To explore the impacts of landscape pattern on water quality at multi-spatial scales, water samples were taken from 15 sampling sites in the main stream of the Liuxi River in June 2020 and January 2021, respectively. The water quality indicators such as water temperature, dissolved oxygen, pH, ammonia nitrogen, nitrate, sulfate and chloride were determined. We then extracted the landscape pattern indices at different spatial scales (sub-basins and riparian buffer zones) based on interpretating Landsat remote sensing image data. Methods such as Bioenv analysis, Mantle test, Variance partitioning and Hierarchical partitioning theory were applied to reveal the influence mechanism of landscape pattern on water quality. The results showed that ammonia nitrogen was the main pollutant in the Liuxi River. The impacts of land use structure and spatial pattern characteristics on the water quality change had the spatial scale effects. Water and built-up land were the main contributor to the change of water quality in the 100 m riparian buffer zone and in other spatial scales, respectively. In the sub-basins, the patch density of forest and built-up land were the main characteristics which affected the change of water quality, while the connectivity characteristics of the water and built-up land and the richness and diversity of forest were the key features affecting the change of water quality in the riparian buffer zones. Overall, the interaction between land use and landscape pattern played a leading role in driving water quality change in the Liuxi River, especially in the 1000 m riparian buffer zone, which had the highest contribution to the water quality. Therefore, strengthening the management of land use in the 1000 m riparian buffer zone and avoiding the large area-development of built-up land are of great significant to the protection of water quality in the basin.