A KALMAN FILTER APPROACH TO ESTIMATING MARINE TURTLE INCIDENTAL TAKE RISK

摘要

A principle U.S. fisheries management concern is ensuring compliance with Endangered Species Act requirements to limit incidental takes of listed species to levels which do not result in potential jeopardy to existing population stocks. Part of the risk assessment process involves an analysis of historic takes (bycatch) of ESA-listed species relative to fishing effort. Sample bycatch per unit of effort (BPUE) ratios are used to gauge the take risk to ESA-listed species posed by future fishing effort.The rare-event nature of protected species takes complicates the interpretation of sample BPUE ratios. For illustration, a total of 23 leatherback turtles were incidentally taken by fishermen in the California-Oregon drift gillnet fishery from the 1990through 2006 fishing seasons in 7733 sets of fishing effort, with from 0 to 5 leatherback takes per season over the entire period. Should one conclude the incidental take risk was significantly higher during the season with 5 leatherback turtle takes than in seasons with 0 takes? To address this and related inference problems, we specify a state space model which describes the pattern of observed takes as a Poisson process conditional on the level of fishing effort. A time-varying Poisson rate parameter, expressed as BPUE, measures the time-varying level of take risk. A Kalman filter for Poisson processes is used to estimate the intertemporal variation in take risk from California-Oregon drift gillnet fishery observer data. A key question regards attributing the observed pattern of takes to randomness which drives fluctuations in observed takes versus intertemporal variation in the level of take risk.