Remote sensing of foliar chemistry of inundated rice with imaging spectrometry

The applicability of imaging spectrometry to the measurement of foliar chemistry of herbaceous wetland vegetation was assessed for cultivated rice. Multiple samples were obtained from five California rice fields for analyses of nitrogen and lignin concentrations, and Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) images were obtained in August 1992. Multiple linear regression (MLR) was used to develop calibration equations for nitrogen and lignin based on AVIRIS reflectance (R) spectra transformed as the first difference of absorbance [log(1/R)]. Transformed AVIRIS absorbances were not as well correlated to lignin (rsu2 = 0.44) as to nitrogen (r2 = 0.74); the relative standard error of calibration (SEC) for nitrogen was 11%. To further test the predictive ability of this technique with respect to nitrogen concentration, four new calibration equations were derived based on each possible set of three fields. With each new subset of the data, MLR selected different wavelengths for the nitrogen calibration equation. Coefficients of multiple determination (r2) ranged from 0.69 to 0.85, with relative standard SECs ranging from 8.4% to 11.4%. Each of these equations was then used to predict the nitrogen concentration of samples in the omitted field, resulting in relative standard errors of prediction (SEP) ranging up to 37%. Results from this study suggest that calibration equations derived from multiple linear regression of AVIRIS spectra can be used to predict the nitrogen concentration of rice in the field. It was not possible to derive a general equation for the detection of nitrogen concentration with AVIRIS because the calibration equations developed with MLR were based on a different set of wavelengths for each subset of the data.