Predictive Landslide Susceptibility Modelling Using Statistical Approaches in Mantsa River catchment, Dawro Zone, South-western Ethiopia
This study evaluates landslide susceptibility in the Mantsa River Catchment of southwestern Ethiopia using Frequency Ratio and Logistic Regression models, identifying key conditioning factors such as land-use and lithology, and demonstrating that the Frequency Ratio model achieves slightly higher predictive accuracy (AUC = 0.874) to support risk reduction and sustainable land-use planning.
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
In the highlands of Ethiopia, where the land rises sharply into rugged mountains and deep valleys, the ground is alive with a slow, dangerous movement. When heavy rains soak the soil, steep slopes can lose their grip, sending rock, mud, and debris sliding down toward villages and roads. This phenomenon, known as a landslide, is not just a sudden event but the result of a complex conversation between the shape of the land, the type of rock beneath the surface, the amount of rain that falls, and how humans use the soil. For communities living in these steep regions, understanding where these slides are most likely to happen is a matter of survival. It allows planners to know which hillsides are safe for building and which must be left alone, turning a chaotic natural force into a manageable risk.
A team of researchers set out to map this risk in the Mantsa River catchment, a specific area in the Dawro Zone of southwestern Ethiopia. This region is a landscape of dramatic contrasts, with elevations ranging from deep river valleys up to high mountain peaks, all carved by water and shaped by ancient volcanic activity. The ground here is made of layers of volcanic rock, some hard and some soft, sitting under a climate that brings intense seasonal rains. To create a reliable map of danger, the scientists first built a detailed record of the past. They combined satellite images with boots-on-the-ground fieldwork to locate 280 specific spots where landslides had occurred, ranging from slow-moving earth flows to sudden rock falls. They then split this list into two groups: one to teach their computer models what a landslide looks like, and another to test if the models could correctly predict where new ones might happen.
The researchers used two different statistical methods to analyze the data, treating the landscape like a puzzle where each piece represents a factor that could trigger a slide. They looked at thirteen different elements, from the steepness of the slope and the type of soil to how close a spot is to a road or a stream. One method, known as the frequency ratio, simply counted how often landslides appeared in each category of these factors. For instance, it checked if landslides happened more often on steep slopes than on flat ones, or more often near roads than far away. The second method, logistic regression, was more complex, calculating the probability of a slide by weighing all these factors together at once to see how they interacted. Both approaches aimed to produce a single score for every square meter of the land, indicating how likely it was to fail.
The results painted a clear picture of danger and safety across the catchment. The models showed that the most influential factor was not the steepness of the hill or the type of rock, but how the land was being used. Areas where the land was moderately cultivated—often meaning crops were grown on steep slopes with the natural forest removed—were far more likely to slide than forested areas or untouched woodland. The removal of trees and the disturbance of the soil by farming made the ground much weaker. The type of rock also played a major role; the area is covered in volcanic rocks that, when exposed to the hot, humid climate and heavy rain, break down into soft, clay-rich soil that holds very little strength. When this clay gets saturated with water, it becomes slippery and heavy, leading to the most common type of slide in the region: a slow, rolling earth flow.
The maps generated by the study revealed that the highest risk is concentrated in specific pockets of the landscape. The most dangerous zones are found in the southwest, northeast, and southeast parts of the catchment, where steep volcanic slopes meet dense networks of cracks in the earth and high rainfall. These areas are also where human activity, such as farming and road building, has pushed against the limits of the terrain. In contrast, the northern and central parts of the region, which are generally flatter and less disturbed, showed a much lower risk. The models were able to distinguish between safe and dangerous ground with high accuracy, correctly identifying the locations of past landslides in the vast majority of cases.
Ultimately, the study provides a practical tool for the people living in the Dawro Zone. By identifying the specific combination of steep slopes, weak volcanic rock, heavy rain, and human land use that leads to failure, the maps offer a guide for the future. They suggest that protecting the remaining forests and being cautious about farming on steep hillsides are critical steps to reducing risk. The research confirms that while the landscape is naturally prone to instability, human choices can either worsen the danger or help keep the ground stable. With this new understanding, local authorities can now make informed decisions about where to build, where to warn residents, and how to plan for a safer coexistence with the shifting earth.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.