Automated Coastal Ice Mapping with SAR Can Inform Winter Fish Ecology in the Laurentian Great Lakes

Abstract

Many freshwater lakes in the temperate zone undergo annual freeze-thaw cycles. Climate change has disrupted these patterns and altered habitat for many species including ecologically, economically, and culturally valuable fish species. To understand the relationship between ice cover and aquatic species, suitable data can be derived from remote sensing. We developed a novel ice classification method with minimal user input using freely available Sentinel-1 data and an adjacent and time-coincident validation dataset. Using image object segmentation and a random forest classifier, ice conditions were classified correctly with >85% overall accuracy. Our ice mapping efforts coincided with a telemetry dataset of tagged Walleye (Sander vitreus) and Northern Pike (Esox lucius) in Hamilton Harbor in western Lake Ontario. Between years with low and high ice covers (2017 and 2019, respectively), we found Walleye appeared to reduce their area of movement when the harbor was covered in ice. Our ice mapping tool can provide a quick and consistent method for agencies to adopt for freshwater resource management as well as provide ice cover information in coastal areas that are important overwintering habitat for many fishes.