Abstract
The management of wintering North American waterfowl is based on the premise that the amount of foraging habitat can limit populations. To estimate carrying capacity of winter habitats, managers use bioenergetic models to quantify energy (food) availability and energy demand, and use results as planning tools to meet regional conservation objectives. Regional models provide only coarse estimates of carrying capacity because habitat area, habitat energy values, and temporal trends in population-level demand are difficult to quantify precisely at large scales. We took advantage of detailed data previously collected on wintering waterfowl at Edwin B. Forsythe National Wildlife Refuge and surrounding marsh, New Jersey, USA, and created a well-constrained local model of carrying capacity. We used 1,223 core samples collected between 2006 and 2015 to estimate available food. We used species-specific 24-h time-activity data collected between 2011 and 2013 to estimate daily energy expenditure, morphometrically corrected for site- and day-specific thermoregulatory costs. To estimate population-level energy demand, we used standardized monthly ground-surveys (2005–2014) to create a migration curve, and proportionally scaled that to fit aerial survey data (2005–2014). Crucially, we also explicitly incorporated estimates of variance in all of these parameters and conducted a sensitivity analysis to diagnose the most important sources of variation in the model. Our results indicated that at estimated mean levels of supply (2.34 × 109 kcal) and cumulative demand (3.4 × 109 kcal), refuge resources were depleted before the end of the wintering season. However, at one standard error greater in supply and one standard error less in demand, 1.33 × 109 kcal remained on the landscape at the end of winter. Variation in model output appeared to be driven primarily by uncertainty in food abundance in high marsh habitats. This model allows for relative assessment of biases and uncertainties in carrying capacity modeling, and serves as a framework identifying critical science needs to improve local and regional waterfowl management planning.
Introduction
Management of North American waterfowl is often touted as a premier example of a successful wildlife conservation program (Williams and Castelli 2012). The North American Waterfowl Management Plan (and its revisions) provides an overarching framework for waterfowl conservation, establishes regional population goals, and tasks interagency Migratory Bird Joint Ventures (partnerships established under the North American Waterfowl Management Plan to help conserve the continent's waterfowl populations and habitats; hereafter, JVs) with meeting these goals (U.S. Fish and Wildlife Service 2012). During the nonbreeding period, these policy frameworks typically assume food is the limiting factor for populations. Therefore, wintering JVs focus on meeting population objectives by providing enough foraging habitat. To estimate how many ducks the habitat can support, some JVs have built bioenergetic models that incorporate population goals, waterfowl energetic demand, and the foraging values of different habitat types (e.g., Gulf Coast Joint Venture [Esslinger and Wilson 2001] and Central Valley Joint Venture 2006).
Within the boundaries of any particular JV, State and National Wildlife Refuges (“refuges”) provide a substantial amount of wintering habitat and can support regionally and continentally significant proportions of waterfowl populations. Managers of large public refuges are interested in the same questions that are being addressed at regional scales by JVs: 1) how many ducks are using the habitat, and 2) how many ducks can the refuge theoretically support? As a step-down approach, these questions may actually be easier to answer at local (refuge) scales, where population abundances and habitat areas can be estimated more precisely. Building robust, refuge-specific carrying capacity models helps inform management of those areas, which can comprise a large percentage of JV habitat and population goals. Additionally, well-constrained local models can identify sources of bias and uncertainty to help improve stepped-up regional models of carrying capacity.
One refuge ideally suited for building such a model is the Edwin B. Forsythe National Wildlife Refuge, (hereafter, “Forsythe”) located in coastal New Jersey, which provides a core wintering site for Atlantic Flyway waterfowl populations. Greater than 20,000 ha in size, most of the Forsythe refuge consists of salt marsh with several freshwater impoundments created to provide moist-soil seeds for wintering waterfowl. The most abundant dabbling duck wintering at Forsythe is the American black duck Anas rubripes, a species of conservation concern since their population began a steady decline in the 1950s (Rusch et al. 1989; Conroy et al. 2002; Sauer et al. 2014). There has been a great deal of waterfowl research conducted at Forsythe over the past decade, much of which has sought to address information gaps in black duck ecology. These include detailed behavioral studies of waterfowl movements (Ringelman et al. 2015b), energy expenditure (Cramer 2009; Jones 2012), and food availability (Cramer et al. 2012; Ringelman et al. 2015a; Goldstein et al. 2016). In addition, several different population surveys are available for Forsythe, including the annual Midwinter Waterfowl Survey (MWS), standardized ground surveys from 2005–2014, and aerial transect surveys for black ducks from 2011–2014.
In short, this accumulation of refuge-specific data afforded us a unique opportunity to build a well-parameterized waterfowl carrying capacity model for Forsythe. We explicitly addressed various sources of sampling error stemming from model assumptions, estimates of food supply, and estimates of energetic demand. We explicitly modeled the variation in parameter estimates and used a sensitivity analysis to diagnose how each contributed to overall model uncertainty to help waterfowl biologists identify the most important avenues for future research and planning at both local and regional scales.
Study Site
Forsythe spans 77 km of the Atlantic Coast of southern New Jersey (3927′N, 74°24′W). For the purposes of this study, we considered the study area to be portions of MWS segments 11, 13, and 14, which span the lower third of Forsythe as well as Absecon and Great Bay Boulevard Wildlife Management Areas (22,091 ha; Figure 1). For simplicity, in this manuscript we refer to our study area as “Forsythe,” although the area boundaries extend beyond the refuge. This study area exactly corresponded to that used by Jones (2012) to estimate black duck energy demand. Within Forsythe, there were a range of foraging habitats available to black ducks, including managed freshwater marshes, forested wetlands, salt marsh that is regularly (low marsh) or irregularly (high marsh) flooded by the tide, mudflats, and shallow subtidal habitats.
Methods
To build a waterfowl carrying capacity model, at a minimum, researchers must estimate both the energy supply on the landscape available to waterfowl and energetic demand of the birds. To estimate supply, this requires quantifying 1) the amount of each type of habitat available, and 2) energy value of those habitat types. Estimating waterfowl energetic demand requires quantifying 1) the daily energy requirement of each species, and 2) how many individuals of each species are present each day during the time period in question. We address each of these points in turn. All analyses were performed in R (version 3.0.1; R Core Team 2013). We present means ± SE unless otherwise noted.
Quantifying energy supply
We used the U.S. Fish and Wildlife Service National Wetlands Inventory layer in ArcMap 10.2 (ESRI 2012) to identify habitats available to foraging dabbling ducks Anas spp.. This layer was slightly modified by Black Duck JV scientists to split (where tractable) unclassified salt marsh into high marsh, low marsh, and mudflat. We also included freshwater impoundments, subtidal habitats shallower than 1 m (identified by National Oceanic and Atmospheric Administration bathymetry data), and forested wetlands in our analysis (see Ringelman et al. 2015b for additional information). We assumed that sand and open water consistently deeper than 1 m had no foraging value for dabbling ducks. To quantify the amount of each habitat type present, we clipped the National Wetlands Inventory layer with the outside perimeter used by Jones (2012) and calculated the total number of hectares of each habitat type (Table 1).
We used data from Cramer et al. (2012), Goldstein et al. (2016), Fino et al. (2017), and newer unpublished data to evaluate foraging resources in high marsh, low marsh, mudflats, subtidal habitats, forested wetlands, and freshwater impoundments within our study area. We used an average value for salt marsh habitats to estimate energy availability in unclassified salt marsh. These data sets excluded foods not known to be eaten by black ducks, as well as ribbed mussels Geukensia demissa and killifish Fundulus spp. deemed too large (>21.4mm) to be consumed by black ducks (Cramer et al. 2012). Despite this, outlier data points clearly remained in the core sampling data. These remaining outliers could be the result of naturally variable densities of food items or the result of inadequate removal of biological outliers. Therefore, we also employed the statistical removal of food items following Ringelman et al. (2015a), hereafter referred to as outlier-removed data. Rather than statistically removing outliers from each habitat in each study, we took a conservative approach and pooled all data and then removed outliers from each habitat type (n = 80 outliers removed in total; Table 1). For most of our analyses, we used the outlier-removed data to estimate available energy (Tables S1 and S2, Supplemental Material), but present outlier-included data for comparison. Ringelman et al. (2015a) counterintuitively found no evidence for food depletion in any habitat at Forsythe; however, because potential replenishment of coastal marsh foods remains an active research question, we developed both constant-supply and monthly depletion models.
Quantifying demand
Daily energy expenditure
Existing carrying capacity models use rough approximations for daily energy expenditure (DEE). Some use single-species estimates of metabolic rate and linearly scale that value to other species based on body size (Gulf Coast Joint Venture; e.g., Esslinger and Wilson 2001), while others use allometric equations (Miller and Eadie 2006) and loosely estimate the average body size across all species to obtain a single value (Central Valley Joint Venture 2006). For our analysis, we used DEE values from Jones (2012), who used 24-h behavioral observations, species-specific allometric estimates of resting metabolic rate (Miller and Eadie 2006), and metabolic multipliers (Wooley Jr and Owen Jr 1978) to estimate baseline daily energy expenditure. Following McKinney and McWilliams (2005) and Livolsi (2015), we used morphometric measurements and site- and day-specific weather data to correct these baseline values to account for the cost of thermoregulation. Below is the complete DEE equation we used from Livolsi (2015), where ai is the activity-specific multiplier of RMR, CT is the cost of thermoregulation (kcal/bird/h), and Ti is the amount of time engaged in a given activity:
Population estimates
The MWS has been performed annually in the United States since the 1930s to track population trends for wintering ducks, but has been criticized for lack of sampling design and estimates of variance (Eggeman and Johnson 1989; Reinecke et al. 1992). However, the MWS is well-suited to the Atlantic coast, where black ducks are widely distributed and easily counted from the air (Conroy et al. 1988). Indeed, Conroy et al. (1988) used 4 y of Atlantic Flyway aerial transect data to confirm that the MWS fell within the 95% confidence interval (CI) of their transect estimate for all 4 y. To verify that the MWS was appropriate at our smaller scale, we established 9 aerial transects (1.3–22.2 km long) spaced 2 km apart over our study area. Transects were flown at an elevation of 45 m, a speed of 90 knots, and observers surveyed a 152 m-wide area on each side of the plane. Transects were flown 3 times/y during early January 2011–2013 by the same crew and aircraft that conducted the MWS. We calculated the mean number of black ducks per transect for each year, and estimated the population size using a combined-ratio estimator (Cochran 1977; Conroy et al. 1988). We compared these estimates with MWS totals from segments 11, 13, and 14 in each year. Midwinter survey estimates for black ducks fell within the 95% CI of the black duck transect survey in all but 1 y, indicating that the MWS is a reasonable index of black duck abundance (Table 2). In our analyses, we used MWS estimates because the transect survey only counts black ducks (not other species) and also to take advantage of the longer term MWS data set to build a more robust migration chronology.
Migration chronology
A major challenge in quantifying energy demand is estimating how many birds are using a refuge during each day of the wintering season. Here, we used standardized diurnal ground-count surveys (2005–2014) conducted at Forsythe to estimate waterfowl migration chronologies. Three to four times per month throughout the year, trained observers counted waterfowl along a 12.9-km unpaved road that affords viewing access to both salt marsh and freshwater impoundments. We used the average monthly count from this survey to construct an approximate migration index for waterfowl during the nonbreeding season (August–March; e.g., black ducks; Figure 2). Because ground observers can only see a portion of the ducks present, we then converted raw ground-counts to proportions of peak abundance. We used the January population estimates from the MWS as an anchor point, and scaled our ground-count migration chronology to a population-level chronology using proportions (Figure 2; Table S3, Supplemental Material).
Carrying capacity modeling
Commensurate with our migration chronology, we modeled depletion of resources on a monthly timescale with the following equations (specific to habitats and species):
To estimate variance in supply and demand, we ran the above equations using values of the mean ± 1 SE for kcalhabitat, DEEspecies, and monthly midpoint populationspecies. We also created a model with no depletion, where supply was held constant. Our model assumed omniscient foragers, no interference competition, complete utilization of the resources, and no other sources of depletion.
Sensitivity analysis
To determine the extent to which uncertainty in each parameter contributes to the model outcome (time until resource depletion), we conducted a sensitivity analysis. We randomly drew parameters for species-specific DEE, species-specific abundance in each month, and habitat-specific energy totals. We drew parameters from a normal distribution with empirically derived mean and variance, but we discarded parameter draws outside ± 1 SE to be comparable with our bounding conditions in our other analyses. We calculated the total depletion over the wintering period following the equations above, and ran 100,000 simulations. We then used linear regression to model total depletion as a function of every input variable, and used standardized regression coefficients to evaluate the strength of each in determining the response variable (Miller et al. 2014).
Results
Energy supply
Our study area was 22,091 ha, primarily consisting of high marsh (51%) and subtidal habitats shallower than 1 m (45%; Table 1). Freshwater habitats rich with moist-soil seeds had a high energy value (∼250,000 kcal/ha), but comprised only 1.6% of the habitats at Forsythe (Table 1). Animal food outliers in the data resulted in estimates of food availability that were 1–2 orders of magnitude greater in saline habitats (Table 1). After removing outliers, the total energy potentially available to foraging waterfowl was 2.34 ± 1.49 × 109 kcal, most of which was derived from high marsh (82%). Managed freshwater impoundments comprised <4% of the available energy for wintering waterfowl, whereas high marsh habitats accounted for >82% of available energy. When the 80 core-sampling outliers were included, landscape-level estimates of food availability increased by an order of magnitude to 14.20 ± 7.89 × 109 kcal.
Energy demand
Black ducks were the most abundant waterfowl observed during ground surveys, followed by mallard Anas platyrhynchos, northern pintail Anas acuta, and green-winged teal Anas crecca. American wigeon Anas americana, gadwall Anas strepera, and northern shoveler Anas clypeata were observed in low numbers throughout the nonbreeding season. Black duck abundance peaked in December (Figure 3) as birds migrated through to points further south for the winter. As expected, northern pintail and green-winged teal migrated through Forsythe earlier (in October) than mallards (in November), and few teal remained at Forsythe for the duration of the winter (Figure 3).
We used a DEE of 440.49 ± 4.46 kcal/bird/d for American black ducks, 395.88 ± 13.21 for mallards, 353.49 ± 16.07 for northern pintail, 176.61 ± 17.21 for green-winged teal, and 221.36 ± 17.78 for northern shoveler. Too few American wigeon and gadwall were observed at Forsythe to compute DEE values, so we used values estimated by Livolsi (2015) from ducks in Delaware for these species (272.59 ± 40.42 for American wigeon and 232.99 ± 18.96 for gadwall). After scaling migration chronologies up to estimate populations and daily energetic demand, black ducks comprised the majority of energetic demand at Forsythe (Figure 3).
Carrying capacity
We summed daily energy expenditure for all species for each month and modeled food energy depletion over the nonbreeding season. For comparison, we also modeled resources that were not depleted (sensu Ringelman et al. 2015a; Figure 4). In the depletion model, at average values for food abundance and energetic demand, the model indicated that refuge resources were fully depleted by 15 December (Table 3). However, there was substantial uncertainty in this result (Figure 4). If energy supply was one standard error higher, and demand was one standard error lower than the mean, then resources would last at least throughout the wintering season; if energy supply was one standard error lower and demand was one standard error higher, food would run out by early October (Table 3). The nondepletion model, although not intuitively realistic, could represent potential replenishment of resources in a tidal system and it also indicated that our area could support a much larger waterfowl community than it currently does. Finally, we also modeled landscape-level food resources based on the full set of core samples, including outliers (Figure 5). At average levels of supply and demand, this model showed >10.8 × 109 kcal remained on the landscape at the end of the nonbreeding season.
Sensitivity analysis
At the coarsest scale, we modeled total energy depletion as a function of species-specific DEEs, total populations over the wintering season, and total food availability. This analysis confirmed initial results (Figure 4): food depletion at Forsythe was driven primarily by the amount of total energy available (β = 0.974), as opposed to waterfowl abundance (β = −0.221) or species-specific DEEs (all β < |−0.020|). Our fully parsed model showed that the energy density of high marsh was by far the most sensitive parameter; waterfowl abundance in any month (except August) was more sensitive than any estimate of DEE (Table 4).
Discussion
Carrying capacity models are ubiquitous tools for managing nonbreeding waterfowl, and used by JVs to prioritize regional habitat delivery goals. Out of necessity, such large-scale models are typically built using simplified or approximate parameter estimates for both supply (habitat quantity and quality) and demand (populations, migration chronologies, and DEE estimates). Because of the inherent error associated with these approximations, existing models are unable to explicitly incorporate statistical error in parameter estimates; thus, model results (and downstream management actions) are based only on mean values. It is well-known that these regional models are plagued by uncertainty, but we lack the biological information at the appropriate scale to satisfactorily diagnose sources of error. Simply put, we are unable to estimate how much we do not know.
Our goal was to constrain and explicitly model sources of uncertainty by building a carrying capacity model with supply and demand data of unprecedented resolution, and provide rigorous estimates of carrying capacity for an important wintering area in the Atlantic Flyway. On the supply side, we used the best available refuge-specific geographic information system layers to estimate the amount of each habitat type, and estimated food density using >1,200 core samples. Ringelman et al. (2015a) recommended removing core sample outliers, especially those with large animal foods, from the data. Indeed, our model indicates that failure to remove outliers results in an astonishingly high estimate of black duck food supply. Certainly, the treatment of outliers in core sampling data is of paramount importance in estimating landscape-level energy availability, but continues to receive little attention from waterfowl biologists and managers. Using outlier-removed data to estimate supply, we found that depletion of food resources was more sensitive to initial food abundance than energetic demand.
Indeed, our model of carrying capacity was highly sensitive to the amount of food available—primarily in high marsh, which comprised the majority of energy available to dabbling ducks. Despite analyzing data from 347 high marsh core samples, and removing 24 outlying samples, the resulting spread of energy values effectively precluded estimating carrying capacity (time to depletion) with any reasonable level of certainty. Adjusting depletion curves by one standard error from the mean yields an estimate of either all food being depleted by 6 October, or there being >1.3 × 109 kcal left on the landscape in March. This variation in supply may be real; that is, no amount of core sampling will reduce estimates of variance in a habitat that is inherently variable. In cases such as ours where managers believe the estimated variation is not due to insufficient sampling (Ringelman et al. 2015a), making decisions based on the mean or median value may be the best course of action (Straub et al. 2012). Further research on the underlying drivers of this variation in habitat quality—and mechanisms to manipulate them—may prove to be fruitful investments.
On the demand side of the equation, we used species-specific 24-h time-activity budgets corrected for the cost of thermoregulation to estimate daily energy expenditure. This represents a substantial improvement over existing models that use either a single value for all species (Central Valley Joint Venture 2006) or linearly scaled estimates based on mallards (e.g., Esslinger and Wilson 2001). Our sensitivity analysis indicated that estimates of daily energy expenditure were not the most important determinants of carrying capacity at Forsythe. However, it is important to note that we only explored the parameter space associated with measurement error in DEE (which was well-constrained). Additionally, there are lingering concerns regarding the validity of metabolic multipliers (Miller et al. 2014), and further refinements to these are in progress (C.K. Williams, unpublished data).
We used ground-counts to estimate migration chronologies and then scaled that to population estimates from the Mid-winter Waterfowl Survey. This represents an advancement over linear migration chronologies (Central Valley Joint Venture 2006), and may be a broadly useful method if obtaining frequent population-level estimates (e.g., from aerial transects; Esslinger and Wilson 2001) is cost prohibitive. Our counts were conducted in sanctuary habitat, and we had no way to estimate if and to what extent individual sampling counts were influenced by hunting in surrounding marshes. However, we believe that by averaging weekly ground-counts by month over 9 y, our chronology is robust to these perturbations. Model outcomes were sensitive to waterfowl abundance during peak periods; refining these estimates should be a high priority for waterfowl managers, especially if habitat energy values cannot be refined further. Indeed, as waterfowl distributions and migration chronologies continue to shift with changes in landscape use and climate, tracking the dynamics of waterfowl abundance will become increasingly critical. This is especially true in areas where waterfowl habitat availability depends on well-timed flooding of moist-soil and agricultural fields.
Several factors influencing supply and depletion were not included in this model. For example, black ducks are known to eat snails Melampus bidentatus on marsh grass and killifish Fundulidae spp. in the water column, which are poorly sampled with coring; therefore, our estimates of food supply were likely biased low. We did not model uncertainty in the energy content of waterfowl foods, which could increase variation in energy supply by ≥10% and likely closer to 30–40% (Livolsi et al. 2015). We did not account for niche differentiation in habitat use, nor species-specific dietary preferences, but focused on habitat and food types selected by black ducks (Cramer et al. 2012); this probably had a minimal influence on the outcome of the model given that black ducks comprise the majority of energetic demand at Forsythe. We also did not model depletion through degradation or consumption by nontarget species such as shorebirds or include foraging effectiveness or thresholds (Central Valley Joint Venture 2006, Miller et al. 2014; Hagy and Kaminski 2015). Given substantial variation in observed foraging thresholds (10–459 kg/ha; Hagy and Kaminski 2015), the fact that these thresholds have never been studied in coastal systems and most recent evidence shows no depletion in our system, we were reluctant to include these in our model (Ringelman et al. 2015a). Nevertheless, as a post hoc analysis, we were interested in how food availability in our salt marsh system compared with foraging thresholds observed in inland wetlands. We used an average true metabolizable energy content for waterfowl foods of 1.55 kcal/g (Livolsi et al. 2015) to back-calculate an average kg/ha over our entire study site (outliers-removed data). Our high (mean + 1 SE) estimate of food availability roughly equates to 112 kg/ha, which is below the observed foraging threshold in many other systems (Naylor 2002; Greer et al. 2009; Hagy and Kaminski 2012). Simply stated, the threshold at which ducks cannot (or will not) forage (profitably) could appear as a horizontal line anywhere on Figure 4. Clearly, landscape-level resource dynamics that emerge from the foraging ecology of waterfowl deserve closer scrutiny in coastal systems.
The most useful aspect of our modeling exercise was in explicitly incorporating uncertainty in parameter estimates, which has been largely ignored in carrying capacity modelling. Our model shows that Forsythe is either well above or well below carrying capacity given variation in supply and demand. We believe there is value in understanding this variation and its underlying sources for several reasons: 1) it helps managers understand and plan for a range of scenarios; 2) it identifies critical knowledge gaps to help direct funding agencies and applied researchers; and 3) transparency about model uncertainty builds trust with regulatory agencies and the general public. For example, at Forsythe it is clear that energy supply, not energetic demand determines carrying capacity. Given our apparent inability to satisfactorily reduce variation in energy estimates for salt marsh by core sampling, new focus should be given to understanding (and managing) the drivers of habitat quality. Energetic supply aside, refining estimates of peak black duck populations will have the greatest impact in reducing uncertainty in bioenergetic models. Managers should continue pairing frequent, inexpensive ground surveys with occasional aerial surveys to measure waterfowl abundance.
Quantitative biologists from JVs whose task it is to step-down continental population objectives to regional levels may view our results with a sense of dismay: if the uncertainty in carrying capacity models is so great even in a well-known, well-constrained system, what hope is there of developing reliable regional models? We believe the value of carrying capacity models is best viewed in relative terms. One strategy used by wintering JVs to increase carrying capacity is to provide more hectares of the habitats that on average (or range) provide more energy per hectare, and the currency used in judging the value of these investments is often “duck-use days.” Our model suggests we are far from accurately estimating duck-use days, but the relative value of habitat management or restoration actions as they affect energy supply may still be evaluated in this currency. Meanwhile, other potential metrics such as the number of ducks present during a given interval, body condition, overwinter survival, or hunter days afield may provide alternative or complementary measures of the success of waterfowl management actions, and merit further consideration.
Supplemental Material
Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.
Table S1. Habitat-specific core sampling data collected from 2010 to 2014 at Edwin B. Forsythe National Wildlife Refuge, New Jersey, USA.
Found at DOI: http://dx.doi.org/10.3996/122016-JFWM-089.S1 (11 KB XLSX),
Table S2. Estimated energy potentially available to waterfowl in various habitats at Edwin B. Forsythe National Wildlife Refuge, New Jersey, USA.
Found at DOI: http://dx.doi.org/10.3996/122016-JFWM-089.S2 (13 KB XLSX),
Table S3. Monthly ground-survey abundance indices (2005–2014), scaled population-level data, and depletion calculations for waterfowl wintering at Edwin B. Forsythe National Wildlife Refuge, New Jersey, USA.
Found at DOI: http://dx.doi.org/10.3996/122016-JFWM-089.S3 (30 KB XLSX),
Reference S1. Central Valley Joint Venture. 2006. Central Valley Joint Venture implementation plan—conserving bird habitat. Sacramento, California: U.S. Fish and Wildlife Service.
Found at DOI: http://dx.doi.org/10.3996/122016-JFWM-089.S4 (16 MB PDF).
Reference S2. Esslinger CG, Wilson BC. 2001. North American waterfowl management plan, Gulf Coast Joint Venture: Chenier Plain initiative. Albuquerque, New Mexico: North American Waterfowl Management Plan.
Found at DOI: http://dx.doi.org/10.3996/122016-JFWM-089.S5 (10 MB PDF).
Acknowledgments
Funding for this study was provided by the U.S. Fish and Wildlife Service Edwin B. Forsythe National Wildlife Refuge, Office of Migratory Bird Management, Populations Survey Branch; New Jersey Division of Fish and Wildlife Federal Aid in Wildlife Restoration P-R Grant W-68-R; and University of Delaware. We also thank O. Jones, K. Duren, C. Meyer, J. Wortham, M. Koneff, E. Silverman, M. Conroy, J. Heise, and countless field and lab technicians for collecting, sorting, and organizing the data used in this manuscript. We thank several anonymous reviewers and D. Haukos for their comments on earlier drafts of this manuscript. The authors have no conflict of interest to declare.
Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
References
Author notes
Citation: Ringelman KM, Williams CK, Castelli PM, Sieges ML, Longenecker RA, Nichols TC, Earsom SD. 2017. Estimating waterfowl carrying capacity at local scales: a case study from Edwin B. Forsythe National Wildlife Refuge, New Jersey. Journal of Fish and Wildlife Management 8(1):209-218; e1944-687X. doi:10.3996/122016-JFWM-089
The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.