At the Junction Between Deep Learning and Statistics of Extremes: Formalizing the Landslide Hazard Definition

Abstract The most adopted definition of landslide hazard combines spatial information about landslide location (susceptibility), threat (intensity), and frequency (return period). Only the first two elements are usually considered and estimated when working over vast areas. Even then, separate models constitute the standard, with frequency being rarely investigated. Frequency and intensity are dependent on each other because larger events occur less frequently and vice versa. However, due to the lack of multi‐temporal inventories and joint statistical models, modeling such properties via a unified hazard model has always been challenging and has yet to be attempted. Here, we develop a unified model to estimate landslide hazard at the slope unit level to address such gaps. We employed deep learning, combined with extreme‐value theory to analyze an inventory of 30 years of observed rainfall‐triggered landslides in Nepal and assess landslide hazard for multiple return periods. We also use our model to further explore landslide hazard for the same return periods under different climate change scenarios up to the end of the century. Our model performs excellently (with an accuracy of 0.78 and an area under the curve of 0.86) and can be used to model landslide hazard in a unified manner. Geomorphologically, we find that under climate change scenarios (SSP245 and SSP585), landslide hazard is likely to increase up to two times on average in the lower Himalayan regions (Siwalik and lower Himalayas; ≈110–3,500 m) while remaining the same in the middle Himalayan region (≈3,500–5,000 m) whilst decreasing slightly in the upper Himalayan region (≳5,000 m) areas.