Modeling the effects of fish migration on bias and variance in area-swept estimates of biomass: a vector-based approach

Abstract

Estimates of fish abundance from trawl and acoustic surveys are typically imprecise and may often be biased because of uncontrollable factors such as fish migration. In this paper, I present a model for the bias in area-swept biomass estimates when fish migration occurs as the survey is conducted. Bias in acoustic and trawl survey biomass estimates is treated as a function of the direction and speed of the vessel and the fish in each survey transect and results mainly from distortions in the observed area across fish aggregations. I use the model to evaluate the magnitude and direction of bias and interannual variability in biomass estimates from fish migration in systematic transect survey designs. For even relatively low fish migration velocities (<0.5 m/s), bias in estimated fish biomass can be very large (>500%). Furthermore, relatively small interannual variability in fish migration rates and vessel speeds (e.g., CV = 0.1 for each) can result in very large interannual error variability in biomass estimates (e.g., CV > 0.5). Designs with the least bias and variance in biomass estimates had the fastest vessel speeds along transects, more survey vessels, shorter transects, and transects aligned roughly parallel to the direction of fish migration.