MACHINE LEARNING SYSTEM TO PREDICT LANDSLIDES IN WESTERN GHATS

NEWS –  Researchers at the National Institute of Technology Karnataka (NITK), Surathkal have developed an advanced, machine learning–based early warning framework to predict landslides in the Western Ghats—one of India’s most landslide-prone regions.

HIGHLIGHTS

Why It Matters?

• Western Ghats account for ~60% of reported landslides in India
• Most are triggered by intense and prolonged rainfall
• Recent incidents (e.g., 2024 Wayanad landslide) exposed limitations of rainfall-only alert systems
• Need for accurate, site-specific and reliable warning mechanisms

About the New System

Slope Vulnerability and LandSlide Assessment (SVALSA)

• Developed using:
  • Rainfall analysis
  • Real-time soil behaviour monitoring
  • Surface deformation tracking
  • Machine learning
• Designed to:
  • Reduce false alarms
  • Provide reliable community alerts

Beyond Rainfall Threshold Systems

• Traditional alerts rely only on rainfall levels
• SVALSA integrates:
  • Hydrological data
  • Soil strength parameters
  • Sub-surface deformation signals
• More than 90% landslides in Western Ghats occur in residual soils, where soil moisture & suction critically influence slope stability

How the System Works

• Funded by Department of Science & Technology (DST), IMPRINT, and National Technical Textiles Mission
• Uses a three-stage warning mechanism
• Powered by Python-based processing
• Employs K-Nearest Neighbour (KNN) machine learning model
• Currently under patent application

Significance

• Enhances disaster preparedness
• Strengthens community resilience
• Supports government decision-making
• Aligns with climate risk adaptation strategies