How did the Chinese Loess Plateau turn green from 2001 to 2020? An explanation using satellite data

•Drivers of vegetation greening were examined at the inter-annual and spatial scales.•Human activities were the dominant driver of variation in vegetation during 2001–2020.•Forests following row-belt distribution with low coverage is more suitable for application in drylands. Vegetation growth of the Chinese Loess Plateau (CLP) has improved since the launch of the Grain to Green Programme. This study examined the mechanism of spatial differentiation of the Normalized Difference Vegetation Index (NDVI) and the relative contributions of climate variations (CV) and human activities (HA) to variations in inter-annual NDVI in the CLP for 2001–2020. The response of NDVI to afforestation was examined and the trend of projected future NDVI was identified using the Moderate Resolution Imaging Spectroradiometer and other correlative data. The results showed that: (1) precipitation, sunshine duration, relative humidity, solar radiations, and vegetation type were the dominant factors affecting the spatial distribution of NDVI. All interactive explanatory powers of two factors exceeded those of any individual factor, and the interactions between natural and anthropogenic factors had significant effects on NDVI. (2) Variation in NDVI in 88.77% of the total area was influenced by the combined effect of HA and CV, and HA was the dominant factor with relative a contribution of 68.44%. (3) There was an increasing (2001–2005) followed by a decreasing (2006–2019) pattern of annual area of afforestation in each province. The seven provinces showed extremely significant correlations (p < 0.01) between the cumulative area of afforestation and NDVI. (4) The trend in the change in NDVI exhibited the unsustainability of development of the CLP, with 63.20% of regions at risk of vegetation degradation. Finally, the present study suggests that forests in arid and semi-arid areas following a row-belt distribution with low coverage composed of narrow forest belts (covering an area of 15–25%) and wide natural vegetation restoration (covering an area of 75–85%) may be more appropriate for restoration of the regional ecology.