The sparse crown along both riversides of the Tarim River plays an important role in firming the sand and restraining the
desertification. It is very difficult to obtain the spectrum information from the remotely sensed data because of the low
percentage of coverage of the sparse vegetation, which affects the classification accuracy of the identification of ground
objects and the extraction of vegetation biophysics. It is a key obstruction in developing the quantification of the RS
technology. Taking the sparse vegetation at the Tarim River Basin as the research object, this paper predicts the surface
bidirectional reflectance of the discontinuous plant canopies in the extremely arid based on the observed ground
spectrum. Two different approaches are presented for the tree and the shrub. The first is to simulate the spectrum of the
tree with the Geometric Optical-Radiative Transfer model based on ground observation. In the second approach,the
spectral responses of sparse shrub and bare soil have been simulated using the linear Geometric Optical (GO) model.
Comparing the simulated bidirectional reflectance with actual remote sensing data (EO-1), the spectral differences of
these data are analyzed.
In this paper, the goal is to found indices best for Cab estimation with leaves and heperion pixels. There are several indices chosen, which showed best results for Cab estimation at both leaf and canopy levels in other studies. Forty-eight typical leaves were sampled in middle and lower reach of the Tarim River, Xinjiang, China. Leaf reflectance and Chlorophyll of leaves collected. Result demonstrated that Indices such as red edge and derivative indices R750/R710, R740/R720, (R734-R747)/(R715+R720), Blog(1/R737), D715/D705,(R734-R747)/(R715+R726), (R694-R680)/(R732-R760) were shown to be the good indicators for Cab estimation at leaf. Hyperion data were acquired for Aqike section in the middle reaches of the Tarim River in Nine 28, 2006. Field data were collected at same day to coincide with the Hyperion, including Chlorophyll of each tree, LAI, green vegetation cover. LAI derived from scanopy 2006. Inventory field plots were 120m×120m quadrants, and Chlorophyll of pixel is deduced from field data of 360 trees. Generally good results are found for Cab estimation at pixel level with indices such as, (R734-R747)/(R715+R726), Blog(1/R737), (R694-R680)/(R732-R760), TCARI, TCARI/OSAVI, MCARI/OSAVI and so on. It was found that (R734-R747)/(R715+R720), Blog(1/R737), D715/D705, (R734-R747)/(R715+R726), (R694-R680)/(R732-R760),R740/R720 were successfully test on leaves and piexls. On the other hand, the "modified" indices (TCARI, MCAVI, TCARI/OSAVI, MCARI/OSAVI) already give good results at the piexl level.