植被理化参数与许多有关植物物质能量交换的生态过程密切相关,定量分析植被反射光谱对理化参数的敏感性是遥感反演理化参数含量的前提。本文采用EFAST(Extended Fourier Amplitude Sensitivity Test)全局敏感性分析方法,利用PROSAIL辐射传输模型分析了冠层疏密程度对叶片生化组分含量、冠层结构以及土壤背景等多种参数敏感性的影响,并对植被理化参数反演所需先验知识的精度问题进行了初步探讨。研究表明:(1)对于较为稠密的冠层,可见光波段的冠层反射率主要受叶绿素含量的影响,近红外和中红外波段的冠层反射率主要受干物质量和含水量的影响;(2)对于稀疏的冠层,LAI是影响400—2500 nm波段范围内冠层反射率的最重要参数,土壤湿度次之,叶片生化参数对冠层反射率的敏感性较低;(3)在已知稀疏冠层LAI的情况下进一步确定土壤的干湿状态,可显著提高冠层反射率对叶绿素含量的敏感度,有助于稀疏冠层叶绿素含量的反演。

Analyzing the effect of vegetation biochemical and biophysical variables to canopy reflectance and defining the importance of variables are useful in constructing reasonable spectral indices and inverse vegetation biochemical and biophysical variables accurately. In this study, we qualitatively calculated the sensitivity of canopy reflectance to leaf biochemical variables (chlorophyll, carotenoid, water content, and dry matter), canopy structure parameter (Leaf Area Index (LAI)), and the background of soil (soil moisture). Considering that canopy reflectance easily becomes saturated with the increase of LAI, the sensitivity change of those parameters in different canopy density scenes is analyzed. We also discussed the precision of a priori knowledge used in vegetation biophysical and biochemical parameter inversion.
In this study, PROSAIL model (coupled with PROSPECT-5 leaf optical model and 4SAIL canopy radiative transfer model) was used to obtain adequate data, including vegetation variables and the corresponding canopy reflectance spectrum, which are impossible to obtain with in-site measurement. The adopted sensitivity analysis method was the extended Fourier amplitude sensitivity test (EFAST). This method first defined a search curve to scan the multidimensional space of model input parameters, and then the samples of model input parameters were generated by searching each axis of the multidimensional space at different frequencies. These samples were entered into the models to obtain the model output value. Fourier decomposition was used to compute the first-order and total-order index.
The result shows that the sensitivity of canopy reflectance to vegetation parameters is strongly related to canopy density. For medium and high canopy density, the canopy reflectance of VIS is mainly affected by C_ab, and C_m and C_w explain the bulk variation of canopy reflectance in NIR and SWIR. The canopy reflectance is slightly responsive to the variation in LAI and soil moisture. Results indicate that the requirement for accurate estimation of LAI is particularly urgent for very thick vegetation. For low canopy density, LAI is the most important variable that influences the canopy reflectance in NIR and SWIR regions, and the contribution of C_m and C_w is covered by LAI. Results show that estimating equivalent water thickness and dry matter content is difficult when LAI is low. Given that the canopy is sparse, the background of soil has a significant effect on canopy reflectance.
With regard to the accuracy requirement of priori knowledge, the result shows that the priori knowledge, which may be able to distinguish dry or wet condition of soil, is enough to obtain the valid inversion result of vegetation biochemical and biophysical variables for low canopy density.
In this study, we quantitatively analyzed the effect of canopy density on the sensitivity of canopy reflectance on various vegetation and background parameters with PROSAIL radiative transfer model and EFAST global sensitivity analysis method. Results obtained in this study can be used to choose and improve the inverse methods according to the real condition of the study area. In addition, we discussed in general terms the accuracy requirement of priori knowledge only for the spare canopy region.