Correlating Leaf Area Index of Ponderosa Pine with Hyperspectral CASI Data

Leaf area index (LAI) estimates collected from a Ponderosa pine stand in Oregon were correlated with the hyperspectral data acquired using a Compact Airborne Spectrographic Imager (CASI). Eight LAI values ranging from 0.87 to 2.72 were measured using an LAI-2000 Plant Canopy Analyzer at the study site. First- and second-order spectral derivatives of reflectance spectra from the CASI data were used to suppress the effects of the soil background on the forest spectral reflectances. A piece-wise multiple regression procedure was then used to explore the relationships between the LAI and the CASI data. This procedure produces multivariate linear equations and their associated goodness-of-fit (GOF) values (coefficients of determination) and standard errors (SE) for LAI estimation. Results show that the spectral derivative technique can increase the correlations between LAI and the derivative spectra of CASI data as compared to those between LAI and the reflectance spectra of CASI data when environmental variability such as background soil and atmospheric effects vary at a lower rate when compared with signal spectra. Therefore, the spectral derivative approach leads to improved accuracies of LAI estimation. For instance, the highest GOF obtained for single-channel LAI prediction is 0.681 with a SE of 0.345. These values have been considerably improved to 0.904 and 0.189, and 0.898 and 0.195 after taking the first- and second-order derivatives, respectively