Study On Above-ground Biomass Estimation Of Caragana Microphylla Shrub In Xilingol Grassland By Remote Sensing

In recent years,shrub-encroached has become one of the most important ecological problems in grassland ecosystems and has received widespread attention from scholars at home and abroad.Accurate estimation of above-ground biomass(AGB)of shrub vegetation is the basis for assessing carbon stocks and developing sound prevention and control measures in grassland ecosystems,which are essential for regional ecosystem conservation and system management.However,because of the relative dwarfism and spatial heterogeneity of shrub community plants,it is difficult to extract information on individual shrub from medium-and high-resolution satellite imagery and significantly increases the uncertainty of the AGB of shrub-encroached grassland vegetation.In recent years,the rapid development of unmanned aerial vehicle(UAV),which can acquire centimetre-scale optical images and high-density Light Detection and Ranging(Li DAR)point cloud data,has the potential advantage of observing vegetation at multiple scales,providing a new way of thinking in estimating shrub vegetation AGB studies on an individual and spatial basis,but there are bottlenecks in their efficiency when applied to large areas.Therefore,there is an urgent need to propose a low-cost and efficient"air-sky-ground"synergistic remote sensing estimation method to accurately assess the AGB of shrub vegetation in shrub-encroached grassland ecosystems.For Boarder Yellow Banner and Zhengxiangbai Banner in Xilingol League in2021-2022,in this study,firstly,based on the direct easy-to-measure factors such as height obtained by Real-time Kinematic(RTK)and a variety of complex factors derived from it,the study proposes an AGB estimation model for the single plant of Caragana microphylla in the study area,and achieves the indirect measurement of AGB of shrub vegetation under non-destructive conditions.Then,on the basis of a breakthrough in the extraction techniques for identifying the extent of C.microphylla based on Digital Orthophoto Map and object-oriented framework,the geometric and height variables and other features extracted from UAV Li DAR data and Digital Orthophoto Map were used as predictors of shrub vegetation AGB,and a UAV-scale model for estimating the AGB of C.microphylla was studied and constructed.Finally,an AGB estimation study of shrub vegetation in the study area at the satellite scale was carried out based on GF6-WFV remote sensing imagery,based on statistical models such as linear and non-linear.The main conclusions were the following:(1)In the comparative analysis of direct measurement methods for ground-based observations,the“standard branch”method(r=0.95,P<0.01)predicted shrub vegetation AGB more accurately and with less damage to the shrub than the“standard quadrat”method(r=0.84,P<0.01).In the correlation analysis of the Real-time Kinematic extracted indirect measures,there was a strong relationship between the single factor crown width,irregular canopy and perimeter of irregular canopy,r>0.98,P<0.01;the relationship between height(H)and the other individual factors was average,r=0.67-0.72,P<0.05.However,the canopy and height product factor,the canopy perimeter and height product factor and the volume factor,which are derived from the height factor to represent volume,can improve prediction accuracy;crown width was the best performer with AGB among the single factors(r=0.92,P<0.01),and irregular volume had the highest correlation among the complex factors(r=0.92,P<0.01).The research builds a prediction model for AGB of using a linear function with crown width as the independent variable and a power function with irregular volume as the independent variable,respectively,with the power function model giving the best estimation results.(2)Based on UAV imagery and the object-oriented nearest neighbor(KNN)algorithm can accurately identify and extract the range of single shrub outlines with an overall classification accuracy of 92.87%and an average Kappa coefficient of 0.83.The results of AGB estimation of single shrub vegetation based on UAV Li DAR and Digital Orthophoto Map(DOM)showed that canopy area(S),canopy perimeter(C),and long and short canopy width(A1,A2)were the most important predictors of AGB for single shrub vegetation,followed by elev＿canopy＿relief＿ratio and first one density＿metrics.Therefore,a prediction model for AGB of A.minor was constructed based on random forest regression using each importance factor of planar geometric features,height features and density features as variables,with a model prediction accuracy of R~2=0.84 and RMSE=310.14 g/plant,indicating that the prediction model constructed based on UAV data has good applicability for the estimation of AGB of single shrub vegetation.(3)Based on the band reflectance,vegetation index and spectrum features of the GF6-WFV multispectral images,three methods,random forest regression,stepwise regression and partial least squares regression,were used to construct a sample-scale AGB model of C.microphylla.Among the variables selected by the different methods,Band2,DVI,NDREI and Entropy index were all the most important factors in predicting the shrub AGB of the sample area.The results of the comparative model analysis showed that the R~2=0.81 for the RF model,R~2=0.45 for the SR model and R~2=0.55 for the PLSR model,with the largest coefficient of determination R~2and smaller RMSE and MAE values for the non-linear model RF,indicating higher prediction accuracy,and therefore the RF model is the best model for building the above-ground biomass of C.microphylla at large scales.In this study,an"air-sky-ground"collaborative estimation model based on ground-based measurements,UAV and satellite remote sensing data was constructed to accurately assess the AGB of of C.microphylla vegetation in shrub-encroached grassland ecosystems.The results of the study can provide new ideas for the quantitative assessment of AGB in arid and semi-arid grassland shrub vegetation,and provide a scientific basis for the accurate assessment of carbon stocks in grassland ecosystems and the formulation of reasonable prevention and control measures.