Habitat selection and potential fitness consequences of two early‐successional species with differing life‐history strategies

Abstract Habitat selection and its relationship to fitness is a fundamental concept in ecology, but the mechanisms driving this connection are complex and difficult to detect. Despite the difficulties in understanding such intricate relationships, it is imperative that we study habitat selection and its relationship with fitness. We compared habitat selection of least terns (Sternula antillarum) and piping plovers (Charadrius melodus) on the Missouri River (2012–2014) to examine the consequences of those choices on nest and chick survival. We hypothesized that plovers and terns would select habitat that minimized the chance of flooding and predation of eggs, chicks, and adults, but that plovers would also select habitat that would provide foraging habitat for their chicks. We developed an integrated habitat selection model that assessed selection across multiple scales (sandbar and nest scales) and directly modeled the effect of selection on nest and chick survival. In general, the species selected habitat in keeping with our hypotheses, such that predation and flooding, in particular, may have been reduced. Sandbar selection had either a negative or no appreciable effect on nest survival for both species across years. Nest‐site selection in 2012 had a generally positive effect on nest survival and chick survival for both terns and plovers, and this trended toward a negative effect by 2014. This result suggested that early selection decisions appeared to be adaptive, but we speculate that relatively high site fidelity and habitat degradation led to reduced benefit over time. Our results highlight the complex nature of habitat selection and its relationship to fitness.

complex ecological and evolutionary trade-offs can obscure signals and result in mismatches between predicted fitness relationships and actual outcomes (Chalfoun & Schmidt, 2012). Continued comparative study, however, can aid in understanding the mechanisms behind these relationships.
Habitat selection is the product of a complex suite of selective pressures and behavioral choices, and perhaps nowhere is this complexity more evident than in the choice of a nest site for a bird. Adults choose a location that not only protects their reproductive investment, but often also their own safety. This choice can carry consequences for survival (Amat & Masero, 2004;Miller, Grand, Fondell, & Anthony, 2007), nest success (Murray & Best, 2014;Stokes & Boersma, 1998), and potentially fitness (Braden, McKernan, & Powell, 1997;Clark & Shutler, 1999;Martin, 1988;Orians & Wittenberger, 1991). Although nest-site selection and its relationship with reproductive success is a widely researched topic (Jones, 2001), incongruences between selection and success are the norm and not the exception (Chalfoun & Schmidt, 2012). Despite this, knowledge of habitat selection can have important implications for guiding conservation and management, particularly for rare or sensitive species and in rapidly changing habitats.
Ground-nesting birds and their nests are vulnerable to a variety of threats during the breeding season, and their choices reflect complex trade-offs . Predation often is a primary cause of nest failure (Fletcher, Aebischer, Baines, Foster, & Hoodless, 2010;Smith, Gilchrist, & Smith, 2007), and the risk of predation may change with distance to predator habitat (Espie, Brigham, & James, 1996) and degree of concealment (Swaisgood et al., 2018), which may be evidenced in avoidance or preference for vegetative cover.
Individual nest success may not be the only selective pressure shaping selection. Thus, assuming random or maladaptive selection in the face of mismatched predictions may overlook other important fitness correlates (Chalfoun & Schmidt, 2012). For precocial birds whose chicks must feed themselves soon after hatch, proximity to foraging habitat can have a profound effect on habitat selection Walker et al., 2019) and chick survival (Cohen, Houghton, & Fraser, 2009;Gibson, Blomberg, Atamian, & Sedinger, 2017;Loegering & Fraser, 1995), and these selective pressures must be weighed against other concerns (Chalfoun & Schmidt, 2012).
Adult birds also must balance their own safety with that of their nests, and these trade-offs may obscure the relationship between selection and fitness (Chalfoun & Schmidt, 2012;Gomez-Serrano & Lopez-Lopez, 2014;Guilherme, Burnside, Collar, & Dolman, 2018).
Despite the similarities in their ecology, there are key life-history differences between the species. Plovers are territorial and their young are precocial, following their parents to wet sand feeding locations that the parents often defend from conspecifics (Catlin, Fraser, & Felio, 2015;Elliott-Smith & Haig, 2004). These feeding territories have an important role in successfully raising a brood and how adults select nesting territories (Cohen et al., 2009;Loegering & Fraser, 1995;Walker et al., 2019). Terns, however, nest colonially, defend their colonies against predators, typically lay fewer eggs, and their young are semi-altricial. Though the young are mobile soon after hatching, they rely on their parents to deliver them small fish until they are able to fish for themselves (Thompson et al., 1997).
Terns also have a faster breeding cycle with a 21-day nesting period and reaching flight at about 20 days, as opposed to plovers with a 34-day nesting period and at least 25 days before flight (Catlin, Felio, & Fraser, 2013;Elliott-Smith & Haig, 2004;Thompson et al., 1997). At least for plovers, the selection of a nest site can contribute to higher nest success in some cases (Espie et al., 1996;Prindiville-Gaines & Ryan, 1988), but the relationship with chick survival is more uncertain (Cohen et al., 2009), and studies that directly relate habitat to fitness and studies that compare selection between the species are lacking. Given the similarities in gross habitat selection, one might assume that factors affect least terns similarly, but these critical tests are largely lacking, and differences between the way the two species select habitat and its effects on fitness could help our understanding of the benefits of habitat selection.
The objectives of this study were (a) to compare the second-and third-order habitat selection of two, ground-nesting species that use F I G U R E 1 We studied habitat selection and its effects on nest and chick survival for (a) least terns (Sternula antillarum) and (b) piping plovers (Charadrius melodus) on the Missouri River (2012-2014)   (Espie et al., 1996;Sidle, Carlson, Kirsch, & Dinan, 1992), and thus, we hypothesized that birds would select habitat that minimized flooding (e.g., farther from the waterline, in dry rather than wet sand). The selection of habitat that fosters feeding of young is a key aspect to plover nest-site selection (Walker et al., 2019), and thus, we hypothesized that plovers would select sandbars and nest sites that provided these opportunities (e.g., higher proportion of wet sand, nearer to wet sand). In particular, we were interested in how the differences in life-history between plovers and terns would affect selection and how those differences in selection would affect reproductive success.

| Study area
The study took place in the Missouri National Recreational River  , with high water precluding nesting and territory establishment for both species on GVP. Following the flooding in 2011, however, there was a nearly 10-fold increase in open and sparsely vegetated sand, and plovers and terns resumed breeding at these sites (Hunt, Fraser, Friedrich, Karpanty, & Catlin, 2018).

| Field methods
We surveyed open and sparsely vegetated areas of emergent sandbars during the plover and tern breeding season (April-August) in search of nests. We located nests by grid-searching potential nesting habitat, using spotting scopes to look for incubating birds, and recognizing behavioral cues of adult birds (e.g., territorial and distraction displays). Upon discovery, we logged each nest location with a handheld GPS unit (Trimble Geo XT, Trimble Navigation, Ltd.). Geographic coordinates had a horizontal accuracy of ±15 cm. We floated the eggs to estimate developmental stage (Westerskov, 1950) and calculate the expected hatch date.
We attempted to check each nest every 2 days throughout the incubation period to determine nest fate, increasing our visit frequency within 3 days of the estimated hatch date when possible.
If we observed ≥1 chicks or if ≥1 eggs disappeared within 2 days of the estimated hatch date without material evidence of failure (e.g., eggs washed out of nest bowl, predator tracks at nest, bloody eggshell), we considered a nest successful Hunt et al., 2018;Nefas, Hunt, Fraser, Karpanty, & Catlin, 2018).
When no obvious signs of failure were present, but eggs disappeared >2 days before estimated hatch date, we considered the nest failed due to unknown cause.  Catlin et al., 2015;. Although the placement of wire predator exclosures (Melvin, Macivor, & Griffin, 1992) around plover nests is a common management practice to increase nest success (Johnson & Oring, 2002;Tan, Buchanan, Maguire, & Weston, 2015), none of the nests in this study were exclosed.
To assess chick survival, we banded all chicks with a unique combination of color bands (plovers) or a numbered, metal band (terns).
We searched for these individuals approximately every 2 days until all birds would have been fledged or the end of the season. Plovers were resighted from a distance or recaptured  to assess survival, but terns had to be physically recaptured to read their band. Common chick predators included the previously described suite of nest predators but also included predators such as great-horned owls (Bubo virginianus; Kruse et al., 2001). There had been some predator removal at these sites prior to the flooding in 2011 , but there was no predator control at these sites during the study.

| Image collection and classification
Pan-sharpened multispectral satellite imagery (5 m resolution) was

| Sandbar, nest, and random point attributes
We extracted habitat data related to each plover and tern sandbar (average value for the sandbar or minimum distance from any point on the sandbar, depending on the data type) and nest point from the classified land cover datasets. We assembled attributes that we predicted were related to predation (vegetation, isolation), flooding (wet sand, proximity to water), and foraging (proximity of wet sand and waterline; see Table 1 for all extracted values and specific hypothesized relationships with selection). The foraging hypothesis was specific to plovers, which have precocial young that need to feed themselves in situ soon after hatch. We did not have sufficient information on the quality of tern foraging habitats (open water) to perform similar analyses, but we did apply this hypothesis to tern sandbars and nest locations for comparison with the plover results.
In each year, we sampled approximately five unused sandbars for each used sandbar and eight random nest locations for each used nest for use in our habitat selection models. We used ArcGIS 10.5 (ESRI) to generate random points and to derive habitat data from the classified imagery. There were no limitations placed on the selection of unused sandbars and random points other than that they were within the boundaries of the river and were not in open water. All habitat covariates were standardized prior to inclusion in the analyses.

| Model
We developed an integrated habitat selection model that tied second-(sandbar) and third-order (nest-site) selection and their predicted values to a model of nest success and a model of chick survival. For sandbar (i) selection, we modeled use as a multinomial, where the outcomes (j) were: (a) only used by plovers, (b) used by both plovers and terns, and (c) unused. Very few sandbars were only used by terns (n = 6) so we lumped these into the second category for analysis.
Categories 1 and 2 contributed to selection for plovers, while only category 2 contributed to selection for terns. Thus, all sandbars chosen by plovers (regardless of tern behavior) were used to estimate plover sandbar selection. Likewise, only sandbars that terns selected (regardless of plover behavior) were used to estimate tern selection. For ease of description, we will refer to tern sandbars (i.e., used by both terns and plovers, but estimates for tern site selection only) and plover sandbars (only used by plovers).
We used our data to estimate year (y) and habitat-specific probabilities of sandbar occupancy. Each covariate coefficient (β j,x ) was assigned a diffuse normal distribution, centered at 0, and all covariates appeared in a single, global model (see Appendix S1 for code).
Then for each random and nest location (k), we modeled the probability that a specific nesting location was used as the joint probability that the location was a suitable nest site (q k ) and that the probability that the sandbar was occupied by the species (m).
As with the sandbar selection model, we used normal distributions for priors on the habitat coefficients (γ x,m ) to identify important parameters.
To estimate the effect of yearly sandbar and nest-site selection on daily nest survival and thus nest success, we used a year and species-specific logistic exposure model Hunt et al., 2018;Rotella, Taper, & Hansen, 2000;Shaffer, 2004). We chose to use the combined selection coefficients from the second-and thirdorder selection models rather than the specific habitat variables because such relationships often are complex, and correlation and misspecification can contribute to failures to detect associations between selection and success (Chalfoun & Schmidt, 2012). We used a standard, diffuse normal prior for the effects of sandbar and nestsite selection.
For chick survival, we used a state-space Cormack-Jolly-Seber model (Kéry & Schaub, 2012) to estimate daily age-specific chick survival, whose format was essentially the same as for nest success except that we modeled the effect of age (a, days since hatch): TA B L E 1 Variables used to estimate habitat selection at the sandbar and nest-site scale for piping plovers and least terns, nesting on the Missouri River (2012)(2013)(2014) and their relationship to predicted fitness correlates. In addition, we have summarized the mean and range for these variables across scales and species at used and unused sites. Mean ± 1 SD with only tern nests, so we lumped them with sandbars that had both tern and plover nesting for analysis. b Sandbars that only were used by plovers (n = 42) and plover nesting sites (n = 326). c Sandbars that were used by neither terns nor plovers (n = 588) and randomly selected, unused nest sites (n = 5,457). d Each factor was associated with three broad categories, habitat selection that reduces (a) predation and (b) flooding, (c) or provides access to foraging (plovers only, but terns were analyzed for comparison). For each, "+" indicates a hypothesis that the factor is positively correlated with selection, and "−" indicates negative correlation.
Because of sparse recapture data for terns, we set the agespecific parameters to 0 for that species (i.e., survival and resight did not depend on age for terns). For full details of model specification, see Appendix S1. In addition to estimating nest and chick survival, we derived an overall measure of predicted reproductive success as the product of predicted occupancy (sandbar and nest-site selection), nest success, and chick survival, which was standardized for comparison and can be projected over the habitat surface.

| Model specification
We specified models within R (R Core Team, 2012) using the package "jagsUI" to call JAGS (Plummer, 2003) and export model results back to R. After assessing the performance of a series of exploratory model runs, we ran five chains of 201,000 with an adaptive phase of 1,000 runs and a burn-in period of 1,000 iterations, thinning by 10 for 100,000 samples from the posterior distribution. We determined parameter convergence using the Brooks-Gelman-Rubin criterion (R) (Brooks & Gelman, 1998) and by examining posterior plots, and we considered the model converged if it had an R < 1.1 at each parameter node across the entire nested model. We used effect size, the standard deviation of the posterior, and the proportion of the posterior on one side of 0 ("f") to assess each parameter.

| RE SULTS
We collected information on habitat and bird use at 714 sandbar-

| Sandbar selection
Sandbars varied in their characteristics (Table 1). Our modeling indicated that plover and tern sandbars shared similar habitat characteristics (Figure 3), and their selection agreed with our hypotheses in general, particularly regarding predator and flooding avoidance (Table 2). For predation, both plover and tern sandbars tended to be islands rather than connected sandbars and they had a lower proportion of moderate canopy cover ( Figure 3, Table 2).
Tern sandbars also were in wider sections of the river, had a lower proportion of low canopy cover, and were farther from the nearest cover than unselected sandbars, though they also tended to be closer to the bank than unselected bars, the opposite of plover sandbars ( Figure 3, Table 2). Plover sandbars tended to be farther from the nearest trees than unused bars, but tern sandbars were not different ( Figure 3, Table 2). Both types of occupied sandbars tended to be smaller than unused bars, contrary to our hypothesis for both predator and flood avoidance. Both tern and plover sandbars, however, were composed of less wet sand and tended to have more dry sand than unused bars, in support of our flooding hypothesis, but not in line with our plover foraging hypothesis ( Figure 3, Table 2).

| Nest-site selection
Plovers and terns also used similar habitat characteristics when selecting their nest sites (Table 1). In general, plovers and terns selected habitat in keeping with our hypotheses relative to predator and flooding avoidance and that could optimize foraging, relative to unused sites ( Figure 4, Table 2). For the predation hypothesis, nests of both species were farther from trees and terns were farther from cover than randomly selected sites, but they also tended to be closer to the flood plain than the random locations ( Figure 4, Table 2). In terms of flooding, both species were more likely to nest in dry sand and farther from the waterline than random locations, but both species also were closer to wet sand than would be expected randomly, which agrees with the foraging hypothesis for plovers ( Figure 4, Table 2). Overall, the strength of effect (magnitude of effect size) was greater for terns than for plovers ( Figure 4, Table 2).

| Effect of habitat selection on nest success and chick survival
Sandbar selection had either a negative effect or no appreciable effect on daily nest survival for both species in all years. In particular, plover sandbar selection appeared to have a negative TA B L E 2 Parameter estimates from a Bayesian multi-step regression analysis of habitat selection at the sandbar and nest-site scale for piping plovers and least terns, nesting on the Missouri River (2012-2014) and their relationship to predicted fitness correlates Note: All variables were standardized prior to analysis so that effect sizes could be compared across estimates. For each estimate, we provide mean ± 1 SD from the posterior as well as the f value, or proportion of the posterior on one side of 0 in parentheses. a Sandbars that were used by both terns and plovers (n = 84) and tern nesting sites (n = 424). At the sandbar scale, there were few sandbars (n = 6) with only tern nests, so we lumped them with sandbars that had both tern and plover nesting for analysis. b Sandbars that only were used by plovers (n = 42) and plover nesting sites (n = 326). c Each factor was associated with three broad categories, habitat selection that reduces (a) predation and (b) flooding, (c) or provides access to foraging (plovers only, but terns were analyzed for comparison). For each, "+" indicates a hypothesis that the factor is positively correlated with selection, and "−" indicates negative correlation.

| D ISCUSS I ON
Both least terns and piping plovers selected sandbars and nest sites within those sandbars that could reduce the chances of predation and flooding, and nest sites that would improve foraging opportunities for plover chicks. For chick survival, these decisions led to improved outcomes for both terns and plovers early in the study, but that diminished through time. The pattern for nest survival was less clear, suggesting that tern selection of sandbar and nest sites had less of a positive effect on their reproductive output than plovers.

| Congruence between selection and success
The fact that selection was only partially predictive of success was perhaps not surprising. Both plovers and terns experienced exceptionally high nest and chick survival in the years following the floods Nefas et al., 2018), suggesting that predation and flooding were of little concern or that the densities were low enough that all birds were able to select relatively good habitat in this environment. It is unlikely that habitat was limiting the population F I G U R E 4 Nest-site selection coefficients for nest sites occupied by (a) least terns (Sternula antillarum, goldenrod) and (b) piping plovers (Charadrius melodus, blue) on the Missouri River. The estimates were derived from a species-specific logistic regression, comparing to unused sites. Negative and positive values indicate selection against or for, respectively, a factor relative to unoccupied sites. The dashed line indicates the origin, or no effect immediately after the flooding , and thus, it is possible that there were few if any birds that had to select marginal habitat. Similarly, cavity-nesting species, which tend to have higher nesting success overall (Martin & Li, 1992), generally had the least congruence between selection and success among groups of birds, presumably because the pressure to select better habitat was less at such high success rates.
In their review of congruence in habitat selection studies and success, Chalfoun and Schmidt (2012) found that only 6.4% of studies showed complete congruence, 37.2% showed partial congruence, and 56.4% of studies showed no congruence at all, citing a series of anthropogenic, methodological, and ecological hypotheses for this pattern. Some of these mismatches were hypothesized to be related to anthropogenic changes in habitat that alter the relationship between habitat cues and the resulting fitness. For example, piping plovers nesting on Lake Sakakawea, a managed reservoir on the upper stretches of the Missouri River, chose nesting locations that led to higher than expected flooding of nests (Anteau et al., 2012), presumably because water fluctuations in the reservoir no longer resemble historical and ecological flows or selection for anti-predator behavior is stronger than that against flooding, which may only exert selective pressure occasionally. Nest flooding was rare in the post-flood population , but it can severely affect certain populations in some years . Sandbar height is positively correlated with the height of the water during a flood (Catlin et al., 2010), and the 2011 flood was the highest recorded since the closing of the dams . Thus, it is possible that much of the available nesting habitat in this study was high enough that flooding was not an issue in these years, though plovers and terns would still select habitat to lessen the chances of flooding.
Both plovers and terns are relatively long-lived species where chick survival tends to contribute more to population growth than nest survival (Catlin et al., 2016;Kirsch, 1996). Thus, we might expect to see more concordance between selection and chick survival since it is the more important value for fitness. Indeed, our results did show congruence between selection and chick survival in year 1 of our study, though it tended to decrease through time ( Figure 6).
Piping plovers have remarkably high site fidelity from 1 year to the next Friedrich, Hunt, Catlin, & Fraser, 2015), dispersing <200 m between years on average, such that early selection decisions can have lifelong consequences. However, as piping plover productivity generally declines through time because of vegetation growth or erosion of sandbars without some intervention , we speculate that an increase in discordance between habitat selection and chick survival may be a natural process. Least terns also show moderate to high site fidelity, though less than plovers (Atwood & Massey, 1988;Renken & Smith, 1995), and colony locations were regular but variable sizes during our study (D. Catlin, personal observation). If flooding reduced the number of and cover for mammalian and avian predators on sandbars, but they continued to colonize through time, selection of a presumably good site early with high site fidelity could alter the effect on success. and other factors such as survival of the adults likely also contribute to a bird's selection of nesting habitat, further complicating the detection of fitness benefits (Chalfoun & Schmidt, 2012). Although we did not explicitly look at adult survival relative to habitat selection, avoidance of vegetative cover  and perch trees could contribute to adult survival, particularly raptor perch trees as avian predation appeared to be the greatest observed cause of adult mortality at our site Catlin et al., 2015). For piping plovers, adult survival has the largest effect among the population parameters on population growth (Larson, Ryan, & Murphy, 2002;McGowan, 2005;Plissner & Haig, 2000), which likely is similar for interior least terns given their long life span and limited annual fecundity. Thus, we may expect the congruence between reproductive output and selection to be clouded somewhat by decisions that will optimize adult survival at the cost of lost reproductive output. We know from other studies that plovers will forego breeding in some years when the risk to their own survival may be high, but there is a complex interplay between breeding and survival (Weithman et al., 2017). Failure to detect a relationship between success and selection, therefore, could be a result of unmeasured fitness correlates that obscure relationships in some years.

| Differences between the species
Although there are key differences in their life-history characteristics (e.g., precocial vs. altricial, insectivorous vs. piscivorous), the two species experience similar pressures from predators and flooding.
Thus, it would stand to reason that their selection would be similar with perhaps the exception of proximity of foraging (i.e., wet sand), but even that has some similarities since least tern adults tended to return to the wetted, moist sand with fish for young (D. Catlin, personal observation). Although there have been relatively few studies of nest selection for least terns, they tend to select open, sandy areas, often near inlets on the coast or on mid-channel sandbars on prairie rivers (Burger & Gochfeld, 1990;Kirsch, 1996), which aligns with the more extensive work done to assess habitat suitability for piping plovers Cohen et al., 2009;Prindiville-Gaines & Ryan, 1988). Least terns are colonial (Thompson et al., 1997), although less so in the Great Plains than on the Atlantic Coast (D. Catlin, personal observation), which could have contributed to the relatively stronger signal for selection in terns than plovers (i.e., larger selection coefficients, Table 2).
At the sandbar scale, plover sandbars were farther from the bank, in narrower stretches, and closer to cover than would be expected randomly, and tern sandbars were closer to the bank, in wider sections, but farther from cover and tended to be islands rather than bank-connected bars (Figure 3). Piping plover adults, chicks, and fledglings preferentially forage in backwater channels (Le Fer, Fraser, F I G U R E 6 Standardized predicted reproductive success for least terns (Sternula antillarum; a) and piping plovers (Charadrius melodus; b) in 2012 from an integrated habitat selection model that assessed the effect of sandbar and nest-site selection on both nest success and chick survival. Standardized reproductive success was the product of the probability of a location being occupied, the predicted nest success, and predicted chick survival, which was standardized for comparison. Blue hues indicate lower reproductive success, while red hues indicate higher reproductive success. Black indicates water & Kruse, 2008) similar to the low wave energy bayside mudflats that they prefer on the Atlantic Coast (Cohen et al., 2009), but mid-channel island sandbars far from the bank often have deep channels on either side of them with little or no wetted sand on their peripheries.
This need for wetted habitat could explain the differences we saw relative to the river width and island versus point bar parameters, as bank-connected sandbars would likely have more wet sand than mid-channel islands. For terns, which do not forage on moist sand, avoidance of cover and potential predators likely had a greater effect on their selection than plovers.
At the nest scale, with the exception of magnitude, plovers and terns used similar cues to select habitat (Figure 4), suggesting that they experience similar predation and flooding pressures at that scale. From our hypotheses, it is unclear why least terns nested nearer to wet sand than random and that the effect size was larger than for plovers that regularly display this behavior, presumably to optimize chick growth and survival (Cohen et al., 2009;Walker et al., 2019). It is possible that these locations facilitate feeding of chicks as mentioned above, but further research into these unexpected findings will be needed to move beyond speculation.

| CON CLUS IONS
In general, terns and plovers made similar decisions when selecting sandbars and nest sites, decisions which should contribute to less predation and flooding, and to a degree, better foraging opportunities for plover chicks. Failure to detect congruence between these selective factors and reproductive output is common and found in over half of the studies that attempt to make those links (Chalfoun & Schmidt, 2012). Despite the difficulties in identifying the fitness benefits to selection, continued study of the effect of habitat selection on fitness is needed to understand the complex relationships and improve selection theory.
In this study, we present an integrated habitat selection and fitness model that can be modified further to accommodate a range of species and fitness correlates (e.g., adult survival). Continued application of similar models will help advance our knowledge of selection and its evolutionary underpinning. Ultimately how and why an animal selects its breeding habitat is of paramount importance in ecology and conservation. Thus, enhanced understanding of these factors will contribute to the conservation of imperiled species.

ACK N OWLED G M ENTS
Funding was provided by the U.S. Army Corps of Engineers, the U.S. Fish and Wildlife Service, the National Park Service, and Virginia Tech. We thank cooperators, South Dakota Department of Game, Fish, and Parks, Nebraska Game and Parks Commission, and the Missouri River Institute for support throughout the project. We acknowledge the tireless efforts of our many technicians that collected these data. We thank two anonymous reviewers, the Associate Editor, and the Editor in Chief for helpful comments on this manuscript. This work was conducted under Institutional Animal Care and Use Committee permits 11-027 and 14-003 and U.S. Fish and Wildlife Service Threatened and Endangered Species permit TE103272-3.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data can be found on figshare, https ://doi.org/10.6084/m9.figsh are.9992333. Model code is in Appendix S1.