Side-Scan Sonar Imagery for Invasive Carp Population Modeling

For his project, MS student Cade Roach has collected side-scan sonar data before and after invasive carp removal efforts conducted by Missouri Department of Conservation and USFWS in the Lamine River, a tributary of the Missouri River, and by Illinois Natural History Survey in Pools 16-19 of the Upper Mississippi River. He is utilizing an assortment of computer vision and image segmentation methods to identify and enumerate fish from the sonar imagery. After compiling counts, Cade is estimating abundance using N-mixture models, a hierarchical technique which accounts for imperfect detection. As these models have primarily been used for sparse count data rather than dense populations, he is experimenting with different modeling frameworks, addressing overdispersion and spatial autocorrelation, and using spatial simulations to assess accuracy.

Cade has presented preliminary abundance estimates from the Lamine River at the Midwest Fish and Wildlife Conference in January 2025 and the 155th American Fisheries Society Meeting in August 2025. He continues to refine the carp enumeration process, expanding upon image segmentation techniques and exploring object detection via convolutional neural networks. Cade aims to compile this workflow into a deployable web-based tool that fisheries managers can use to obtain estimates of silver carp abundance from side-scan sonar data inputs. The main goal of this project is to develop a reliable and efficient method of abundance estimation for invasive silver carp populations, applicable throughout their range in the Upper Mississippi River Basin, in order to monitor these populations over time and evaluate removal success.