Assessing extinction risk, conservation reliance, and down-listing potential for two endangered Hawaiian waterbirds
 
Évaluation du risque d’extinction, fiabilité de la conservation et rétrogradation potentielle pour deux oiseaux aquatiques hawaïens menacés

Charles B. van Rees ¹ , Chris S. Elphick ² and J. Michael Reed ³ 

• Abstract
• Introduction
• Methods
  • Study system
  • Goal 1: Estimating equilibrium population density using Bayesian time-series models
  • Goal 2: Baseline population viability analyses
  • Goal 3: Conservation reliance using PVA Sensitivity analysis
• Results
  • Goal 1: Estimating equilibrium population density using Bayesian time-series models
  • Goal 2: Baseline population viability analyses
  • Goal 3: Conservation reliance using- PVA Sensitivity analysis
• Discussion
• Responses to this article
• Author contributions
• Acknowledgments
• Literature cited

INTRODUCTION

An idealized view of endangered species management is that vulnerable species are investigated, their threats identified, and management is enacted to eliminate or mitigate the threats such that the species is no longer endangered. This scenario is central to the application of many endangered species laws, such as the U.S. Endangered Species Act (Goble 2009, Neel et al. 2012), but it presumes that the threats to a species can be eliminated. For some species, however, threat elimination is not possible and continued management is required to prevent species decline toward extinction (Scott et al. 2005, Scott et al. 2010; Goble et al. 2012). Such conservation reliance (also known as conservation dependence) is increasingly widespread (Averill-Murray et al. 2012, Wiens and Gardali 2013, Wilsey et al. 2014, Brown et al. 2017, Connor et al. 2019, Scott et al. 2020), and has been defined for the IUCN Green Status of Species assessments as the short-term (10 years or 3 generations, whichever is longer) reduction in recovery that would occur if management of a population were to cease (Akçakaya et al. 2018).

Conservation reliance is often described as a binary condition, but more realistically exists on a continuum. Reed et al. (2012) categorized 33 endangered Hawaiian bird species into five groups within this spectrum based on their degree of conservation reliance. These groups ranged from those requiring “modest, often intermittent” or minimal management, to those species that were considered “possibly too late” to recover; only one species was not considered conservation reliant. Developing a system to quantify the degree of conservation reliance would improve prioritization of management actions within and across species following delisting (Bottrill et al. 2008, Goble et al. 2012, Brown et al. 2017); this is one function of the IUCN Green Status of Species assessments (Akçakaya et al. 2018). Most assessments of population viability focus on metrics such as the likelihood of persistence for some targeted period of time (Doak et al. 1994, Walters et al. 2002, Traill et al. 2007), with sensitivity analyses identifying combinations of vital rates that would allow the most rapid population growth (Beissinger and Westphal 1998, Mills et al. 2001, Fefferman and Reed 2006, Aiello-Lammens and Akçakaya 2017). The difference between the current value for a vital rate, if the population has a high probability of persistence, and the point at which a population’s probability of extinction increases to high (>0.8) levels, making the likelihood of persistence low, may be particularly useful for evaluating conservation reliance (e.g., Mortensen and Reed 2016, van Rees and Reed 2018). This is one measure of a population’s resistance (Pimm 1984) to extinction, and proximity to this point is relevant to management needs prior to and following delisting. Similarly, Hutchings (2015) identifies population sizes at which species have impaired positive responses to management efforts (an Allee-effects threshold), and de Silva and Leimgruber (2019) identified vital rate values to look for a tipping point between positive and negative population growth. The effort required to keep vital rates above the threshold between increasing and decreasing could be a useful metric of the degree of conservation reliance.

The catastrophic loss of endemic island species is a global problem largely driven by introduced species, overharvest, and habitat loss (e.g., Milberg and Tyrberg 1993, Wood et al. 2017). Although efforts to remove introduced species have grown, and are increasingly successful (e.g., Russell and Holmes 2015), in many cases a return to historical conditions is not possible (Scott et al. 2020). The Hawaiian avifauna is an excellent system to explore species loss and recovery of endangered species in light of these problems (Boyer 2008, Pratt et al. 2009) because it has lost 71 of 113 endemic species, and most that remain are considered globally threatened (Leonard 2008). The extant Hawaiian waterbirds, by contrast, are generally considered to be conservation success stories because management has not only slowed declines but resulted in population increases (Reed et al. 2011, Underwood et al. 2013) and subsequent calls for down-listing some species from endangered to threatened status under the U.S. Endangered Species Act. Nonetheless, concerns remain that any reduced protections could leave these species vulnerable to the same risks that caused the original declines, resulting in the loss of hard-earned progress.

Here we use population viability analysis (PVA), perturbation sensitivity analysis, and Bayesian state-space modeling to estimate current extinction risk, the potential to achieve specific down-listing criteria, and degree of conservation reliance for two Hawaiian waterbird taxa with contrasting life histories and levels of scientific understanding (for convenience, we refer to both taxa as “species” following the legal definition embodied in the Endangered Species Act.) The Hawaiian Stilt (Himantopus mexicanus knudseni), which is currently being considered for down-listing to threatened, has a small clutch size, is relatively long-lived, and has a long history of detailed demographic study (Reed et al. 1998a, Reed et al. 2015, Pratt and Brisbin 2020). In contrast, the Hawaiian Coot (Fulica alai) has a large clutch size, shorter lifespan, and is poorly studied, with little published information on vital rates (Pratt and Brisbin 2020). Both are considered to be conservation reliant because of our current inability to remove threats caused by exotic invasive predators and plants that make habitat unsuitable (USFWS 2011, Reed et al. 2012, Underwood et al. 2013).

For either species to be down-listed from endangered to threatened, three major criteria must be satisfied: (A) a PVA must be conducted to determine the population size necessary for long-term viability, and (B) statewide surveys must show a stable or increasing trend with (C) population sizes above 2000 adult birds for at least 5 consecutive years (USFWS 2011). The last criterion was set in the original recovery plan for the species (USFWS 1978) through expert opinion (Ron Walker, lead author of the original plan, pers. comm. to JMR). Accordingly, the goals of this study were to: 1) quantitatively assess the probability of the statewide populations of either species being sustained at levels at or above the down-listing threshold of 2000 individuals, 2) conduct individual-based PVAs for both species to assess their long-term viability, and 3) use sensitivity analysis of vital rate values within the PVA model to estimate the degree of conservation reliance in these species.

METHODS

Study system

The Hawaiian Stilt and Hawaiian Coot are year-round residents of coastal wetlands, and they are endangered because of wetland loss, invasion by exotic plants that alter their habitat, and introduced predators and competitors (Coleman 1981, Griffin et al. 1989, USFWS 2011). One result of human activities on the Hawaiian Islands has been a reduction in wetland cover and degradation in habitat quality (Shallenberger 1977, Griffin et al. 1989, van Rees and Reed 2014). Habitat degradation and pressures from anthropogenic impacts like altered hydrology and invasive species require frequent and continuous management action. Management in the form of flooding, mowing, or predator trapping takes place on the scale of days to weeks to adjust to rapidly changing water levels, spreading invasive emergent vegetation, and incursions by invasive predators. Both Hawaiian Stilts and Coots are present on the seven largest Hawaiian Islands (Hawaiʻi, Maui, Molokaʻi, Oʻahu, Kauaiʻi, Lanaʻi, Kahoʻolawe; Engilis and Pratt 1993, Pratt and Brisbin 2020), but are absent from smaller islands which have scarce rainfall and little surface water to support wetlands. Higher numbers of both species are reported now than during the 1960s, with annual counts of coots occasionally exceeding 2000 individuals in the last 30 years (Fig. 1; Banko 1987, Engilis and Pratt 1993, Reed et al. 2011).

Goal 1: Estimating equilibrium population density using Bayesian time-series models

Biannual surveys of Hawaiian Stilts have never reported range-wide values above 2000 individuals, and Hawaiian Coot populations are only seldom above this value (Reed et al. 2011, USFWS 2011, Camp et al. 2016), so we analyzed long-term time series of statewide abundances using a Bayesian state-space model (Delean et al. 2013, van Rees et al. 2020) to estimate range-wide equilibrium population sizes for the two species and determine if this down-listing criterion is currently achievable. We used the posterior probability distribution of parameter estimates of equilibrium population size to assess whether current habitats within the state of Hawaiʻi are capable of sustaining population levels above the down-listing threshold (goal 1 of this study). This method provides a conservative quantitative metric of the populations’ achievement of the down-listing criterion of 2000 individuals by measuring the probability that this abundance could theoretically be achieved. We assessed the long-term viability and conservation reliance of both species’ populations using PVA methods outlined in later sections.

We used a Bayesian framework for model-fitting because it has the advantage of producing easily communicable results that can be described using real probabilities derived from the posterior probability distributions. Results presented as actual probabilities are easier to relay to decision-makers for endangered species listings. Bayesian approaches also provide the opportunity to use a priori knowledge about other aspects of a species’ biology to guide parameter estimation, such that empirical data on better-studied parameters can facilitate the estimation of an unknown parameter (Hooten and Hobbs 2015). This approach is broadly applicable to recovery planning where there are sufficient time-series data, even if vital rate information is missing.

We used Bayesian state-space models to estimate the statewide equilibrium population size for each species with the specific interest of evaluating whether current population dynamics were consistent with an equilibrium population size of at least 2000 individuals. We parameterized these models using time series of abundance data from summer and winter counts of the state of Hawaii’s biannual waterbird survey (Fig. 1), following the approach used by van Rees et al. (2020) for the Hawaiian Stilt. Biannual waterbird survey data consist of both summer (August) and winter (January) counts along consistent routes in all known major habitats for the subspecies (Engilis and Pratt 1993, Camp et al. 2014, Hawaii Division of Forestry and Wildlife 2017). We conducted all analyses separately for winter and summer data.

Bayesian state-space models are considered among the best approaches to estimating parameters such as the intrinsic rate of population growth (r_max) and equilibrium population size (sometimes treated as extant carrying capacity, K) using time-series datasets (Delean et al. 2013). Such analyses explicitly account for detection error which can otherwise bias model outcomes (Dennis et al. 2006, Freckleton et al. 2006, Ahrestani et al. 2013). Van Rees et al. (2020) found that state-space population models did not perform better than non-state-space models for the Hawaiian Stilt, but detection rates may be lower for the Hawaiian Coot, which can occupy more dense habitats than the stilt (Chang 1990). We thus employed both state-space and non-state-space versions of all our nonlinear population models for estimating equilibrium population size to evaluate if fit differed between model types.

For both species, we fit Bayesian state-space and non-state-space versions of five population models (exponential or density independent, Ricker, theta-logistic, Gompertz, and a conditional model; Table A1) as well as one-, two-, and three-year time-lagged versions of the Ricker and conditional models, and compared them using leave-one-out cross-validation (cf., Hooten and Hobbs 2015).

The conditional model was designed to exhibit no density-dependent impacts on population growth (i.e., exponential growth) until a threshold population size was passed, and then Ricker density-dependent dynamics after that (as in van Rees et al. 2020). We selected the level of this threshold by comparing frequentist versions of conditional models for a range of threshold values via AICc and used the threshold value for which AICc was minimized (van Rees et al. 2020). An alternative approach would be to set the population threshold as an additional parameter in the model to be explored via Markov-chain Monte Carlo (MCMC) sampling, but models did not converge with this additional parameter. We used the posterior probability distributions of the top (best fit) model for each taxon to estimate the probability that the K parameter (equilibrium population size) was at least 2000. All models were fit using MCMC sampling in JAGS (version 4.3.0; Plummer 2003) using the packages Runjags (Denwood 2016), and jagsUI (Plummer 2018, Kellner et al. 2019) in R (version 3.5.1; R Core Team 2018).

We used vague priors for K (uniform, bounded at 1000–3000 individuals) to achieve an unbiased estimate of carrying capacity under current conditions (equilibrium population size) for both species, and loosely informative priors for observation and process variation (Table A2) to take advantage of prior knowledge of the species’ biology from independent data sources. The minimum and maximum values for K were based on observed population fluctuations over the last 30 years. For both taxa, priors for r_max were defined by a normal distribution with mean and standard deviation determined by a deterministic demographic model in Vortex (described in Lacy et al. 2017, van Rees et al. 2020, Table A1). This amounted to running a deterministic version of the model with all parameters set to the maximum observed values for all vital rates. This simulation generated a realistic estimate of a biologically possible maximum growth rate for each species around which to base prior distributions for r_max in Bayesian population models. Importantly, this approach served to discourage the search algorithm from using biologically unreasonable values for this parameter.

Priors for process variation in population growth rate and observation error were defined using gamma distributions, and shape parameters were selected to ensure that realistic and observed values for both parameters in both species were within the given distribution, but that larger nonzero values were possible and would be searched by the MCMC algorithm. All priors and initial values used to conduct MCMC sampling are shown in Table A2. We based process variation in growth rate for the stilt on variance in population growth rate from Reed et al. (1998a). Because we lacked such prior information for coots we used priors that were higher and less informative (more spread) to reflect higher observed apparent variation across time and greater uncertainty for that species. Parameters of prior distributions for observation error (i.e., detection error) for both species were based on plausible detection error from Chang (1990), again with the possibility of much higher errors for coots (Table A2). These were defined using a gamma distribution, which allowed for non-negative values (i.e., only allowing undercounting), the shape of which was determined using a graphing calculator and values from Chang (1990) for the order of magnitude of likely values. Unlike the Hawaiian Gallinule (Gallinula galeata sandvicensis), detection error in both species is likely to be very low because of their tendency to occupy more open parts of wetland habitats, whereas stilts use open sparsely-vegetated areas and coots are often found in open water.

Goal 2: Baseline population viability analyses

We used an individual-based simulation model conducted in Vortex 10 (Lacy and Pollak 2014) to address goals (2) and (3) of this project: assess the long-term viability of the two focal species and estimate the strength of conservation reliance by testing the effects of changing parameter values on predicted viability. Using Vortex models, we generated population projections for both species until 2100, an approximately 80-year time frame similar to that used in van Rees and Reed (2018) for the Hawaiian Gallinule. Extensive inter-island movements by marked Hawaiian Stilts and Hawaiian Coots (Udvardy 1960, Banko 1988, Reed et al. 1998b, Dibben-Young 2010, Riggs 2016) and high levels of gene flow in the latter (Sonsthagen et al. 2018) suggest little population subdivision, so we modeled both species as single statewide populations.

We evaluated two persistence criteria: probability of extinction, which was defined as when individuals of only one sex remained, and the quasi-extinction probability (e.g., Holmes et al. 2007), estimated as the probability of the population size dropping below 50 adults. We selected the latter criterion because it is generally considered to be small enough that a population is at significant risk of extinction due to demographic stochasticity (Burgman et al. 1993, Lande 1998). Baseline models were run for 1000 replicates; initial population sizes were set from recent biannual waterbird survey data at 1700 for the stilt and 1280 for the coot, starting with an even sex ratio and stable age distribution. We evaluated viability from these models based on (1) probability of extinction by the year 2100, (2) mean population size by the year 2100, and (3) the mean stochastic growth rate estimated in Vortex. We assumed no correlation between reproduction and survival for either species because nest or chick loss rarely involves adult mortality.

The reproductive systems of both taxa were modeled as long-term monogamous based on current knowledge and observations of the species (Byrd et al. 1985, Reed et al. 1998a). Although cooperative breeding has been observed in the stilt (A. Dibben-Young pers. comm.), this was only in a single small group of birds in a highly habitat-limited context, and we have no evidence that it is a sufficiently widespread phenomenon to affect population dynamics. For Hawaiian Stilts, we set a maximum age of 30 based on a large mark-recapture dataset (Reed et al. 2015) and the maximum age of reproduction for both sexes at 30 because there is no evidence of reproductive senescence (Table 2). We took all other reproductive parameter estimates for Hawaiian Stilts from Reed et al. (1998a). For Hawaiian Coots, we used an age of first breeding of 1 given anecdotal observations (M. Silbernagle, pers. comm.) and data from the American Coot Fulica americana (Brisbin and Mowbray 2020) and assigned a maximum lifespan of 9 based on a mark-recapture dataset assembled and analyzed in this study. We assumed a 1:1 sex ratio at hatch for both species.

We followed the approach of van Rees and Reed (2018) for implementing density-dependent population dynamics in models for both species, setting the Allee effect parameter to zero in the absence of evidence for such effects, and setting the slope parameter to 20 so that density-dependent impacts were not introduced until population size was above 80% of the model’s population ceiling parameter, called carrying capacity K in Vortex, but distinct from the equilibrium population size estimated in our Bayesian time-series models. We assumed density dependence would arise based on territorial dynamics in Hawaiian Coots (Riggs 2016) and recent evidence from Hawaiian Stilts (van Rees et al. 2020). We assumed no mate monopolization and that all males of breeding age were in the breeding pool.

For Hawaiian Stilts, we used the reproduction dataset from Reed et al. (1998a) to define the distribution of brood sizes (number of hatched young) and number of broods per year. Lacking similar information on the Hawaiian Coot we followed the methods of van Rees and Reed (2018) to generate a distribution of brood frequencies using observed nest-failure rates from Chang (1990) and the binomial equation, assuming up to four breeding attempts per breeding season irrespective of fate (M. Silbernagle, pers. comm.). We specified a distribution of brood sizes based on data from Chang (1990).

We estimated mean and standard deviation of juvenile (age 0-1) and first-year (age 1-2) mortality for stilts using draws from the posterior probability distribution produced by Reed et al. (2015), and modeled percent mortality for each subsequent year until age 30 according to the best fit line for annual adult survival from the same study: 22.4 - (0.40*(A-2)), where A is bird age, and the standard deviation of the posterior probability distribution of adult survival was used to account for uncertainty and natural variability in annual survival.

We estimated mean annual apparent survival of adult Hawaiian Coots using standard Cormack-Jolly-Seber (CJS) models in program MARK (version 8.2, White and Burnham 1999) and mark-resight data consisting of >1000 resightings of 596 individuals over 12 years (A. Dibben-Young, unpublished data). Years were defined by calendar date for computational convenience. We compared the likelihood of a suite of potential survival models using AICc: (1) a null model with constant survival (φ) and detection (p); (2) a model with constant survival, but variable detection among years; (3) a model where survival varied by wetland, but detection was constant; (4) a model where both survival and detection varied among years; two models where survival varied by wetland, and detection either (5) varied by year, or (6) varied by wetland; and four models where detection varied by wetland, year, and their interaction and survival was either (7) constant, or varied by (8) year, (9) wetland, or (10) year, wetland, and their interaction). We implemented "goodness-of-fit" tests using program U-CARE (Choquet et al. 2009).

Given the limited information on within-year variation in adult survival for Hawaiian Coots, we used an estimate for American Coots, which has nearly identical values for apparent adult survival (Arnold et al. 2016). We estimated first-year survival rates from Chang (1990) and Byrd et al. (1985). Because we have little to no information on variance in juvenile survival for Hawaiian Coots we chose a value that resulted in population dynamics qualitatively similar to those observed in recent decades using the trajectories of simulated populations.

We set the population ceiling parameter K for Hawaiian Stilts as the equilibrium population size estimated by the top time-series population model from van Rees et al. (2020). For Hawaiian Coots, however, there is high inter-annual variation in estimated population size (Engilis and Pratt 1993), so we set this parameter at a higher level than the estimated carrying capacity and used the highest observed population size since 1990 (Table 2). We adjusted the shape parameter B in Vortex’s density dependence equation such that density-dependent feedback would be exhibited at population sizes approaching actual equilibrium size. This allowed simulated Hawaiian Coot populations to fluctuate similarly to observed statewide counts over the last several decades.

Goal 3: Conservation reliance using PVA Sensitivity analysis

We assessed the effects of changes to vital rate parameters, a proxy for different hypothetical levels of management, on the outcomes of PVAs for both species to provide a measure of conservation reliance (goal 3). While few data exist to compare vital rates between managed and unmanaged populations of either species, a general examination of the vital rate values linked to elevated extinction risk, and the proximity of those values to present estimates for our target species, provides a means of assessing conservation reliance. Examining this relationship in conjunction with the associated uncertainty in parameter values due to measurement error and natural variation provides further guidance on where critical management-relevant information is missing and where research effort should be directed.

We tested the influence of selected vital rate parameter values on model output using two approaches. First, we employed logistic regression to assess the relative contribution of each vital rate parameter to the probability of extinction. We used the Latin hypercube sampling method in Vortex with 10 iterations each for 1000 combinations of parameter values sampled across the range of observed values (Mortensen and Reed, 2016; Table 1). All parameters were varied simultaneously from random distributions based on the ranges provided (Table 2) and a logistic regression was performed relating these values as covariates to the probability of extinction estimated for each simulation run. This method broadly assessed the relative importance of different vital rate parameters to population viability.

Second, we used perturbation analysis to find vital rate parameter values at which large, abrupt increases in the probability of extinction occurred (Reed et al. 1998a, Reed et al. 2011, van Rees and Reed 2018), to highlight vital rates that might be driving conservation reliance. For each perturbation analysis, we held values of all parameters not being assessed at baseline levels and compared the probability of extinction for each simulation. Perturbation analysis is useful in detecting ranges of parameter values where the probability of extinction shifts from being low to high (or vice-versa), providing minimum vital rate values to be maintained by management. We perturbed values of the mean and standard deviation of juvenile survival, adult survival, nest-failure rate, mean brood size, proportion of females breeding at low population densities, and carrying capacity (values and intervals in Table 1). We also perturbed sex ratio at hatch for coots to examine the potential impacts of a population with 69% males, which has been reported on O‘ahu (Riggs 2016).

We used the proximity of empirical parameter estimates for vital rates based on field data, and their margin of uncertainty, to observed parameter values with high (≥80%) probability of extinction from our perturbation analysis as a quantitative test for conservation reliance. Because most of the vital rate information collected for the species in this study is derived from field studies on managed populations, we judged this comparison as a reasonable heuristic for degree of conservation reliance. Specifically, where the confidence interval around observed values for a given vital rate encompassed a high extinction risk value, or was 10% of its present value to that point, and if that value was likely to be affected by cessation of management, we judged a given trait as indicative of conservation reliance. If management were to cease, that vital rate would likely be pushed to a value that caused a dramatic shift in population size.

RESULTS

Goal 1: Estimating equilibrium population density using Bayesian time-series models

Among Bayesian population time-series models, the top performing coot model was the non-state-space version of the conditional model using summer data with a one-year lag (Table 2). The next ranked model was the non-state-space version of the Ricker model with a two-year lag run on winter data, which also had a large difference in cross-validation value from the third best model. Among the top five Bayesian models assessed for Hawaiian Coots, four of which were non-state-space (did not include a detection parameter), four were density-dependent models, and four used summer count data.

The best fit model for both species indicated a probability of equilibrium population size (carrying capacity under current conditions) at or above 2000 of 0.06 or lower (Table 2). The posterior probability distributions of the top five models for the Hawaiian Coot indicated that the probability of the equilibrium population being above the goal of 2000 was between 0.04-0.11, while the probability for Hawaiian Stilts lay between 0.03-0.26 (Table 2).

Goal 2: Baseline population viability analyses

Our baseline population models yielded low probabilities of extinction (0.0 and 0.11) and quasi-extinction (0.0 and 0.16), for Hawaiian Stilt and Hawaiian Coot, respectively. Both species showed similar mean stochastic population growth rates (0.11 and 0.094, respectively). The Hawaiian Stilt population model resulted in no extinctions. For the small number of simulation runs ending in extinction for the Hawaiian Coot, the mean time to extinction was 53.5 years (95% CI: 28.5–78.5); for quasi-extinction, it was 40 years (95% CI: 22–73). The mean ending population size at year 80 for the Hawaiian Stilt was 1588, with the entire 95% confidence interval (1321–1855) below the down-listing goal of 2000 individuals. The mean for coots was 1617, also below the down-listing goal, but the confidence interval (0.0–3639) was much broader, encompassing both extinction and the target population size. The mean stochastic growth rate for Hawaiian Stilts was 0.11 (95% CI: -0.27–0.49), and for Hawaiian Coots, it was 0.09 (95% CI: -1.03–1.21). The maximum deterministic growth rate (r_max) based on the highest observed reproduction parameter values was 0.52 for stilts and 1.2 for coots.

Goal 3: Conservation reliance using- PVA Sensitivity analysis

Logistic regression

Sensitivity analysis by logistic regression showed mean adult mortality, percentage of adult females in the breeding pool, mean nest-failure rate, and mean juvenile mortality as the most important parameters affecting probability of extinction for Hawaiian Stilts. In contrast, the top four parameters for Hawaiian Coots all related to survival (Table 3).

Perturbation

For both species, sharp increases between low and high probabilities of extinction were identified for most vital rate values (Figs. 1 and 2; Table 4). In Hawaiian Coots, probability of extinction increased abruptly from 0 to near 1 when annual adult mortality exceeded approximately 60%. Uncertainty around this parameter estimate from observations in managed populations encompassed probability of extinction values up to 1. Probability of extinction also transitioned rapidly from 0 to 1 for mean juvenile mortality rates above 74%; no data on the uncertainty of this estimate were available, but the estimate derived from the literature was close to the point of rapid increase in probability of extinction. Nest failure rates above 20% showed a steady increase in probability of extinction, reaching 0.5 at a 34% failure rate and 1 at around 75%. The range of empirical estimates for nest-failure rate produced probabilities of extinction from 0.25–1, with the mean parameter estimate from field observations between 0.9–1. Probability of extinction for Hawaiian Coots was 1 for mean brood sizes below 2.6 chicks and declined to 0 by 4.4 chicks. Observed variation in this parameter value encompassed a full range of probabilities of extinction from 0–1.

For Hawaiian Coots, the probability of extinction increased linearly with variation in annual adult mortality, reaching 1 at a standard deviation of around 25% variation in annual mortality (Fig. A1). Variation in juvenile annual mortality showed a similar trend, with extinction probability increasing with values around 15% and reaching a maximum between 30% and 35% (Fig. A2). Probabilities of extinction decreased rapidly from 1 to near 0 at K above 250 individuals, showing a persistent small, but nonzero, probability of extinction when K exceeded 1000 individuals (Fig. A3).

Increasing the proportion of male coots at hatch generated a gradual increase in the probability of extinction, approaching 1 when there were >70% of males at hatch (Fig. A4), a value that is similar to the sex bias observed by Riggs (2016) on O‘ahu. Probability of extinction was 1 for populations with percentages of females reproductively active below 50%, and declined for higher values, dropping to near 0 at 95% of adult females in the breeding pool (Fig. A5).

For Hawaiian Stilts, mean annual adult mortality rates above 20% led to increased probabilities of extinction, reaching a value of 1 at ~41% mean annual mortality (Fig. 3; Table 4). Current uncertainty around mean annual adult mortality rates in managed areas corresponds with a probability of extinction below 0.15. Mean annual juvenile mortality showed an increase in probability of extinction for values above 40%, with a rapid rise in probability of extinction from 55–78%, reaching 1 at 79% (Fig. 3). The larger margin of uncertainty around parameter estimates for this vital rate in managed areas showed that the higher end of this estimate corresponded with probabilities of extinction on the order of 0.35–0.5. Simulated Hawaiian Stilt populations showed a rapid decline in probability of extinction for mean brood sizes averaging greater than 0.8 chicks. Probability of extinction dropped to zero above a mean brood size of 1.5. Uncertainty and natural variation in observed brood size from managed populations included probabilities of extinction from 0–1. Nest-failure rates above 50% drove a rapid increase in probability of extinction, reaching 1 at failure rates beyond 75%. The empirical parameter estimates for this value, based on observations in managed habitats, was below this point of increase in probability of extinction, corresponding with a probability of 0.0-0.5.

The variation in mean annual adult mortality for Hawaiian Stilts showed a sharp increase in quasi-extinction risk after 20%, reaching a maximum by 45% (Fig. A1; Table 4). Variation in juvenile mortality, in contrast, did not affect probability of quasi-extinction at values below 80% (Fig. A2). Probability of extinction dropped rapidly for simulations with carrying capacities above 550 individuals, approaching 0 as carrying capacities exceeded 1600–2000 individuals (Fig. A3). For the percentage of adult females in the breeding pool, probability of extinction decreased rapidly above 25%, slowing and approaching zero at around 75% (Fig. A5).

Importantly, the increases in extinction risk associated with varying vital rate values for mean brood size, mean annual juvenile mortality, and nest-failure rates for Hawaiian Coots were all near to, or within the range of uncertainty of, current estimates for those parameters (Fig. 2; Table 5). This was also true for juvenile mortality and mean brood size in Hawaiian Stilts, but not for other vital rates for that species (Fig. 3; Table 4).

DISCUSSION

Using the best available data on vital rates and individual-based population models we find that both the Hawaiian Stilt and Hawaiian Coot have a high likelihood of persisting to the end of the century under current habitat and management conditions, and assuming no major loss of habitat, as might be expected from development or sea level rise. However, almost all vital rate data used in population viability models came from populations that were managed to control predators and maintain suitable nesting conditions. Moreover, there are few individuals and no reproductive strongholds of these species in unprotected areas. Although few empirical data exist from unmanaged populations, the precipitous declines of these two species prior to management and widespread field surveys, and their coincidence of population recovery with the establishment of protected areas (Reed et al. 2011), strongly suggest that cessation of management would reduce the vital rates identified in this study as most important to population persistence in both species: nest success and both adult and juvenile survival rates. We anticipate that the vital rates used in our models are unlikely to be maintained with reduced management, and our analyses of conservation reliance through the perturbation of individual vital rate values suggest that risk of extinction could increase dramatically if reduced management resulted in less favorable rates.

Although both mortality and reproductive parameters were important for both coots and stilts, their relative importance differed. For Hawaiian Coots, likelihood of extinction was more sensitive to changes in mean and variation in mortality of adults and juveniles than to any reproductive parameters. In contrast, although adult mortality was the most influential variable in extinction risk for Hawaiian Stilts, reproductive parameters - especially the proportion of females nesting and nest-failure rate - ranked highly. This difference between species aligns with life-history differences, with coots having shorter life spans and a much larger reproductive potential. Management activities that focus on both adult and juvenile survival are thus likely to be most important for coots, while activities that focus on maximizing the number of successful nests, both by ensuring that females breed and that their nests survive, will be more important for stilts. Predator management, through removal and exclusion, will likely benefit all of these variables.

The confidence intervals around vital rate parameter estimates for the stilt are considerably smaller than those for the coot due to more abundant data and lower natural variation, but in the case of adult and juvenile mortality as well as brood size, the boundaries of these intervals contain or abut values associated with a high probability of extinction. The proximity of plausible vital rate values in both species to projected high-extinction-probability states suggests that even small shifts in some vital rates would significantly increase extinction risk. Even at some managed sites, vital rate estimates based on field observations fall within the range of values associated with high probability of extinction (e.g., Morin 1998). Although we do not have such contrasting data for most vital rates, field observations of these species and published findings for an ecologically related species (Hawaiian Gallinule: Nagata 1983, van Rees et al. 2018) suggest that reduced management for Hawaiian Stilts and Hawaiian Coots would have harmful impacts and easily push some of these parameters toward high-risk values. This concern is supported by a few studies outside protected areas where vital rates are reduced for the Hawaiian Gallinule, implying that analogous risks would be incurred with the cessation of invasive plant management and predator control.

The generalizable nature of our perturbation sensitivity analysis of vital rate values provides a quantitative approach to assessing conservation reliance in data-limited species. Particularly where empirical data comparing managed to unmanaged populations are scarce, this approach helps distill relevant information on current estimates of parameter values, uncertainty in those values due to observer variation and measurement error, and the proximity of plausible values to levels at which long-term population viability declines below acceptable levels. Although this is no replacement for quantitative comparisons using baseline data, controls, and replicated observations of wild populations, it provides a formal and organized way to use existing knowledge to inform decision-making with a quantitative illustration of potential risk, especially in situations where key information is impossible to collect without conducting questionable experiments (e.g., by removing protections to test their effects).

Our assessments of extinction risk are likely to be conservative because, in our simulations, only one vital rate needs to change for the worse for a species to decline to extinction, but reduced management would likely impact multiple rates simultaneously (Griffin et al. 1989). Reduced exotic invasive plant control, for example, would rapidly eliminate available habitat (Cox 1999, Bantilan-Smith et al. 2009), reducing number of nests, nest success, and average brood sizes (via competition for space). Similarly, unmanaged mammalian predators would simultaneously impact nest-failure rate, brood size, juvenile survival, and adult survival (Coleman 1981, Ohashi and Oldenburg 1992, Underwood et al. 2014). Thus, although predicted probability of extinction for both species is low, the evidence suggests that this prognosis is precarious and subject to dramatic change in the absence of management. To view this conclusion in a positive light, these two management strategies simultaneously affect a suite of vital rates, and specific management techniques are not needed to address these rates separately.

In the absence of empirical data on the impact of reduced or no management on our focal taxa, data on other species may be informative regarding the potential impacts on vital rates. For example, Schüttler et al. (2009) found nest-failure rates of 80-95% due to unmanaged invasive predators in solitary, ground-nesting waterbirds on an island off Chile. Nest failure rates of this magnitude would lead to a high probability of extinction for both of our modeled taxa, even without additional effects on juvenile and adult survival. A recent review of the efficacy of predator management for bird conservation ranked the removal and control of mammalian predators on islands, generally to protect nesting bird communities, as having the strongest empirical evidence for efficacy and conservation benefit (Sutherland et al. 2019), implying strong support for the need to continue management for the focal taxa in our study.

We did not include catastrophic decreases in vital rates in our viability assessment, although it is known that catastrophes can increase extinction risk (Mangel and Tier 1994, Dale et al. 1998). Very little is known about what catastrophes can or do regularly affect the island-wide populations of Hawaiian Stilts or Hawaiian Coots. Although there are hurricanes that hit Hawaii, occasional botulism outbreaks in Hawaiian wetlands, as well as occasional flooding from rain events (Coleman 1981, Fletcher et al. 2003, Work et al. 2010), such events are not known to have catastrophic effects on our target species.

Our predictions assume maintenance of an ecological status quo for both species, with both continued management and no large-scale systematic environmental changes. We know, however, that Hawaiian wetlands are likely to be strongly affected by climate change and sea level rise (Kane et al. 2015, Htun et al. 2016, van Rees and Reed 2018). Like Hawaiian Gallinules, which are typically found only in areas of low salinity, Hawaiian Coots are found primarily in fresh water, but appear to better tolerate brackish water (Byrd et al. 1985, Engilis and Pratt 1993). Although it has not been studied in Hawaii, American Coots show reduced breeding density and reproductive success in brackish water (Kantrud in Allen 1985). In contrast, Hawaiian Stilts use wetlands with a wide variety of salinities, from fresh water to hypersaline (Coleman 1981, Engilis and Pratt 1993). Their apparent tolerance for increased salinity may buffer the potentially negative impacts of saltwater intrusion on coastal wetland habitats under sea level rise, given that new habitats may be created, and salinized wetlands may not necessarily become uninhabitable to both species. Shorebirds as a group, however, appear to require access to fresh water, particularly for chicks (Rubega and Robinson 1997), so this should be investigated for Hawaiian Stilt management. Accordingly, the type of simple analyses conducted by van Rees and Reed (2018) to estimate loss in available habitats for Hawaiian Gallinules would not be as informative for these species, for which population responses are likely to be more nuanced, requiring hydrological modeling for changes in habitat abundance and quality. Such research should be considered a high priority and perhaps an additional criterion for down-listing of these species, as sea level rise is a potentially critical factor affecting extinction risk (Underwood et al. 2013, Kane et al. 2015). The degree to which individual wetlands of importance for these species have a clear topographic pathway for inland and upland migration under sea level rise will be a major factor in such analyses.

Beyond population viability, we also used a Bayesian hierarchical modeling approach to estimate equilibrium population size, which we used as an assessment of whether the down-listing criterion of 2000 individuals is attainable for either species. Given the low probability that the equilibrium population size is above 2000 individuals in either species, our analysis suggests that more habitats will have to be protected and managed to achieve this goal. Although both species use coastal wetlands, they differ significantly in their microhabitat affiliations, so managing for one species would reduce habitat for the other. Many protected wetlands are not currently managed (e.g., large portions of Kawainui marsh on O‘ahu) and they could likely support much higher populations with some investment.

One possibility for supplementing protected wetlands for waterbirds is through agricultural wetlands (van Rees 2018, Harmon et al. 2021). Taro lo‘i is a traditional agricultural system that is readily used for nesting or foraging by both species (Shallenberger 1977, USFWS 1978, Stone et al. 1989, Rader 2005, Gee 2007). It is possible that restoration of Indigenous land practices such as taro lo‘i, which create large tracts of artificial wetland, could support waterbird populations currently reliant upon managed areas (Winter et al. 2020). Little is known, however, about how to manage taro lo‘i to meet the needs of any of the Hawaiian waterbirds while also meeting agricultural goals. The potential benefits of jointly managing taro lo‘i for waterbirds and taro harvest suggests the need for investigation (Hawaii DLNR 2005). Ideally, such work would involve replicated, controlled experimental assessments (Macnab 1983), similar to studies of rice field management during the non-growing season to support waterbirds in other parts of the world (e.g., Elphick et al. 2010). A key question would be to determine whether there is a risk that taro fields could become ecological traps (Robertson and Hutto 2006).

A complementary alternative method for creating new waterbird habitats and increasing available habitat statewide is via the construction and ecological maintenance of freshwater and coastal natural infrastructure (Da Silva and Wheeler 2017). The restoration and creation of wetland ecosystems is increasingly recognized as an option for improving water security and climate resilience (Thorslund et al. 2017), and strategically locating mitigation or restoration sites could lead to ecological benefits, including increased waterbird habitats, connectivity between habitats, and the protection of offshore marine ecosystems (van Rees and Reed 2015, van Rees 2018). These benefits are especially important in urban communities like the population centers around and including Honolulu, which will be increasingly threatened by sea level rise and intensifying storm activity (Chausson et al. 2020, Seddon et al. 2020).

What do these results mean for delisting or down-listing possibilities for the species, particularly with respect to the concern about their conservation reliance? Given the small possibility of removing all of the threats to Hawaiian Stilt and Hawaiian Coot persistence on the islands, intensive management involving predator control and habitat management is critical for the foreseeable future. In the absence of active management, it is clear that these species are likely to decline rapidly. Conservation reliance is not explicitly addressed by the U.S. Endangered Species Act, so the implications of conservation reliance are not addressed in recovery planning (Scott et al. 2020). Therefore, allowing de-/down-listing in the presence of a mechanism for ongoing management likely would require changing the Act. If either species is down-listed, current legal requirements for management would not change. Delisting would be another matter because the mandate for management would be lifted. The proximity of several vital rate values to thresholds below which population persistence becomes much less likely suggests that easing management could result in either species becoming endangered once again. Since current estimates of all vital rates come from managed populations, even this precarious status quo is the current best-case scenario. Data for the Hawaiian Stilt are more abundant, replicated, and informative than those for the coot, so the viability assessment for that species engenders more confidence. Despite the predicted high chance of persistence, down-listing the species would require additional conservation actions to ensure that the minimum population size criterion of >2000 individuals is met. Such actions would need to be designed in ways that ensure that other endangered waterbirds are not adversely affected (e.g., Griffin et al. 1989).

LITERATURE CITED

Ahrestani, F. S., M. Hebblewhite, and E. Post. 2013. The importance of observation versus process error in analyses of global ungulate populations. Scientific Reports 3:3125. https://doi.org/10.1038/srep03125

Aiello-Lammens, M. E., and H. R. Akçakaya. 2017. Using global sensitivity analysis of demographic models for ecological impact assessment. Conservation Biology 31:116-125. https://doi.org/10.1111/cobi.12726

Akçakaya, H. R., E. L. Bennett, T. M. Brooks, M. K. Grace, A. Heath, S. Hedges, C. Hilton-Taylor, M. Hoffmann, D. A. Keith, B. Long, D. P. Mallon, E. Meijaard, E. J. Milner-Gulland, A. S. L. Rodrigues, J. P. Rodriguez, P. J. Stephenson, S. N. Stuart, and R. P. Young. 2018. Quantifying species recovery and conservation success to develop an IUCN Green List of Species. Conservation Biology 32:1128-1138. https://doi.org/10.1111/cobi.13112

Allen, A. W. 1985. Habitat Suitability Index Models: American Coot. Biological Report 82, U.S Fish and Wildlife Service, Washington, DC, USA.

Arnold, T. W., C. N. De Sobrino, and H. M. Specht. 2016. Annual survival rates of migratory shore and upland game birds. Wildlife Society Bulletin 40:470-476. https://doi.org/10.1002/wsb.669

Averill-Murray, R. C., C. R. Darst, K. J. Field, and L. J. Allison. 2012. A new approach to conservation of the Mojave Desert tortoise. Bioscience 62:893-899. https://doi.org/10.1525/bio.2012.62.10.9

Banko, W. E. 1987. History of Endemic Hawaiian Birds. Part I. Population Histories - Species Accounts Freshwater Birds: Hawaiian Coot `Alae-`ke`oke`o in A. H. R. C. Cooperative National Park Resources Studies Unit, editor. University of Hawaiʻi at Manoa, Honolulu, Hawaii.

Banko, W. E. 1988. History of Endemic Hawaiian Birds. Part I. Population Histories - Species Accounts Freshwater Birds: Hawaiian Stilt Ae`oin A. H. R. D. Cooperative National Park Resources Studies Unit, editor. University of Hawaiʻi at Manoa, Honolulu, Hawaii.

Bantilan-Smith, M., G. L. Bruland, R. A. MacKenzie, A. R. Henry, and C. R. Ryder. 2009. A comparison of the vegetation and soils of natural, restored, and created coastal lowland wetlands in Hawai‘i. Wetlands 29:1023-1035. https://doi.org/10.1672/08-127.1

Beissinger, S. R., and M. I. Westphal. 1998. On the use of demographic models of population viability in endangered species management. Journal of Wildlife Management 62:821-841. https://doi.org/10.2307/3802534

Bottrill, M. C., L. N. Joseph, J. Carwardine, M. Bode, C. N. Cook, E. T. Game, H. Grantham, S. Kark, S. Linke, E. McDonald-Madden, R. L. Pressey, S. Walker, K. A. Wilson, and H. P. Possingham. 2008. Is conservation triage just smart decision making? Trends in Ecology & Evolution 23:649-654. https://doi.org/10.1016/j.tree.2008.07.007

Boyer, A. G. 2008. Extinction patterns in the avifauna of the Hawaiian islands. Diversity and Distributions 14:509-517. https://doi.org/10.1111/j.1472-4642.2007.00459.x

Brisbin, I., and T. Mowbray. 2020. Mowbray (2020). American Coot (Fulica americana), version 1.0. in A. Poole, and F. Gill, editor. Lab of Ornithology, Ithaca, New York. https://doi.org/10.2173/bow.y00475.01

Brown, D. J., C. A. Ribic, D. M. Donner, M. D. Nelson, C. I. Bocetti, and C. M. Deloria-Sheffield. 2017. Using a full annual cycle model to evaluate long-term population viability of the conservation-reliant Kirtland's warbler after successful recovery. Journal of Applied Ecology 54:439-449. https://doi.org/10.1111/1365-2664.12776

Burgman, M., S. Ferson, and R. Akçakaya. 1993. Risk Assessment in Conservation Biology. Chapman & Hall, London, UK.

Byrd, G. V., R. A. Coleman, R. J. Shallenberger, and C. S. Arume. 1985. Notes on the breeding biology of the Hawaiian race of the American Coot. `Elepaio 45:57-62.

Camp, R. J., K. W. Brinck, P. M. Gorresen, and E. H. Paxton. 2016. Evaluating abundance and trends in a Hawaiian avian community using state-space analysis. Bird Conservation International 26:225-242. https://doi.org/10.1017/S0959270915000088

Camp, R. J., K. W. Brinck, E. H. Paxton, and C. Leopold. 2014. Monitoring Hawaiian waterbirds: evaluation of sample methods to produce reliable estimates in T. R. HCSU-049, editor. Hawai`i Cooperative Studies Unit, University of Hawai'i, Hilo.

Chang, P. R. 1990. Strategies for Managing Endangered Waterbirds on Hawaiian National Wildlife Refuges. University of Massachusetts, Amherst, USA.

Chausson, A., B. Turner, D. Seddon, N. Chabaneix, C. A. J. Girardin, V. Kapos, I. Key, D. Roe, A. Smith, S. Woroniecki, and N. Seddon. 2020. Mapping the effectiveness of nature-based solutions for climate change adaptation. Global Change Biology 26:6134-6155. https://doi.org/10.1111/gcb.15310

Choquet, R., J. D. Lebreton, O. Gimenez, A. M. Reboulet, and R. Pradel. 2009. U‐CARE: Utilities for performing goodness of fit tests and manipulating CApture-REcapture data. Ecography 32:1071-1074. https://doi.org/10.1111/j.1600-0587.2009.05968.x

Coleman, R. A. 1981. The Reproductive Biology of the Hawaiian Subspecies of the Black-necked Stilt, Himantopus mexicanus knudseni. Pennsylvania State University, State College, USA.

Connor, W. P., K. F. Tiffan, J. A. Chandler, D. W. Rondorf, B. D. Arnsberg, and K. C. Anderson. 2019. Upstream migration and spawning success of chinook salmon in a highly developed, seasonally warm river system. Reviews in Fisheries Science & Aquaculture 27:1-50. https://doi.org/10.1080/23308249.2018.1477736

Cox, G. W. 1999. Alien Species in North America and Hawaii. Island Press.

Dale, V. H., A. E. Lugo, J. A. MacMahon, and S. T. Pickett. 1998. Ecosystem management in the context of large, infrequent disturbances. Ecosystems 1:546-557. https://doi.org/10.1007/s100219900050

Da Silva, J. M. C., and E. Wheeler. 2017. Ecosystems as infrastructure. Perspectives in Ecology and Conservation 15:32-35. https://doi.org/10.1016/j.pecon.2016.11.005

de Silva, S., and P. Leimgruber. 2019. Demographic tipping points as early indicators of vulnerability for slow-breeding megafaunal populations. Frontiers in Ecology and Evolution 7:171. https://doi.org/10.3389/fevo.2019.00171

Delean, S., B. W. Brook, and C. J. A. Bradshaw. 2013. Ecologically realistic estimates of maximum population growth using informed Bayesian priors. Methods in Ecology and Evolution 4:34-44. https://doi.org/10.1111/j.2041-210x.2012.00252.x

Dennis, B., J. M. Ponciano, S. R. Lele, M. L. Taper, and D. F. Staples. 2006. Estimating density dependence, process noise, and observation error. Ecological Monographs 76:323-341. https://doi.org/10.1890/0012-9615(2006)76[323:EDDPNA]2.0.CO;2

Denwood, M. J. 2016. Runjags: an R Package Providing Interface Utilities, Model Templates, Parallel Computing Methods and Additional Distributions for MCMC Models in JAGS. Journal of Statistical Software 71:1-25. https://doi.org/10.18637/jss.v071.i09

Dibben-Young, A. 2010. Thirty years of banding Hawaiian Coot and Hawaiian Moorhen, 1977-2008. `Elepaio 70:57-63.

Doak, D., P. Kareiva, and B. Kleptetka. 1994. Modeling population viability for the desert tortoise in the western Mojave desert. Ecological Applications 4:446-460. https://doi.org/10.2307/1941949

Elphick, C. S., O. Taft, and P. M. Lourenço. 2010. Management of rice fields for birds during the non-growing season. Waterbirds 33:181-192. https://doi.org/10.1675/063.033.s114

Engilis, A., Jr., and T. K. Pratt. 1993. Status and population trends of Hawaii's native waterbirds, 1977-1987. Wilson Bulletin:142-158.

Fefferman, N. H., and J. M. Reed. 2006. A vital rate sensitivity analysis for nonstable age distributions and short-term planning. Journal of Wildlife Management 70:649-656. https://doi.org/10.2193/0022-541X(2006)70[649:AVRSAF]2.0.CO;2

Fletcher, C., J. Rooney, M. Barbee, S.-C. Lim, and B. Richmond. 2003. Mapping shoreline change using digital orthophotogrammetry on Maui, Hawaii. Journal of Coastal Research pp. 106-124.

Freckleton, R. P., A. R. Watkinson, R. E. Green, and W. J. Sutherland. 2006. Census error and the detection of density dependence. Journal of Animal Ecology 75:837-851. https://doi.org/10.1111/j.1365-2656.2006.01121.x

Gee, H. K. 2007. Habitat Characteristics of Refuge Wetlands and Taro lo'i Used by Endangered Waterbirds at Hanalei National Wildlife Refuge, Hawai'i. South Dakota State University, Brookings, USA.

Goble, D. D. 2009. The Endangered Species Act: What we talk about when we talk about recovery. Natural Resources Journal 49:1-44. https://doi.org/10.5860/rusq.49n1.18

Goble, D. D., J. A. Wiens, J. M. Scott, T. D. Male, and J. A. Hall. 2012. Conservation-reliant species. Bioscience 62:869-873. https://doi.org/10.1525/bio.2012.62.10.6

Griffin, C. R., R. J. Shallenberger, and S. I. Fefer. 1989. Hawaii's endangered waterbirds: a resource management challenge. Pages 1165-1175. Freshwater Wetlands and Wildlife. USDOE Office of Scientific and Technical Information, Oak Ridge, Tennessee, USA.

Harmon, K. C., K. B. Winter, N. Kurashima, C. H. Fletcher, H. H. Kane, and M. R. Price. 2021. The role of indigenous practices in expanding waterbird habitat in the face of rising seas. Anthropocene 34:100293. https://doi.org/10.1016/j.ancene.2021.100293

Hawaii Division of Forestry and Wildlife. 2017. Hawaii Waterbird Database. State of Hawaii in Division of Forestry and Wildlife, Department of Land and Natural Resources, editor, Honolulu, Hawaii.

Hawaii DLNR. 2005. Hawaii's Comprehensive Land and Resource Strategy. Hawaii Department of Land and Natural Resources, Honolulu, Hawaii.

Holmes, E. E., J. L. Sabo, S. V. Viscido, and W. F. Fagan. 2007. A statistical approach to quasi‐extinction forecasting. Ecology Letters 10:1182-1198. https://doi.org/10.1111/j.1461-0248.2007.01105.x

Hooten, M. B., and N. T. Hobbs. 2015. A guide to Bayesian model selection for ecologists. Ecological Monographs 85:3-28. https://doi.org/10.1890/14-0661.1

Htun, H., S. A. Gray, C. A. Lepczyk, A. Titmus, and K. Adams. 2016. Combining watershed models and knowledge-based models to predict local-scale impacts of climate change on endangered wildlife. Environmental modelling & software 84:440-457. https://doi.org/10.1016/j.envsoft.2016.07.009

Hutchings, J. A. 2015. Thresholds for impaired species recovery. Proceedings of the Royal Society B-Biological Sciences 282:20150654. https://doi.org/10.1098/rspb.2015.0654

Kane, H. H., C. H. Fletcher, L. N. Frazer, and M. M. Barbee. 2015. Critical elevation levels for flooding due to sea-level rise in Hawai ‘i. Regional Environmental Change 15:1679-1687. https://doi.org/10.1007/s10113-014-0725-6

Kellner, K., M. K. Kellner, and J. SystemRequirements. 2019. Package ‘jagsUI’.

Lacy, R., and J. Pollak. 2014. Vortex: a Stochastic Stimulation of the Ecological Process. IUCN SSC Conservation Planning Specialist Group & Chicago Zoological Society, Brookfield, Illiois, USA.

Lacy, R. C., P. S. Miller, and K. Traylor-Holzer. 2017. Vortex 10 User's Manual. 6 September 2017 updatein a. C. Z. S. IUCN SSC Conservation Breeding Specialist Group, editor, Apple valley, Minnesota, USA.

Lande, R. 1998. Demographic stochasticity and Allee effects on a scale with isotropic noise. Oikos 83:353-358. https://doi.org/10.2307/3546849

Leonard, D. L., Jr. 2008. Recovery expenditures for birds listed under the US Endangered Species Act: The disparity between mainland and Hawaiian taxa. Biological Conservation 141:2054-2061. https://doi.org/10.1016/j.biocon.2008.06.001

Macnab, J. 1983. Wildlife management as scientific experimentation. Wildlife Society Bulletin 11:397-401.

Mangel, M., and C. Tier. 1994. Four facts every conservation biologists should know about persistence. Ecology 75:607-614. https://doi.org/10.2307/1941719

Milberg, P., and T. Tyrberg. 1993. Naive birds and noble savages - a review of man-caused prehistoric extinctions of island birds. Ecography 16:229-250. https://doi.org/10.1111/j.1600-0587.1993.tb00213.x

Mills, L. S., D. F. Doak, and M. J. Wisdom. 2001. Elasticity analysis for conservation decision making: reply to Ehrlen et al. Conservation Biology 15(1):281-283. https://www.jstor.org/stable/2641669

Morin, M. P. 1998. Endangered waterbird and wetland status, Kaloko-Honokohau National Historical Park, Hawai'i Islandin P. I. S. C. USGS-BRD, editor. Cooperative National Park Resources Studies Unit, University of Hawai`i at Manoa.

Mortensen, J. L., and J. M. Reed. 2016. Population viability and vital rate sensitivity of an endangered avian cooperative breeder, the white-breasted thrasher (Ramphocinclus brachyurus). Plos One 11(2):e0148928. https://doi.org/10.1371/journal.pone.0148928

Nagata, S. E. 1983. Status of the Hawaiian Gallinule on Lotus Farms and a Marsh on Oahu, Hawaii. Thesis for Colorado State University, Fort Collins, Colorado, USA.

Neel, M. C., A. K. Leidner, A. Haines, D. D. Goble, and J. M. Scott. 2012. By the numbers: how is recovery defined by the US Endangered Species Act? Bioscience 62:646-657. https://doi.org/10.1525/bio.2012.62.7.7

Ohashi, T. J., and J. G. Oldenburg. 1992. Endangered species in the Pacific Islands: the role of animal damage control. Proceedings of the Vertebrate Pest Conference.

Pimm, S. L. 1984. The complexity and stability of ecosystems. Nature 307:321-326. https://doi.org/10.1038/307321a0

Plummer, M. 2003. JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling. Pages 1-10. Proceedings of the 3rd international workshop on distributed statistical computing. Vienna, Austria.

Plummer, M. 2018. rjags: Bayesian graphical models using MCMC. R package version 4.8. in.

Pratt, H., and J. Brisbin, IL. 2020. Hawaiian Coot (Fulica alai), version 1.0. in A. Poole, and F. Gill, editor. Birds of the World. Lab of Ornithology, Ithica, New York, USA. https://doi.org/10.2173/bow.hawcoo.01

Pratt, T. K., C. T. Atkinson, P. C. Banko, B. L. Woodworth, and J. D. Jacobi. 2009. Conservation Biology of Hawaiian Forest Birds: Implications for Island Avifauna. Yale University Press, New Haven, Connecticut, USA.

R Core Team. 2018. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Rader, J. A. 2005. Response of vegetation and endangered waterbirds to habitat management techniques at Kealia Pond National Wildlife Refuge. Thesis M.S. South Dakota State University, Brookings, South Dakota, USA.

Reed, J. M., D. W. Desrochers, E. A. Vanderwerf, and J. M. Scott. 2012. Long-term persistence of Hawaii's endangered avifauna through conservation-reliant management. Bioscience 62:881-892. https://doi.org/10.1525/bio.2012.62.10.8

Reed, J. M., C. S. Elphick, E. N. Ieno, and A. F. Zuur. 2011. Long-term population trends of endangered Hawaiian waterbirds. Population Ecology 53:473-481. https://doi.org/10.1007/s10144-011-0262-9

Reed, J. M., C. S. Elphick, and L. W. Oring. 1998a. Life-history and viability analysis of the endangered Hawaiian Stilt. Biological Conservation 84:35-45. https://doi.org/10.1016/S0006-3207(97)00077-3

Reed, J. M., C. R. Field, M. D. Silbernagle, A. Nadig, K. Goebel, A. Dibben-Young, P. Donaldson, and C. S. Elphick. 2015. Application of the complete-data likelihood to estimate juvenile and adult survival for the endangered Hawaiian stilt. Animal Conservation 18:176-185. https://doi.org/10.1111/acv.12156

Reed, J. M., M. D. Silbernagle, K. Evans, J. A. Engilis, and L. W. Oring. 1998b. Subadult movement patterns of the endangered Hawaiian Stilt (himantopus mexicanus knudseni). Auk 115:791-797. https://doi.org/10.2307/4089431

Riggs, R. 2016. Movement and sexual dimorphism of the endangered Hawaiian Coot,(Fulica alai), on Oahu. University of Hawaii at Manoa, Honolulu

Robertson, B. A., and R. L. Hutto. 2006. A framework for understanding ecological traps and an evaluation of existing evidence. Ecology 87:1075-1085. https://doi.org/10.1890/0012-9658(2006)87[1075:AFFUET]2.0.CO;2

Rubega, M. A., and J. A. Robinson. 1997. Water salinization and shorebirds: emerging issues. International Wader Studies 9:45-54.

Russell, J. C., and N. D. Holmes. 2015. Tropical island conservation: rat eradication for species recovery. Biological Conservation 185:1-7. https://doi.org/10.1016/j.biocon.2015.01.009

Schüttler, E., R. Klenke, S. McGehee, R. Rozzi, and K. Jax. 2009. Vulnerability of ground-nesting waterbirds to predation by invasive American mink in the Cape Horn Biosphere Reserve, Chile. Biological Conservation 142:1450-1460. https://doi.org/10.1016/j.biocon.2009.02.013

Scott, J. M., D. D. Goble, A. M. Haines, J. A. Wiens, and M. C. Neel. 2010. Conservation-reliant species and the future of conservation. Conservation Letters 3:91-97. https://doi.org/10.1111/j.1755-263X.2010.00096.x

Scott, J. M., D. D. Goble, J. A. Wiens, D. S. Wilcove, M. Bean, and T. Male. 2005. Recovery of imperiled species under the Endangered Species Act: the need for a new approach. Frontiers in Ecology and the Environment 3:383-389. https://doi.org/10.1890/1540-9295(2005)003[0383:ROISUT]2.0.CO;2

Scott, J. M., J. A. Wiens, B. Van Horne, and D. D. Goble. 2020. Shepherding Nature: The Challenge of Conservation Reliance. Cambridge University Press, Cambridge, UK. https://doi.org/10.1017/9781108377898

Seddon, N., A. Chausson, P. Berry, C. A. J. Girardin, A. Smith, and B. Turner. 2020. Understanding the value and limits of nature-based solutions to climate change and other global challenges. Philosophical Transactions of the Royal Society B: Biological Sciences 375:20190120. https://doi.org/10.1098/rstb.2019.0120

Shallenberger, R. J. 1977. An Ornithological Survey of Hawaiian Wetlands, vol 1. Ahuimanu Productions, Honolulu, Hawaii.

Sonsthagen, S. A., R. E. Wilson, and J. G. Underwood. 2018. Interisland genetic structure of two endangered Hawaiian waterbirds: The Hawaiian Coot and Hawaiian Gallinule. The Condor: Ornithological Applications 120:863-873. https://doi.org/10.1650/CONDOR-18-98.1

Stone, C. P., D. B. Stone, L. W. Cuddihy, and M. E. Lane. 1989. Conservation Biology in Hawai'i. University of Hawaii Cooperative National Park Resources Studies Unit.

Sutherland, W. J., L. V. Dicks, N. Ockendon, S. O. Petrovan, and R. K. Smith. 2019. What Works in Conservation. Open Book Publishers, Cambridge, UK. https://doi.org/10.11647/OBP.0179

Thorslund, J., J. Jarsjo, F. Jaramillo, J. W. Jawitz, S. Manzoni, N. B. Basu, S. R. Chalov, M. J. Cohen, I. F. Creed, R. Goldenberg, A. Hylin, Z. Kalantari, A. D. Koussis, S. W. Lyon, K. Mazi, J. Mard, K. Persson, J. Pietro, C. Prieto, A. Quin, K. Van Meter, and G. Destouni. 2017. Wetlands as large-scale nature-based solutions: Status and challenges for research, engineering and management. Ecological Engineering of Sustainable Landscapes 108:489-497. https://doi.org/10.1016/j.ecoleng.2017.07.012

Traill, L. W., C. J. A. Bradshaw, and B. W. Brook. 2007. Minimum viable population size: A meta-analysis of 30 years of published estimates. Biological Conservation 139:159-166. https://doi.org/10.1016/j.biocon.2007.06.011

Udvardy, M. D. F. 1960. Movements and concentrations of the Hawaiian Coot on the island of Oahu. `Elepaio 21:20-23.

Underwood, J. G., M. Silbernagle, M. Nishimoto, and K. Uyehara. 2013. Managing conservation reliant species: Hawai'i's endangered endemic waterbirds. Plos One 8:e67872. https://doi.org/10.1371/journal.pone.0067872

Underwood, J. G., M. Silbernagle, M. Nishimoto, and K. J. Uyehara. 2014. Non-native mammalian predator control to benefit endangered Hawaiian waterbirds. Proceedings of the Vertebrate Pest Conference 26:32-39. https://doi.org/10.5070/V426110653

U.S. Fish and Wildlife Service (USFWS). 1978. Hawaiian Waterbirds Recovery Plan, Portland, Oregon, USA.

U.S. Fish and Wildlife Service (USFWS). 2011. Recovery Plan for Hawaiian Waterbirds, second revision, Portland, Oregon, USA.

van Rees, C. B. 2018. Wetland conservation in Hawaii and the need for flagship species. `Elepaio 78:37-41.

van Rees, C. B., P. R. Chang, J. Cosgrove, D. W. DesRochers, H. K. Gee, J. L. Gutscher-Chutz, A. Nadig, S. E. Nagata, M. Silbernagle, and J. G. Underwood. 2018. Estimation of vital rates for the Hawaiian Gallinule, a cryptic, endangered waterbird. Journal of Fish and Wildlife Management 9:117-131. https://doi.org/10.3996/102017-JFWM-084

van Rees, C. B., and J. M. Reed. 2014. Wetland loss in Hawai’i since human settlement. Wetlands 34:335-350. https://doi.org/10.1007/s13157-013-0501-2

van Rees, C. B. and J. M. Reed. 2015. Water diplomacy from a duck's perspective: wildlife as stakeholders in water management. Journal of Contemporary Water Research & Education 155:28-42. https://doi.org/10.1111/j.1936-704X.2015.03193.x

van Rees, C. B. 2018. Wetland conservation in Hawai'i and the need for flagship species. `Elepaio 78:37-41.

van Rees, C. B., and J. M. Reed. 2018. Predicted effects of landscape change, sea level rise, and habitat management on the extirpation risk of the Hawaiian common gallinule (Gallinula galeata sandvicensis) on the island of O ‘ahu. PeerJ 6:e4990. https://doi.org/10.7717/peerj.4990

van Rees, C. B., G. Surya, and J. M. Reed. 2020. Multiple sources of evidence for density dependence in the endangered Hawaiian stilt (Himantopus mexicanus knudseni). Population Ecology 62:207-219. https://doi.org/10.1002/1438-390X.12037

Walters, J. R., L. B. Crowder, and J. A. Priddy. 2002. Population viability analysis for Red-cockaded Woodpeckers using an individual-based model. Ecological Applications 12:249-260. https://doi.org/10.1890/1051-0761(2002)012[0249:PVAFRC]2.0.CO;2

White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46:S120-S139. https://doi.org/10.1080/00063659909477239

Wiens, J. A., and T. Gardali. 2013. Conservation reliance among California's at-risk birds. Condor 115:456-464. https://doi.org/10.1525/cond.2013.120086

Wilsey, C. B., J. J. Lawler, D. Cimprich, and N. H. Schumaker. 2014. Dependence of the endangered Black-capped Vireo on sustained cowbird management. Conservation Biology 28:561-571. https://doi.org/10.1111/cobi.12176

Winter, K., T. Ticktin, and S. Quazi. 2020. Biocultural restoration in Hawaiʻi also achieves core conservation goals. Ecology and Society 25(1):26. https://doi.org/10.5751/ES-11388-250126

Wood, J. R., J. A. Alcover, T. M. Blackburn, P. Bover, R. P. Duncan, J. P. Hume, J. Louys, H. J. Meijer, J. C. Rando, and J. M. Wilmshurst. 2017. Island extinctions: processes, patterns, and potential for ecosystem restoration. Environmental Conservation 44:348-358. https://doi.org/10.1017/S037689291700039X

Work, T. M., J. L. Klavitter, M. H. Reynolds, and D. Blehert. 2010. Avian botulism: a case study in translocated endangered Laysan ducks (Anas laysanensis) on Midway Atoll. Journal of Wildlife Diseases 46:499-506. https://doi.org/10.7589/0090-3558-46.2.499