A Spectral Unmixing Approach To Leaf Area Index (LAI) Estimation At The Alpine Treeline Ecotone - Chapter X

• Main Author: Daniel G. Brown
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Published: Kluwer 2001
• Subjects: Bare Rock; Tundra
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.25.488; http://www.umich.edu/~danbrown/research/unmix.pdf

The objective of this research was to develop methods for mapping arboreal leaf area index (LAI) at the alpine treeline ecotone in Glacier National Park, Montana using Landsat Thematic Mapper (TM) imagery. A three-stage approach was tested for addressing the problem of mixed pixels in biophysical value estimation. This paper illustrates a proof of concept for this method. First, we used spectral unmixing to obtain estimates of the percentage of each pixel that was composed of tree, tundra, bare rock and shadow. Spectral signatures obtained through image interpretation were used for mixture modeling. The second step involved adjusting the pixel vegetation index (VI) values so that they represented the VI of the tree-only portion of the pixel, assuming an average VI for background components. Finally, the adjusted VI was regressed against leaf area index (LAI) measured in the field using a LiCor LAI-2000 and spatially referenced through differential GPS. Results using the adjusted VI values were compared with unadjusted VI, as were the results obtained using the normalized difference vegetation index (NDVI) and the simple ratio (SR). The results indicate that adjusted NDVI can be used to predict the LAI of trees within mixed pixels much better than does unadjusted NDVI, which overestimates the LAI because it includes non-arboreal vegetation. NDVI provided better results than did SR. A number of issues affect the accuracy of LAI estimates in practice: the accuracy of estimates of endmember proportions obtained through unmixing, the adequacy of the endmember average NDVI for removing the effects of non-arboreal NDVI contributions, the non-synchronous nature of the satellite flight and field work, and the accuracy of field estimates of LAI made using the LAI-2000. 2