Warning and evacuation are among the most effective ways for saving human lives and properties from landslide dam hazards. A new warning decision model for landslide dam break is developed by using Influence Diagrams to minimize the total losses. An Influence Diagram is a simple visual representation of a decision problem. It analyzes the qualitative (causal) relationships between the variables via logic diagram and determines the quantitative relationships via conditional probability and Bayes’ theorem. The model is applied for the warning decision making of the 2008 Tangjiashan landslide dam. The new model unifies the dam failure probability, evacuation, life loss and flood damage in a Influence Diagram. Besides, a warning criterion is proposed for efficient decision making. The model is more advanced than decision tree since the inter-relationships of influence factors are qualitatively analyzed with causality connections and quantitatively analyzied with conditional probabilities. It is more efficient than a dynamic decision making model (DYDEM) as it can directly calculate the three types of flood loss (i.e., evacuation cost, flood damage and monetized life loss) as well as the expected total loss. Moreover, the probabilities of the influence factors leading to known results can be obtained through inversion analysis based on Bayesian theory. The new warning decision model offers an efficient way to save lives from landslide dam breaking and to avoid unnecessary expenses from premature warning and evacuation.

中文翻译：

---

基于影响图的滑坡溃坝预警决策

预警和疏散是从滑坡大坝灾害中拯救人类生命和财产的最有效方法之一。通过使用影响图开发了一种新的滑坡溃坝预警决策模型，以最大限度地减少总损失。影响图是决策问题的简单直观表示。它通过逻辑图分析变量之间的定性（因果）关系，并通过条件概率和贝叶斯定理确定定量关系。该模型应用于2008年唐家山滑坡大坝的预警决策。新模型在影响图中统一了大坝破坏概率、疏散、生命损失和洪水灾害。此外，还提出了一个警告标准，以实现高效决策。该模型比决策树更先进，因为影响因素的相互关系通过因果关系进行定性分析，并通过条件概率进行定量分析。它比动态决策模型（DYDEM）更有效，因为它可以直接计算三种类型的洪水损失（即疏散成本、洪水损失和货币化生命损失）以及预期总损失。此外，通过基于贝叶斯理论的反演分析，可以得到导致已知结果的影响因素的概率。新的预警决策模型提供了一种有效的方法，可以在滑坡大坝破裂时挽救生命，并避免因过早预警和疏散而产生的不必要的费用。