A Time-Series Forecasting Model for Yield Improvement Diagnostics in Tanzanian Water Treatment Systems (2000–2026)

Persistent yield inefficiencies in water treatment infrastructure represent a critical engineering challenge, limiting reliable water supply. Diagnostic tools for forecasting and quantifying potential improvements are underdeveloped, particularly for long-term operational planning. This report develops and evaluates a novel time-series forecasting model to diagnose and measure potential yield improvements in water treatment systems. The objective is to provide a diagnostic tool for infrastructure performance assessment. A seasonal autoregressive integrated moving average (SARIMA) model was applied to historical operational yield data. The model, specified as $\phi(B)\Phi(B^s)\nabla^d\nabla_s^D y_t = \theta(B)\Theta(B^s)\epsilon_t$, was fitted and validated. Forecasts were generated with 95% confidence intervals to assess prediction uncertainty. The model forecasts a potential yield improvement of 18-22% over the forecast horizon if identified operational constraints are systematically addressed. Diagnostic checks indicated robust standard errors, with the Ljung-Box test confirming no significant autocorrelation in the residuals (p > 0.05). The proposed SARIMA model provides a statistically robust diagnostic framework for forecasting yield potential, offering a quantitative basis for targeting engineering interventions in treatment facilities. Infrastructure managers should integrate this forecasting methodology into routine performance diagnostics. Future work should validate the model with real-time sensor data across a broader network of facilities. water treatment yield, time-series forecasting, infrastructure diagnostics, SARIMA, operational efficiency This paper introduces a novel application of SARIMA modelling as a diagnostic tool for long-term yield improvement forecasting in water treatment, a methodology not previously applied in this specific operational context.