J Water SRT Ð Aqua Vol. 48, No. 6, pp.

227±237, 1999

A review of performance indicators for real losses from water

supply systems
A. O. Lambert*, T. G. Brown{, M. Takizawa{ and D. Weimer}, *International Water Data Comparisons Ltd; {AWWA,
North American Representative to UFW Task Force; {Tokyo Metropolitan Waterworks Japan; and }Neckarwerke Stuttgart AG, Germany

ABSTRACT: The IWA's Task Force on Unaccounted-for Water (UFW) had two key objectives. The
®rstÐRecommendations for a standard international terminology for calculation of real and apparent losses
from water balanceÐis presented as a Blue Pages [1]. As the secondÐto review performance indicators (PIs)
for international comparisons of losses in water supply systemsÐis only brie¯y mentioned in the Blue Pages,
this AQUA paper explains the technical basis for the task force's recommendations on PIs for real (physical)
losses. Traditional PIs were checked against several key local factors which constrain performance in
managing real losses. `Number of service connections' was found to be the most consistent of the traditional
PIs over the greatest range of density of service connections, and is recommended as the preferred basic
traditional technical indicator for real losses (TIRL). However, TIRL does not take account of several key
local factors. To overcome this de®ciency, TIRL should be compared with an estimate of unavoidable
annual real losses (UARL). An auditable component-based approach is developed and satisfactorily tested
for predicting UARL for any system, taking into account the local factors and using international data. The
infrastructure leakage index (ILI), calculated as the ratio of TIRL to UARL, is a nondimensional PI, which
enables overall infrastructure management performance in control of real losses to be assessed indepen-
dently of the current operating pressures; minimum achievable operating pressures are usually constrained
by local topography and standards of service.

INTRODUCTION . show the derivation of the parameter values used to predict

UARL for individual systems;
The annual volume of water lost is an important indicator of
. introduce the infrastructure leakage index (ILI), being the
water distribution eciency, both in individual years, and as a
ratio of TIRL to UARL;
trend over a period of years. High and increasing water losses
. test the UARL predictions and ILI calculations against a
are an indicator of ine€ective planning and construction, and of
wide range of international data;
low operational maintenance activities. The recommended
. explain how TIRL, UARL and ILI can be used as improved
terminology and method of calculation of real and apparent
diagnostic performance indicators;
losses for international comparisons is explained in the Blue
. provide examples showing how to calculate TIRL, UARL
Pages [1]. However, once these volumes have been calculated,
and ILI.
which performance indicators should be used to decide whether
The study uses a reference data set of 27 diverse water distribu-
real losses are `high' or `low', and how can rational national and
tion systems in 20 countriesÐAustralia, Brazil, Denmark,
international comparisons be made in a wide variety of di€er-
Finland, France, Germany, Gibraltar, Greece, Iceland, Japan,
ent situations?
Maltese Islands, Netherlands, New Zealand, Singapore, Spain,
The objectives of this paper are to:
Switzerland, Sweden, UK, USA, and West Bank (Palestine)Ð
. identify key local factors which may constrain technical
together with published data from other international sources
performance in managing real losses;
listed in the references.
. review the extent to which traditional PIs take account of
The methodology described here is an improvement of earlier
these key local factors;
draft versions presented and discussed at workshops and
. identify the preferred basic traditional PI with largest range
symposia in Portugal, UK, Brazil, USA and Australia, during
of application, and its limitations;
1997 and 1998, as the methodology and terminology for UARL
. propose an auditable component-based methodology for
and ILI were developed and re®ned.
calculating unavoidable annual real losses (UARL) for any
system, taking key local factors into account;
. describe the general relationship between UARL and eco-
nomic leakage levels;

# 1999 IWSA 227

228 A. O. Lambert et al.

REVIEW OF TRADITIONAL PERFORMANCE Continuity of supply is often assumed, but this is not the case
INDICATORS FOR REAL LOSSES in many countries. In situations of intermittent supply, the
percentage of time for which the distribution system is pres-
Key local factors in¯uencing real losses
surised is an important parameter to be included in PIs for real
The type of soil/ground can in¯uence the frequencies of leaks losses. This is easily achieved by expressing the annual volume
and bursts, and the speed with which leaks and bursts become of real losses as a volume per day `when the system is
visible at the ground surface. However, correct selection and pressurised' (w.s.p.). The average operating pressure should
laying of pipe materials, and modern leakage control methods also be calculated over the period when the system is pres-
(e.g. night ¯ows) can reduce these in¯uences signi®cantly. surised.
There are ®ve other local factors which constrain perfor- Because variations, in pressure (and leakage rates) over 24-h
mance in managing real losses, which can vary widely between periods are often substantial, it is preferable to express losses
individual distribution systemsÐcontinuity of supply, length of derived from annual water balance on a `per day' basis rather
mains, number of service connections, location of customer than `per hour'.
meters on service connections, and average operating pressure.
Note that `number of service connections' should be used in
PIs for real losses, rather than `number of properties'. This is Limitations of basic traditional performance indicators
because there is no standard international de®nition of `proper- The basic traditional PIs for real losses which are most widely
ties'; real losses are calculated up to the ®rst metering point, and used in di€erent parts of the world to make comparisons of the
in cities the service frequently splits into several separate pipes annual volume of real losses are:
serving individual domestic or commercial properties after the . % of input volume;
®rst metering point. . volume lost per length of mains per unit time;
In approximately half of the 27 systems in the reference data . volume lost per property per unit time;
set, customer meters were located close to the edge of the street. . volume lost per service connection per unit time;
In the remainder, customer meters were located up to 30 m . volume lost per length of system per unit time (where length
from the edge of the street. Rational PIs for real losses need to of system = length of mains + length of service connections
allow for such substantial di€erences. up to point of customer metering)
Density of service connectionsÐexpressed as a number per Traditional PIs for real losses appear to be selected on the basis
km of mainsÐvaried widely in the reference data set, from 24/ of the simplicity of calculation, or country tradition, or avail-
km to 114/km, with a median of 47/km, and more extreme ability of data for the calculation, or even the PI which produces
values are known to exist. This factor has a major in¯uence on the best impression of performance. However, the di€erences
real losses. Note that use of the qualitative terms `urban' or can be substantial [5]. The proper basis of selection should be
`rural' to imply ranges of connection densities is misleading in the PI which gives the most rational technical basis for compar-
an international contextÐit is recommended that connection isons. Table 1 shows the limited extent to which each of the
densities should always be quoted on a `per km of mains' basis. traditional PIs take into account the key local factors (other
Because operating pressures are constrained by local topo- than ground conditions) which in¯uence real losses.
graphy and minimum standards of service (to customers or for
®re-®ghting) average operating pressures vary widely between
systemsÐfrom 30 m to over 100 m (median 45 m) in the The traditional PI with the greatest range of applicability
reference data setÐand more extreme values are known to
Table 1 shows that real losses expressed as a percentage of
exist. Many countries recognise pressure control as a technique
system input does not take account of any of the key local
for managing leakage, but there are local limits to the lowest
factors; instead, under continuous supply conditions, the
acceptable average pressures which can be achieved. The
average rate of consumption (which is not a primary explana-
average frequency with which new leaks occur, and rates of
tory parameter) dominates the calculated value [1]. If real losses
¯ow of individual leaks, are very sensitive to operating pres-
average 100 L/service connection/dayÐwhich is a good perfor-
sures. The observed relationship between pressure and leakage
mance for a system with average operating pressures and
rate for individual small sectors of distribution systems varies
density of connectionsÐthen real losses as percentage of
widely [2,3] because the areas of some types of leakage paths
system input would be:
vary with pressure [4]. The weighted average relationship for
large systems appears to be that leakage rates vary with
29% for consumption of 250 L/conn/d (e.g. Maltese Islands)
pressure approximately to the power 1.15, so the simplifying
17% for consumption of 500 L/conn/d (e.g. UK, Netherlands)
assumption that leakage rate varies linearly with operating 9% for consumption of 1000 L/conn/d (e.g. German cities)
pressure is likely to be reasonably satisfactory for performance 2% for consumption of 5000 L/conn/d (e.g. Scandinavian city)
comparisons of real losses for large systems, except at very high 1% for consumption of 8000 L/conn/d (e.g. Singapore)
or very low pressures

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

 Performance indicators for real losses 229

Table 1 Do traditional performance

indicators for real losses allow for key local Location of
factors? Number of customer Average
Basic traditional Continuity Length of service meters on operating
PI for real losses of supply mains connections services pressure

% of Volume input No No No No No
Litres/property/day No No Only if No No
1 property/conn
Litres/service No No Yes No No
connection/day
m3/km mains/day No Yes No No No
m3/km of system/day No Yes Possibly Yes No

Also, considerable confusion is introduced when interpreting losses, to be referred to as the `technical indicator real losses'
percentage losses data in intermittent supply situations [6]. (TIRL) is:
Accordingly, over the last 30 years this measure has consis- Litres/service connection/day, when the system is pressurised
tently been rejected by National Technical CommitteesÐin the (w.s.p)
UK [3,7], Germany [8] and South Africa [9]Ðand more recently However, the Task Force recommended further interpreta-
by the UK Economic Regulator OFWAT [10] and the IWA tion of the calculated TIRL value for an individual system by
UFW Task Force [1]. comparing it with a calculated value for unavoidable annual
Of the remaining basic traditional PIs in Table 1, `number of real losses (UARL), using a methodology which takes account
service connections' is logically preferable to `number of of the local factors of density of connections, location of
properties', which can be rejected for reasons previously customer meters on service connections, and average operating
explained. It might also appear logical to assume that `length pressure. The component-based calculation of UARL is
of system' allows for more of the key factors than `number of described in the next section of the paper. The ratio of TIRL
connections' or `length of mains'. However, it was the experi- to UARL becomes a nondimensional Infrastructure Leakage
ence of all the Task Force members, and other experienced Index (ILI), which allows overall infrastructure management
practitioners who o€ered views, that (except at low density of performance to be assessed independently of the in¯uence of
connections) in well-run systems the majority of leaks and current operating pressure.
bursts (and of the annual volume of real losses) occurs on Figure 1 shows values of real losses in litres/service connec-
service connections rather than mains, with most frequent tion/day w.s.p. for each system in the reference data set.
problems in the section of the service connection between the Figure 2 shows the values of ILI for each of the systems. The
main and the edge of the street. results are discussed later in the paper, after the concept and
The Task Force therefore recommended [1] that the basic calculation of UARL is explained.
traditional PI with the greatest range of applicability for real

Fig. 1 Technical indicator for real losses

(TIRL) in L/service connection/day for 27
water supply systems in 20 countries.

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

230 A. O. Lambert et al.

Fig. 2 Infrastructure leakage index (ILI)

values for 27 supply systems in 20
countries.

UNAVOIDABLE ANNUAL REAL LOSSES: the average real losses (x-axis) reduce asymptotically towards
CONCEPT AND CALCULATION some base level, the annual cost of the lost water decreases as
the average volume of real losses falls. The economic level of
The concept of unavoidable annual real losses
losses occurs when the total cost curve (A' 4 B' 4 C'), which is
Leakage management practitioners recognise that it is impos- the sum of the cost of lost water and the cost of active leakage
sible to eliminate real losses from a large distribution system. control, is at a minimum (point B' in Fig. 3). If, in simplistic
There must therefore be some value of `unavoidable annual real terms, we assume that:
losses' (UARL) which could be achieved at the current operat- . the infrastructure is in good condition;
ing pressures if there were no ®nancial or economic constraints. . point A represents the technical `state of the art' for intensive
If the UARL volume for any system can be assessed, taking active leakage control; and;
into account key local factors, then the ratio of technical . all detectable leaks and bursts are identi®ed and repaired
indicator real losses (TIRL) to UARL o€ers the possibility of rapidly and e€ectively.
an improved performance indicator for real losses. The real losses for point A therefore correspond to unavoidable
annual real losses (UARL). Actual or economic levels of real
losses should always lie at, or to the right, point A. The
Relationship between UARL and economic levels of losses
infrastructure leakage indexÐthe ratio of actual or economic
A simpli®ed economic approach [11] to determining an appro- real losses to UARLÐshould always exceed 1.0.
priate intensity of active leakage control for dealing with
unreported leaks and bursts is outlined in Fig. 3. As the
A component-based approach to assessing unavoidable annual
intensity of active leakage control increases (C 4 B 4 A),
real losses
causing the annual cost of leakage control (y-axis) to increase,
The `BABE' (background and bursts estimates) [12] approach
for calculations of components of real losses, successfully used
in a number of speci®c studies in di€erent countries (including
World Bank projects), considers real losses in three categories
for modelling and calculation purposes:
. Background losses from undetectable leaks (typically low
¯ow rates and long durations);
. Losses from reported leaks and bursts (typically high ¯ow
rates and short durations);
. Losses from unreported bursts (typical ¯ow rates and dura-
tions depend on the method and intensity of active leakage
control).
Using the BABE technique, it is possible to predict with
reasonable overall accuracy, for each individual system, what
the average UARL would be for various components of
infrastructure at any speci®ed pressure. The simpli®ed compo-
nents of infrastructure used for this study have been selected for
Fig. 3 Relationship between unavoidable annual real losses and ease of calculation in diverse international situations. Para-
economic level of real losses meters which are required for these BABE calculations are

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

 Performance indicators for real losses 231

Table 2 Parameters required for

calculation of unavoidable annual real Background
losses (UARL) Component of (undetectable) Reported Unreported
infrastructure losses bursts Bursts

Mains Length Number/year Number/year

Pressure Pressure Pressure
Min loss rate/km* Average ¯ow rate* Average ¯ow rate*
Average duration Average duration
Service Number Number/year Number/year
connections, Pressure Pressure Pressure
main to edge Min loss rate/conn* Average ¯ow rate* Average ¯ow rate*
of street Average duration Average duration
Service Length Number/year Number/year
connections Pressure Pressure Pressure
after edge of Min loss rate/km* Average ¯ow rate* Average ¯ow rate*
street Average duration Average duration

* At some speci®ed standard pressure.

Table 3 Parameters values used for

calculation of unavoidable annual real Background
losses (UARL) Infrastructure (undetectable) Reported Unreported
component losses bursts bursts

Mains 20 L/km/h* 0.124 bursts/km/year 0.006 bursts/km/year

at 12 m3/h* at 6 m3/h*
for 3 days duration for 50 days duration
Service 1.25 L/conn/h* 2.25/1000 conn/year 0.75/1000 conn/year
connections to at 1.6 m3/h* at 1.6 m3/h*
edge of street for 8 days duration for 100 days duration
Service 0.50 L/conn/h* 1.5/1000 conn/year 0.50/1000 conn/year
connections after at 1.6 m3/h* at 1.6 m3/h*
edge of street{ for 9 days duration for 101 days duration

*All ¯ow rates are quoted at 50 m pressure.

{For 15 m average length.

shown in Table 2. No UARL allowance is given for service The calculated values of UARL for each component of
reservoir leakage or over¯ows, or for pipework located above infrastructure, using the Table 3 values, are shown in Table 4.
ground. An example of the calculation process, for the average annual
losses from reported bursts on mains, is as follows:

Calculating components of unavoidable annual real losses UARL component = Burst frequency 6 Average flow
rate 6 Average duration
The parameter values used to calculate the Table 2 UARL = 0.124 bursts/km/year
components for di€erent sections of infrastructure are based on 6 (12 6 24 h) m3/day 6 3 days
published international data (summarised in Table 3) for = 107 m3/year per km mains at 50 m
minimum background loss rates, typical burst ¯ow rates and pressure
frequencies [13±15] for infrastructure in good condition. = 293 L/km/day at 50 m pressure
Average durations assumed for unreported bursts are based = 5.8 L/km/day/m pressure
on intensive active leakage control, approximating to night
¯ows (or water balance) once per month on highly sectorised It can of course be argued that not all systems with good
distribution networks. infrastructure condition would experience the same burst

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

232 A. O. Lambert et al.

Table 4 Calculated components of

Infrastructure Background Reported Unreported UARL unavoidable annual real losses (UARL)
component losses bursts bursts total Units

Mains 9.6 5.8 2.6 18 L/km mains/

day/m of pressure
Service connections, 0.60 0.04 0.16 0.80 L/conn/day/
meters at m of pressure
edge of street
Underground pipes 16.0 1.9* 7.1* 25 L/km u.g. pipe/
between edge day/m of pressure
of street and
customer meters

* Assuming an average length of 15 m/connection for calculation purposes.

frequencies and average ¯ow rates as assumed in Table 3. Table 5 UARL values in L/service connection/day, for customer
However, the `background' loss components of UARL dom- meters located at edge of street. `Add-on' values for underground
pipes distant from edge of street shown at foot of table
inate the calculated values, and sensitivity testing shows that
di€erences in assumptions for parameters used in the `bursts'
Density of Average operating pressure (m)
components have relatively little in¯uence on the UARL `total connections
losses' values (5th column of Table 4). (per km mains) 20 40 60 80 100
The `UARL total' values, in the units shown in Table 4,
provide a rational yet ¯exible basis for predicting UARL values 20 34 68 112 146 170
for a wide range of distribution systems, taking into account 40 25 50 75 100 125
continuity of supply, length of mains, number of service 60 22 44 66 88 110
connections, location of customer meters, and average operat- 80 21 41 62 82 103
ing pressure. An example calculation using Table 4 values is 100 20 39 59 78 98
shown at the end of the paper. Add on, for each 0.5 1.0 1.5 2.0 2.5
The Table 4 values can also be presented as a wide variety of metre of pipe (per
equations, look-up tables, graphs and spreadsheets, in any connection) between
selected combination of metric or imperial measurement units. edge of street
In the most basic form, UARL in L/day is and customer meter

UARL = (18 6 Lm + 0.80 6 Nc + 25 6 Lp) 6 P

where Lm is mains length in km, Nc is number of service

Graphical presentation of UARL predictions
connections, Lp is the total length in km of underground pipe
(between the edge of street and customer meters), and P is To demonstrate important features of relationships between
average operating pressure in metres. This basic equation can UARLs and local key factors, and to test the validity of the
be manipulated into many other forms and units, for example UARL predictions, the Table 4 `Total Losses' values have been
into a look-up table (Table 5) or graphs (Figs 4 and 5). used to produce graphs where the x-axis value is density of
UARL values for each individual system can be read o€ or service connections (per km of mains) and the y-axis is the
interpolated from Table 5. For example, a system with connec- UARL value in:
tion density 40 per km mains at 60 m average pressure has a . L/service connection/day/m of pressure (Fig. 4);
UARL of: . L/km of mains/day/m of pressure (Fig. 5).
. 75 L/service connection/day for customer meters located at The three lines on each of Figs 4 and 5 show the UARL
the edge of the street; losses on mains only (bottom line), on mains plus service
. 90 (= 75 + 1.5 6 10) L/conn/day for meters located 10 m connections for customer meters located at the edge of the
from the edge of the street. street (middle line), and (upper line) on mains plus service
Table 5 demonstrates very clearly why it has previously connections where there are 15 m of underground pipe (per
proved impossible to quote a reliable single value for unavoid- service connection), between the edge of the street and the
able real losses, even when the best of the traditional perfor- customer meter. So, for example, for a system with density of
mance indicators is used, because of the wide range of local key connections of 70 per km mains, the UARL values can be read
factors experienced internationally. o€ from Figs 4 and 5 as:

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

 Performance indicators for real losses 233

Fig. 4 Unavoidable annual real losses in L

per service connection per day per metre of
pressure, vs. density of service connections.

Fig. 5 Unavoidable annual real losses in L

per km of mains per day per metre of
pressure, vs. density of service connections.

. 1.06 L/conn/day/m pressure, or 74 L/km mains/day/m pres- . USA, 2.4±7.1 m3/km/day (1000±3000 US gallons/mile/day)
sure, for customer meters located at the edge of the street; [16];
. 1.43 L/conn/day/m pressure, or 100 L/km mains/day/m pres- . Germany, 1±5 m3/km/day depending on ground type, for
sure, for customer meters located 15 m from the edge of the density of connections between 35 and 50 per km [7];
street. . France, 1.5±7 m3/km/day for `rural' to `urban' situations
The ratio of these ®gures, which is 1.35, is very close to the [17].
ratio of values for real losses published by OFWAT [10] in The wide ranges and limiting constraints of these ®gures have
England and Wales for customer meters at, or around 15 from, severely limited their application to speci®c situations outside
the edge of the street, for an average density of connections of their country (or region) of origin. In the international refer-
70 per km of mains. Further tests of the validity of the UARL ence data set, individual density of service connections varied
predictions are described below. from 24/km to 114/km. `urban' connection densities for
German cities (around 45/km) were around twice those of
Scandinavian cities, half of those for Japanese/Brazilian/UK
TESTING THE VALIDITY OF THE UARL cities, but similar to the values for the most rural of the England
PREDICTIONS and Wales water companies.
Comparisons of UARL predictions with ranges of `unavoidable Assuming typical operating pressures between 30 and 60 m,
losses' customer meters at an average of 7.5 m from the edge of the
street, and density of service connections typically between 20
Examples of previously published values for `unavoidable and 100 per km mains, Fig. 5 can be used to predict that the
losses' are:

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

234 A. O. Lambert et al.

Table 6 Infrastructure leakage index (ILI)

Unavoidable Technical values for well-managed systems in four
annual real indicator countries
Connection Location of Average losses real losses ILI =
Supply density per customer pressure (UARL) (TIRL, (TIRL/
system km mains meters (m) (L/conn/day) L/conn/day) UARL)

A 86 ES* 39 39.4 69.9 1.77

B 47 ES + 30 m 57 111 146 1.31
C 38 ES + 10 m 40 60.9 73.8 1.21
D 39 ES + 11 m 40 61.3 107 1.75

* ES = edge of street.

typical range of UARL values, in m3/km mains/day are close to OTHER ASPECTS OF UARLS
published ranges for `unavoidable losses' in USA, Germany
What does the UARL approach tell us about traditional
and France.
performance indicators?
. Lower (at 20 conns/km): 37 L/km/day/m 6 30 m pressure =
1.1 m3/km/day; The shape of the lines in Fig. 4 show that for a wide range of
. Upper (at 100 conns/km): 117 L/km/day/m 6 60 m pressure values of connection densities (30 to over 100), the UARL in L/
= 7.0 m3/km/day. conn/day/m pressure is within + 15% of the value at the
However, the UARL approach has the advantage that it median connection density of 47 per km in the reference data
gives a speci®c value for `unavoidable losses' for each system set. This is because, at connection densities greater than around
depending upon its own local environment factors. The next 20 per km, over 47% of the UARL occurs on service connec-
test uses this feature. tions rather than mains. Conversely, the UARL in L/km mains/
day/m of pressure (Fig. 5) varies widely over the whole range of
connection densities.
Figures 4 and 5 provide strong technical support for the Task
Force recommendation (initially based on experience world-
Comparison of UARL values for four well-managed systems wide) that `per service connection' is preferable to `per km
Four supply systems requiring active leakage control were mains' as a basic technical PI for real losses for international
selected from the reference data setÐtwo from the Asia Paci®c comparisons, for a large range of connection densities exceed-
region, and two from western Europe. Each has a good ing 20 per km. UARL losses which are expressed `per km of
national and international reputation for technical leakage system/day/m of pressure' [18] can also be seen to be slightly
management with sectorised networks. The systems (Table 6) less consistent than on a `per service connection' basis.
cover a diverse range of operating pressures, density of connec-
tions and customer meter locations. If the true leakage manage-
ment performance of these four systemsÐwhich should be
similarÐis assessed in terms of traditional PIs, the rank orders
Situations where UARL calculations are unlikely to be valid
are di€erent for each PI, and the values range from:
. 70±146 L/service connection/day. The basic assumptions used in the UARL predictions may
. 2.8±6.8 m3/km/day; break down in situations where intensive active leakage control
. 1% to 23% of system input volume. to locate unreported leaks is not possible, or not necessary. For
If the UARL predictions are reasonably representative, and example, in situations where pressures are signi®cantly less than
the ratio of TIRL/UARL (the infrastructure leakage index) is a around 20 m, sonic detection of hidden leaks may not be
reliable PI of overall infrastructure leakage management at possible with some pipe materials and some depths of cover.
current operating pressures, the ILI values in Table 6 should all In some types of soil, where all signi®cant new leaks and
be moderately greater than 1.0, and reasonably similar to each bursts become rapidly visible at the ground surface, the Table 4
other. The actual values of infrastructure leakage index in values will inevitably over-estimate the attainable level of
Table 6 pass this test. The variation (approximately + 20%) UARL where there is good infrastructure and rapid good
from the average value ILI of 1.50 is well within the likely range quality repair of all visible leaks and bursts. For example, in
of error in assessing real losses from water balances on systems the German DVGW technical recommendations [8], sandy soils
with low levels of real losses [1]. have the lowest `lower limit' for losses.

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

 Performance indicators for real losses 235

CALCULATING AND INTERPRETING THE repairs. Figure 2 shows the range of ILIs for the reference data
RECOMMENDED PERFORMANCE set, which vary from around 0.7 to just over 10.
INDICATORS The e€ect on real losses of managing operating pressuresÐ
increasing pressures to meet minimum standards of service, or
Recommended calculation procedure
decreasing them to reduce excess pressures in parts of the
The systematic step-by-step procedure for calculating recom- system, or at speci®c times of dayÐcan and should be assessed
mended performance indicators for real losses is detailed in [1]. separately from the ILI calculation. A simple initial assumption
Current real losses are calculated as an annual volume (m3/ for calculations is that real losses in large systems will increase
year), then expressed in m3/day when the system is pressurised, and decrease linearly with average pressure, over small ranges
then in terms of the Task Force's recommended technical of pressure.
indicator for real losses, TIRLÐlitres per service connection
per day w.s.p.
DISCUSSION OF RESULTS FROM
Next, the unavoidable annual real losses (UARL) are calcu-
REFERENCE DATA SET
lated at the current operating pressure for up to three compo-
nents of infrastructure (depending on customer meter location). Operators of all of the systems which have ILIs between 1.0 and
The infrastructure leakage index (ILI) is then calculated as the the median value of 2.9 in the reference data set (Fig. 2) make
ratio of TIRL to UARL. Simpli®ed examples of these calcula- substantial e€orts to manage and maintain their infrastructure,
tions are shown at the end of the paper. ensure that all detected leaks and bursts are promptly repaired,
and undertake active leakage control on a continuous or semi-
continuous basis. Those which have ILIs in the range 1.0±2.0
Interpreting the TIRL, UARL and ILI values
also have good reputations in technical leakage management.
The technical indicator for real losses (TIRL) in litres per The lowest ILI reading, of 0.7 is both from a country (Nether-
connection/day w.s.p. is the traditional basic performance lands) where the ground conditions favour leaks showing
measure with the greatest range of applicability. However, rapidly at the surface and little active leakage control is
individual values of TIRL may still be in¯uenced by operating required (so assumptions for calculating UARL are likely to
pressure, location of customer meters and low density of produce over-estimates to some extent).
connections. Figure 1 shows the values of TIRL for the 27 Almost all systems which have ILIs signi®cantly greater that
systems in the reference data set, which vary from 29 to 832 L/ the median value of 2.9 have problems associated with old or
connection/day w.s.p.Ða range of 28±1. poor infrastructure, or a relatively relaxed active leakage
control policy. In some cases, because of relatively low
The unavoidable annual real losses (UARL) is a prediction of pressures and high consumption per connection, the previous
what the real losses would be for any speci®c system if all use of percentages as a traditional performance indicator
infrastructure was in good condition, with intensive `state of the appears to have masked opportunities for further reductions
art' active leakage control, and all detectable leaks and bursts in real losses.
are repaired quickly and e€ectively. It takes account of length Comparison of Figs 1 and 2 shows that the omission of some
of mains, number of service connections, location of customer of the key local factors from the basic recommended TIRL
meters, continuity of supply, and average operating pressures may, in certain situations, compromise the assessment of true
(when the system is pressurised) between 20 and 100 m. It is not performance in managing real lossesÐfor example in systems 5,
necessarily economic to achieve the UARL. The ability to 11, 13±16, 22±24, 26 and 27. Although the more diagnostic
calculate reasonably reliable values of UARL has several approach based on UARL and ILI requires assessment of
applications in leakage management studies, but this paper estimates of density of service connections, meter location, and
considers only performance indicators. The UARLs of the average operating pressure, the extra e€ort is likely to be
reference data set vary from 32 to 153 L/connection/day justi®ed.
w.s.p.Ða range of 5±1.
CONCLUSIONS
The infrastructure leakage index (ILI) is the ratio of the
technical indicator real losses (TIRL) to the value of UARL The main messages of this paper are:
calculated for current pressures and continuity of supply. It is a . Key local factors which constrain achievable annual volume
nondimensional performance indicator of the current overall of real losses have been identi®ed.
management of the infrastructure for leakage control purposes. . Traditional PIs have been checked against these key
The greater the amount by which the ILI exceeds 1.0, the factorsÐcontinuity of supply, mains length, number of
greater the potential opportunity for further management of service connections, location of customer meters, and
real losses by infrastructure management and maintenance, average operating pressure.
more intensive active leakage control, or speed and quality of . The common practice of expressing Real Losses as a percen-

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

236 A. O. Lambert et al.

tage of volume input has to be rejected as a technical PI; it Unavoidable annual real losses (UARL) components:
takes none of these factors into account, and is unduly Mains = 18 L/km/d/m 6 1500 km 6 (0.9 6 365) days
in¯uenced by consumption. 6 30 m/106 = 266 6 103 m3/year
. In most well-run systems, the greatest proportion of real Connections to edge of street = 0.8 L/conn/d/m 6 60 000
losses volume occurs on service connections. 6 (.09 6 365) days 6 30 m/106 = 473 6 103 m3/year
. The recommended basic technical indicator for real Losses Edge of street to customer meter = 25 L/km/d/m
(TIRL) is therefore the annual volume of real losses in litres 6 (60 000 6 6/1000) 6 (0.9 6 365) days 6 30 m/106
per service connection per day, when the system is pres- = 87
surised (w.s.p) rather than real losses per km of mains per day
(w.s.p). Total unavoidable annual real losses, UARL = 266 + 473 + 87
. The TIRL does not take account of the local key factors of = 826 6 103 m3/year
density of connections, location of customer meters and = 826 6 103 6 103/(60 000 6 0.9 6 365)
average operating pressure. In the international reference = 42 L/SERVICE CONNECTION/DAY W.S.P.
data set, these factors varied widely.
. An approach which takes these local factors into account has Infrastructure Leakage Index (ILI)
been developed and tested, to assist in interpreting the = TIRL/UARL = 202/42 = 4.8.
calculated TIRL values.
. The improved approach is based on predicting components
of unavoidable annual real losses (UARL) for each indivi- BIBLIOGRAPHY
dual system, taking into account these local factors. 1 Lambert A, Hirner WH. Losses from Water Supply Systems.
. The ratio TIRL/UARL becomes a nondimensional infra- Standard Terminology and Performance Measures. IWSA Blue
structure leakage index (ILI). Pages; 1999, in preparation.
. The infrastructure leakage index approach provides an 2 Ogura. Jap Water Wks Assoc J 1979; June.
improved basis for technical comparisons, which separates 3 DoE/NWC. Leakage control policy and practice. DoE/NWC,
aspects of infrastructure management performance (pipe Standing Technical Committee Report 26. ISBN 0 904561 95
X; 1980.
selection/installation/maintenance/renewal/replacement,
4 May J. Pressure dependent leakage. World Water Environ
speed and quality of repairs, and e€ectiveness of active
Engng; October 1994.
leakage control policy) from aspects of pressure manage- 5 IWSA International Report. Unaccounted±for water and the
ment. economics of leak detection. IWSA World Congress, Copenha-
gen; 1991.
6 Seba Messtechnik GmbH & Co. KG; 1996. Consulting services
ACKNOWLEDGEMENTS within the project water distribution network rehabilitation in
Sarajevo, Bosnia Herzegovina. NRW Draft Report, October
To Laurent Hecquet (France) who was an initial member of the 1996.
Task Force. To all water suppliers who contributed data to the 7 Managing Leakage Report B. Reporting comparative leakage
study, and to Dr Wolfram Hirner and the many practitioners performance. ISBN 1 898920 07 9; 1994.
world-wide who assisted with their comments and constructive 8 Deutscher Verein des Gas- und Wasserfaches (DVGW) e. V
criticisms. Technische Mitteilungen W391, Wasserverluste in Wasserver-
teilungen, S.8. (DVGW Standard Regulations, Note-Paper
W391: Water Losses in Water Distribution Systems: Identi®ca-
EXAMPLE CALCULATIONS tion and Assessment) October; 1986.
9 Code of practice for the management of potable water distribu-
Example: A distribution system has 1500 km mains and 60 000 tion systems. South African Bureau of Standards, Private Bag
service connections with customer meters located (on average) X191, Pretoria 0001, S. Africa; 1998, in press.
6 m from the edge of the street. The system is pressurised for 10 Report on leakage and water eciency. Oce of Water Services.
90% of the time, and the average pressure (when pressurised) is ISBN 1 874234 42 6; 1997±98.
30 m. The current annual real losses in the above system, 11 Managing water leakage. Economic and technical issues. Finan-
cial Times Energy. ISBN 1 84083 011 5; 1998.
calculated from annual water balance, are 4000 6 103 m3/year.
12 Lambert AO. Accounting for lossesÐthe bursts and back-
Calculate the technical indicator for real losses (TIRL), una-
ground estimates concepts. J Institution Water Environ Mngnt
voidable annual real losses (UARL) (using Table 4) and the 1994; 8(2): 205±214.
infrastructure leakage index (ILI). 13 Managing leakage report E. Interpreting measured night ¯ows.
ISBN 1 898920 10 9; 1994.
Technical indicator for real losses (TIRL) 14 Heide G. Private communication of data attributed to Dr Hoch
= 4000 6 103 6 103/(60 000 6 0.9 6 365) of Germany.
= 202 L/service connection/day w.s.p. 15 Sattler R, Einfuhrung der bundesweiten DVGW-Schadenssta-

# 1999 IWSA, J Water SRTÐAqua 48, 227±237

 Performance indicators for real losses 237

tistik Wasser-verteilung. gwf Wasser Special 1997; 138(13): 17 Agence l/Eau Rhone Mediterranee Corse. Le Diagnostic des
S27±S31. reseaux d'eau potable, 1990.
16 AWWA. Minutes of American WaterWorks Association Leak 18 England & Wales Environment Agency. Demand management
Detection and Water Accountability Committee. 21 Septem- bulletin, Issue 33, February; 1999.
ber; 1998.

# 1999 IWSA, J Water SRTÐAqua 48, 227±237