华南地区亚热带树木叶面积指数的高光谱反演研究

Hyperspectral Inversion of Leaf Area Index of Subtropical Vegetation in South China

摘要:

为构建树种叶面积指数的估算模型，以NDVI、RVI、FREP、CI_Green、CI_Red-edge、MSAVI₂为高光谱特征变量，通过统计分析，确定反演树种叶面积指数的最佳光谱特征变量，构建华南农业大学校园内50种亚热带树木的叶片反射率和叶面积指数（LAI）模型。结果表明，6种高光谱特征变量与树种叶面积指数间都具有极显著相关性，其中红边位置反射率（FREP）和比值植被指数（RVI）与LAI的拟合方程的R²都大于0.8，决定系数分别为0.820和0.811。经过精度验证，FREP估算的均方根误差（RMSE）只有0.13，该回归模型为估测亚热带典型树种的叶片LAI最佳模型。从高光谱遥感的角度结合亚热带植被的群落结构特点来看，建立的红边位置光谱反射率与叶面积指数的回归模型普遍具有较高的拟合度，所以利用高光谱特征变量反演亚热带树木叶片的叶面积指数等植被参数的应用前景较好。

Abstract:

In order to determine the optimal hyperspectral characteristic variables of subtropical tree species and construct the estimation model of leaf area index (LAI), the leaf reflectance and LAI of 50 tree species in the campus of South China Agricultural University (SCAU) were measured. At the same time, the relationship model of LAI with the six hyperspectral characteristic variables, including NDVI, RVI, FREP, CI_Green, CI_Red-edge and MSAVI₂, were constructed through statistical analysis, respectively. The results showed that there were significant correlations between the six hyperspectral characteristic variables and LAI of tree species. The R² of fitting equations between LAI with red edge position reflectivity (FREP) and the ratio vegetation index RVI were more than 0.8 with correlation coefficients for 0.820 and 0.811, respectively. The root mean square error (RMSE) of FREP estimation is only 0.13, so the regression model is the best model for estimating the LAI of typical subtropical tree species. Combining subtropical vegetation community structure and hyperspectral remote sensing, the regression model between red edge position reflectivity and leaf area index generally has a high fitting degree. Therefore, using hyperspectral characteristic variables inverted subtropical leaves of the leaf area index and other vegetation parameters had better application prospects.