Municipal infrastructure asset management in many developing nations lacks robust, data-driven methodologies for long-term performance forecasting. This creates significant challenges for capital planning and maintenance budgeting, often leading to suboptimal asset yield and service delivery. This case study aims to methodologically evaluate existing municipal infrastructure asset systems and develop a validated time-series forecasting model to predict asset yield improvements for informed engineering and fiscal decision-making. A case study methodology was employed, integrating historical asset performance data with engineering condition assessments. A seasonal autoregressive integrated moving average (SARIMA) model, specified as SARIMA (1, 1, 1) (1, 1, 1) ₁₂, was fitted to the yield data. Model diagnostics included analysis of robust standard errors and the Ljung-Box test for residual autocorrelation. The forecasting model indicates a positive trajectory for aggregate infrastructure yield, with a projected increase of approximately 18% over the forecast horizon. The 95% confidence intervals for the forecasts widen notably after the third year, reflecting increased uncertainty in longer-term predictions. The developed SARIMA model provides a technically sound and transferable methodological framework for forecasting municipal infrastructure yield, demonstrating the utility of quantitative time-series analysis in asset management contexts with limited data. Municipal engineers should adopt similar forecasting techniques to supplement traditional condition assessments. Furthermore, establishing centralised, standardised asset data registries is critical for improving model accuracy and long-term infrastructure planning. asset management, infrastructure yield, time-series forecasting, SARIMA, municipal engineering, predictive modelling This study presents a novel application of the SARIMA modelling framework to forecast the aggregate yield of heterogeneous municipal infrastructure assets, providing a replicable methodology for data-scarce environments.
Nalwadda et al. (Sun,) studied this question.