Skip to main content
ARS Home » Plains Area » College Station, Texas » Southern Plains Agricultural Research Center » Aerial Application Technology Research » Research » Publications at this Location » Publication #314825

Title: Feasibility study on Huanglongbing (citrus greening) detection based on WorldView-2 satellite imagery

item LI, XIUHUA - University Of Florida
item LEE, WON SUK - University Of Florida
item LI, MINZAN - China Agricultural University
item EHSANI, REZA - University Of Florida
item MISHRA, ASHISH - University Of Florida
item Yang, Chenghai

Submitted to: Biosystems Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/6/2015
Publication Date: 5/10/2015
Citation: Li, X., Lee, W., Li, M., Ehsani, R., Mishra, A., Yang, C. 2015. Feasibility study on Huanglongbing (citrus greening) detection based on WorldView-2 satellite imagery. Biosystems Engineering. 132:28-38.

Interpretive Summary: Huanglongbing (HLB), also known as citrus greening, is a devastating worldwide citrus disease without any known cure. To explore a fast way of monitoring HLB in large citrus groves, multispectral satellite images were evaluated for HLB detection. Image analysis and ground verification showed that the satellite image was able to distinguish HLB-infected trees from healthy trees with an average accuracy of 65%, very close to that based on airborne image (70%) in a previous study. These results indicate that high resolution satellite imagery can be a useful tool for monitoring citrus greening over large geographic areas.

Technical Abstract: Huanglongbing (HLB) is a devastating citrus disease worldwide, without any known cure. Since this disease shows visible symptoms on newly developed canopies, some remote sensing methods have been used as a tool for detection. In order to explore a fast way to monitor HLB in large citrus groves, a satellite multispectral (MS) image with a 2-m resolution acquired by the WorldView-2 was evaluated for HLB detection in this study. Ground truthing was conducted and two spectral libraries were constructed. Library 1 was based on RTK GPS locations of infected trees, and Library 2 involved manual screening according to the ground spectral features. In both libraries, HLB and healthy classes showed differences on either spectral signatures or vegetation indices. Several supervised classifications of the satellite image were performed. The validation accuracies were 50-77% with an average of about 65%, very close to that for the airborne MS image (70%) in a previous study. Libraries 1 and 2 yielded very similar results in the validation set, and Library 2 didn’t perform any better even though it yielded better accuracy in the training set. The main reason was that the lower spatial resolution of the satellite image had mixed up the spectral features and made those characteristics harder to explore. Simpler methods such as parallelepiped and minimum distance showed higher accuracies of over 70%. Overall, the average accuracies of the satellite MS image were very close to those of the airborne MS image, indicating that it has the potential for monitoring citrus greening.