Accounting for Sampling Uncertainty in Species Apportionment of Hydroacoustic Data

Pairing hydroacoustic and traditional fisheries surveys provides a spatial description of fish populations while estimating abundance. However, apportioning species in mixed communities is a substantial challenge. Target strength (TS) variability and limited traditional sampling complicate partitioning hydroacoustic fishes into species and size groups. We propose an apportionment method that accounts for sampling uncertainty, and apply this to Lake Erie walleye. We paired a hydroacoustic survey with an ODNR-DOW annual fall gillnet survey in Lake Erie’s western and Sandusky sub-basins during 2014, collecting hydroacoustic data from 34 5-minute grids and 29 gillnet sites. We used a Bayesian hierarchical logistic regression and gill net catches to estimates the probability any fish was a walleye, given its length within each 5-minute grid. We quantified uncertainty in mean TS estimates from tracked fish using a normal distribution and used Love’s (1971) multispecies equation to estimate the probability that a fish was a specific TL given its TS. We used a Markov chain Monte Carlo simulation to sample the product of these distributions returning the probability that an acoustically sampled fish was a walleye. This method provided spatially explicit abundance estimates with associated uncertainty, enhancing the information available to fisheries managers from paired surveys.