An international research team has developed an automated early warning system to forecast slope failures in Idukki, as heavy monsoon rains continue to trigger landslides across the Western Ghats. According to officials, researchers from the University of Mississippi, Michigan Technological University and the University of Kerala developed the system, and the study was published in a recent edition of the journal Environmental and Engineering Geoscience.

K.S. Sajinkumar, associate professor at the University of Kerala's Geology department, who was associated with the study, said the system — called the Google Earth Engine-based Landslide Early Warning System, or GLEWS — combines real-time satellite data, weather forecasts and live ground-level rainfall measurements to issue colour-coded public safety alerts. He said predicting landslides in mountainous terrain like Idukki has long been hindered by gaps in rainfall data, as satellites frequently underestimate intense, localised downpours while sparse ground weather stations cannot capture rainfall patterns across an entire district. GLEWS addresses this by blending multiple data streams on the Google Earth Engine cloud platform, he added.

The system draws on three-hour precipitation measurements from NASA's Global Precipitation Measurement mission and hourly rainfall forecasts from the Global Forecast System, along with real-time readings from eight automatic weather stations. Using a mathematical technique called "conditional merging," it calibrates satellite data against ground-gauge observations. Rainfall metrics are then overlaid onto a high-resolution, 30-metre slope stability map that calculates a Factor of Safety, with areas rated 1.5 or lower flagged as highly susceptible to collapse and zones above 2.0 considered stable.

To test its reliability, the team evaluated GLEWS against two major landslide events in Idukki. Dr. Sajinkumar said that during the August 15, 2018 floods and landslides, uncalibrated satellite data and standard coarse gridded rainfall maps detected only 8.9% of landslide conditions, but calibration using point-level observations from five India Meteorological Department rain gauges raised detection sensitivity to 72.6%. In a second-phase test analysing the July 7, 2023 landslide along the Gap Road stretch of Munnar, raw satellite data recorded 20 mm of daily rainfall, which tripled to 63.9 mm after ground-gauge calibration, he said. Originally prototyped in JavaScript, GLEWS has been ported to Python to enable automated data ingestion from weather stations, allowing it to run autonomously through Kerala's monsoon seasons, he added.