Multiscale impacts of landscape metrics on water quality based on fine-grained land use maps

Quantifying the impact of landscape metrics on water quality can offer scientific supports for water conservation and land use planning. However, previous studies mainly relied on coarse land use maps, and were lack of understanding of effects from physiographic metrics. Here, based on the in-situ water quality monitoring data in the Fujiang river basin, we used redundancy analysis, variation partitioning analysis, and Shapley Additive exPlanations methods to assess the impact of landscape metrics on water quality. We use these analyses in the dry and wet season, in circular buffer zone, in riparian buffer zone, and at the sub-basin scale, we are able to analyze and understand the complex interactions between landscape features and water quality, as well as spatial and temporal scale effects. The results indicated that the impact of landscape metrics on water quality variation can be ranked in the following order: landscape composition (15.8%–32.2%) > landscape configuration (1.2%–19.5%)> physiographic metrics (−2.0%-0.6%). Forests and grasslands improved water quality, whereas farmland and impervious surfaces degraded water quality. At a finer scale of land use types, closed broadleaf evergreen forests improved water quality, while rainfed cropland had the opposite effect. The 1500 m circular buffer was the key scale with the highest rate of interpretation. The relationship between landscape metrics and water quality was marginally stronger during the wet season than the dry season. Water quality was improved by large relief amplitude and slope standard deviation. The water quality is not significantly affected by the river network density, the length of the river, or the basin area. These conclusions could provide science-informed information and support to the study between landscape metrics and water quality.