Nepal experiences one of the highest landslide frequencies in the Himalayan region because of its steep topography, active tectonics, fragile geology, and intense monsoon rainfall. Landslides account for more than 31% of disaster-related fatalities in the country, underscoring the urgent need for reliable operational early warning systems. This study presents the evolution and application of a multi-scale rainfall threshold framework for landslide prediction, integrating national, provincial, district, and municipal-scale thresholds to improve landslide early warning across Nepal.
The national empirical rainfall threshold, first developed for the Nepal Himalaya (Dahal and Hasegawa, 2008) using historical landslide inventories and rainfall records, established an intensity-duration relationship of I = 73.90D⁻⁰·⁷⁹, indicating that rainfall exceeding approximately 144 mm within 24 hours represents a critical condition for widespread landslide occurrence. This threshold has been adopted by the Government of Nepal as the basis for regional landslide early warning.
To enhance regional forecasting accuracy, the rainfall threshold framework was extended from the national to provincial, district, and municipal scales using spatio-temporal landslide inventories, IMERG satellite rainfall, Automatic Weather Station data, field investigations, Google Earth interpretation, local records, and rainfall reconstruction. In Bagmati Province, more than 600 documented landslides were analyzed to derive rainfall thresholds under different non-exceedance probabilities, with the 20% threshold showing the best predictive performance through cross-validation and real-time testing. More than 1,200 landslides were inventoried across Sindhupalchok and Rasuwa districts, with over 200 events field-verified. Local thresholds, including I = 41.029D⁻⁰·⁶⁵¹ for Helambu and I = 52.476D⁻⁰·⁷⁴³ for Panchpokhari Thangpal, significantly improved prediction accuracy, reduced false alarms, and strengthened operational landslide early warning systems in Nepal.

关键词