Skip to Main Content
A top-down model inversion method of estimating leaf reflectance from hyperspectral remote sensing measurements has been tested with an empirical approach used to estimate chlorophyll content. Leaf reflectance is obtained by inverting a geometric-optical model, 5-Scale, validated using hyperspectral AVIRIS data. The shaded scene fractions and the M factor, which includes both the multiple scattering effect and the shaded components, are computed for inverting canopy reflectance into leaf reflectance. The inversion is based on the look-up tables (LUTs) approach. The simulated leaf reflectance values are combined in hyperspectral indices for leaf chlorophyll retrieval and compared against the measured leaf chlorophyll content in the Greater Victoria Watershed District (GVWD), British Columbia (BC). The results demonstrate that the modeled canopy reflectance and AVIRS data are in good agreement for all locations. The regressions of the modified simple ratio [(R728 - R434)/(R720 - R434)] and modified normalized difference index [(R728 - R720)/(R728 + R720 -2R434)] against chlorophyll content exhibit the best fit using second-order polynomial functions with the root-mean-square errors (RMSE) of 4.434 and 4.247, and coefficients of determination of 0.588 and 0.588, respectively. Larger RMSE are observed when the direct canopy-level retrieval, using canopy-level generated vegetation indices, is considered, suggesting the importance of the proposed canopy-to-level reflectance inversion step in chlorophyll retrieval based on hyperspectral vegetation indices. This approach allows for estimation of leaf level information in the absence of leaf spectra field measurements, and simplifies further applications of hyperspectral imagery at the regional scale.