The role of landscape features and density dependence in growth and fledging rates of Piping Plovers in North Dakota, USA

Abstract

For species with precocial young, survival from hatching to fledging is a key factor influencing recruitment. Furthermore, growth rates of precocial chicks are an indicator of forage quality and habitat suitability of brood-rearing areas. We examined how growth and fledging rates of Piping Plover (Charadrius melodus) chicks were influenced by landscape features, such as hatchling density (hatchlings per hectare of remotely sensed habitat [H ha⁻¹]), island vs. mainland, and wind fetch (exposure to waves) at 2-km segments (n = 15) of Lake Sakakawea, North Dakota, during 2007–2008. Hatchling growth was comparable with published estimates for other habitats. Models for fledging rate (fledged young per segment) assuming density dependence had more support (w_i = 96%) than those assuming density independence (w_i = 4%). Density-dependent processes appeared to influence fledging rate only at densities >5 H ha⁻¹, which occurred in 19% of the segments we sampled. When areas with densities >5 H ha⁻¹ were excluded, density-dependence and density-independence models were equally supported (w_i = 52% and 48%, respectively). Fledging rate declined as the wind fetch of a segment increased. Fledging rate on mainland shorelines was 4.3 times greater than that on islands. Previous work has indicated that plovers prefer islands for nesting, but our results suggest that this preference is not optimal and could lead to an ecological trap for chicks. While other researchers have found nesting-habitat requirements to be gravelly areas on exposed beaches without fine-grain substrates, our results suggest that chicks fledge at lower rates in these habitats. Thus, breeding plovers likely require complexes of these nesting habitats along with protected areas with fine, nutrient-rich substrate for foraging by hatchlings.

Resumen

Para especies con pichones precociales, la supervivencia desde la eclosión hasta el abandono del nido, es un factor clave para el reclutamiento. Además, las tasas de crecimiento de pichones precociales son indicadoras de la calidad del alimento y la idoneidad del hábitat en áreas de anidación. Examinamos como el crecimiento y las tasas de abandono de nido de pichones de Charadrius melodus, estuvieron influenciados por características del paisaje, tales como la densidad de eclosión (pichones que abandonaron el nido por ha, en hábitats teledetectados remotamente [H ha⁻¹], isla vs. tierra firme, y exposición a las olas en segmentos de 2-km (n = 15) en el Lago Sakakawea, Dakota del Norte, durante 2007–2008. El tamaño de pichones recién eclosionados es comparable con estimaciones publicadas para otros hábitats. Los modelos para la tasa de abandono de nido (volantones que abandonaron el nido por segmento) asumiendo denso-dependencia tuvieron mayor respaldo (w_i = 96%) que aquellos que asumían denso-independencia (w_i = 4%). Los procesos denso-dependientes parecen influenciar las tasas de abandono del nido únicamente a densidades de >5 H ha⁻¹, lo que ocurrió en el 19% de los segmentos que muestreamos. Cuando las áreas con densidades >5 H ha⁻¹ fueron excluidas, los modelos denso-dependientes y denso-independientes fueron respaldados de igual manera (w_i = 52% y 48%, respectivamente). Al aumentar la exposición a las olas en un segmento, disminuyó la tasa de abandono del nido. Investigaciones previas han indicado que los chorlitos prefieren islas para nidificar, pero nuestros resultados sugieren que esta preferencia no es la óptima, y podría llevar a trampas ecológicas para los pichones. Mientras otros investigadores han encontrado que playas expuestas con áreas rocosas sin substratos de granos finos, son los hábitats requeridos para anidar, nuestros resultados sugieren que las tasas de abandono de nido por pichones disminuye en estos hábitats. Por lo tanto, los chorlitos reproductores probablemente requieren complejos de estos hábitats para nidificar, junto con áreas protegidas con sustratos finos, ricos en nutrientes para el forrajeo de los volantones.

Palabras clave: condición corporal, supervivencia de pichones, demografía, Río Missouri, reclutamiento, correlimo, ave acuática

Introduction

For species with precocial young, survival from hatching to fledging is a key factor influencing recruitment (Wisdom and Mills 1997, Hoekman et al. 2002, 2006, Coluccy et al. 2008). Prior to fledging, precocial young have relatively limited mobility compared with adults; thus, their survival is influenced by the landscape that was selected by adults during nesting (Rotella and Ratti 1992a, Krapu et al. 2000). For many bird species, factors related to nest-site selection or increased nest success may be different from those related to fledging success. Accordingly, understanding how habitat influences fledging success, in addition to nest-site selection and nest success, is an important step toward species conservation. A collective understanding of these phenomena may reveal a need for the conservation of complexes of nesting and brood-rearing habitat to support adequate recruitment (Dwyer et al. 1979, Stephens et al. 2005), or may identify ecological traps (Schlaepfer et al. 2002, Anteau et al. 2012a). Furthermore, identifying landscape-scale habitat features that influence fledging success may improve demographic models through spatially explicit estimation of recruitment rates.

Growth rates of hatchlings can be indicators of forage quality and availability and general habitat suitability of brood-rearing areas (Lepage et al. 1998, Audet et al. 2007, Fondell et al. 2011). Growth rates can also be directly related to fledging success by influencing the duration of the flightless period, when young are most vulnerable to predation (Mauser et al. 1994, Pearse and Ratti 2004). In addition, growth rates relate to fledging success because hatchlings requiring more foraging time to acquire resources, or those foraging in areas without accessible cover, may be more vulnerable to predation (Mace et al. 1983, Flint et al. 2006, Mainguy et al. 2006, Miller 2010).

Habitat availability clearly is an important factor influencing fledging success (Rotella and Ratti 1992b, Krapu et al. 2000), but habitat availability can also influence the density of conspecifics and nest success (Johnson and Grier 1988, Stephens et al. 2005, Anteau et al. 2012b, Shaffer et al. 2013). Accordingly, the influence of available habitat is intrinsically linked with the number of hatchlings in a given area, and thus with the potential for density-dependence during brood-rearing (Gunnarsson et al. 2006). Identifying the hatchling density at which density-dependence appreciably reduces fledging success would be useful for understanding habitat limitations and prioritizing habitat conservation.

When habitat is unavailable or unsuitable, individuals may have to move greater distances to find adequate habitat (Mace et al. 1983). Furthermore, nesting habitat may differ from brood-rearing habitat, requiring precocial hatchlings to relocate from nesting to brood-rearing sites (Blomqvist and Johansson 1995). Such movements can negatively affect hatchling growth rates and fledging success by increasing energy expenditure, unfamiliarity with location of resources, movement through marginal habitats, or increased predation rates (Rotella and Ratti 1992a, Pearse and Ratti 2004).

Piping Plovers (Charadrius melodus, hereafter “plovers”) have been the focus of many conservation programs and research studies since their federal listing in 1986 (Elliott-Smith and Haig 2004). Plovers typically nest in areas that are unvegetated, flat, and sandy, with occasional scattered pebbles or gravel, and nest sites and territories typically are unevenly or patchily distributed across the landscape (Burger 1987, Prindiville Gaines and Ryan 1988, Elliott-Smith and Haig 2004, Sherfy et al. 2008, Anteau et al. 2012b). Plover broods typically stay near natal sites (within ∼800 m) until fledging at about 25 days old (Knetter et al. 2002, Elliott-Smith and Haig 2004, Harris et al. 2005, Haffner et al. 2009, M. T. Wiltermuth and M. J. Anteau personal observation); thus, selection of nesting territories has implications for chick growth and survival.

Historically, plovers nested on coastal beaches, alkali wetlands, and riverine sandbars (Burger 1987, Prindiville Gaines and Ryan 1988, Elliott-Smith and Haig 2004, Sherfy et al. 2008). Much of the area where plovers breed has been greatly changed by water-management practices (Elliott-Smith and Haig 2004, Anteau 2012, Anteau et al. 2012a). Processes that drive nest-site selection and reproductive fitness may be more disjunct than they were historically, due to water management and other anthropogenic changes (Anteau et al. 2012a). Since the 1950s, with the construction of dams on large river systems in the Northern Great Plains, plovers have begun nesting on reservoir shorelines, both on mainland and island shorelines (Elliott-Smith and Haig 2004). In 2005, 64% of Missouri River plovers were reported in reservoir habitats (Shaffer et al. 2013, C. J. Huber personal communication), and 29% of the Northern Great Plains population used reservoir habitat during the summer of 2006 (Elliott-Smith and Haig 2004, Elliott-Smith et al. 2009). Further, in 2005, 43% of Missouri River adult plovers were reported on Lake Sakakawea, a large (163,800 ha), main-stem reservoir of the Missouri River in northwestern and central North Dakota (Anteau et al. 2014, C. J. Huber personal communication; Figure 1).

There is concern that plovers nesting in manipulated environments may not have access to both quality nesting and foraging habitats (Le Fer et al. 2008a, Sherfy et al. 2012). Shorelines receiving greater wave action due to prevailing winds (i.e. greater wind fetch) experience greater erosion of fine sediments (Anteau et al. 2014). These areas can become prime plover nesting habitat once water levels decrease (Anteau et al. 2012b, 2014). However, these areas may provide less productive benthic conditions for the invertebrates that plovers eat. Plovers select islands over mainland areas (Shaffer et al. 2013, M. J. Anteau personal observation). Although islands may provide better isolation of chicks from predators, they may not provide adequate foraging habitats. We used remotely sensed habitat characteristics to evaluate the potential influences of hatchling density (number of eggs hatched per area of suitable habitat), landform (island vs. mainland), and wind fetch on growth and fledging rates of plover chicks at Lake Sakakawea during 2007 and 2008. Specifically, we evaluated predictions about the influence of the following factors on the growth of plover chicks: hatchling density, mean age-adjusted movement rates of chicks, landform, and wind fetch. We also evaluated predictions about the influence of the following variables on fledging rate: hatchling density (competing density-dependent and density-independent functions), landform, and wind fetch.

Methods

Study Area

Our study area included all mainland and island habitats of Lake Sakakawea from Garrison Dam near Riverdale, North Dakota, to White Tail Bay, North Dakota, USA (Figure 1). We refined our study area to include only the area between the elevation of the water and the maximum flood elevation of the reservoir (565 m MSL [mean sea level]). The reservoir shoreline is irregular and dissected, and consists of numerous substrates, slopes, and aspects. The distribution and area of these features vary annually as lake elevation changes in response to precipitation, melt of Rocky Mountain snowpack, and releases from Garrison Dam (the dam impounding Lake Sakakawea) and Fort Peck Dam (an upstream reservoir in Montana). Lake Sakakawea periodically floods near its maximum holding elevation (e.g., in 1976, 1985, 1998, and 2011); typically, several years of successive drawdowns in water level follow these events, which exposes potential plover habitat (Anteau et al. 2014). Preflooding topography and hydrologic processes have created diverse habitat conditions, including wide beaches where shoreline slopes are gradual, narrow beaches with a terracing pattern of slopes, and cut banks or bluffs of varying elevations. Detailed descriptions of plover brood-rearing habitat, vegetation communities and dynamics, and substrate composition are provided in Anteau et al. (2014).

Sampling Design and Allocation

We used a stratified random sampling design based on historic plover nest counts (Shaffer et al. 2013, C. J. Huber personal communication) to allocate our nest survey efforts. We used National Agricultural Imagery Program aerial photographs from 2004 to delineate the shoreline of Lake Sakakawea, because 2004 represented an approximate midpoint between the full reservoir and the lowest observed water levels. The shoreline was divided into 544 segments of ∼2 km in length (e.g., Figure 1). The majority of segments were mainland (n = 506, 95%), but there were 38 segments that we classified as island because they were detached from the mainland under all of the observed water levels. We classified segments into strata based on counts of plover nests from the U.S. Army Corps of Engineers' (USACE) Least Tern (Sternula antillarum) and Piping Plover monitoring program from 1998 to 2005. We defined strata as low (<2 nests), medium (2–9 nests), or high (>9 nests) use. The classification resulted in 403 low-, 88 medium-, and 53 high-use segments. We determined sampling intensity in each stratum using Neyman allocation (Thompson et al. 1998), and randomly selected 30 unique 2-km segments for surveys in years 2007 and 2008 (see Anteau et al. 2012b). However, we restricted our analyses to the segments in which at least one egg hatched (n_total = 26, n₂₀₀₇ = 14, n₂₀₀₈ = 12; n_unique = 15); these segments were split relatively evenly among island (42%) and mainland (58%) landforms.

Nest and Chick Monitoring

We searched study segments for nests every 2–3 days throughout the nesting season (April–July) during 2007–2008. We systematically searched for nests (a scrape containing ≥1 egg) and also used adult behavior consistent with nest defense to locate nests. Upon discovery of a nest, we recorded a GPS location (post-processed differential correction; Trimble model GeoXT, Trimble Navigation Limited, Sunnyvale, California, USA) and floated eggs to estimate incubation stage (Hays and LeCroy 1971). We revisited nests during each nest search, except we visited nests daily beginning one day prior to expected hatch or if we observed pipped eggs. Frequent nest visits improved our ability to weigh and band chicks soon after hatching.

We captured, weighed (with a digital balance to ±0.1 g), and banded chicks with uniquely identifiable color bands (Shaffer et al. 2013). We periodically captured and reweighed chicks during the brood-rearing period. We weighed chicks at least 2 times before they fledged, unless they died prior to 20 days of age. We considered a plover chick to have fledged if it survived to 20 days (Shaffer et al. 2013), and calculated the fledging rate as the number of chicks on a segment that were observed alive after 20 days of age. Due to intensive chick resighting efforts and visits every 2 days to our segments, our chick detection rates were very high. For example, the probability of detecting a chick during the fledging period (20–25 days) was 0.99 (Shaffer et al. 2013, Roche et al. 2014). Accordingly, we did not adjust for detectability in our analyses. We also assumed that we banded all hatchlings on our segments due to our daily visits to nests near the time of hatching.

When we observed marked broods, we recorded the location of the brood on a GPS using an offset point that incorporated directional and distance estimates (using a compass and hand-held laser range finder or ocular estimate). We calculated movements for each brood, limiting estimates of movement to observations with ≤2-day return intervals. Movement of broods was correlated with age up to 5 days of age, thus we controlled for age-related variation in our calculation of relative movement. We included all ≤2-day-interval brood-movement estimates in a linear regression with brood age (tanh[age/2]); we then calculated the mean of the residuals for each brood to represent the age-standardized brood movements.

Defining Available Habitat and Wind Fetch

We used 3 data sources to estimate the amount of nesting habitat on Lake Sakakawea: a Digital Terrain Model (DTM) with vertical accuracy <1 m, high-resolution optical satellite imagery (2.5-m pixels), and field observations for model training and evaluation. We acquired Probationary System of Earth Observation satellite 5 (SPOT-5, hereafter SPOT) imagery (Satellite Imaging Corporation, Magnolia, Texas, USA) during 2007 and 2008. Collection of imagery occurred in mid- to late-summer to correspond with field habitat observations and the brood-rearing period. We classified each 2.5-m grid as plover habitat if it had: 1) <15% bare-substrate obstruction; 2) <10% slope; and 3) ≥75% of adjoining pixels with <15% bare-substrate obstruction (Shaffer et al. 2013, M. J. Anteau personal observation). These classification thresholds were selected to conform to the scale at which these features were selected and/or avoided based on a nest-site selection study (Anteau et al. 2012b), but were also relevant for brood-rearing habitat (M. T. Wiltermuth and M. J. Anteau personal observation). Our bare-substrate obstruction classifications were 76% (2007) and 75% (2008) accurate, and omission and commission errors were equal (12%; Shaffer et al. 2013). We calculated the total area (ha) of potential habitat for each segment by summing the area of pixels classified as plover habitat.

The amount of exposure to wind and wave erosion can vary on reservoir shorelines due to the complex shoreline shape. Reservoir shorelines often include long bays where tributaries join with the main-stem river; these bays can have relatively little exposure to wave action relative to shorelines on the main body of the reservoir that face the prevailing wind. We calculated mean wind fetch for each study segment. We defined wind fetch as the linear distance from the shoreline of the segment, across the open water, to the nearest shoreline following the direction of the prevailing wind. We took the mean of 5 measurements evenly spaced along each segment. Weather data from 1950 to 2010 indicated a northwest prevailing wind.

Statistical Analyses

Fledging rate

We found no evidence of correlations among our covariates (between wind fetch and hatchling density or hyperbolic-tangent-adjusted hatching density; |r| < 0.06; PROC CORR; SAS Institute 2002). We identified factors that correlated with fledging rates of plovers by comparing 9 models using an information-theoretic approach (Burnham and Anderson 2002; Table 1). Our study was repeated across years and we specified segment as the subject in a mixed-model analysis of variance (PROC MIXED; SAS Institute 2002). We used a maximum-likelihood estimation method and specified the covariance structure for year as compound symmetry. We examined the residuals of our full model for normality, examined plots of residuals for each continuous variable, and concluded that the model satisfied assumptions.

To consider abundance of habitat and potential interactions with number of hatchlings, we calculated hatchling density using segment habitat summaries. If fledging rate were density independent, we would expect a positive linear relationship between fledging rate and hatchling density. However, if fledging rate were density dependent, we would expect fledging rate to increase with the number of hatchlings up to a point and then plateau and remain constant with an increasing density of hatchlings. We allowed for density dependence in our modeling by iteratively fitting our full model with hyperbolic tangent (TANH) transformations of hatchling density at a number of density (D) quotients (e.g., D/1, D/5, D/6.5, D/8, and D/10). TANH creates an asymptotic curve; dividing the density by a constant shifts the inflection point of the curve on the x axis. The best-fitting curved transformation of hatchling density was TANH(D/6.5) based on the value of Akaike's Information Criterion corrected for small sample size (AIC_c).

Our a priori model set comprised 9 models, 4 of which assumed density independence, 4 of which assumed density dependence, and a null model (Table 1). We did not include the fixed effect of year in any model, because density-dependent and density-independent full models fit better without the fixed effect of year (>4.0 ΔAIC_c). However, year was retained in the repeated statement. We also did not include any interactions, because our data were insufficient to support more complex models. All candidate models converged; they ranged in complexity from 3 to 6 parameters. We ranked models by AIC_c and interpreted results from the most parsimonious models (Burnham and Anderson 2002). Estimates from our top model were made by holding all other effects at their median level; we calculated 95% confidence limits for estimates. The majority (81%) of our segments had hatched chick densities of fewer than 5 hatchlings per hectare. Accordingly, we reran our models using only data from segments with <5 hatchlings per hectare, so that we could evaluate if there was evidence of density dependence at more common densities of hatchlings.

Growth rate

We ln-transformed mean mass of chicks within a brood for each capture event (Le Fer et al. 2008a). Because we found no support for a year effect on growth rates of plover chicks (ln(mean chick mass) = age age-by-year; PROC MIXED; SAS Institute 2002), we used a regression analysis to develop a single growth curve for chicks (PROC REG; SAS Institute 2002).

We also examined how landform, wind fetch, and chick density influenced chick growth rates at our 2-km segment scale. For this analysis, we used only the mean chick mass for each brood from the capture occasion when the brood was oldest. This approach should have maximized effects due to posthatch growth over effects that were maternal contributions through eggs. We regressed mean chick mass on age and used the residuals of that analysis to index age-adjusted chick growth. We then averaged all age-adjusted growth mass values for each segment and used those values to index chick growth rate for each segment. We also computed an index for the distance of brood movements (hereafter relative movement) on the segment by averaging all of the age-adjusted brood movements for that segment.

We found no evidence of strong correlations (|r| < 0.33) among our covariates (PROC CORR; SAS Institute 2002). We compared 8 models (Table 2) using an information-theoretic approach (Burnham and Anderson 2002). We did not include the fixed effect of year in any models, because our full model fit better without year (ΔAIC_c > 4). We also did not include the repeated statement (as above), because the covariance parameter estimate was near zero (−0.0005) and the model fit better without it (ΔAIC_c > 10). All models were run with the Mixed Procedure in SAS using maximum-likelihood estimation (PROC MIXED; SAS Institute 2002). We examined the residuals of our full model for normality, examined plots of residuals for each continuous variable, and concluded that the model satisfied assumptions. All candidate models converged; they ranged in complexity from 4 to 8 parameters. We ranked models by AIC_c and interpreted results from the most parsimonious models (Burnham and Anderson 2002). Estimates from our best model were made by holding all other effects at their median level, and we calculated 95% confidence limits for estimates.

Results

We found 103 and 83 nests and banded 94 and 81 chicks, representing 30 and 36 unique broods, during 2007 and 2008, respectively. Broods were observed at a mean interval of 2.3 ± 1.5 SD and 2.1 ± 1.4 SD days during 2007 and 2008, respectively. Hatchlings per segment averaged 6.7 (range = 1–28, CV [SD/mean] = 0.85). The number of chicks fledged per segment averaged 2.1 (range = 0–9; CV = 1.14). The amount of habitat at surveyed segments with chicks that hatched averaged 5.7 ha and ranged from 0.2 to 19.2 ha (CV = 0.87). The density of chicks hatched per segment averaged 3.8 chicks ha⁻¹ and ranged from 0.2 to 27.1 chicks ha⁻¹ (CV = 1.70). Mean wind fetch for segments averaged 1.5 km and ranged from 0.0 to 10.4 km (CV = 1.49).

Fledging Rate

Our top model of chick fledging rate accounted for 96% of the model weight; it assumed density dependence and included landform and wind fetch (Table 1). Fledging rate estimates from the density-dependent model increased with the density of hatchlings, but the slope stabilized between approximately 5 and 15 hatchlings per hectare (Figure 2). When we reran our analyses on segments with <5 hatchlings per hectare (n = 21), there was considerable model uncertainty between the full density-dependent model (LL = −21.7, AIC_c = 61.3, ΔAIC_c = 0.0, and w_i = 0.52) and the full density-independent model (LL = −21.8, AIC_c = 61.5, ΔAIC_c = 0.2, and w_i = 0.48). No other models had an Akaike weight ≥ 0.01. While estimates of the number of fledglings per segment diverged between density-dependent and density-independent models (Figure 3), their model fits were nearly identical. The slope of the density parameter in the density-independent model was shallower when the model was fitted to the entire dataset (slope = 0.2, SE = 0.02, n = 26) than when the model was fitted to the restricted dataset (slope = 0.6, SE = 0.01, n = 21), indicating nonlinearity in the relationship between hatchlings and fledglings at densities above 5 hatchlings per hectare.

While holding wind fetch and density at median values, the fledging rate was 4.3 times greater on the mainland (x̄ = 3.0, 95% CL = 2.9–3.2) than it was on islands (x̄ = 0.7, 95% CL = 0.5–0.9). Chick fledging rate declined by −0.7 ± 0.04 SE fledglings per 2-km segment for every km increase in wind fetch.

Growth Rate

Mean chick mass (ln-transformed) was positively correlated with age (intercept = 1.717, Β̂ = 0.080, SE = 0.003, R² = 0.86, n = 139). The residuals of this regression were averaged by segment and used to index chick growth. There was considerable uncertainty in model selection to examine relative chick growth (Table 2). Our null model and the relative movement model had equal AIC_c values, but we ranked the relative movement model second because it had one more parameter (Table 2). Based on our second-ranked model, relative chick growth was negatively correlated with relative movement (Β̂ = −0.276, SE = 0.170, R² = 0.10; Figure 4).

Discussion

We found little support for density-dependent processes influencing the fledging rates of chicks when densities were <5 hatchlings per hectare, which comprised >80% of areas we studied. However, there was support for density dependence influencing the fledging rates of chicks at densities >5 hatchlings per hectare. Our data suggest that fledging rates no longer increase with increasing hatchling density greater than about 10–15 hatchlings per hectare. At 2 beaches near Long Island, New York, USA, there was no evidence to support density dependence affecting Piping Plover chicks; the greatest density observed during 13 years of data collection was 1.05 nesting pairs per hectare (Cohen et al. 2009), which is comparable to the densities that we observed at 80% of our study areas, given that brood size at hatching is typically 4. However, comparisons of density-dependent limitations among breeding areas may be problematic because variations in predator communities and food abundance influence habitat quality (Murphy et al. 2001). Regardless, our findings suggest that quantity or quality of habitat at Lake Sakakawea did not appreciably limit the fledging rates of plovers in most areas during our study years. Further, if our study years are representative of long-term habitat quality, then 5–15 hatchlings per hectare could be used to help define a brood-rearing habitat goal for plovers at Lake Sakakawea. However, habitat abundance varies markedly interannually (Anteau et al. 2014), thus further study is needed to understand how or if habitat quality or predator communities vary among years.

We have little information to speculate about mechanisms driving density-dependent processes. This is largely because we observed only a few areas (19%) where density-dependent processes appeared important. There was no support for the notion that chick density influenced chick growth rates, suggesting that food resources may not have been limiting, even in dense chick-rearing areas. Regardless of density, we observed no agonistic interactions among broods or adults during our visits. Accordingly, further study is needed to understand what processes drive density dependence in fledging rates of plovers chicks.

We documented that the fledging rate for plovers was lower on islands than on the mainland. We are unsure what may have caused this; during data collection we did not observe any obvious factors that may have led to lower fledging rates on islands. However, predation rates of chicks may be higher on islands because gulls nest at and frequent islands at Lake Sakakawea (Dwernychuk and Boag 1972, Sayler and Willms 1997). Piping Plovers selected islands over the mainland for nesting (Murphy et al. 2001, Shaffer et al. 2013, M. J. Anteau personal observation), and nest survival did not vary by landform (Anteau et al. 2012a). Accordingly, selection for island breeding sites on Lake Sakakawea during our study appeared to be a suboptimal decision and further supports the hypothesis that altered habitats, such as this reservoir, can create ecological traps, because cues used for selection no longer positively correlate with realized recruitment (Anteau et al. 2012a, 2012b).

We found modest support for a tradeoff between chick movements and growth rates. We suspect that chicks that must move more to find adequate foraging habitat grow at a slower rate. This suggests that the amount of brood-rearing habitat or its juxtaposition in relation to nesting areas was inadequate for some individuals. The range of relative movement distances of plover broods mostly overlapped with respect to landform (5^th, 25^th, 50^th, 75^th, and 95^th percentiles: for islands = −0.04, −0.02, 0.00, 0.05, 0.12; for the mainland = −0.04, −0.01, 0.01, 0.06, 0.32). Thus, relative movement rates cannot explain the lower plover fledging rates observed on islands.

Growth rates of chicks generally reflect food resources, serving as indicators of habitat suitability (Martin 1987, Lepage et al. 1998). Plover growth rates in our study were similar to those for alkali wetlands in North Dakota and those reported for Lake Sakakawea during 2001–2003 (Le Fer et al. 2008a). However, the growth rates that we documented were greater than those reported for plovers on sandbars of the Missouri River near Bismarck, North Dakota, and Yankton, South Dakota, during 2001–2003 (Le Fer et al. 2008a). These comparisons suggest that foraging conditions at Lake Sakakawea are similar to those at other areas that have been studied and perhaps are of no greater management concern than at other plover breeding sites. However, our chick growth rate models were unable to explain much of the observed variability. In addition, we did not obtain data on growth rates from island segments after chicks reached 10 days of age. Accordingly, further research on foraging resources and growth rates on islands would be prudent to explore the effects of foraging limitations, especially because we documented much lower fledging rates on islands. A more intensive study of growth rates at a finer spatial scale (e.g., the brood's home range prior to recapture) may better explain our observed variation in chick growth rates, and may reveal whether foraging limitations are an important factor behind the decreased fledging rates on islands.

Our results indicated that fledging rates were lower in areas that were more exposed to wind and waves than in areas that were more protected. On the Atlantic coast, areas protected from wind and waves produced more invertebrate forage for Piping Plovers, were used more by Piping Plover chicks, and had higher survival rates of Piping Plover chicks than exposed areas (Loegering and Fraser 1995, Elias et al. 2000, Fraser et al. 2005). On Lake Sakakawea, exposed areas experienced more beach erosion from wave action, which led to expanses of gravelly substrate (Anteau et al. 2014). We suspect that exposed areas generally had fewer food resources than those that were more protected, because nutrients tend to leach out of coarse substrates in exposed areas, therefore these substrates likely support fewer invertebrates than fine soils (Anderson and Day 1986, Sayler and Willms 1997, Le Fer et al. 2008b). Wave action on exposed areas may deposit macroinvertebrates on the shoreline that plovers could eat; however, it is unlikely that many macroinvertebrates wash up on the shoreline of Lake Sakakawea because of the lake's abundant fish populations that likely deplete macroinvertebrates in the water column (Sayler and Willms 1997, Wooster 1998, Anteau et al. 2011). Furthermore, foraging in the wave wash zone on exposed shorelines may require more energy and may expose plover chicks to extreme weather conditions.

Plovers are attracted to exposed gravel patches for nesting (Anteau et al. 2012b), even though we found evidence that these types of areas with greater wind fetch have lower fledging rates than more protected areas. These circumstances result in an ecological tradeoff, because there appear to be conflicting factors influencing habitat selection and recruitment. However, certain shoreline features, such as points and peninsulas, may provide both exposed areas for nesting and protected areas for brood rearing. While not quantified, we observed an apparent preference for nesting on points and peninsulas. Accordingly, we suspect that complexes of exposed gravelly areas and protected shorelines with fine substrate in close proximity may provide the habitats and resources needed for successful plover recruitment.

Our findings on plovers have implications relevant to the conservation of other precocial birds. Tradeoffs may exist between habitat selection and suitability for nesting and brood-rearing periods. These tradeoffs can be especially important when species are utilizing novel or highly modified habitats (see Schlaepfer et al. 2002, Anteau et al. 2012a). We contend that understanding selection cues, habitat suitability, and tradeoffs could help to inform conservation efforts in order to provide complexes of habitat that would be selected and beneficial for both nesting and brood-rearing.

Acknowledgments

This study was funded by 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. We are grateful for technical support from the U.S. Geological Survey (USGS) Northern Prairie Wildlife Research Center Missouri River Least Tern and Piping Plover Research Team, including Marsha Sovada, Larry Strong, Jennifer Stucker, and Erin Roche. We especially thank Melisa Bernard, Betty Euliss, and Nickolas Smith for their work on remote sensing and GIS. We thank Phil Brown, Deb Buhl, Tom Buhl, Colin Dovichin, Anthony Hipp, Coral Huber, Casey Kruse, Michael Morris, Brandi Skone, Nickolas Smith, and Ryan Williamson for help with project planning and logistics, and the many field technicians for their assistance with data collection. We thank Courtney Amundson and anonymous reviewers for valuable comments on earlier versions of this manuscript. Our field protocols were approved by the USGS Northern Prairie Wildlife Research Center Animal Care and Use Committee. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Literature Cited