Mapping of powdery mildew using multi-spectral HJ-CCD image in Beijing suburban area

Powdery mildew is one of the most serious diseases, which has a significant impact on the production of winter wheat. As an effective alternative to traditional sampling methods, remote sensing can be a useful tool in disease detection. This study examines the potential of a moderate resolution multispectral satellite image in disease monitoring at regional scale. At the suburban area around Beijing, a large size ground survey sample (n = 90) and the corresponding HJ-CCD image were acquired at the grain filling stage of winter wheat. A number of spectral features were found to be sensitive to powdery mildew through an independent t-test. Based on these spectral features, classification models were established using both spectral information divergence (SID) and spectral angle mapper (SAM), respectively. The results showed that the overall accuracies of disease identification and severity estimation were moderate. The estimation of normal and seriously infected samples yielded higher accuracies than slightly infected samples. The single phase HJ-CCD can only be used for locating the infected areas of powdery mildew, whereas is unable to discriminate the severity levels of disease. The presence of several stressors and disturbances other than disease is a possible reason of the unsatisfactory performance of disease monitoring models. Therefore, the integration of multi-phase onboard data and some relevant ancillary data is necessary to improve the accuracy and reliability of disease monitoring at regional scale.