Remote Sensing Estimation Of Dominant Plant Community Biomass At Poyang Lake Wetland

Natural grasses in Poyang Lake wetland are mainly distributed in Poyang Lake National Nature Reserve and Poyang Lake Nanjishan Wetland National Nature Reserve.They are important foraging ground and habitat for Siberian wintering migratory birds.Understanding the spatial distribution of wetland plant communities and accurately quantifying wetland plant biomass are of great practical significance for protecting rare and endangered migratory birds and their habitats,monitoring wetland vegetation changes,and maintaining the carbon cycle and sustainable development of Poyang Lake region.In this study,grass flat vegetation in these two national nature reserves was taken as the research object.Based on field survey and multi-source remote sensing data,different remote sensing feature variables were combined for random forest classification,and the optimal combination was selected to complete the extraction and mapping of dominant plant community information in Poyang Lake Wetland reserve.Combining traditional sampling and remote sensing above-ground biomass inversion modeling,the optimal inversion model of aboveground biomass of wetland plant community was discussed,and the above-ground biomass of wetland plant community was estimated in dry season.The main conclusions are as follows:(1)Through field investigation and literature investigation,the land features of Poyang Lake Wetland Reserve were divided into six types: water,mudflat,Reed and Miscanthus,Carex and Phalaris arundinacea,Zizania and Polygonum and other types.According to the time series curves of different plant communities on key feature variables,it was found that distinguishing wetland plant communities needed to combine multi-source and multi-temporal data for feature selection.From the perspective of feature selection,"optimal spectral feature + optimal water-vegetation index + optimal texture feature + optimal backscattering coefficient" produced the highest classification accuracy(OA=90.78%,Kappa=0.8989).The red-edge index has more advantages than the common vegetation index in the classification of wetland plant community.The collaboration of Sentinel-1 and Sentinel-2 data can effectively improve the classification accuracy of wetland plant communities.(2)Combining the measured above-ground biomass and 23 remote sensing feature variables,a biomass inversion model was established using cubic polynomial,stepwise regression and random forest regression models.The results show that for a single feature variable,the cubic polynomial regression model has a certain simulation ability,but when the number of samples is small,the stepwise regression model is better than the cubic polynomial model.Random forest regression has a better performance in retrieving above-ground biomass of vegetation in Poyang Lake wetland reserve.(3)At the regional scale,the addition of Sentinel-1data was able to improve the precision of the model in the inversion of biomass over wetland vegetation sites,and the red edge index based on Sentinel-2 data was more advantageous than the common vegetation index in the overall estimation of biomass over wetland vegetation sites.The above ground biomass inversion model of different plant communities was built at the community scale,and its model fitting effect was better than the overall modeling.Obtaining the fine spatial distribution patterns of wetland plant communities is extremely important for the precise inversion of aboveground biomass of wetland plant communities.(4)The average aboveground biomass of Poyang Lake Wetland Reserve in 2022 dry season was 814.86g/m~2,and the total biomass density was about 1.1×106t.The aboveground biomass distribution showed obvious spatial heterogeneity,showing a trend of "high in southeast and low in northwest".The disc-shaped sub-lakes,such as Dahu Pond,normal lake pond and middle lake pond,showed an obvious phenomenon of "low in the center and high in the periphery",and the rest showed an irregular distribution.The order of aboveground biomass of the four dominant plant communities is: Reed and Miscanthus community(1250.51g/m~2)>Zizania community(1111.69g/m~2)>Polygonum community(659.49g/m~2)>Carex and Phalaris arundinacea community(626.32g/m~2).