Although monitoring data for sea ducks (Tribe Mergini) are limited, current evidence suggests that four of the most common species wintering along the eastern coast of the United States—long-tailed duck Clangula hyemalis, white-winged scoter Melanitta fusca, surf scoter Melanitta perspicillata, and black scoter Melanitta americana—may be declining, while the status of American common eider Somateria mollissima dresseri is uncertain. The apparent negative trends, combined with the fact that sea duck life histories are among the most poorly documented of North American waterfowl, have led to concerns for these species and questions about the impacts of human activities, such as hunting, as well as catastrophic events and environmental change. During winter, thousands of sea ducks are found along the U.S. Atlantic coast, where they may be affected by proposed wind-power development, changes to marine traffic, aquaculture practices, sand mining, and other coastal development. Possible impacts are difficult to quantify because traditional winter waterfowl surveys do not cover many of the marine habitats used by sea ducks. Thus, the U.S. Fish and Wildlife Service conducted an experimental survey of sea ducks from 2008 to 2011 to characterize their winter distributions along the U.S. Atlantic coast. Each year, data were collected on 11 species of sea ducks on >200 transects, stretching from Maine to Florida. In this paper, we describe distribution of common eider, long-tailed duck, white-winged scoter, surf scoter, and black scoter. Densities of the two species with the most northerly distribution, white-winged scoter and common eider, were highest near Cape Cod and Nantucket. Long-tailed duck was most abundant around Cape Cod, Nantucket Shoals, and in Chesapeake Bay. Surf scoter also concentrated within Chesapeake Bay; however, they were additionally found in high densities in Delaware Bay, and along the Maryland–Delaware outer coast. Black scoter, the most widely distributed species, occurred at high densities along the South Carolina coast and the mouth of Chesapeake Bay. Spatial patterns of high-density transects were consistent among years for all species except black scoter, which exhibited the most interannual variation in distribution. The distance from land, depth, and bottom slope where flocks were observed varied among species and regions, with a median distance of 3.8 km from land along the coastal transects and 75% of flocks observed over depths of <16 m. Common eider and long-tailed duck were observed closer to shore and over steeper ocean bottoms than were the three scoter species. Our results represent the first large-scale quantitative description of winter sea duck distribution along the U.S. Atlantic coast, and should guide the development of sea duck monitoring programs and aid the assessment of potential impacts of ongoing and proposed offshore development.
The 15 species of North American sea ducks (Tribe Mergini) are the least understood group of waterfowl protected under the Migratory Bird Treaty Act (Bellrose 1980; Goudie et al. 1994; Sea Duck Joint Venture 2003; Sea Duck Joint Venture Management Board 2008). Monitoring data for these species are of limited scope and uncertain quality, and little is known about their distributions and habitat preferences (Zipkin et al. 2010). Available data suggest that 4 of the 12 species that winter along the east coast of the United States are declining—long-tailed duck Clangula hyemalis, white-winged scoter Melanitta fusca, surf scoter Melanitta perspicillata, and black scoter Melanitta americana—but causes of the declines are not known (Caithamer et al. 2000; Sea Duck Joint Venture 2003, 2007). Populations of American common eider Somateria mollissima dresseri, an iconic species and the most commonly hunted sea duck in eastern Canada and the United States, are thought to be stable, but the species may be declining in the southern portion of its range (Environment Canada Working Group 2011; North American Waterfowl Management Plan Committee 2012).
Sea ducks exhibit delayed sexual maturity and have long life spans and low annual recruitment (Caithamer et al. 2000; Sea Duck Joint Venture 2003), making their population dynamics sensitive to adult survival and slow to recover from catastrophic events, environmental degradation, and anthropogenic impacts (Di Giulio and Scanlon 1984; Ohlendorf and Fleming 1988; Piatt et al. 1990; Guillemette and Larsen 2002; Larsen and Guillemette 2007). Sea duck populations are harvested (Krementz et al. 1996, 1997; Caithamer et al. 2000) and human activity is expanding in both their northern breeding and coastal wintering areas. Along the U.S. Atlantic coast, an important sea duck wintering area, energy production (e.g., proposed wind farms), coastal development, sand mining, shipping, and aquaculture all have the potential to alter sea duck habitats and affect migrating and wintering birds.
Existing information on the distribution of sea ducks along the Atlantic coast comes from previous U.S. Fish and Wildlife Service (USFWS) surveys of near-shore habitats (i.e., 0.25 nautical mi, or 0.46 km, from coast; Zipkin et al. 2010). These data do not lend themselves to estimation of intraannual spatial or temporal variation, or provide information for offshore areas, where many sea ducks are observed to aggregate and where wind-energy development is proposed.
The USFWS conducted several experimental offshore surveys of limited scope in the mid-Atlantic region during the 1990s and early 2000s; these efforts highlighted concerns about reliance on near-shore surveys to monitor sea duck populations (D.J. Forsell, USFWS, unpublished data). To improve our understanding of offshore sea duck distribution and estimate population sizes of wintering sea ducks, the USFWS initiated the Atlantic Coast Wintering Sea Duck survey in 2008. This experimental survey was aimed at developing an operational survey for all sea duck species wintering along the full extent of the U.S. Atlantic coast.
In this paper, we analyze data from the Atlantic Coast Wintering Sea Duck survey to characterize the winter distribution of common eider, long-tailed duck, white-winged scoter, surf scoter, and black scoter along the U.S. Atlantic coast, and to identify the relationships between sea duck occurrence and distance from land, water depth, and bottom slope. A quantitative description of species-specific distribution will aid the design of an efficient survey for abundance estimation by guiding the definition of survey boundaries and strata. Identifying associations among sea duck occurrence and physical features such as distance to land and water depth will, in turn, provide a framework for understanding observed distribution and for setting survey boundaries, and is a first step toward defining research efforts necessary to identify offshore resources that are critical for sea ducks.
The Atlantic Coast Wintering Sea Duck survey was flown between late January and early March in 2008–2011. The survey design varied among years with the 2008 design differing most substantially from the other years (Table 1; see Silverman et al. 2010, 2011, 2012a for more detailed description of study design). The design was changed in 2009 based on lessons learned in 2008 and consisted of east–west transects spaced at 5′ intervals of latitude, except in the northern part of the Chesapeake Bay where transects were spaced at 10′ intervals. These transects extended east from the coastline to the longer of two distances: 14.8 km or the distance to 16-m depth (hereafter referred to as off-coast transects). The 16-m depth boundary was based on previous offshore observation (D.J. Forsell, personal observation), and the 14.8-km boundary ensures that most transects run much deeper, unless coastal bathymetry is shallow for many miles offshore. Transects were also located at 5′ intervals over Nantucket Shoals, and across major coastal bays and sounds (e.g., Chesapeake Bay, Delaware Bay, Cape Cod, and Long Island Sound, see Figure 1). In 2009–2011, as feasible, crews flew transects from north to south, then returned north and replicated every other line, flying south. Direct comparison of the 2008 data with the 2009–2011 is problematic, because the 2008 survey differed in both its northern and offshore extent, and lacked replication; therefore, 2008 analyses are not included here.
Surveys were conducted using USFWS fixed-winged aircraft flown at 204 km/h and 70 m altitude. Ten crews flew Cessna 185/206 aircraft, one flew the Partenavia P68C (standard instrument panel), and one flew the Quest Kodiak 100. The Partenavia P68C has been used in the past for USFWS aerial waterfowl surveys and the cockpit field of view is similar to that of the Cessna 185/206 (M.D. Koneff, personal observation; see also Koneff et al. 2008). During surveys, an observer and pilot-observer counted all sea ducks and other aquatic birds along transects, from the closest observable distance to the center line (∼50 m) to 200 m on their side of the aircraft.
Due to the vagaries of field operations, transects and replicates differed somewhat among years. We only included transects that were sampled consistently among years (e.g., same latitude and similar area surveyed in ≥2 y; Figure 1). We present results from 253 unique transects and 151 unique replicated transects, representing 170 off-coast transects placed according to the 14.8-km distance–16-m-depth design rule, 9 transects placed to cover Nantucket Shoals, 3 transects covering shoals off the North Carolina coast, and 71 transects spanning shore-to-shore across bays and sounds. Table 1 includes the total number of transects and replicates by year.
Sea duck observations
Pilot-observer and observer records were entered directly into laptop computers as sound files using a hardware–software recording platform developed by the USFWS (Hodges 2003). Each computer was linked to the aircraft global positioning system (GPS), enabling simultaneous recording of both observations and their coordinates in linked sound and ASCII (American Standard Code for Information Interchange) files. Locations recorded for each observation do not represent the location of the birds; rather, they correspond to the location of the aircraft when an observation is recorded. Individual records include the 1) location and time, 2) species, and 3) number of birds seen at the location. In this paper, we refer to the individual records as flocks and the count associated with the record as the flock size, although flock boundaries are not well-defined and the counts are restricted to the area within the transect boundary. According to this definition, flocks can consist of a single bird. The three scoter species can be difficult to identify in the field, leading to a large number of scoters recorded only to genus (Melanitta spp.). In addition to recording observations, the software records aircraft location at least every 15 s while surveying to form a track file or flight path (Hodges 2003).
Following each survey, observers transcribed the observation data from sound files to ASCII files and simultaneously attributed each observation record with appropriate geographic coordinates using another in-house software application (Hodges 2003). We postprocessed the observation files and flight paths using scripts written in the R statistical computing environment (R Development Core Team 2011) and ArcGIS 10.0 (Environmental Systems Research Institute 2011). Processing included deleting bad GPS fixes and observations located far from transects (>1.85 km) or on land (>0.93 km inland), as well as correcting records (e.g., mistyped species codes, missing ending or starting locations, etc.).
Physical features and tides
We measured distance from land as the Euclidean distance between sea duck observations and the nearest edge of the Atlantic coastline obtained from the National Oceanic and Atmospheric Administration's Coastal Geospatial Data Project (National Oceanic and Atmospheric Administration 2010) using ArcGIS 10.0. We extracted water depths from the U.S. Geological Survey digital elevation model of the Atlantic Ocean obtained from the National Elevation Dataset (Gesch et al. 2002; Gesch 2007). We also used the digital elevation model to create an interpolated surface of slope (i.e., the steepness of the ocean bottom based on changes in water depths, measured in degrees) for the Atlantic Ocean.
To characterize the tidal cycle for each sea duck observation, we consulted the National Oceanic and Atmospheric Administration's Center for Operational Oceanographic Products and Services (National Oceanic and Atmospheric Administration 2012). First, we divided the off-coast areas of the survey into 14 regions, each consisting of 20 contiguous transects, and defined separate regions for transects in bays and over Nantucket Shoals. We then estimated the mean time of low tide by region and survey date using the three coastal reporting stations that were closest to the region's northern boundary, southern boundary, and center point. We subtracted the mean low tide time from the time of each sea duck record in the region to estimate time since low tide.
Spatial patterns in species density
We calculated sea duck densities for each transect by dividing total number of birds observed for each species by area surveyed. Area surveyed is the distance surveyed by each observer, multiplied by 150 m (the transect half-width [200 m] less the area under the plane that is not visible to observers [50 m]). For transects replicated in a given year, we averaged densities across replicates by summing all birds counted on both replicates and dividing by the total replicated area. For species-specific scoter density calculations, unidentified scoters on individual transects were apportioned among the three species based on the composition of the identified scoters on the transect and its northern and southern neighbors in the survey year, extending north and south if none of the neighboring transects included speciated scoter records. Species-specific proportions on neighboring transects are positively correlated. This calculation assumes that the probability of identifying scoters does not differ by species and deserves further investigation. However, the survey crews had substantial experience with aerial waterfowl observation and sea ducks; moreover, using only scoters identified to species did not alter our results.
To define distinct coastal regions of high and low density for each of the five species (hereafter, density regions), we applied a spatially constrained grouping algorithm available in ArcGIS 10.1 Spatial Statistics toolkit (Environmental Systems Research Institute 2012) to the 3-y average densities for the 253 transects (area weighted average, as above, with 2-y averages for some transects). The algorithm, called SKATER for Spatial ‘K’luster Analysis by Tree Edge Removal (Assunção et al. 2006), identifies groups using a minimum spanning tree, based on 1) a spatial constraint matrix constructed from the transect locations, and 2) transect densities, which represent the tree's node values. We defined a transect's neighbors to be those directly to the north and south, as well as any contiguous transect to the east or west. Transects located to the east or west that were not contiguous, but within 55–93 km over water or within 18.5–55 km over land, were treated as half-neighbors and weighted by 0.5, instead of 1 in the spatial constraint matrix. The algorithm calculates an r2 value (i.e., the sum of squares for the group means divided by total sum of squares), and pseudo-F statistic for 2–15 groups. The pseudo-F represents the test statistic for a 1-way analysis of variance (ANOVA) comparing group means, and the largest F-value indicates the number of groups with the lowest P-value, allowing selection of an “optimal” number of groups. If the pseudo-F statistic increased monotonically from 2 to 15 groups, we also ran SKATER with >15 groups specified, to explore the possibility of a larger optimal number of groups.
The method, as implemented, does not allow for specification of a minimum group size, so, in some instances, the optimal grouping resulted in small clusters. This problem is exacerbated by the high degree of right-skew present in the count data and can result in optimal groupings consisting of one large cluster and many 1–2-transect clusters. We therefore ran the clustering algorithm on log-transformed average annual densities. To handle zero values, counts were adjusted prior to transformation by adding 0.25, which is one half of the minimum number of birds needed to record a positive count. With this correction, we assume that if a bird is more than halfway within the transect, the bird would be counted and, if less than halfway within, the bird would not be counted, where a value of ≥0.5 is rounded to one bird and a value of <0.5 is rounded to zero birds. This correction resulted in variable densities for transects of different lengths with zero birds, where longer transects with zero birds have smaller transformed densities than do shorter transects with zero birds.
For all species except black scoter, we identified a southern range limit and ran the spatial clustering only on data from transects north of this latitude. When low- or no-density southern transects are included, the clustering algorithm is dominated by the range edge and we learn less about the spatial patterns in the core of the species' winter ranges. We set the southern limit for the clustering analysis at the latitude below which <1% of the all birds of the species were counted for common eider, long-tailed duck, and surf scoter. Because a few white-winged scoter were sometimes seen quite far south of their main wintering areas in the company of surf and black scoter, we used a cutoff of 5% for white-winged scoter: running the algorithm with a 1% cutoff for white-winged scoter resulted in a cluster boundary at the 5% cut-off latitude, and a huge, predominately zero-valued southern region. In addition to identifying a southern range limit, we also established an eastern range limit, or distance from coast, in order to exclude unoccupied areas.
To provide a more detailed picture of density variation within and among the 3-y average density regions, we ran the transformed 3-y average densities through SKATER with no spatial constraints on cluster membership in order to classify transects into six density categories by species (1 = low to 6 = high) and overlaid these results on the density regions. The unconstrained SKATER algorithm is equivalent to k-means clustering (Hartigan and Wong 1979; Assunção et al. 2006).
To explore annual variation, we created annual density regions for each species and compared these results with the regions based on the 3-y average. We also ran the unconstrained clustering on all 3 y of densities (unaveraged) simultaneously to define six density categories; we quantified interannual variation in density for each transect by subtracting its minimum density category value from its maximum. For example, if a given transect was assigned to cluster 6 (high density) in 2009 but to cluster 1 (low density) in 2011, the difference of 5 would indicate a large interannual change in density. We overlaid these density change values, which range from 0 to 5, on the 3-y average density regions to identify the regions and species with more or less variable distributions.
Occurrence relative to physical features and tides
We measured distance from land, water depth, and bottom slope for each sea duck flock, with the goal of contrasting characteristics of sea duck locations with the surrounding physical environment, and comparing patterns among species and years. These analyses are complicated by the coastal bathymetry, which changes with latitude as the shallow, gradual depth gradient in the south transitions to a steep, sharp drop in the north: observed differences in the physical characteristics of sea duck locations can result either because the birds are at different latitudes or because they are in different locations at the same latitude. To explore how the features where birds are found differ by year within and among similar physical environments, we first classified the off-coast transects by their bathymetry, and then compared the characteristics of flock locations by species and year in the resulting transect types. These analyses describe distributional differences with respect to distance, depth, and slope in areas with broadly similar bathymetry.
We applied the unconstrained SKATER algorithm to three variables that characterize the transect profile: distance to 16-m depth, start depth at 0.46 km from the coast (or 0.25 nautical mi, the point where survey protocol calls for the near-shore start of each transect), and gradient (end depth – start depth / distance flown). The clustering procedure resulted in definition of three off-coast transect types—steep, moderate, and flat bottom gradients—corresponding roughly to the northern, mid-, and southern coast (Figure 2). The three off-coast transect types, along with the 1) Nantucket Shoals transects and 2) bay or sound transects, define five transect types for three analyses of species-specific associations with the physical features, and annual changes in these associations.
First, we compared the distribution of the three variables—distance from land, water depth, and bottom slope—by species and transect type, with the characteristics of the surveyed transects: we determined the proportion of flocks within each transect type that were 1) > or <7.4 km from the land, 2) over bottom depths of 0–6 m, 6–12 m, 12–18 m, or >18 m, and 3) over bottom slopes of <0.1°, 0.1–0.5°, 0.5–1.0°, and >1.0°. We then calculated the proportion of surveyed area within these categories for each transect type. When the range of a species did not include all transects of a given type, we used only transects that overlapped the species range to calculate surveyed area. Thus, surveyed areas, and the proportion surveyed in each distance, depth, and slope category, differ somewhat by species within transect type. We highlight categories where the proportion of flocks observed is greater than the proportion expected based on the available surveyed area, testing for significant patterns with simple chi-square goodness-of-fit tests. We present results based on an overall significance level of 0.05, applying the Bonferroni correction to adjust for the multiple transect-type tests for each species × physical variable.
Second, we compared the characteristics of the flock locations by species for each of the three variables and five transect types using 1-way ANOVA, followed by Tukey's multiple-comparisons test with α = 0.05. Diagnostic QQ- and residual plots showed residuals that were generally somewhat right-skewed, but results were not sensitive to outlying values.
Finally, for the off-coast transects, we explored annual patterns in the distance, depth, and bottom slope associated with flock locations, using 2-way ANOVA with year and transect type as factors. Because none of the species were found in all three off-coast transect types in sufficient numbers, these comparisons were restricted to 1) steep and moderate gradient regions for common eider and long-tailed duck, 2) moderate and flat gradient regions for black scoter, and 3) moderate gradient regions for surf and white-winged scoter (i.e., 1-way ANOVA with year as the factor was used for these two species). We checked ANOVA assumptions using standard diagnostic plots, including QQ- and residual plots, and, in cases where there appeared to be influential points, ran the models with and without these observations included. We tested for significant interactions between year and transect type by comparing the residual sum of squares for the full model with the main-effects-only model using an F-test and present results for the main-effects-only models when the interactions were not significant (P > 0.05).
To explore whether variability in flock location on the off-coast transects might be explained by the tide cycle and time of observation, we estimated the effect of time and tide on distance from coast for each species using 2-way ANOVA. Time of day was treated as a categorical variable, with two levels: morning (0001–1159 hours) and afternoon (1200–2400 hours). Results for finer divisions of the day were the same as for two intervals; we also checked for patterns by plotting distance from coast vs. time of day. We similarly treated time since low tide as a categorical variable with three levels: flocks recorded within 1.5 h of low tide were classified as occurring at low tide, those within 1.5 h of high tide were classified as occurring at high tide; flocks observed between 1.5 and 4.5, or 7.5 and 10.5 h after low tide were classified as occurring during rising or falling tide. We took this approach because we anticipated the greatest differences in location between low and high tide and did not want to limit the model to fitting a linear relationship between tide and location, and because the regional tide times were inexact. We also checked our results by plotting distance against the original estimated time since low tide. We tested for significant effects by comparing the residual sum of squares for the previous year and transect type model with the model with tide × time effects included using an F-test at α = 0.05.
The data used for all analyses, and the spatial neighborhood matrix, are available in Data A1, Archived Material in Dryad, http://dx.doi.org/10.5061/dryad.m9t12.
During the four survey winters, crews observed 11 species of sea ducks: common eider, long-tailed duck, surf scoter, white-winged scoter, black scoter, common goldeneye Bucephala clangula, bufflehead Bucephala albeola, common merganser Mergus merganser, red-breasted merganser Mergus serrator, hooded merganser Lophodytes cucullatus, and harlequin duck Histrionicus histrionicus. For the five focal species, 121,882 birds were observed comprising 9,286 flocks with flock sizes in the transect area ranging from 1 to 5,000 individuals. The composition by species was 26% long-tailed duck, 26% common eider, 23% black scoter, 11% surf scoter, 4% white-winged scoter, and 10% unidentified scoters.
The southern range boundaries were 40°46′N latitude for common eider, 40°21′N latitude for white-winged scoter, and 35°06′N latitude for long-tailed duck and surf scoter. The eastern range limit for each transect (except the nine transects over Nantucket Shoals) was 28 km from its western endpoint, because no sea ducks were detected farther from the coastline. Using the 3-y average densities for 2009–2011 within these boundaries, we identified 12 spatially distinct density regions for common eider (r2 = 0.76; Figure 3A), 2 for white-winged scoter (r2 = 0.36; Figure 3B), 4 for long-tailed duck (r2 = 0.42; Figure 3C), 3 for surf scoter (r2 = 0.24; Figure 3D), and 15 for black scoter (r2 = 0.50; Figure 4). Because the optimal number of groups for black scoter was 15, we also ran SKATER with the number of black scoter clusters set to 16–25; the procedure resulted in fragmented, small groupings that did not provide any additional illumination into black scoter distribution.
Although there was some variability in the annual density regions, the optimal number and location of the clusters were similar among years for all five species. For common eider, high densities were observed around Cape Cod and Maine coastal islands (Figure 5A). Common eider densities were consistent across years (Figures 5B and 6A–C), with 42% of transects shifting less than two density categories (50% of the occupied transects; Table 2). Common eider had a compact distribution along the U.S. coast and only 17% of the surveyed transects in their range were unoccupied in all surveyed years (Table 2). Long-tailed duck was observed at highest densities around Cape Cod and Nantucket Shoals, followed by Chesapeake Bay, Long Island Sound, and the Maine coast (Figure 7A). Their densities were second to common eider as the most consistent between years, with 28% of 192 transects (or 40% of occupied transects) shifting one or fewer density categories (Table 2; Figures 7B and 8A–C).
White-winged scoter, the least abundant of the five species, had highest densities in Cape Cod Bay, over Nantucket Shoals, and at the eastern end of Long Island (Figure 5C). Surf scoter was found at high densities within Chesapeake and Delaware Bays, and along the Maryland–Delaware coast with smaller areas of high density around Nantucket Island and the southern end of Pamlico Sound (Figure 7C). Black scoter had a wide range, with high densities along the South Carolina coast, in Pamlico Sound, at the mouth of Chesapeake Bay, and around Cape Cod, and variably high densities around Delaware Bay, interspersed with low densities between these regions (Figure 9).
The three scoter species had more variable annual densities than did common eider and long-tailed duck (Figures 5D, 6D–F, 7D, 8D–F, and 9), with only 13–16% of transects shifting one or fewer density categories (or 22–29% of occupied transects; Table 2). Fourteen to 21% of occupied transects shifted four or more categories (calculated from Table 2, removing unoccupied transects). Not unexpectedly, densities were more consistent in low-density regions (Figures 5D and 7D). Densities of black scoter were the most variable among years (Figure 9 and Table 2), shifting from the mouth of the Chesapeake Bay and Pamlico Sound in 2009, south to the South Carolina coast and north to Cape Cod in 2010–2011. In 2008, black scoter distribution was intermediate between these patterns, with high numbers in the mouth of the Chesapeake Bay, in Delaware Bay, and in Cape Cod, as well as along the southern Georgia coast (results not shown). The three scoter species also had a higher percentage of unoccupied transects than did common eider and long-tailed duck (33–44%; Table 2). This was likely a consequence of their lower overall abundance relative to the latitudinal extent of each species' winter range.
Table 3 presents the mean distance to 16-m depth, mean western start depth, and mean gradient for the three off-coast transect types, Nantucket Shoals, and the bays and sounds. Although the three profile types were interspersed throughout the survey area, the steep profile was most common along the Maine coast and the outer coast of Cape Cod, with transects starting at >16-m depth (i.e., the cutoff depth for our survey design), the moderate profile occurred mainly south of Cape Cod to the southern boundary of North Carolina, and the flat profile predominated south of North Carolina (Figure 2).
Sea duck flocks were not distributed randomly along transects with respect to the covariates (Table 4; χ2 test unadjusted P-values were < 0.001 in 42 cases, between 0.001 and 0.050 in 7 cases, and > 0.05 in 8 cases). When the last category (>18-m depth) was excluded, common eider, long-tailed duck, and white-winged scoter flocks were distributed randomly with respect to depth along the steep profile transects (P = 0.36, 0.62, and 0.77, respectively), but this result did not hold for the other transect types.
In our comparisons among species, the results for distance from land were the strongest and most consistent across transect types: common eider was found closest to shore, followed by long-tailed duck (Table 5; see also Table 4). Both species were closer to shore on steep profile transects than on moderate profile transects (Table 5). We also observed common eider over the steepest bottom slopes, followed by long-tailed duck. Differences among the species in depth were less significant and varied by transect type: long-tailed duck was found in deeper water in areas with steep profiles, while this was not the case for common eider. The locations of common eider and long-tailed duck flocks, as measured by distance from land and depth, were more variable than those of the scoters (Table 5).
The scoter species were significantly farther from shore than were common eider and long-tailed duck, with no consistent differences among the three species (Tables 4 and 5). Like common eider, white-winged scoter flocks were farther from shore, and in somewhat deeper waters, on moderate profile transects compared with steep profile transects. Both surf and black scoter were in deeper water on moderate profile transects than flat profile, but surf scoter was farther from shore in flat areas, while black scoter was closer. Overall, black scoter was found at the shallowest depths, and white-winged scoter at the deepest. Scoters were also found, without distinction, over the flattest ocean bottom slopes. The bottom slopes underneath black scoter and long-tailed duck locations were the most variable of the five species (bold font, Table 5).
We found significant yearly differences in distance from land for all five species, and the annual effect differed by transect type only for black scoter (common eider: F3,225 = 9.42, P < 0.001; long-tailed duck: F3,283 = 3.55, P = 0.02; white-winged scoter F2,82 = 16.07, P < 0.001; surf scoter F2,150 = 6.47, P = 0.002; black scoter F5,540 = 16.61, P < 0.001). White-winged and surf scoter were closer to shore in 2010, as was black scoter on moderate profile transects (Table 6). Common eider and long-tailed duck shifted approximately 1 km offshore in 2011, and the shift was consistent across transect types, while white-winged scoter moved substantially farther. (Note that the common eider and long-tailed duck shifts in 2011 were smaller when several influential points were removed: common eider effect size = 0.5 km, SE = 0.2, P = 0.02; long-tailed duck effect size = 0.6 km, SE = 0.4, P = 0.12.) In contrast, in 2011 black scoter moved closer to the coast on flat profile transects (Table 6).
As flock distance to land increased or decreased, so did the locations' depth, but the annual effects were weaker (Table 6). Surf scoter was observed at shallower depths in 2010. In 2011, white-winged scoter was found at deeper locations, as was common eider on steep profile transects and long-tailed duck on all transect types (but only if one influential 2010 point was excluded). Common eider was also found in significantly deeper water on moderate profile transects in 2009 (5.0 m, SE = 2.3, P = 0.03, compared with mean 2010–2011 moderate profile locations). Annual changes in locations relative to bottom slope were weaker still; this measurement is substantially more right-skewed than distance or depth, and model results were sensitive to influential points.
The effects of time and tide on distance from coast differ by species. Long-tailed duck and surf scoter were significantly closer to shore at low tide (long-tailed duck low-tide effect = −1.4 km, SE = 0.5, P = 0.003; F2,283 = 4.10, P = 0.02; surf scoter low-tide effect = −2.5 km, SE = 0.8, P = 0.003; F3,150 = 5.90, P < 0.001). White-winged scoter, in contrast, was closer to shore in the morning during both high and low tides (F5,82 = 3.84, P = 0.004), while black scoter was farther from shore in the morning (morning effect = 1.5 km, SE = 0.4, P < 0.001; F5,542 = 8.68, P < 0.001). There was no time or tide effect on common eider's distance from land (F5,225 = 1.25, P = 0.29).
The 2008–2011 experimental sea duck survey conducted by the USFWS generated the largest and most comprehensive data set available to characterize the winter distribution of common eider, long-tailed duck, white-winged scoter, surf scoter, and black scoter along the U.S. Atlantic coast. Our analyses provide the first quantitative description of important coastal regions, variation in density and abundance, and the characteristics of locations where sea ducks occur. These five species of sea ducks have overlapping distributions that concentrate in several critical regions, most particularly around Cape Cod, Nantucket Shoals, and the mouth of Chesapeake Bay (Figures 3 and 4). Nantucket Shoals is an especially important wintering area for long-tailed duck, with 50% of all observations for this species occurring on these nine transects; large numbers of white-winged scoter can also be found over the Shoals (e.g., 29% of all white-winged scoter observations in 2010). Other regions of significance for individual species include the Maine coastal bays and islands for common eider, Chesapeake Bay for long-tailed duck, Long Island Sound for white-winged scoter, the Maryland coast and Delaware Bay for surf scoter, and Pamlico Sound and South Carolina coast for black scoter.
Despite substantial inter- and intra-annual variation, high-density transects and regions were similar between years. This result supports conclusions from the previous USFWS coastal surveys (Zipkin et al. 2010), although the order of species with respect to the strength of annual consistency differed. The differences may reflect the limitations of near-shore surveys in characterizing distribution. Black scoter shifted the most between high-density regions, with large numbers in Delaware Bay in 2008, at the mouth of Chesapeake Bay in 2009 (and 2012, M.D. Koneff and W.E. Rhodes, USFWS, personal communication), and the southern coast in 2010 and 2011. It is not unexpected that the species with the largest winter range would exhibit the most annual variation in distribution, but black scoter occurrence and abundance along the southern coast was particularly variable (Figure 9; only 8% of southern transects shifted one or fewer density categories). We have additional, more intensive survey data for the South Carolina and Georgia coasts from 2011 to 2012, which, in combination with information from black scoter outfitted with satellite transmitters (Loring 2012; Sea Duck Joint Venture 2012), may provide further insight into this species' distribution along the southern coast.
Although coastal regions of high and low density were broadly similar between years, all five species exhibited significant interannual shifts in their distance from shore and, to a lesser extent, their association with sea bottom depth. Previous analyses suggested that near-shore abundance increases for scoters, and decreases for common eider and long-tailed duck, when the North Atlantic Oscillation (a measure of climate over the Atlantic Ocean) is negative (Zipkin et al. 2010). Winter 2009–2010 recorded the lowest North Atlantic Oscillation since measurements became available in 1825 (−2.4 [Osborn 2011; Taws et al. 2011]), and scoter species were significantly closer inshore along the mid-coast. The pattern for 2011 is more complicated and ambiguous: December 2010 saw the second lowest North Atlantic Oscillation on record (−4.62, Taws et al. 2011), which might explain the common eider and long-tailed duck shift farther offshore, but the January–February 2011 average North Atlantic Oscillation was positive (0.71 [Climatic Research Unit, University of East Anglia, Norwich, UK, http://www.cru.uea.ac.uk/∼timo/datapages/naoi.htm]), which could equally explain why scoters had moved back offshore. Detailed analyses that consider regional and monthly variation in winter conditions are needed to understand how climate may be affecting sea duck distribution.
Depth increases with distance from land; therefore, we expect some correspondence between annual distance and depth changes. The weaker annual depth effect and lack of significant annual changes in bottom slope, however, suggest that movements on and offshore may be moderated by selection for specific depth and slope characteristics (or other features with which depth and slope covary). For all five species, the characteristics of flock locations were not random with respect to the available bathymetry. Some of the patterns may result because transects were extended beyond the eastern range of the species. For example, the depth distributions for species found commonly along the steep profile transects—common eider, long-tailed duck, and white-winged scoter—were similar to the surveyed depth profile when the deepest offshore category is excluded. However, clear species-specific differences in distance, depth, and slope in areas of similar bathymetry suggest that the nonrandom associations represent meaningful patterns. Common eider was found more often in steeper areas nearer to shore, scoters in flatter areas farther from land, and long-tailed duck intermediate between these two. Zipkin et al. (2010) found similar patterns, with common eider and long-tailed duck more abundant in near-shore survey segments with steeper slopes and scoter more abundant in flat-bottomed segments. These relationships likely reflect associations with preferred prey and substrate: common eider has been shown to concentrate over shallow reefs (Guillemette et al. 1993), surf and black scoter over sandy substrates (Stott and Olson 1973; Loring 2012), with long-tailed duck exhibiting the most varied habitat use (Stott and Olson 1973).
The relationships between time of observation, tidal cycle, and sea duck location on the off-coast transects were inconsistent among the five species. We might expect sea ducks to move offshore to deeper, otherwise inaccessible, locations during low tides, but, for long-tailed duck and surf scoter, there was evidence that birds were closer to shore during low tide. White-winged scoter and black scoter exhibited opposite patterns with respect to time of day, and no clear tide effects. At the scale of this survey, tide cycle and time do not appear to explain the distributions we found, nor suggest changes to survey protocols to account for daily movements. Overall, the causes of species-specific patterns of abundance associated with distance, depth, and slope; and annual changes in these patterns; remain to be explained. A full understanding will require additional survey data, as well as the collection of other covariate measurements and finer scale observations.
The results of the current analysis are essential to finalizing the design of an operational survey that can provide efficient winter population estimates and guide offshore development planning. We have used the density regions and eastern range boundary identified in these analyses to define 10 preliminary survey strata, and we explored stratified sample allocation strategies within these strata based on the densities of the individual species, and all five species combined (Silverman et al. 2012b). The next step is to use the distance and depth patterns described here to adjust the eastern survey boundary based on coastal bathymetry and to define near- and offshore strata. Because sea ducks show clear annual changes in their distribution relative to shore, shore-based and offshore surveys that do not account for these shifts are likely to confound real annual variation with estimation errors.
In addition to improving the definition of survey and stratum boundaries, an understanding of the relationship between the species and their physical environment should aid in building efficient models to estimate winter abundance. The density patterns we observed suggest that any survey focused on abundance estimation for one of the five species will necessarily sacrifice information about the others, while an omnibus survey is unlikely to achieve the level of precision possible from more targeted monitoring efforts. Further consideration of annual changes in distribution and association with physical covariates, however, may assist in developing an efficient multispecies survey to estimate wintering population sizes. We view our preliminary assessment of associations among sea duck occurrence and the physical covariates examined as a first step toward specifying and testing hypotheses about critical resources exploited by sea ducks and modeling these relationships to improve survey design, estimation, and predictions about the effect of environmental change on these species.
Regulation of waterfowl harvest requires estimates of population abundance or harvest rate. Current breeding surveys are poorly timed for sea ducks and do not cover the distributions of common eider, long-tailed duck, white-winged, surf, and black scoter, so population estimates are not available, while banding programs are insufficient to estimate harvest rates. In the absence of a feasible and cost-effective breeding survey, we are developing estimates using the winter survey data (Silverman et al. 2012b). Human activities are also increasing in sea duck wintering habitat along the U.S. Atlantic coast, prompting interest in planning for sea duck habitat conservation. The characterization of winter sea duck distribution is a critical component of these efforts, and will allow for the identification of diurnal concentration areas. Data from this survey are being used by USFWS partners as part of offshore development planning, including evaluation of the potential effects of sand mining and wind development.
The USFWS is coordinating with other agencies responsible for management of the marine environment to determine how an operational survey might assist decision-making, and in considering the needs of all stakeholders in the determination of the operational survey design. The available experimental data, and a future operational survey, could also support critical research by informing development, and subsequent testing, of hypotheses about factors affecting sea duck wintering distributions. Understanding mechanistic relationships would help refine survey design and improve model-based estimation procedures, and will be required to predict the effects of development activities or other environmental influences on sea duck habitats and, ultimately, sea duck populations. Advancing this management-oriented research agenda is beyond the capacity of any individual agency or organization and will require broad and well-coordinated collaboration. Offshore planning envisioned by the evolving U.S. National Ocean Policy (http://www.whitehouse.gov/administration/eop/oceans/implementationplan, National Ocean Council 2011) should provide further impetus for such collaboration.
Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any archived material. Queries should be directed to the corresponding author for the article.
To cite this archived material, please cite both the journal article (formatting found in the Abstract section of this article) and the following recommended format for the archived material.
Silverman ED, Saalfeld DT, Leirness JB, Koneff MD. 2013. Data from: Wintering sea duck distribution along the Atlantic coast of the United States, Journal of Fish and Wildlife Management, 4(1):178–198. Archived in Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.m9t12.
Data A1. All data for the analyses are contained in the zipped folder titled DataACWSDsurvey.zip. The file contains three comma delimited data files (CluterWeightsMatrix.csv, ObservationData.csv, and TransectInformation.csv), as well as a ReadMe text file describing all data fields for each data file.
Reference A1. Caithamer DF, Otto M, Padding PI, Sauer JR, Haas GH. 2000. Sea ducks in the Atlantic flyway: population status and a review of special hunting seasons. Laurel, Maryland: U.S. Fish and Wildlife Service.
Reference A2. National Ocean Council. 2011. Draft national ocean policy implementation plan.
Reference A3. Sea Duck Joint Venture. 2003. Species status report. Continental Technical Team.
Reference A4. Sea Duck Joint Venture. 2007. Recommendations for monitoring distribution, abundance, and trends for North American sea ducks.
Reference A5. Sea Duck Joint Venture. 2012. Atlantic and Great Lakes migration study progress report—March 2012.
Reference A6. Sea Duck Joint Venture Management Board. 2008. Sea Duck Joint Venture strategic plan 2008–2012. Anchorage, Alaska and Sackville, New Brunswick, Canada.
Reference A7. Silverman E, Koneff M, Fleming K, and Wortham J. 2010. 2009 Atlantic coast wintering sea duck survey. Laurel, Maryland: U.S. Fish and Wildlife Service.
Reference A8. Silverman E, Koneff M, Fleming K, and Wortham J. 2011. 2010 Atlantic coast wintering sea duck survey. Laurel, Maryland: U.S. Fish and Wildlife Service.
Reference A9. Silverman E, Leirness J, Saalfeld D, and Richkus K. 2012a. 2011 Atlantic coast wintering sea duck survey. Laurel, Maryland: U.S. Fish and Wildlife Service.
Reference A10. Silverman ED, Leirness JB, Saalfeld DT, Koneff MD, and Richkus, KD. 2012b. Atlantic coast wintering sea duck survey, 2008–2011. Laurel, Maryland: U.S. Fish and Wildlife Service.
The authors would like to thank the many pilots and observers for collecting the data for this study: pilots J. Bredy, J. Bidwell, C. Ferguson, M. Koneff, T. Liddick, W. Rhodes, F. Roetker, and J. Wortham; observers N. Carle, S. Earsom, D. Forsell, T. Jones, T. Lewis, H. Obrecht, P. Padding, M. Perry, J. Solberg, and T. White. We thank K. Fleming and K. Luke for data and GIS support, and the Journal Subject Editor and three anonymous reviewers for their thoughtful comments.
Data collection and analysis were funded in part by the Sea Duck Joint Venture, the Atlantic coast Joint Venture, and the Atlantic Marine Assessment Program for Protected Species, administered by the National Oceanic and Atmospheric Administration.
Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
Silverman ED, Saalfeld DT, Leirness JB, Koneff MD. 2013. Wintering sea duck distribution along the Atlantic coast of the United States. Journal of Fish and Wildlife Management 4(1):178-198; e1944-687X. doi: 10.3996/122012-JFWM-107
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.