Peer review

Remote sensing has great potential to serve as a useful means in crop disease detection at regional scale. With the emerging of remote sensing data on various spectral settings, it is important to choose appropriate data for disease mapping and detection based on the characteristics of the disease. The present study takes yellow rust in winter wheat as an example. Based on canopy hyperspectral measurements, the simulative multi-spectral data was calculated by spectral response function of ten satellite sensors that were selected on purpose. An independent t-test analysis was conducted to access the disease sensitivity for different bands and sensors. The results showed that the sensitivity to yellow rust varied among different sensors, with green, red and near infrared bands been identified as disease sensitive bands. Moreover, to further assess the potential for onboard data in disease detection, we compared the performance of most suitable multi-spectral vegetation index (MVI)-GNDVI and NDVI based on Quickbird band settings with a classic hyperspectral vegetation index (HVI) and PRI (photochemical reflectance index). The validation results of the linear regression models suggested that although the MVI based model produced lower accuracy (R2 = 0.68 of GNDVI, and R2 = 0.66 of NDVI) than the HVI based model (R2 = 0.79 of PRI), it could still achieve acceptable accuracy in disease detecting. Therefore, the probability to use multi-spectral satellite data for yellow rust monitoring is illustrated in this study.

Text in English