Flood Monitoring In Poyang Lake Area For Rice Inundation Assessment

The Poyang Lake region in the south of China is an important rice production area,but the region is cloudy and rainy all year round,with frequent flooding.This situation inevitably leads to a large area of surrounding farmland being submerged,causing incalculable agricultural economic losses.Microwave remote sensing has the ability to penetrate clouds and fog,and the use of this technology to obtain information on the extent changes of flooded water bodies and the distribution of rice cultivation in long time series is of great significance for the assessment of agricultural disasters.This paper presents a flood monitoring method based on microwave data for water body and rice extraction.The method firstly addresses the problem that it is difficult to obtain optical images and SAR images are interfered by the shadow of mountains in the southern region during the flood season,takes Sentinel-1 SAR data as the main data source,takes Poyang Lake area as the study area,fuses topographic features,and combines decision tree,random forest and U-Net models for comparative analysis of water body extraction to achieve accurate extraction of flooded water bodies and dynamic monitoring of their changes.Then,for the problem of limited application of long time series optical remote sensing in Poyang Lake area,based on Sentinel-1 SAR time series data and Sentinel-2 data,combining time series SAR features,red edge band,EVI and LSWI index features,random forest algorithm is used to construct rice extraction model and obtain accurate rice planting information.Finally,the spatio-temporal analysis of flooded water bodies based on the best water body extraction method is carried out,and the results of rice extraction are used to analyze the damage of rice and achieve a preliminary assessment of rice damage.Finally,the spatio-temporal analysis of flooded water bodies based on the best water body extraction method is carried out,and the results of rice extraction are used to analyze the damage of rice and achieve a preliminary assessment of rice damage.The research shows that:(1)The overall accuracy of the results of the three water body extraction methods proposed in this study incorporating topographic features is above 88%,among which the water body extraction method based on the U-Net model has the highest accuracy and can quantitatively perform fast and accurate extraction and long time series monitoring of a wide range of water bodies,which can provide theoretical reference and technical support for the accurate extraction of water bodies in southern floods.(2)The random forest rice extraction method based on multi-source datasets can more effectively improve the extraction accuracy of rice planting information,and the overall accuracy of 92.67%and a Kappa coefficient of 0.91.These results can serve as a valuable reference for extracting rice planting information in regions of southern cloudy and rainy weather,with potential practical applications.(3)The scope of water body in Poyang Lake area shows the trend of"rapid expansion and slow receding"during the flooding period,in which the area of water body reached the peak of4080 km~2on July 14,2020,and the key affected areas of rice in this flooding are mainly concentrated in Yongxiu County,Poyang County,Xinjian District and Yugan County,and the area of affected area can be obtained for the preliminary assessment of agricultural disaster.The flood monitoring method based on water bodies and rice has high accuracy and feasibility.Compared with traditional monitoring methods based on water level and rainfall,this method is not limited by weather and water level measurement equipment,and enables more accurate monitoring and damage assessment of large area flooding.