Objective  This paper aims to explore the spatial and temporal evolution of net primary productivity (NPP) of vegetation in the Loess Plateau over many years and the combined impact of human activities and natural factors in the same period on it, so as to provide reference for the local ecological restoration planning and implementation.

  Method  CASA model was used to calculate and analyze the NPP and its spatial and temporal distribution pattern of vegetation in the Loess Plateau from 2001 to 2019, and the driving factors and mechanisms of vegetation NPP were analyzed based on Geodetector.

  Result  (1) The vegetation NPP in the Loess Plateau showed an overall significant increasing trend from 2001 to 2019, with an average annual increase rate of 5.59 g/(m²·year). The areas with significant increase in NPP were mainly distributed in the central gully and hilly gully areas of the Loess Plateau. The analysis of the spatial center of gravity of NPP based on the center of gravity model showed that the migration of NPP center of gravity in the Loess Plateau exhibited a periodic change feature, with the average NPP increment and growth rate in the southern part of the center of gravity being higher than in the northern part for most years. The significant NPP increased was detected mainly in the central gully area and hilly gully area of the Loess Plateau. Among different land use types, forest land had the highest mean NPP. Land use transfer was mainly transformed between cultivated land and grassland, accounting for 75% of the total change area, leading to the highest rate of linear changes in NPP. (2) The results of geographic detectors showed that annual precipitation and dryness index were the dominant natural factors affecting vegetation NPP in the Loess Plateau, followed by land use type, annual average temperature and slope. The results of the interaction detector showed that the interaction of factors was mainly bi-factor enhancement or nonlinear enhancement, and the factors that had no significant impact on vegetation NPP had a significant impact on NPP when interacting with other factors. The optimal range of vegetation NPP identified by the risk detector varied among different land use types. The optimum annual precipitation range for NPP of most land use types was 500−1000 mm. Except for unused land, the suitable temperature range for NPP of other land use types was between 10 and 14 ℃. The suitable altitude ranged for NPP of cultivated land and other land use types were 19.62−548.43 m and 1000−2500 m, respectively. The suitable slope range for NPP of the woodland was relatively spanned, and the suitable slope aspects varied among different land types.

  Conclusion  The results of this study show that vegetation restoration on the Loess Plateau from 2001 to 2019 contributes significantly to ecosystem NPP. The interactions between environmental factors enhance the influence of single factors on vegetation NPP. Moreover, the optimal accumulation ranges of NPP of different environmental factors are different for varied land use types. Our study provides a theoretical basis for the vegetation restoration and management practices in this area.