doi.org/10.1016/J.ECOLMODEL.2007.03.040. In this issue.
Grifth, B., Scott, J.M., Carpenter, J.W., Reed, C., 1989. Translocation as a species conservation tool: Status and strategy. Science 245, 477–480.
Hector, D.P., 1987. The decline of the Aplomado Falcon in the United States. Am. Birds 41, 381–389.
Holling, C.S., 1986. Adaptive environmental management. Environ.: Sci. Policy Sustain. Dev. 28, 39.
Hughes, B.B., Beas-Luna, R., Barner, A.K., Brewitt, K., Brumbaugh, D.R., Cerny-Chipman, E.B., Close, S.L., Coblentz, K.E., De Nesnera, K.L., Drobnitch, S.T., Figurski, J.
D., Focht, B., Friedman, M., Freiwald, J., Heady, K.K., Heady, W.N., Hettinger, A., Johnson, A., Karr, K.A., Mahoney, B., Moritsch, M.M., Osterback, A.M.K.,
B.W. Rolek et al.

Global Ecology and Conservation 38 (2022) e02226
16
Reimer, J., Robinson, J., Rohrer, T., Rose, J.M., Sabal, M., Segui, L.M., Shen, C., Sullivan, J., Zuercher, R., Raimondi, P.T., Menge, B.A., Grorud-Colvert, K.,
Novak, M., Carr, M.H., 2017. Long-Term studies contribute disproportionately to ecology and policy. BioScience 67, 271–278.
Hunt, W.G., Brown, J.L., Cade, T.J., Coffman, J., Curti, M., Gott, E., Heinrich, W., Jenny, J.P., Juergens, P., Macías-Duarte, A., Montoya, A.B., Mutch, B., Sandfort, C.,
2013. Restoring aplomado falcons to the United States. J. Raptor Res. 47, 335–351.
Jenny, J.P., Heinrich, W., Montoya, A.B., Mutch, B., Sandfort, C., Hunt, W.G., 2004. Progress in restoring the aplomado falcon to Southern Texas. Wildl. Soc. Bull. 32,
276–285.
Johnson, J.A., Stock, A., Juergens, P., Mutch, B., McClure, C.J.W., 2021. Temporal genetic diversity and effective population size of the reintroduced Aplomado
Falcon (Falco femoralis) population in coastal south Texas. J. Raptor Res. 55.
Katzner, T., Milner-Gulland, E.J., Bragin, E., 2007. Using modeling to improve monitoring of structured populations: are we collecting the right data? Conserv. Biol.
21, 241–252.
Kellner, K. 2016. jagsUI: A wrapper around “rjags” to streamline “JAGS” analyses. R package version 1.4.2.
K´
ery, M., Schaub, M., 2012. Bayesian Population Analysis Using WinBUGS: A hierarchical perspective. Elselvier Inc., Oxford, UK.
Kiff, L., Peakall, D.B., 1980. Eggshell thinning and organochlorine residues in the bat and aplomado falcons in Mexico. Congr. Int. Ornithol. 12, 949–952.
Lieury, N., Gallardo, M., Ponchon, C., Besnard, A., Millon, A., 2015. Relative contribution of local demography and immigration in the recovery of a geographically-
isolated population of the endangered Egyptian vulture. Biol. Conserv. 191, 349–356.
Ligon, J.S., 1961. New Mexico Birds and Where to Find Them. University of New Mexico Press, Albuquerque, New Mexico, USA.
Lindenmayer, D., Likens, G., 2010. The science and application of ecological monitoring. Biol. Conserv. 143, 1317–1328.
Lindenmayer, D.B., Likens, G.E., 2009. Adaptive monitoring: a new paradigm for long-term research and monitoring. Trends Ecol. Evol. 24, 482–486.
MacArthur, R.H., Wilson, E.O., 1967. The Theory of Island Biogeography. Princeton University Press, Princeton, New Jersey.
Makowski, D., Ben-Shachar, M.S., Lüdecke, D., 2019. bayestestR: describing effects and their uncertainty, existence and signicance within the Bayesian framework.
J. Open Source Softw. 4, 1541.
McClure, C.J.W., Pauli, B.P., Mutch, B., Juergens, P., 2017a. Assessing the importance of articial nest-sites in the population dynamics of endangered Northern
Aplomado Falcons Falco femoralis septentrionalis in South Texas using stochastic simulation models. Ibis 159, 14–25.
McClure, C.J.W., Rolek, B.W., Hayes, T.I., Hayes, C.D., Thorstrom, R., Curti, M., Anderson, D.L., 2017b. Successful enhancement of Ridgway’s Hawk populations
through recruitment of translocated birds. Condor 119, 855–864.
McClure, C.J.W., Brown, J.L., Schulwitz, S.E., Smallwood, J., Farley, K.E., Therrien, J.F., Miller, K.E., Steenhof, K., Heath, J.A., 2021. Demography of a widespread
raptor across disparate regions. Ibis 163, 658–670.
McGowan, C.P., Runge, M.C., Larson, M.A., 2011. Incorporating parametric uncertainty into population viability analysis models. Biol. Conserv. 144, 1400–1408.
Morris, W.F., Doak, D.F., 2002. Quantitative Conservation Biology. Sinauer Associates, Sunderland, Massachusetts, USA.
Newton, I., 1979. Population Ecology of Raptors. Buteo Books. Vermillion, South Dakota, USA.
Paquet, M., Knape, J., Arlt, D., Forslund, P., P¨
art, T., Flagstad, Ø., Jones, C.G., Nicoll, M.A.C., Norris, K., Pemberton, J.M., Sand, H., Svensson, L., Tatayah, V.,
Wabakken, P., Wikenros, C., Åkesson, M., Low, M., 2021. Integrated population models poorly estimate the demographic contribution of immigration. Methods
Ecol. Evol. 12, 1899–1910.
Perkins, D.W., Vickery, P.D., Shriver, W.G., 2008. Population viability analysis of the Florida Grasshopper Sparrow (Ammodramus savannarum oridanus): Testing
recovery goals and management options. Auk 125, 167–177.
Plummer, M., 2003. JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling.
R Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
Robinson, R.A., Green, R.E., Baillie, S.R., Peach, W.J., Thomson, D.L., 2004. Demographic mechanisms of the population decline of the song thrush Turdus philomelos
in Britain. J. Anim. Ecol. 73, 670–682.
Rodríguez-Ruíz, E.R., G´
omez-Rodríguez, J.F., Trevi˜
no Carre´
on, J., Ter´
an-Trevi˜
no, S., Enríquez Medina, F.A., S´
anchez Gonz´
alez, A., Garza-Torres, H.A., P´
aramo
Mendoza, C., G´
omez Ramírez, H., 2012. Nuevos avistamientos de aves para Tamaulipas, M˜
A. Huitzil 13, 162–168.
Royle, Andrew J., 2008. Modeling individual effects in the Cormack–Jolly–Seber model: A state–space formulation. Biometrics 64 (2), 364–370.
Sæther, B.-E., Bakke, Ø., 2000. Avian life history variation and contribution of demographic traits to the population growth rate. Ecology 81, 642–653.
Saunders, S.P., Cuthbert, F.J., Zipkin, E.F., 2018. Evaluating population viability and efcacy of conservation management using integrated population models.
J. Appl. Ecol. 55, 1380–1392.
Schaub, M., Abadi, F., 2011. Integrated population models: a novel analysis framework for deeper insights into population dynamics. J. Ornithol. 152, 227–237.
Schaub, M., Fletcher, D., 2015. Estimating immigration using a Bayesian integrated population model: choice of parametrization and priors. Environ. Ecol. Stat. 22,
535–549.
Schaub, M., Kery, M., 2021. Integrated population models: theory and ecological applications with R and JAGS. Academic Press.
Schaub, M., Gimenez, O., Sierro, A., Arlettaz, R., 2007. Use of integrated modeling to enhance estimates of population dynamics obtained from limited data. Conserv.
Biol. 21, 945–955.
Seddon, P.J., Armstrong, D.P., Maloney, R.F., 2007. Developing the science of reintroduction biology. Conserv. Biol. 21, 303–312.
Sherrod, S.K. 1982. Hacking: a method for releasing peregrine falcons and other birds of prey. The Peregrine Fund, Boise, ID, USA.
United States Department of Interior Fish and Wildlife Service, 1986. Endangered and Threatened Wildlife and Plants; Determination of the northern Aplomado
Falcon to be an Endangered Species, 51. Federal Register, pp. 6686–6690.
Verderber, E.M. 2015. Restoration strategies to mitigate shrub encroachment into coastal prairies. The University of Texas at Rio Grande Valley.
Watson, P.A., Alexander, H.D., Moczygemba, J.D., 2019. Coastal prairie recovery in response to shrub removal method and degree of shrub encroachment. Rangel.
Ecol. Manag. 72, 275–282.
White Jr., T.H., Brown, G.G., Collazo, J.A., 2006. Articial cavities and nest site selection by Puerto Rican Parrots: a multiscale assessment. Avian Conserv. Ecol. 1, 5.
Wiens, J.A., 1984. The place of long-term studies in ornithology. Auk 101, 202–203.
Wiley, J.W., Wunderle, J.M., 1993. The effects of hurricanes on birds, with special reference to Caribbean islands. Bird. Conserv. Int. 3, 319–349.
Williams, B.K., Nichols, J.D., Conroy, M.J., 2002. Analysis and Management of Animal Populations. Academic Press, New York, NY, USA.
Wolf, C.M., Garland, T., Grifth, B., 1998. Predictors of avian and mammalian translocation success: reanalysis with phylogenetically independent contrasts. Biol.
Conserv. 86, 243–255.
B.W. Rolek et al.