Landslide hazard zonation in Namasigue and El Triunfo, Southern Honduras

Abstract

Hurricane Mitch was considered one of the strongest hurricanes of the century. While passing through Honduras and Nicaragua, it weakened and generated intense rain that triggered widespread landslides and flooding. The objective of this thesis was to test two models to determine the feasibility of creating landslide hazard maps. Data were used to determine how landslide occurrence was affected by the variables in the model. Four easily observable variables were used for both models: slope, aspect, stream proximity and land cover type. A heuristic adaptation approach and a logistic regression approach were developed based on the data from Namasigue watershed. Once landslide hazard maps were developed and tested for accuracy using the Namasigue data, they were applied to El Triunfo. The degree of slope influences the occurrence of landslides. Land cover seems to have an influence on landslide activity. Forest and shrub fallow appear to decrease the probability of land being affected by landslides. The landslide hazard map suggests that forests and shrub fallows might also offer a protective cover and, thus, reduce landslide activity. The percentage of land affected by landslides increases at a second degree polynomial rate as slope increases in bare soil, crops and grass fallow cover types. Therefore, vegetation removal seems to increase landslide activity. The heuristic adaptation approach seems to be a consistent method to assess landslide activity since the validation process showed similar results in terms of hazard category distribution. The logistic regression approach results did not show consistency in terms of the distribution of landslides by hazard category. Further research and adjustments need to be performed to assess transportability of the logistic regression approach.