Land use patterns change the potential supply of regional ecosystem services, which in turn has an important impact on regional ecological environment and sustainable economic development. Taking the Guangxi section of the Pearl River-West River Economic Belt in China as an example, the Land Use Change (LUC) and Ecosystem Services Value (ESV) in 1990, 2005 and 2018 were calculated. Then the spatial and temporal characteristics of ESV under the influence of LUC were discussed by using the ESV profit and loss flow table and the bivariate spatial autocorrelation. The results showed that(1) The total values of ESV in the study area have not changed significantly in 28 years, and the ESV decreased in recent years compared with that in 2005. The area of construction land increased the most, the area of cultivated land and forestland decreased first and then increased, and land use type transfers were mainly based on the mutual transfer of forestland and cultivated land. (2) Forestland was the most important ecological land in the study area. Biodiversity maintenance, hydrologic regulation, soil conservation, gas regulation and climate regulation constituted the main body of the ESV in the study area. The high-value areas of ecosystem services value intensity (ESVI) were mainly distributed in the northwest, east and north of Guangxi, while the low-value areas were located in the hills and plain areas in the middle of Guangxi with relatively flat terrain. The ESVI of each district and county showed significantly positively spatial autocorrelation (P<0.05), with a high degree of spatial aggregation. High-high aggregation and low-low aggregation were highly coincident with high-value and low-value regions of ecosystem services value intensity. (3) The mutual transfers between forestland and cultivated land during the study period were the main reasons for the increase and decrease of ESV. The land use degree and ESVI's bivariate spatial autocorrelation map had similar spatial characteristics in the three periods, showing a significantly spatially negative autocorrelation pattern (P<0.05).