Research On Data Dependency Of Natural Disaster Events

In recent years,global change has exacerbated the intensity and frequency of natural disasters.With the continuous improvement of data acquisition capabilities,the role of data in disaster reduction research has become more prominent.Scientific research on disaster events has gradually shifted from single disciplines to multiple disciplines,and multidisciplinary data resources have become effective support.The data dependence of natural disaster events refers to the contribution of multidisciplinary data to specific disaster types or disaster events.Study on data dependence of natural disasters is helpful to build scientific plans for specific disaster events research.Focus on the the difficulties in providing adequate and scientific supporting data for the research of natural disaster mitigation,this paper explored the correlation between natural disaster events and their supporting data using a literature-based knowledge discovery model,accurate and quantitative analysis was also conducted on multidisciplinary data for natural disasters.Three earthquake events(Tangshan,Wenchuan and Haiti)and flood events were selected as typical study cases.The data dependence of natural disasters were quantitatively studied from multi-disciplinary view and earth observation perspective.The specific requirement of scientific researchers for data were modelled by personalized data recommendation approach.The concept ontology of natural disaster events was also constructed.The main content and innovations include:1)The theory of mechanisms of natural disasters,classification of natural disaster types and data management for disaster research were investigated,we concluded the background knowledge,which including life cycles of natural disasters,data management flow for disaster mitigation and the data dependency of natural disasters.2)The literature-based knowledge discovery approach was adopted in our study,the research ideas and technical routes were proposed for typical disaster events.Taking the example of three earthquake events(Tangshan earthquake,Wenchuan earthquake and Haiti earthquake)and flood events as study cases,the characteristics of multidisciplinary data for typical disaster events were studied.Also earth observation data for flood study topics was analyzed in detail.The results showed that: the most frequently used data in flood research were meteorological data and hydrological data,followed by earth observation data.Most of the articles in Tangshan earthquake study used Geophysical data and geological data,while Wenchuan earthquake study mainly adopted ground measurement data and earth observation data.However,studies on Haiti earthquake made extensive use of socio-economic data,clinical data and observational data.It was also found that the obvious spatio-temporal variation in data for natural disaster research.3)The influence factors of multidisciplinary data for natural disaster research were analyzed based on the statistical results.The key factors including timeliness index,user index,geographical index as well as comprehensive index were proposed considering the spatio-temporal features and user relevant features.At last,taking the flood literature published between 1990-2015 and literature sources of three earthquake study as data sources,the data dependency of natural disasters were quantitatively analyzed from sptio-temporal view and user perspective.4)Personalized data recommendation model for scientific researchers was built according to the collaborative-filtering recommendation algorithm.The research content and numerical fields on authority,popularity,timeliness,geography and subject were selected for vector construction of academic literature.The scientific user vector was built based on different aspects of a user's personalized preference.Personalized data recommendation model was proposed by incorporating the multi-criteria decision making approach and recommendation methods.The flood research articles were selected as experiment data source.5)According to the natural disaster classification criteria and concept extraction results for natural disaster events from literature,the concept ontology for natural disaster events oriented to scientific research was constructed.Hierarchical ontology building method and class meta-information description was adopted to solve the semantic heterogeneity and retrieval problems.According to the investigation,the current literature-based knowledge discovery methods are widely used in biomedical and chemical fields,but there has no research to extend this method in natural disaster domain recently.In addition,most of studies on the data dependence of natural disasters were from a single-disciplinary perspective based on the academic background of the researcher.To the best of our knowledge,no such study have been found that using literature-based approaches to analyze the multidisciplinary data dependence of natural disasters.Therefore,the research objectives and methods used in this article are groundbreaking.The technical methods and models proposed in this study can be extended to other natural disaster events,which is of great significance for enriching the knowledge category of natural disaster research and comprehensive disaster mitigation research.