Relationship between narrow band normalized deference vegetation index and rice agronomic variables

The objective of this article is to determine spectral bands that are best suited for characterizing rice agronomic variables. The data for this study comes from ground-level hyperspectral reflectance measurements of rice during different stages of the 2002 growing period. Reflectance was measured in discrete narrow bands between 350 and 2500 nm. Observed rice agronomic variables included wet biomass and leaf-area index. Narrow band normalized difference vegetation index (NBNDVI) involving all possible two-band combinations of discrete channels was tested. A rigorous search procedure to identify the best NBNDVI predictors of rice agronomic variables was described. Special narrow-band lambda (lambda(1)) vs. lambda (lambda(2)) plots of R-2 values illustrates the most effective wavelength combinations (lambda(1) and lambda(2)) and band width (Deltalambda(1) and Deltalambda(2)) for predicting rice agronomic variables at different development stages. The best of the NBNDVI models explained 53-83% variability of rice agronomic variables at different development stages. A strong relationship with rice agronomic variables is located in red-edge (700-750 nm), in the longer portion of red (650-700 nm), in moisture-sensitive NIR (950-1000 nm), in the longer portion of the blue band (450-500 nm), in the longer portion of the green (1550-600 nm), in the intermediate portion of short-wave infrared (SWIR) (1600-1700 nm), and in the longer portion of SWIR (2150-2250 nm).