Quantification of water loss indices in a water distribution network: a case of water and sewerage company of Nairobi City, Kenya Open Access

Globally, water utilities are struggling with persistent gaps to ensure the reliable and sustainable circulation of water to users (Adedeji et al. 2022). Most water utilities in sub-Saharan Africa (SSA) are mandated by the roles and obligations for potable water provision and managing consistent supply and demand for potable water consumption (Adedeji et al. 2022). This aligns with the sixth Sustainable Development Goal (SDG6) of the United Nations (UN), which can be achieved through accurate and precise quantification of water loss indices and the provision of long-term solutions to bridge the current gaps.

Most urban areas are currently struggling to supply potable water to residences, despite the challenges of non-revenue water (NRW). Urbanization pressure, steady population growth, dilapidated water infrastructure, and unfavourable climatic conditions are some of the reasons that contribute to NRW (Jang 2018). Water loss, irrespective of its nature, is a serious problem that requires concerted efforts by all stakeholders to devise appropriate interventions to reduce it, as approximately 60% of potable water is unaccounted for during conveyance to users (Jang 2018).

Water losses in water distribution networks (WDNs) negatively impact the socio-economic development of countries and regions. By discovering, detecting, and minimizing leaks and other forms of water loss promptly, a platform is needed to manage WDNs more effectively and ensure that consumers have access to quality and reliable water supply (Jung & Kim 2018). Over the last two decades, researchers have invested considerable effort in developing efficient leakage management techniques to lower water losses in WDNs, even though some significant breakthroughs have been realized in the global north (Kanakoudis & Gonelas 2014, 2016; Kanakoudis & Muhammetoglu 2014; Gonelas & Kanakoudis 2016); however, this has not been achieved in the Global South (Adedeji et al. 2017).

Leakage is a significant problem in systems and networks that distribute water. Water utilities can manage leaks and minimize water loss by integrating data from diverse transmission routes. Currently, most leakage sensorizations adopt manual procedures (Soldevila et al. 2022). Quantification of these losses can offer a huge potential route for the viability of water utilities because it can guarantee the security of extensive water distribution systems, reliability of supply, detection of leaks and illegal connections, monitoring of water quality and quantities, managing pressure variations, and minimizing NRW (Shabangu et al. 2020). In addition to issues of water leaks, other concerns regarding water supply directly influencing the efficient delivery of potable water in WDNs are the management of water demand, scheduling and oversight of operations, pump scheduling, pressure control, corrosion of metallic pipes and other metallic devices, water quality degradation, and water security, which pose great challenges to water utilities globally (Kanakoudis 2004; Bello et al. 2018).

The annual water loss volume in the world is considerably high, estimated 126 billion m³, which costs around US$ 39 billion annually (Liemberger & Wyatt 2019). A large proportion of water loss (WL) occurs because of water leaks in WDNs, which are sufficient to serve over 200 million people (Norouzi et al. 2019). Different approaches have been proposed to estimate WLs that occur in WDNs (Al-Washali et al. 2018). Apparent losses (ALs) and real losses (RLs) are the key components of WLs and are normally assessed by applying different methodologies such as top-down, bottom-up, and component analysis of leakages (Al-Washali et al. 2016, 2021; Al-Washali et al. 2018; Taha et al. 2020). These approaches have been applied in various contexts and at varying scales to approximate WL components, and various measures have been proposed (Kanakoudis & Tsitsifli 2010; Kanakoudis et al. 2013a; Al-Washali et al. 2018; Norouzi et al. 2019).

The Nairobi City Water and Sewerage Company (NCWSC) cannot operate profitably under the current dynamic conditions without managing and quantifying WL and determining their indices. This necessitates higher operating costs and additional investment. In addition to serving as a benchmark for future research on the NCWSC WDN and other water utilities across the nation and region, this study is the first attempt to assess, evaluate, and measure WL indices/indicators. Based on past successful experiences, water industry authorities should promote the adoption of suitable structural policies and rules.

The research work assessed, analysed, and quantified the volume of water lost by NCWSC in Nairobi. The WL indices of the company were then compared with the international standards set by the International Water Association (IWA), the American Water Works Association (AWWA), and the World Bank Institute (WBI) banding system (Grievson et al. 2022; Alegre et al. 2000, 2006, 2016). It critically analysed interventions for real-time detection and correction of damaged pipes, resulting in the minimization of NRW. The study analysed recorded data (2014–2023) from the NCWSC for one decade for better and more insightful perspectives (Ociepa et al. 2018, 2019).

The total volume of water supplied to the network, water used for social welfare, water for both productive and non-productive use, length of the network, the total length of water supply connections, materials, and network age, number of end users/recipients, population, average annual pressure in the network, and number of failures in individual years were the specific parameters that were investigated. These data were utilized to compute various indices suggested by the AWWA, IWA, and WBI banding systems, as well as the percentage ratio of WL (WS) to unit WL per capita (Q_los) and the hydraulic load index (q_o) (Ociepa et al. 2019). These include the infrastructure leakage index (ILI), real leakage balance (RLB₂), unavoidable annual real losses (UARLs), and NRW basic index (NRWB). Metrics of WL, which are commonly used in the literature as well as the recent development of decision support systems, which are comprehensive tools used to check pipe failures, water quality, water demand, water interruptions, and customer complaints (Kanakoudis et al. 2011, 2013b, 2015a, b; Ociepa et al. 2018, 2019; Hoțupan et al. 2019), were adopted for computations as per Equations (1)–(10).

The system input volume (SIV) is the water supplied to the network in m³/year, unbilled authorized consumption (UAC) is the water used by the company, and billed authorized consumption (BAC) is the actual amount of water sold. All units are in m³/year.

IN is the number of inhabitants using water from the water supply system.

Figure 2 presents comprehensive data on water production and sales volumes for the NCWSC from 2014 to 2023. To compute the water balance, information about the volumes of water supplied to the network is required; thus, SIV, BAC, and UAC are the volumes of water sold and volumes used for internal use by the company, respectively. The CARLs in the WDN were calculated using Equation (1) and are shown in Figure 2. The calculated water balances (Figure 2) and UARL, computed using Equation (5), show the values of the assessed WDN. These values will help in further computation and analysis of the WDN WL indices and performance status. The main idea of water balance characterization (Figure 2) is to provide concise and precise insights into water utility management on the viability of utility operations through evidence-based analysis of actual production volumes and billed volumes.

In contrast to the WL index (WS), computed based on Equation (2), the NRWB, calculated based on Equation (4), as recommended by the IWA, is utilized for a more accurate evaluation of WLs. This index does not consider the amount of water the company uses for its requirements, which helps prevent inaccuracies caused by certain firms purposefully overstating the amount of water they use for their own needs.

The values for WS and NRWB for the NCWSC WDN indicate inconsistent and insignificant reductions in WLs in the analysed years from 2014 to 2023. In 2014, WS was 38.98%, and in 2023, it was reduced to 34.87%, which shows a reduction of 4.11% over the last 10 years. For the NRWB, a value of 45.64% was reached in 2014, which was also the highest recorded value. By 2023, a value of 38.86 was reached, indicating a 6.78% drop over the last 10 years. It is also important to note that, in 2016, the WS value of 39.76% was the highest, with a corresponding value of 43.24% for NRWB. This shows an extremely high level of WL compared with the global systems recommended by the IWA standards, which should be below 25%, as indicated by WBI banding systems and AWWA conditions. It is worth noting that the indices were lowered by more than 4% over the entire 10 years of analysis. This means that the utility is making progress towards reducing the level of NRW, but a more concerted effort is still required, especially in the modernization of the WDN.

The WS and NRWB values were comparatively higher for this network compared with the WL levels in the Netherlands, which fell within the range of 3 − 7% and were much higher than the average values for the USA (15%), Canada (13.8%), Italy (42%), and Greece (34.9%), where the mean values for WS fell within the range of 13 − 35% (Kanakoudis & Tsitsifli 2014; Ociepa-Kubicka et al. 2024). This confirms that the WS and NRWB of the assessed water company are higher.

In contrast, it should be noted that because water supply networks differ in terms of their hydraulic load, material, total running length, age, pressure, and number and length of connections, they lack sufficient concrete evidence or are incorrect for comparing and evaluating loss levels solely based on the percentage of WL index. In addition, the amount of water utilized for the water supply utility's internal needs, which is an estimate, affects the index score. It is not advisable to apply it in comparison to other WLs in various reticulation networks because of the aforementioned factors. The assessment of WL variability over an extended period within a single distribution system is the only recommended and most accurate approach (Kwietniewski 2013).

The NRWB, as per Equation (4), was used to determine WLs in accordance with the IWA recommendations for a more reliable evaluation of the technical condition of the network. Because its value ignores the amount of water used for a water supply system's internal demands, this system's index is greater than that of the WS. The NRWB index, as per the calculations, recorded a decline of 6.78% over the past decade, as shown in Figure 3.

The IWA also recommends using the unique RLB₂ to compare the states of the WDNs. RLB₂ calculates the volume of water lost per day per water supply connection. The RLB₂ for the NCWSC reticulation network for the duration under study (Figure 3) has shown a significant decline of RLB₂ of 310.54 dm³/(connection·day), decreasing from 1015.73 dm³/(connection·day) recorded in 2016 to 705.19 dm³/(connection·day) in the year 2023. However, there was some inconsistency; for example, in 2014, the index was 1000.66 dm³/(connection·day). Other studies have shown that the RLB₂ index in Poland in 2015 was estimated to be between 150 and 200 dm³/(connection·day), whereas, according to reference values for retail systems in Portugal, the quality of services is good if RLB₂ < 100 dm³/(connection·day) (WAREG Report 2017). In New Zealand, this value ranges from 100 to 290 dm³/(connection·day) (Ociepa 2021). Generally, in Western European countries, 100 dm³/(connection·day) is regarded as the highest allowable RLB₂ (Ociepa 2021). Comparing the values obtained for the last 10 years from the NCWSC water distribution system in Kenya, the values are considered to be almost 10 times higher than the expected average values, despite showing positive downward trends. Given the above comparison, the WDN conditions for WL reduction and management are considered unfavourable (Ociepa et al. 2019).

Another computation for the WL index suggested by the IWA is the unit WL per capita or inhabitant per day (Q_los), which was computed using Equation (7). The WDN of the NCWSC was subjected to this index from 2014 to 2023. The results shown in Figure 3 indicate that the highest value of 77.08% was realized in 2016, while the lowest value of 49.31% occurred in 2023, with a reduction of 27.77%.

Inconsistency was noted, as in 2014, the index was 76.68%, which increased in 2016; thereafter, a positive declining trend was recorded until 2023. This index is typically used to gauge the technical state or condition of a water company by measuring the unit WL per inhabitant, where high values indicate poor performance. The stakeholders in the water sector, such as water utilities, financial and regulatory agencies, and organizations that plan water resources, are urged by the AWWA to cease using and adopting percentage loss performance measures (Horbatuck & Beruvides 2024). Instead, the AWWA recommends using normalized metrics that may be calculated for a certain period, such as per capita, service connection per day, or the length of a network (AWWA 2019).

Another analysed index was the unit coefficient of the WL index per kilometre of the WDN, determined based on Equation (8), and is normally denoted as q_s. The city water company (NCWSC) WDN from the data provided, analysed and presented in Figure 3, shows a range of 24.2 m³/(h·km) in 2014 to 16.09 m³/(h·km) in 2023. This demonstrates a significant downward trend of 8.11 m³/(h·km) with uniform consistency over the last 10 years. However, the index is not within the recommended standard range because the values or ratio between the CARL and the total length of the pipelines are quite high.

Additionally, the hydraulic load index, or the unit coefficient of the WL index, computed as per Equation (9) and represented by the symbol q_o in accordance with the IWA guidelines, was examined for 10 years under evaluation, and the results are shown in Table 1. In a distribution system, the WL is correlated with the hydraulic load of the network. Therefore, WL evaluations must consider the hydraulic load of the network, in addition to the number of losses, to obtain accurate findings. It is best to compare systems with comparable loads to assess the WL in reticulation networks. Generally, less water is lost when the network is under a low strain. The reliability in relation to the acceptable or permissible requirements is noteworthy, with the hydraulic load index for the network ranging from 0.17 in 2014 to 0.13 in 2023.

Conversely, it is challenging to determine a direct correlation between WL and load owing to the assessment of the hydraulic load indices of the water provision systems, which display modest variations, as indicated in Table 1.

In addition to the current annual real WL CARL, the IWA recommends calculating the annual UARL for distribution systems. This calculation should be considered the network length and the average operating pressure, which should vary over time as NRW decreases and the number and length of connections, as per Equation (5). From a technological perspective, it is extremely impossible to completely eradicate or eliminate inevitable losses, and from an economic perspective, doing so would be unprofitable. The minimal loss that can be attained in a correctly functioning water delivery system is thought to be caused by unavoidable losses (Lambert 2009).

The analysis and comparison of the current real WLs and the technical state of the network over the last 10 years against the UARLs are presented in Figure 6. The NCWSC's total current RLs and unavoidable losses were analysed, and the results indicate that it is best to reduce WLs even though there was a notable decrease in CARL of 16,608,489.3 m³/year from 2016, when current losses were recorded at 79,655,347.3 m³/year, to 2023, and when losses were recorded at 63,046,858 m³/year, indicating a relatively stable CARL. The UARL of networks is increasing. It was 11,722,718.1 m³/year in 2014 and increased to 14,697,267 m³/year in 2023, representing an increase of 2,974,548.9 m³/year. Technically, a firm should gear up more stringent and achievable actions for further loss reduction, which will enable it to anticipate economic impacts based on the assessment and evaluation results.

The ratio of total losses to unavoidable or inevitable losses, computed as per Equation (6), is the ILI and is regarded by the IWA as a trustworthy comparison measure for a variety of networks or systems and is highly recommended for network assessment and evaluation. Its usage is recommended in cases where the network pressure is at least 25492.9 mmH₂O, there are more than 5,000 connections to a WDN, and the network density is greater than 20 connections per kilometre (Ociepa 2021). These requirements were satisfied by the assessed WDN. Table 2 shows the evaluation criteria for the ILI as per the AWWA scores, the WBI banding system, and IWA conditions.

Table 2

Categories of the assessment of NCWSC water supply systems according to the ILI (Lambert 2009)

View Large

The green lines at 2.0 and 4.0 (Figure 7) show the maximum desired limits of good and well operated water utility conditions according to the IWA, the WBI banding system, and the AWWA for developed countries and developing countries, respectively.

From the literature reviewed, the results of surveys of 16 water reticulation systems in the European countries (Austria, Belgium, Bulgaria, Denmark, England, France, Germany, Italy, Malta, Portugal, Scotland, Serbia, and Croatia) indicate an extensive range of ILI values from 0.7 to 5.8 (Dohnalik & Jędrzejewski 2004; Lambert 2009; Merc 2017). ILI values in the range of 6.444 in 2014 to 4.29 in 2023 indicate positive improvement and the need to put more effort and interventions to increase the profitability of the company through efficient pressure management, active leakage control, and network maintenance (Ferrari & Savic 2015).

The water volume losses of the NCWSC WDN and other associated parameters relevant to the quantification of losses were assessed, analysed, and evaluated over the last 10 years. The findings demonstrated that the company introduced significant strategic interventions and efforts in the operation and management of effective WL reduction. Over the period under consideration and analysis, the NCWSC has expanded its efforts to actively reduce, minimize, and manage WLs through the implementation of the 10-year strategic plan, which also coincided with the study period from 2014 to 2023. The plan places strong emphasis on the modernization and digitization of the distribution network as a key strategy through the deployment of emerging technologies and innovations to reduce or minimize WL.

There was variability in the network density of service connections, with values varying between 0.067 and 0.062. The failure intensity index value for 2014 was 1.38 failure/km/y, the WS index value was 38.98%, the Q_los index value was 76.68 dm³/(inhabitant·day), the RLB₂ index value was 1000.66 dm³/(connection·day), the NRWB index value was 45.64%, the q_s value was 24.2 m³/(km·h), the q_o value was 0.17 m³/km·days, and the ILI value was 6.444. In 2023, the failure intensity index decreased to 0.64 failure/km/y, the WS index dropped to 34.87%, the Q_los index dropped to 47.51 dm³/(inhabitant·day), the RLB₂ index reduced to 693.64 dm³/(connection·day), the NRWB index dropped to 38.86%, the q_s index decreased to 16.11 m³/(km·h), the q_o index decreased to 0.13 m³/km·days and the ILI was reduced to 4.29. Most of these indices significantly declined and were beginning to stabilize, although they were still far below the internationally accepted standards recommended by the WBI banding system, the AWWA scope, and IWA conditions.

Network monitoring greatly benefits from the deployed geographical information system (GIS) by the water utility. The use of the GIS to gather typical network data, such as pipe sizes and layouts, fitting locations, material types, and failure intensities, may be swiftly achieved with a high level of precision. It can quickly collect data describing the source and type of failure, in addition to specifics such as the diameter, construction year, and length. This makes it possible to examine the level of failure in a supply zone or specific network segments. Further significant reductions in WLs can be achieved by regularly swapping outdated and broken pipes with new ones, as well as the restoration and rehabilitation of water supply pipelines.

It is worth noting that despite the desire or aspiration of the firm to modernize the network, it has not fully embraced the use of emerging technologies and innovations in the detection and management of leakages and other forms of WL, such as the use of acoustic leakages using geophones, loggers, correlators, and contemporary stethoscopes, application of the internet of things (IoTs), artificial intelligence, machine learning, data analytics, and electroacoustic diagnostic instruments. The use of sensors for the detection of acoustic leaks to record noise from water supply networks at night or day can be of great importance in the localization of failures. In addition, the active use of GIS technology can enhance and improve the efficiency and speed of data collection, storage, and processing, in addition to enabling the construction of maps for controlled regions.

Based on the assessment, analysis, and quantification of the volumes of water lost along the distribution network over the last 10 years, it is correct to state that the water supply company is not yet fully devoted to effectively executing the best WL management interventions and best practices to address and reduce NRW and other forms of WL in their network, as most of the processes are still executed manually. Detecting and restoring leaks in a water supply system is a labour-intensive task that does not always provide the desired results if conventional approaches are applied. The company is required to implement extensive and consistent actions to decrease WL. Inconsistency and, in some cases, the total lack of data that is critical for the assessment, analysis, and quantification of WLs is a huge gap that management must address. Some of the data used in the computation of the PIs were estimates provided by the company staff, as documented records could not be traced. For instance, the average pressure in the network and the volumes of water used by the company were estimated. However, it is important to note that the research and development department at the NCWSC has made tremendous efforts to document key data and information that are critical in assessing and evaluating PIs for the water company. The company is also working towards modernizing its key operations.

Adopting action plans for the reduction of WLs and implementing them consistently over the past few years has resulted in a substantial reduction in WL, considering the figures in 2014 and the current figures in 2023. This study recommends the execution of water pressure and flow monitoring and management, as well as the systematic replacement of the majority of old pipes that are prone to failure as a significant step leading to a decrease in WLs. To precisely oversee network operations and, above all, reduce the time between the onset of a fault and its removal, it is imperative that current monitoring gadgets be installed (smart water network devices).

Partitioning the network into metered sections and analysing the losses for each area are critical tasks. By dividing the supply area into zones, it becomes possible to limit the quantity of water wasted and speed up the localization of leaks, in addition to the deployment or application of emerging technologies and innovations, such as IoT solutions and sensors, in the overall monitoring and management of the water network. The company should execute a viable and successful technique for continuously controlling the pressure in designated locations and minimizing it to optimal levels. Based on data from the water utility, the company's key operations over the past 10 years, to reduce apparent and real WLs to some extent, have produced desirable results, as some of these initiatives have shown promising outcomes in terms of reducing WLs.

In conclusion, based on the evidence presented, it is clear that urgent action is needed to reduce and manage the high levels of NRW experienced by the company and further investigate some of the strategies put in place. Consequently, the final analysis based on the findings of the study shows that the assessment and quantification of NCWSC WL indices and water balance from 2014 to 2023 have made it possible to better understand the patterns and trends pertaining to WL in the company's WDN. This is novel because it is the first study conducted on this network that uses WL indices to evaluate the technical state of the distribution system.

The results show that the WL of the network is much greater than that recommended by the IWA guidelines. This may be due to the inefficiency and ineffectiveness of the utility management's WL reduction actions and procedures, particularly in the region (SSA). Based on the strict recommendations of the IWA standards, the ILI rating indicates that the NCWSC distribution system is rated as unacceptable and inadmissible. The WDN's ILI indicates a satisfactory technical state and condition under both the WBI banding system for developing nations and AWWA standards. The specific dynamics of these indices were found to be significantly higher than the ILI levels recommended by the AWWA, the WBI banding system, and the IWA. The accuracy of WL evaluations and calculations, and thus the efficacy of the analyses, is seriously threatened by data ambiguity and unavailability. The findings indicate that the water company does not utilize the ILI or other indices to measure its performance.

An analysis of the water balance and WL indices over an extended period is required for a more factual, trustworthy, and legitimate quantification and determination of the amount of WL. Evaluations that are predicated on outcomes from a single year are highly deceptive, as well as taking into account the fact that low NRWB and WS indices do not necessarily signify acceptable or perfect network conditions. In addition, the computed values of UARL and ILI were greatly affected by inaccurate mean operating pressure estimation by the water utility, as no proper record was available. This causes underestimation and overestimation of the ILI owing to the overestimation or underestimation of the UARL. It is more prudent to develop a uniform approach for calculating the mean operating pressure in any WDN. When UARL is overestimated and the ILI is underestimated, the service connection length is overestimated. Therefore, it is vital to know whether a water supply provider has recorded the length of service connections from the property line to the metre or from the mains to the metre.

Finally, estimating the mean pressures in extensive WDNs, such as NCWSC, can be difficult. Consequently, it is advised that the network be divided into smaller district metres and/or pressure control areas and that the ILI be determined for each of these distinct zones. In WDN where the ILI in developing countries (Kenya) is higher than 8.0 (4.0 < ILI ≤ 8.0) as per the WBI banding or 5.0 (3.0 < ILI ≤ 5.0) as per the AWWA condition, or 3.5 (ILI ≥ 3.5) as per the IWA scoring, then further WL control interventions be urgently implemented and this is confirmed by the values obtained from the study which are far above the IWA recommended ones (inadmissible), even though satisfactory to both AWWA and WBI banding system scopes.

The scope of this study was limited to the NCWSC in Nairobi, Kenya. It is important to undertake more in-depth and comparative studies on the assessment, analysis, and quantification of WLs in a WDN, especially for water companies in the SSA region, as most studies have only been conducted in the global north.

We are grateful to the following: Nairobi City Water and Sewerage Company, Kenya; National Commission for Science Technology and Innovation, Kenya; and the University of Johannesburg, South Africa for providing the necessary information, support, productive discussion, guidance, and a conducive environment during the study period.

P.O. conceptualized and conducted the research. S.R. and M.D. supervised and co-wrote the manuscript. P.O. conducted field data gathering, synthesized, and analysed all related data and wrote the original draft. S.R. and M.D. reviewed and edited the manuscript and were the research supervisors. All authors discussed the results and contributed to the preparation of the manuscript.

This study was partially supported by the Artificial Intelligence for Development (AI4D) Africa Scholarship grant from the International Development Research Centre (IDRC) and the Swedish International Development Cooperation Agency (SIDA), administered through the African Centre for Technology Studies (ACTS).

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

The authors declare there is no conflict.