Multimodal Temporal AI Reasoning for Climate-Driven Landform Change Detection via Remote Sensing

Climate-driven landform evolution is governed by complex deformation processes resulting from extreme climatic events and long-term environmental transitions. Conventional remote sensing approaches that rely on static imagery or single-sensor observations often fail to distinguish permanent geomorphic change from short-term spectral or seasonal variability. This chapter presents a multimodal temporal reasoning framework that integrates optical, SAR, LiDAR, and hyperspectral data within a unified sequence-learning architecture conditioned by climatic drivers. Modality-aware spatial encoding and attention-based temporal modeling enable coherent reconstruction of landform dynamics across riverine, coastal, glacial, and deltaic environments. The proposed approach enhances robustness to missing data, improves temporal consistency, and advances monitoring from retrospective change detection toward predictive geomorphology, supporting early-warning systems and climate-resilient land-use planning.