Search

The hazard potential of Sri Lanka's landslide-prone areas has been evaluated based on a well-defined decision process involving primary and secondary attributes. The relative contribution (weights) of both types of attributes and the severity ratings of secondary attributes have been defined subjectively on the same linguistic severity scale ('very high' to 'very low') based on expert judgement and past statistics, respectively. However, when field evaluations are performed to determine the ultimate hazard potential indices on a scale of 0-100, the subjective weights and ratings are replaced with conveniently defined single numeric values. It had been long felt that significant revision of this somewhat vaguely defined evaluation scheme was essential since the zoning maps developed based on it have not agreed well with field maps produced from actual landslide events. Hence, a research project was initiated to (1) quantify the ratings and weights used in the hazard potential evaluation more objectively and (2) use a rational scheme to uncover the confounding relative contributions of the primary attributes. To achieve those objectives, the weights and ratings are more appropriately defined as fuzzy numbers, and a technique resembling the Monte-Carlo simulation is developed to assemble the fuzzy ratings with their fuzzy weights. Then, the fuzzy hazard potential computed for a given site is transformed back to (1) a more conceivable linguistic descriptor on the above linguistic severity scale and (2) a landslide potential index (LPI). The reliability of the evaluation process is estimated by comparing the LPI at each site with the corresponding actual landslide intensities computed from past data from one province of Sri Lanka. Finally, the weights attached to the primary attributes are refined systematically through an optimisation process until the (LPI) predictions match the field observations reasonably well. It is shown how fuzzy sets improve the reliability of the evaluation procedure by capturing the subjectivity in the ratings and weights and providing an analytical base for determining the weights that relate better to field data. [ABSTRACT FROM AUTHOR]