Based on the water quality monitoring data from 26 monitoring sections of Poyang Lake urban agglomeration in 2021, this study extracted landscape pattern metrics under six spatial scales, including riparian buffer zones ranging from 200 m to 2000 m and corresponding sub-watersheds, using the GlobeLand30 dataset. Redundancy analysis and other statistical approaches were adopted to explore multi-scale correlations between landscape patterns and surface water quality. The results showed that the water quality at all monitoring sections met the good grade or above. Water bodies and cultivated land constituted the dominant land use types across riparian buffer zones, while forestland and cultivated land occupied the largest proportions at sub-watershed scale. Construction land,cultivated land and wetland exhibited source-landscape characteristics, whereas forestland exerted positive effects on water quality conservation. In most scenarios, the largest patch index and contagion index had the strongest impact on water quality variations. Human-induced landscape transformation and resultant landscape fragmentation potentially elevated the risk of pollutant transport into rivers. Landscape pattern within 1000 m riparian buffer zone presented the strongest correlation with surface water quality. Accordingly, this study proposes optimizing landscape pattern allocation at 1000 m riparian scale, strengthening pollutant discharge regulation and sewage treatment, and prioritizing the ecological functions of dominant landscape patches. Keywords: landscape pattern; spatial scales; surface water quality; redundancy analysis; Poyang Lake urban agglomeration 〖FL