基于Landsat的高山松地上生物量动态变化估测模型构建

以香格里拉市高山松为研究对象，基于1987—2017年的国家森林资源清查固定样地和对应年的Landsat TM/OLI数据提取对应的遥感因子，计算30 a期间生物量和遥感因子的5、10 a定期变化量、年平均变化量、5、10 a变化率；采用多元线性回归（MLR）和随机森林（RF）方法构建地上生物量模型，提高高山松森林AGB动态变化遥感估测的精度。结果表明：基于5种变化量类型的RF模型效果均优于相应的MLR模型，RF模型中采用5 a变化率效果最好，其拟合R²为0.956，RMSE为0.664 t/(hm²·a)，预测结果中RMSE为2.285 t/(hm²·a)；纹理因子在高山松地上生物量变化量建模中贡献最大。采用遥感因子变化率构建的高山松地上生物量动态变化估测模型，提高了生物量变化的估测精度，可为森林地上生物量的动态估测研究提供参考。

Estimating the Dynamic Changes of Aboveground Biomass of Pinus densata Based on Landsat

In this research, Pinus densata in Shangri-La was taken as the research object. Based on permanent sample plots of National Forest Inventory data in 1987–2017 and the corresponding Landsat TM/OLI data, the remote sensing factors were extracted. Based on these factors, their variables were calculated for aboveground biomass modelling using multilinear regression (MLR) and random forest (RF) methods. The variables include, the average annual variation, the 5-years-interval and 10-years-interval variations and the change rate. The results showed that model effect of RF was better than that of MLR in the 5 variables. In RF models, the best modeling effect was to use the change rate of 5 years, the R2 was 0.956, the RMSE in the modeling effect was 0.664 t/(hm2·a), and the RMSE in the prediction effect was 2.285 t/(hm2·a). The texture factors are the most contributing factor in the aboveground biomass change of P. densata modeling. The dynamic change estimating model of aboveground biomass of P. densata based on the remote sensing factors' change rate can improve the estimation accuracy of biomass change, and it provides reference for the dynamic estimation of forest aboveground biomass.