Author(s): Abbas El Hachem; Martin Henning; Christian Noss

Linked Author(s): Martin Henning

Keywords: Upstream migration; Individual-based model; Deep learning; LSTM; CFD

Abstract: Given the urgent need for solutions to upstream and downstream fish migration at barriers, there is a growing demand for studies on fish behavior. Since lab and field experiments are time and cost intensive, results of such studies are used to develop (individual based) models to assess fish movement based solely on hydraulic scenarios. This study uses deep neural networks to reproduce fish tracks from experiments with two small bodied potamodromous fish species at a fishway entrance. A long short -term memory model was applied to model the relationship between fish movement and hydraulic conditions observed using high-resolution data. The results show that the model performed well, with simulated tracks matching observed movement patterns (fish entering, turning around or being drifted by the flow jet) and similar entrance success ratio. The results support the utility of the model for a better analysis of fishway entrance designs.

DOI: https://doi.org/10.64697/iahr.proc.ise2026.abs.201

Year: 2026