Optimal visible and near-infrared waveband used in hyperspectral indices to predict crop variables of rice [Oryza sativa]

In order to select suitable wavebands to predict crop variables, canopy reflectance and crop variables was regularly measured over rice canopy during growing season in Java Island, Indonesia. Two crop variables such as leaf area index (LAD and SPAD values, three rice cultivars and four nitrogen levels were involved in this study. Optimal and paired-wavebands for predicting crop variables tested not only in traditional VIs (we used Reflectance Indices, RIs, instead of VIs in this paper to avoid confusion) such as NDRI (normalized difference reflectance index), also tested them in RRI (ratio reflectance index), RDRI (renormalized difference reflectance index) and SARI (soil adjusted reflectance index) with involving all possible waveband combinations to obtain best fitted two-pair waveband related to crop variables. A first derivative reflectance (FDR) spectrum was calculated and analysis from all possible paired-waveband combinations used in RIs was investigated with 6,786 combinations attributed to LAI and SPAD. The R**2 value of paired-wavebands used in RIs for LAI ranged from 0.883 to 0.908, and for SPAD ranged from 0.667 to 0.771. The significant relationships (R**2) between optimal single band with crop variables achieved by red edge region wavebands (735 and 720 nm) of FDR attributed to LAI and SPAD respectively. The highest relationships between LAI and optimal paired-waveband combinations of near-infrared (835 nm) and red edge region (720 nm) obtained by RDRI (R**2=0.908), while for SPAD was red edge regions (715 and 710) attained by SARI (R**2=0.771). Validation of measured and predictive value using FDR implied better accuracy to estimate LAI (R**2=0.859) than using reflectance data (R**2=0.797), meanwhile using either reflectance (R**2=0.709) or FDR (R**2=0.702) data to predict SPAD demonstrated close value. Furthermore, validation using SARI denoted the highest values for predicting LAI (R**2=0.852) and SPAD (R**2=0.658), respectively. These results giving fundamental information on monitoring growth progress and enable to estimate grain production. In addition this information provides recommendation for practical use in studying hyperspectral remote sensing for growth assessment, to support rice farming management for broader assessment using satellite.