Fine Monitoring of Wetlands at Provincial Large-Scale Using Object-Based Technique and Medium-Resolution Image

Abstract

Due to extreme similarity of wetland spectra, a significant uncertainty lies in the accuracy of the traditional pixel-based classification, which is a bottleneck in the extraction of wetland information. Object-based image analysis (OBIA) has brought opportunity for fine monitoring of wetland information. However, previous studies on monitoring wetlands have focused mainly on exploratory experiments involving high-resolution images for small areas. The application of OBIA in various medium-resolution images for large areas needs further verification. Here, Landsat and China's HJ-CCD images, a new OBIA mixed binary decision tree, tasseled cap transformation, and field samples were used to refine monitoring of changes in wetlands in Hubei Province. The results showed that while the overall accuracy and Kappa coefficient of the extracted wetland information for 2000 were, respectively, 88.98% and 0.87, the overall accuracy and Kappa coefficient of the detected change were 94.75% and 89.41. This indicated that OBIA performed well with medium-resolution HJ-CCD and Landsat images in monitoring changes in wetlands at provincial scale. The area of wetlands in Hubei increased by 171.03 km2 during 2000–2010. Lakes and reservoirs/ponds increased the most in the province, with respective contributions to the total wetland area of 40.21% (108.97 km2) and 59.17% (160.37 km2). At administrative unit scale, Shiyan Prefecture (157.53 km2) and Fangxian County (317.33 km2) contributed the most to the increase of wetland area. The main reason for the increase in wetland area in Hubei in 2000–2010 was the implementation of major ecological projects during that decade.