库布齐沙漠人工灌木林生物量的遥感估测模型构建

Construction of Remote Sensing Model for Biomass Estimation of Artificial Shrubbery in Kubuqi Desert

以库布齐沙漠沙柳(Salix psammophila)、柠条锦鸡儿(Caragana korshinskii)、沙棘(Hippophae rhamnoides)、杨柴(Corethrodendron fruticosum var.mongolicum)4种主要人工灌木林为研究对象,设置柠条锦鸡儿样地69块、沙棘样地53块、沙柳样地59块、杨柴样地39块,对每个样地选取1株标准株进行样品收集,计算全株生物量(含地下生物量)及样地生物量;以样地人工灌木林全株生物量为评价指标,以2017年陆地卫星8号携带的运营性陆地成像仪(Landsat 8 OLI)卫星遥感影像数据的遥感影响因素6个波段反射值(蓝光波段(B_(2))、绿光波段(B_(3))、红光波段(B_(4))、近红外波段(B_(5))、短波红外波段1(B_(6))、短波红外波段2(B_(7))),依据样地调查数据计算的植被影响因素5个植被指数(归一化植被指数(I_(V,ND))、增强型植被指数(I_(V,E))、比值植被指数(I_(V,R))、土壤调整植被指数(I_(V,SA))、修正的土壤调节植被指数(I_(V,MSA)))为影响因素,应用逐步回归法,构建库布齐沙漠4种人工灌木林生物量(W)的遥感估测模型,探索提升估测库布齐沙漠人工灌木林生物量及生态系统碳储量精度的方法。结果表明:与柠条锦鸡儿林生物量相关性较高的遥感影响因素为B_(2)波段反射值(ε_(B2))、B_(6)波段反射值(ε_(B6)),与沙棘林生物量相关性较高的遥感影响因素为I_(V,R)、I_(V,MSA),与沙柳林生物量相关性较高的遥感影响因素为I_(V,ND)、I_(V,SA),与杨柴林生物量相关性较高的遥感影响因素为B_(6)波段反射值(ε_(B6))。经相关分析,筛选出与生物量相关性较高的影响因素,应用逐步回归法,建立了柠条锦鸡儿、沙棘、沙柳、杨柴4种灌木林生物量的遥感估测模型;柠条锦鸡儿林生物量的遥感估测模型为多元线性回归方程,沙棘、沙柳和杨柴均为一元归线回性模型。柠条锦鸡儿林生物量的遥感估测最优模型为W_(NT)=4.364-0.008ε_(B2)+0.002ε_(B6),决定系数(R^(2))为0.49,预估精度为72.1%;沙棘林生物量的遥感估测最优模型为W_(SJ)=-3.368+8.027 I_(V,MSA),R^(2)为0.51,预估精度为62.5%;沙柳林生物量的遥感估测最优模型为W_(SL)=-10.803+23.853 I_(V,SA),R^(2)为0.47,预估精度为76.5%;杨柴林生物量的遥感估测最优模型为W_(YC)=1.643-0.0003ε_(B6),R^(2)为0.41,预估精度为72.2%。

Four typical artificial shrubbery(Caragana korshinskii,Hippophae rhamnoides,Salix psammophila and Corethrodendron fruticosum var.mongolicum)in Kubuqi desert area were selected as the research object,and 69 plots,53 plots,59 plots,and 39 plots were set up,respectively.One standard plant was selected from each plot for sample collection,and the whole plant biomass(including underground biomass)and plot biomass were calculated.The biomass of whole plant of artificial shrubbery in the sample plot was taken as the evaluation index,with the operational land imager carried by Landsat 8 in 2017(Landsat 8 OLI)satellite remote sensing image data are affected by the reflection values of six bands(blue light band(B_(2)),green light band(B_(3)),red light band(B_(4)),near-infrared band(B_(5)),short-wave infrared band 1(B_(6)),short-wave infrared band 2(B_(7)),according to the sample survey data calculation of vegetation factors five vegetation index(normalized difference vegetation index(I_(V,ND)),enhanced vegetation index(I_(V,E)),the ratio vegetation index(I_(V,R)),soil adjusted vegetation index(I_(V,SA)),modified soil vegetation index(I_(V,MSA))for influence factors of the stepwise regression method.A remote sensing estimation model for four kinds of artificial shrubbery biomass(W)in the Kubuqi desert was constructed,and a method to improve the estimation accuracy of artificial shrubbery biomass and ecosystem carbon storage in the Kubuqi Desert was explored.The remote sensing influencing factors with higher correlation with Caragana korshinskii forest biomass were B_(2)band reflection value(ε_(B2))and B_(6)band reflection value(ε_(B6)),the remote sensing influencing factors with higher correlation with Hippophae rhamnoides forest biomass were I_(V,R)and I_(V,MSA),the remote sensing influencing factors with higher correlation with Salix psammophila forest biomass were I_(V,ND)and I_(V,SA).And the reflection value of band B_(6)(ε_(B6))was the influencing factor with high correlation with Corethrodendron fruticosum var.mongolicum forest biomass.Based on the correlation analysis,the influencing factors with high correlation with biomass were screened,and the biomass estimation models of Caragana korshinskii,Hippophae rhamnoides,Salix psammophila and Corethrodendron fruticosum var.mongolicum were established by using stepwise regression method.The estimation model of biomass of Caragana korshinskii was a multiple linear regression equation,and the regression model of Hippophae rhamnoides,Salix psammophila and Corethrodendron fruticosum var.mongolicum was a single linear regression model.The optimal model for estimating biomass of Caragana korshinskii forest was W_(NT)=4.364-0.008ε_(B2)+0.002ε_(B6),the coefficient of determination(R^(2))was 0.49,and the prediction accuracy was 72.1%.The optimal model for estimating Hippophae rhamnoides forest biomass by remote sensing was W_(SJ)=-3.368+8.027 I_(V,MSA),R^(2)was 0.51,and the prediction accuracy was 62.5%.The optimal model for estimating Salix psammophila forest biomass by remote sensing was W_(SL)=-10.803+23.853 I_(V,SA),R^(2)was 0.47,and the prediction accuracy was 76.5%.The optimal remote sensing estimation model of Corethrodendron fruticosum var.mongolicum forest biomass was W_(YC)=1.643-0.0003ε_(B6),R^(2)was 0.41,and the prediction accuracy was 72.2%.