Modelling capture probability of Atlantic salmon (Salmo salar) from a diverse national electrofishing dataset: Implications for the estimation of abundance

Abstract

Estimates of juvenile fish abundance are commonly used to assess fish population health and to underpin assessment tools used in environmental and fisheries management such as the EU Water Framework Directive and EU Habitats Directive. Accurate and precise estimates of capture probability are important for estimating fish abundance from electrofishing data. This paper presents the findings of a maximum likelihood approach for modelling spatio-temporal variability in capture probability of Atlantic salmon in a large (2,837 samples) and diverse national electrofishing dataset. Capture probability was related to explanatory variables indicative of (1) equipment, personnel and protocols, (2) fish size, (3) fish behaviour, (4) habitat, and (5) time. A linear modelling approach allowed for flexible model specification, including smoothers and spatial effects, and rapid model specification, fitting and selection. The use of GIS variables as proxies for directly observed habitat characteristics allowed capture probability to be modelled using aggregated data where there were no common meta-data or habitat recording protocols. Modelled capture probabilities provided more precise estimates of abundance than sample-wise estimates and more accurate estimates than those obtained when assuming constant capture probability. Fisheries and environmental managers should routinely model capture probability when estimating fish abundance from electrofishing data, especially where datasets are geographically widespread, cover a large environmental range or have been collated across multiple organisations. Capture probability models can also facilitate the legitimate use of single pass electrofishing data in assessments of fish abundance providing that there is sufficient spatio-temporal coverage of multi-pass data.