Developing a prototype landslide early warning system for Darjeeling Himalayas using SIGMA model and real-time field monitoring

摘要

Rainfall thresholds are commonly utilized to forecast landslides using the historical relationship between occurrence of slope failures and rainfall in an area. SIGMA (Sistema Integrato Gestione Monitoraggion Allerta) is a rainfall threshold model, which uses the statistical distribution of rainfall for forecasting the occurrence of landslides. The threshold curves are functions of standard deviation of the cumulated rainfall data, taking into account both long-term and short term-rainfall. To overcome the limitations of statistical rainfall threshold, the real-time monitoring data from MicroElectroMechanical Systems (MEMS) tilt sensors have been integrated with SIGMA model using a decisional algorithm for a test site (Kalimpong) in Darjeeling Himalayas, in the northeastern part of India. Three different models, the SIGMA model, tilt meter readings and the combination of both are compared quantitatively using the precipitation and landslide data of Kalimpong town between July 2017 and September 2020. The results indicate that the integration of tilt meter readings has lowered the number of false alarms issued by SIGMA model from 70 to 38 in the studied period, with an increase in the likelihood ratio from 18.10 to 20.23. The Receiver Operating Characteristic (ROC) curves indicate that the combined approach has the best performance among the models considered in this study, with an area under the curve 0.976. The proposed method was found to have better performance than the other rainfall thresholds derived for Kalimpong region so far, and the prototypal model can be further fine-tuned to develop an operational Landslide Early Warning System (LEWS) for the region.