Des chercheurs de l’université de Nanjing – Chine – (University of Information Science & Technology) ont récemment publié un article qui fait le point sur la détection de l’oïdium du blé tendre par télédétection. Ces travaux restent préliminaires puisqu’ils se concentrent sur la signature spectrale des feuilles contaminées mesurées au plus près de la végétation (proxy-détection). Ces travaux mettent toutefois en évidence l’intérêt de deux indices pour détecter avec une relativement bonne précision (R² = 0.84) le feuilles de blé infestées. Ces chercheurs mettent en évidence l’intérêt de quatre bandes spectrales pour définir des indices permettant la détection de la maladie :   570–590 nm,  536–566 nm, 568–592 et 528–570 nm. Remarquons toutefois que l’essai ne considère que des feuilles de blé saines ou contaminées exclusivement avec l’oïdium. Il est donc difficile de dire, pour le moment si ces bandes spectrales sont spécifiques ou non de l’oïdium.
Résumé de l’article :  In this study, we investigated the possibility of using ground-based remote sensing technology to estimate powdery mildew disease severity in winter wheat. Using artificially inoculated fields, potted plants, and disease nursery tests, we measured the powdery mildew canopy spectra of varieties of wheat at different levels of incidence and growth stages to investigate the disease severity. The results showed that the powdery mildew sensitive bands were between 580 and 710 nm. The best two-band vegetation index that correlated with wheat powdery mildew between 400 and 1000 nm wavelength were the normalized spectrum 570–590 and 536–566 nm bands for the ratio index, and 568–592 and 528–570 nm for the normalized difference index. The coefficients of determination (R²) for both were almost the same. The optimum dual-green vegetation index was constructed based on a calculation of the ratio and normalized difference between the normalized spectrum within the two green bands. The coefficients of determination (R²) of DGSR (584, 550) (dual-green simple ratio) and DGND (584, 550) (dual-green normalized difference) were both 0.845. The inverse models of disease severity performed well in the test process at the canopy scale, and indicated that, compared with the traditional vegetation indices of Lwidth, mND705, ND (SDr, SDb), SIPI, and GNDVI, the novel dual-green indices greatly improved the remote sensing detection of wheat powdery mildew disease. Following these results, combined disease severity and canopy spectra were shown to be of enormous value when applied to the accurate monitoring, prevention, and control of crop diseases.
Références : Feng, W., Shen, W., He, L., Duan, J., Guo, B., Li, Y., … & Guo, T. Improved remote sensing detection of wheat powdery mildew using dual-green vegetation indices. Precision Agriculture, 1-20.