A Spatiotemporal Physics-Motivated State-Space Model of Lake Temperature Profiles

Area of Science:

• Environmental Science
• Ecology
• Data Science

Background:

• Lake temperature stratification is crucial for dissolved oxygen and freshwater ecosystem health.
• Predicting thermal profiles is essential for ecological process understanding.
• Existing models face trade-offs between physical consistency and computational efficiency.

Purpose of the Study:

• To develop an accurate and interpretable model for predicting lake water temperature profiles.
• To integrate physics-based principles with data-driven approaches for enhanced prediction.
• To assess the impact of climate change on lake thermal structures.

Main Methods:

• A mixture-of-experts framework combining a physics-motivated neural network and a deep neural network.
• Physics-motivated component enforces physically consistent vertical profiles.
• A gating mechanism dynamically integrates expert predictions.

Main Results:

• The framework accurately predicts the timing, magnitude, and vertical structure of thermal stratification across diverse lakes.
• The model successfully reproduces empirical patterns while maintaining physical consistency.
• Future climate change projections indicate a warming trend in near-surface lake temperatures.

Conclusions:

• The developed AI framework offers a powerful tool for predicting lake thermal dynamics.
• This approach improves our ability to understand and manage freshwater ecosystems.
• The findings highlight potential ecological consequences of climate change on lakes.