What question did this study set out to answer?

The study aims to develop a hybrid model to predict water quality parameters in the River Ganges to enhance sustainable management.

March 21, 2026Open Access

Spatiotemporal forecasting of river Ganges water quality using a hybrid ARIMA LSTM-Random Forest model: Enabling data-driven sustainable water management

Key Points

The study aims to develop a hybrid model to predict water quality parameters in the River Ganges to enhance sustainable management.
Integrated ARIMA, LSTM, and Random Forest models for forecasting.
Utilized a 14-year dataset from 23 monitoring stations.
Implemented spatiotemporal imputation to address data gaps.
Predicted DO with strong accuracy (pooled R²=0.88) and BOD (pooled R²=0.80).
Identified persistent pollution hotspots for future monitoring.
Reported low pH levels (6.5–6.9) and hypoxic DO levels (~2 mg/L) at certain sites.

Abstract

The River Ganges in Uttarakhand, India, faces pollution pressures from urban expansion, industrial effluents, and agricultural runoff, rendering traditional monitoring approaches inadequate. This study develops and validates a hybrid predictive framework integrating Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), and Random Forest (RF) models to forecast key water quality parameters (pH, Temperature, BOD and DO). This approach is designed to handle challenges of inconsistent, long-term environmental data. Utilizing a 14-year dataset (2011–2024) from 23 monitoring stations, data gaps were addressed through spatiotemporal imputation. The model demonstrated strong predictive accuracy, particularly for DO (pooled R²=0.88) and BOD (pooled R²=0.80). Forecasts to 2040 identify persistent pollution hotspots, with site XVIII exhibiting low pH (6.5–6.9), and sites XVI, XVIII and XXII with DO levels nearing hypoxic conditions (~2 mg/L) and BOD concentrations exceeding regulatory standards (up to 31.9 mg/L). Elevated temperatures (>26 °C) suggest localized thermal pollution. The highest prediction errors for BOD (RMSE=4.41) coincided with these hotspots, reflecting challenges in modelling event-driven contamination. The framework proves to be a powerful tool for proactive, data-driven river management, pinpointing future critical zones to guide targeted interventions, while highlighting that reliable forecasting of complex parameters like BOD requires improved real-time monitoring.

Perguntar à IA

Bookmark

View Full Paper