PROBABILISTIC LANDSLIDE RISK MANAGEMENT USING BAYESIAN BELIEF NETWORKS

Abstract

Development of a landslide hazard assessment based on a Bayesian belief network. Methodology. Landslide risk assessment with basic systematic use (after heavy rains, snowfalls, etc.) of expert fragments of basic knowledge implemented in graphical probabilistic models – Bayesian Belief Networks (BBN), resolved objective and selective probabilities and knowledge. We considered possible relationships (dependencies and independences) for uncertainty and risk management. Results. Risk management during the closure of drainage systems involves the use of various deterministic and probabilistic methods. One of the risk management procedures is its assessment. Hazardous natural phenomena (landslides, mudflows, water and wind erosion, flooding, waterlogging and land erosion) can threaten the lives and health of the population and lead to signifi cant damage. Resource limitations (time, financial, etc.) often do not allow for full-fledged monitoring and additional research using GIS and other means. To minimize risk and make timely management decisions to protect people and territories, it is necessary to use methods and models based on expert knowledge (subjective probabilistic assessments) about cause-and-effect relationships and distribution laws of integral characteristics of the manifestation of characteristic conditions (for example, in the case of landslide hazard – the slope stability coefficient FS). As a result of using the proposed methodology in the first example, restrictions were obtained that allow using different methods to maintain the slope. Conclusions. The proposed ideology of probabilistic landslide risk assessment based on the system of using expert knowledge in Bayesian belief networks is well applied in the context under consideration. The factors affected by the landslide zone are universal and can be used to assess risks in other areas where slope processes are developing.