Using Joint Models of Fisher Targeting and Resource Abundance to Account for Preferential Sampling in Fisheries

• Activity Number: 559 - Preferential Sampling of Environmental and Ecological Processes
• Type: Invited
• Date/Time: Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): James Turner Thorson* and Paul Conn and Devin Johnson and John Best
• Companies: Northwest Fisheries Science Center, National Marine Fisheries Service, NOAA and Marine Mammal Laboratory, Alaska Fisheries Science Center, NOAA, NMFS and Alaska Fisheries Science Center (NOAA) and School of Aquatic and Fishery Sciences, University of Washington
• Keywords: Spatio-temporal model; preferential sampling; fisheries science; stock assessment; catch-per-unit-effort; multivariate

Abstract

Commercial fishers record catch and fishing effort, and fisheries scientists analyze catch-per-unit-effort data (CPUE) to identify trends in fish abundance. CPUE data are increasingly analyzed using spatio-temporal models, which involve assuming that the location of data are ignorable. However, fishers are likely to target areas with high fish density (known as "preferential sampling", PS), which will bias predictions of density. In this talk, we discuss efforts to ameliorate bias from PS by jointly modeling population density and sampling intensity. We first discuss a recent simulation-experiment involving a log-Gaussian Cox process. We show that ignoring moderate levels of PS can cause a 40% bias in estimates of average population density. By contrast, jointly modeling sampling intensity and target density can largely mitigate this bias. We then discuss ongoing research using multivariate spatio-temporal models to mitigate bias for abundance-indices used in stock assessments. This research seeks to demonstrate that joint models are available for use by fisheries scientists worldwide, to routinely explore the potential importance of PS in stock assessments.