Spatiotemporal Characteristics Of Flood Disasters In Karst Cluster Depression Areas Of Southwest China Using Sentinel-1 Time Series Data

Karst area in southwest China is located in the monsoon area.Short-time heavy rainfall and the special geological conditions in this region make the flood disaster in this region more special than that in the same latitude region.The severity of flood disaster has a great impact on agricultural production and people's life.Therefore,an accurate and effective flood monitoring method is of great significance for the protection of agricultural production and people's life in this area.In this study,a large number of sample data of the occurrence range,occurrence time and duration of flood disasters were extracted by combining field collection and remote sensing visual interpretation.Sentinel-1 remote sensing time series data were used as the research data source.The distribution characteristics of statistical features of radar image time series under different disaster conditions are analyzed,and the correlation between these statistical features and the duration of flood disaster is analyzed,which proves the feasibility of extracting time series by using statistical features of time series.Since the time resolution of remote sensing images is always longer than the duration of some flood disasters,the Adaptive Neuro-Fuzzy Inference System was used in this study to make fuzzy approximation.Combined with the statistical characteristics of time series and other auxiliary data sets,the accuracy of duration monitoring of flood disasters is improved.At the same time,based on the physical characteristics of water bodies in microwave reflection,mutation information of radar remote sensing time series was extracted through Mann-Kendall trend analysis to monitor the occurrence time of flood disasters in karst areas in Southwest China.The analysis of the results shows that the variance and skewness of the time series statistical characteristics of the backscattering rate of radar images play an important role in distinguishing the flood-affected areas from the non-flood-affected areas,and the mode has an obvious relationship with the duration of the flood-affected areas.Through the comprehensive utilization of these features in the adaptive fuzzy logic system,the accuracy of flood duration extraction is improved effectively.These statistical features can reduce the impact of other surface changes and mountain shadows,and the extraction accuracy of flood disaster duration can reach about 3-4 days.The results obtained by the method of this study show that the occurrence time of flood disasters in this region is mainly from May to July,and the areas with the duration of flood over the middle period(more than 15 days)can account for about 1% of the region,and the severe areas are mainly concentrated in the central region,such as Pingguo County and Long 'an County in Guangxi Province.At the same time,the spatial distribution characteristics of flood disasters are closely related to meteorological factors,soil conditions and geological conditions in this area.