Accuracy assessment of four global land cover datasets in China

This study aims to examine the suitability of four global land cover datasets (UMD, IGBP-DISCover, MODIS and GLC2000) for their accuracies in mapping and monitoring cropland across China. For that, four global land cover products were firstly compared with the national land cover dataset 2000 (NLCD-2000) at provincial, regional and national scales to evaluate the accuracies of estimation of aggregated cropland area in China. This was followed by a spatial comparison to assess their accuracies in estimating the spatial distribution of cropland across China. The results showed that there were varying levels of apparent discrepancies in estimating China’s cropland among these four global datasets, and that both aggregated areas and spatial agreement between them varied from region to region. MODIS and GLC2000 datasets had a relatively higher accuracy in depicting China’s cropland than UMD and IGBP-DISCover datasets. The coarse spatial resolution and per pixel classification approach, as well as landscape heterogeneity, are the main reasons for large discrepancies between these four global land cover datasets and NLCD-2000 dataset.