Comparison of two satellite imaging platforms for use in land use/land cover classification in agricultural regions

Date
2014-12
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract

ABSTRACT Images used in remote sensing applications may be collected through several different imaging platforms. Satellite based platforms are widely used because of the large ground area covered in each pass. One of the largest limitations to satellite imagery is the spatial resolution. Many spaceborne platforms have contained spatial resolutions that are classified as low or medium resolution, while those with high spatial resolution are very expensive. For this project the mid-spatial resolution Landsat-8 was compared to the high-spatial resolution WorldView-2. The objectives of this research were to 1) compare the capabilities of WorldView-2 to those of Landsat-8 in an agricultural environment and determine whether or not one has an advantage over the other, by performing Land Use/Land Cover classification, and 2) visually inspect the images for advantages in identifying environmentally sensitive sites within an agricultural environment. Images were collected in 2012 and 2013 in the Texas Southern High Plains and Southeast Arkansas. Images from the WorldView-2 satellite were collected in 2012 and Landsat-8 images were collected in 2013. Each location was split into smaller field locations based upon existing fields, natural land formations, and farming practices of the region. The Texas Southern High Plains location contained 113 field locations while the Southeast Arkansas location contained 91 field locations. Ground truth data provided information on ground cover present and aided in defining training locations. Ground truth data was collected through field visits in 2012 and meetings with the producers in 2013. Classification was conducted with ERDAS IMAGINE 2013 using maximum likelihood classifiers with a minimum distance parametric rule and no non-parametric rule. Due to issues associated with the cost of WorldView-2 imagery, image collection was obtained in December for the Texas Southern High Plains location and in September for the Southeast Arkansas location. Landsat-8 images were collected during the growing season for both location in 2013. In the Texas Southern High Plains map accuracy for WorldView-2 was 48.7% while accuracy for Landsat-8 was 57.6%. Southeast Arkansas saw a reversal in satellite accuracies. Map accuracy for WorldView-2 was 87.5%, while accuracy for Landsat-8 was 53.5%. Identification of environmentally sensitive regions was conducted in the Southeast Arkansas location as water bodies were not present in the Texas Southern High Plains location. Classification was able to identify most of the water bodies with WorldView-2, but could only identify water bodies used as training sites with Landsat-8. Buffer strips were identified by both satellites through the classification process. Identification with Landsat-8 was inconsistent and no buffer strips were identified in entirety. WorldView-2 did not achieve complete identification in its entirety but did identify portions of all buffer strips.

Description
Keywords
Satellite, WorldView-2, Landsat-8
Citation