Climate change is reshaping hydro-climatic extremes by disturbing the long-term stability of hydrological systems, a pattern known as non-stationarity. This non-stationarity is far more pronounced wherever watersheds are physically disturbed. Wildfire is one of the most rapid forms of such disturbance. It strips vegetation, alters soil properties, and reduces the land's ability to absorb water. When heavy rainfall follows, floods expected once every 100 years can become up to three times larger in the first year after a fire. This risk is intensifying. Warming raises the likelihood of severe fire seasons followed within a year by extreme rainfall. Compound climate hazards are also becoming more frequent in fire-prone regions such as the Mediterranean. Most models, however, evaluate hazards one at a time. They rarely capture how hazards interact and cascade. This leaves the added complexity that a warming climate brings to post-fire hazard planning largely unaccounted for. This leaves the added complexity that a warming climate brings to post-fire hazard planning largely unaccounted for. This session aims to bring together researchers exploring how Earth observation, machine learning, and deep learning methods can model and monitor these compound and cascading hydro-climatic extremes under climate change. It welcomes hydrologists, remote sensing specialists, climatologists, and machine learning researchers to discuss practical approaches for understanding and predicting these evolving fire-to-flood extremes. By fostering interdisciplinary discussions, we aim to form new collaborations and identify outstanding research questions like: Which climatic regimes and conditions foster this fire-to-flood interaction? What role do geographical and anthropogenic influences play in this? What role can explainable AI play in detecting and providing early warnings for these non-stationary compound hydro-climatic extremes? How can data-scarce regions be better modeled using deep learning and Earth observation?