Ecosystem models remain limited by how they represent key ecological and biogeochemical processes and their responses to climate variability, extremes, and long-term environmental change. Persistent challenges include representing vegetation demography and trait diversity, acclimation and adaptation, stress responses and disturbances, carbon allocation, and coupled carbon–nutrient–water cycles. Uncertainties and biases in these process representations limit our ability to predict ecosystem dynamics and further climate feedbacks. Meanwhile, rapidly expanding Earth observations, data-synthesis efforts, and advances in artificial intelligence (AI) are creating new opportunities for ecosystem process development, including identifying missing process connections, emulating computationally demanding components, and complementing certain processes.
This session brings together scientists working on process-based, data-driven, and hybrid approaches to advance process understanding in land ecosystem modelling. Contributions may address model development, benchmarking and evaluation, uncertainty reduction, or the integration of ecological knowledge and observations into models.
We invite abstracts addressing one or more of the following themes:
1. Advances in representing ecosystem responses to climate variability, extremes, disturbances, or long-term environmental change;
2. Improved representation of vegetation demography and dynamics, species interactions, functional diversity, acclimation, adaptation, and carbon allocation;
3. Advances in modelling biogeochemical processes and interactions, including coupled carbon, nutrient, and water cycles across ecosystems and scales;
4. Applications of AI and machine learning to ecosystem modelling, including process emulation or replacement, parameter estimation, and discovery of missing or poorly represented processes;
5. Emergent constraints, causal inference, and other observation-based approaches for reducing uncertainty in ecosystem responses and land–climate feedbacks;
Session format
The session will include oral and poster presentations. As conveners from five universities, we aim to create a forum to discuss transferable model developments, evaluate emerging datasets and relationships, and share knowledge across modelling frameworks. We plan to host a session dinner to facilitate discussion and networking further, and we will announce more details later.