Comparison of Sentinel-2, Landsat, and airborne imagery for identification of cotton fields for boll weevil eradication

Authors: Yang, Chenghai; Suh, Charles

Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: 1/23/2021
Publication Date: 5/5/2021
Citation: Yang, C., Suh, C.P. 2021. Comparison of Sentinel-2, Landsat, and airborne imagery for identification of cotton fields for boll weevil eradication. National Cotton Council Beltwide Cotton Conference. pp. 234-239. CD-ROM.

Interpretive Summary: Previous studies demonstrated that airborne imagery and 10-m Sentinel-2 satellite scenes are effective for identification of cotton fields, which is important for advancing the boll weevil eradication program in Texas. In this study, both 10-m Sentinel-2 and 30-m Landsat scenes were evaluated and compared with airborne imagery for identifying cotton fields during cotton squaring to blooming stages. Airborne images were acquired over a 10 km by 11 km study area near Snook, Texas in June 2020. One Sentinel-2 scene and one Landsat 7 scene acquired over the study area within a week of the airborne image acquisition were selected for comparison and analysis. Accuracy assessment on the classification maps from the three types of imagery showed that both Sentinel-2 and Landsat imagery had similar performance to airborne imagery for distinguishing cotton from other crops. This result is important because Sentinel-2 and Landsat scenes covers large geographic areas and are freely available to end users. Efforts are underway to evaluate these types of satellite imagery for identification of cotton fields at early growth stages.

Technical Abstract: Our previous studies have demonstrated that both submeter airborne imagery and 10-m Sentinel-2 satellite scenes are effective for identification of cotton fields, which is important for the boll weevil eradication program in Texas. In this study, both 10-m Sentinel-2 and 30-m Landsat images were evaluated and compared with airborne imagery for identifying cotton fields during cotton squaring to blooming stages. Airborne color and near-infrared images were acquired over a 10 km by 11 km study area near Snook, Texas on 17 June 2020. Among the Sentnel-2 and Landsat images, two cloud-free scenes were identified for the study area within a week of the airborne image acquisition, including one Sentinel-2 scene acquired on 10 June and one Landsat 7 scene acquired on 16 June. The airborne images from the study area were mosaicked, and the airborne mosaic and the two satellite images were each classified into different crop types using multiple classification techniques. Accuracy assessment were performed to compare the classification maps with ground surveyed field maps. Preliminary results showed that both Sentinel-2 and Landsat imagery had similar performance to airborne imagery for distinguishing cotton from other crops. Efforts are underway to determine if cotton fields can be detected at earlier growth stages with satellite images collected over the study area in May 2020. The results and methodologies from this study will provide boll weevil eradication program managers with some guidelines to identify cotton fields during early and mid-growth stages over large geographic areas.