Realistic assessment of slope reliability for effective landslide hazard mangement

Abstract

Probabilistic methods have been employed as a practical aid for slope management in respect of the stochastic nature of landsliding. The effect of deterioration on the reliability of slopes of various periods of service, however, has seldom been considered explicitly in the current reliability evaluation. The first part of this paper will present an empirical procedure, using past performance of an extensive set of cut slopes in Hong Kong as an example, to model probabilistically the deteriorating effect on the failure probability with age for a population of slopes. The procedure extends further to predict the failure probability of slopes over different future service periods. Although this probability would serve as an estimate of average slope reliability to assist landslide hazard management by planners, a more accurate site-specific failure probability would be preferred for individual projects. The second part of this paper will adopt a Bayesian procedure to map the calculated probability of a given slope to a more realistic estimate of failure probability. The procedure enables performance records, the effect of system deterioration and other unaccountedfor factors to be systematically incorporated for the estimation. Finally, a procedure is developed to assess the uncertainty associated with the updated probability, resulting from limited observed performance data available for the probability mapping process.