Grassland degradation in the Qinghai–Tibet Plateau (QTP), has attracted considerable concern because of its negative influence on the development of the local economy and the ecological security of China. Climate and human activities are considered as the main driving forces of grassland degradation. However, distinguishing their respective contributions to grassland degradation is a challenge. This study used the Carnegie–Ames–Stanford Approach model, which coupling remote sensing (e.g. NDVI, LAI, near and mid-infrared bands) and meteorological data (precipitation, temperature and radiation), was adopted to simulate the actual and potential NPP in the QTP from 2001 to 2013. The difference between potential NPP and actual NPP was used to represent the influence of human activities. Results showed that nearly 38.8% of the total grassland area underwent degradation, whereas 61.2% experienced restoration. Furthermore, 56.7% of the degraded grassland areas were influenced by climate, and 19.9% were affected by human activities. The restored areas induced by human activities, climate variation, and the combination of the two factors accounted for 28.6%, 12.8% and 19.9% with an increases in NPP of 5923.4, 3188.1 and 5959.2 GgC, respectively. Therefore, climate was the principal driving force of grassland degradation, whereas human activities were the dominant factor in grassland restoration. Climate and human activities, as the potential driving force in grassland NPP variations, should be fully understood by a long term monitoring and the main causes exploring in its dynamics. In addition, the uncertainty of the driving forces should be clarifying immediately in the future, and provide scientific basis for policies and plans making in grassland management.