Using conventional methods and tools for determining detailed zonal water balances and performance indicators involves considerable efforts in manual data collection and processing. Hence, water balances mostly are determined for a whole network only. Unmetered components are often neglected or based on rough estimates. This article presents an approach to the automated determination of zonal water balances and its implementation as a software tool. The approach is demonstrated for the practical case of the Pforzheim water utility (Stadtwerke Pforzheim GmbH & Co. KG).

INTRODUCTION

Compilation of a detailed zonal water balance is a prerequisite in efficient planning and implementation of steps of water loss management (US EPA 2010). Water balances are calculated to determine how much water is lost in a distribution system. For this purpose, system input and the different components of water consumption in the period under consideration are broken down for a system (Puust et al. 2010). An appropriate basis is provided by the standardized water balance published by Lambert & Hirner (2000). This balance is applied by numerous water utilities worldwide and has been adopted in the guidelines of many national water associations; see, e.g., DVGW (2003), AWWA (2009), ÖVGW (2009), and Lambert & Taylor (2010).

Lambert & Hirner (2000) as well as the guidelines referred to above recommend a balancing period of 12 months or less. They also recommend that the water balance ideally be established not only for the complete system, but separately for individual supply zones. This allows for the identification of zones with a higher priority of implementing countermeasures. The results of the balance are then used to establish performance indicators (PIs), which allow water losses to be compared and the decisive causes to be analysed (Farley & Liemberger 2005). Establishing water balances and determining PIs will be summarised below by the term of water loss analysis.

However, there are no standardised approaches to determining those components which cannot be measured, such as water used to flush pipes or fight fires. As a consequence, those components are taken into account either not at all or only on the basis of lump sum values (Klingel & Knobloch 2015). Moreover, only one water balance and PIs for the entire system are established in most cases, due to the amount of work required for manual data acquisition in a variety of IT systems and their allocation to the respective zones. Klingel & Knobloch (2015) also found that most software tools available to support the generation of water balances require the users to collect, process, and enter the necessary data manually. Consequently, it is mainly the absence of proper approaches to determining unmeasured components and the large expenditure that restrict zonal determination of complete water balances and PIs.

This paper describes an approach to the automated calculation of water balances and PIs. Automated data acquisition and calculation reduce the expenditure involved and support the establishment of zonal water loss analysis and analyses covering shorter periods of time. Moreover, methods to determine unmeasured components were developed and incorporated into the approach in order to allow complete water balances to be established. To determine the amount of real losses attributable to undetected leakage and to reported and repaired leaks, as recommended by Liemberger & Farley (2004) and several other authors, like Lambert & Taylor (2010) and Charalambous & Hamilton (2012), a method has been developed and integrated, which analyses the minimum night flow (MNF) and information about defect reports. This makes the method a combination of component analysis (Lambert 1994) and bottom-up assessment (Farley & Trow 2003; AWWA 2009).

The approach to the automated calculation of the water balance was already outlined by Knobloch et al. (2014). In the following section, it is described in greater detail together with the calculation of PIs. Data management is outlined briefly. In the ‘Results and discussion’ section, the implementation and application of this approach for pilot purposes is reported for the case of the water distribution system (WDS) run by the Stadtwerke Pforzheim GmbH & Co. KG (SWP). Conclusions are given in the subsequent section.

APPROACH

Data management

The central level of integration in this approach is the utility's geographic information system (GIS), which contains the network information. The input data from secondary IT systems of a utility required for water loss analysis are connected to the GIS by means of flexible interfaces. Connection is achieved by logic relations in a database system defined by means of a query stored in the database management system. Whenever there is a query, reading access to the database is established to show the current data inventory. Items which can be connected include the supervisory control and data acquisition (SCADA) system, the billing system, the meter management system, the work order tracking system (WTS), the defect database, and the hydraulic model. Isolated databases can be connected as well, such as the documentation of firefighting missions or manual meter read-outs.

The input data are checked in the GIS for quality and consistency, homogenised, and made available for further processing in an exchange database with a set structure. The tool for water loss analysis is connected to the GIS via the exchange database. Once parameterised, all the necessary input data of the exchange database are automatically accepted into the tool.

Topology

For water loss analysis, the logical interconnection of the system to be analysed must first be derived in a way relevant to the analysis. For this purpose, the WDS is converted into abstract terms and modelled by standardised components. The three compulsory components are system inputs/exports, balancing zones, and bulk meters. They can be used to model the topology of simple systems.

System inputs/exports model systems boundaries at which water is fed into the system or exported from it. Bulk meters establish connections between two other components for which measurements of the volumes fed into the system or into individual balancing zones exist. Balancing zones model a network segment for which a water balance is established. A balancing zone combines all elements of the piping system of the segment with information about pipes, valves, customers, and customer meters as well as pipe defects. Each balancing zone may be assigned various types of water consumption, such as water consumption by customers, flushing, or water for firefighting. Balancing zones are established for as small as possible hydraulically discrete network segments in order to achieve meaningful water balance results. However, balancing zone inputs and exports must be known. Hence, the number and size of the balancing zones are determined by the existing bulk meters.

Other components, such as pumps and valves, can be included into the topology to model temporary opening and closing of a connection between two balancing zones. Where applicable, missing bulk meters can be replaced by so-called virtual meters which reconstruct the associated flows from known level changes in a tank.

Figure 1 illustrates how a small supply area with three zones is translated into the necessary topology. If, for example, the bulk meters at the pressure reducing valves 28 and 27 do not have sound records, balancing zones 8 and 65 will be merged in one zone.
Figure 1

Detail of a water distribution system (left) and the topology derived (right).

Figure 1

Detail of a water distribution system (left) and the topology derived (right).

Calculation

On the basis of the modelled topology of the WDS, water losses are calculated automatically for pre-selected periods and balancing zones. Figure 2 indicates the components considered by the approach presented here. With the exception of the column on the right, the terms used are taken from the IWA water balance by Lambert & Hirner (2000), while the symbols are those used in the water balance of the DVGW (2003). The water losses determined are evaluated by means of a number of PIs. The results of the analyses are transferred to the exchange database and visualised in the GIS. The individual approaches to calculating the components and the PIs are described below.
Figure 2

Water balance components analysed.

Figure 2

Water balance components analysed.

System input QN

The first step in calculating the water balance of a balancing zone determines the system input QN. The associated bulk meters and virtual meters of the zone are identified unambiguously on the basis of topology. This allows their measured data to be read in from the exchange database. The total system input QN.1 and the system export QN.2 are used to calculate the daily system input QN and summed up over the entire period of observation.

Billed authorised consumption QAI

The input data serving to determine billed metered consumption QAI.1 originate from the billing system. GIS functionalities allow all customer meter data to be allocated to a specific balancing zone. In this way, the billed metered consumption QAI.1 in the period under review can be determined for each balancing zone. Should the utility have customers whose water consumption is not measured, but billed on the basis of a flat rate, the assumed water volume is taken into account as billed unmetered consumption QAI.2.

Unbilled authorised consumption QAN

Withdrawals at waterworks and tanks QAN.1

A certain amount of the drinking water produced is used by the utilities in waterworks and storage tanks to clean filters, tanks, and other plant components. Where the utility collects information about the date and amount of water withdrawal, these data can be written into the exchange database by way of the GIS, and the QAN.1 component can be quantified.

Withdrawals for flushing QAN.2

Pipe sections in which the water stagnates or reaches only low flow velocities are regularly flushed by the utilities. The water volumes used for flushing normally are not measured. However, for reasons of operation, the beginning and end points of flushing and the clear identification (ID) of the hydrant flushed are often recorded in the WTS. Where the system input QN is known in intervals of minutes for a zone, the input curve can be analysed over the flushing period known from the WTS in order to estimate withdrawals for flushing QAN.2. Figure 3 illustrates the analysis for the example of a single flushing event. The exact beginning and end times of flushing are identified on the basis of high positive (beginning) and negative (end) gradients of the input curve. In Figure 3, flushing starts at 03:02 p.m. and ends at 03:19 p.m. Withdrawals for flushing QAN.2 are determined approximately via the difference between the curve in the flushing period and a reference curve. The reference curve is produced from the averages of the readings on the same day of the week over the past 3 and the future 3 weeks.
Figure 3

Illustration of the determination of withdrawals for flushing QAN.2.

Figure 3

Illustration of the determination of withdrawals for flushing QAN.2.

Withdrawals for pipe works QAN.3

Construction work in the piping system is conducted either by the utility or outsourced to construction companies. Normally, construction companies are provided with meter standpipes. Water withdrawal is billed and, thus, can be taken into account in the QAI component.

Work performed in-house normally does not include metering of water withdrawal. However, information is documented in the WTS about the date, place, and type of measure. To determine the amount of water withdrawn at least approximately, the pipe section involved is automatically identified in the GIS and, on the basis of the topology of the piping system, all valves that must be shut to isolate this pipe section are determined. Afterwards, the volume of the pipes so isolated can be calculated. Water withdrawal QAN.3 is assumed to be six times the calculated volume. This takes into account emptying of the pipe section prior to the intervention and the amount of flushing water of five times the pipe content after the intervention in accordance with DVGW (2000).

Withdrawals for firefighting QAN.4

Withdrawal of water for fighting fires and for drills as a rule is not measured by the firefighters. On the basis of literature research and an inquiry among members of the firefighting forces, it is safe to assume that the amount of water used to fight small and medium-sized fires is less than 5 m3 (Korkmazer 2013). It is not possible, however, to make such a universal statement for large fires, as this category includes very different fires ranging from 5 m3 to 4,600 m3 of water (Korkmazer 2013).

As no general statements can be made, the volume of water used to fight fires QAN.4 is determined by a technique largely corresponding to the approach used to determining the amount of flushing water QAN.2. However, it is sufficient to know the input into the pipe system QN at hourly intervals, because firefighting missions involving larger fires take more than one hour as a rule. Information about the address, date, start, and end of mission are taken from the mission documents of the firefighters and included in the exchange database.

Apparent water losses QVS

Annualisation errors QVS.1

Errors in data handling result from different times of reading bulk meters and customer meters. Errors in the water balance frequently are also caused by the fact that the billed authorised consumption QAI is determined from data taken from the billing system without any further correction. However, consumption data of a customer may be incorrect, due to book transfers of higher or lower consumptions in the previous year.

In order to be able to quantify at least approximately the amount of annualisation errors QVS.1, water consumption is calculated on the basis of customer meter readings and compared with the billed metered consumption QAI.1. Two approaches described by Renaud et al. (2015) are used as a function of the level of detail of the input data available to correctly convert water consumption over the period of review on the basis of the meter readings (annualisation).

Bulk meter inaccuracies QVS.2 and customer meter inaccuracies QVS.3

Bulk meter measurements are either transmitted directly to the SCADA system or read by the operating staff of the utility at regular intervals, e.g. monthly, and then entered into the respective database. To approximately determine meter inaccuracies QVS.2, all measurements of a bulk meter are compared with the measuring range of the respective unit. The sum total of the products of each individual reading with the associated percentage meter deviation (obvious from the error curve) results in the bulk meter inaccuracy QVS.2 over the period of review. Bulk meters, for which only hourly readings or monthly manual readings are available, are not taken into account in the analysis of meter inaccuracies.

Where smart meters are used, the approach described above can also be employed to determining customer meter inaccuracies QVS.3. Alternatively, the results of random sampling of customer meters can be used to determine QVS.3.

Dripping losses QVS.4 and unauthorised consumption QVS.5

Dripping losses QVS.4 are discharges of less than the starting point of the water meter which, for that reason, are not registered. As the status of in-house installations (minimal leakages, dripping taps, dripping toilet flushing systems) has a decisive influence on the production of dripping losses, the level of QVS.4 cannot be approximated even with information about the park of water meters. Moreover, the unauthorised water withdrawal QVS.5 cannot be simulated or determined by approximation. Consequently, utility operators must use data based on experience or standard data for QVS.4 and QVS.5, as proposed, e.g., by DVGW (2003) and AWWA (2009).

Real water losses QVR

Overflow of storage tanks QVR.1

Water losses caused by overflow from storage tanks QVR.1 are quantified by summing up documented events and the associated (estimated) water losses. Alternatively, warning signals from the SCADA system are evaluated, which are transmitted by level meters in case of tank overflow. In this case, the amount of water loss QVR.1 is determined from the difference between the input and withdrawal meters of the tank for the period of overflow.

Leakage of storage tanks QVR.2

Leakages of tanks can be found in volumetric tests only. In this case, measurements must be performed for each tank compartment over a period of at least 24 hours to determine the lowering of the water level with the inlet and outlet valves closed (Farley 2001). To determine QVR.2, the leak rates per compartment as determined in tests are extrapolated to the period under consideration.

Leakage on mains QVR.3 and on service connections QVR.4 due to known failures

To determine QVR.3 and QVR.4, all defects in pipes, for which the time of repair is documented in the WTS or in the defect database, are attributed to the specific pipes by interconnection in the GIS first. In this way, it is possible to determine the defects that occurred in the mains and service connections and their locations.

If the system input is known in at least hourly intervals, the MNF curve of the zone can be determined. For this purpose, the minimum system input is determined for each day within the period of assessment between 1:00 a.m. and 5:00 a.m. Each value is rounded and compared with the previous value. In case the value is higher than the previous value, the value is also compared with the two subsequent values. The value is neglected, if those three values also differ. Hence, upward outliers caused by increased nocturnal consumption are identified and disregarded.

The resulting smoothed stepped curve is matched to the point in time at which defects were repaired, which is known from the WTS, as is shown in Figure 4. If a repair time coincides with the decline in the MNF, a connection is assumed to exist. In that case, the difference in MNF before and after repair is registered and the point of time is determined, at which the level established after repair occurred last. In this way, the run time and the flow rate of the leak are determined approximately. The total of leakage volumes arising from all defects ultimately results in the losses from known defects in mains QVR.3 and service connections QVR.4.
Figure 4

Illustration of the determination of leakage on mains QVR.3 and service connections QVR.4 due to known failures.

Figure 4

Illustration of the determination of leakage on mains QVR.3 and service connections QVR.4 due to known failures.

Leakage due to unknown failures QVR.5

The difference between the system input QN and the total sum of authorised consumption QA, apparent water losses QVS, and the components QVR.1 to QVR.4 of the real water losses indicates the level of water losses from unknown failures QVR.5.

Performance indicators

The zonal water balances are used to calculate various PIs, including specific real water losses qVR (m3/h/km) and their evaluation in accordance with DVGW (2003), and the infrastructure leakage index ILI (-), which establishes a relationship between current annual real losses CARL and systems-specific unavoidable annual real losses UARL (Lambert et al. 1999; Lambert & Hirner 2000). These PIs allow water losses to be evaluated and compared in different zones in order to set priorities for countermeasures.

RESULTS AND DISCUSSION

Case study

The WDS of the city of Pforzheim, Germany comprises 30 supply zones, 458 km of mains, 212 km of service connections, 19,725 customer meters, and 88 bulk meters. The WDS was used as a case study for pilot implementation and testing of the approach. The SCADA system, the consumption billing software, and the WTS in which also pipe defects are documented were linked to the GIS of the SWP. Moreover, a database with manual bulk meter readings and the mission documentation of the Pforzheim firefighters were linked to the GIS. A pilot area with a total of 18 supply zones (60%) was chosen, which covers 71% of the mains, 78% of the service connections, 65% of the customer meters, and 47% of the bulk meters. Sixteen balancing zones were established and analysed automatically within the pilot area. The water balances and PIs were determined for 2012 (Table 1). Selected results are shown and discussed in the sections below in order to demonstrate the functionality of the applied approaches.

Table 1

Results of the water balance for 2012 in the balancing zones of the pilot area

ZoneQN (m3)QAI (m3)QAN (m3)QVS (m3)QVR (m3)QVR (%)
176,408 185,148 292 −1,837 −7,195 −4 
204,605 178,572 334 1,260 24,439 12 
7 + 64 1,056,638 825,409 11,465 39,162 180,602 17 
8 + 65 93,375 71,290 231 1,392 20,462 22 
13 534,363 501,980 3,396 −23,502 52,489 10 
14 381,940 356,849 831 4,522 19,738 
15 51,846 62,760 363 −3,566 −7,711 −15 
16 49,935 46,165 55 1,213 2,502 
18 424,426 348,958 2,955 −20,596 93,109 22 
19 148,975 94,702 710 3,834 49,729 33 
28 45,905 50,655 71 −1,160 −3,661 −8 
31 141,496 134,931 146 −2,350 8,769 
32 64,620 56,254 40 −774 9,100 14 
58 182,642 164,977 88 −4,968 22,545 12 
60 1,464,661 1,189,343 4,746 −30,189 300,761 21 
61 95,677 35,910 104 9,109 50,554 53 
ZoneQN (m3)QAI (m3)QAN (m3)QVS (m3)QVR (m3)QVR (%)
176,408 185,148 292 −1,837 −7,195 −4 
204,605 178,572 334 1,260 24,439 12 
7 + 64 1,056,638 825,409 11,465 39,162 180,602 17 
8 + 65 93,375 71,290 231 1,392 20,462 22 
13 534,363 501,980 3,396 −23,502 52,489 10 
14 381,940 356,849 831 4,522 19,738 
15 51,846 62,760 363 −3,566 −7,711 −15 
16 49,935 46,165 55 1,213 2,502 
18 424,426 348,958 2,955 −20,596 93,109 22 
19 148,975 94,702 710 3,834 49,729 33 
28 45,905 50,655 71 −1,160 −3,661 −8 
31 141,496 134,931 146 −2,350 8,769 
32 64,620 56,254 40 −774 9,100 14 
58 182,642 164,977 88 −4,968 22,545 12 
60 1,464,661 1,189,343 4,746 −30,189 300,761 21 
61 95,677 35,910 104 9,109 50,554 53 

Water balance

Unbilled authorised consumption QAN

Analysis of the information from the WTS shows that a total of 2,003 hydrants were flushed in the pilot area in 2012. Depending on the quality of the bulk meter measurements available, between 14.3 and 76.9% of the flushing processes in the individual balancing zones were identified. The mean water withdrawal for flushings QAN.2 amounted to 2.6 m3. Problems associated with the automatic quantification of flushing volumes are attributable to major fluctuations of system input in large zones and to the lack of time synchronisation between the WTS and the SCADA system. The three major fires in the pilot area during the balancing period were identified. The withdrawal for firefighting QAN.4 was between 17 and 55 m3. On the other hand, only some 35% of medium-sized fires were identified. The results also show that water withdrawal for pipe works QAN.3 and for firefighting QAN.4 makes up a negligible proportion of unbilled authorised consumption QAN.

Apparent water losses QVS

The annualisation errors QVS.1 were below 1.0% in three zones. In the other zones, QVS.1 was between −6.9 and +3.9% of billed authorised consumption QAI. Complete meter characteristics were available for 14 bulk meters in the pilot area. Comparison with the measuring data shows that three bulk meters may have been over-dimensioned. The meter inaccuracy QVS.2 determined was between −0.1 and +1.4%. Results show that the annualisation error QVS.1 is the dominating component in apparent water losses QVS and has a significant influence on the level of the real water losses QVR determined.

Real water losses QVR

Analysis of the alarms in the SCADA system showed that there were 19 overflow events in six tanks in the pilot area over the balancing period. Their duration was between a few minutes and more than 17 hours, resulting in loss quantities QVR.1 between 0.2 and 58.3 m3. This means that total losses due to tank overflow QVR.1 amounted to approx. 0.5% of the real water losses QVR only.

As a result of the merging of MNF and defect repair data from the WTS, it was possible to automatically identify 27 of 72 defects (37.5%) in the mains and 11 of 28 defects (39%) in service connections. No identified defects may be due to two reasons: either the flow rate of the leak was very small or the leak had a very high flow rate and was localised and repaired very quickly. Thus, the repair of the defect was not associated with a lowering of the MNF. For the identified leaks, the associated leak run times and leak flow rates could be calculated. The average runtime was 118 days. On an average, leakages continued for 114 days, before they were detected and then repaired after an average of four days. The average flow rate was between 1.6 m3/h in service connections and 2.6 m3/h in mains. For the pilot area, this results in losses from known defects in mains QVR.3 of 182,545 m3 and in service connections QVR.4 of 42,211 m3. This corresponds to 3.6% and 0.8%, respectively, of the system input QN. Zonal assessment of water losses due to unknown defects QVR.5 indicated that the share of QVR.5 in the different zones turned out to be very different. On an average, the share was approx. 70% of the real water losses QVR.

Real water losses QVR of −15 to 53% of the system input QN were calculated (Table 1). The negative values in zones 5, 15, and 28 and the quite high values in zones 18, 19, 60, and 61 do not seem to be plausible. The negative values are probably due to open valves separating zones, incomplete bulk meter measurement series, or wrong allocation of consumers to zones. The high values might be due to the same reasons or the uncertainty of the input data. The uncertainty is discussed in the next section.

Performance indicators

Once the water balances have been calculated for the balancing zones, PIs are determined which, in most cases, include secondary data, such as pipe lengths for specific real water losses qVR.. Figure 5, by way of example, shows the results for the specific real water losses qVR in the balancing zones, their evaluation in accordance with DVGW (2003), and the uncertainty of the results. The uncertainties are determined by assigning specific uncertainties to the input data and considering the error propagation. Evaluation of the uncertainty associated with the PIs determined provides information about the plausibility and accuracy of the results. In the case of the specific real water losses qVR shown in Figure 5, the results obtained for zones 18 and 61 must be seen critically. The large uncertainties in these cases are due to the lack of separation between the transmission and the distribution systems, which leads to a relatively large total system input QN.1 and export QN.2 in those zones compared to the consumption QA. Thus, the deviations of the bulk meters measuring QN.1 and QN.2 have a large impact on the uncertainty of the calculated water losses QV. In zones 19 and 60, the water losses determined are unrealistically high. The high specific system input qN indicates that the measurements of the bulk meters are incorrect, leading to an overestimation of the system input QN and, hence, to high results for water losses QV. The relatively large proportion of results that are not plausible (7 of 16) is due to erroneous input data (zones 5, 15, 18, 19, and 60) and to the system structure (zones 18 and 61) and, hence, cannot be attributed to the approach presented.
Figure 5

Specific real water losses qVR in the pilot area in 2012 and associated uncertainties and evaluations in accordance with DVGW (2003).

Figure 5

Specific real water losses qVR in the pilot area in 2012 and associated uncertainties and evaluations in accordance with DVGW (2003).

CONCLUSIONS

This article describes an approach to automated water loss analysis, which has been implemented in a GIS-based tool and tested in a case study. Input data are used, which are transferred from a GIS to an exchange database to calculate water balances by zones. Unmeasured components are determined approximately by partly newly developed procedures.

For 14.3 to 76.9% of the flushing processes in the individual balancing zones, withdrawals QAN.2 were quantified with the developed procedure by subtracting the integral of a reference curve from the integral of the system input curve during the event. The main reasons for unidentified flushing processes were major fluctuations of system input in large zones and the lacking time synchronisation between the WTS and the SCADA system. All withdrawals for fighting major fires were successfully quantified, but only 35% of medium-sized fires were identified. However, withdrawals for firefighting QAN.4 and for pipe works QAN.3 seem to make up a negligible proportion of unbilled authorised consumption QAN. The annualisation errors QVS.1 which seem to be the dominating component in apparent water losses QVS according to the case study results, were calculated successfully. Analysis of the individual components of real water losses QVR shows that water losses caused by tank overflow (QVR.1) can be quantified approximately on the basis of records from the SCADA system. The results also illustrate that the combined analysis of MNF and the date of defect repair allow for an approximate determination of leak flow rates and leak run times (QVR.3 and QVR.4). It was possible in small balancing zones (QN < 200,000 m3/a) to identify defects with leakage rates of just 0.5 m3/h. This approach allows new findings to be obtained with respect to the response time of a utility to defects and the extent of water losses QVR.5 resulting from unreported leaks and undetectable background leakage. Moreover, the results can also serve to determine night consumption in a zone and, thus, define an alarm level for on-line monitoring and control of the MNF.

Certain requirements must be met, if valid results are to be obtained. A GIS is required, which carries correct information about the pipe system and the locations of customers, of bulk meters, and tanks as well as of zone boundaries. Allocation of bulk meters and customers to databases with complete measurement and billing data must be possible. Detailed water balances taking into account withdrawals for flushing and firefighting, etc., require databases or IT systems with the appropriate basic information to be connected. Analysis of water losses from known defects needs additional data from the defect database or the WTS. Defining the topology and establishing the interfaces of the different systems for the GIS takes considerable time, but only have to be carried out once.

ACKNOWLEDGEMENTS

The authors are grateful to the project partners, 3S Consult GmbH, COS Systemhaus OHG, and Stadtwerke Pforzheim GmbH & Co KG, for their cooperation and to the German Federal Ministry of Education and Research for funding the work (funding code: 02WQ1254C).

REFERENCES

REFERENCES
AWWA
2009
AWWA Manual M 36: Water Audits and Loss Control Programs
. 3rd edn,
American Water Works Association (AWWA)
,
Denver
,
USA
.
Charalambous
B.
Hamilton
S.
2012
Water Balance – The Next Stage
. In
Proceedings of the Water Loss 2012
,
26–29 February 2012, Manila
,
International Water Association
,
London
,
UK
.
DVGW
2000
DVGW-Arbeitsblatt W 291: Reinigung und Desinfektion von Wasserverteilungsanlagen (DVGW Worksheet W 291: Cleaning and disinfection of water distribution facilities)
.
DVGW German Technical and Scientific Association for Gas and Water
,
Bonn
,
Germany
.
DVGW
2003
DVGW-Arbeitsblatt W 392: Rohrnetzinspektion und Wasserverluste – Maßnahmen, Verfahren und Bewertungen (DVGW Worksheet W 392: Pipe network inspection and water losses – measures, techniques and assessment)
.
DVGW German Technical and Scientific Association for Gas and Water
,
Bonn
,
Germany
.
Farley
M.
2001
Leakage Management and Control – A Best Practice Training Manual
.
World Health Organization, Water Supply and Sanitation Collaborative Council
,
Geneva
,
Switzerland
.
Farley
M.
Liemberger
R.
2005
Developing a non-revenue water reduction strategy: planning and implementing the strategy
.
Water Science & Technology: Water Supply
5
(
1
),
41
50
.
Farley
M.
Trow
S.
2003
Losses in Water Distribution Networks
.
IWA Publishing
,
London
,
UK
.
Klingel
P.
Knobloch
A.
2015
A review of water balance application in water supply
.
Journal AWWA
107
(
7
),
339
350
.
Knobloch
A.
Guth
N.
Klingel
P.
2014
Automated water balance calculation for water distribution systems
.
Elsevier Procedia Engineering
89
(
2014
),
428
436
.
Korkmazer
A.
2013
Ermittlung der Löschwasserentnahme aus Wasserversorgungsnetzen (Determining water supply network withdrawals for firefighting)
.
BSc thesis
,
Hochschule Esslingen – University of Applied Sciences, Esslingen, and Karlsruhe Institute of Technology (KIT), Institute for Water and River Basin Management
,
Karlsruhe
,
Germany
.
Lambert
A.
1994
Accounting for Losses: The bursts and background estimates concepts
.
Water and Environment Journal
8
(
2
),
205
214
.
Lambert
A.
Hirner
W.
2000
Losses from Water Supply Systems: Standard Terminology and Recommended Performance Measures. The blue pages, October 2000
.
Lambert
A.
Taylor
R.
2010
Water Loss Guidelines
.
Water New Zealand – The New Zealand Water & Wastes Association
,
Wellington
,
New Zealand
.
Lambert
A. O.
Brown
T. G.
Takizawa
M.
Weimer
D. A.
1999
Review of performance indicators for real losses from water supply systems
.
Journal of Water Supply – AQUA
48
(
6
),
227
237
.
Liemberger
R.
Farley
M.
2004
Developing a nonrevenue water reduction strategy Part 1: Investigating and assessing water losses
. In
Proceeding of the 4th World Water Congress
,
19–24 September 2004, Marrakech
,
International Water Association
,
London
,
UK
.
ÖVGW
2009
ÖVGW-Richtlinie W 63: Wasserverluste in Trinkwasserversorgungssystemen (ÖVGW Guideline W 63: Water losses in drinking water supply systems)
.
Austrian Association for Gas and Water (ÖVGW)
,
Vienna
,
Austria
.
Puust
R.
Kapelan
Z.
Savic
D. A.
Koppel
T.
2010
A review of methods for leakage management in pipe networks
.
Urban Water Journal
7
(
1
),
25
45
.
Renaud
E.
Lapuyade
F.
De Grissac
D.
Bremond
B.
2015
Annualizing metered volumes in water balance
.
Water Utility Journal
10
(
2015
),
5
14
.
US EPA
2010
Control and mitigation of drinking water losses in distribution systems
,
Report
,
United States Environmental Protection Agency (US EPA)
,
Washington, DC
,
USA
.