基于Landsat 8新邵县森林碳密度遥感反演研究

Forest carbon density remote sensing inversion research based on Landsat 8 in Xinshao county

准确估算森林碳密度是研究森林生态系统的核心。基于Matlab工作平台,以森林资源连续清查(湖南省第七次复查)及同期Landsat 8影像为本底,建立非线性回归模型、RF随机森林模型和RBF径向基神经网络模型进行森林碳密度反演。结果表明:RBF神经网络精度最高,决定系数为0.96,均方根误差为1.33 t·hm^-2,很好的拟合了样地实测碳密度;RF随机森林优于非线性回归模型,拟合精度、均方根误差分别为0.91、2.50 t·hm^-2;非线性回归模型精度最低,决定系数和均方根误差分别为:0.62、3.87 t·hm^-2。故应用RBF神经网络对森林碳密度的反演具有很好的效果。

Accurate estimation of forest carbon density is the research core on forest ecosystems. Based on the Matlab work platform, and the forest resources were continuously inspected of ‘The seventh reexamination of Hunan province’ and the Landsat 8 image was used as the corresponding period as the copy, to establish the nonlinear regression model, RF random forest model and RBF radial basis neural network model for the purpose of forest carbon density inversion. The results show that: the RBF neural network has the highest accuracy, and the determination coefficient is 0.96, and the root mean square error is 1.33 t·hm^-2. It matches the measured carbon density sample beautifully. The RF random forest superior to non-linear regression model which are 0.91 and 2.50 t·hm^-2 respectively. The nonlinear regression model has the lowest accuracy, and the determination coefficient and the root mean square error are 0.62 and 3.87 t·hm^-2 respectively. Therefore, the application of RBF neural network to the inversion of forest carbon density has a good effect.