A review of nonrevenue water assessment software tools

Abstract Several software tools are available that can assess the performance of nonrevenue water (NRW) in water distribution networks and plan for reduction measures. Of the 21 tools that have been reported in the literature, 12 are freely available. The creation of these many tools and different versions of each individual tool indicates the promising future of NRW software development. This review comprises 12 freely available tools for water balance establishment, NRW performance assessment, and NRW reduction planning. Most of the tools have been developed to establish standard annual water balances and recommended performance indicators (PIs) for the entire network. Some tools have been developed to intervene and reduce the leakage in a district metered area. Key features increasingly being included in NRW software include uncertainty analysis, recognition of supply intermittency, and accommodation of a guidance matrix and benchmarks. Leakage assessment is fully recognized, and leakage reduction analyses are increasingly growing in the software tools. However, much less attention has been paid to assessing and options for reducing apparent losses. Although a comprehensive NRW management tool for monitoring, planning, and intervention is not currently available, developing a comprehensive tool is worthwhile, in the form of one package or a kit of smaller tools. Toward this goal, the article provides insights and recommendations addressing topics of intermittency, normalization, multi‐method assessment, planning for the reduction of apparent and real losses, and estimation of the economic level of water loss. This article is categorized under: Engineering Water > Planning Water Engineering Water > Methods

SigmaLite (ITA, 2000), Leaks suite (Lambert, 2015b), Prototype (Halfawy & Hunaidi, 2008), NAIS (Heydenreich & Kreft, 2004), Netbase (Netbase, 2019), and ÖVGW spreadsheet (ÖVGW, 2009). The 12 freely available tools have been designed for water loss assessment, water balance establishment, and NRW PIs evaluation. Table 1 presents these tools and their approaches, which were developed in response to the establishment of standard terminology, standard water balance (WB) methods (Lambert & Hirner, 2000), and recommended NRW PIs (Alegre, Hirner, & Baptista, 2000). The main functionalities of the tools are: (a) the use of a top-down water audit to estimate or assume AL, from which RL and NRW PIs are calculated. The top-down water balance is usually conducted for a period of 1 year and encompasses the whole system (i.e., global); (b) the assessment of the RL based on the bottom-up approach using minimum night flow (MNF) analysis in a DMA; and (c) the use of the burst and background estimates (BABE) for the whole network and in a DMA-scale. Only one tool-component analysis-analyses different leakage reduction options. Table 1 summarizes these tools and their finicalities. Brief descriptions of each of these tools are also provided in a Supporting Information of this article. Lambert and Hirner (2000) suggested the standard terminology for a standard water balance in water distribution networks (Figure 1). Deducting the volume of billed consumption (BC) from the system input volume (SIV) gives the volume of the NRW. Deducting the volume of the UAC from NRW gives the volume of water loss. Breaking down water loss into AL and RL involves four methods. The top-down methods start by estimating the volume of AL and then calculating the volume of RL. The bottom-up methods analyze the leakage volume based on field measurements or available records. In the top-down water audit, the AL components are estimated. To determine customer meter inaccuracies, a representative sample is tested in the laboratory at different flow rates that represent the field conditions (Arregui, Cabrera, Cobacho, & García-Serra, 2006;Arregui, Cabrera Jr, & Cobacho, 2007;Walter, Mastaller, & Klingel, 2018). Data handling and billing errors are estimated by investigating historical billing records and trends (AWWA, 2016;Mutikanga, Sharma, & Vairavamoorthy, 2011). Estimating the amount of unauthorized use is challenging, and therefore it is commonly assumed arbitrarily AWWA, 2016;Klingel & Knobloch, 2015;Mutikanga et al., 2011;Seago, Bhagwan, & McKenzie, 2004;Vermersch et al., 2016). After estimating the components of AL, RL can be calculated. Afterward, the International Water Association (IWA) standard water balance in Figure 1 can be established. Another top-down method is the water and wastewater balance method (AL-Washali et al., 2018), which assumes that AL enters the sewer network. Analyzing the WWTP inflows and comparing it to the BC enables the estimation of the volume of AL, from which RL are then calculated (AL-Washali et al., 2018;Al-Washali et al., 2020). These calculations establish the IWA water balance, after which best practice NRW PIs can be calculated for target monitoring and leakage benchmarking (Alegre et al., 2000;Alegre et al., 2016). Table 2 shows the recommended key PIs of NRW. Historically, the fundamental indicator for monitoring and benchmarking NRW was presenting NRW as a percentage of the SIV, using Equation (1). Nevertheless, consistent feedback from field data revealed that relying on this indicator for monitoring and benchmarking NRW progress is rather misleading. This is because it is strongly influenced by water consumption (Lambert et al., 2014), favors less water supply over more supplied water (AL-Washali, Sharma, AL-Nozaily, Haidera, & Kennedy, 2019), a zero-sum indicator for BC, NRW, and SIV (Lambert, 2019), and because when used, the denominator in the first part of Equation (1) should be a cause of change in the numerator (Alegre et al., 2016). Tackling this problem, the PIs in Table 2 were proposed, to give meaningful input and inform perception about the status of NRW progress. The units of the PIs in Table 2 indicate vividly the intuition of each indicator and how it should be calculated. For the infrastructure leakage index (ILI), it can be calculated based on Equations (2) and (3) (Lambert et al., 2014).

| NRW ASSESSMENT
F I G U R E 1 IWA Standard water balance (Lambert & Hirner, 2000). IWA, International Water Association UARL Litres serv:conn: day where NRW (m 3 /year) is nonrevenue water, SIV (m 3 /year) is the system input volume, BC (m 3 /year) is the billed consumption, AL (%) is the apparent losses, RL (%) is the real losses, UAC (%) is the unbilled authorized consumption, ILI (unitless) is the infrastructure leakage index, CARL is the current annual RL (m 3 /year), UARL (m 3 /year) is the unavoidable annual RL, N c is the number of service connections, L m is the length of mains in km, L p is the length of private underground connection pipe in m between the edge of the street and the customer meters, and P ave is the network's average operating pressure. On the other hand, the bottom-up methods are only for estimating the RL. MNF analysis is carried out for a DMA during night time when most customers are inactive. Flow and pressure measurements are analyzed and night flows should indicate the volume of the RL (Farley & Trow, 2003;Puust et al., 2010). The leakage rate can be estimated through this method using Equations (4)-(6) (Alkasseh, Adlan, Abustan, Aziz, & Hanif, 2013;Farley & Trow, 2003). Notably, this method can only be carried out for a DMA and scaling it up for the entire network is very uncertain .
where L DMA@t-MNF is the leakage rate in the DMA (m 3 /hr) at the time hour of MNF, Q MNF is the minimum flow rate, Q LNC is the legitimate night consumption in the DMA at the MNF time, Q DMA-daily is the daily leakage in the DMA, F ND is the night day factor, and i is hours of the day. After the RL are estimated through one of the above methods, it can be broken down into its subcomponents, using the BABE analysis (Lambert, 1994). Although most RL are avoidable, some are unavoidable, even in a new and well- constructed network. Background leaks may be too small to detect by the available detection technology. In contrast, bursts are big enough to be reported for repair by customers or by the utility crew. Unreported bursts are usually detected by the leakage detection surveys (i.e., the active leakage control-ALC; AWWA, 2016; Puust et al., 2010). Unavoidable annual RL (UARL) can be estimated using a recommended empirical equation presented in Equation (3) (Lambert et al., 2014;Lambert, Brown, Takizawa, & Weimer, 1999). The BABE analysis is useful because it enables water utilities to understand the nature of RL and plan reduction measures. Finally, because the water balance is associated with uncertainties, it is usually accompanied by uncertainty analysis (Lambert et al., 2014;Thornton, Sturm, & Kunkel, 2008). The uncertainties of the water balance can be calculated straightforward using the error propagation theory (Taylor, 1997). As the water balance problem is a process of adding and subtracting, the general equation of the error propagation theory in Equation (7) can be simplified as in Equation (8). The error propagation analysis is simple and sufficient for the water balance problem. It produces the same results with other advanced methods (Al-Washali et al., 2020) such as Monte Carlo simulation (Rubinstein & Kroese, 2016). However, some of the tools do use the variance analysis based on the statistical principles of the root-meansquare method for the normally distributed data (Thornton et al., 2008), which generates the same uncertainties. In this case, the higher the variance of the water balance component, the more significant its uncertainty becomes. The variance for each water balance component can be calculated using Equation (9) for Gaussian distribution whose curve density is represented by Equation (10) (Thornton et al., 2008).
where X and Y are independent and measurable quantities that are used to obtain a value of a calculated quantity Z; δZ/δ is the partial derivative of the variable Z with respect to an independent parameter (X or Y), and ΔX and ΔY are the uncertainties of the variables X and Y.
where σ 2 is the variance, Q is the amount of the water balance component (m 3 /year), Z is the 95% confidence limit, μ is the mean, and σ is the standard deviation.

| TOOLS FOR WATER BALANCE ESTABLISHMENT
Tables 1 and 3 shows that nine tools are basically water balance tools: AquaLite, AWWA Water Audit, BenchLeak, BenchLoss, CalcuLEAKator, CheckCalcs, Component Analysis, EasyCalc, and WB-PI Calc-UTH. The main focus of these tools is to establish the standard water balance and NRW PIs. Basic system data such as number of service connections, mains, and pressure data are input as well as the water balance data, and the main output is the standard water balance and NRW PIs. However, some tools have more or deeper features. EasyCalc remains the most detailed (Table 4), straightforward and comprehensive tool for the water balance establishment. It contains detailed input for system data, pressure data, water balance data, a historical comparison of water balances, and brief what-if scenarios. CheckCalcs, AquaLite, and BenchLoss come next in tolerating essential details about a particular case study. AWWA Water Audit and Component Analysis are tools that are more standardized for water utilities in USA and North American countries, where the input of key figures and water balance components are briefly condensed in a sole or limited input. The limited input has eventually an impact on the sensitivity and the accuracy of the tool. However, the tools (AWWA Water Audit and Component Analysis) have complementary mini-tools for data report, collection and validation. Similarly, WB-PI Calc-UTH and BenchLeak are tools that are locally focused. BenchLeak is one of the first water balance tools and now, is generally outdated and  substituted by AquaLite. While all the water balance tools use only the recommended IWA PIs for NRW, WB-PI Calc-UTH has 170 PIs that cover broad aspects of the water service in general. It also has a unique feature of recognizing the overbilling practice, which overestimates the BC and subsequently underestimates the NRW. This is the impact of charging a customer a minimum BC (e.g., 10 m 3 /month) even though a customer does not consume this amount. The impact of this overbilling practice occurs only if the billing system cannot charge the monetary minimum consumption bill unless the volumetric real data in the billing system is altered (manually or automatically; AWWA, 2016). This process is rather destructive for essential costly data of a water utility and causes avoidable opacity of the utility performance. However, when this practice exists in a water utility, the overbilling should be considered in the standard water balance itself and this is the intuition of the WB-PI Calc-UTH tool (Tsitsifli & Kanakoudis, 2010).

| Intermittency and normalization
Another key issue is the applicability of these tools in intermittent water supply systems, where water is not available in the network 24/7 and customers adapt to this situation by setting up local storage tanks in their premises. In such a case, the volume of water loss is highly influenced by the volume of the supplied water. The greater is the volume of the water supplied into the network, the more will be the volume of water losses and the higher will be the PIs of NRW, indicating worse performance while it is not necessarily the case. To tackle this issue, the volume of water loss and its PIs have to be normalized and adjusted as if the supply is continuous (AL-Washali, Sharma, . This normalization process enables monitoring and benchmarking the performance of water loss management in intermittent supply, which is an issue of increasing interest. However, only four tools recognize the intermittency in the tools' input: AquaLite, BenchLeak, CheckCalcs, and EasyCalc, where AquaLite and EasyCalc are relatively more detailed. Although normalizing the volume of NRW, AL and its PIs for monitoring and benchmarking is intuitive and critical, it is still a recent highlight that is not considered yet in the four tools. These tools normalize only the RL, explicitly in EasyCalc and AquaLite, and implicitly in CheckCalcs.

| Uncertainty analysis
Introducing the uncertainties of the water balance components to the users of the tools is a significant achievement in raising the awareness of the water balance limitations and the practical way to improve them. The uncertainty analysis points out which input data should be more verified in order to minimize the uncertainty of the interesting output. AquaLite, BenchLoss, CalcuLEAKator, and EasyCalc use the variance analysis (Equation 9) for identifying the uncertainty of the water loss components. AquaLite has this feature for the water balance components but also for NRW PIs, which is an important gesture of AquaLite. CheckCalcs use the uncertainty analysis, with a similar approach to error propagation theory, to generate uncertainties for identifying the opportunity of pressure management and its influence on reducing the leaks, the bursts, and the consumption (N 1 , N 2 , and N 3 , respectively). It is, however, a key limitation for CheckCalcs and the remaining tools that they do not recognize the uncertainties of the water balance because the water balance is critically influencing all aspects of water loss management intervention and has a great implication on the estimated benefits of each intervention (AL-Washali, . It should be noted that CheckCalcs is just one free tool of many commercial tools that form one package (LeaksSuit) for leakage management. Relevant to the uncertainty analysis is the use of validation score for the input and output of the tool, to determine the validity and reliability of the tool's output. This is a unique feature of the AWWA Water Audit whose validity score triggers changes in data acquisition rules when a low validity score is recorded. It is equivocal why AWWA Water Audit incorporates the qualitative validity score approach instead of the commonly used uncertainty analysis. Al-Washali et al. (2020) found that uncertainty analysis helps to improve the outputs of water loss assessment methods, although it did not demonstrate the accuracy or the validity of the methods. Even so, using the uncertainty analysis to improve the output of the tools is not questionable and strongly recommended (Alegre et al., 2016;AWWA, 2016;Lambert et al., 2014).

| Water loss assessment approach
The approach used to establish the water balance in all the above nine tools, except CalcuLEAKator, is the top-down water audit methodology, where AL are estimated and then RL are calculated. For CalcuLEAKator, the approach used is the MNF analysis (Equations 4-6) for one specific DMA and then water balance and NRW PIs are generated for this particular DMA. The tool enables data entry and water balances for 20 DMAs and based on these DMA mini-balances, a global water balance, and NRW PIs for the whole network are compiled and created.
As can be noticed from the above description, the top-down water audit is the main approach used to establish the water balance. The MNF analysis and BABE are usually used as complementary analyses for the top-down water audit. However, using more than a method for establishing the water balance for the entire network is recommended , because it can improve the accuracy of the tool significantly and assist in establishing more reliable and system-wide balances. For the DMA-scale, MNF analysis remains a powerful methodology to establish the water balance in DMAs.

| TOOLS FOR WATER LOSS REDUCTION PLANNING
Out of the 12 available tools, many tools touch on several aspects of planning for water loss reduction for the whole network: five tools provide guidance matrix for leakage reduction intervention, two tools accommodate economic analysis, and three tools indicate the opportunity of global pressure management.

| Guidance matrices
The common guidance matrix is shown in Table 5. The matrix was developed for the World Bank Institute as a target matrix and a banding system for leakage performance categories. The limits of categories for low and mid-income countries were set as twice the allowance of high-income countries (Lambert, 2015a), to set feasible targets for water utilities in low and mid-income countries. Having the volumetric leakage level or through the ILI, the leakage category of a certain utility is easily defined in Table 5. Based on the categories A1, A2, B, C, and D, different recommendations are provided (Liemberger and Partners, 2018): A1: small potential for further NRW reductions; A2: further NRW reduction may be uneconomic unless there are water shortages or very high water tariffs; B: potential for marked improvements; establish a water balance, consider pressure management, ALC, better network maintenance, improve customer meter management, review meter reading, data handling and billing processed and identify improvement potentials; C: poor NRW record; tolerable only if water is plentiful and cheap; even then, analyze level and causes of NRW and intensify NRW reduction efforts; and D: Highly inefficient; a comprehensive NRW reduction program is imperative and high-priority.
The matrix presented in Table 5 is provided in EasyCalc, BenchLoss, and CalcuLEAKator. EasyCalc provides this matrix but also another similar matrix for the total volume of NRW. CheckCalcs has a similar leakage matrix but with splitting the B, C, and D categories into two sub-categories for each category, following practices in Australia and Malaysia, with exactly the same approach of A1 and A2 in Table 5. AWWA Water Audit has, in turn, preset targets for the ILI and coupled with technical and financial considerations. These matrices are useful and commonly applied, however, they are developed based on mere experience, not a scientific foundation nor published materials with deeply studied data.

| Economic leakage detection
EconoLeak estimates the total volume of the leakage, background losses, and their costs. Afterward, it aggregates the leakage reduction cost in terms of sounding cost, leak correlation cost, MNF cost, repair cost, and finally administration and supervision cost. A simple cost-benefit analysis enables plotting of the curve of the short-run ELL or the economic leakage detection. Figure 2 represents the main output of EconoLeak, where the x-axis represents the leakage level and also the number of days required to survey the whole network using the leakage detection techniques. The ALC curve (green curve) shows that the detection survey becomes dramatically costly when the survey period of the whole system is less than 1 year and becomes more economic when it is more than a year . The lowest point in the total cost curve (pink curve) in Figure 2 corresponds to the economic survey period, which is almost annually in this example. The dotted vertical lines represent the base, economic and unavoidable levels of leakage of this example system. Any further leakage reduction after the economic (accepted) threshold in Figure 2 becomes basically uneconomic.
Obviously, this approach requires extensive cost data for all single elements of intervention. The tool in fact does not create the (long-run) economic level of the leakage but only the economic intervention frequency of the leakage detection survey. It does not consider other interventions such as pressure management, assets management and response time minimization. The economic level of the leakage should be determined based on all possible intervention for a specific case and not only the leakage detection survey. Alternatively, the tool Component Analysis has a more matured feature of defining the economic intervention frequency. It defines how frequent should the leakage detection F I G U R E 2 Economic level of leakage detection. Source: EconoLeak software v1.a survey complete the entire network. This is estimated based on the variable cost of water, intervention cost, and also the rate of rise of unreported leakage, as shown in Equations (11) and (12) (Lambert & Fantozzi, 2005).
where EIF (months) is the economic intervention frequency through the leakage detection surveys, CI is intervention cost ($/Km), CV is variable cost ($/m 3 ), RR is the rate of rise of unreported leakage (m 3 /Km mains/day/year), and EP is economic percentage of system to be surveyed annually. The limitation of Equation (11) is the variable RR, which is sensitive and difficult to estimate. However, guidance for estimating this factor is available in Fanner and Lambert (2009). Another concern of Equation (11) is its probable overestimation of the leakage detection potential, as discussed in . Interestingly, the tool Component Analysis does not only estimates the frequency of the leakage detection survey, but also gives an opportunity in doing cost-benefit analysis of pressure management and response time minimization for the whole network. Firstly, the tool enables setting the exponent of the pressure-leakage relationship (N 1 ; Equation 13) and then the tool estimates the potential volumetric and monetary savings of pressure reduction. It is worth mentioning that the leakage discharge from pressurized pipes varies with pressure during the day as highlighted by the concept fixed and variable area discharges (FAVAD; Lambert, 2001;May, 1994), which overcomes the limitations of Torricelli equation for plastic pipes. Equation (13) presents the leakage-pressure relationship that is reconciled with intensive empirical research, using Japanese and UK data. The leakage exponent N 1 in Equation (13) varies from 0.5 for rigid pipe materials to 1.5 for variable leakage area in flexible (plastic) pipe material, whose leaks' split varies with pressure.
where L is the leakage volume, P is the pressure, and N 1 is the leakage-pressure exponent. Similarly, the Component Analysis tool estimates the potential volumetric and monetary saving of minimizing the repair-response time, based on the direct cut of the leaks' run-time. Expectedly, the tool does not recognize the economic aspect of assets management, perhaps because estimating the benefits of the assets management is complex and it is typically not cost-effective. This tool is probably the most economically comprehensive tool for leakage management that is freely available. Example of the output of this tool is presented in Al-Washali et al. (2020) and .

| Global pressure management opportunity
There are many network simulation and hydraulic modeling tools that consider pressure management and pressure control for water networks. This is basically based on the concept that optimizing the pressure in the network triggers significant leakage reduction. Incorporating pressure management with DMA demarcation is commonly utilized by such tools. An example of these tools is the freely available tool, EPANET and its possible add-ins modules. Several (commercial) tools are also available, including: WaterGYMS, InfoWater, WDNetXL, H 2 O MAPWater, KYPIPE, and other tools. However, there are three freely available tools that snapshot the pressure management opportunity for the whole (global) network and estimate the feasibility of pressure management without the need of hydraulic simulation of the network pipes and appurtenances, which is the focus of this review. These tools are: CheckCalcs, Component Analysis, and EasyCalc. The principle is similar and intuitive. Based on defining the exponent of the leakage-pressure relationship (N 1 ) in Equation (13), the reduction on the leakage volume as a consequence of reduced pressure can be estimated, and then its monetary value can be easily derived. This is what EasyCalc exactly does. However, CheckCalcs further analyses the impact of pressure reduction on the reduction of bursts frequencies as well as the customers' consumption (N 2 and N 3 , respectively). The Component Analysis, however, enables estimating the benefit of pressure management and comparing it to its cost as well as the costs and benefits of other leakage interventions.

| TOOLS FOR WATER LOSS MANAGEMENT INTERVENTION
The previous two sections tackle the assessment of water losses and planning for its management. However, when it comes to working on the ground during the intervention phase, only the tools that are focused on DMA-scale are hands-on. These tools are SanFlow and PresMac.

| Active leakage detection
Similar to CalcuLEAKator which is discussed in Section 4, SanFlow is a powerful tool that also analyses MNF in a DMA within the network using Equations (4)-(6). It is a suitable and operational tool to intervene and reduce the leaks in a DMA. It estimates the legitimate night use, the leakage and the unavoidable background losses using the BABE approach. After deducing the unavoidable background losses, the tool transforms the leakage volume for each DMA into an estimated equivalent number of bursts in the DMA. Notably, the tool does not assume the presence of bursts within each DMA, but use this as an index to rank the DMAs based on their leakage level. Comparing this equivalent number of bursts for all the DMAs assists in prioritizing the DMAs for leakage minimization interventions. SanFlow estimates the hourly leakage only during the MNF time and does not provide estimates on a daily leakage level, let alone annually. Figure 3 gives an insight into the output of SanFlow. Figure 3 implies that there was significant leakage in the DMA at the beginning (the yellow area) and after some time, when the leaks were fixed, the pressure in the DMA enhanced, causing more unavoidable background leakage in the DMA that cannot be sensed by the available leakage detection technology. So the blue line, which shows the estimated equivalent service pipe bursts, represents the priority level of intervention in this DMA, and can be compared with other DMAs in the network which is a main function in this tool.

| Zonal pressure management
The potential for pressure management is consistently promising. Reducing the pressure at a critical point in a DMA by a certain level triggers a more pressure reduction at the inlet of the DMA (Mckenzie & Langenhoven, 2001). PresMac, therefore, assesses the monetary savings of pressure reduction due to installing fixed-outlet and time-modulated PRVs in a specific DMA. The tool does not rely on hydraulic representation of the DMA's network, rather, it compromises MNF analysis, BABE analysis, N 1 exponent estimation, and further analyzing pressure measurements, friction factors (K) and F I G U R E 3 Leakage, background losses, and night use in an example DMA. Source: SanFlow v4.6. DMA, district metered area head losses (H L ) in three key points in the DMA: the inlet point, the average (elevation) zone point, and the critical point(s) in the DMA whose pressure is the lowest in the DMA during the course of the day. The fixed-outlet PRV dictates a fixed pressure during all the hours of the day (Figure 4b). The time-modulated controller reduces the outlet pressure at certain times of the day when the demand is basically reduced (Figure 4c). In a critical point modulation, the pressure is sensed at the critical point and communicated to the PRV at the inlet which, in turn, adjusts the inlet pressure to maintain the minimum pressure at the critical point during the course of the day. This process is effective, but further investment is required. In PresMac, the flow-modulated controller (Figure 4d) dictates the inlet pressure in accordance with the instantaneous demand and the excessive pressure at the critical point in the DMA. While firefighting flows cannot be met using the time-modulated controller, it can be satisfied using the flow-modulated controller. The use of PRVs optimizes the pressure in the DMA and achieve the minimum leakage. PresMac, however, analyses only the benefits of fixed-outlet and time-modulated PRVs. If the less expensive time-modulated controller is economically justified, the flow-modulated controller can provide greater benefits. A limitation of pressure reduction is the potentially small reduction in customer demand. In this regard, PresMac assumes that leaks are pressure-dependent and consumption is pressure-independent, although some consumption is in fact pressure-dependent, such as washing hands, brushing teeth and garden irrigation. The approach used in PresMac is iterative. Firstly, using MNF measurements and BABE analysis, the tool estimates the background leaks and solves Equation (13) to estimates the value of N 1 as shown in Equation (14). Secondly, the tool splits the inflow into the DMA into pressure-dependent and pressure-independent flows. Then, the tool estimates the friction factor (K) using Equation (15) where K is head loss coefficient (m −5 hr 2 ), H L is head loss in m, and Q is flow in (m 3 /hr). The tool then selects a fixed-outlet pressure and recalculates pressure in the average zone point and the critical point using K factors and the corresponding zone inflow. This process is iterative until the minimum acceptable pressure at the critical point is achieved. This process is carried out for the fixed-outlet PRV which saves significant leakage volume. Nonetheless, the achieved minimum acceptable pressure at the critical point is only for the peak hour and is still higher than the required pressure during the remaining hours of the day, when the demand is lower. Therefore, the timemodulated controller can further reduce the leakage by changing the times and switching from high to low-pressure periods during the day, and the benefits of this option are particularly analyzed in PresMac. The benefits of adopting flow-modulated PRV in the DMA are unfortunately not provided in PresMac. Finally, it is worth mentioning that recent researches on N 1 exponent (Lambert, Fantozzi, & Shepherd, 2017;Van Zyl & Cassa, 2014) revealed that N 1 itself is affected by the changing pressure during the day and this particular issue is still not considered in PresMac. Table 3 summarizes the modules and gaps for each individual software tool. The functions of these tools fall into the above three categories: (a) water balance establishment; (b) water loss reduction planning; and (c) water loss management intervention. Although asset management is a main pillar of RL reduction, it is still not recognized in the NRW reduction tools. Besides, while AL is a major concern in low-and mid-income countries, it is not sufficiently recognized in the input data for most of the tools. Providing more detailed input for AL results in more accurate results. Table 4 shows the level of details of the tools in terms of AL and other key parameters that affect the accuracy of the water balance. Table 4 shows that only one tool, the WB-EasyCalc, has a sufficiently detailed module for AL. Regarding AL reduction, Table 3 shows that AL reduction planning has garnered relatively little attention among software developers, which is a major limitation in the industry in general. A good start in this regard would consist of developing a tool or a module to optimize policies for customer meter replacement in the distribution network. It is worth to highlight that the reviewed tools are mainly developed by practitioners in the field, aiming to be applicable in real-world and for different network sizes. The reviewed tools use hydraulic analysis and hydraulic modeling to a certain extent, however, they do not incorporate hydraulic representation of the network's pipes and appurtenances neither for the whole network nor for a DMA in the network. Figure 5 presents a fit-for-purpose guide for selecting an NRW software tool to establish the water balance and generate NRW PIs. When using the top-down water balance, if the water supply is intermittent, then three tools can be used: WB-EasyCalc, Check-Calc, and Aqualite. If the water supply is continuous and overbilling of consumption is of concern, then WB-PI Calc-UTH can be used. Otherwise, Water Audit, WB-EasyCalc, Check-Calc, Aqualite, and BenchLoss can be used. When the water balance needs to be established using MNF, then CalcuLEAKator can be used. Similarly, Figure 6 presents a fit-for-purpose guide for selecting an NRW software tool when assessing water loss to plan for reduction measures. Planning for RL reduction requires conducting a BABE analysis (i.e. CAL). Planning for RL reduction can be accomplished using two scales: the whole network or within a DMA. For the whole network, a BABE analysis can be conducted using the tool Component Analysis. To analyze the potential for leakage reduction when carrying out pressure management interventions, the tools Component Analysis, Check-Calc, and WB-EasyCalc can be used. To analyze the ELL, EconoLeak can be used. To analyze the potential for leakage reduction when minimizing the response times for reported bursts or when carrying out active leakage detection, the Component Analysis tool can be used. To plan for RL reduction on a DMA-scale, Sanflow is first used to conduct a BABE analysis. Afterward, Sanflow is again used to prioritize between DMAs for active leakage detection and then PresMac can be used for pressure management analysis. Unfortunately, planning for AL reduction is still missing and tools need to be developed for this purpose. The availability of several NRW software tools and many iterations of each individual tool indicate the promising future of NRW software development. This article reviewed 12 freely available NRW software tools for assessing NRW in distribution networks. The tools use hydraulic analysis in their modules, but they do not hydraulically model a representation of the network's pipes and appurtenances. Most (nine) of the tools have been developed to establish the standard annual water balance and NRW recommended PIs for the whole network. Some (three) tools, however, have been developed to intervene and reduce leakage within a DMA.

| GUIDANCE FOR THE USE OF NRW SOFTWARE TOOLS
The main aspects of the software tools that are currently improving or need to be improved further in the software tools are: uncertainty analysis, consideration of supply intermittency, and loss reduction analyses. Although the importance of uncertainty analysis is widely recognized in the industry, only five tools have a module for uncertainty analysis. Intermittency of the water supply is a subject of increasing interest, yet it is considered in only four tools. Whenever intermittency is of concern, normalizing the PIs to a continuous supply is necessary, and this has been included in only four tools (which consider only real losses). Normalizing the NRW and apparent losses remain unaddressed. To plan for leakage reduction, the tool Component Analysis Model is relatively comprehensive as it analyses the potential benefits of pressure management, active leakage detection, and minimizing response and repair time of bursts. However, apparent losses have still not been adequately recognized in the industry.
Although a comprehensive NRW management tool for monitoring, planning, and intervention is currently not available, such a tool can be developed in one complete package or in a kit of several tools. The modules presented in Table 3 should be included in such a tool. However, certain critical aspects should be emphasized. Firstly, recognizing the intermittency of the supply will widen the use of NRW software tools. For this normalizing all the volumes and PIs of the NRW, apparent losses, and real losses is essential. Currently, normalization is only considered for real losses and not for apparent losses and NRW. Secondly, distinguishing the scales of the tool is crucial. Planning for loss reduction should first be performed for the whole network; then planning for specific interventions should be done on a DMA-scale. In both scales, a comprehensive tool that considers different reduction measures for both real and apparent losses should be useful and effective. Thirdly, greater focus needs to be placed on the apparent loss estimation and minimization; the reduction of apparent losses is cost-effective and it is relevant to, and considered a priority for, many water utilities in low-and mid-income countries. Fourthly, a comprehensive tool for NRW would definitely benefit from including capabilities for uncertainty analysis, guidance matrix and assessing NRW components using different methods. Finally, NRW software tools should include the capability to determine the economic level of water loss based on estimating and combining the economic levels of leakage and apparent losses.

ACKNOWLEDGMENTS
Developers of the reviewed tools are herewith acknowledged for making them available. This work was funded by NICHE 027YE Project: The Dutch Organisation for Internationalisation in Education (NUFFIC), Water and Environment Centre in Sana'a University-Yemen, and MetaMeta Research, The Netherlands.

CONFLICT OF INTEREST
The authors have declared no conflicts of interest for this article.

RESEARCH RESOURCES
The reviewed tools for nonrevenue water management.

ORCID
Taha M. AL-Washali https://orcid.org/0000-0003-2113-8837 RELATED WIRES ARTICLES Leak Detection and Water Loss Control Leak detection in pipeline networks using low-profile piezoceramic transducers Arriving at the Economic Level of Leakage: Environmental Aspects A Realistic Basis for an International Comparison of Real Losses from Public Water-Supply Systems