在植被光合有效辐射吸收率(FAPAR)遥感估算中被广泛采用的植被指数法,其估算精度往往受到"红波段吸收峰"峰值点光谱反射率易饱和特征的影响。考虑到高光谱吸收特征参数能较好地诠释地物光谱吸收特征的细节信息,基于微分法与包络线去除法研发"高光谱曲线特征吸收峰自动识别法"识别对FAPAR敏感的特征吸收峰,再结合连续统去除法以及光谱吸收指数(SAI)提取FAPAR的高光谱吸收特征参数,构建估算天然草地冠层水平FAPAR的高光谱吸收特征参数模型。结果表明:(1)天然草地冠层FAPAR与高光谱吸收特征参数具有很好的相关性,其中,"红波段吸收峰"SAI对FAPAR变化最为敏感,在植被覆盖度较高时,其饱和性相比"红波段吸收峰"峰值点反射率与归一化植被(NDVI)值有较大的提升。(2)以"红波段吸收峰"SAI为变量的对数方程为FAPAR的最佳估算模型,在植被覆盖度处于中与高时,其FAPAR预测精度比NDVI模型有不同程度的提高。研究采用的高光谱吸收特征参数一定程度上弥补了部分植被指数因饱和问题在估算FAPAR时的不足,可作为植被FAPAR反演的新参数,适用于中、高覆盖度的天然草地FAPAR监测。

The Vegetation Index method is often used in remote sensing estimation of vegetation Fraction of Absorbed Photosynthetically Active Radiation(FAPAR), but the estimation accuracy of the Vegetation Index method is often affected by "saturation" of spectral reflectance in the"red band absorption peak." The article is aimed at developing new parameters to improve the estimation precision of FAPAR, particularly in average or high vegetation coverage.
Given that the hyperspectral absorption characteristic parameters can be used to interpret the spectral absorption feature details of ground object, we develop the automatic recognition method of the characteristic absorption peakof the hyperspectral curve based on the differential and envelope removal methods to identify thecharacteristic absorption peaks sensitive to FAPAR. We extract the hyperspectral absorption characteristic parameters based on the continuum removal method and Spectral Absorption Index(SAI). By combining the characteristic absorption peaks with hyperspectral absorption characteristic parameters, we build a hyperspectral absorption characteristic parameters model to estimate the FAPAR of natural grassland canopies.
The results are listed below:(1) The hyperspectral absorption characteristic parameters have a high correlation with the FAPAR, and the SAI of the "red band absorption peak" is the most sensitive parameter to the change of FAPAR. Compared with the reflectivity of the "red band absorption peak" and Normalized Difference Vegetation Index(NDVI), saturation in the high vegetation coverage level was significantly improved.(2) The best estimation model is the logarithmic equation, which takes the SAI of the "red band absorption peak" as a variable. Compared with the NDVI model, the prediction accuracy of FAPAR exhibits varying degrees of improvement while in the average or high vegetation coverage level.
The hyperspectral absorption characteristic parameters can remedy, to some extent, the defects caused by the "saturation" problem of the Vegetation Index in estimating FAPAR, can be used as a new inversion parameter of vegetation FAPAR, and can monitor the FAPAR of natural grassland in the average or high vegetation coverage level.