Eco- and Earth system simulation remains challenging due to heterogeneous, high-dimensional, nonlinear, and interacting processes across scales. Traditional process-based models struggle with scalability, incomplete representations, and calibration at large scales. Meanwhile, purely data-driven machine learning approaches can lack physical consistency, interpretability, and robustness under sparse, shifting, or out-of-sample conditions.

Knowledge-guided machine learning (KGML) has emerged as a promising paradigm that integrates scientific understanding with machine learning to address the above challenges. Recent advances have demonstrated broad success in ecosystem and Earth sciences, spanning carbon-water-nitrogen cycling, weather and climate prediction, and land-atmosphere interactions. We invite contributions advancing KGML across agricultural and natural ecosystems and broader Earth system domains, including, but not limited to, hybrid and differentiable models, causal learning, data assimilation, uncertainty quantification, scientific discovery, and decision support. This session aims to broaden participation, accelerate ecosystem and Earth sciences, and help shape future directions in AI for science.

• A - Atmospheric Sciences
• EP - Earth and Planetary Surface Processes
• GC - Global Environmental Change
• H - Hydrology