Improving Lake Temperature Estimates for Midwestern Fisheries with Process-Guided Deep Learning

Improved estimates of lake water temperatures can benefit managers of midwestern fisheries. Water temperature controls growth and reproduction of fish, and water temperature measurements are commonly collected as part of aquatic monitoring campaigns to provide a measure of the ambient temperature environment. However, most lakes are unobserved or lacking consistent sampling during the multiple seasons and years necessary to understand change in fish communities. Our research team has developed new methods that combine the deep learning (the most advanced class of machine learning methods) with traditional process-based models in order to improve the accuracy and transferability of water temperature predictions. These new Process-Guided Deep Learning (PGDL) models have been shown to outperform existing models even when data are sparse or nonexistent. This webinar will provide background information on how these new techniques were developed, share use-cases for management decisions, and discuss future efforts to apply PGDL models in lakes and streams.