Landslide Representation Strategies in Susceptibility Studies using Weights-of-Evidence Modeling Technique |
| |
Authors: | S Poli S Sterlacchini |
| |
Institution: | (1) Department of Environmental Science (DISAT), University of Milan-Bicocca, Piazza della Scienza 1, Milan, 20126, Italy;(2) Institute for the Dynamic of Environmental Processes, National Research Council History (CNR-IDPA), Piazza della Scienza 1, Milan, 20126, Italy |
| |
Abstract: | This paper is focused primarily on how to represent landslide scarp areas, how to analyze results achieved by the application
of specific strategies of representation and how to compare the outcomes derived by different tests, within a general framework
related to landslide susceptibility assessment. These topics are analyzed taking into account the scale of data survey (1:10,000)
and the role of a landslide susceptibility map into projects targeted toward the definition of prediction, prevention, and
mitigation measures, in a wider context of civil protection planning. These aims are achieved by using ArcSDM (Arc Spatial
Data Modeler), a software extension to ArcView GIS useful for developing spatial prediction models using regional datasets.
This extension requires a representation by points of the investigated problems (landslide susceptibility, aquifer vulnerability,
detection of mineral deposits, identification of natural habitats of animals, and plants, etc.). Maps of spatial evidence
from regional geological and geomorphological datasets were used to generate maps showing susceptibility to slope failures
in two different study areas, located in the northern Apennines and in the central Alps (Italy), respectively. The final susceptibility
maps for both study areas were derived by the application of the weights-of-evidence (WofE) modeling technique. By this method
a series of subjective decisions were required, strongly dependent on an understanding of the natural processes under study,
supported by statistical analysis of the spatial associations between known landslides and evidential themes. Except for maps
of attitude, permeability, and structure, that were not available for both study areas, the other data were the same and comprised
geological, land use, slope, and internal relief maps.
The paper illustrates how different representations of scarp areas by points (in terms of different number of points) did
not greatly influence the final response map, considering the scale of this work. On the contrary, some differences were observed
in the capability of the model to describe the relations between predictor variables and landslides. In effect, a representation
of the scarp areas using one point every 50 m led to a more efficient model able to better define relationships of this type.
It avoided both problems of redundancy of information, deriving by the use of too many points, and problems related to a random
positioning of the centroid. Moreover, it permitted to minimize the uncertainty related with identification and mapping of
landslides. |
| |
Keywords: | Weights of evidence landslides ArcSDM |
本文献已被 SpringerLink 等数据库收录! |
|