AI-driven opportunities and challenges in lake remote sensing

Lakes act as sensitive climate change indicators and critical regulators in global biogeochemical cycles. Yet, the monitoring and understanding of the responses of lake systems to climatic and anthropogenic stressors are constrained by data and general models. The artificial intelligence (AI) revolution has catalyzed paradigm shifts in lake remote sensing, achieving global-scale monitoring capabilities and enabling retrievals of non-optically active parameters through advanced models from multi-source datasets. Current challenges emerge in two key dimensions: (1) disconnection between radiative transfer mechanisms and black-box AI models hindering cross-regional transferability, and (2) overreliance on complete observations constraining future prediction. To bridge these gaps, we propose an integrated framework emphasizing: (i) Coupling physical mechanism and AI models synergizing radiative transfer theory with deep learning architectures to enhance interpretability, (ii) Unified data cubes harmonizing satellites, drones, and Internet of Things (IoT) sensor networks through adaptive fusion, and (iii) Embedding causal AI for real-time water quality forecasting and century-scale climate scenario projections. Implementing this roadmap will establish a new generation of intelligent observation-prediction systems, providing critical scientific and practical support for lake management under accelerating global change. • AI advances continental-to-global scale lake dynamic observations. • AI enables the capabilities of retrieving non-optical active matters in lakes. • A roadmap integrating multi-source data is presented to bridge monitoring-prediction gaps. • A coupling mechanism and AI algorithm paradigm for lake remote sensing is proposed.