Landslide scars in the Kenyan highlands Physical and chemical topsoil changes and landslide susceptibility assessment under tropical conditions

Detta är en avhandling från Stockholm : Department of Physical Geography, Stockholm University

Sammanfattning: Farmers living in the Kenyan highlands are exposed to the risk of being affected by landslides and they are vulnerable because of lack of resources. Vulnerability can be reduced by increasing knowledge and awareness of the risk. This thesis aims to increase knowledge of landslide-related problems. The first objective is to investigate topsoil development and to assess land cover changes in landslide scars over time (paper I). Topsoil samples were taken in previously investigated landslide scars, and analysed for chemical and physical characteristics. The results show that soil has recovered well during the 20-30 years that have passed. However, there are still higher levels of total soil organic carbon in reference topsoil than in scar topsoil. The short time for plants to produce organic material, in combination with low pH, which inhibits decomposing microorganisms, may be the explanation for this. Previous research, including photographs of the landslide scars, forms the basis for comparison over time regarding land cover changes. The results show changes in land cover where forests have been replaced by cultivated land and thereafter by plantations of Eucalyptus sp. trees or cultivation. Secondly, this study investigates the applicability of GIS-based analysis for landslide assessments in this part of the World (paper II). Commonly used data were collected and analysed in ArcGis 10.0. Documentation and field measurements from historical landslides were used as reference. The results show that (1) much information was unavailable, (2) the resolution of the Digital Elevation Model (DEM; 30 m) was too coarse to capture relevant morphometric terrain parameters and (3) land use changes were difficult to map because of low quality historical air photos, and because of lack of information about the precise timing of landslide events. Hence, landslide prone areas could not be recognised. However, we could with reasonable certainty identify sensitive conditions: landslides are most likely to occur in slopes with a gradient of 25° or more, within 80 m from roads, and during the rainy seasons.