Monitoring and analysis of 2020 flooding at Poyang Lake based on Sentinel-1B SAR data

Flooding is one of the most destructive and sudden natural disasters, and it is of great significance to quickly and accurately extract the flood inundation range and assess the disaster situation for emergency rescue and disaster loss reduction. Flood monitoring using SAR backward scattering coefficient often needs to extract the pre-disaster water body and the water body in the disaster firstly, and then carry out the phase subtraction operation in order to determine the flood inundation range. In this paper, based on the object-oriented classification method, we constructed a rapid extraction method of flood range with joint backward scattering coefficient and interferometric coherence. Taking the July 2020 flooding in Poyang Lake Basin, Jiangxi Province as a test area, the 2-view Sentienl-1B GRD and 2-view Sentinel-1B SLC dual-polarized data before and during the disaster are used as the data source, and firstly, the RGB pseudo-color composite display is carried out by introducing the interferometric coherence to analyze whether the flood disaster information can be accurately reflected, which provides a basis for the subsequent selection of the flood samples, and then, the flood extent is quickly extracted by using the object-oriented classification method. Then, the feature fusion is carried out by using pre-disaster and in-disaster dual-polarization backward scattering coefficients and pre-disaster-in-disaster interferometric coherence, and finally, the direct classification extraction and segmentation of flood, normal water, and non-water ranges are accomplished based on the object-oriented Support Vector Machine Classification method, and combined with the data of the AI Earth China 10-meter Feature Classification product to carry out the disaster assessment. The experimental results show that the overall accuracy and Kappa coefficient of the method in this paper are 95.79% and 0.88, respectively, of which the extraction accuracy of the flood range reaches 94.57%. The method constructed in this paper is fast and direct compared with the traditional flood extraction method, and can provide new ideas for flood emergency response.