Niu, Z.-G./ Y.-X. Shan, / H.-Y. Zhang 2012: Accuracy assessment of wetland categories from the GlobCover2009 data over China. - Wetland Science10(4): 389-395. [RLL List # 231 / Rec.# 34668] Keywords: Accuracy assessment/ GlobCover2009/ Remote sensing data/ Wetlands in china
Abstract: At present, few specific data sets on wetland distribution exist at a global scale and many researchers have to acquire wetland information from the existed global land cover data products. GlobCover2009, with 300 m spatial resolution, issued by ESA, becomes one of the important choices. However, up to now, there is no evaluation of it in detail made on the precision of the wetland category. In view of this situations, China wetland map in 2008, which is produced through visual interpretation by IRSA (Institute of Remote Sensing Application, Chinese Academy of Sciences), is employed to examine the wetland accuracy of GlobCover2009 from area accuracy, spatial agreement, confusions among different categories and different regions across China. The results highlight a very low precision of wetland from GlobCover2009. The area consistency coefficients between the wetland from the global data set and the reference data is only 46%, overall accuracy is only 32% and Kappa 0.13. Furthermore, the producer's accuracy of marshlands (0.05%) is much lower than the one (53.34%) of water. However, the user's accuracy of water is 90.18%, which highlights that the water from the GlobCover 2009 product could meet some requirements at large scale. The categories where errors occurred frequently mainly include bare areas, closed to open herbaceous vegetation, mosaic vegetation(grassland, savannas or lichens/mosse)/cropland, rainfed croplands, and part of mosaic croplands/vegetation (grassland/shrubland/forest). The most omission error of wetlands occurred in the region of Southwest and Northeast China where the mires widely distributed. No specific definition and classification of wetlands in GlobCover2009 may be responsible for the low precision. In addition, different spatial resolution and image" acquisition date between these two data sets also contribute to it. Overall, the unsatisfactory results highlight more attention should be paid to the application of the global data products, especially in wetland-relevant categories over China. Furthermore, specific global wetland classification and products are urgent to develop in order to meet requirements of the global changes research and the ecological environment research.