Mechanisms of Population Heterogeneity Among Molting Common Mergansers on Kodiak Island, Alaska: Implications for Genetic Assessments of Migratory Connectivity

Abstract.

Quantifying population genetic heterogeneity within nonbreeding aggregations can inform our understanding of patterns of site fidelity, migratory connectivity, and gene flow between breeding and nonbreeding areas. However, characterizing mechanisms that contribute to heterogeneity, such as migration and dispersal, is required before site fidelity and migratory connectivity can be assessed accurately. We studied nonbreeding groups of Common Mergansers (Mergus merganser) molting on Kodiak Island, Alaska, from 2005 to 2007, using banding data to assess rates of recapture, mitochondrial (mt) DNA to determine natal area, and nuclear microsatellite genotypes to assess dispersal. Using baseline information from differentiated mtDNA haplogroups across North America, we were able to assign individuals to natal regions and document population genetic heterogeneity within and among molting groups. Band-recovery and DNA data suggest that both migration from and dispersal among natal areas contribute to admixed groups of males molting on Kodiak Island. A lack of differentiation in the Common Merganser's nuclear, bi-parentally inherited DNA, observed across North America, implies that dispersal can mislead genetic assessments of migratory connectivity and assignments of nonbreeding individuals to breeding areas. Thus multiple and independent data types are required to account for such behaviors before accurate assessments of migratory connectivity can be made.

Resumen.

Cuantificar la heterogeneidad genética poblacional dentro de agregaciones no reproductivas puede brindar información sobre patrones de fidelidad a los sitios, conectividad migratoria y flujo génico entre áreas reproductivas y no reproductivas. Sin embargo, antes de poder evaluar con precisión la fidelidad al sitio y la conectividad migratoria es necesario caracterizar los mecanismos que contribuyen a la heterogeneidad, como la migración y la dispersión. Estudiamos grupos no reproductivos de Mergus merganser que se encontraban mudando en la isla Kodiak, Alaska, desde 2005 a 2007, usando datos de anillado para evaluar las tasas de recaptura, ADN mitocondrial (mt) para determinar el área natal y genotipos de microsatélites nucleares para evaluar la dispersión. Utilizando información de base de haplotipos de ADNmt de haplogrupos de ADNmt diferenciados a lo largo de América del Norte, pudimos asignar a los individuos a regiones natales y documentar la heterogeneidad genética poblacional dentro y entre grupos de muda. Los datos de recuperación de anillos y de ADN sugieren que tanto la migración desde las áreas natales como la dispersión entre ellas contribuyen a la conformación de grupos entremezclados de machos que mudan en la isla Kodiak. La falta de diferenciación en el ADN nuclear (de herencia biparental) de M. merganser observada a lo largo de América del Norte, implica que la dispersión puede confundir las evaluaciones genéticas de conectividad migratoria y la asignación de individuos no reproductivos a áreas reproductivas. Por lo tanto, antes de que se puedan hacer evaluaciones precisas de conectividad migratoria, se necesitan datos múltiples e independientes para explicar estos comportamientos.

Introduction

The migration of animals to nonbreeding areas for staging, molting, or wintering is a common phenomenon. Nonbreeding areas may be located far from the breeding grounds and offer sources of nutrition needed for subsequent migrations and reproduction as well as protection from predators and weather (Alcock 1993). Enhancements to fitness gained in one portion of the annual cycle may be carried over to subsequent life-history stages (Marra et al. 1998, Weitkamp and Neely 2002). Quantifying these various effects is useful for interpreting population dynamics and trends of animal aggregations at a specific point in the annual cycle but requires an understanding of migratory connectivity between breeding and nonbreeding areas (Webster et al. 2002). That is, do most individuals from one breeding area move to the same nonbreeding area or are the populations in nonbreeding areas heterogeneous, composed of individuals from multiple breeding populations? Heterogeneity within nonbreeding groups is important to quantify because if the group consists of demographically heterogeneous components, inferences about the species' life history drawn from it may be biased (e.g., Burnham and Rexstad 1993).

Aggregations of animals at nonbreeding areas are often more amenable to study than animals on breeding areas, where individuals of many species are widely dispersed at lower densities. Nonbreeding waterfowl (family Anatidae) are particularly easy to capture and study while flightless during molt. Many avian species maintain their flight capability throughout a prolonged molt (Langston and Rohwer 1996, Voelker and Rohwer 1998), but most species of waterfowl shed their flight feathers simultaneously, necessitating a 20- to 50-day period of complete flightlessness (Hohman et al. 1992). The capture and marking of molting waterfowl has been useful for quantifying annual survival rates (Flint et al. 2000, Iverson and Esler 2007), movement and site fidelity during molt (Bollinger and Derksen 1996, Flint et al. 2004, Nicolai et al. 2005), examining disease transmission (Hollmén et al. 2003), and determining the energetic requirements of molt (Guillemette et al. 2007).

The degree of population heterogeneity within nonbreeding groups of waterfowl has not often been examined. Understanding heterogeneity and the mechanisms that contribute to admixture, such as migration or dispersal, however, is useful for assessing the degree of migratory connectivity between breeding and nonbreeding areas. Additionally, mechanisms that contribute to heterogeneity may vary by sex and age, as well as across broad temporal and spatial scales (Avise et al. 1992, Fedy et al. 2008). Previous studies have used various methods to assess population heterogeneity, such as mark—recapture (Bollinger and Derksen 1996, Dau et al. 2000) and genetic (Pearce et al. 2000) and morphological criteria (Scribner et al. 2003). Fewer studies have combined inferences from multiple data sources to assess heterogeneity and also infer mechanisms that contribute to population admixture (Bjorndal and Bolten 2008, Fedy et al. 2008). This combination of data types is useful not only when single markers (genetic or mark—recapture data) lack resolution sufficient to characterize variation within a group but also when populations are highly diverged for a single data type (Pearce and Talbot 2006, Pearce et al. 2008).

The Common Merganser (Mergus merganser) is a large sea duck with a holarctic distribution. Mark—recapture and satellite-telemetry data suggest that male Common Mergansers migrate to molting areas whereas females molt near where they breed (Little and Furness 1985, Pearce and Petersen 2009). Thus molting flocks of Common Mergansers may be composed of individuals from multiple natal and breeding areas. Identification of nonbreeding Common Mergansers' source populations appears possible with mitochondrial (mt) DNA. Hefti-Gautschi et al. (2008) observed substantial differentiation of mtDNA among European breeding populations of the Common Merganser (Goosander) and genetically heterogeneous groups of nonbreeding birds during winter.

Although the determination of natal or breeding-area origins of nonbreeding waterfowl appears straightforward on the basis of baseline populations divergent in mtDNA, such assignments may not reflect the current breeding area of an individual because of prior dispersal (Fig. 1). If male Common Mergansers migrate directly between natal and molting areas, then mtDNA infers population membership correctly (Fig. 1A). In contrast to other bird families, however, in waterfowl males typically disperse, whereas females tend to be philopatric (Greenwood 1980, Anderson et al. 1992). Therefore, dispersal away from a natal area for subsequent breeding means that population membership based on mtDNA may be misleading. Thus types of data aside from mtDNA are needed to assess if multiple mechanisms, such as migration and dispersal, are contributing to the composition of flocks at nonbreeding areas. Nuclear DNA has often been used to infer male dispersal that results in gene flow (Avise 2004). This is because there is often a lack of concordance between patterns of mtDNA and nuclear DNA variation in species with malebiased dispersal patterns. In such species, such as waterfowl, higher levels of differentiation are often observed for maternally inherited mtDNA, whereas lower levels are found with biparental nuclear DNA (Scribner et al. 2001).

Some degree of population structure has been hypothesized for the Common Merganser on the basis of band-recovery data (Pearce et al. 2005), but nothing is known about patterns of genetic differentiation for the Common Merganser across North America. Similarly, little is known about levels of migratory connectivity between the species' breeding and nonbreeding areas. In this study, we characterized mtDNA sequences and nuclear genotypes for Common Mergansers across North America, used mtDNA group membership to determine whether molting Common Mergansers on Kodiak Island, Alaska, were derived from multiple natal origins, and examined whether group membership varied temporally and spatially across the island. We also used nuclear DNA and band-recovery data as an independent method to infer relative frequencies of migration and dispersal to explain any population heterogeneity observed within flocks. For this objective, we hypothesized that if males migrate between natal and molting areas directly, then levels of breeding-population structure in biparentally inherited nuclear DNA should be similar to those observed with maternally inherited mtDNA (Fig. 1A). If males are dispersing away from natal areas prior to the molt, however, then nuclear DNA should show little if any population structure among breeding areas in comparison to that shown by mtDNA (Fig. 1B).

Materials and Methods

Study Area and Data Collection

To obtain a baseline of breeding populations that could be used to determine natal origins of each molting bird via mtDNA and levels of male gene flow via nuclear microsatellite genotypes, we collected DNA samples from Common Mergansers across North America (Appendix). All samples of breeding birds were collected between March and August and came from young before fledging, attending adult females, or single adult females. The only adult males in the breeding sample were eight from the Columbia River, Washington, from a total of 20. To include samples from southeastern Alaska, we used nine putatively breeding samples of Common Mergansers (two juvenile males, five juvenile females, and two adult females) collected in September and October from Prince of Wales Island, Alaska.

To understand the distribution of molting Common Mergansers across Kodiak Island, we mapped locations of Kodiak Island, both freshwater and marine, where >50 molting birds were observed from 1994 to 2007 (Fig. 2). From these locations, we selected five that were within our logistical constraints to conduct captures and DNA sampling. These locations included the freshwater Frazer and Karluk lakes and marine areas of Uyak, East Arm Uganik, and Terror bays (Fig. 2). We captured and sampled flightless male Common Mergansers from 2005 to 2007: at Karluk Lake 12–21 July 2005, 11–13 July 2006, and 20–22 July 2007, at Frazer Lake 20 July 2007, and in the three marine areas 12–14 July 2007). Locations for 2007 sampling were selected because of their proximity to Karluk Lake to maximize recapture of birds previously banded at that lake. We captured mergansers by using small boats to herd flocks of flightless birds into a net enclosure on shore. We surveyed by boat each year near the shores of Karluk Lake before and after captures to estimate the total number of molting males present. All capture and sampling procedures were approved by the Institutional Animal Care and Use Committee at the U.S. Geological Survey (USGS), Alaska Science Center, and by the University of Alaska, Fairbanks. We identified all captured birds as males by plumage, cloacal exams, and a molecular sexing technique (below). We marked each bird with a metal leg band and recorded standard measurements to assess the birds' size and stage of molt: exposed culmen, total tarsus, mid-wing (distal end of the radius to the proximal end of the ulna measured on the dorsal surface), ninth primary length, and body mass. For DNA samples, we collected two or three small (≤2 cm) emerging feather quills from the secondary coverts.

Dna Extraction, Sex Verification, and Mtdna Sequencing

We extracted DNA from all samples by following methods described in Pearce et al. (2004) and verified that all sampled birds were males with the P8 and P2 polymerase chain reaction (PCR) primers for the CHD gene (Griffiths et al. 1998). We amplified the CHD gene by PCR in a final volume of 10 µL, containing 1.5 µL DNA extract, 10.0 mM Tris-HCl (pH 8.3), 1.5 mM MgCl₂, 50.0 mM KCl, 0.01% gelatin, 0.01% NP40, 0.01% Triton-X 100, 0.2 mM deoxyribonucleotide triphosphate (dNTP), 3.6 pmoles unlabeled forward P8 primer, 4.0 pmoles unlabeled reverse P2 primer, 0.4 pmoles fluorescence-labeled P8 primer, 0.1 µL-¹ bovine serum albumin, and 0.75 units Taq polymerase (United States Biochemical, Cleveland, OH). The PCR cycling was performed on a Stratagene 96 Robocycler (La Jolla, CA) with a profile of 94° C for 90 sec (1 cycle), 40° C for 45 sec, 72° C for 45 sec, 94° C for 30 sec (40 cycles), 48° C for 60 sec (1 cycle), and 72° C for 5 min (1 cycle). The PCR products were visualized on 6% polyacrylamide gels by means of a LI-COR 4200 DNA sequencer (LI-COR Biosciences, Lincoln, NE).

To characterize mtDNA for samples of both breeding and molting birds, we amplified and sequenced a 425-base-pair fragment of the control region (domain I) of mtDNA by using primers MMCRL F and MMCRL R designed for the Common Merganser in Europe (Hefti-Gautschi et al. 2008) and methods identical to those described in Pearce et al. (2008). We aligned all sequences with the program AlignIR version 2.0 (LI-COR) and organized multiple sequences into unique haplotypes by using the COLLAPSE function of the software FaBox (Villesen 2007). All mtDNA sequences have been deposited in GenBank (accession numbers FJ190670-FJ191172) under location codes provided in the Appendix.

Mitochondrial Dna Analyses

To assess levels of genetic differentiation among breeding locations (Alaska and western and eastern North America), we examined neighbor-joining trees from 10000 bootstrap replicates in MEGA version 4 (Tamura et al. 2007) and calculated overall and pairwise levels of F_ST or φ_ST by means of ARLEQUIN version 3.11 (Excoffier et al. 2005). These analyses incorporated the Tamura and Nei (1993) model of nucleotide substitution as identified by MODELTEST (Posada and Crandall 1998) as the best fit to our data.

To determine the most likely natal origin of each molting Common Merganser captured on Kodiak Island, we used COLLAPSE to compare the sequence of each bird to the breeding haplotypes. Group assignment of novel haplotypes was determined through a comparison to the breeding haplotypes by means of a neighbor-joining tree with 10 000 bootstrap replicates in MEGA. Because our samples from breeding areas from western and eastern North America were separated by a wide gap, we restricted our assignment of individuals to either Alaska or elsewhere in North America rather than to specific areas within these regions. A European Common Merganser sequence (GenBank accession #EF675705) from Hefti-Gautschi et al. (2007) was used as an outgroup.

We tested for annual and spatial differences in the number of haplotypes from each group by using exact tests of population differentiation via F_ST with the program ARLEQUIN. Two sets of analyses were performed: (1) annual samples from Karluk Lake (2005–2007) and (2) site-specific samples from all additional areas where birds were captured in 2007, plus a composite sample from Karluk Lake that contained the frequencies of all haplotypes observed 2005–2007, recaptured individuals excluded. We computed F_ST values by using conventional F-statistics from haplotype frequencies and assessed significance by comparison to values generated from 10000 random permutations. ARLEQUIN was also used to compute haplotype (h) diversity for each sampling location.

Nuclear Dna Analysis

To examine if genetic heterogeneity among molting males at Kodiak Island resulted from migration, dispersal, or both, we obtained nuclear microsatellite genotypes for 140 breeding samples collected across North America (Appendix) by using seven microsatellite loci: APH04 and APH08 (Maak et al. 2003), CRG (A. Baker, pers. comm. F 5′GTAGGCAAAGCAAGTCTGAAGTT-3′, and R 5′- GCAACCACCAGCAGTCACTACAA-3′), Hhiµ5 (Buccholz et al. 1998), MM01 and MM02 (Gautschi and Koller 2005), and HrU2 (Primmer et al. 1995). Gautschi and Koller (2005) isolated nine loci for the Common Merganser from Europe, but all except MM01 and MM02 were monomorphic or yielded no PCR product, likely because of the relatively deep divergence (5–7% sequence divergence) between mergansers of the two continents (see Hefti-Gautschi et al. 2008). PCR amplification of all nuclear loci during screening and data collection involved identical reagent cocktails as described in Pearce et al. (2004) except that all were amplified with the same PCR temperature profile (94° C for 2 min followed by 40 cycles of 94° C for 2 min, 50° C for 1 min, and 72° C for 1 min) by means of an MJ Research PTC-200 thermal cycler. PCR products were visualized on 6% polyacrylamide gels with a LI-COR 4200 DNA sequencer. Genotypes were scored according to allele size on the basis of an initial comparison to an M13 DNA sequence ladder and then to samples established as size standards that were run on each subsequent gel.

Using ARLEQUIN, for each microsatellite locus we calculated allelic frequencies, number of alleles per locus (A), and observed (H₀) and expected (H_e) heterozygosities. Following the method of Guo and Thompson (1992), we used GENEPOP (Raymond and Rousset 1995) to conduct exact probability tests for deviations from Hardy—Weinberg equilibrium in each sampling area. We further assessed deviations from Hardy—Weinberg by estimating Wright's inbreeding coefficient, FIS, across all loci for each sampling area by using the program FSTAT, version 2.9.3 (Goudet 1995). Positive values of FIS indicate heterozygotic deficiency, a signal of inbreeding or population admixture (i.e., Wahlund effect), whereas negative values indicate heterozygotic excess. We tested for differences among breeding samples by using FST in ARLEQUIN. To examine differentiation among areas further, we used the Bayesian clustering method of Pritchard et al. (2000) in the program STRUCTURE, version 2.2 (Pritchard et al. 2000). We excluded population information from the analysis and assessed the likelihood that the entire data set was composed of individuals from K populations (the admixture model), each of which may be characterized by a unique set of multilocus allele frequencies. We estimated the probability of each K (ln P(X | K)) from 1 to 5. Five runs of each value of K were conducted to examine variation in the probability estimates of K. Results are based on 30 000 Markov-chain Monte Carlo iterations following a burn-in period of 30 000 iterations.

Results

Mitochondrial Dna Population Structure

In mtDNA we found 29 haplotypes among North American breeding samples, and these had a pronounced pattern of population structure. North American samples clustered into three major clades of haplotypes (Fig. 3A; Appendix): (1) comprises samples from interior and western Alaska (we designate it “Beringia”); (2) comprises samples from south-central Alaska, Prince of Wales Island in southeastern Alaska, and British Columbia (“Alaska, British Columbia, and Prince of Wales Island”); (3) comprises samples from Washington state, western Ontario, Vermont, Maine, and New Brunswick (“western and eastern North America”). The overall difference among sampling areas was high (φST = 0.857, P < 0.001), and all pairwise tests were similarly high and significant (range 0.622–0.919, P < 0.001). Differentiation between western and eastern North American samples was also significant (φST = 0.622), and this was the lowest pairwise difference among all comparisons.

Mark-Recapture and Band Recovery

From 2005 to 2007, we captured and banded 176 molting Common Merganser males at Karluk Lake (Table 1). The annual rate of return of previously banded males appeared to be low, as we recaptured only four marked birds in subsequent years on Karluk Lake, even though our capture effort averaged 41% (Table 1). We recaptured no additional birds in 2007 in marine or freshwater areas adjacent to Karluk Lake. MtDNA haplotypes of three of the four recaptured birds clustered with group 2 (Alaska, British Columbia, and Prince of Wales Island), the fourth with group 1 (Beringia).

Four males banded during the molt on Kodiak Island (2 in 2005 and 2 in 2007) were subsequently recovered by hunters outside of Alaska: two in Washington in January 2006, one in California in January 2008, and one in Oregon in November 2008 (Table 1). MtDNA from one of these birds clustered with group 3 (western and eastern North America). Additionally, a male captured during molt at Karluk Lake in 2005 was originally banded in April 2004 in south-central Alaska near Seward (60° 05′ N, 149° 25′ W).

Population Membership

We obtained mtDNA sequence data for 190 samples across Kodiak Island (Table 2). Seventy-seven percent of samples were identical to one of the 29 haplotypes identified in the North American breeding samples, with the remainder yielding 25 new haplotypes. Novel haplotypes were observed in each year at Karluk Lake and all other sites except Uyak Bay. Haplotypic diversities were high in all areas, even in Uyak Bay, where few samples (n = 7) were collected (Table 2).

Our analysis of temporal variation in mtDNA group membership at Karluk Lake from 2005 to 2007 suggested no difference in the proportions of each of the three groups (FST = 0.002, P = 0.451), with group 2 haplotypes (Alaska, British Columbia, and Prince of Wales Island) being the most common (Fig. 3B). Greater variation in the frequency of group 2 haplotypes was observed in our analysis of spatial variation in membership across all five sampling sites (Fig. 3C). In this analysis, group membership varied significantly by site (FST = 0.071, P = 0.001) as a result of differences between two often pairwise comparisons (Karluk Lake vs. Terror Bay and Terror Bay vs. Uganik Bay). A greater proportion of group 3 haplotypes (western and eastern North America) was present in Uyak and Terror bays.

Nuclear Microsatellite Data Among Breeding Areas

In the 140 samples from the breeding range analyzed for seven loci, average variation was from two to ten alleles but was consistent for each locus across all sampling areas (Table 3). Significant deviations from Hardy—Weinberg proportions (P < 0.05) were detected in four of the 28 area-by-locus combinations. No consistent pattern of deviation was noted across all sampling areas, except for the MM02 locus, which had significant heterozygote deficiencies in all areas but western North America (Table 3). Exact tests of Hardy—Weinberg equilibrium showed significant values (P < 0.05) in two of 84 comparisons, but pairs of loci in these two tests were not the same (data not shown). In none of the four sampling areas were FIS values significantly large, an alternate indicator of heterozygotic deficiency (Table 3). The overall estimate of FST across broadly distributed breeding areas was low and nonsignificant (F_ST = 0.002, P = 0.73). Results from the program STRUCTURE were similar to those via F-statistics. Model likelihood was at a maximum with K equaling a single population, and the log likelihood associated with this value of K (-2030.0) was the same across all five runs. Smaller (more negative) and more variable likelihood estimates were observed across the five runs for each K > 1. Thus both sets of analyses suggested less differentiation for nuclear data that we conclude are likely the result of males' dispersal and gene flow among breeding areas.

Discussion

Our results suggest that Common Mergansers sampled on breeding areas have a high degree of population structure in mtDNA, likely the result of female philopatry, and little differentiation in biparentally inherited nuclear microsatellite loci, which we attribute to male dispersal and gene flow. On the basis of nuclear and mtDNA, groups of molting male Common Mergansers on Kodiak Island, regardless of year and area, appear to be composed of individuals from multiple natal and breeding areas. Our band-recovery data, albeit somewhat limited, suggest that some males migrate directly from natal areas to molting areas, whereas data from nuclear microsatellite loci suggest that male dispersal among breeding areas prior to capture on Kodiak Island also contributes to genetic heterogeneity within molting flocks (i.e., both mechanisms described in Figure 1). Over the three years of the study we recaptured few previously banded birds, suggesting limited molt-site fidelity, low recapture probability, or poor annual survival (see below).

The lack of concordance between mitochondrial and nuclear DNA is expected for species with female philopatry and male dispersal (Greenwood 1980, Avise et al. 2004). Such patterns are typical of both waterfowl (Scribner et al. 2001, Tiedemann et al. 2004) and animals other than birds (Lyrholm et al. 1999, Ngamprasertwong et al. 2008). Similar results for nuclear and mtDNA were also noted among European Common Mergansers (Hefti-Gautschi et al. 2008). Additionally, nonmolecular data for the Common Merganser have also suggested female philopatry and male dispersal (Pearce and Petersen 2009) and population structure (Pearce et al. 2005). Thus we find greater evidence for sex-biased dispersal tendencies to explain the lack of congruence between nuclear and mitochondrial molecular markers than do alternative explanations, such as incomplete sorting of nuclear alleles since population divergence or the theoretical fourfold lower effective population size of mitochondrial DNA (Avise 2004).

The implication of male-mediated nuclear gene flow for nonbreeding groups is that population genetic heterogeneity at Kodiak Island cannot be attributed to direct connectivity between Kodiak Island and natal areas. Instead, nuclear DNA suggests that male dispersal within and among breeding areas of North America is continuing, whereas band-recovery data suggest migration to Kodiak Island. Thus on breeding areas males are likely to be as heterogeneous as observed on molting areas (Fig. 1B). As a result, inferring levels of migratory connectivity (sensuWebster et al. 2002) is problematic not only in this and other waterfowl species whose dispersal is sex-biased but also in species in which dispersal from natal sites is common. In other words, levels of population differentiation among breeding areas should be established before migratory connectivity to wintering areas can be assessed. Limited evidence from band recoveries and mtDNA suggests that some males at Kodiak are migrating between natal and molting areas. Four birds banded on Kodiak Island during molt were recovered while wintering in Washington, Oregon, and California (Table 1). A DNA sample from one of these was most closely related to the western and eastern North American group (Fig. 2), suggesting this bird migrated to Kodiak for molt before returning to its natal area to winter. Long-distance migration to molting areas has also been documented by banding for the Common Mergansers in Europe (Little and Furness 1985).

The long-distance migrations implied by our band-recovery data could also be interpreted as annual migration north to breeding and molting areas in Alaska followed by a return migration to wintering areas farther south. Recoveries of Common Mergansers banded elsewhere in North America suggest a migration for some areas (Pearce et al. 2005), but it is unknown what proportion of Alaska breeding birds undertake such a migration. Common Mergansers winter throughout Alaska along ice-free rivers and coasts. A larger set of band recoveries or deployment of satellite transmitters to track the annual movements of individual males would improve our understanding of the species' migration. Additionally, an assessment of mtDNA group affiliation of breeding males across North America would yield information about the proportion of immigrant haplotypes within each breeding group. Last, stableisotope analysis of feathers (reviewed in Hobson 2005) could yield information on the recent breeding or wintering locations used by molting male Common Mergansers.

Our three-year mark—recapture analysis at Karluk Lake suggests that male Common Mergansers return to this molting area at a low rate. This finding could be due to low probabilities of recapture, annual survival, or molt-site fidelity. Although we captured an average of 41% of all birds molting on Karluk Lake each year, our recapture rate may be low. The survival rate of birds molting at Karluk Lake is unknown but was low (< 0.51) for other groups of marked Common Mergansers (Pearce et al. 2005). Despite the low rate of recapture, the proportional representation of mtDNA groups at Karluk Lake was maintained each year (Fig. 2B). Given the low level of recapture, one would expect greater annual variation in mtDNA group proportions at Karluk Lake. Additional years of sampling across Kodiak Island are needed to clarify if the consistent mtDNA group proportions observed over three years at Karluk Lake are due to sampling bias or other factors. For example, in aggregations of nonbreeding Green Turtles (Chelonia mydas), Bjorndal and Bolten (2008) found little annual variation in population membership, as inferred by mtDNA haplotype, in a series of 4 years but substantial variation over 10 years. These authors identified factors such as breeding-colony productivity as contributing substantially to the annual variation of different mtDNA groups in nonbreeding aggregations. These grouns may vary temporally within the year at each site on Kodiak Island. We sampled molting birds at Karluk Lake in early July each year, but flightless Common Mergansers are observed on the lake until early September (D. Zwiefelhofer, unpubl. data), and proportions of the three groups toward the end of the molting period may be different. We observed no difference, however, in the average lengths of new ninth primaries among 115 birds from each of the three mtDNA groups (Kruskal—Wallis χ² = 0.291, df = 2, P = 0.86, data not shown), suggesting similar timing of molt among all birds.

Our study examined the behavioral mechanisms that contribute to population heterogeneity within molting flocks, but not the factors involved in the Common Merganser's use or selection of a particular area for molting. To understand such mechanisms, more information is needed on molting Common Mergansers' foraging behavior, duration of the molt, and habitat variables associated with molting locations. Additionally, a large number of satellite transmitters deployed on male Common Mergansers would be useful to examine patterns of annual migration to molting areas. This would, in turn, require the quantification of habitat variables and foraging ecology of mergansers that use other molting sites away from Kodiak Island as they are discovered.

Such research may yield insights into our low rates of recapture of birds molting on Kodiak Island. In Scotland Hatton and Marquiss (2004) observed greater levels of site fidelity for molting female Common Mergansers than we observed for males, but also some switching of sites. Apparent patterns of molt-site fidelity among sea ducks vary greatly with measurement scale and methodology (Flint et al. 2000, 2004, Iverson et al. 2004, Mehl et al. 2004, Phillips et al. 2006). We do not conclude that apparent low rate of return of banded birds results in heterogeneity within molting flocks of Common Mergansers but instead view the two behaviors as decoupled. For example, Steller's Eiders have shown high levels (> 95%) of molt-site fidelity to specific beaches in Alaska (Flint et al. 2000) but little connectivity between molting and breeding areas (Dau et al. 2000). Similarly, Bollinger and Derksen (1996) found high site fidelity (95%) of molting Black Brant (Branta bernicla) from multiple nesting colonies throughout Alaska and Canada to specific lakes on the north slope of Alaska. Thus, as argued previously (Pearce and Talbot 2006, Pearce et al. 2008), site fidelity is an inconsistent measure of within-group population composition or demographic independence among groups.

Acknowledgments

We thank the staff, seasonal technicians, and volunteers of the Kodiak National Wildlife Refuge and D. Derksen, S. Talbot, K. Sage, J. Gust, P. Flint, M. Marquiss, and D. Berthiaume. We also thank the numerous individuals who assisted with the collection of DNA samples across North America, including the staff of the Togiak National Wildlife Refuge, K. McCracken, K. Winker, R. Wilson, and S. Sonsthagen (University of Alaska, Fairbanks), staff of the U.S. Forest Service (Tongass Ranger District), S. Birks (Burke Museum of Natural History and Culture), K. Timm, R. McNeil, and B. Braune (Environment Canada, Wildlife Toxicology Division, Specimen Bank), J. Peters and K. Omland (University of Maryland), L. Savoy (BioDiversity Research Institute), and M. Schwitters. Logistical and funding support was provided by the North American Sea Duck Joint Venture, Kodiak National Wildlife Refuge, and the U.S. Geological Survey, Alaska Science Center. P. Flint, K. McCracken, M. Lindberg, K. Winker, J. Peters, D. Derksen, P. Unitt, and two anonymous reviewers offered comments on earlier versions of the manuscript. Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. government.

Literature Cited