On spatial uncertainty in hazard and risk assessment

摘要

Exploiting the mathematical framework of favourability function modelling and of software designed for spatial target mapping, experiments are discussed to measure the effects of uncertainty of spatial support. A database that has been the focus of landslide hazard prediction and risk assessment of buildings, roads and land uses is reanalyzed gradually modifying the spatial characteristics of supporting patterns of the proposition of "presence of landslide occurrences." Boundary fuzziness of categorical mapping units and filtering of continuous fields, represent weakening of spatial relationships. Comparisons are made of ranges of ranks representing uncertainties of class membership in target patterns obtained by iterative cross-validations. The impacts of the different spatial uncertainties on risk assessment patterns are visualized using progressive combinations of uncertainties of target patterns to resolve risk equations and study the consequent changes in risk values.