Abstract: Studying spatial-temporal evolution of land use and ecosystem services value (ESV) is of great significance to the optimization of land use structure and sustainable development of ecological environment in dry valleys. Based on high-precision land use remote sensing interpretation data, spatial statistical analysis and Geodetector were used to study the spatial-temporal evolution and differentiation characteristics of land use and ESV of Panxi area (typical dry valleys) and its inner regions from 2000 to 2020. The results showed that forest land was dominant in Panxi area and its internal divisions, and water area and construction land increased rapidly from 2000 to 2020. Except for the southern canyon industrial and mining area, ESV of Panxi area and its internal divisions showed an increasing trend. Contribution rate of forest land to ESV of Panxi area was the largest, but obstructive effect of construction land on ESV growth was gradually increasing. ESV generally showed high in the northwest and central and low in the northeast and south. The loss area was concentrated in the urban construction and forest degradation area, while the increasing area was concentrated in the construction of reservoirs, hydropower stations and farmland conversation area. The high-value area gradually expanded while the low-value area gradually decreased. During the study period, Moran's I of ESV and its changes were more than 0, and showed a significant spatial autocorrelation (P < 0.001), but the spatial aggregation of ESV decreased slowly. Human activities had a significant impact on the spatial differentiation of ESV in Panxi area and its internal divisions, and Human Active Index was over 48.45%.