| Identifying Regions at Risk for Landslides Using Combined GIS and Genetic Algorithm Procedures
ABSTRACT Evaluating the susceptibility of land to various natural hazards (such as landslides, wildfire, flooding, and so on) is an obvious factor in evaluating land use suitability. In a GIS environment, such hazard susceptibility evaluations are typically accomplished by using standard statistical techniques to find relationships between expertly-appraised hazard rating levels and a variety of cartographic variables (such as elevation, soil types and conditions, proximities to other important physiographic features, etc.). Using standard GIS techniques, these cartographic variables can then be derived for regions that have not been hazard-rated by experts, and estimated hazard ratings for these regions can be obtained using these cartographic variables and the statistical relationship derived earlier. There are a number of limitations to procedures such as this, not the least of which is the inherent limitations of most statistical procedures when applied to spatial data. This study will evaluate the alternative of using genetic algorithms instead of statistical techniques to derive relationships between cartographic variables and hazard ratings. Using a database describing landslide hazard ratings as an example, both statistical and genetic algorithm techniques will be used to identify relationships between hazard ratings and cartographic variables. The absolute and relative accuracies of both techniques’ predictions will be compared, and the strengths and weaknesses of both techniques will be identified. |