The Western Ghats account for almost 60% of the landslides reported in India, most of which are caused by heavy and prolonged rainfall. Photo credit: K. MURALI KUMAR
Researchers at the Karnataka Institute of Technology (NITK) in Suraskar have developed an integrated landslide early warning framework specifically designed for the Western Ghats, one of the most landslide-prone regions in India.
The system, called Slope Vulnerability and LandSlide Assessment (SVALSA), combines rainfall analysis, real-time monitoring of soil behavior and surface movement, and machine learning to provide reliable landslide warnings while reducing false alarms.
Why existing alerts often fail
The Western Ghats account for nearly 60% of the landslides reported in India, most of which are caused by heavy and prolonged rainfall. Recent disasters, including the Wayanad landslide in July 2024, have highlighted the limitations of existing warning systems and the need for more accurate site-specific warnings.
Currently, landslide warnings in India are primarily based on rainfall thresholds, with warnings being issued if rainfall exceeds certain intensity or duration limits. Although these systems are useful, they often cannot explain what is happening inside the slope itself. As a result, poor soil conditions, even with moderate rainfall, can lead to frequent false alarms or missed impending failures.
Beyond rain-only warnings
The SVALSA framework addresses this gap beyond rain-only alerts. It integrates hydrological data, soil strength behavior, and visible surface deformation into a single decision system that reflects how the slopes of the Western Ghats actually fail. More than 90% of landslides in this region occur in residual soils formed from weathered rocks, where changes in water content and soil suction play an important role in slope stability.
The SVALSA device is currently patent pending. The study was developed by Varun Menon under the supervision of Sreevalsa Kolathayar with financial support from the Department of Science and Technology (DST), IMPRINT (Impactful Research Innovation and Technology) and National Technology Textiles Mission (NTTM) under the Ministry of Textiles.
3 stage alarm system
The system operates through a three-stage alert mechanism implemented as a Python-based algorithm on a compact processing unit.
In the first stage, government-recorded rainfall data and historical landslide data are analyzed using a machine learning technique called K-Nearest Neighbors (KNN). The model compares current rainfall to events that caused previous landslides and filters out low-risk situations to reduce unnecessary warnings. Tests have shown this method to be highly accurate.
If rainfall conditions appear to be hazardous, the second step is to assess soil stability using a modified version of the simplified Bishop method that takes into account soil moisture and suction based on unsaturated soil mechanics. Laboratory tests confirmed that slope stability decreases as the soil absorbs more water.
The final step is to monitor surface motion through image analysis using particle image velocimetry (PIV). It has been found that a sudden increase in ground movement is a precursor to an impending landslide, often before visible collapse occurs.
According to the researchers, using all three metrics in combination significantly increases the reliability of the warning system.
The SVALSA framework also includes a deployable low-power monitoring device that integrates rainfall sensors, soil moisture probes, imaging units, and a compact processor that can generate real-time alerts and communicate remotely with authorities.
Where it can be used
Researchers say the framework is particularly suitable for hilly roads, highways, railway cuts, villages located on steep slopes, and critical infrastructure corridors across the Western Ghats. Its implementation could improve disaster preparedness, support timely evacuation, and significantly reduce loss of life and property.
issued – Dec 28, 2025 7:56 PM IST
