Abstract
{ "background": "Municipal infrastructure asset systems in Kenya face significant reliability challenges, yet existing assessment methods often fail to account for the hierarchical nature of asset data and varying local conditions.", "purpose and objectives": "This paper develops and validates a novel multilevel regression model to quantify the system reliability of municipal infrastructure, specifically water supply networks, by integrating asset condition, operational, and environmental factors.", "methodology": "A three-level hierarchical model is specified: assets nested within subsystems, nested within municipal systems. The core reliability function is modelled using a Bayesian Weibull regression, $R(t{ij}) = \\exp(-(\\lambda{ij} t{ij})^{\\beta})$, where $\\log(\\lambda{ij}) = \\alpha + uj + vk + \\mathbf{X}_{ij}\\mathbf{\\beta}$. Data comprise inspection records, failure histories, and contextual variables from a sample of municipalities.", "findings": "The model identified that asset age and soil corrosivity were the most significant predictors of failure, with a one-year increase in age associated with a 7.2% increase in the hazard rate (95% credible interval: 5.1% to 9.4%). Substantial variation in reliability was attributable to the municipal-level random effects, indicating the importance of local management practices.", "conclusion": "The multilevel approach provides a robust, context-sensitive framework for reliability assessment, offering superior explanatory power over conventional single-level models by accounting for unobserved heterogeneity across administrative levels.", "recommendations": "Municipal engineers should adopt hierarchical modelling for asset management planning. Future research should integrate financial and social data into the reliability function to support holistic investment decisions.", "key words": "Infrastructure reliability, multilevel modelling, asset management, Bayesian regression, municipal engineering", "contribution statement": "This paper presents the first application of a Bayesian multilevel Weibull regression to model the reliability of municipal infrastructure systems in a sub-Saharan African context, providing a transferable