The body of remotely sensed observations is growing, while Earth system models move to km-scales. Together, these tendencies require land surface models to integrate heterogeneous observations, address incomplete process knowledge, and overcome computational constraints. Machine learning (ML) offers tools for addressing these challenges, and has demonstrated a strong ability to learn complex non-linear relationships from data when high-quality data is available. However, many land surface applications do not meet this condition and must instead rely on sparse, irregular, or imbalanced observations or represent processes that are only partially observable. Hybrid and AI-enabled models combine the flexibility of ML with physics-based knowledge to produce predictions that are both data-informed and physically constrained. The spectrum of physics-ML integration ranges from a physical model core that is extended by ML for parameterization, data assimilation, emulation, or bias correction, to an ML model core that is constrained through physical pretraining, regularization, causal pathways, or explicit equations. When and how physics-ML fusion is useful is highly use-case specific, and its potential for operationalisation is an open question. This session brings together researchers from ecosystem and land surface modeling, boundary layer meteorology, hydrology, Earth observation, and Earth system data science across this spectrum.

We welcome contributions including but not limited to:
hybrid process-ML models and learned process representations;
emulators, surrogate models, and reduced-order land model representations;
discovery of governing scientific equations from data;
differentiable and end-to-end trainable modelling frameworks;
AI-based models hard- or soft-constrained by physical principles or causal structure;
model-data integration, data assimilation, and uncertainty quantification.

We also encourage work on neural operators, foundation models, digital twins, and other AI-enabled approaches to land, ecosystem, and hydrological dynamics - especially studies that systematically compare modelling strategies, assess generalization and physical consistency, or identify conditions when physics-ML integration succeeds or fails.