Identifying Landslide Hazards in a Tropical Mountain Environment, Using Geomorphologic and Probabilistic Approaches

Title of Document: IDENTIFYING LANDSLIDE HAZARDS IN A TROPICAL MOUNTAIN ENVIRONMENT, USING GEOMORPHOLOGIC AND PROBABILISTIC APPROACHES José Gregorio Roa-Lobo PhD, 2007 Directed By: Prof. Michael Kearney, Department of Geography The objective of this study is the performance, assessing, comparison and validation of a set of three landslide hazard maps: The geomorphological, the multicriteria evaluation (MCE) and the probabilistic (weights of evidence); in order to evaluate its accuracy, advantages and limitations, and finally state its reliability. These approaches were tested in a tropical mountain environment located in the central Venezuelan Andes, particularly in the Río Chama basin (2820.63 Km2), where the complexity and variety of the landscape provides a special geographical framework to address the landsliding process as natural hazard. The scale of this study is regional. For doing this a GIS data base was built up to collect and manipulate the landslide inventory map and the cartography of the main landslide passive factors found in this study area. The landslide inventory map was generated through the manual interpretation of 300 aerial photographs and by the processing of two sets of Landsat imagery via contrast-widening color composite, given as result the outline of 493 landslide polygons. The landslide passive factors represent the physical features of the study area associated to landslide occurrences, as those found in the topographical, geological and physiographical settings. In that sense, given the main role played for a digital elevation model (DEM) as data input, a DEM for the study area was built through remotely sensed data obtained from the shuttle radar topographical mission (SRTM) and optical stereographic imagery provided by the advanced spaceborne thermal emission and reflection radiometer (ASTER) system. Because of the comparative nature of this study, these data was preliminary processed via density analysis in order to establish a common background on the landsliding process – passive factors relationship, which was used later to set up the criteria applied in the geomorphological and multicriteria evaluation (MCE) approaches. All the three landslide hazard mapping approaches were fully benefited from the use of GIS, improving the processing and manipulation of the spatial information, even in procedures considered subjective as the geomorphological mapping, as well as in the use of non-areal statistics measures for the weights of evidence procedure where the Kappa index proved to be a useful index to assess the level of independence between factor maps. As a way of validation, the accuracy and error rate of the three landslide hazard maps were performed by its comparison to the landslide inventory map. Hence through the use of contingency tables and the success rate curve, was concluded that although the geomorphological approach achieved a better landslide predictive power for this study area at a regional scale, the remaining procedures can play a complementary role, for example the MCE plays a crucial role in an early assessment of landslide hazard which highlights the needs and improving necessary to achieve a better probabilistic approach, which can be later incorporated in a more objective geomorphological assessment. Results also showed that any methodology can be improved and even empowered by the development of better and more integrated standards for geographical data collection rather that the simplification of them, in that sense, satellite data improves the spatial and temporal consistence of data used for landslide hazard purposes, potentially allowing the integration of useful geographical data as those Holdridge life’s zones and geomorphometric generated and used in this study. Hence, further studies at regional scale must explore the remotely sensed imagery capacities for generation of data bases addressing regional susceptibility to landsliding process. IDENTIFYING LANDSLIDE HAZARDS IN A TROPICAL MOUNTAIN ENVIRONMENT, USING GEOMORPHOLOGIC AND PROBABILISTIC APPROACHES