Landslide susceptibility and mobilization rates in the Mount Elgon region, Uganda

Abstract

Mount Elgon in Eastern Uganda is one of the most landslide-prone regions in Africa. This extinct shield volcano is characterized by steep slopes, intense precipitation, and fertile lands supporting a dense human population. As a result, landslides frequently cause damage and many fatalities. Apart from the need for a landslide susceptibility assessment, insight into the landslide mobilization rates [ton/km²/y] is required to assess the geomorphological importance of landslides. Such quantitative information is scarce for many regions around the world and particularly for Africa. We therefore compiled a calibration dataset of 653 landslides (514 slides and 139 rockfalls). Additionally, a second dataset of over 400 landslides was independently collected for validation purposes. To determine a logistic landslide susceptibility model, we used Monte Carlo simulations that selected different subsets of the calibration dataset to test the significance of the considered environmental variables. Susceptibility maps for all landslide types and for rockfalls were constructed for the Mount Elgon region in Uganda. In both maps, topography is by far the most significant factor controlling landslide susceptibility. Including lithology and soil moisture, further improved the model predictions. The models explain about 55% and 85% of the observed variance in landslide occurrence for all landslide types and for rockfalls respectively. The average calculated landslide frequency and mobilization rate for the landslide affected area are respectively 0.04 landslides/km²/y and 750 ton/km²/y. Landslide size is only weakly positively correlated with landslide susceptibility. Therefore, observed larger landslide mobilization rates correlating with higher landslide susceptibilities result from larger landslide numbers rather than from larger landslides. Our research highlights the relevance of detailed and long-term landslide mapping at the regional scale in data-scarce areas for regional planning and risk reduction strategies, as an improvement to continental and global susceptibility models, but also to assess the geomorphological importance of landslides as an erosion process.