Hyperspectral estimation of canopy chlorophyll density in winter wheat under stress of powdery mildew

Abstract: The diagnosis of crop health and growth status under various stresses is an important step in precision agriculture. In order to enhance the timeliness in monitoring the growing state and evaluating disease influence in wheat production by using hyperspectral remote sensing, three experiments were conducted with different disease degree treatments across different growth phases, and different cultivars were chosen including four cultivars of Yanzhan 4110, Yumai34, Yumai 49 and Xinong 979 considered as susceptible variety types, and three cultivars of Zhoumai 18, Zhengmai 366 and Aikang58 considered as medium resistant variety types to strengthen wide applicability of monitoring models. The spectrum reflectance and canopy chlorophyll density (CCD) infected with powdery mildew were measured in winter wheat diseased in the nursery and the field during the experiments' periods in 2009-2011. The relationships of spectra reflectance, the first derivative of reflectance, and existing spectral indices to CCD were analyzed respectively, and the monitoring models of the CCD of diseased wheat were established and tested. The results indicated that the canopy chlorophyll content resulted in an obvious declining trend in two different resistant varieties of wheat, and the powdery mildew damage was heavy for the susceptible variety Yanzhan4110 and slighter lower for the medium resistant variety zhoumai18. The correlations were most significant between CCD and spectral reflectance in red wave bands of 600-630 nm and red edge wave bands of 690-718 nm, and the first derivative spectrum in red edge wave bands of 718-756 nm. The models of SDr/SDb and VOG3 among existing spectral indices had the best fitting precision with a determination coefficient (R2) of 0.752 and 0.723 respectively, and the lowest relative errors (RE) between measured and estimated CCD as 18.0% and 18.6%. A new red edge angle index (REAI) selecting optimal bands of 680, 718 and 756 nm was found to be a good indicator for CCD, with fitting R2 of 0.783 for NDAI (α,β) and 0.776 for RAI (α,β), and test RE of 16.8% for NDAI (α,β) and 17.5% for RAI (α,β). The overall results indicated that hyperspectral vegetation indices can be used to reliably estimate CCD of wheat infected with powdery mildew, and CCD models based on a red edge angle index has great practical and application value for monitoring photosynthetic potential productivity and disease influence evaluation on wheat.