Monitoring leaf pigment status with hyperspectral remote sensing in wheat

Leaf pigment status within a canopy is a key index for evaluating photosynthetic efficiency and nutritional stress in crop plants. Non-destructive and quick assessment of leaf pigment status is needed for growth diagnosis, yield prediction, and nitrogen (N) management in crop production. The objectives of this study were to analyse quantitative relationships of leaf pigment concentration on a dry weight basis and leaf pigment density per unit soil area to ground-based canopy hyperspectral reflectance and derivative parameters, and to establish estimation models for real-time monitoring of leaf pigment status with key hyperspectral bands and indices in wheat (Triticum aestivum L.). Two field experiments were conducted with different N application rates and wheat cultivars across two growing seasons, and time-course measurements were taken on canopy hyperspectral reflectance over 350-2500 nm, leaf pigment concentrations, and leaf dry weights under the various treatments at different growth stages. The results showed that different pigment concentrations and densities of chlorophyll a (Chla), chlorophyll b (Chlb), total chlorophyll (Chla+b), and carotenoids (Car) in two cultivars, Ningmai 9 (low grain protein concentration) and Yumai 34 (high grain protein concentration), tended to increase with increasing N rates, and differed with genotypes and growth stages. The analyses on the relationships between vegetation indices and leaf pigment concentrations and densities indicated that the pigment concentrations and pigment densities, respectively, were highly correlated with eight spectral parameters selected. The leaf chlorophyll concentrations were highly correlated with red edge position, with highest coefficients of determination (R²) for REPLE, while R² between Car and spectral indices decreased. The chlorophyll densities were highly correlated with VOG2, VOG3, RVI(810,560), Dr/Db, and SDr/SDb, but the correlation was also reduced for carotenoids. Testing of the monitoring equations with independent datasets indicated that the red edge position was the best hyperspectral parameter to estimate leaf pigment concentrations, with no significant difference between REPLE and REPIG for Chla, Chla+b, and Car, although better performance with REPIG than with REPIE for estimation of Chlb. The VOG2, VOG3, Dr/Db, and SDr/SDb were the best hyperspectral parameters to estimate leaf pigment densities, but with lower estimation accuracy for Chlb and lower estimation precision for Car. The overall results suggested that the pigment concentrations and densities in wheat leaves, especially for Chla and Chla+b, could be reliably estimated with the hyperspectral parameters established in this study.