Bayesian species recognition and abundance estimation: unravelling the mysteries of salmonid migration in the Teno River

Abstract

In Teno River, annual sonar monitoring is used to estimate the abundance of three salmonid species: Atlantic salmon, pink salmon, and sea trout. However, the size distribution of these species is partially overlapping making species recognition impossible from plain sonar data. A Bayesian model was developed to tackle this problem and to estimate abundance and migration timing for these three species. The model integrates multiple sources of data including catch, video count, daily average school sizes, and expert knowledge. Given the limited catch and video statistics for 2021, the use of school size data and expert knowledge on migration intensity enhanced the estimation when other data sources were unavailable. The model estimated a median of 11.8 thousand Atlantic salmon, 6.6 thousand sea trout, and 52.0 thousand pink salmon migrating into the river during 2021. These findings offer a more accurate representation of species distribution, support future conservation and management efforts, and provide a modelling-based solution for distinguishing similarly sized species from sonar counting data.