AI for Lake Ecosystem Modeling

The Challenge: AI is transforming science, but applying it to environmental problems is uniquely demanding. Natural systems are complex, with heterogeneous, often limited data. Environmental science also values explainability – understanding processes, not just making predictions. This requires specialized AI approaches that work effectively with environmental data while supporting scientific interpretation.

​

Our Approach: EDI partners with environmental and computer scientists to advance knowledge-guided machine learning (KGML), an AI approach tailored for scientific discovery. KGML combines domain expertise with deep learning algorithms to create models that are both accurate and interpretable. EDI supports this work by publishing high-quality environmental data, developing workflows for data aggregation and formatting, and creating benchmark datasets to advance the broader AI community.

​

Impact and Outcomes: These collaborations have produced models that predict lake temperature and dissolved oxygen more accurately than traditional approaches. Scientists are also working with EDI to develop foundation models for lake ecosystems that can be adapted for applications like water quality management, habitat assessment, satellite monitoring, and disease forecasting. By enabling these AI-driven solutions, EDI supports better-informed decisions for sustainable management and conservation of freshwater systems.