Comparison of hyperspectral retrievals with vegetation water indices for leaf and canopy water content

Leaf and canopy water contents provide information for leaf area index, vegetation biomass, and wildfire fuel moisture content. Hyperspectral retrievals of leaf and canopy water content are determined from the relationship of spectral reflectance and the specific absorption coefficient of water over the wavelength range of a water absorption feature. Vegetation water indices such as the Normalized Difference Water Index [NDWI = (R850 - R1240)/(R850 + R1240)] and Normalized Difference Infrared Index [NDII = (R850 - R1650)/(R850 + R1650)] may be calculated from multispectral sensors such as Landsat Thematic Mapper, SPOT HRG, or MODIS. Predicted water contents from hyperspectral data were much greater than measured water contents for both leaves and canopies. Furthermore, simulated spectral reflectances from the PROSPECT and SAIL models also had greater retrieved leaf and canopy water contents compared to the inputs. Used simply as an index correlated to leaf and canopy water contents, hyperspectral retrievals had better predictive capability than NDII or NDWI. Atmospheric correction algorithms estimate canopy water content in order to estimate the amount of water vapor. These results indicate that estimated canopy water contents should have a systematic bias, even though this bias does not affect retrieved surface reflectances from hyperspectral data. Field campaigns in a variety of vegetation functional types are needed to calibrate both hyperspectral retrievals and vegetation water indices.