Synchrony of Piping Plover breeding populations in the U.S. Northern Great Plains

Abstract

Resumen

Introduction

Declines in North American shorebird species, coupled with the potential impacts of climate change and sea level rise, have led to increasing prioritization of research on the status and trends of shorebird populations (Brown et al. 2001, Thomas et al. 2006, Bart et al. 2007, Iwamura et al. 2013, Galbraith et al. 2014). Understanding the population structure and dynamics of migratory species such as shorebirds throughout their annual cycle could aid in identifying threats and implementing conservation actions (Esler 2000). Lower-latitude breeding shorebird species exhibit broad habitat selection, enabling them to utilize a variety of habitat types (Colwell and Oring 1988, Haig et al. 1998, Piersma 2007). In the Great Plains of the U.S., this means that there is variability in both shorebird habitat use (e.g., breeding and migratory staging; Skagen and Knopf 1994, Niemuth et al. 2006, Skagen et al. 2008) and conditions (e.g., wet–dry cycles [Euliss et al. 2004, Johnson et al. 2004], river water management [Anteau et al. 2012a], agriculture and grazing [Higgins et al. 2002], and wetland drainage [Niemuth et al. 2006]). As a result, breeding populations of the same species may experience very different site conditions across a narrow geographic range.

The risk of extinction for a single population is elevated when the dynamics of multiple local populations are synchronized, making the detection of synchrony and its interaction with dispersal a subject of conservation interest (Harrison and Quinn 1989, Matter 2001). Synchronous populations are characterized by spatial and temporal covariance in population size. When individuals comprising separate local populations experience the same environmental conditions in a year (i.e. the ‘Moran effect'; Moran 1953a, 1953b), this may lead to synchrony in population growth rates (e.g., Koenig 2002). Synchrony also may develop among populations of one species due to trophic interactions with another species (Bjørnstad et al. 1999), such as predator–prey or host–parasite relationships (e.g., Jones et al. 2003). Lastly, dispersal between demographically distinct populations can induce synchrony (Bjørnstad et al. 1999, Koenig 1999, Liebhold et al. 2004, Goldwyn and Hastings 2008), particularly when dispersal is density dependent (Schaub et al. 2015), or can decouple previously synchronized populations (Abbott 2011).

Understanding population synchrony and dispersal-induced connectivity can offer insights into the long-term stability and persistence of threatened shorebird populations. If the types of threats faced vary by habitat type and local populations span multiple habitat types, we would expect these populations to fluctuate asynchronously. In turn, if these asynchronous local populations are interconnected, we would expect the overall stability of this population of local populations (or ‘metapopulation') to be greater than the individual stability of any one of the local populations.

The Piping Plover (Charadrius melodus) is a federally Threatened North American shorebird species for which population synchrony might have important implications for population stability. Piping Plovers are ground-nesting, capable of intra- and interannual dispersal (Haig and Oring 1988, Rioux et al. 2011), and breed across a diverse range of riverine, reservoir, and alkaline lake habitats in the Great Plains, USA (U.S. Fish and Wildlife Service [USFWS] 1985, 2009). The species faces different anthropogenic threats depending upon the habitat in which it nests. For example, water management can cause nesting habitat loss and reproductive failure at rivers and reservoirs (Anteau et al. 2012a), while in alkaline lake habitat, nest loss and disturbance may occur via livestock, agricultural practices, and periodic climate fluctuations (Prindiville Gaines and Ryan 1988, Anteau 2012, McCauley et al. 2015). In light of these differing anthropogenic threats, we would expect local populations nesting in habitats connected through water management, such as a river and its reservoirs, to exhibit greater synchrony with each other than with those nesting in unconnected habitats such as alkaline lakes. We examined the degree to which local breeding populations of Piping Plovers in the Great Plains exhibited population synchrony and stability. We also identified environmental factors associated with changes in the size of local breeding populations.

Methods

Study Area

Nine local populations of Piping Plovers were distributed across alkaline lake, reservoir, and riverine nesting habitat stretching from central South Dakota through North Dakota into northeastern Montana, USA (Figure 1). These populations comprised the ‘U.S. Alkali Lakes' (i.e. the alkaline lake populations) and the ‘Northern Rivers' (i.e. the reservoir and river populations) portions of the Great Plains Piping Plover metapopulation as defined by McGowan et al. (2014). The alkaline lake habitat consisted of ∼150 basin (i.e. lake, pond, or slough) shorelines located in the Missouri Coteau region of North Dakota and Montana (Knetter et al. 2002). The Missouri Coteau is characterized by alternating hummocks and depressions and, when managed for human use, consists primarily of annual planted cropland or rangeland (Knetter et al. 2002). The reservoir shoreline habitat was irregular, dissected, and composed of a diversity of substrate types (Anteau et al. 2012b). For the purposes of these analyses, this habitat consisted primarily of mainland and island shorelines along Lake Oahe (OAH) and Lake Sakakawea (SAK), 2 main stem reservoirs of the Missouri River. The Lake Oahe reservoir extended from the headwaters of Lake Oahe, ∼10 km south of Bismarck, North Dakota, to just north of Pierre, South Dakota. The Lake Sakakawea reservoir (from Garrison Dam near Riverdale, North Dakota, to White Tail Bay, North Dakota; see Anteau et al. 2014a, 2014b) was located ∼80 km from Bismarck, North Dakota. The riverine habitat consisted of the Missouri River's Garrison Reach (GRR), which extended from the Garrison Dam to the headwaters of Lake Oahe. Nesting habitat primarily consisted of mid-channel low- to mid-elevation sandbars with some established woody vegetation (Sherfy et al. 2009).

Figure 1.

Local Piping Plover breeding populations distributed throughout the Great Plains of Montana, North Dakota, and South Dakota, USA. Subscripts indicate the habitat type of each population. Each local breeding population was assigned to a region composed of multiple sites: AUD = Audubon National Wildlife Refuge, managed as part of the Audubon Wetland Management District (n_sites = 43); CROS = Crosby Wetland Management Unit (n_sites = 17); LONG = Long Lake National Wildlife Refuge, managed as part of the Long Lake Wetland Management District (n_sites = 36); LOST = Lostwood National Wildlife Refuge, managed as part of the Lostwood Wetland Management District (n_sites = 35); MED = Medicine Lake National Wildlife Refuge, managed as part of the Montana Wetland Management District (n_sites = 44); TNC = The Nature Conservancy's John E. Williams Preserve (n_sites = 12); OAH = the Lake Oahe region of the Missouri River (n_sites = 1084); GRR = the Garrison River reach region of the Missouri River (n_sites = 38); and SAK = the Lake Sakakawea region of the Missouri River (n_sites = 548).

Open in new tabDownload slide

Count Data

Alkaline lakes

The U.S. Fish and Wildlife Service (USFWS) has counted adult Piping Plovers located in alkaline lake habitat in North Dakota and Montana annually since the late 1980s. Six USFWS regions are responsible for monitoring and surveying for Piping Plovers at ∼150 basins (hereafter, sites) historically used by Piping Plovers for nesting: (1) Audubon National Wildlife Refuge, managed as part of the Audubon Wetland Management District (AUD; n_sites = 43); (2) Crosby Wetland Management Unit (CROS; n_sites = 17); (3) Long Lake National Wildlife Refuge, managed as part of the Long Lake Wetland Management District (LONG; n_sites = 36); (4) Lostwood National Wildlife Refuge, managed as part of the Lostwood Wetland Management District (LOST; n_sites = 35); (5) Medicine Lake National Wildlife Refuge, managed as part of the Montana Wetland Management District (MED; n_sites = 44); and (6) The Nature Conservancy's John E. Williams Preserve (TNC; n_sites = 12). We considered each one of these 6 USFWS regions (and the sites comprising them) a local Piping Plover breeding population.

During a 2-week period in mid-June, from 1993 to 2011, USFWS personnel surveyed ∼150 sites for adult Piping Plovers. Surveyors walked or boated past available nesting habitat and recorded the number of adult Piping Plovers observed at each site. Each site included in this annual June census (hereafter, June Census) was surveyed once. Mid-June is a period of the breeding season when plovers typically have either nests or chicks, thus the adult Piping Plovers observed are assumed to be breeding individuals.

For the purposes of our analyses, we considered each individual basin surveyed a ‘site,' and each of the 6 USFWS regions referenced above a ‘local breeding population.' For each of the 6 USFWS regions, we totaled the number of adult Piping Plovers counted at each site in each year (1993–2011; but see ‘Missing Data' section). We considered these counts an index of breeding population size for each of the 6 local breeding populations located in alkaline lake nesting habitat.

Reservoir and river

Following the 1985 federal listing of the Piping Plover (USFWS 1985), the U.S. Army Corps of Engineers (USACE) created the Tern and Plover Monitoring Program (TPMP), which was tasked with monitoring the terns and plovers nesting along the rivers and around the reservoirs of the Missouri River system (USACE 1993). Throughout the breeding season (approximately mid-May through mid-August), crews surveyed available and historically used shoreline and sandbar habitat for Piping Plover nests along both reservoir and riverine sections of the Missouri River.

For the purposes of our analyses, we considered a consecutive stretch of 4 river miles of reservoir or riverine habitat a ‘site.' We considered the Lake Oahe (OAH; n_sites = 1,084), Garrison River reach (GRR; n_sites = 38), and Lake Sakakawea (SAK; n_sites = 548) regions of the Missouri River to represent 3 local Piping Plover breeding populations. For each year (1993–2011), we totaled the number of nests found across sites on Lake Oahe, the Missouri River's Garrison reach, and Lake Sakakawea, respectively. We considered these counts to be annual indices of breeding population size for the 3 local breeding populations, 2 of which were located in reservoir habitat (OAH, SAK) and 1 in riverine habitat (GRR).

Missing Data

USACE crews attempted to search all available riverine and reservoir habitat in each year. To estimate correlation coefficients and coefficients of variation, we generated cumulative annual counts for the Missouri River regions by summing the total number of nests counted each year across all river miles in each of the 3 regions. However, we could not take this approach with the adult census data collected by the USFWS in alkaline lake habitat. Although the annual June Census was conducted in all years from 1993 to 2011, censuses were not conducted at all known sites (i.e. basins) in all years. Values were more likely to be missing during the early years of the June Census. This resulted in the following percentage of missing values in our dataset (percentage of missing values out of total values [n] is presented by USFWS alkaline lakes region): AUD_{n= 817} = 42%; CROS_{n= 323} = 22%; LONG_{n= 684} = 72%; LOST_{n= 665} = 28%; MED_{n= 1,026} = 41%; and TNC_{n= 228} = 28%. In an average year, this meant that ∼42% (± 11% SD) of the possible values were missing, with the following breakdown by USFWS region: AUD = 42% (± 26% SD), CROS = 22% (± 13% SD), LONG = 72% (± 30% SD), LOST = 28% (± 13% SD), MED = 42% (± 11% SD), and TNC = 28% (± 6% SD).

Using these counts, we built zero-inflated negative binomial models for each USFWS region, treating year and site as fixed effects, thereby assuming that year and site did not interact within a region. We then used these models to generate predicted counts for all year and site combinations, which generated ‘counts' for sites at which no data were collected in a given year. We assessed the predictive ability of our models by regressing the predicted counts generated by the models against the actual counts for each region. The associations were significant in all cases, with the following correlations between predicted and actual counts for each USFWS region (n = 249): r_AUD = 0.81; r_CROS = 0.76; r_LONG = 0.59; r_LOST = 0.78; r_MED = 0.79; and r_TNC = 0.89. Due to the extensive number of missing values in the LONG USFWS alkaline lakes region and the low predictive ability of our zero-inflated negative binomial model for this region, we excluded this USFWS region from our correlation analyses. Finally, for all USFWS alkaline lakes regions excluding LONG, we used the predicted counts from our models as the ‘counts' for each site–year combination and summed these predicted counts across sites for each year to generate annual cumulative counts for each USFWS region (Gelman and Hill 2006).

Cross-Correlation Coefficients: Population Synchrony

We assessed the synchrony in total abundance among populations by calculating the Pearson moment correlation for log-transformed annual cumulative count data (Bjørnstad et al. 1999). We computed Pearson correlation statistics for pairwise combinations of count data to assess the degree to which the 9 Piping Plover local breeding populations could be considered to have been behaving ‘synchronously' during 1993–2011. Cross-correlations between successive between-year differences in log-transformed counts place emphasis on population growth rates and thus synchrony in change rather than in total abundance (Bjørnstad et al. 1999). Thus, we subtracted log-transformed counts in year t from log-transformed counts in year t + 1 and used the resulting time series to calculate Pearson moment correlations (Bjørnstad et al. 1999).

To determine whether our measures of population synchrony could be explained by the distances between local populations, we regressed the Pearson moment correlation coefficients for the log-transformed annual cumulative count data and cross-correlations of annual cumulative count data against pairwise distances between local populations (Ranta et al. 1995). Pairwise distances between populations were calculated between the breeding population centroids. We considered relationships significant if P < 0.05.

Coefficients of Variance: Population Stability

Increased temporal variability in population size estimates is associated with an elevated risk of extinction for populations (Inchausti and Halley 2003, Legendre et al. 2008). We considered the degree to which the abundances of local populations changed over time to be an indicator of population stability. We used the coefficient of variation for counts of either plover nests or plover adults for each local population to represent the change in local population abundance. The coefficient of variation is defined as the ratio of the standard deviation to the mean and is a unitless measure of variability, thus allowing comparisons between different units of measurement (e.g., counts of plover nests and plover adults).

Zero-Inflated Negative-Binomial Regressions: Environmental Factors

We identified several environmental factors that we believed could influence Piping Plover breeding population size, and represented each factor with measurable site- or habitat-specific covariates (Table 1). For each environmental factor, we present a prediction of its influence on population size as well as a proposed rationale for the associated prediction (Table 2). We also present descriptions and summary statistics for the covariates used in the analyses (Table 1). To ensure model convergence, all covariates were standardized to a mean of 0 and a standard deviation of 1, with the exception of the covariates ‘Within' and ‘Within.prev' (see Table 1). We restricted the analysis to sites for which all covariate values were available in all years (AUD: n_sites = 36; CROS: n_sites = 16; LONG: n_sites = 25; LOST: n_sites = 29; MED: n_sites = 35; TNC: n_sites = 11; OAH: n_sites = 1,083; SAK: n_sites = 545; and GRR: n_sites = 37).

Table 1.

Open in new tab

Descriptions, means, and standard deviations of covariates used in zero-inflated Poisson regressions to explore relationships between environmental factors believed to influence Piping Plover breeding population size by habitat type from 1993 to 2011.

Table 1.

Open in new tab

Descriptions, means, and standard deviations of covariates used in zero-inflated Poisson regressions to explore relationships between environmental factors believed to influence Piping Plover breeding population size by habitat type from 1993 to 2011.

Table 2.

Open in new tab

Environmental factors potentially related to Piping Plover breeding population size in the alkaline lake, reservoir, and riverine habitats of the U.S. Northern Great Plains, 1993–2011. Shown are the region that the environmental factor was believed to influence, a prediction for how it would influence Piping Plover breeding population size in this region, and a proposed rationale for the relationship. Covariates used to represent these environmental factors in zero-inflated Poisson regressions are indicated in parentheses and further described in Table 1.

Table 2.

Open in new tab

Environmental factors potentially related to Piping Plover breeding population size in the alkaline lake, reservoir, and riverine habitats of the U.S. Northern Great Plains, 1993–2011. Shown are the region that the environmental factor was believed to influence, a prediction for how it would influence Piping Plover breeding population size in this region, and a proposed rationale for the relationship. Covariates used to represent these environmental factors in zero-inflated Poisson regressions are indicated in parentheses and further described in Table 1.

Failure to account for ‘zero inflation' in count data characterized by a high proportion of zero values can cause biased parameter estimates (Martin et al. 2005). The zero-inflated negative binomial (ZINB) is a distribution commonly chosen when modeling ecological count data (e.g., Zipkin et al. 2014). We built hierarchical ZINB regression models to explore relationships between environmental factors believed to influence Piping Plover breeding population size (Table 2). We constructed the ZINB distribution as a zero-inflated nonhomogeneous Poisson distribution in which the rate parameter followed a gamma distribution (Hilbe 2011). We built and ran ZINB models for each of the 3 habitat types (alkaline lake, reservoir, river) in a Bayesian framework using program Jags and package ‘R2Jags' in program R (Plummer 2003, R Core Team 2014, Su and Masanao 2014). In each case, the response variable (C) was a year-by-site (i) index of population size:

$$

$$

$$

where each basin was considered a ‘site' for the alkaline lake habitat type and each consecutive 4-river-mile stretch was considered a ‘site' for the reservoir and riverine habitat types.

For each habitat type, we built a ZINB regression model consisting of all environmental factors by adapting code from Kéry (2010:184). We did not add covariates to the zero-inflated logistic portion (ϕ_i) of the regression. However, the λ_i component of the negative binomial portion of the model consisted of 7 fixed-effect regression coefficients and two random effects. We used the Kuo and Mallick (1998) method to assess variable importance (O'Hara and Sillanpää 2009), which required that we add an indicator variable (I) consisting of a prior drawn from a Bernoulli distribution to each fixed effect (j). We modeled site-by-year specific counts as follows:

$$

$$

$$

$$

$$

where β_int, β_Avail, β_Ppt, β_Ppt.prev, β_Within, and β_Within.prev are the fixed-effects regression coefficients estimated based on the covariate values of x_Avail, x_Ppt, x_Ppt.prev, x_Within, and x_Within.prev, respectively, for each site-by-year observation (i). Following Kéry and Schaub (2012), we represent the extra variation associated with each site,⁠, and year,⁠, for each year-by-site observation (i) as:

$$

and

$$

We used priors that were intended to be uninformative for all parameters, and ran 3 Markov chain Monte Carlo (MCMC) chains for 5,000 iterations; we discarded 1,000 iterations as burn-in and then ‘thinned' every 4 iterations. We visually monitored chains for convergence and used the R̂ statistic for confirmation (Gelman and Hill 2006), judging that convergence was acceptable when the R̂ for each parameter was ≤ 1.1. We assessed model fit using posterior predictive checks and report posterior medians and 95% credible (confidence) intervals. We assessed variable importance using the indicator variables (I; Kuo and Mallick 1998). We report values for all regression coefficients and binomial indicators, but only interpret regression coefficients with indicator values of I ≥ 0.75 (Mutshinda et al. 2013).

Results

Population Stability

We interpreted population-specific estimates of the coefficient of variation (CV) as measures of population stability, with greater stability reflected by lower values. Estimates of CV were lower for the counts of Piping Plover adults collected by USFWS surveyors during the June Census in the alkaline lakes habitat (CV = 55%, n = 13) than for the Piping Plover nest counts collected by USACE crews in riverine (CV = 63%, n = 13) and reservoir habitat (CV = 83%, n = 13). The values for the region-specific coefficients of variation were: CV_AUD = 37%; CV_CROS = 23%; CV_LOST = 32%; CV_MED = 27%; CV_TNC = 36%; CV_OAH = 82%; CV_SAK = 84%; and CV_GRR = 63%.

Population Synchrony

Log-transformed counts of Piping Plover nests found in riverine habitat were positively correlated with log-transformed counts of Piping Plover nests found in reservoir habitat (r_{GRR, SAK} = 0.60, P = 0.007, n = 19; r_GRR, OAH = 0.57, P = 0.01, n = 19). The counts at the 2 reservoirs also were positively correlated with each other (r_{OAH, SAK} = 0.88, P < 0.001, n = 19). Log-transformed counts of adult Piping Plovers observed at alkaline lakes during the June Census either showed no relationship with each other or were positively correlated (e.g., r_{AUD, CROS} = 0.46, P = 0.05, n = 19; r_{AUD, TNC} = 0.72, P < 0.001, n = 19). However, supported correlations with log-transformed counts of Piping Plover nests found in reservoir habitat were all negative (r_{AUD, SAK} = −0.52, P = 0.02, n = 19; r_{LOST, SAK} = −0.56, P = 0.01, n = 19; r_{LOST, OAH} = −0.51, P = 0.03, n = 19; r_{SAK, TNC} = −0.61, P = 0.006, n = 19; r_{OAH, TNC} = −0.49, P = 0.03, n = 19). There was no correlation between the log-transformed counts of adult Piping Plovers at alkaline lakes and log-transformed counts of Piping Plover nests found in riverine habitat (Table 3A).

Table 3.

Open in new tab

Pearson correlation coefficients for (A) log-transformed counts and (B) cross-correlations of log-transformed counts of Piping Plover nests (GRR, SAK, OAH) and adults (AUD, CROS, LOST, MED, TNC) in the Northern Great Plains, USA, 1993–2011 (n = 19). AUD = Audubon National Wildlife Refuge, managed as part of the Audubon Wetland Management District; CROS = Crosby Wetland Management Unit; LONG = Long Lake National Wildlife Refuge, managed as part of the Long Lake Wetland Management District; LOST = Lostwood National Wildlife Refuge, managed as part of the Lostwood Wetland Management District; MED = Medicine Lake National Wildlife Refuge, managed as part of the Montana Wetland Management District; TNC = The Nature Conservancy's John E. Williams Preserve; OAH = the Lake Oahe region of the Missouri River; GRR = the Garrison River reach region of the Missouri River; and SAK = the Lake Sakakawea region of the Missouri River. Bolded values reflect correlation coefficient values with P ≤ 0.05; P-values are reported in parentheses beneath each correlation coefficient. Cross-correlations were calculated by subtracting log-transformed counts in year t from log-transformed counts in year t – 1.

Table 3.

Open in new tab

Pearson correlation coefficients for (A) log-transformed counts and (B) cross-correlations of log-transformed counts of Piping Plover nests (GRR, SAK, OAH) and adults (AUD, CROS, LOST, MED, TNC) in the Northern Great Plains, USA, 1993–2011 (n = 19). AUD = Audubon National Wildlife Refuge, managed as part of the Audubon Wetland Management District; CROS = Crosby Wetland Management Unit; LONG = Long Lake National Wildlife Refuge, managed as part of the Long Lake Wetland Management District; LOST = Lostwood National Wildlife Refuge, managed as part of the Lostwood Wetland Management District; MED = Medicine Lake National Wildlife Refuge, managed as part of the Montana Wetland Management District; TNC = The Nature Conservancy's John E. Williams Preserve; OAH = the Lake Oahe region of the Missouri River; GRR = the Garrison River reach region of the Missouri River; and SAK = the Lake Sakakawea region of the Missouri River. Bolded values reflect correlation coefficient values with P ≤ 0.05; P-values are reported in parentheses beneath each correlation coefficient. Cross-correlations were calculated by subtracting log-transformed counts in year t from log-transformed counts in year t – 1.

Between-year differences in log-transformed counts of Piping Plover nests found in riverine habitat were positively correlated with the Lake Sakakawea (i.e. reservoir) Piping Plover nest counts (r_{GRR, SAK} = 0.63, P = 0.005, n = 19). However, between-year differences in log-transformed counts of Piping Plover nests showed no relationship between the Lake Oahe and Lake Sakakawea reservoir habitats. Between-year differences in log-transformed counts of adult Piping Plovers at the alkaline lakes mostly showed no relationship with each other. There were significant positive correlations in only 3 instances (r_{AUD, CROS} = 0.54, P = 0.02, n = 19; r_{AUD, LOST} = 0.63, P = 0.005, n = 19; r_{LOST, MED} = 0.61, P = 0.007, n = 19). Between-year differences in log-transformed counts of adult Piping Plovers at alkaline lakes showed no correlations with the between-year differences in log-transformed counts of Piping Plover nests in reservoir habitat (Table 3B).

We found no significant relationships between the Pearson moment correlation coefficients of log-transformed cumulative count data and the pairwise population distances (Figure 2).

Figure 2.

Pearson correlation coefficients for (A) log-transformed counts and (B) cross-correlations of log-transformed counts of Piping Plover nests (GRR, SAK, OAH; see Figure 1 for abbreviation definitions) and adults (AUD, CROS, LOST, MED, TNC) in the Northern Great Plains, USA, 1993–2011, plotted according to distance (km) between local populations. Distances between alkaline lake populations are represented by open circles, reservoir populations by gray squares, alkaline lake and river populations by open triangles, alkaline lake and reservoir populations by black circles, and reservoir and river populations by black triangles. Cross-correlations were calculated by subtracting log-transformed counts in year t from log-transformed counts in year t + 1.

Open in new tabDownload slide

Environmental Influences on Population Synchrony

Site-specific annual counts of adult Piping Plovers at alkaline lakes were associated with changes in standardized precipitation evapotranspiration indices (SPEI; ‘Avail' covariate). In general, counts of Piping Plover adults were higher in years with hotter and drier conditions (i.e. more negative SPEI values; Table 4). Site-specific annual counts of Piping Plover nests in riverine habitat were associated with the between-year change in river flow; between-year increases in river flow were associated with decreased nest counts (Table 4). None of the covariates included in our model were strongly associated with changes in site-specific annual counts of Piping Plover nests in reservoir habitat (Table 4).

Table 4.

Open in new tab

Mean regression coefficient estimates with 95% credible intervals for reduced zero-inflated negative binomial regressions of alkaline lake habitat adult Piping Plover counts (n_sites = 152), reservoir habitat nest counts (n_sites = 1,628), and riverine habitat nest counts (n_sites = 37). Regression coefficient estimates (β̂) are subscripted with the covariates of the effect that they estimated; see Table 1 for a description of covariates used to represent environmental factors and Table 2 for predictions and explanations for relationships between each environmental factor and breeding population size. For each covariate we also report the value of the binomial indicator (I) used in model selection; bolded values highlight covariates with I ≥ 0.75 (i.e. a high probability of inclusion in the model).

Table 4.

Open in new tab

Mean regression coefficient estimates with 95% credible intervals for reduced zero-inflated negative binomial regressions of alkaline lake habitat adult Piping Plover counts (n_sites = 152), reservoir habitat nest counts (n_sites = 1,628), and riverine habitat nest counts (n_sites = 37). Regression coefficient estimates (β̂) are subscripted with the covariates of the effect that they estimated; see Table 1 for a description of covariates used to represent environmental factors and Table 2 for predictions and explanations for relationships between each environmental factor and breeding population size. For each covariate we also report the value of the binomial indicator (I) used in model selection; bolded values highlight covariates with I ≥ 0.75 (i.e. a high probability of inclusion in the model).

Discussion

Heterogeneity in the spatial distribution and habitat over which local populations are distributed is considered critical to population stability (Scheuring 2000, Matter 2001, Briggs and Hoopes 2004). We suspected that the proclivity of low-latitude breeding shorebird species such as Piping Plovers to utilize a diversity of habitat types while nesting (Piersma 2007, Anteau et al. 2012a, 2012b) could lead to a decrease in the probability of synchrony among breeding populations and thus contribute to population stability. In this study, we closely examined the dynamics of 2 portions of the Great Plains Piping Plover metapopulation; specifically, the components referred to by McGowan et al. (2014) as the ‘U.S. Alkali Lakes' and the ‘Northern Rivers.' Our results suggested that, during 1993–2011, the U.S. Alkali Lakes component (i.e. local populations nesting in alkaline lake habitat) and the Northern Rivers component (i.e. local populations nesting in reservoir and riverine habitat) of the Great Plains Piping Plover metapopulation did not share a common environmental driver. We believe that these results support the original decision by McGowan et al. (2014) to treat the U.S. Alkali Lakes and Northern Rivers local populations as separate components of the Great Plains Piping Plover metapopulation.

Overall, there was very little synchrony detected, either among the populations nesting in alkaline lake habitat, or between the populations nesting at alkaline lakes and populations nesting in reservoir or riverine habitat. This suggests that the dynamics of the local breeding populations in these areas were likely driven by different environmental factors. Populations nesting in alkaline lake habitat were more stable during 1993–2011 than populations nesting in riverine and reservoir habitats (as measured by the coefficient of variation). Unlike the riverine and reservoir habitat types which, although large, are connected through water management, the basins comprising the alkaline lakes are generally spatially separate. Although nesting habitat faces similar pressures across the alkaline lakes (e.g., agriculture, drainage, and grazing), not all basins experience the same conditions at the same time. As a result, habitat change that has occurred across alkaline lake habitat has, to date, been more gradual and less all-encompassing than that on the Missouri River. We suspect that the potential for alkaline lake basins to experience independent habitat conditions has contributed to the lack of synchrony and overall stability of Piping Plover populations nesting in alkaline lake habitat.

In contrast to populations nesting in alkaline lake habitat, populations of Piping Plovers nesting in riverine and reservoir habitats (i.e. the habitat comprising the Northern Rivers) exist in a highly managed system. The historical nesting habitat and foraging base of the Missouri River have been altered (Catlin et al. 2010, Poff and Zimmerman 2010, Anteau et al. 2012b), and changes in reservoir water levels can alter the amount of available reservoir and riverine nesting habitat annually (Anteau et al. 2014b). The geographic distribution of the populations located in reservoir and riverine habitats is such that actions undertaken upstream (e.g., the release of water out of the Garrison Dam) have synchronous repercussions for the riverine and reservoir habitats downstream. While the populations nesting in riverine and reservoir habitat demonstrated annually correlated total abundance (i.e. population synchrony), only the populations nesting in Lake Sakakawea reservoir habitat appeared to undergo population growth in the same years as the populations nesting in the Garrison Reach riverine habitat. The latter result is of particular interest as it suggests that the correlations between the populations located in reservoir and riverine habitats could have been induced by correlated demographic rates (e.g., recruitment rates into the breeding populations) in addition to shared environmental circumstances (Bjørnstad et al. 1999).

There was both within- and between-year dispersal among the local populations at the alkaline lakes, reservoirs, and river (Roche et al. 2015). Changing nesting habitat conditions, such as reservoir water level rise, have been shown to induce the dispersal of Piping Plover breeding adults (Roche et al. 2012), although some pairs will renest locally following flooding (Claassen et al. 2014). Thus, we originally suspected that we might see a relationship between annual reservoir water level change and the abundance of Piping Plovers nesting at the alkaline lakes. We found that years of high abundance for populations nesting in reservoir habitat were years of low abundance for at least 3 of the populations nesting in alkaline lake habitat; specifically, in the USFWS alkaline lakes regions closest to the Lake Sakakawea reservoir habitat. This is suggestive of annual movement of Piping Plovers between the Northern Rivers and U.S. Alkali Lakes. However, we found no evidence for correlated growth rates between local populations nesting in alkaline lake habitat and populations nesting in reservoir habitat.

McGowan et al. (2014) concluded that the Great Plains Piping Plover metapopulation had a 3% chance of extinction within the next 50 yr. The authors noted that as dispersal between the components of the metapopulation increased, decreased metapopulation stability and consequently an increased risk of extinction resulted (McGowan et al. 2014). Depending on the degree of synchrony, dispersal can either increase the stability of a metapopulation by ‘rescuing' local populations as they trend toward extinction or increase the degree of connectivity among populations such that they fluctuate in complete synchrony (Matter 2001, Abott 2011); the latter is not conducive to the long-term stability and persistence of metapopulations (Hanski 1998). Our results suggest that the asynchronous relationship between populations nesting in alkaline lake habitat vs. riverine and reservoir habitats means that dispersal between the U.S. Alkali Lakes and Northern Rivers populations could result in greater stability and persistence for the overall Great Plains Piping Plover metapopulation.

At the time of the McGowan et al. (2014) study, there were limited available demographic data for the Northern Rivers and U.S. Alkali Lakes components of the Great Plains Piping Plover metapopulation. Thus, each of the 4 components of the metapopulation (i.e. Prairie Canada, U.S. Alkali Lakes, Northern Rivers, and Southern Rivers) was parameterized to respond to dispersal in the same way, largely informed by demographic data from the Southern Rivers component. Our results suggest that (1) it is unlikely that dispersal will affect all components of the metapopulation in the same way, and (2) the model may be improved by reparameterization to reflect the dynamics of dispersal between riverine and reservoir habitats. We found that the populations in the riverine and reservoir habitats of the Northern Rivers were highly synchronous; this synchrony could serve as a destabilizing force if both habitats experience the same conditions in a single year. During the Missouri River flood of 2011, exactly this situation unfolded; high water levels on both the river and reservoirs resulted in essentially zero productivity in these areas. Haig et al. (1998) stated that during years of major flooding on the Missouri River, wetlands in the Missouri Coteau (i.e. alkaline lake habitat) were used by Piping Plover adults that would normally have bred in riverine and reservoir habitats. Mark–recapture data collected after the 2011 flood support this observation (Roche et al. 2015). In light of the stabilizing influence of dispersal under asynchronous population dynamics, our results suggest that the ability to disperse to available alkaline lake nesting habitat could serve as a stabilizing influence on the Great Plains Piping Plover metapopulation.

Acknowledgments

We are indebted to M. Borgreen, D. Borgreen, K. Brennan, N. Kadrmas, M. Rabenberg, L. Richardson, E. Rosenquist, and numerous USFWS biologists and managers who preceded these individuals for providing us access to and interpretation of 19 yr of Piping Plover June Census data for the alkaline lakes. R. K. Murphy was instrumental in developing original field and data protocols for the monitoring of alkaline lakes. We thank A. Albrecht, C. Aron, B. Hill, C. Huber, K. Kreil, C. Kruse, G. Pavelka, M. Ring, and D. Toy for help with project planning, logistics, analysis, or access to and interpretation of nesting data for the Missouri River. We also thank many USACE, TNC, North Dakota Game and Fish, South Dakota Game and Fish, Montana Fish, Parks and Wildlife, and USFWS field technicians and crew leaders who collected the data. C. Amundson, K. Brennan, C. Brown, and P. Doherty provided helpful comments on this manuscript. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Funding statement: This study was funded by the Lostwood Complex of the U.S. Fish and Wildlife Service and the U.S. Army Corps of Engineers' Missouri River Recovery Program through financial and logistical support from the Corps' Omaha District Threatened and Endangered Species Section and Garrison Project Office. None of the funders had any input into the content of the manuscript, nor required approval of the manuscript prior to submission or publication.

Ethics statement: Our field protocols were approved by the U.S. Geological Survey Northern Prairie Wildlife Research Center Animal Care and Use Committee.

Author contributions: E.A.R. and T.L.S. analyzed the data and wrote the paper. M.J.A., C.M.D., M.H.S., and M.T.W. formulated the question, provided key insights for analytical directions, and substantially edited this manuscript.

Literature Cited