Remote Sensing Estimation Of Grassland Carbon Storage During Dry Season At Shengjin Lake

Wetland carbon sequestration and storage has attracted more and more international attention under the background of global climate change.The global wetland area just covers only about 3.2 to 9.7 percent of the earth surface,whereas wetlands account for about 18 to 30%of the global terrestrial carbon pool.Wetland vegetation is the key driving force of wetland ecosystem dynamic change,which affects the structure and function of ecosystem.Wetland vegetation biomass is a critical index to assess the carbon sequestration capacity and plays an important role in regional and global carbon cycles.Accurate quantification of aboveground and underground biomass of wetland vegetation helps to understand the dynamic changes of wetland ecosystem.The large area of seasonal grassland,in the middle and lower reaches of the Yangtze River floodplain during the dry season,not only provides high quality habitat and food resources for large number of wintering waterbirds,but also is an important part of the carbon sequestration function of lake wetland,which has important practical significance.Biomass estimation by traditional measurement methods are time-consuming and laborious,the fusion of multi-source remote sensing data brings a new approach to estimate biomass of wetland grassland.This paper took the grassland vegetation of Shengjin Lake wetland during the dry season as the research object,combined the Sentinel-1（S1）synthetic aperture radar（Sar）data and Sentinel-2（S2）multispectral（MSI）remote sensing data to extract remote sensing feature variables,including bands,backscattering coefficient,vegetation indices and textures.Random forest regression algorithm（RF）and extreme gradient boosting algorithm（XGBoost）were used to build the inversion model of above-ground biomass with the different combination of remote sensing features and the measured biomass data.In addition,we predicted the underground biomass via the allometric models.Finally,we obtained the carbon storage of the grassland in the Shengjin Lake wetland by using the measured organic carbon content of the aboveground and underground biomass in the laboratory.The main results are as follows:1.Both models have achieved good performance by combining all the features to model the aboveground biomass.There is better inversion accuracy with a Root Mean Square Error（RMSE）of 29.858 g.m^-2,R² of 0.842,and RMSE%of 0.100 for the RF model than that for the XGBoost model with a RMSE of 24.856 g.m^-2,R² of 0.889,and RMSE%of 0.084.And the comparative analysis showed that the XGBoost regression algorithm performed better in estimating the AGB of grassland vegetation with the lower RMSE,RMSE%values and higher R² values.Furthermore,the XGBoost algorithm is more effective in dealing with the overestimation and underestimation problems.2.The S2 red-edge vegetation indices have better performance for estimating grassland AGB than the common vegetation indices.The introduction of S2 textures has a positive effect on estimating AGB of grassland vegetation,whereas the performance is poor when used alone.The integration of S1 and S2 remote sensing data can effectively improve the modeling accuracy of wetland grassland AGB inversion.3.The UGB of grassland vegetation is more than the AGB in the Shengjin Lake wetland during early dry season.There is a significant allometric relationship between AGB and UGB of grassland vegetation,and the power function model fitted well（R²=0.842,P<0.001）.4.The average carbon content of the UGB is slightly higher than that of AGB in the Shengjin Lake wetland（the average carbon content of aboveground biomass is40.79%,and that of the underground biomass is 41.00%）.The carbon storage of AGB is 2479.18 t,and the carbon storage of UGB is 8230.52 t.This paper explored an effective way to estimate the vegetation carbon storage of seasonal flood lake wetland,which provided a demonstration for quantifying the vegetation carbon storage of the lake wetlands in the middle and lower Yangtze River floodplain,and has a positive significance for the rapid assessment of carbon sequestration potential of lake wetlands.