Landslide susceptibility evaluation and validation at a regional scale

Tese de doutoramento, Geografia (Geografia Física), Universidade de Lisboa, Instituto de Geografia e Ordenamento do Território, 2014

This dissertation aims to deepen the knowledge about the causes that influence the spatial and temporal occurrence of slope instability at a regional scale. The study area, located 90km north from Lisbon, comprises three sub‐catchments, named Arnoia, Tornada and Alfeizerão (275.9 km2). These sub‐catchments were chosen for their geological and geomorphological features and because it is an area prone to slope instability. Many methods have been proposed worldwide to evaluate landslide hazard. In this dissertation there are presented and performed two approaches that, according to Guzzetti (2002), are the most promising: physically‐based methods; and statisticallybased methods. Methodologies were applied for acquisition of their input data. In addition, within the physically‐based methods was also implemented a temporal dynamic approach, which simulated the hydrology over time and evaluated its effects on slope stability. Thereby, in order to obtain the overall goal the following 10 specific objectives were stated: 1) Acquisition of multi‐temporal landslide inventories; 2) Acquisition and production of new themes based on modeling and field observation (e.g. detailed lithological map, morpho‐structural map, DEM, soil depth); 3) Acquisition of soil characteristics according to the hydrogeological and geotechnical properties of soils (through field work, laboratory measurements and back analysis; 4) Landslide susceptibility assessment using the hydrological model coupled to slope stability model under static temporal conditions and its validation through the quantification of the degree of prediction rate; 5) Acquisition, processing and modeling of long term climatic data (e.g., rainfall and temperature); 6) Landslide susceptibility assessment using hydrological model coupled to slope stability model under dynamic temporal conditions and its validation through the quantification of the degree of prediction rate; 7) Comparison between physically base models: static and dynamic approach; 8) Sensitivity analysis and hierarchy of the landslide predisposing factors; 9) Landslide susceptibility assessment using statistically‐based method (Information Value Method) and its validation through the quantification of prediction and success rate; 10) Comparison between statistically and physically static models. Some input data, of extreme importance for every method used in this dissertation, proved to be very difficult to obtain. It is worth mention the case of the geological map, which was only available at a 1:50,000 scale. Thus, being aware that landslides are greatly conditioned by the lithological properties of the terrain, a detailed lithological map at a 10,000 scale was performed through the stereoscopic interpretation of aerial photographs and field work validation.The quality of the landslide inventory is of crucial importance for any prediction model. Thus, a multi‐temporal landslide inventory was achieved through aerial photo‐interpretation, orthophotomaps interpretation and field work. Since the models obtained through the Infinite Slope method aim to predict the areas susceptible to shallow translational slides, a validation was made based on the shallow translational slides validated through field work. Further, all the landslides types from each landslide inventory were modeled through a bivariated statistical method (Information Value Method). A comparison, between the shallow translational slides susceptible model obtained from different approaches was also performed. The main conclusions of the work are the following: 1) The detailed lithological map has a better discriminating power than the original lithological map; 2) Trough a spatial and temporal dynamic physically‐based method it is possible to inferred the possible conditions that triggered shallow translational slides; 3) the static physically‐based method, presented better skills for predicting the spatial occurrence of shallow translational slides over the study area than the statistically‐based method.

Fundação para a Ciência e a Tecnologia (FCT, SFRH/BD/46816/2008)