Identification of cotton fields using Sentinel-2 satellite imagery for boll weevil eradication

Authors: Yang, Chenghai; Suh, Charles

Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: 1/22/2020
Publication Date: 5/7/2020
Citation: Yang, C., Suh, C.P. 2020. Identification of cotton fields using Sentinel-2 satellite imagery for boll weevil eradication. National Cotton Council Beltwide Cotton Conference. p. 128-133.

Interpretive Summary: Early identification of cotton fields is important for advancing the boll weevil eradication program in Texas. Our previous work demonstrated that high-resolution airborne imagery was effective for this purpose, but large numbers of airborne images are needed to cover large geographic regions. As 10-m Sentinel-2 satellite imagery is available at no cost and has large area coverage, this study evaluated the feasibility of this type of imagery for identifying cotton fields during the 2019 growing season. Preliminary results showed that Sentinel-2 imagery in conjunction with appropriate classification techniques was feasible for distinguishing cotton from other crops. The first year of this multi-year study provides useful information for future evaluations and practical applications of Sentinel-2 and other satellite imagery to identify cotton fields over large geographic areas at relatively early growth stages.

Technical Abstract: Early identification of cotton fields is important for advancing the boll weevil eradication program in Texas. Remote sensing has long been used for crop identification, but limited work has been reported on identification of cotton fields when cotton plants are small. Our previous work demonstrated that high-resolution airborne imagery was effective for this purpose, but large numbers of images taken along multiple flight lines are needed to cover large geographic regions. As 10-m Sentinel-2 satellite imagery is available at no cost and has large area coverage, this type of imagery was evaluated for identifying cotton fields before cotton plants start to bloom in this study. Three cloud-free scenes acquired on June 11, July 11, and August 15, 2019 were selected to identify cotton fields over a 10 km by 11 km cropping area. The images were classified into different crops and cover types using multiple supervised classification techniques. Preliminary results showed that Sentinel-2 imagery in conjunction with the maximum likelihood classifier was feasible for distinguishing cotton from other crops. However, the excessive rainfall in April and May delayed the planting of some cotton fields and the wet areas within crop fields affected the classification accuracy for the June 11 and July 11 scenes. Nevertheless, the methodologies presented in this study provide boll weevil eradication program managers with a tool to identify cotton fields over large geographic areas at relatively early growth stages.