Improving landslide multi-hazard early warnings in Eastern Himalayas using a Bayesian belief network

研究的主要目的是通过了解滑坡与其他自然灾害在不同时间或空间尺度上的相互作用，找出如何进行多灾害滑坡风险评估，以完善东喜马拉雅地区的预警系统。目的(O)和问题(Q)：阶段1(基于文献综述、主管指导和访谈)O 1。总结在增加/降低山体滑坡和其他自然灾害风险方面突出的物理和社会变量，这些灾害要么触发山体滑坡，要么由山体滑坡引发，或者通常与山体滑坡重合。总结不同的自然灾害在空间和时间上是如何相互关联的。探索和总结如何使用现有的贝叶斯网络来预测不同的危险。确定O1中的哪些变量已被用于多灾害滑坡风险评估的贝叶斯网络的节点，以及哪些变量可能适用于本研究。理解并总结变量是如何相互联系的，以准备贝叶斯网络的联系。学习专家启发的原则，以期将其作为一种方法应用于构建多危险滑坡风险评估的贝叶斯网络。对于选定的研究区域(最有可能是在东锡金或印度大吉岭)，总结O4.O7中确定的主要变量的可用数据。使用从O6中确定的变量收集的数据和专家启发，通过应用贝叶斯网络方法创建一个模型来量化多种危险的滑坡风险。将该模型与已有的滑坡易感性图O9进行对比。在另一个地区使用O7中开发的模型。针对东喜马拉雅制定一般指导方针，说明如何将该模型用作预警系统的一部分

The main aim of the research is to find out how to carry out a multi-hazard landslide risk assessment in order to improve the early warning system in Eastern Himalayas by understanding the interaction between landslides and other natural hazards at different temporal or spatial scales.Objectives (O) and questions (Q):PHASE 1 (based on literature review, guidance of supervisors, and interviews)O1. Summarize the physical and social variables that are prominent in increasing/decreasing the risk of landslides and other natural hazards that either trigger landslides, are triggered by landslides, or are commonly coincident with landslides.O2. Summarize how the different natural hazards are interrelated spatially and temporally.O3. Explore and summarize how existing Bayesian networks are used to forecast different hazards.O4. Determine which variables from O1 are already commonly in use for the nodes of a Bayesian Network for multi-hazard landslide risk assessment and which might be appropriate for this study.O5. Understand and summarize how the variables are interconnected to prepare the linkages of Bayesian Network.O6. Learn principles of expert elicitation with the view of applying this as one methodology for constructing a Bayesian Network for multi-hazard landslide risk assessment.PHASE 2O6. For the study area chosen (most likely in East Sikkim or Darjeeling, India) summarize which data is available for the major variables identified in O4.O7. Using data gathered from variables identified in O6, and expert elicitation, create a model to quantify the multi-hazard landslide risk by applying the Bayesian Network approach.O8. Confront the model against already present landslide susceptibility map O9. Use the model developed in O7 in another region.O10. Produce general guidelines of how this model can be used as a part of Early Warning system, specific for Eastern Himalayas