Dynamic Rainfall Thresholds for Landslide Early Warning System in Progo Catchment, Java, Indonesia

Landslide occurrences are governed by precondition factors and triggering factors. Hence, it is desirable to include physical parameters representing precondition factors in determining thresholds over which landslides are likely to occur. In the case of rainfall-triggered landslides, such parameters include soil properties and land cover information. However, high-resolution data required for a physical-based approach are rarely readily available for a large area, especially in developed countries. Therefore, in developing a landslide early warning system (LEWS) for a large area, rainfall thresholds are derived by optimizing the usage of rainfall datasets.

This study aims to derive rainfall thresholds from a meteorological perspective regarding rainfall event characteristics (e.g., cumulative rainfall, intensity, duration) that result in trigger the landslides in Progo Catchment in Java, Indonesia.  We explore various hourly rainfall datasets, including rain gauge measurements and satellite-based rainfall products (e.g., the Japan Aerospace Exploration Agency’s Global Satellite Mapping of Precipitation/GSMaP and the Climate Prediction Center/National Oceanic and Atmospheric Administration’s morphing technique/ CMORPH), to derive the thresholds. The effect of rainfall event characteristics is assessed by clustering the rainfall event types and preceding conditions associated with different triggering mechanisms leading to the landslide occurrences. The rainfall thresholds are then derived using the frequentist method for each group, hence “dynamic.”