Machine Learning and Geospatial Mapping of the Water Quality Index for Cleaner Water Management in Peri-Urban Sub-catchments of the National Capital Region, India

Rapid urbanisation and largely untreated discharge have degraded surface water across the districts bordering Delhi, yet assessments there remain fragmented, with microbiological contamination less well characterised than chemical loading and geospatial mapping seldom applied to an integrated index. This study addresses that gap by consolidating physical, chemical, and microbiological measurements into a single Water Quality Index (WQI) and predicting it across the peri-urban National Capital Region of India, in support of Sustainable Development Goal 6. Surface water was sampled at 118 sites across seven industrially active districts, Ghaziabad, Gautam Buddha Nagar, Faridabad, Gurugram, Jhajjar, Sonipat, and Baghpat, that hydrologically influence the National Capital Territory of Delhi. Ten parameters were measured, including total and faecal coliforms as microbiological indicators. Four decision tree-based regressors, RT, REP Tree, AR_REP Tree, and BG_RT, were benchmarked, and BG_RT was the most accurate, returning CC of 0.9903, NSE of 0.9705, RMSE of 25.5894, MAE of 17.9543, and rRMSE of 8.59% on the testing data. Sensitivity analysis by the Cosine Amplitude Method ranked total suspended solids (R ij = 0.9735) and total solids (R ij = 0.9384) as the dominant controls on WQI, with total and faecal coliforms contributing moderately (R ij = 0.7232 and 0.6743). Geospatial interpolation of the predicted index located severe degradation in the western and eastern zones (WQI > 563.66) and better quality in the less urbanised south (WQI ≤ 86.72). The framework offers a low-cost, scalable tool for prioritising source control, decentralised wastewater treatment, and safe water reuse.