A SMART GEOSPATIAL APPLICATION INTEGRATING AI AND TIME SERIES ANALYSIS FOR WATER QUALITY DECISION SUPPORT
Pooja Deepak Pawar, Vidula Sohoni
Abstract: Urbanization and release of untreated effluents have largely deteriorated the quality of water in rivers in the rapidly expanding cities like Pune, India. The Mula-Mutha River Basin is affected by constant spatial heterogeneity and seasonal variations as well as non-linear dynamics of the pollutants, which make the use of the traditional monitoring systems and policy-based interventions less effective. It is a research of a smart geospatial decision support framework combining artificial intelligence (AI) and time-series analysis of GIS-based analytics to high-resolution spatio-temporal characterization and prediction of important water quality indicators. The framework uses data on long-term monitoring of the MulaMutha River Basin, and applies Random Forest Regressor models to predict pH, DO (mg/L), BOD(mg/L), FC MPN/100ml, TC MPN/100ml and a Random Forest Classifier is used to classify the overall water quality. Non-linear dynamics are included by use of temporal encoding (month and year), spatial location encoding and seasonal trend modeling. Model validation indicates high predictive capability with R 2 of 0.79 with TC, 0.70 in the case of DO and BOD, respectively and 0.57 in the case of FC. The classification accuracy of the model was 95.5%. Comparative analysis shows that the explanatory power of the key indicators by ensemble-based AI models is more than 40% higher than that of linear regression. Spatial analytics also point to a high level of variability between the upstream reservoir area and urban downstream areas, where the concentration of BOD and coliform is higher around congested settlements. The combined system allows providing early-warning information and policy-supportive decision-making by geospatial visualization and predictive forecasting. The suggested framework illustrates how AI-driven geospatial intelligence can be used to manage river basins sustainably and govern the environment based on evidence.
Keywords: Artificial intelligence; Geospatial analytics; Time-series analysis; Random Forest; Water quality prediction; Mula-Mutha river basin; Decision support system
DOI: 10.24874/PES08.03.011
Recieved: 10.06.2026 Revised: 22.07.2026 Accepted: 19.08.2026
UDC:
Reads: 3 
