American Woodcock Symposium

A Bayesian Hierarchical Model for Estimating American Woodcock Harvest

Todd W. Arnold

Department of Fisheries, Wildlife and Conservation Biology

DOI: https://doi.org/10.24926/AWS.0106

---

Abstract

Estimates of total harvest help inform harvest management decisions, but such data are also useful for estimating population size and composition in demographic models. Historical estimates for U.S. harvest of American woodcock (Scolopax minor; hereafter woodcock) are available from 2 separate surveys: the 1964−2001 duck stamp survey (DSS) that sampled woodcock hunters who also hunted waterfowl, and the 1999−2016 Harvest Information Program (HIP) that sampled all licensed woodcock hunters. During overlap years (1999−2001), HIP estimates of total woodcock harvest were approximately twice as large as DSS estimates, but with only 3 years of overlap there was little potential to develop robust correction factors for historical DSS data. I developed a model of historical woodcock harvest that posited 3 groups of woodcock hunters, including those who always, sometimes, or never hunted waterfowl. During the HIP survey all 3 groups were included in harvest surveys; during the DSS years, however, only woodcock hunters who always hunted waterfowl were reliably sampled during all years, but I used annual duck stamp sales as a covariate to help predict harvest by woodcock hunters who hunted waterfowl irregularly. Using a reverse-time (2016 to 1964) model that assumed these 3 proportions of harvest remained constant through time, I was able to estimate total harvest in all years by estimating the latent component of harvest by non-waterfowl hunters. Averaged over all harvest jurisdictions, this model estimated that hunters who always, sometimes, or never hunted waterfowl contributed 43%, 32%, and 25% of the total woodcock harvest. Using these relationships, I estimated total harvest during all years (1964−2016) using data from both harvest surveys, although estimates based only on DSS data had greater uncertainty. In combination with band recovery data and harvest composition from the Parts Collection Survey, analysts could use estimates of historical harvest to estimate population size, composition, fecundity, and survival dating back to the initiation of harvest surveys in 1964.