Making do with less: Extending an acoustic-based time series of euphausiid abundance using an uncrewed surface vehicle with fewer frequencies

Abstract

Survey time series are used to track species- and ecosystem-level trends over time to support ecosystem-based fishery management. However, these recurring survey efforts are subject to unpredictable cancellations and reductions in effort. This occurred in 2020 when the COVID-19 pandemic forced the cancellation of the research vessel-based 2020 eastern Bering Sea (EBS) acoustic-trawl survey, which has provided an estimate of euphausiid abundance and distribution since 2004. As a partial replacement for this lost effort, three uncrewed surface vehicles (USVs) were used to collect acoustic data. In contrast to the standard vessel-based survey, which provides 4-frequency acoustic data, the USVs collected acoustic data at only two frequencies. This presented a challenge given that four frequencies are currently used to identify euphausiids in this time series. Here, we first evaluated two methods to provide comparable euphausiid abundance estimates using fewer acoustic frequencies. We found that a random forest classifier was able to produce abundance estimates comparable to those obtained in the vessel-based time series. This method was then used to estimate euphausiid abundance and distribution from the 2020 USV survey. We additionally estimated the increase in survey uncertainty due to the use of the random forest classifier and changes in the acoustic instruments. Together, this allowed for the EBS euphausiid abundance time series to be extended with fewer acoustic frequencies.