Methodological Evaluation of South African Water Treatment Facilities Using Time-Series Forecasting Models for Yield Improvement Analysis

Water treatment facilities in South Africa face challenges in maintaining consistent yield due to variable input parameters such as raw water quality and operational conditions. The study employs a hybrid ARIMA (AutoRegressive Integrated Moving Average) model with robust standard errors to forecast yield improvements based on historical data from selected facilities. A significant improvement in predicted yield was observed when using the ARIMA model, particularly for raw water quality fluctuations up to ±20%. The hybrid ARIMA model demonstrated its potential in forecasting and enhancing yield stability of South African water treatment facilities, offering a practical tool for policy-makers. Policy recommendations include pilot testing the model across different regions and incorporating feedback loops into facility management protocols. The maintenance outcome was modelled as Yᵢₜ=β₀+β₁Xᵢₜ+uᵢ+εᵢₜ, with robustness checked using heteroskedasticity-consistent errors.