Recent advances in artificial intelligence (AI) have transformed how we analyze and predict complex systems. Deep learning (DL), with its multi-layer neural networks, enables the discovery of intricate patterns beyond traditional methods, assuming that the needed data/evidence will never be complete and that new knowledge will always emerge. In the natural sciences, data-driven DL, along with self-supervised and transfer learning, leverages large unlabeled datasets to improve predictions. Integrating human-in-the-loop frameworks ensures that computational power is guided by expert knowledge, resulting in systems that are both accurate and trustworthy. Foundation models trained across scales offer new opportunities to simulate ecosystem dynamics under diverse conditions. This session will highlight advances in AI applications for understanding complex ecosystems, particularly under data limitations, and showcase their use in both basic and applied research.

• GC - Global Environmental Change
• H - Hydrology