Landslide susceptibility, mobilization rates and link with sediment yield at regional to global scale

Landslides are a key process in landscape evolution, a considerable sediment generator and a major geohazard affecting livelihoods and causing fatalities in many areas around the world. This study aims to quantify the landslide susceptibility (LSS) and mobilization rate (LMR [m³/km²/y]) and to investigate the link between landsliding and sediment yield (SY), from regional to global scale. More specifically we focus our research on four different study areas: Mount Elgon in Uganda, Romania, Africa and the entire world. Along with this increasing spatial scale (103 - 108 km²) we build up in terms of data requirements and (quantitative) depth of the analyses. This way, we gradually grow our expertise and insights to facilitate and improve the analyses at the largest spatial scales. For the Mount Elgon region in Uganda we started from the basis of landslide research with the mapping of landslides in the field. We used the compiled dataset of 653 landslides to calibrate and validate logistic models for landslides and rockfalls. We find that topography, lithology and soil moisture best explain the observed patterns of landslide occurrence across the region. Apart from the need for a rigorous LSS assessment in this landslide-prone region, we used our landslide inventory to quantify the landslide frequency and LMR. We observe an average annual LMR of 750 ton/km² and a landslide frequency of 0.04 landslides/km². Ultimately, we applied the landslide susceptibility map and frequency assessment in combination with population density data, to estimate the landslide risk at the parish level. In areas at highest risk landsliding can potentially affect hundreds of people per year per square kilometre. In Romania our focus is on the impact of landslides on SY. Few studies have investigated this effect at larger spatial scales outside the seismically most active regions. We test the value of a LSS map, to predict patterns of SY. We delineated 133 catchments for which SY was measured and derive catchment characteristics including several indicators derived from the LSS map. We observe that LSS is a significant and better predictor of SY in Romania than estimates of average sheet and rill erosion. The explanatory power of LSS for SY is mainly driven by regional variations in lithology and seismicity that might exert a strong control on landslide occurrence. Accounting for the landslide to river connectivity does not result in stronger correlations. Our study for Africa combines the analyses from the two previous studies to present a first landslide dataset and LSS map for the entire continent and to assess whether LSS is also useful to predict SY at the continental scale. We compiled over 18000 landslides by reviewing the literature and by mapping landslides in Google Earth in underrepresented regions. We then calibrated a LSS model that explains 80% of the observed variance in landslide occurrence and is determined by topography, seismicity and lithology. Based on the compiled dataset and our model we demonstrate major research gaps in various African countries prone to landsliding (e.g. Sierra Leone and Eritrea). LSS shows also a highly significant correlation with measured SY at continental scale. However, given the small portion of Africa susceptible to landsliding (10% moderately to very highly susceptible), we do not expect that landslides are the main driver of sediment production at this continent. We suspect this strong relation is at least partly explained by autocorrelation with other erosion processes. Our global study combines all previous insights and analyses by collecting time-specific and volumetric landslide data from the literature to construct a global LMR model and to confront it with SY measurements for Africa and Europe. Based on 116 globally distributed LMR observations, we calibrated a model that explains 62% of the variance in LMR controlled by topography, seismicity and lithology. Applying this model, we estimate an annual global landslide production of 100 ± 10 gigaton (56 ± 5 billion m³), with large intercontinental and interregional differences. Asia accounts for 68% of this rate due to major mountain ranges such as the Himalaya and Karakoram, while Africa only contributes 2.5% to the global LMR. We find strong correlations between LMR and SY for Europe (R²: 0.42), while they are weaker for Africa. Nevertheless, LMR are typically higher than SY, which indicates the importance of internal catchment dynamics. Through these different studies we advanced our quantitative understanding of landsliding and enhanced its comparability worldwide. We provide first models at regional to global scale and show the multifunctionality of landslide data and its derived products. Our analyses and tools can be directly used by scientists and policy makers: (1) to assess the geomorphic impact of landslides, (2) to reduce its harmful effects on people and infrastructure and (3) to support the choice of future study areas for landslide research. We underline the importance of long-term landslide mapping in all environments and especially in data-scarce regions. In this way, we can further constrain the global and regional geomorphic importance of landslides and better assess their associated risks.

Publication year:2019

Accessibility:Open