Comparison of Different Hyperspectral Vegetation Indices for Estimating Canopy Leaf Nitrogen Accumulation in Rice

Abstract

Variations in water and soil backgrounds can affect canopy spectral reflectance, complicating canopy N status estimation. We created rice (Oryza sativa L.) canopies with varying levels of leaf area index (LAI) and water and soil backgrounds using different rice varieties and a combination of different N rates and planting densities. The quantitative relationships between hyperspectral vegetation indices and leaf nitrogen accumulation (LNA) were analyzed to derive new spectral indices and models for estimating LNA. The sensitive spectral region of LNA significantly differed from that of leaf nitrogen concentration (LNC). All two‐band hyperspectral vegetation indices derived from the systematic combinations of bands in the 400 to 2500 nm range were correlated to canopy LNA. The new, simple vegetation index SR(R770, R752) exhibited the highest correlation with LNA, with R2 0.88 for model calibration and with R2 0.80 for model validation, respectively. The SR(R770, R752) was modified by incorporating a coefficient of soil/water line parameter, θ, yielding the simple vegetation index SR2(R770, R752). This modified index provided slightly better estimates of rice LNA, with a calibration R2 0.90 and a validation R2 0.80. Datasets obtained for different sensor heights before and after canopy closure confirmed the superior performance of SR2(R770, R752). Therefore, the SR(R770, R752) and SR2(R770, R752) can be used to estimate rice LNA. Since SR2(R770, R752) is less dominated by soil background, this index is recommended for estimating LNA in rice under various cultivation conditions.