Land cover plays a significant role in the earth system science, which reflects the influence of human activities and environmental changes. In China, many land use/cover maps can be used in recent years. How to combine these date effectively to better produce land cover map is a key question. This paper presents a method to map land cover classification using Dempster-Shafer evidential reasoning (DS). DS theory is a non-parametric classifier and can handle data from any number of sources at any scale of measurement and has an explicit mechanism for handling information uncertainty through the use of the concept of ignorance. The 1km land use data of China in 2000 from China landuse database, the 1∶1000000 vegetation map, the 1km glacier map from 1∶100000 glacier map, the 1∶1000000 wetland map and the MODIS land cover classification product (MOD12Q1) as multi-source of evidence to support each IGBP land cover class. These evidences are then combined using Dempster′s Rule of combination. Finally, a new land cover map with higher accuracy is generated by a decision rule based on maximum belief function and could be representative of the Chinese land-cover status in 2000. A simple compare validation is taken and result shows that the accuracy of each land cover class of new land cover map has be strongly promoted relative to IGBP DISCover and MOD12Q1 land cover map especially for the accuracy of the croplands, urban, glacier, wetland and water bodies class. The evidence theory is can be used for fusing multi classification information. The fusion result is sensitive for the basic probability assignment of input data. The results will be validated further and the basic probability for multi-source input data will be assigned more scientific.