Dynamic rainfall thresholds for landslide early warning in Progo Catchment, Java, Indonesia

R. Satyaningsih, V. Jetten, J. Ettema, Ardhasena Sopaheluwakan, L. Lombardo, Danang Nuryanto

Research output: Working paperPreprintAcademic

21 Downloads (Pure)


High spatiotemporal resolution satellite data have been available to provide rainfall estimates with global coverage and relatively short latency. On the other hand, a rain gauge measures the actual rain that falls to the surface, but its network density is commonly sparse, particularly those that record at sub-daily records. These datasets are extensively used to define rainfall thresholds for landslides. This study aims to investigate the use of GSMaP-GNRT and CMORPH-CRT data along with automatic rain station data to determine rainfall thresholds for landslides in Progo Catchment, Indonesia, as the basis for landslide early warning in the area. Using the frequentist method, we derived the thresholds based on 213 landslide occurrences for 2012-2021 in the Progo Catchment. Instead of relying on a fixed time window to determine rainfall events triggering landslides, we consider a dynamic window, enabling us to adapt to the rainfall event responsible for landslides by extending or shortening its duration depending on the persistence of the rainfall signal. Results indicate that both GSMaP-GNRT and CMORPH-CRT products fail to capture high-intensity rainfall in Progo Catchment and overestimate light rainfall measured by rain gauge observations.
Nevertheless, when accumulated to define the rainfall threshold, the overall performance of GSMaP-GNRT and automatic rain station data in Progo Catchment is comparable. The rainfall measured at the stations performed slightly better than the estimated rainfall from GSMaP-GNRT, particularly at a probability exceedance level below 15%. In contrast, CMORPH -CRT performed the worst for all exceedance probabilities. The suitable exceedance probability for early warning purposes in Progo Catchment is 10% if it is based on the automatic rain station data. At this exceedance probability level, the threshold can adequately discriminate triggering/non-triggering rainfall conditions and produces the minimum false alarms and missed events.
Original languageEnglish
PublisherEarth ArXiv
Number of pages27
Publication statusPublished - 20 Jan 2023


Dive into the research topics of 'Dynamic rainfall thresholds for landslide early warning in Progo Catchment, Java, Indonesia'. Together they form a unique fingerprint.

Cite this