Forest Structural Variables Retrieval Using EO-1 Hyperion Data in Combination with Linear Spectral Unmixing and an Inverted Geometric-Optical Model
基于高光谱遥感数据提取森林结构参数的研究

The potential of EO-1 Hyperion data combined with linear spectral unmixing and an inverted geometric-optical model for the retrieval of forest structural variables in the Longmenhe broadleaved forest natural reserve,located in the Three Gorges region(China),is studied in this paper.Based on the principle of Li-Strahler geometric-optical model,we derive the per-pixel reflectance as being a linear combination of four scene components(sunlit canopy/sunlit background and shaded canopy/shaded background).The fraction of each component is subsequently related to several forest structural attributes.With the advantage of having hyperspectral data,we use linear spectral unmixing to separate the above classes present in an atmospherically corrected Hyperion image with support of extensive in situ measurements.In addition,we include DEM derived parameters(slope and aspect) and measured canopy structural parameters for different forest communities to invert the geometric-optical model and retrieve the pixel-based variables forest crown closure(CC) and crown diameter(CD).In total 37 sample plots collected in the Longmenhe study region are used for validation,and the results of the above parameters show a good agreement(e.g.,R2CC=0.61/RMSE=0.046;R2CD=0.39/RMSE=0.984).