基于Landsat 8和Sentinel-1的广东青云山自然保护区森林生物量反演

Forest Biomass Inversion in Qingyunshan Nature Reserve,Guangdong Province Based on Landsat 8 and Sentinel-1 Da

森林是陆地生态系统的重要组成部分,对全球碳循环具有重要意义。森林生物量研究有助于深化对森林生态系统的理解。以广东省青云山省级自然保护区的森林为研究对象,基于Landsat 8光学遥感数据和Sentienl-1合成孔径雷达(SAR)数据,结合固定样地数据等,采用随机森林(random forest,RF)和支持向量机(support vector machine,SVM)方法分别构建森林地上生物量反演模型。结果表明,1)在单一遥感数据源和联合2种遥感数据源情况下,随机森林算法的预测精度均高于支持向量机;2)不同数据源的预测精度顺序是Landsat 8+Sentinel-1>Landsat 8>Sentinel-1;3)青云山自然保护区的平均生物量为132 t/hm^(2),平均生物量范围为93~171 t/hm^(2),总生物量为892466 t。研究认为,Landsat 8+Sentienl-1联合遥感数据源和随机森林方法在森林生物量估算中具有良好的应用价值。

Forest is an important part of terrestrial ecosystem and plays an important role in global carbon cycle.Forest biomass research contributes to the deeper understanding of forest ecosystems.The inversion models of forest above-ground biomass for the forests occurring in the Qingyunshan Nature Reserve(QNR)in Wengyuan,Guangdong Province were constructed by employing methods of random forest(RF),support vector machine(SVM)based on such remote sensing data as Landsat 8 and Sentienl-1,and permanent sample plot data.The results showed that 1)the prediction accuracy of RF method was higher than that of SVM in the case of using either Landsat 8 or Sentienl-1,or both.2)From the angle of comprehensive accuracy of remote sensing estimation,the order of data sources was cooperative remote sensing data>optical remote sensing data>synthetic aperture radar data.3)The average biomass of QNR was 132 t/hm^(2),ranging from 93 t/hm^(2) to 171 t/hm^(2),and the total biomass was 892466 t.The results indicates that collaborative Landsat 8 and Sentienl-1 data sources and RF method plays important roles in the procedure of forest biomass estimation.