ESTIMATION OF LEGUME CONTENT IN GRASS-LEGUME MIXTURES FROM HYPERSPECTRAL MEASURMENT

摘要

Legume content in grass-legume mixtures is a key parameter for deciding the forage quality and the amount of fertilizer application to the pasture due to nitrogen (N) fixation. This study investigated the ability of field hyperspectral radiometer (400-2350 nm) with genetic algorithm partial least squares (GA-PLS) regression for estimating legume content in a mixed sown pasture of Hokkaido, Japan. Canopy reflectance data and plant samples were obtained from 50 selected sites in two seasons (n = 100); spring (May) and summer (July) 2007. The predictive accuracy of GA-PLS was compared with that of multiple linear regression (MLR), and standard full-spectrum PLS (FS-PLS) for spring, summer, and combined datasets. The predictive ability of the model was assessed by the coefficient of determination (R~2) and the root mean squared error of cross validation (RMSECV). Overall, the highest R~2 and the lowest RMSECV were obtained in the GA-PLS models for all datasets (R~2 = 0.62-0.86, RMSECV = 4.10-7.20%). In MLR, selected wavebands in the final models were blue (400-456 nm) and red bands (659-670 nm) in visible wavelength, red-edge region (704-724 nm), near infrared regions (813, 937, and 1121 nm), and shortwave infrared regions (2303-2344 nm) that are mainly linked to known biochemical components such as chlorophyll, nitrogen, lignin and cellulose. These results suggest that legume content in grass-legume mixtures can be predicted from field hyperspectral measurements, and that the predictive accuracy of the model can be improved by waveband selection using the GA-PLS.