Identification of the control factors affecting water quality variation at multi-spatial scales in a headwater watershed.

Understanding the effect of landscape characteristics on water quality can provide insight into mitigating water quality impairment. However, there is no consensus about the key controlling factors influencing water quality. This paper examined the combined effects of land use and topography on water quality across multi-scale, and identified the key controlling factors determining water quality variation in the headwater watershed of the Hengxi reservoir in Eastern China. Water quality impairment (WQI), expressed as a composite variable, was established to measure the overall water quality. We used the partial least squares (PLSR) method to explore the combination of landscape metrics and identify the key controlling factors. Results showed that the optimal PLSR model at 50-m, 100-m, and 150-m buffer scales and catchment scale explained 77%, 63%, 60%, and 56% of variability in WQI, respectively. At catchment scale, patch density, the percentage of paddy field, and hypsometric integral were the key controlling factors impacting water quality. At buffer scales, the slope gradient, the percentage of forest land, and topographic wetness index were more effectively determined WQI variation. Thus, the key controlling factors depend on spatial scales. Both spatial scales and corresponding key controlling factors should be considered in the adjustment of land use composition and planning of landscape configuration to better protect water quality.