Geographic Information Systems (GIS) are more and more used for the management and the prevention of risks, either natural or man-made. Most of these studies concern flood risk mapping [1] and earthquake susceptibility assessment [2]. During the last ten years, most progress has been also accomplished for landslide susceptibility assessment, at both medium (1:25.000) and large (1:10.000) scales. At these scales, three types of methods exist to locate and characterize the landslide-prone areas (ie., that gather the ground conditions favourable to slope instabilities) [3], [4], [5]: o the qualitative approach, based on expert knowledge [6]; o the deterministic approach, taking into account geotechnical data to calculate a safety factor of the hillslopes [7], used for very large scale (1/5.000, and less); o the statistical approach (bivariate, multivariate), used for coarser scales [3], [8]. The underlying concept of these GIS-based (raster) methods is the "terrain unit", or "homogeneous unit" (ie., the pixel) which is defined as the portion of land surface which contains a set of ground conditions which differ from the adjacent units across definable boundaries [9]. With the statistical approach, the objective is to forecast the occurrence in space of a landslide on the basis of the values of a set of variables or factors. The aim is to predict the behaviour of a dependent variable (landslide susceptibility) on the basis of a set of known independent variables (ground conditions). The final result is an estimate of a spatial correlation between the landslide distribution and each ground conditions. Several statistical methods have been developed (multivariate analysis, conditional analysis, weight of evidence analysis). A comprehensive review can be found in [4]. By definition, as the simulations are dependent on the quality of the input data (map scale, typology, precision), it is necessary to evaluate the propagation errors and to define a set of information sources adapted with the quality expected for the final results. The terrain unit comprises five categories of input dataset [4]: the geomorphology (landslides), the topography (elevation, slope angle, slope aspect,) the geology (lithology, structure, superficial deposits), the land-use (human activities, vegetation, infrastructures), and the hydrology (hydrographic network, swamps). Moreover, the terrain-unit must be mappable at effective cost over the entire region through criteria which are as objective as possible. Recent studies have shown that the use of a similar data acquisition technique (analysis of aerial photography) by different experts can lead to distinct results. For, instance, Van Carrara et al. [3] and Westen et al. [10] observed respectively 75% and 78% of difference in the location of landslides, carried out by distinct persons on different areas. The expert knowledge of the scientist and the type of data acquisition influence the final results.
Towards the construction of a spatial database to manage landslides with GIS in mountainous environment
Sterlacchini S;
2003
Abstract
Geographic Information Systems (GIS) are more and more used for the management and the prevention of risks, either natural or man-made. Most of these studies concern flood risk mapping [1] and earthquake susceptibility assessment [2]. During the last ten years, most progress has been also accomplished for landslide susceptibility assessment, at both medium (1:25.000) and large (1:10.000) scales. At these scales, three types of methods exist to locate and characterize the landslide-prone areas (ie., that gather the ground conditions favourable to slope instabilities) [3], [4], [5]: o the qualitative approach, based on expert knowledge [6]; o the deterministic approach, taking into account geotechnical data to calculate a safety factor of the hillslopes [7], used for very large scale (1/5.000, and less); o the statistical approach (bivariate, multivariate), used for coarser scales [3], [8]. The underlying concept of these GIS-based (raster) methods is the "terrain unit", or "homogeneous unit" (ie., the pixel) which is defined as the portion of land surface which contains a set of ground conditions which differ from the adjacent units across definable boundaries [9]. With the statistical approach, the objective is to forecast the occurrence in space of a landslide on the basis of the values of a set of variables or factors. The aim is to predict the behaviour of a dependent variable (landslide susceptibility) on the basis of a set of known independent variables (ground conditions). The final result is an estimate of a spatial correlation between the landslide distribution and each ground conditions. Several statistical methods have been developed (multivariate analysis, conditional analysis, weight of evidence analysis). A comprehensive review can be found in [4]. By definition, as the simulations are dependent on the quality of the input data (map scale, typology, precision), it is necessary to evaluate the propagation errors and to define a set of information sources adapted with the quality expected for the final results. The terrain unit comprises five categories of input dataset [4]: the geomorphology (landslides), the topography (elevation, slope angle, slope aspect,) the geology (lithology, structure, superficial deposits), the land-use (human activities, vegetation, infrastructures), and the hydrology (hydrographic network, swamps). Moreover, the terrain-unit must be mappable at effective cost over the entire region through criteria which are as objective as possible. Recent studies have shown that the use of a similar data acquisition technique (analysis of aerial photography) by different experts can lead to distinct results. For, instance, Van Carrara et al. [3] and Westen et al. [10] observed respectively 75% and 78% of difference in the location of landslides, carried out by distinct persons on different areas. The expert knowledge of the scientist and the type of data acquisition influence the final results.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
