Grassland is the largest terrestrial ecosystem in China. The accurate assessment of grassland vegetation carbon stocks plays an important role in maintaining national ecological security and guiding the development of animal husbandry. Vegetation biomass and grassland area are the key parameters to the estimation of grassland carbon storage. With the development of remote sensing technology, the estimation accuracy and efficiency of grassland biomass and area have been significantly improved. Various remote sensing estimation models of grassland biomass and land cover products have been developed, with high accuracy of estimation results in flat areas. However, in complex terrain area, due to the high heterogeneity in ecosystem structure and functions caused by the geometry and consequently water and heat distribution, it is difficult to accurately make out vegetation types as well as the biomass and area of each type. Therefore, it is difficult to adopt remote sensing methods suitable for flat land directly to estimate grassland biomass and area in complex terrain, affecting the accuracy of grassland vegetation carbon storage determination. This paper reviews remote sensing methods and the key parameters in estimation of vegetation carbon storage of grasslands in complex terrain. It points out that LAI (Leaf Area Index) inversion is moderately affected by topographic effect at slope above 30° and introduction of topographic parameters obviously promotes the accuracy of NPP (Net Primary Productivity) estimation as compared to that with NDVI (Normalized Difference Vegetation Index) alone. In process models based on remote sensing, topography affected the determination of key parameters including optimal temperature of photosynthesis, soil water content, grazing intensity, vegetation type and phenology, and carbon allocation. Ignoring rolling topography underestimates grassland area especially with slope above 30°. With a thorough analyses of the fundamental issues, including "topographic effect removal and scale selection of remote sensing image", "selection of vegetation indexes and topographic parameters", "calibration of vegetation growth parameters in process model", "estimation of grassland area", "matching of meteorological data with microclimate in complex terrain", the paper proposes corresponding solutions. Among the diverse vegetation indexes, EVI (Enhanced Vegetation Index) is more sensitive to topographic effect, which is better used in smooth surface with high plant coverage. NDVI is recommended for terrains with slope less than 25° and moderate plant coverage. However, all the vegetation indexes should be corrected in terms of topographic effect in rough terrains. For topographic data, TWI (topographic wetness index) or indexes of terrain complexity is needed to characterize rough terrain. For climate data, it is recommended to combine fine DEM and re-analysis of climate data to fit micro-climate. The paper emphasizes the importance of characteristic length scale of remote sensing image and suggests it is much larger than the mean distance among the ridges in rough terrains. To dampen topographic effect, C correction is proposed to be a simple and effective method that is applicable to estimation of vegetation carbon storage in grasslands on complex terrains.