Estimating harvest when hunting bag data are reported by area rather than individual hunters: A Bayesian autoregressive approach

Harvest estimation is a central part of adaptive management of wildlife. In the absence of complete reporting, statishods are required to extrapolate from the available data. We developed a Hierarchical Bayesian framework tailored for partial reporting where hunting areas covered by reporting hunting teams are available. The framework accounts for autocorrelation at the national, county, and hunting management precinct levels. We derived and evaluated an approximation for the probability of harvest on non-reported areas under a non-linear relationship between harvest area per team and harvest rate. We applied the framework to reports of red fox (Vulpes vulpes), wild boar (Sus scrofa), common eider (Somateria mollissima), and grey partridge (Perdix perdix) harvest in Sweden from the hunting years 1997/1998–2019/2020. The approximation was evaluated and determined to be sufficiently accurate. We showed that accounting for autocorrelation in harvest reduced uncertainty and increased predictive accuracy, particularly for game hunted in low numbers and variably between teams. The analysis also revealed that hunting rate has a sub-linear relationship with a team’s area for all considered species. Further, the framework revealed substantial differences across regions in terms of parameters modeling the distribution of huntable land across teams as well as parameters modeling harvest rates. We conclude that the framework is useful to detect shifts in hunting rates and/or practices.