Forest ecosystems have a great potential in mitigation of carbon concentration in the atmosphere. Thus, generating its spatially explicit estimates at national, regional and global scale becomes very important. In Southern China, mapping forest carbon is often conducted by combining ground plot data from national forest inventory and remotely sensed images from Landsat and MODIS (Moderate Resolution Imaging Spectroradiometer) with variable spatial resolutions. However, the inconsistency of sample plot sizes with spatial resolutions of images will lead to a great challenge for mapping and accur...