基于复杂网络的历史自然灾害链式规律与演化机理研究

The research on the chain regularity and evolution mechanism of natural disaster is mainly concentrated on the direction from single-source disaster to its secondary disaster within short periods, and the recent case studies have shown that the disaster evolution is generally accompanied by the complexity of multi-disaster association and long periods of space migration. This project plans to use complex network to achieve the modeling of historical natural disaster events, and in-depth research the chain regularity and system dynamics evolution mechanism of long time series of disasters. First of all, on the basis of the analysis of spatial-temporal characteristics and relevance of historical natural disaster, then to determine the network key parameters and abstract standards and establish a multidimensional complex network model; Then, the map area of network model is divided into multiple regions, and from "co-occurrence analysis", "time correlation" and "space correlation" to study the disaster chaining probability and temporal-spatial trends of each region, and then by regional correlation analysis to determine the overall composite mechanism of the disaster; On the basis of the above work, then to execute the dynamic simulation and explore the key elements and common regularity of disaster evolution, so as to achieve the purpose of predicting disasters effectively. This project contributes to a better understanding of the formation law of disaster chain and evolution mechanism of disaster system, and could provide theoretical basis and decision support for the effective implementation of government disaster prevention and mitigation work.

自然灾害的链式规律与演化机理研究主要集中在短时间的单源灾害与次生灾害方向上，近年的案例研究证明灾害演化具有多灾种关联的复杂性和长时间的空间迁移性。本项目拟利用复杂网络对历史自然灾害事件进行建模，深入研究长时间序列的灾害链式规律和系统动力学演化机理。首先，在历史自然灾害的时空特征及关联性分析的基础上，确定网络关键参数和抽象标准，建立一个多维的复杂网络模型；然后，在网络中进行地图区域划分，从“共现分析”，“时间关联”，“空间关联”三个方面来研究各区域的灾害成链概率与时空趋势，进而通过区域相关性分析来确定灾害的整体复合作用机制；在此基础上，进行动力学仿真，挖掘灾害演化的关键要素和共性规律，达到有效预测灾害的目的。本项目的研究有助于更好地理解灾害链的生成规律和灾害系统的演化机理，为政府防减灾工作的有效开展提供理论依据和决策支持。

自然灾害的关联性及演化过程所体现出的灾害间的同源性、链发性以及韵律性，对灾害预测具有极高价值。本项目利用复杂网络对历史自然灾害事件进行建模，深入研究长时间序列的灾害链式规律和系统动力学演化机理。对历史灾害进行3维时空网络建模，通过寻找网络中的相似空间向量来挖掘灾害节点间的链式关联关系，同时通过将向量发现问题转化为一种路径优化问题来利用蚁群算法进行求解，仿真结果表明该方法具有较高的精确性和实用性。提出了基于共现分析的自然灾害的关联研究方法。在Jaccard指数与归一化方法的基础上提出一种新的灾害共现率的计算方法，即CO指数，并以此计算灾害间相互引发的概率。基于灾害时序信息提出一种灾害区域关联复杂网络构建方法。通过相关转化使得可以利用压缩感知理论对节点关联系数进行求解，从而实现对网络节点关系拓扑的构建。实现对网络区域灾害链特征以及灾害链间复合演化关系的概率模型处理，并提出一种结合TSP模型和蚁群算法的灾害复杂网络社团划分方法。