Landslide susceptibility in the Western Ghats controlled by terrain and tourism driven urbanisation mapped with remote sensing and machine learning
This study combines remote sensing and machine learning to demonstrate that tourism-driven urbanization, alongside terrain characteristics, significantly increases landslide susceptibility in the Western Ghats, providing quantitative evidence to support regulatory actions against construction in high-hazard zones.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
The Western Ghats, a towering mountain chain running along India's western coast, is a place of breathtaking beauty and terrifying instability. This ancient range, recognized as a UNESCO World Heritage Site, is home to some of the most intense rainfall on Earth during the monsoon season. For decades, scientists have understood that heavy rain is the immediate spark that sets off landslides here. When the soil becomes saturated, gravity takes over, and the earth slides down the steep slopes. However, a growing body of evidence suggests that rain is not the only actor in this drama. The way humans change the land—by cutting down forests to build roads, houses, and hotels—may be loading the gun, making the terrain far more likely to fail when the rain finally comes. This shift in perspective moves the conversation from purely natural disasters to a mix of nature and human choice, asking whether we can manage our construction to keep people safe.
In a recent study focused on this critical region, researchers set out to measure exactly how much tourism-driven construction contributes to landslide risk. The team, led by Vinay Kellengere Shankarnarayan, examined four specific hotspots known for frequent landslides: Wayanad in Kerala, and the Kodagu, Sakleshpur, and Kolhapur–Satara belts in Karnataka and Maharashtra. These areas share a common geography of steep slopes and thick, weathered soil, but they differ in how quickly they are being developed. The researchers wanted to see if they could use satellite images and computer models to separate the influence of the natural landscape from the influence of human building. They were particularly interested in the rapid expansion of resorts and homestays, which often replace deep-rooted forests with concrete and cleared land, potentially weakening the ground's ability to hold together.
To build their picture of the landscape, the team turned to a vast archive of satellite images taken over two decades, from 2005 to 2026. They used free, high-resolution images from the Landsat satellites, which act like a camera in the sky, capturing the Earth in different colors of light. By analyzing these images, the researchers could see how the land cover changed over time. They looked for specific signs of change: a drop in the greenness of the vegetation, which indicates forest loss, and a rise in the brightness of the ground, which signals the spread of buildings and paved surfaces. They combined this visual data with a detailed digital map of the terrain's shape, including its steepness and roughness. This created a rich dataset that described not just where the mountains were, but how humans had altered them.
The researchers then fed this data into powerful computer programs designed to learn from patterns, similar to how a student learns to recognize shapes by studying many examples. They taught these programs to recognize the difference between slopes that had slid and slopes that had remained stable. The computer was shown 516 documented landslides from official government records, along with an equal number of safe spots, and asked to figure out what made the dangerous spots dangerous. The goal was to see if the computer could predict where landslides would happen based on the terrain and the land use. The results were striking. The computer models, particularly one called Random Forest, became highly accurate at predicting risk, correctly identifying the location of nearly 90% of the known landslides when tested on new data.
The most revealing part of the study was understanding what the computer learned. The model confirmed that the natural shape of the land is the most important factor; steep, rough slopes are inherently more likely to slide. However, the study also found that human construction plays a significant and measurable role. The presence of buildings and cleared land ranked as the third most important factor in predicting a landslide, coming in just behind the steepness of the slope. This suggests that the act of building on these fragile hills is not just a passive background detail but an active driver of risk. The data showed a clear trend: between 2005 and 2026, the area of dense forest in the Wayanad region dropped from 90% to 79%, while the area covered by plantations and buildings grew. This conversion of forest to built-up land created a direct link between tourism development and increased instability.
The findings provide a scientific backbone for recent policy decisions in the region. Following a catastrophic landslide in July 2024 that claimed over 300 lives in the Mundakkai–Chooralmala corridor, local authorities ordered the demolition of seven resorts located in a high-risk zone. The computer-generated maps from this study independently confirmed that these specific resorts were situated in areas of very high susceptibility, validating the decision to remove them. The research indicates that while we cannot control the monsoon rains, we do have control over where and how we build. By regulating construction in ecologically sensitive areas, particularly on steep slopes where forests have been removed, it is possible to reduce the likelihood of future disasters. The study concludes that managing the built environment is just as crucial for safety as forecasting the weather, offering a practical path forward for protecting lives in one of the world's most beautiful and vulnerable mountain ranges.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.