Developing scalable spatial stream network models to support freshwater fish conservation efforts

Dr. Jess Kunke (Dept. of Mathematical Sciences, MSU)

08/28/2025 3:10 pm

Abstract: 

This talk will focus on scalable spatial models for fish populations, for which count and occurrence data have correlations induced by the stream network. Large-scale (regional and national) population estimates are impossible with existing spatial stream network models developed over the last two decades due to the computational costs of preprocessing, estimation, and prediction. We combine computationally efficient stream network preprocessing with nearest neighbor Gaussian processes and a fast spatial bootstrap (BRISC) to make spatial stream networks scalable.  We demonstrate our new method, S3N, on the Ohio River Basin. This is joint work with Julian Olden and Tyler H. McCormick.