Optimizing water quality monitoring networks through temporal and spatial analysis

Efficient design of water quality monitoring networks is vital for improving watershed-scale pollution control, particularly in rapidly urbanizing river basins. This study aims to establish a spatiotemporal optimization framework that integrates autocorrelation analysis, local spatial clustering (Local Moran’s I), K-means clustering, and random forest interpretation. Monthly water quality data from 65 stations in the Tanjiang River Basin (southern China) during 2018–2024 were used to evaluate temporal persistence and spatial aggregation of 12 key parameters. Temporal autocorrelation analysis revealed strong annual periodicities for nitrate nitrogen (NO₃⁻-N), water temperature (WT), chloride (Cl⁻), and sulfate (SO₄²⁻), suggesting their suitability as indicators for long-term trend detection. Spatial analysis identified significant local clusters of NO₃⁻-N, SO₄²⁻, fluoride (F⁻), and ammonia nitrogen (NH₃-N), reflecting pollution hotspots tied to anthropogenic land use. Based on these spatiotemporal patterns, K-means clustering stratified stations into three categories—“Add”, “Keep”, and “Merge”. A random forest model was then applied to evaluate the relative importance of each parameter, identifying Cl⁻, NO₃⁻-N, and SO₄²⁻ as the most influential variables. The model also showed high classification consistency with the K-means result (95.0 % accuracy), indicating strong agreement between unsupervised grouping and feature-driven interpretation. This integrated method supports strategic adjustment of monitoring networks by reducing redundancy while retaining representativeness. It offers a scalable solution for data-driven environmental governance, especially under resource constraints. Future work should incorporate real-time data, cost-efficiency evaluation, and adaptive scheduling to further enhance network responsiveness.