• Characteristic scale;
  • Independence test;
  • Land cover/use;
  • Restoration;
  • Stream ecosystem

The aim of our study was to investigate the most effective predictors and scales to explain water quality and macroinvertebrate in the Sha River Basin, China. Predictive variables included land use/cover and physical characteristics at a fine gradient of spatial scales (60, 120, 240, 480, and 960 m and whole catchment). Multivariate regression results indicated agriculture and urban land were both primary predictors for water quality at the catchment scale in the Sha River Basin. Partial Mantel tests further showed most of correlation between agriculture and water quality could be attributed to urban land and spatial covariate. Macroinvertebrate exhibited no direct association with most land cover percentage at all scales, but significant relationship with in-stream physical variables. Increasing forest cover could help to improve water quality and macroinvertebrate biodiversity. Among all scales, the catchment scale was the most effective scale to detect water eutrophication in the Sha River Basin, with urban land as the primary cause, while in-stream habitat scale was the most effective scale for macroinvertebrate restoration. Overall, empirical models combined with independence tests could be used to clarify the influence of landscape change on strongly disturbed stream ecosystem, and may serve for guiding effective management and restoration.