Ilorin metropolitan water supply infrastructure (WSI) asset management under limited data Open Access

Water supply is the process of efficiently transporting water from its source to the point of usage through a meticulously designed and engineered network of pipes. Urban water supply systems are complex and dynamic. Their management is increasingly challenging due to urbanization, climate change, fluctuating consumer demands, and limited water resources (Adedeji et al. 2022). Efficient management of these systems is essential to provide a sustainable water supply that can accommodate increasing consumer demand. This emphasizes the necessity for intelligent systems to aid in operational management (Adedeji et al. 2022).

Water supply infrastructure is a system of structures, facilities, and components that are designed and built to provide a community or region with a reliable, continuous, and safe supply of water for various purposes. According to Alegre & Coelho (2012), the overall functionality of the system is maintained through the gradual renewal of individual components, rather than complete replacement. This functional definition characterizes infrastructure assets as systems that are continuously revitalized.

Infrastructure is facing two known common challenges: ageing and climate change effects. All infrastructure has a designated lifespan, making the ageing of infrastructure a natural process. Ageing is described as the deterioration process that begins more than 5 years after the start of the operational phase, whereas any deterioration before this period is attributed to design, construction, or operational deficiencies (Perera et al. 2021).

A compelling case of water supply infrastructure challenges is illustrated through a study conducted by Ibrahim et al. (2018) in the local government area of Ilorin West. The research conducted by Ibrahim et al. (2018) focused on evaluating public access to piped water, which is a vital indicator of the effectiveness of water infrastructure. While nearly 500 public water points were identified, suggesting a broad network, the study uncovered a significant concern: nearly 40% of these points were non-functional. The data from Ibrahim et al. (2018) highlights specific challenges faced by Ilorin due to its struggling water infrastructure. The significant number of inoperable water points to issues such as ageing pipes, insufficient treatment capacity, and an outdated distribution network. These limitations hinder the system's ability to meet the demands of a rapidly growing metropolis, underscoring the need for effective management of these infrastructure assets.

In resource-limited settings, water infrastructure management has often relied on reactive approaches, addressing failures as they arise, without a systematic understanding of the network's condition. While these methods provide immediate solutions, they are cost-intensive, disruptive, and inefficient for long-term planning. Condition-based monitoring, which involves periodic inspections to assess performance, has also been employed but is often hindered by incomplete records and limited access to buried components. The Asian Development Bank (2013) highlights the challenges faced by water utilities in developing countries, particularly the lack of comprehensive asset registers and effective maintenance schedules, which lead to system inefficiencies and frequent failures. Similarly, the International Reference Centre for Community Water Supply and Sanitation (IRC 2015) emphasizes the importance of infrastructure asset management in rural water systems, advocating for structured asset data collection and analysis to prioritize repairs and optimize resource allocation. Building on these insights, this study employs advanced statistical methods, such as Monte Carlo simulations and Bayesian updating, to address the limitations of traditional approaches and enable effective management of water infrastructure even in data-scarce settings.

To ensure the long-term reliability and functionality of water supply systems, it is crucial to consider the design lifespan of their critical components. Water treatment plants, designed to purify and distribute clean water, typically have a design life ranging from 25 to 50 years, contingent on the technologies employed and the maintenance practices implemented. The design lifespan of pipes, which serve as essential conduits in networks for water distribution, is influenced by the different materials used in their construction. For example, cast iron pipes may have a design lifespan of 75–100 years, while materials such as polyvinyl chloride (PVC) or high-density polyethylene (HDPE) pipes can have design lifespans of over 100 years. Pumps, integral to water movement within the system, generally have a design life ranging from 20 to 30 years, subject to factors such as usage intensity and regular maintenance. To ensure the longevity of those assets, there should be a continuous yet effective management practice in place.

Asset management involves the integration of design, construction, maintenance, rehabilitation, and renovation processes to optimize benefits and reduce costs. It is a strategy for managing an organization's infrastructure based on a decision-making process guided by a specified standard of service. The term asset management describes business principles focused on balancing risk while minimizing the life-cycle costs of physical assets, including pipes, roads, structures, and equipment. Asset management is also employed as a tool for municipalities to assess the condition of their infrastructure (Bloetscher et al. 2017).

Overseeing the maintenance of water infrastructure assets is particularly challenging, as crucial elements such as primary distribution pipes and large control valves are typically buried underground. These elements, essential for the distribution and regulation of water flow, are out of sight and difficult to monitor due to their deep placement (Bloetscher et al. 2017). This necessitates disruptive and costly excavations for inspections or repairs, often involving significant digging, which is expensive, time-consuming, and disruptive to public life Uncovering subsurface infrastructure is challenging and typically requires excavation efforts (Bloetscher et al. 2017). Without direct visibility, assessing which parts of the system need maintenance or replacement becomes a daunting task. Moreover, water supply systems are vast and intricate, covering extensive areas with numerous interconnected components, adding layers of complexity to their management and maintenance.

The scenario is further aggravated by limited data availability, largely a result of dependency on traditional, manual inspection methods that fall short when it comes to monitoring these inaccessible, deeply buried systems. Addressing these issues necessitates a shift toward innovative technological solutions that allow for indirect monitoring and data collection. The use of sparse data and statistical methods becomes a practical solution. By applying predictive models and analyzing any available records or expert insights, it's possible to make more informed decisions about the infrastructure (Bloetscher et al. 2017). This approach identifies areas where maintenance is most urgently needed, shifting from reactive to strategic management of water supply systems. Even with limited data, it focuses efforts effectively, enhancing both the system's resilience and reliability.

Asset management encompasses the entire lifecycle of infrastructure, planning, construction, maintenance, restoration, and renewal to enhance efficiency and reduce costs (Bloetscher et al. 2017). It serves as a decision-making framework that guides organizations in managing their infrastructure through processes driven by defined service levels. Effective asset management minimizes wasteful or misdirected expenditures while addressing the health and environmental requirements of the community. Organizations that implement asset management programs generally experience longer asset lifespans, as this strategy enables well-informed decisions regarding restoration and maintenance while also adhering to customer expectations and regulatory standards (Bloetscher et al. 2017).

Ensuring sustainability and reliability in urban water systems necessitates effective asset management. Traditionally, infrastructure maintenance strategies have been reactive, addressing issues as they arise. While this approach may seem straightforward, it often results in higher long-term costs and frequent service disruptions (Alegre & Coelho 2012). In contrast, contemporary asset management strategies leverage technological advancements and data analytics to shift toward predictive maintenance and optimization, aiming to improve service reliability and minimize operational expenses (Amadi-Echendu et al. 2010).

Traditional asset management strategies focus on addressing issues as they become apparent, often relying on manual inspections and assessments to gauge asset conditions. These strategies typically follow a sequence of stages: initial design, construction, operation, maintenance, and eventually, rehabilitation or replacement. In mature infrastructures, these stages coexist, reflecting a cycle that repeats throughout the asset's lifespan. However, the reactive nature of this approach can lead to higher costs and reduced asset lifespans, as issues are only addressed after they manifest visibly (Alegre & Coelho 2012).

Innovative strategies adopt a proactive mindset, focusing on the early identification of potential failures and the optimization of maintenance and rehabilitation efforts to extend asset lifespans and enhance efficiency (Bloetscher et al. 2017). This forward-thinking approach advocates for using statistical methods, such as Bayesian analysis, to manage the uncertainty inherent in asset condition assessments, particularly when direct evaluations are impractical, such as with buried infrastructure. This strategy incorporates advanced modeling and simulation techniques to evaluate various scenarios, identify potential vulnerabilities, assess risk mitigation strategies, and explore alternative asset management approaches.

The study by Bloetscher et al. (2017) introduces an innovative approach to address data scarcity in asset management. Their study proposes utilizing Bayesian information theory to infer the condition of water infrastructure assets, even with limited data. This method acknowledges the inherent uncertainties in asset conditions and leverages them to enhance decision-making processes. The study highlights the importance of incorporating both qualitative expert judgments and quantitative data, recognizing the multifaceted nature of evaluating infrastructure health.

Cantos & Juran (2019) provide a comprehensive framework for addressing the challenges posed by ageing infrastructure within water distribution systems (WDS). Their study offers advanced asset management strategies to meet the complexities faced by metropolitan governments and water operators dealing with deteriorating infrastructure. Fundamental to their methodology is the employment of statistical and stochastic methods to examine spatial data related to infrastructure failures, based on a 74-year historical dataset from Wattrelos, France. This dataset underpins a decision support system designed to facilitate strategic planning for infrastructure rehabilitation and replacement.

At the core of Cantos and Juran's research is the development of a risk assessment method that integrates statistical and stochastic models to evaluate the degradation rates of network micro-zones. This approach enables optimized asset management by prioritizing rehabilitation efforts based on a nuanced understanding of the risks associated with ageing infrastructure. The authors demonstrate how an experience-based risk matrix can effectively prioritize network sections for preemptive maintenance, significantly reducing rehabilitation costs and ensuring the sustained delivery of critical water services.

Karasneh & Moqbel (2024) introduce a priority-based decision model for water network rehabilitation, particularly addressing the challenges posed by limited data. Their work presents the fuzzy analytical hierarchy process (FAHP) as a systematic approach to evaluating infrastructure needs. This model integrates quantitative data with qualitative expert judgments, establishing a detailed framework for infrastructure assessment that enhances decision-making processes. Central to Karasneh et al.'s methodology is the application of FAHP to assess a range of factors critical to infrastructure health and operational efficiency. These factors are categorized into five key areas: physical condition, operational efficiency, socio-economic impact, environmental considerations, and quality of service. Each category is meticulously evaluated and weighted based on inputs from experts and stakeholders, ensuring a comprehensive assessment of infrastructure priorities, even amidst data constraints.

Tabesh et al. (2017) tackle the problem of real losses in WDS, which are intensified by insufficient data and the complexities involved in water distribution network operations. They employ Bayesian networks (BNs) to model various factors contributing to these losses, with a focus on real losses that significantly impact non-revenue water. BNs are particularly effective in handling data uncertainties and combining expert knowledge with quantitative data, providing a deeper understanding of real losses. Tabesh et al. (2017) study uses a detailed questionnaire to identify factors such as incorrect pipe installation, poor worker training, and operational inefficiencies. The BN model developed from this data reveals how these factors interconnect and impact real losses. Key findings highlight the significant effects of improper installation and inadequate training on losses. While the study's approach is rigorous, it acknowledges the subjectivity in expert judgments and suggests a structured process to improve objectivity. A case study in Tehran's District 4 reveals that ‘improper and non-standard installation of pipes and devices’ along with ‘insufficient training for workers and experts’ are significant contributors to real losses, informing effective intervention strategies.

Borzi (2023) introduces a methodology for evaluating the vulnerabilities of essential water supply infrastructure concerning climate change and environmental threats, including natural disasters like landslides. The study highlights the significance of safeguarding water supply infrastructures to promote community well-being and facilitate social and economic development. Vulnerability analyses are essential components of international risk management programs aimed at safeguarding critical infrastructure from climate change impacts. A multiple-indicator methodology for vulnerability assessment is introduced by Borzi, utilizing a learning-from-experience approach to identify specific indicators. This methodology encompasses eight indicators divided into four categories: land characteristics, service inefficiencies resulting from infrastructure failures, pipeline route features, and the physical attributes of the aqueduct pipe. The research highlights the growing occurrence of natural disasters, such as landslides, resulting from climate change and their potential to harm both the environment and man-made infrastructure. It explores how climate-induced changes in soil settings, moisture and temperature variations, and seasonal physical failures of pipeline materials contribute to the vulnerability of water supply systems. Borzi's research offers a thorough framework for assessing and mitigating risks to water supply infrastructure, enabling the identification of critical vulnerability factors and aiding decision-making in the management, planning, and design of resilient water supply systems.

Jayawickrama et al. (2022) developed a BIM-based 3D asset database for municipal infrastructure, focusing on a 1.3-km road segment in Colombo, Sri Lanka. The methodology involved four phases: conducting a literature review and data collection, identifying relevant assets and parameters, developing a 3D asset database using InfraWorks 360, and applying this database to a case study. The database was enriched with data from GIS files, AutoCAD drawings, and expert interviews, incorporating parameters like pipe dimensions, material, and maintenance activities.

In the final phase, the database was applied to the selected road segment, integrating detailed asset information such as pipe IDs, diameters, and maintenance schedules. This comprehensive 3D model was then uploaded to the BIM 360 cloud platform, allowing municipal engineers to navigate and utilize the data for informed decision-making in infrastructure management and maintenance. The study highlights the potential of BIM-based databases to enhance asset management practices by providing a rich, integrated view of municipal infrastructure.

The National Population Commission estimated Ilorin's population to be approximately 532,088 in 1991 and 781,933 in 2006. Ilorin is predominantly a Yoruba town, with the Yoruba people making up approximately 60% of the population. The remaining 40% of the population is comprised of the Fulani, Hausa, Nupe, and other ethnic minorities. This diversity in language and ethnicity contributes to the uniqueness of its culture (Mas'ud et al. 2020). Ilorin experiences variability in rainfall both over time and across different areas. The region receives an average annual rainfall of approximately 1,200 mm. The city's relative humidity is around 65% during the dry season and ranges from 75 to 80% in the wet season. The average monthly temperature in the area ranges from 25 to 28.9 °C (Shiru et al. 2015). Ilorin is home to numerous financial institutions, including banks, insurance companies, and cooperative societies, among others. Various educational institutions are present in Ilorin, including universities, polytechnics, colleges of education, nursing schools, and many primary and secondary schools. The city hosts media outlets, including television and radio stations, as well as correctional services (Mas'ud et al. 2020). Ilorin's infrastructure includes notable financial, educational, and media institutions, in addition to a comprehensive water supply system. This infrastructure features the Asa, Sobi, and Agba dams, a thorough water treatment facility, embankment dams for regulation, pumps for water distribution, and a wide-ranging network to guarantee the delivery of clean water across the city, highlighting its significance in urban water management.

This study introduces a systematic framework for overseeing water infrastructure assets in Ilorin, tailored for scenarios where data availability is constrained. Moving beyond traditional, disruptive methods, this approach adopts a proactive stance toward asset management, structured across four main phases: data collection, statistical analysis, predictive modeling, and dynamic assessment.

Due to the absence of comprehensive records, this study generated simulated data based on limited information from the Asa Dam area. The dataset incorporated essential attributes, including intake structure specifications, dam details, treatment plant features, storage capacity, distribution networks, and maintenance history, with a specific focus on the intake pipe. Key variables such as pipe breaks, diameter, age, soil type, burial depth, and material were generated using the Monte Carlo method, providing a controlled dataset that mirrors real-world scenarios for subsequent analysis.

The variables selected for the simulation were based on their established relevance to water infrastructure asset management, as supported by prior studies and expert consultations:

• (i) Pipe breaks: A critical indicator of asset condition, as frequent breaks often signal structural weaknesses or material degradation. This variable was modeled using a Poisson distribution, which is well-suited for count-based events.

• (ii) Diameter (Dia): Larger diameters may experience internal pressure fluctuations that could contribute to material fatigue. This was assigned a uniform discrete distribution reflecting standard pipe diameters used in the study area (e.g., 50–100 mm).

• (iii) Age: Ageing infrastructure is a known risk factor for failures. This was generated using a uniform distribution within the range of 30–46 years, reflecting the typical lifespan of pipes in the region.

• (iv) Material: The material type (e.g., Ductile, GI, PVC, AC, and HDPE) influences both durability and susceptibility to environmental factors. This was modeled as a categorical variable with equal probabilities, as precise empirical distributions were unavailable.

• (v) Soil type: Soil characteristics (e.g., sand and clay) directly affect pipe degradation rates. This variable was represented as a categorical distribution to account for the binary nature of soil types in the study area.

• (vi) Bury depth: The burial depth (shallow or deep) impacts pipe exposure to external forces. This was similarly modeled as a categorical variable.

Monte Carlo methods use repeated random sampling to generate numerical results. Here's a general outline of the procedures for conducting the Monte Carlo method:

• (i) Define the problem: Clearly state the objective and define all relevant variables and their relationships.

• (ii) Generate random inputs:

• • • Identify the variables that are not fixed and have variability.

  • • Assign appropriate probability distributions to represent this variability.

  • • Generate random samples from these distributions.

• (iii) Simulate the model:

• • • Run simulations using the generated random inputs.

  • • Record the results of each simulation run.

While Monte Carlo methods offer a comprehensive framework for generating synthetic datasets, it is essential to acknowledge potential limitations. Simulated datasets rely on assumed probability distributions and input parameters, which may not fully capture the complexities of real-world conditions. For example, pipe failure rates or material degradation patterns may be influenced by unique local environmental factors or operational anomalies that are difficult to model precisely. These deviations could affect the accuracy of the predictive models and the applicability of the findings in some contexts.

To address these concerns, the assumptions used in the simulations were validated through expert consultations and calibrated based on limited historical records from the study area. Additionally, the subsequent integration of Bayesian updating allowed for dynamic refinement of the model predictions using sparse observational data, ensuring the results remain grounded in real-world evidence. Despite these measures, the findings should be interpreted with caution, particularly in scenarios where the underlying assumptions deviate significantly from actual conditions. Future studies could focus on integrating hybrid approaches that combine limited empirical data with simulated datasets to further improve accuracy and applicability.

To ensure the validity of the logistic regression model, it is important to consider its key assumptions. Logistic regression assumes a linear relationship between the log-odds of the dependent variable and the independent variables. This assumption was met by designing the simulated dataset to reflect real-world variability, including key attributes like pipe breaks, material type, and soil characteristics. Additionally, balancing the categories of good, fair, and bad pipe conditions minimized potential classification bias, ensuring a more accurate predictive framework.

However, the limited availability of real-world data introduced challenges, such as reduced model generalizability and the risk of overfitting to simulated scenarios. These concerns were addressed through expert consultations to validate simulation assumptions and parameters. Moreover, Bayesian updating, discussed in the next section, was employed to dynamically refine predictions by incorporating sparse observational data, further mitigating the limitations of relying solely on simulated datasets.

This formula recalculates the probability of the infrastructure being in a particular condition after considering the new observation. Bayesian updating assumes that prior probabilities and observational data can be combined to refine predictions. This method was particularly well-suited to the simulated dataset, as expert input was used to define the priors, ensuring the model adapts dynamically even with limited empirical data. Sparse observational data were incorporated into the framework to recalibrate predictions, bridging the gap between simulated and real-world scenarios. By dynamically integrating these sparse data points, Bayesian updating mitigates potential inaccuracies and enhances the reliability of the predictive framework.

This approach mitigates some of the limitations of the simulated dataset by incorporating new observational data to recalibrate model predictions. However, it is important to recognize that the accuracy of this method still depends on the quality and representativeness of the observational data used for updating.

For this study, a dataset for intake pipes was simulated using the Monte Carlo method, controlled by the limited information from the Asa Dam treatment plant, Ilorin. Various scenarios were constructed to build a predictive model. The dataset, summarized in Table 1, includes the following columns:

• (i) Asset: Identifies the type of asset, focusing on the intake pipe.

• (ii) Breaks in 10 years: Number of pipe failures over a decade, generated using a Poisson distribution.

• (iii) Dia (mm): Pipe diameter, randomly selected between 50 and 100 mm.

• (iv) Age: The age of the pipe, uniformly distributed between 30 and 46 years, reflecting major maintenance in 2009 and the plant's establishment in 1978.

• (v) Soil type: Represents the type of soil surrounding the pipe, categorized as sand or clay.

• (vi) Trees and no trees: Presence or absence of trees around the pipe.

• (vii) Burial depth: Specifies the depth at which the pipe is buried, categorized as either shallow (under 6 feet) or deep.

• (viii) Material types: Includes ductile iron, galvanized iron, PVC, asbestos cement, and HDPE.

• (ix) Current condition: Categorizes the pipe's state as good, fair, or bad, with adjustments for age-related deterioration in pipes older than 40 years.

Table 1 provides a sample of the simulated data, reflecting the diverse scenarios used to develop the predictive model. This table is critical for understanding how the synthetic dataset was constructed and highlights the variables used in the analysis.

The correlation between pipe diameter and current condition is weak and slightly positive (r = 0.04), implying that larger diameter pipes may be in slightly better condition but not significantly enough to be a major factor.

Material types show generally weak correlations with the current condition, with HDPE (r = 0.06) indicating slightly better conditions due to resistance to corrosion and durability, whereas PVC and AC pipes may deteriorate faster under certain conditions. Soil type (sand or clay) and tree presence (trees or no trees) exhibit very weak correlations, suggesting minimal impact on pipe condition. Similarly, burial depth (shallow under 6 or deep bury) shows a weak correlation, indicating it does not significantly affect pipe condition.

The findings reveal that pipe age and the number of breaks are the most critical risk factors, whereas other factors, such as material type and burial depth, show weaker correlations. Prioritizing maintenance and replacement activities based on these insights can help improve the overall reliability and performance of the water supply infrastructure.

In the bad condition category (0), the model correctly predicted 78 out of 86 actual cases, demonstrating high accuracy, with eight instances misclassified as good. Notably, no bad-condition pipes were misclassified as fair. For the good condition category (1), the model correctly predicted 65 out of 75 actual cases, indicating strong predictive accuracy, though there were eight instances misclassified as bad and two as fair. The fair condition category (2) presented more challenges, with the model correctly predicting 35 out of 39 actual cases, but four instances were misclassified as good. Overall, the model shows strong predictive accuracy for both good and bad conditions, with fewer classification errors. However, the accuracy for predicting fair conditions is lower, indicating an area for potential improvement. The confusion matrix highlights the model's strengths and areas needing refinement.

In evaluating the state of the intake pipe, Bayesian methods were applied to dynamically update the probability of the pipe being in good condition based on expert observations.

Case scenario:

The initial prediction from the logistic regression model indicated that the intake pipe is in good condition (1) with a prior probability (P (Condition)) of 0.6.

An expert inspection revealed no negative observations. The expert provided the following estimates:

• (i) Likelihood (P (Observation|Condition)): The likelihood of not detecting any negative indicators, such as cracking or rusting, when the pipe is functioning properly was estimated to be 0.8.

• (ii) Overall probability (P (Observation)): The probability of observing no negative signs, taking into account all potential states of the pipe, was estimated to be 0.7

  

  

  

  

  

The updated probability that the intake pipe is in good condition, given the expert's observation of no negative signs, is approximately 0.686. This updated probability represents an increase in confidence from the initial prior probability of 0.6. To formulate a prioritized asset management plan, the updated probability of the intake pipe being in good condition (0.686) serves as a key factor.

The findings of this study underscore the significance of adopting proactive asset management strategies for urban water supply infrastructure, particularly in resource-constrained settings. By leveraging Monte Carlo simulations and Bayesian updating, the study provides actionable insights into critical risk factors such as pipe age and break frequency, which are often challenging to address using traditional management approaches.

This study builds upon earlier research that has highlighted the challenges faced by water utilities in developing regions. For instance, the Asian Development Bank (2013) emphasized the inefficiencies caused by incomplete asset registers and the absence of predictive tools, while the International Reference Centre for Community Water Supply and Sanitation (IRC 2015) advocated for structured data collection to prioritize repairs in rural systems. However, these studies primarily focused on static methods and lacked dynamic predictive capabilities.

In comparison, this study introduces a dynamic framework that integrates statistical modeling with Bayesian methods, enabling real-time adjustments to predictions based on new observational data. For example, unlike the findings of Ibrahim et al. (2018), which qualitatively identified ageing pipes as a key issue in Ilorin's water infrastructure, this study quantifies the relationship between pipe age, break frequency, and condition. The use of advanced statistical techniques allows for a more precise understanding of risk factors, providing practical recommendations for targeted maintenance strategies.

The study's results offer a range of practical applications for water utility management:

• (i) Prioritized maintenance: By identifying pipes with frequent breaks or exceeding a critical age threshold, maintenance efforts can be directed where they are needed most, reducing downtime and repair costs.

• (ii) Efficient resource allocation: The predictive model enables resource-limited utilities to focus their financial and manpower resources on high-risk assets, improving overall operational efficiency.

• (iii) Policy formulation: The insights gained from this study can support the development of policies that mandate the adoption of predictive models and routine data collection practices in infrastructure management.

These findings not only address existing gaps in the literature but also provide a practical framework that can be adapted for similar regions with limited data availability. By emphasizing both predictive accuracy and real-world applicability, this study contributes to the evolving field of infrastructure asset management and offers a roadmap for future research.

The study successfully achieved its objectives through a structured approach tailored for managing Ilorin's water supply infrastructure assets, with a specific focus on the intake pipe. A comprehensive simulated dataset was generated for the intake pipe, based on limited available information. This dataset included key attributes such as the number of breaks, pipe diameter, age, soil type, presence of trees, burial depth, and material type. The simulated data provided a solid foundation for further analysis and modeling, facilitating a detailed understanding of the intake pipe's characteristics and potential vulnerabilities.

Correlation analysis identified significant risk factors, including pipe age and the number of breaks. These factors were crucial in assessing the vulnerabilities based on the location and physical characteristics of the intake pipe. A predictive model for asset failure was developed using logistic regression, which demonstrated strong predictive accuracy, especially for pipes in good and bad conditions. Bayesian methods were then employed to dynamically update the probabilities of asset conditions based on expert observations, leading to more accurate and current assessments. The updated probabilities informed a prioritized asset management plan, optimizing maintenance, resource allocation, and strategic planning for the intake pipe.

To translate these findings into actionable outcomes, the following recommendations are proposed for local stakeholders and water authorities in Ilorin:

• (i) Implement prioritized maintenance plans: Utilize the predictive model to identify high-risk assets, particularly pipes with frequent breaks or exceeding critical age thresholds, to minimize failures and optimize repairs.

• (ii) Establish routine data collection frameworks: Regularly collect data on pipe performance and environmental conditions using affordable technologies, such as IoT sensors, to ensure real-time updates and enhance predictive accuracy.

• (iii) Advocate for policy development: Encourage the adoption of data-driven infrastructure management practices through local policies, including budgetary support for predictive modeling and routine asset monitoring.

• (iv) Integrate into regional infrastructure management frameworks: Collaborate with national and regional agencies to scale the methodology, fostering a unified approach to water infrastructure management in resource-constrained settings.

By adopting these recommendations, water authorities in Ilorin can improve the resilience and sustainability of their water supply systems, ensuring reliable service delivery for the city's growing population. Future studies could explore hybrid approaches that combine limited empirical data with simulated datasets to further enhance accuracy and applicability in asset management.

This study demonstrates the potential of combining simulated datasets and advanced statistical methods to manage water supply infrastructure under data constraints. However, certain limitations must be acknowledged:

• (i) Data scarcity and model assumptions: The simulated datasets relied on assumed probability distributions and expert inputs due to the limited availability of empirical data. While these assumptions were validated through expert consultations and limited historical records, they may not fully capture the complexities of real-world conditions. Factors such as localized pipe degradation patterns influenced by environmental or operational anomalies remain challenging to model accurately.

• (ii) Generalizability of findings: The study's findings are tailored to the specific context of Ilorin's water supply infrastructure. Applying the proposed methodology to other regions may require recalibration of simulation parameters and model assumptions to account for variations in environmental, operational, and socio-economic factors.

• (iii) Uncertainty in expert observations: Bayesian updating integrates expert observations, which are subject to human judgment and potential biases. While this approach improves predictions in data-scarce environments, more structured and objective methods of incorporating expert input could enhance accuracy.

Building on the insights and limitations of this study, future research could focus on the following areas:

• (i) Hybrid approaches: Explore methodologies that combine limited empirical data with simulated datasets to improve the robustness and accuracy of predictive models. For instance, integrating sparse observational data with machine learning techniques could refine the framework further.

• (ii) Real-time monitoring: Incorporate IoT-enabled sensors to provide real-time data on pipe performance and environmental conditions. This would enhance the dynamic assessment capabilities of the framework, allowing immediate responses to emerging risks.

• (iii) Scalability and adaptability: Extend the framework to accommodate larger and more complex infrastructure systems beyond Ilorin. This would involve developing scalable algorithms and conducting comparative studies across different geographical and socio-economic contexts.

• (iv) Stakeholder collaboration: Engage local stakeholders, including policymakers, water utility managers, and community representatives, to refine the framework and integrate it into existing infrastructure management practices. Such collaborations could also facilitate the adoption of predictive models at a regional or national level.

• (v) Advanced data collection techniques: Invest in affordable and innovative data collection methods, such as unmanned aerial vehicles or ground-penetrating radar, to gather more comprehensive information on buried infrastructure components.

By addressing these limitations and pursuing the suggested future directions, the proposed framework can evolve into a more comprehensive and adaptable tool for managing water supply infrastructure in resource-constrained settings.

Conceptualization by O.G. and T.S.; analysis by L.O., O.G.; initially draft by L.O.; written, reviewed, and edited by O.G.; oversight by A.W. All authors have read and agreed to the published version of the manuscript

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.