Remote Sensing Estimation Of Aboveground Biomass In Subtropical Forests Based On Active Remote Sensing Data,Passive Remote Sensing Data,and Spaceborne LiDAR Data-A Case Study Of Chenzhou City,Hunan Province

The aboveground biomass（AGB）of forests is a critical parameter for monitoring their carbon sink capacity and assessing carbon balance.AGB estimation provides essential scientific information and research support for regional ecological development planning,forestry resource management and forestry carbon trading.With the advancement of remote sensing technology,multisource remote sensing data can provide diverse forest information,such as multispectral remote sensing data,Synthetic Aperture Radar（SAR）data,and Light Detection and Ranging（Li DAR）data,etc.The use of multisource remote sensing data to estimate AGB has become a research trend.Thus,this study used the continuous inventory of forest resources in Chenzhou City in 2019 as sample plot data.The study extracted feature factors including Forest Canopy Height（FCH）through Sentinel series active and passive remote sensing data,ICESat-2 satellite-borne Li DAR data,and SRTM digital elevation data.The study employed four models to estimate AGB,including Multiple Stepwise Regression（MSR）,k-Nearest Neighbors（k NN）,Artificial Neural Network（ANN）,and Random Forest（RF）.This study further explored the benefits of multispectral data,SAR data,and their active-passive combination for AGB estimation.The study also incorporated Interferometric Synthetic Aperture Radar（In SAR）technology combined with satellite-borne Li DAR data to extract FCH and analyzed the impact of adding FCH to AGB estimation accuracy.The research results indicated that:（1）This study extracted 57 active remote sensing feature variables such as normalized backscatter coefficient,texture features,polarization decomposition components,and coherence coefficient from Sentinel-1.This study also extracted 223 passive remote sensing feature variables such as band reflectance,texture features,vegetation indices,and principal component analysis bands from Sentinel-2.In addition,7 terrain factors were extracted using SRTM DEM for assistance.Then,the study used stepwise and RF importance ranking methods to screen out appropriate feature variables as model predictors.Among all models,the dominant feature variable was texture,especially the average statistical feature of texture had the most participation in modeling in AGB inversion,which played a vital role.Furthermore,the primary factors involved in Sentinel-2 prediction were texture features of visible light bands（B2,B3,B4）,and vegetation indices such as TSAVI,PSSRa,RVI,and IPVI also played a critical role in Sentinel-2 prediction.For Sentinel-1,topographic factors effectively assisted SAR data in predicting AGB.Moreover,when using Sentinel-1 VV+VH polarization data to predict AGB,more attention should be paid to the information provided by VH polarization data.（2）Sentinel-2 multispectral data can effectively estimate AGB in subtropical regions,and the predictive performance in all models was better than Sentinel-1.Compared with a single-source remote sensing model,the synergistic inversion of Sentinel-1 and Sentinel-2 active-passive data improved the prediction accuracy of AGB,leading to better large-scale estimation of AGB.However,the estimation accuracy of regional AGB using Sentinel-1 SAR data alone needed further improvement.（3）Compared with the four modeling methods of multiple stepwise regression,k-nearest neighbor,artificial neural network,and random forest,the random forest model showed the best predictive performance in both single-source remote sensing and active-passive combined remote sensing.In the model combined with active-passive data,R² reached a maximum of 0.69,RMSE was 24.26 t·ha^-1,and MAE was 30.08 t·ha^-1,achieving remarkable estimation effects of AGB in the subtropical forest region.Hence,random forest could be used as the preferred model for remote sensing estimation of forest AGB in this area.（4）Owing to the short wavelength of the C band,the SAR signal mainly involved the forest canopy information,thereby the ground height obtained from Sentinel-1 In SAR data can be used as the forest surface height.ICESat-2 corrected SRTM digital elevation data was used as DEM.The combination of the two was used for subtraction calculation to obtain the FCH of the study area.There was a significant correlation between the extracted FCH and the average tree height of the sample plot（r=0.54,P≈0<0.01）and a significant correlation with the aboveground biomass（r=0.32,P=0.003<0.01）,which performed better than the downloaded GLAD global forest canopy height product.Adding FCH to the AGB inversion model played a critical role,and the prediction accuracy of biomass from different data sources had improved.（5）The study used active and passive remote sensing data to estimate the forests aboveground biomass of Chenzhou City in 2019 through the random forest model,which total biomass was 69.44×10⁶ t,and the average was 51.87 t·ha^-1.The area of moderate-low biomass percentage was the highest,with percentages of low,moderate-low,moderate,moderate-high,and high biomass areas,respectively,3.44%,39.46%,29.26%,23.10%and 4.74%,biomass percentage was 1.10%,24.54%,30.87%,34.40%and 9.09%,respectively.Overall,the forest ecological protection in Chenzhou City was good,and the spatial distribution of aboveground biomass showed an east-high,west-low trend.