Home | About us | Author | Editor | Reader | Articles Indexed by EI | Network Pre-printed Edition | Top 100 | Agr. Eng. for Sustainability | 中文
Chen Yicong,Shao Hua,Li Yang.Consistency analysis and accuracy assessment of multi-source land cover products in the Yangtze River Delta[J].Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE) ,2021,37(6):142-150
Consistency analysis and accuracy assessment of multi-source land cover products in the Yangtze River Delta
Received:December 24, 2020  Revised:March 17, 2021
Foundation item:国家自然科学基金项目(41601449,41501431);江苏省高校自然科学研究面上项目资助(15KJB420001,16KJD420002)
Author NameAffiliation
Chen Yicong 1.School of Geomatics Science and Technology, Nanjing Tech University, Nanjing 211816, China 
Shao Hua 1.School of Geomatics Science and Technology, Nanjing Tech University, Nanjing 211816, China 
Li Yang 2 Key Laboratory of Virtual Geographic Environment, Nanjing Normal University, Ministry of Education, Nanjing 210023, China
3.Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China 
Hits: 109
Download times: 28
Abstract: Land cover change is one of the most important data sources for global environmental issues, such as the impact assessment of human activities on ecosystems. Since a great number of land cover datasets have been produced at the global and national scale in recent years, it is necessary to scientifically evaluate the reliability of these datasets in specific areas and the consistency between multi-source datasets in the selection of products. There are great changes in land cover, leading to high demand for the land cover data in the Yangtze River Delta, which is one of the most active and open regions in economic development with the strongest innovation capabilities in China. Taking the Yangtze River Delta region as the study area, this study aims to assess the relative accuracy of six land cover datasets (CCI_LC, FROM-GLC, GLC_FCS30, GLCNMO, GlobeLand30, and CGLS_LC) using the accuracy validation and consistency analysis. Five ways were selected to verify the accuracy, including the confusion matrix, overall accuracy, Kappa coefficient, user and producer accuracy. The correlation of land cover types in the study area was analyzed to calculate the proportions of various types in land cover products and the correlation coefficients. Three 30m-resolution land cover products (FROM-GLC, GLC_FCS30, and GlobeLand30) were also compared via the consistency distribution of classification, overall consistency, quantity and allocation disagreement between them. The results show that the overall accuracy of six products in the Yangtze River Delta region were 76.89%, 78.4%, 84.7%, 74.3%, 80.6%, and 85.4%, respectively, indicating that all of land cover products behaved fine accuracies. The correlation coefficients of land cover types were greater than 0.9 in six products. The best correlation was achieved in the GlobeLand30 with the most reliable estimation on the area of land cover types. Three 30m-resolution land cover products had 65.51% similar classified pixels, while, the overall consistencies between the products were 72.23%, 77.99%, and 76.41%, respectively. A fine data support was obtained from three land cover products, although with a little pixel confusion. In addition, the greatest stability was achieved in GLC_FCS30 with high accuracy of land cover types, particularly on the second-level categories for a more detailed description of land cover. High accuracy of classification was obtained in six land cover products for the cropland, woodland, water body, and impervious areas that were widely distributed in most of the research area. Nevertheless, there was a low accuracy of classification for the wetland, bare land, shrub and grassland. There was less influence of topographical fluctuations on the accuracy of classification in topography and geomorphology. Specifically, only a few unstable topographical classifications occurred in the hilly areas of Zhejiang Province. Correspondingly, the economic level dominated the accuracy of classification in urban development. The land cover products also considered the needs of different users. The best performance of CGLS_LC was achieved in cropland, and the similar excellent performance of GLC_FCS30 and CGLS_LC was found in woodland and impervious areas, while FROM was the most suitable for the water body.
KeyWord:land use  remote sensing  land cover data  accuracy analysis  consistency analysis  category confusion
View Full Text   Download reader

Sponsored by:Chinese Society of Agricultural Engineering Address:No. 41 Maizidian Street, Chaoyang District, Beijing

Tel:010-59197078, 59197077, 59197076, 59197079 Fax:010-59197077 Postal code:100125 Email:tcsae@tcsae.org
This system is designed by Tiller, Beijing Science and Technology Development Co., Ltd.