Understanding the quantitative relationship between multi-scale landscape pattern and water quality is of important theoretical significance for rational landscape configuration and non-point source pollution control. Using water quality data at nine monitoring sites on the Baihe River in the upper reaches of the Miyun Reservoir in Beijing, a correlation analysis and a multiple stepwise regression were used to determine the effects of the landscape characteristics on the water quality at different riparian buffer zone scales (100, 200, 300, 500, and 1000 m). The results showed that the impact of the landscape pattern, composed of both landscape composition and configuration, on the surface water quality, varied significantly with spatial scales. The landscape characteristics for the 300 m and 100 m buffer zones appeared to have slightly greater effects on the water quality index TN and TP, respectively. The patch density of cultivated land and the aggregation index of grassland were recognized as the dominant indices influencing TN for the 300 m buffer zone. The area proportion of rural residential at the 100 m buffer zone was the dominant index influencing TP. It is very important to optimize the landscape pattern within a 300 m width of a riparian buffer zone. In particular, the reasonable allocation of cultivated land, forest, and grassland, to improve the connectivity and aggregation of agricultural landscapes, and the control of rural residential areas and pollutant discharge along the river bank, will enhance the ecological function of the water quality of the Baihe River in Beijing. This will ensure drinking water safety from the Miyun Reservoir.
Keywords: landscape metrics; multivariate statistical analysis; riparian buffer zones; spatial scale; water quality.