Inferring benthic fish weights from photographic pixel areas to support automated biomass estimation

Session: Poster session

Jennifer Wardell, US Geological Survey, [email protected]
Peter Esselman, U.S. Geological Survey, [email protected]

Abstract

Photographic methods for quantifying round goby, sculpins, and other benthic species are becoming increasingly common in the Great Lakes.  Common measures derived from photographic and video sources include species presence absence and numeric abundance/density.  However, biomass density (body weight per unit area) is a preferable measure of abundance because it is a more direct measure of the energetic availability of prey for predators, and is a more appropriate value for use in food web  and bioenergetic modeling.  In this presentation we explore alternative measures to length in commonly-used length-weight regressions used to convert body size to weight for fishes.  Specifically, we explore the utility of ‘pixel area occupied’ as a substitute for length in length-weight regressions, in addition to fish body width for fish that may be partially obscured.  We present pixel area to body weight relationships for round goby, and deepwater sculpins—two species that are the target of ongoing efforts to automate quantification of benthic fish species using deep learning algorithms.

1. Keyword
round goby

2. Keyword
benthos

3. Keyword
fish