基于多时相光学和雷达遥感的太平湖生态保护区森林地上生物量反演  

Inversion study of above-ground biomass in Taiping Lake Ecological Reserve forests using multi-temporal optical and radar data

作　　者：卢佶 张国威 吴昊 LU Ji;ZHANG Guowei;WU Hao(East China Survey and Planning Institute,National Forestry and Grassland Administration,Hangzhou 310019,Zhejiang,China)

机构地区：[1]国家林业和草原局华东调查规划院,浙江杭州310019

出　　处：《浙江农林大学学报》2023年第5期1082-1092,共11页Journal of Zhejiang A&F University

摘　　要：【目的】探究光学和雷达卫星遥感对亚热带森林生物量的反演潜力。【方法】利用不同季节时间序列的合成孔径雷达(Senitinel-1)和光学数据(Sentinel-2),对太平湖生态保护区森林地上生物量进行反演。基于后向散射系数、光谱波段、植被指数和生物物理参数,采用回归随机森林算法探究Sentinel-1和Sentinel-2在地上生物量制图中的精度,探究对亚热带森林地上生物量制图的最佳影像采集时期,评估光学和雷达遥感特征参数对提高地上生物量估计精度的贡献。【结果】Sentinel-2对研究区森林地上生物量的估计精度[决定系数(R^(2))=0.68,均方根误差(ERMS)=37.69 Mg·hm^(-2)]要优于Sentinel-1(R^(2)=0.47,ERMS=49.11 Mg·hm^(-2)),但两者联合产生了最佳结果(R^(2)=0.78,ERMS=31.56 Mg·hm^(-2))。生长季(6和9月)的光学数据和旱季(12月)获得的雷达数据结合有利于提高地上生物量估算精度。另外,Sentinel-2提取的叶面积指数(LAI)、光合有效辐射吸收比(fapar)和覆盖度(fcover)与Sentinel-1提取的VH极化和VH+VV指数对地上生物量估算具有重要的贡献度。【结论】通过联合不同季节的光学和雷达数据,明确了6、9、12月与LAI、fapar、VH极化、VH+VV指数是地上生物量反演的最佳时相和预测变量。[Objective]This study,with the inversion of the above-ground biomass(AGB)of forests in Taiping Lake Ecological Reserve using the Senitinel-1(SAR)and Sentinel-2(optical)data with different seasonal time series,is aimed to investigate the inversion potential of optical and radar satellite remote sensing on subtropical forest biomass.[Method]First,based on backscatter coefficients,spectral bands,vegetation indices and biophysical parameters,the regression random forest algorithm was used to explore the accuracy of Sentinel-1 and Sentinel-2 in AGB mapping.Then,an exploration was conducted of the optimal image acquisition period for AGB mapping of subtropical forests which was followed by an evaluation of the contribution of optical and SAR remote sensing feature parameters to the improvement of AGB estimation accuracy.[Result]The accuracy of AGB estimation for forests in the study area using Sentinel-2 data[coefficient of determination(R^(2))=0.68,root mean square error(ERMS)=37.69 Mg·hm^(−2)]was better than that of Sentinel-1(R^(2)=0.47,ERMS=49.11 Mg·hm^(−2)),but the combination of the two produced the best results(R^(2)=0.78,ERMS=31.56 Mg·hm^(−2)).The combination of optical data based on the growing season(June and September)and SAR obtained in the dry season(December)was beneficial to improving the accuracy of AGB estimation.The leaf area index(LAI),photosynthetically active radiation absorption ratio(fapar)and cover(fcover)extracted by Sentinel-2 and the VH polarization and VH+VV indices extracted by Sentinel-1 had important contributions to the AGB estimation.[Conclusion]By combining optical and SAR in different seasons,it was clarified that June,September and December are the best temporal phases for AGB inversion while its best predictors are LAI,fapar,VH polarization and VH+VV index.

关 键 词：地上生物量 Sentinel-1 Sentinel-2 不同季节 随机森林 

分 类 号：S758.5[农业科学—森林经理学]