Construction And Reasoning Method Of The Rainstorm Flood Disaster Chain Ontology

China is one of those countries that devastated by frequent disaster in the world.There are a wide variety of natural disasters in China,disasters often represent as chain structure because of the complex formation mechanism and discipline between the various disasters,so they can cause much more economic loss and life threat than single disaster and it has become a challenge to disaster management and decision support.The occurrence of natural disasters in China shows a trend of incline year by year.so there is an urgent need for an effective disaster description and reasoning mechanism to guide disaster response and decision support.The occurrence of natural disaster is sudden,random and chained.It has complex relation between disaster event and wide conceptual framework.Due to lack of a unified description of hazards,disaster knowledge is un-shareable and non-reusable.So constructing a uniform describing mechanism model is an urgent issue confronted with us.This paper realized structured expression of disaster concept,supported disaster knowledge sharable and reusable.Firstly,the paper proposed a disaster chain domain ontology construction method in order to achieve a unified description and expression of the disaster knowledge.Comparing existing ontology construction methods and referencing the national and industry-related standards,this paper defined the concept,attributes and relation between the disasters,then the uncertainty of disaster chain has been analyzed,and disaster chain has been divided into fine-grained event factors,the events factors are classified as rainfall factor,topography factor,hydrology factor,land use factor and temporal factor,the disaster chain ontology is described by OWL and visualized ontology by protege which is the most popular software in constructing ontology.On this basis,the paper uses the Bayesian network as the disaster chain risk reasoning model,the uncertainty and probability extension has been made to the disaster chain ontology at the description level.To distinguish from the general Bayesian network structure learning,the paper uses the uncertainty-extended disaster chain ontology map into the Bayesian network,which ensure the correct semantic relation between the nodes in reasoning model.The paper collects disaster cases as training samples from National Administration of Surveying,Mapping and Geo-information,calculated the prior probabilities of the Bayesian network model constructed to realize the prediction of disaster chain.This paper selects Hubei Province,one of the regions with high disaster clustering characteristics as the research area by temporal and spatial autocorrelation analysis.Two typical storms floods in May 2016 and June 2016 are selected as test samples,the accuracy of the result was evaluated by the Precious,Recall,Fl-Measure,Kappa and ROC curve,the results show that this method satisfies the precision requirement and has a practical significance.