Landslide susceptibility assessment and mapping using GIS and Analytic Hierarchy Process (AHP). Case study: Atalanti watershed, Central Greece

Abstract

A methodology for landslide susceptibility assessment and mapping in a GIS environment through multicriteria decision analysis (MCDA) is presented. Landslides belong to the severe natural disasters which are often
experienced in Greece and have become a significant concern mainly in the mountainous areas. In this paper, the
Analytic Hierarchy Process (AHP) was applied with a varied weighted linear approach to identify the landslide
potential associated with the terrain aiming at contributing in landslide risk assessment evaluation. The study was
focused on landslide susceptibility mapping in Atalanti catchment located in Central Greece, employing spatial analysis
of factors influencing the landslide occurrence. GIS is a useful tool for the construction of landslide prediction model
and for application in regional planning, risk management and hazard mitigation as well as early warning for the
prioritization of efforts to reduce future landslide hazards. In total, twelve dataset layers including slope angle, slope
aspect, rainfall, geology-lithology, curvature, elevation, land use, proximity to the rivers, faults and roads,
topogragraphic wetness index (TWI) and stream power index (SPI) were selected as the causative factors for the
analysis. Digital elevation model (DEM) of 25x25m resolution was used to extract the topographic, geological,
geomorphological, hydrological, land use and climatic related landslide causative-instability factors. The final
susceptibility score was classified into susceptible rating values based on the factors’ importance and the spatially
generated layers were assembled to produce the final landslide indexed susceptibility assessment map. According to the
map, 30.4% area of the region is moderately susceptible to the occurrence of landslides, 30.0% area is low to
moderate, 19.8% is low, 14.3% is moderate to high and only 3.1% is high to very high susceptible to the landslides’
occurrences. Validation results and sensitivity analysis based on landslide inventory showed that this model could be
used for the prediction of future landslides since almost 81.8% correspond to areas where landslide phenomena were
actually took place. Finally, the analysis of the susceptibility modeling showed the high importance of slope, rainfall,
geology and tectonics parameters.