Quantification of water and exposed lined areas of coal-bed methane water ponds using regular true-color images by developing a novel uniformness based multi-component algorithm

• Published in: Journal of hydrology (Amsterdam) 2019-05, Vol.572, p.645-658
• Main Authors: Zhou, Xiaobing, Zhou, Zhaoming, Miao, Jiaqing, Apple, Martha E., Spangler, Lee
• Format: Article
• Language: English
• Subjects: Chrominance-based K-mean clustering; Denoising; Feature extraction; Image segmentation; L a b color image; True-color image
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2019.03.015

•A multi-component algorithm for water body extraction from true-color images.•The algorithm includes denoising, color space transformation, K-mean clustering and connected component analysis, and extraction of water body.•The algorithm was successfully used for extraction of water and exposed line area of coal-bed methane ponds. The Powder River Basin in Montana and Wyoming supplies about 40 percent of coal in the United States. In recent decades, coalbed methane (CBM) extracted from coal beds has become an important source of energy in the United States and other countries. CBM water is the produced water brought to the surface as a byproduct of gas extraction. The use of CBM water from the Powder River Basin for irrigation is limited by its relatively high sodium adsorption ratio so the water has to be maintained in lined retention ponds. The liners can be detached from the pond bed, breached through normal weathering or from extreme weather events, and can wear out with time. Water level changes due to constant and dynamic evaporation and precipitation of rainfalls and snowfalls and refilling by gas companies. The objectives of this study are: (1) to develop and validate a novel uniformness based multi-component algorithm to quantify the areas of water and exposed (not submerged in water) liners of the retention ponds from regular red-green-blue (RGB) images, and (2) to quantify the areas of water and liners of the CBM retention ponds in the Powder River Basin using high-resolution true-color images such as free Google Earth imagery. The multi-component algorithm is composed of a sequence of processes: georeferencing, denoising, RGB to L * a * b* color space transformation, K-mean clustering analysis, intelligent identification of connected component in each class, extraction of water body based on maximum uniformness, extraction of exposed lined zones based on uniformness or least root-mean-square value of RGB, and eventually calculation of water area and exposed line area. Application of the multi-component algorithm to the CBM retention ponds in the Powder River Basin using high-resolution true-color images shows that the algorithm can extract accurately water area and the exposed area of liners from these images of high spatial resolution (0.20 ∼ 0.34 m), providing a useful tool to use regular RGB images as a resource for monitoring reservoirs, lakes or artificial small ponds.