Cohort-Specific Differential Vulnerability to Capture of Mallards and Wood Ducks in Baited Swim-in Traps

Annual banding directly after the breeding season is a primary component of waterfowl monitoring in North America (Tautin et al. 1999; U.S. Fish and Wildlife Service [USFWS] 2022a). This monitoring program, known as preseason banding for its occurrence directly before hunting seasons, marks thousands of individuals a year; these data are subsequently used in annual harvest management decision-making processes (USFWS 2022a). In addition to regulatory decisions, these data are used to address research questions related to harvest timing, annual fecundity, and space use, among others (Munro and Kimball 1982; Szymanski and Dubovsky 2013). Although capture and recovery data are primarily used for estimating survival or movement (Anderson and Burnham 1976; Brownie et al. 1985; Roberts et al. 2023), data may be used to estimate age ratios, formulate full annual-cycle population models, or expand the suite of available survival and movement models to include joint live and dead encounter models among others (Hostetler et al. 2015; Arnold 2018; Wann et al. 2020).

Managers typically use age ratios of harvested individuals (Paloheimo and Fraser 1981; Carney 1992) or ocular surveys of species where juveniles and adults can be identified in the field (Fronczak and Rigby 2019) to estimate population age ratios of waterfowl and other wildlife species. Because juveniles are often more vulnerable to harvest than adults, band recovery data are needed to adjust harvest age ratios for differential vulnerability to harvest to estimate true age ratios. Alternatively, age ratios during banding operations may be used to determine regional variation in fecundity after addressing effects of differential capture vulnerability, provided immigration and emigration from the area are minimal or do not differ among cohorts. Population-level recruitment can be assessed by pooling vulnerability-corrected age ratios among banding stations. Before fall migration, juveniles may be more susceptible to capture than adults, necessitating an estimate of age-specific vulnerability to capture methods. The estimate of differential vulnerability to capture can be used to correct observed age ratios and estimate the true age ratio at various scales (Zimmerman et al. 2010).

Use of data from annual monitoring programs can be multifaceted and add value to other data streams used by managers. The value of preseason banding data may be further increased by determining if age and sex cohorts of waterfowl are differentially vulnerable to a typical capture method. Our objectives were to estimate the recapture rate of age and sex cohorts and test for differential vulnerability to capture among these cohorts and banding locations using existing data for mallards Anas platyrhynchos and wood ducks Aix sponsa.

We obtained capture and recapture data for mallards from nine locations across the mid-continent and western breeding range (Figure 1). These locations are typical of the annual monitoring effort for mallards for use in regulatory decisions. Many of these locations were in the Prairie Pothole Region of North America (Audubon Wetland Management District [WMD], Fairview, Sand Lake WMD, Brooks, Oak Hammock, and Big Grass Marsh). This area is dominated by agriculture and grasslands with numerous temporary and seasonal wetlands and is the core breeding area of North American waterfowl. We also had two locations in the boreal forest, another primary breeding area for waterfowl (Mills Lake, Hay-Zama). The boreal forest generally has lower densities of waterfowl but some large wetlands host thousands of birds directly after the breeding season. Finally, we used a single location in the Great Basin of the western United States (Summer Lake). This area has fewer wetlands than the mid-continent but is a primary breeding and migration area for birds using the Pacific Flyway. Data were obtained for a single year at each location, either one of 2018 (Mills Lake), 2021 (Audubon WMD, Big Grass Marsh, Fairview, Oak Hammock, Sand Lake WMD, Summer Lake), or 2022 (Brooks, Hay-Zama).

We also obtained in-season recapture data for wood ducks from two locations: Tennessee National Wildlife Refuge during 2012–2022 and Hatchie National Wildlife Refuge during 2021–2022. Both refuges have high densities of breeding wood ducks. Numerous wetlands managed through water control structures, often to promote use by waterfowl, are present on these refuges.

We captured mallards and wood ducks using baited swim-in traps during August and September (USFWS 2022b). Each location consisted of one to three traps at two to eight sites within a 20-km radius. Trap type for mallards was primarily oval or Benning II (Dieter et al. 2009), with variable trap size. Data were collected by the same crew within a location and year. We recorded the first capture of all birds within a year, then all recaptures at every site throughout the capture period. We defined recaptures as birds recaught after initial capture in the same year. Therefore, a bird caught in 2022 that was initially banded in 2021 would not be considered a recapture until it is subsequently recaptured in 2022. Within a year, all mallard traps were operated for some subset of dates over a 52-d period, 4 August to 24 September. Wood duck banding occurred 1 July to 30 September.

All models were formulated in a Bayesian framework and fit in JAGS using the jagsUI package (Kellner 2019) for Program R (R Version 4.2.2, www.r-project.org, 15 January 2023). Mallard and wood duck data were analyzed separately. For each species we ran three chains of 10,000 iterations, with the first 2,000 as burn-in and a thin rate of 5. Thus, parameter estimates were based on 4,800 posterior samples. We assessed model convergence using the Gelman statistic ⁠, assuming values <1.01 indicated adequate convergence (Gelman and Hill 2007). We report estimates as posterior distributions with 95% credible intervals.

For mallards across the nine locations, we obtained capture–recapture history for 5,757 individuals, including 2,860 adult male, 1,774 adult female, 596 juvenile male, and 527 juvenile female birds. Our wood duck data from two sites included 8,704 individuals. Compared with mallards, juveniles were more abundant in wood duck samples, with cohort totals of 462 adult males, 647 adult females, 3,988 juvenile males, and 3,607 juvenile females. Audubon WMD had only a single capture of a juvenile mallard, and no recaptures, so this site was removed for all juvenile recapture rates and ratios.

Apparent survival posterior estimates for mallards were similar among cohorts during the entire banding period. Adult apparent survival estimates for males were 0.83 (0.82–0.85) and 0.85 (0.83–0.87) for females. Juvenile cohorts had mean survival estimates for males of 0.89 (0.87–0.91) and 0.86 (0.84–0.88) for females. Wood duck apparent survival rates were all above 0.90. There was little variation in cohort-specific survival rates between adult males (0.96–0.98), adult females (0.93–0.95), juvenile males (0.96–0.97), and juvenile females (0.96–0.97).

We found that posterior estimates of mallard group- and location-specific recapture rates varied from a mean of 0.004 (0.002–0.008) for adult females at Sand Lake WMD to 0.547 (0.486–0.609) for juvenile females at Mills Lake (Table 1). Among cohorts at each location, the pattern of recapture rates was the same, with some exceptions. Generally, both juvenile cohorts had the largest recapture rates, followed by adult males and then adult females. Ratios of recapture rates among cohorts (differential vulnerability) were variable among locations (Table 2). For example, the adult female-to-adult male ratio ranged from 0.29 (0.14–0.51) at Oak Hammock Marsh to 0.75 (0.67–0.85) at Mills Lake (Table 2, Figure 2). Juvenile female-to-adult female ratio was greater than adult female-to-adult male ratios (Table 2).

Wood duck posterior estimates of group- and location-specific recapture rates were similar among sites for the same cohort (Table 1). The pattern of recapture rates among cohorts was the same as mallard with the exception of adult females, which had greater recapture rates than adult males. Resulting ratios of recapture rates were similar among the two sites (Table 2). Mean wood duck recapture ratios for adult females to adult males and for juvenile females to adult females were both lower than those estimated for mallards, though there was significant variation within the estimates (Table 2).

We found a wide range of recapture rate estimates for mallards across banding locations. This result suggests that location-specific characteristics may have a strong effect on the ability of biologists to capture birds or there is variable emigration of marked birds from banding areas. Other hypotheses we were unable to test regarding variable recapture rates include trapping effort, abundance of available birds for capture, or weather throughout the trapping period or single storm events that may affect recapture rate or banding effort. Differences in estimates of differential vulnerability to capture suggest that no single correction factor can be used for all banding locations. This result indicates that location-specific recapture data would be needed to adjust for differential vulnerability. Given the short time frame of our study and its occurrence after the breeding season, we would expect mortality to be minimal. Our estimates of survival likely are a product primarily of emigration from the trapping location, so the use of recapture data in models must account for an open population. Contrary to mallards, recapture ratios for wood ducks were similar between the two sites, but the smaller sample of sites compared with mallards and relatively small distance between sites make inference to continental wood duck banding limited.

Other estimates of differential vulnerability of ducks to capture are rare. Dieter et al. (2009) tested capture rate among trap types using ratios of number of birds caught per day that were banded or not. They found that trap style affected capture rate, with trap types used in our study, oval and Benning II, being most successful. Arnold (2018) estimated differential vulnerability to capture during preseason banding for northern pintails Anas acuta. In that work, juvenile pintails were more susceptible to capture than adults by a ratio of 2.05:1. That estimate is within the range of our female mallard (juveniles to adult) vulnerability estimates of 2.35:1, but larger than our mallard male (juveniles to adult) estimates of 1.38:1. Our mean wood duck vulnerability estimates for females (1.30, juveniles to adult) and males (1.24, juveniles to adult) were closer to the pintail estimate.

There are multiple hypotheses for why juveniles may be more vulnerable to capture than adults. First, juveniles may have a greater capture availability than adults. Additionally, juveniles may be naïve to the landscape outside of the general fledging locations and reluctant to travel greater distances for food if provided at a nearby trapping location. Juvenile ducks may also be recently separated from brood hens after fledging, leaving them more susceptible to pursue bait with less risk aversion. In contrast to increased relative vulnerability of juveniles, adults may be more capture averse than juveniles. There may be a learned response to capture by adults or general skepticism of anthropogenic trapping features that juveniles have not achieved. Capture probability could also be affected by differing energetic needs among cohorts. An energetic disparity with adults of greater mass or better condition than juveniles could simply make juveniles more apt to accept risk associated with use of trapping sites for bait. Finally, capture vulnerability could also be affected by general capture strategies deployed and represent differences among banding crews. Banding crews operate in a wide variety of conditions, and variables such as access to traps, local logistical support, or crew experience can affect both initial capture and recapture rates.

Management agencies often have limited resources for monitoring, so the ability to monitor multiple demographic parameters using a single program is vital for continued assessment of population status, informing true adaptive resource management and opportunistic research to address science needs and model assumptions. Banding data can provide estimates of survival, harvest rate, distribution and derivation of harvest, fecundity, emigration, and immigration (Brownie et al. 1985; White and Burnham 1999). Knowledge of differential vulnerability to capture greatly increases the utility of continental banding programs (Arnold 2018). If there is minimal movement between breeding efforts and capture events, banding data can complement more intensive research to assess spatiotemporal variation in fecundity (Specht and Arnold 2018; Devries et al. 2023). Most banding sites are operated for a long period, so there is potential for long-term monitoring of true age ratios on a regional basis if assumptions about differential vulnerability can be tested through time. Relative age ratio data from long-term banding sites can provide a data source to test the effects of landscape change and other factors on this vital rate (Osenkowski et al. 2012). A full accounting of recaptures at banding sites can also allow the use of many more capture–recapture models to expand inference of population dynamics including robust design or live–dead models (White and Burnham 1999; Sandercock 2006; Weegman et al. 2020).

We encourage banding program managers to consider collection of in-season recapture data at a subset of banding locations where feasibility allows and consistency over time is likely. Data collection needs should be coordinated among stations to determine collection intensity. Primary factors to consider when designing a larger study are species or cohort of interest, where landscape or other factors may vary, and what are the primary research objectives. Many North American banders do not report within-season live encounters either because the Bird Banding Laboratory historically discouraged such reports (Buckley et al. 1998), or because of the perceived difficulty or time required in collecting this information during banding efforts. Recently, the U.S. Geological Survey Bird Banding Laboratory updated their policy on reporting live in-season recaptures, and data submission methods allow easier submission of those data (Smith 2013). Banders should be aware of the value of live in-season encounters in monitoring populations and modeling dynamics (Arnold 2018). Data management of live recaptures can be challenging and the effort to collect these data may seem burdensome, especially at sites or with species with abundant captures, but there are methods to improve efficiency in data collection. For example, rather than record all recaptures at each trapping location on each day, live recaptures can be recorded at a single site each day or at a subset of banding locations where feasibility is maximized. There are also several concerns to be addressed when attempting to collect these data. Banders must first consider bird welfare and human safety before collecting additional information, and time is often a limiting factor in improving those conditions.

As an alternative to increased effort recording in-season recapture data, we recommend a robust study to record recaptures using all capture methods for all species that are regularly banded. Capture methods such as propelled nets (i.e., rocket nets, pneumatic cannon nets) or night-lighting are often used and likely have varying effectiveness and hence result in different measures of differential vulnerability. Some species, guilds, or cohorts may have different vulnerability to capture using the same methods, so use of preseason banding data to estimate fecundity for those species would require additional live recapture data. Although few capture methods are identified in bander-submitted data (https://www.pwrc.usgs.gov/ BBL/MANUAL/summary.cfm), if recapture data were submitted with detailed information about capture type, biologists could complete a thorough estimation of differential vulnerability to capture on the basis of capture methodology (Arnold 2018). A coordinated assessment of differential vulnerability to capture on the basis of method and species would remove the burden of data collection from operational banding efforts. Finally, we recommend that as part of a robust study assessing in-season recaptures and banding age ratios, comparisons with harvest age ratios and use in population models occur to determine utility in population modeling efforts.

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.

Data S1. An Excel file of all capture and recapture data and associated variable definitions. We obtained capture and recapture data for mallards from nine locations across the mid-continent and western breeding range in the United States and Canada and for wood ducks at two locations in Tennessee and national wildlife refuges. Data were obtained for a single year at each location, either one of 2018, 2021, or 2022. We captured mallards and wood ducks using baited swim-in traps during August and September and defined recaptures as birds recaught after initial capture in the same year. Therefore, a bird caught in 2022 that was initially banded in 2021 would not be considered a recapture until a subsequent recapture in 2022.

Reference S1. Brownie C, Anderson DR, Burnham KP, Robson DS. 1985. Statistical inference from band recovery data: a handbook. Washington, D.C.: U.S. Department of the Interior, Fish and Wildlife Service.

Reference S2. Carney SM. 1992. Species, age, and sex identification of ducks using wing plumage. Washington, D.C.: U.S. Department of the Interior, U.S. Fish and Wildlife Service.

Available: https://www.fws.gov/media/species-age-and-sex-identification-ducks-using-wing-plumage (3,116 KB PDF)

Reference S3. Munro RE, Kimball CF. 1982. Population ecology of the mallard: VII. Distribution and derivation of the harvest. Washington, D.C.: U.S. Department of the Interior, Fish and Wildlife Service Resource Publication 147.

Available: https://pubs.usgs.gov/unnumbered/5230180/report.pdf (11,072 KB PDF)

Reference S4. Smith GJ. 2013. The U.S. Geological Survey Bird Banding Laboratory: an integrated scientific program supporting research and conservation of North American birds. Washington, D.C.: U.S. Geological Survey Open-File Report 2013–1238.

Available: http://pubs.usgs.gov/of/2013/1238/ (2,886 KB PDF)

Reference S5. Szymanski ML, Dubovsky JA. 2013. Distribution and derivation of the blue-winged teal (Anas discors) harvest, 1970–2003. Washington, D.C.: U.S. Department of the Interior, Fish and Wildlife Service, Biological Technical Publication FWS/BTP-R6017-2013.

Reference S6. U.S. Fish and Wildlife Service. 2022a. Adaptive harvest management: 2023 hunting season. Washington, D.C.: U.S. Department of Interior.

Available: https://fws.gov/project/adaptive-harvest-management (1,491 KB PDF)

Reference S7. U.S. Fish and Wildlife Service. 2022b. Western Canada cooperative waterfowl banding program, Final Report.

Available: https://www.fws.gov/library/collections/2022-waterfowl-banding-operations-field-reports (5,734 KB PDF)

We acknowledge the significant effort in collecting recapture data outside normal banding activities by all field crews. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the U.S. Fish and Wildlife Service or other respective agencies. We thank the Associate Editor, T. Riecke, and one anonymous reviewer for their helpful comments in improving this manuscript.