In rural farming regions, water environments such as rice paddies and agricultural reservoirs play a key role in local ecosystems, agriculture, and climate. However, understanding how these water bodies heat up and cool down throughout the day and across seasons is complex. This temperature behavior affects plant growth, irrigation efficiency, and even carbon cycling. In this study, we used a new type of machine learning called Knowledge-Guided Machine Learning (KGML), which combines data-driven techniques with physical and environmental knowledge.

We focused on three applications: (1) simulating how rice paddies change temperature over time, (2) estimating water depth and plant growth conditions based on surface temperature data, and (3) predicting how heat moves vertically in agricultural reservoirs. The third application is especially important for understanding carbon storage and greenhouse gas emissions from these water bodies.

Our research shows that KGML can make reliable predictions while respecting real-world physics. This helps farmers, water managers, and environmental scientists better monitor and manage agricultural water systems under changing climate conditions. By connecting advanced technology with traditional water landscapes, we hope to support smarter, more sustainable farming.