| China is a mountainous country,and flash flood is the natural disasters that causes the most casualties.In 2013-2016,the MWR and the MF carried out the Chinese Flash flood Survey and Evaluation(CFFSE)project to determine the disasters' distribution,magnitude,main causes and warning indices with their threshold values,including flash flood events occurred from 1950 s.The CFFSE dataset is huge.It is estimated that the total data storage exceeds 100 TB and that there are over 100 million individual records.Because of the numerous survey differences,data errors are frequent.Therefore,an effective data quality validation approach is significant for flash flood prevention.Meanwhile,Analysis and data mining need to be carried to study the mechanism of flash flood based on the CFFSE dataset,due to the aim of preventing flash flood.The following topics are studied in this article:Firstly,a DM-Moran-based validation approach is established to perform quality control on the CFFSE dataset.The approach is based on a data model of the flash flood and a spatial data mining algorithm and involves examining logical relationships and detecting anomalous survey units,which effectively integrates the advantages of certainty rules and checking for reasonableness.It resolves the inconsistencies in massive amounts of flash flood survey data that result from inconsistencies.The DMMoran model to validate the quality of the data of CFFSE and meet the requirements of the CFFSE project.Secondly,evaluation of spatio-temporal distribution and driving forces of historic flash flood events.Based on the validation of the quality of CFFSE data,the age variation,annual variation,daily variation,spatial distribution and the driving forces of more than 60,000 flash flood events from 1950 s are analyzed by methods of regression analysis,Geodetector,etc.the results show flash flood has obvious characteristics of spatial and temporal variation,the driving force has significant spatio-temporal heterogeneity.The center of gravity is offset from northeast to southwest;the correlation coefficient of precipitation indicators with the annual variation of flash flood disasters were between 0.426 and 0.513,and the correlation with daily variation of flash flood disasters were between 0.757 and 0.838.The ecological zones can explain 64% of spatial variation of flash flood events,and other factors are sorted by precipitation,human activities,and surface environment,with the power of determinants(PDs)are 0.317-0.440,0.175-0.299 and 0.066-0.193 respectively.Thirdly,evaluation of spatio-temporal heterogeneity of driving forces of flash flood.fifteen possible driving factors related to the precipitation,surface environment and human activities from 1951 to 2015 in 11 ecological zones of China are collected to conduct a quantitative spatial evaluation of variations in driving forces of flash flood by the methods of Getis-Ord Gi*,Factor detector,Interaction detector,etc.The results showed that there is obvious spatial heterogeneity in the forces driving the spatial distribution of flash floods.Precipitation was the major driving force in 8 ecological zones,and the power of determinants(PDs)were 0.176-0.756.Landform was the major driving force in the 3 other ecological zones,and the PDs were 0.401-0.714.The interactive influence of precipitation and landform was much greater than that of any single factor,with PDs of 0.478-0.901.The quantitative analysis of the two periods of 1995-2004 and 2005-2014 in Sichuan province shows that heavy precipitation is the primary driving factor for the spatial distribution of flash floods.The PD of heavy precipitation in those two periods is 0.771 and 0.711,respectively,which shows that the precipitation is the main factor causing the temporal and spatial variability.Followed by elevation and population density,the PD of elevation in the two periods is 0.593 and 0.734,respectively,and the PD of population density in two periods is 0.646 and 0.776,respectively.The analysis shows the daily precipitation causing flash floods has a tendency to decrease in 2005-2014,and the impact of human activities on the spatial distribution of flash floods is strengthened.Fourthly,Hazard analysis of flash flood disaster based on spatial heterogeneity.Based on the study of spatio-temporal distribution and heterogeneity of driving forces of flash flood in China from 1951-2015,hazard analysis of prevention area is conducted by comprehensive analysis method,with different weight of 15 evaluation factors in each ecological zone according to the preamble result of factor detector.The evaluation factors are collected by small watersheds from 10 km2 to 50km2 which cover 3.65 million km2 of mountainous area and acquired by updating original value according to the result of risk detector.GIS map is applied to express the distribution of hazard.Comparison of the analysis result and the existing hazard map shows that the analysis result is more consistent to the occurrence of flash floods.This conclusion is also proved by comparing the distribution of occurrence of flash floods and hazard assessment.Finally,developing a data management and service software platform for National Mountain Flood Investigation and Evaluation dataset.Based on the spatio-temporal data model of flash flood(FFSTDM)and the spatio-temporal data cube,a data management and service software system for flash flood data is designed and developed by using the service-oriented software architecture.The system integrates the analysis model of DM-Moran,GIS-ETL,Geodetector,etc.To serve as a data service and decision support platform for flood prevention,the system realizes 5 modules: data quality audit,data analysis and mining,data display,online statistics,data sharing and service. |
PDF Full Text Request