How Landscape Patterns Affect River Water Quality Spatially and Temporally: A Multiscale Geographically Weighted Regression Approach

Abstract

The water quality of a river can be considered a function of its surrounding landscape. Understanding the relationship between landscape patterns and river water quality is essential for optimizing landscape patterns to reduce watershed pollution and has not yet been solved. A multiscale geographically weighted regression (MGWR) model was used to explore the associations between the landscape patterns and water quality. Our results showed that landscape metrics have varied relationships with the water quality across spatial scales in different seasons. The strongest independent influencing variable for NO3–-N, NH4+-N, and TN was tea gardens, residential land, and varied seasonally, respectively. The impacts of the landscape metrics on the TP were relatively weak throughout the year at the watershed scale. The influence of landscape metrics on NO3–-N was more significant during the flood season, whereas that on NH4+-N was more notable during the non-flood season. Seasonal changes in the influencing landscape metrics of TN were not regular. Although landscape composition more significantly influenced water quality than configuration, the Shannon’s diversity index and patch density were important configuration indices that significantly impacted water quality. Therefore, with limited land availability, it is essential to optimize the landscape spatial configuration without changing the composition of the watershed to reduce the risk of river pol-lution. This study further indicated that the MGWR model can well quantify the effects of landscape pattern on water quality at the watershed scale.