Effect of land use, soil characteristics and rainfall conditions on landslide occurrence in the tropical mountains of NW Rwanda

Landslides are a widespread geo-hydrological hazard that cause loss of life, damage to infrastructure, and long-term disruption of livelihoods. Reducing their impacts requires a better understanding of the processes that govern their occurrence in both time and space. Considerable progress has been made in landslide hazard assessment through systematic mapping, susceptibility analysis, and early warning systems, but such advances remain concentrated in the Global North. In contrast, the Global South continues to face severe data scarcity and methodological gaps, which limit the effectiveness of hazard assessment tools. Motivated by the frequent and severe impacts of landslides in Rwanda, this thesis investigates the role of anthropogenic factors in landslide occurrence, with a focus on the northwestern provinces, the country’s hotspot of recurrent events. The research consists of two complementary studies. The first examines how agricultural terraces influence landslide occurrence during three major rainfall-triggered events in 2016, 2018, and 2020. Terraces are widely promoted in tropical mountains as a soil and water conservation strategy on intensively cultivated hillslopes. However, their impact on slope stability remains poorly understood. Using an inventory of more than 4,600 landslides triggered by these events, I analyzed the relation between landslides, terraces, and other controlling factors. Results indicate that landslides were, on average, about three times more likely on terraced hillslopes than on non-terraced ones. The effect varied across events: it was strongest for the 2020 event, weaker in 2018, and negligible or negative in 2016. Terraces were most influential on moderately to highly susceptible slopes, but had little additional effect where susceptibility was already very high. Landslides on terraces also tended to be smaller in size. These findings demonstrate the complex and context-dependent role of terraces in modulating slope stability, with direct implications for susceptibility mapping and land management. The second study addresses the scarcity of soil moisture (SM) data in tropical Africa by developing and analyzing one of the first large-scale field-based SM datasets in the region. Soil moisture is a critical variable for rainfall-triggered landslides, yet remains rarely measured in the Global South. Between November 2021 and November 2024, I collected 195,026 daily volumetric SM measurements along six representative hillslope transects that covered two soil textures (clayey and sandy loam) and three land management types (forested, cultivated without terraces, cultivated with terraces). The dataset reveals strong spatial and temporal variability in SM linked to soil depth, texture, rainfall conditions, and land use. Clayey soils were generally wetter at 30–60 cm depth, whereas sandy loam soils under cultivation and terraces often reached the highest values at 100 cm depth. Land use effects were soil-dependent: forests on sandy loam were much drier at depth than cultivated slopes, while in clayey soils forests were slightly wetter than cultivated areas. Correlations between SM and antecedent rainfall varied widely with season, soil, and land use. Notably, during the long wet season—the peak landslide period—the predictive power of simple antecedent rainfall indices was low, especially on cultivated and terraced slopes. Together, these studies highlight the strong influence of land management on both the spatial distribution of landslides and the dynamics of soil moisture in tropical mountainous environments. The results challenge the assumption that rainfall alone can serve as a reliable proxy for slope instability and call into question the uniform use of fixed antecedent rainfall thresholds in early warning systems. By demonstrating the context-dependent interactions among rainfall, soil, and land use, this thesis underscores the need for soil- and land use–specific calibration of landslide hazard assessments in the Global South. Ultimately, the work advances understanding of landslide processes in Rwanda and similar tropical settings, with important implications for risk reduction strategies and sustainable land management.