Lilongwe Water Board (LWB) is currently unable to meet Lilongwe City's water demand as evidenced by low supply coverage (65%) and intermittent water supply in the city. One of the major challenges is high levels of unaccounted for water (UFW) reported at 37% (2012), higher than the recommended 23% for developing countries. This study, done in Lilongwe City (Areas 15, 18 and 28), investigated water losses and partitioned UFW into real and apparent losses. Data collection involved data logging for pressures and flows at selected points in the network, meter testing, and water audits. This study estimated an average UFW of 37.5% for Lilongwe City and 33%, 44% and 20%, respectively, in the specific study areas (Areas 15, 18 and 28). The UFW in Lilongwe City was higher than recommended and was also higher than recommended in Areas 15 and 18 but within the acceptable limit for Area 28. High UFW levels in Areas 15 and 18 were mainly driven by real losses. The LWB should consider partitioning of its UFW to establish the main drivers, implement active leak detection programme and active pressure management in areas with high pressures.

INTRODUCTION

In sub-Saharan Africa, the water services sector has for a long time suffered from poor performance of its public water utilities (Kalulu & Hoko 2010). According to WHO/UNICEF JMP (2014), 700 million people still lack access to improved sources of drinking water, of which nearly half are in sub-Saharan Africa. Many of the efforts to solve the water crisis have proved to be unsustainable (Brunson et al. 2013). Kingdom et al. (2006) highlighted that one of the major challenges affecting water utilities in the developing world is the high levels of unaccounted for water (UFW). UFW is the difference between system input volume and authorized consumption and comprises apparent and real losses (RL) (IWA 2000). UFW seriously affects the financial viability of water utilities through lost revenues and increased operational costs (Kingdom et al. 2006). Lost or UFW can be equated to lost or unaccounted revenue (EPA 2010). According to Balkaran & Wyke (2007), UFW ranges between 15% and 30% in the developed world but elsewhere it is more likely to be in the 30%–60% range. The target for UFW, as proposed by Tyanan & Kingdom (2002), is less than 23% for well performing utilities in developing countries.

Water losses in the distribution system is a common problem in Malawi, as in other developing countries worldwide (Kafodya 2010). According to Mulwafu et al. (2002), several water supply schemes in Malawi are old and thus prone to high water losses through leakages, and the average levels of UFW in urban utilities range from 20% to 30% and can go up to 51% in some urban areas served by regional water boards in Malawi. LWB highlighted in its corporate plan report for 2012–17 that high levels of UFW, at 37%, are one of its major challenges in addition to the ever increasing water demand by the city due to urbanization and erratic power supply (LWB 2011). The UFW of 37% is higher than the recommended figure of less than 23% for developing countries. Nkhoma et al. (2005) reported a UFW of 38% for Lilongwe City for the year 2004 and SOGREAH (2010) reported a UFW of 37.68% for Lilongwe City for the year 2008.

The UFW levels reported for various years show that there has not been a great improvement in the reduction of UFW for the utility. Despite these high UFW levels, the city has a low coverage of water, estimated at around 60% of the population. Even at this coverage the city is now threatened with raw water shortages. In response to this, the city intends to implement a number of activities to increase raw water availability and security. However, funding for this is not easy to come by and has largely been through loans and grants from multinational institutions, thus increasing the cost of water provision. LWB has set a target of only 5% losses between abstraction and production by 2015 and only 20% of losses in the distribution system by 2025 (SOGREAH 2010). Thus, reduction of UFW will provide a new source of water and complement efforts to improve water availability.

According to Kingdom et al. (2006), no proper UFW reduction strategy can be planned without the quantification of real and apparent losses (APL). Chipwaila (2009) highlighted that partitioning UFW provides a better understanding of what happens to the water after it leaves the treatment plant. Thornton et al. (2008) recommended the use of water loss management tools in the quantification of UFW and outlined that modelling water losses is an excellent tool in performing water audits and in water loss management planning. The objective of this study was then to investigate the performance of LWB in terms of management of UFW through application of models and to establish the main drivers of UFW.

STUDY AREA

Location

This study was carried out in Lilongwe City and the specific study areas were Areas 15, 18 and 28 which are within the supply area of LWB (Figure 1). Figure 2 is the schematic diagram of the study areas. Lilongwe City was selected because, apart from being the capital city, it is the fastest growing city of Malawi due to rural–urban influx, expansion of the city to the periphery, high birth rate and reduced mortality due to improved health services (NSO 2008).
Figure 1

Location of the study areas: adapted from SOGREAH (2010).

Figure 1

Location of the study areas: adapted from SOGREAH (2010).

Figure 2

Schematic diagram for the study areas.

Figure 2

Schematic diagram for the study areas.

Areas 15, 18 and 28 were selected first because they are district metered areas and had functional bulk meters during the study period which made this research study feasible, as recommended by Farley (2001). Areas 15 and 18 are residential areas while Area 28 is an industrial area. Projected population for 2015 for Areas 15, 18 and 28 are 2,124, 18,846 and 65 with 278, 2,240 and 141 customer connections, respectively (NSO 2008; LWB 2014).

Water supply in the study area

The Lilongwe Water Board (LWB) was mandated, under the Waterworks Act No. 17 of 1995, to provide potable water within the urban and peri-urban areas of the City of Lilongwe. As reported in the 2012–17 five-year corporate plan report, the total water consumption as met by LWB supply system is 55,000 m3/day, spread over 37,000 residential connections, 525 kiosks, 712 institutional connections, 2,481 commercial and industrial connections, and two bulk supplies to the Central Region Water Board (LWB 2011). According to SOGREAH (2010), the water demand for the city is 24 litres per capita per day (LPCD) at a kiosk connection and 81 LPCD at a residential connection.

The LWB has two potable water treatment plants with a combined design capacity of 95,000 m3/day but they are operating below the design capacity with a production of 87,000 m3/day. To augment supplies, LWB has embarked on the Diamphwe Raw Water Supply project designed to meet the drinking water demand of 210,540 m3/day by the year 2045 (SP 2015).

LWB has previously reported an average of 18 hours/day of continuous supply of potable water to over 60% of customers in the distribution system (LWB 2011). However, in the study areas, supply is reported to range from 20 to 24 hours per day and 7 days a week. The disruptions in continuity of supply are the result of power outages in pumping stations, inadequate water supply and frequent pipe bursts (LWB 2011).

MATERIALS AND METHODS

Selection of monitoring points

Monitoring points were selected based on the parameters that are required in performing water audits, analysing minimum night flows (MNFs) using SANFLOW model, analysing system pressures and checking the efficiency of the billing system. Therefore, monitoring points included bulk meters located at zone inlets, critical points (highest points or farthest point from the source), average zone points and some randomly selected points within the zone where pressure loggers were installed. Figures 35 show the monitoring points for data logging for Areas 15, 18 and 28, respectively.
Figure 3

Spatial location of all the logged points for Area 15.

Figure 3

Spatial location of all the logged points for Area 15.

Figure 4

Spatial location of all the logged points for Area 18.

Figure 4

Spatial location of all the logged points for Area 18.

Figure 5

Spatial location of all the logged points for Area 28.

Figure 5

Spatial location of all the logged points for Area 28.

Selection of monitoring parameters and monitoring period

The selected monitoring parameters included volume of water supplied and consumed in a DMA in a particular time (m3/year, m3/month and m3/day), pressures (m), flows (m3/h), percentage meter error and the billing error factor (BEF) (%). Pressure and flows were monitored over a period of 24 hours for a 5–9 day period to check the daily and weekly variations, as recommended by Farley (2001). The monitoring period for bulk meter and customer meter readings was on a monthly basis. Meter readings on the inlet to the DMA were taken on the 16th day of every month for a period of four months (January–April, 2015). This day corresponded to the beginning of the billing cycle of LWB. Historical data on water produced and consumed were monitored on a yearly basis for a period of seven years to establish trends of UFW for Lilongwe City and were collected from utility records. This seven year period was selected because of the availability of complete data sets. A District Metered Area (DMA) is a hydraulically discrete part of the distribution network that is isolated from the rest of the distribution system.

Selection of model

The models deployed in this study were selected mainly because of their capabilities to partition UFW into various components (SANFLOW model and burst and background estimates (BABE) Water Balance approach). The models have been extensively used in water utilities in South Africa and the United Kingdom, respectively (McKenzie 1999; Thornton et al. 2008). The models also complemented one another as data outputs from other models were data input requirements for other models.

Establishment of UFW trends

To establish UFW trends for the whole city, two data sets were required which included data on yearly water production and billed consumption. The overall UFW for the city was computed as the difference between water produced and water billed, as suggested by Lambert (2003). For the specific study areas, UFW was calculated as the difference between water supplied and water consumed.

Partitioning of UFW

The SANFLOW model was used to partition UFW into real and APL from the analysis of MNFs obtained from flow logging. The lowest flows occurring between 12 a.m. and 4 a.m. were taken as MNFs in the analysis, as proposed by Thornton et al. (2008). Although customer demand is minimal at night, there is still a small amount of flow in the system owing to night-time customer demand for such uses as toilet flushing, washing machines and any other night use. In urban situations, about 6% of the population will be active during the minimum night-time flow period (McKenzie 1999). Equations (1)–(3) explain how RL are computed from MNFs. Apart from the MNF data, the SANFLOW model also requires basic infrastructure variables which are presented in Tables 1 and 2.

Table 1

Infrastructure variables used in SANFLOW model

Infrastructure variable Area 15 Area 18 Area 28 
Length of mains (km) 6.8 39.62 10.83 
Number of connections 278 2,240 141 
Number of properties 278 2,240 141 
Estimated population 2,124 18,846 8,063a 
Density of users per km of mains 41 57 13.3 
Infrastructure variable Area 15 Area 18 Area 28 
Length of mains (km) 6.8 39.62 10.83 
Number of connections 278 2,240 141 
Number of properties 278 2,240 141 
Estimated population 2,124 18,846 8,063a 
Density of users per km of mains 41 57 13.3 

Note: The infrastructure variables are based on actual data in the areas from utility records.

aPopulation for Area 28 is the equivalent population obtained from average consumption for Area 28 and average per capita consumption for Area 18 (i.e., population it could serve assuming it was also a residential area with per capita consumption similar to that of Area 18).

Table 2

Leakage parameters used in SANFLOW model as recommended by McKenzie (1999) 

Description Default Area 15 Area 18 Area 28 
Background losses from mains (L/km.hr) 40 40 40 40 
Background losses from connections (L/conn.hr) 
Background losses from properties (L/conn.hr) 
% of population active during night 
Quantity of water used in a cistern (L) 10 10 10 10 
Background losses pressure exponent 1.5 1.5 1.5 1.5 
Burst/leaks pressure exponent 0.5 0.5 0.5 0.5 
Average use for small non-domestic users (L/h) 50 50 50 50 
Use by large non-domestic users (m3/h) 1.2 1.2 1.2 3a 
Description Default Area 15 Area 18 Area 28 
Background losses from mains (L/km.hr) 40 40 40 40 
Background losses from connections (L/conn.hr) 
Background losses from properties (L/conn.hr) 
% of population active during night 
Quantity of water used in a cistern (L) 10 10 10 10 
Background losses pressure exponent 1.5 1.5 1.5 1.5 
Burst/leaks pressure exponent 0.5 0.5 0.5 0.5 
Average use for small non-domestic users (L/h) 50 50 50 50 
Use by large non-domestic users (m3/h) 1.2 1.2 1.2 3a 

aEstimate based on McKenzie (1999) who established an average 20.5 L/property.h of this loss parameter (i.e., 20.5/1,000*141).

RL were then calculated according to Fanner (2004) as follows: 
formula
1
 
formula
2
 
formula
3
The maximum ENF value was used hence the hour/day factor was taken as 24 as recommended by Lambert (2003). Total apparent losses (TAPL) were then calculated as the difference between the average UFW obtained from water audits and the calculated RL from the SANFLOW model. Thus: 
formula
4
The BABE water balance approach was used to further partition RL into bursts (reported and unreported bursts) and background leakages by establishing the unavoidable annual real losses (UARL). UARL is the lowest technically achievable annual volume of RL for well-maintained and well-managed systems (IWA 2000). 
formula
5
where Lm is mains length (km); Nc is number of service connections; Lp is length of unmetered underground pipe from street edge to customer meters (km), (10 m for LWB) and P is average operating pressure at average zone point (m).
Losses due to pipe bursts, Lb, were then calculated as follows: 
formula
6
where RL is real losses, UARL is unavoidable annual real losses (background leakages).

Partitioning of APL

APL were further partitioned into losses due to meter errors, billing anomalies and unauthorized consumption.

APL due to meter errors

To establish losses due to meter errors, a total of 60 customer meters were randomly uninstalled in Areas 15 (15), 18 (25) and 28 (10) and were tested for accuracy using a meter testing bench at LWB's laboratory. Meters were connected in series and a known volume of water was passed through them. The error was then obtained as the difference between the meter reading and the known volume passed through the meters. As proposed by Arregui et al. (2006), meters were tested at three different flow rates, namely low flow rate (30 L/h), medium flow rate (750 L/h) and high flow rate (1,500 L/h). The meter error, E, was established which helped to estimate losses due to meter inaccuracies. For residential consumption (CR) and meter error (E), the volume lost (Lme) due to meter inaccuracy was estimated according to DWSD (2004) as: 
formula
7

APL due to billing anomalies

Losses due to billing anomalies were determined by conducting an independent monthly meter reading exercise on some sampled customers (59, 132 and 37 for Areas 15, 18 and 28, respectively) in the study areas at exactly the same time when LWB meter readers took their readings. The monthly volume computed from the independent meter reading (VIS) was then compared with the volume as read on the bill delivered to the sampled customers from LWB (VLWB) to establish a BEF which was used to estimate losses due to billing anomalies. For a sample size of n customers, the BEF was then estimated by: 
formula
8
Total water lost due to billing anomalies, Lba, is given by: 
formula
9

APL due to unauthorized consumption

The volume lost through unauthorized consumption was estimated based on the assumption that total apparent loss consists of losses due to meter inaccuracies, billing anomalies and unauthorized consumption. Therefore, losses due to unauthorized consumptions were estimated by subtracting losses due to meter errors and billing anomalies from the TAPL. Thus volume lost through unauthorized consumption can be given by: 
formula
10

where TAPL is total apparent losses, Lme is meter error losses, Lba is billing anomaly losses.

Pressure assessment

According to McKenzie (2001), water reticulation systems are designed to provide a minimum working pressure at all points in the system throughout the day and LWB has set this at 10 m. Trifunovic (2006) highlighted that there is a direct relationship between high pressures and leakages in the distribution system and pressures greater than 60–70 m should not be accepted in the distribution system. Pressures were assessed to check the existence of high pressures which could be contributing to high leakages. This helped to determine those DMA's that required pressure management to reduce leakages.

Pressure analysis included 24 hour temporal pressure variations and spatial pressure variations at two critical periods (average night pressure from 0:00 to 04:00 hours and average peak hour pressure from 06:00 to 08:00 hours, respectively).

RESULTS AND DISCUSSION

UFW trends

UFW trends for entire Lilongwe City

Figure 6 is a plot of UFW.
Figure 6

(a) Volumetric UFW from 2008 to 2014; (b) UFW as a % of total water produced; (c) UFW as loss per connection.

Figure 6

(a) Volumetric UFW from 2008 to 2014; (b) UFW as a % of total water produced; (c) UFW as loss per connection.

The results in Figure 6 show that from 2008 to 2014 an average of 11.9 million m3/year of water was unaccounted for Lilongwe City making an UFW of 37.5% on the volume produced. Most customers in the supply area of LWB fall into the consumption block of ‘in excess of 10 m3/month’ and the tariff for this block is USD 0.67/m3 (LWB 2014a). Using this tariff, the average UFW translates to a loss in revenue of USD 7.9 million/year.

Figure 6(b) shows that there was a general decrease in UFW from 42.7% in 2008 to 35% in 2014 although there was no defined pattern since in the years 2012 and 2013, UFW rose again. Figure 6(c) showed that UFW expressed as loss/connection showed a similar trend with an average of 316.19 m3/h/connection. The average UFW of 37.5% is higher than the recommended less than 23% as suggested by Tyanan and Kingdom (2002) for the best performing utilities in developing countries. This analysis concluded that LWB is underperforming in terms of managing its UFW.

UFW trends for the study areas

UFW trends for the study areas were established as explained in the section ‘Establishment of UFW trends’. Areas 15 and 18 have two inlets to the DMA and total volume of water supplied was obtained by summing the volumes obtained from readings on each inlet bulk meter. Area 28 has one inlet into the zone and bulk meter readings were taken at this zone inlet. UFW trends for the study areas were analysed for a period of three months and Table 1 summarizes the water audit results.

Results in Table 3 show that Area 18 registered the highest UFW during the three month period, averaging 44% while Area 28 had the least (20%). The UFW expressed as m3/Km.month also showed highest value in Area 18 and least value in Area 28 registering 987 and 507 m3/Km.month, respectively. This is consistent with the indication drawn from the UFW as a percentage in each of the three areas. It can be concluded that Areas 15 and 18 were underperforming and Area 28 was performing well in terms of management of UFW.

Table 3

Summary of average UFW in the study areas for the period from January to March, 2015

Area Bulk meter readings (m3/month)
 
Billed consumption (m3/month)
 
UFW (m3/month)
 
Average UFW
 
Jan 15 Feb 15 Mar 15 Jan 15 Feb 15 Mar 15 Jan 15 Feb 15 Mar 15 m3/ month % of vol supplied m3/ km montha 
Area 15 14,713 13,922 19,063 9,957 9,957 12,677 4,756 4,756 6,386 5,239 33 770 
Area 18 93,024 89,804 86,770 52,982 50,783 47,953 40,042 39,021 38,817 39,293 44 987 
Area 28 27, 263 29,200 24,935 21,609 23,323 19,981 5,654 5,877 4,954 5,495 20 507 
Area Bulk meter readings (m3/month)
 
Billed consumption (m3/month)
 
UFW (m3/month)
 
Average UFW
 
Jan 15 Feb 15 Mar 15 Jan 15 Feb 15 Mar 15 Jan 15 Feb 15 Mar 15 m3/ month % of vol supplied m3/ km montha 
Area 15 14,713 13,922 19,063 9,957 9,957 12,677 4,756 4,756 6,386 5,239 33 770 
Area 18 93,024 89,804 86,770 52,982 50,783 47,953 40,042 39,021 38,817 39,293 44 987 
Area 28 27, 263 29,200 24,935 21,609 23,323 19,981 5,654 5,877 4,954 5,495 20 507 

aAreas 15 and 18 are residential while Area 28 is industrial. As such, the density per users for Area 28 is affected by fewer connections in the industrial area as the plot sizes are large resulting in a low density of users, hence assessing loss per density of users could give a distortion. Instead, analysis of loss per unit length (per km) has been added.

Partitioning of UFW

Partitioning of UFW into real and APL

SANFLOW model was used to partition UFW into real and APL as explained in the section ‘Partitioning of UFW’. Data logging results in Figures 79 show the average MNF and peak hour demands (PHDs). Both inlets for Area 15 and 18 (residential areas) showed a similar trend where MNFs occurred between midnight and 04:00 hours and PHDs occurred between 06:00 hours and 08:00 hours. The trend for Area 28 (industrial area) was different from the others, where PHDs mostly occurred after midday. This was explained by the possibility that different industries had different times for their PHDs. APL were then calculated as explained in the section ‘Partitioning of UFW’. Table 4 summarizes the results from partitioning of UFW into real and APL.
Table 4

Summary of results from partitioning of UFW

  Avg UFW m3/month Avg MNF m3/h Avg ENF m3/h Avg RL m3/month Avg TAPL m3/month RL as % of UFW APL as a % of UFW 
Area 15 5,239 8.1 5.96 4,221 1,018 81% 19% 
Area 18 39,293 102 47.03 33,862 5,431 86% 14% 
Area 28 5,495 1.98 1.471 1,145 4,436 21% 79% 
  Avg UFW m3/month Avg MNF m3/h Avg ENF m3/h Avg RL m3/month Avg TAPL m3/month RL as % of UFW APL as a % of UFW 
Area 15 5,239 8.1 5.96 4,221 1,018 81% 19% 
Area 18 39,293 102 47.03 33,862 5,431 86% 14% 
Area 28 5,495 1.98 1.471 1,145 4,436 21% 79% 
Figure 7

(a) Flow/pressure variation for: (a) Lingazi inlet – Area 15 and (b) Chitukuko inlet – Area 15.

Figure 7

(a) Flow/pressure variation for: (a) Lingazi inlet – Area 15 and (b) Chitukuko inlet – Area 15.

Figure 8

(a) Flow variations for: Kanengo zone inlet – Area 18 and (b) Mtunthama zone inlet – Area 18.

Figure 8

(a) Flow variations for: Kanengo zone inlet – Area 18 and (b) Mtunthama zone inlet – Area 18.

Figure 9

Flow variations – Area 28 zone inlet.

Figure 9

Flow variations – Area 28 zone inlet.

This analysis concluded that UFW was dominated by RL in Areas 15 and 18, while in Area 28 UFW was dominated by APL. Area 15 had high RL probably because of aged infrastructure (41 years) while high RL for Area 18 may be attributed to high system pressures. Reduced RL for Area 28 may be attributed to a young system of only 20 years and low system pressures.

Partitioning RL into bursts and background leakages

As suggested by Mckenzie (2001), BABE concepts were used to further partition RL into bursts and background leakages by computing the UARL, as explained in ‘Partitioning of UFW. Table 5 summarizes the partitioning of RL into bursts and background losses.

Table 5

Summary of calculations for losses due to bursts (Lb) for all the study areas

Area RL m3/month UARL m3/month Bursts (Lb) m3/montha UARL as % of RLb Lb as % of RL 
Area 15 4,221 415 3,806 10% 90% 
Area 18 33,862 4,260 29,602 13% 87% 
Area 28 1,059 370 689 35% 65% 
Area RL m3/month UARL m3/month Bursts (Lb) m3/montha UARL as % of RLb Lb as % of RL 
Area 15 4,221 415 3,806 10% 90% 
Area 18 33,862 4,260 29,602 13% 87% 
Area 28 1,059 370 689 35% 65% 

aBursts = reported and unreported leakages (Thornton et al. 2008).

bUARL = background leakages.

This analysis concluded that bursts (reported and unreported leakages) dominated RL in all the study areas, contributing 90%, 87% and 65% of the total RL for Areas 15, 18 and 28, respectively. Background leakages were minimal in all the areas.

Partitioning of APL

APL due to meter errors. Meter testing results showed that the average meter errors at low flow rate, medium flow rate and high flow rate tests were 0.005%, 1.2% and 1.1%, respectively. According to Pack (1997), between minimum flow rate, Qmin, and transitional flow rate, Qt, the allowed meter error is ±5% while from transitional flow rate, Qt, up to maximum flow rate, Qmax, the allowed meter error is ±2%. Therefore, it can be concluded that meter errors from the study areas were within the acceptable limits at all the three flow rate tests. Therefore, contribution of meter errors to APL was insignificant as suggested by Pack (1997) and Arregui et al. (2006).

APL due to billing anomalies and unauthorized consumption. An independent meter reading exercise was conducted on 59, 132 and 37 customers for Areas 15, 18 and 28, respectively, at the same time, when LWB meter readers carried out the exercise. This was done to establish a BEF which was used to estimate losses due to billing anomalies as explained in the section ‘Partitioning of apparent losses’. The volume lost through unauthorized consumption was estimated based on the fact that TAPL consist of losses due to meter inaccuracies, billing anomalies and unauthorized consumption, and that contribution of meter errors to APL was insignificant. Table 6 summarizes partitioning of APL in all the study areas.

Table 6

Estimation of APL in all the study areas

Area TAPL (m3/month) BEFavg Lba (m3/month) Lme (m3/month) Luc as a % of TAPL Luc (m3/month) 
Area 15 445.49 7.2% 32.08 92.8% 413.41 
Area 18 5,431.40 32.2% 1,748.91 67.8% 3,682.49 
Area 28 4,706.07 17.6% 828.27 82.4% 3,877.80 
Area TAPL (m3/month) BEFavg Lba (m3/month) Lme (m3/month) Luc as a % of TAPL Luc (m3/month) 
Area 15 445.49 7.2% 32.08 92.8% 413.41 
Area 18 5,431.40 32.2% 1,748.91 67.8% 3,682.49 
Area 28 4,706.07 17.6% 828.27 82.4% 3,877.80 

Lba=losses due to billing anomalies, Lme = losses due to meter errors, Luc = losses due to unauthorized consumption.

System pressure assessment and pressure management

Pressures were assessed as explained in the section ‘Pressure assessment’ to check for existence of high pressures which could be contributing to high leakages. For DMAs with high pressures, the study proposed implementation of an active pressure management as recommended by Mckenzie (2001). This includes installation of pressure reducing valves (PRV) (fixed outlet PRV, time modulated PRV and flow modulated PRV).

Pressure analysis for Area 15

Results for pressure analysis for Area 15 are shown in Figures 10 and 11. These figures show temporal and spatial pressure variations, respectively.
Figure 10

Twenty-four hour pressure variations for all pressure logged points in Area 15 (31/01/2015).

Figure 10

Twenty-four hour pressure variations for all pressure logged points in Area 15 (31/01/2015).

Figure 11

Spatial variation of pressures for all logged points at critical times (31/01/2015): (a) night and (b) peak hour.

Figure 11

Spatial variation of pressures for all logged points at critical times (31/01/2015): (a) night and (b) peak hour.

The 24 hour pressure variations in Figure 10 for each of the eight logged points showed a general similar trend where high pressures were experienced between midnight and 4:00 hours (at a time when water usage was very minimal) and low pressures were experienced between 6:00 hours and 8:00 hours (peak hour demand period). The spatial pressure variations show that both at night and at peak hour, pressures are moderately high mostly ranging from 40 to 50 m during the night (Figure 11(a)) and 30 to 40 m during the peak demand period (Figure 11(b)).

The 24 h pressure trends showed that 60% of the logged points had their pressures above the minimum required of 10 m throughout the day, while 40% experienced pressures lower than the minimum required during some hours of the day. The 24 hour average pressure at the critical point (highest point) was 16 m and the 24 hour average zone pressure was 40 m. This pressure analysis concluded that Area 15 generally experienced moderately high pressures which were within the acceptable limits and hence it was unlikely that high pressures were a major cause of leakages, as suggested by Trifunovic (2006). Pressure management was therefore not applicable for Area 15.

Pressure analysis for Area 18

Results for pressure analysis for Area 18 are presented in Figures 12 and 13. The 24 hour pressure variations for all the ten pressure logged points in Figure 12 showed high pressures between 0:00 hours and 04:00 hours and low pressures at peak hour demand (06:00–0:800 hours). The temporal and spatial pressure trends also showed that all the pressure logged points had their pressures above the minimum required level of service of 10 m throughout the day. The critical point, whose pressures are generally expected to be very low in most water distribution systems (McKenzie 2001), had an average 24 h pressure of 29 m which was far above the minimum required of 10 m for LWB. The average zone pressure for the DMA was 51 m. The spatial pressure variations in Figure 13(a) showed that during the night, unacceptable high pressures ranging from 50 to 80 m were predominant in the DMA. During peak hours (Figure 13(b)), acceptable high pressures (40–50 m) dominated the DMA followed by unacceptable high pressures (50–80 m). This indicated that at both critical periods, the DMA experienced unacceptably high pressures, as highlighted by Trifunovic (2006).
Figure 12

Twenty-four hour pressure variations for all logged points in Area 18 (21/02/2015).

Figure 12

Twenty-four hour pressure variations for all logged points in Area 18 (21/02/2015).

Figure 13

Spatial variation of pressures at night and at peak hour (21/02/2015): (a) night and (b) peak hour.

Figure 13

Spatial variation of pressures at night and at peak hour (21/02/2015): (a) night and (b) peak hour.

Field reconnaissance surveys also indicated frequent occurrences of pipe bursts in the DMA, which were probably attributed to high system pressures, among other reasons. Therefore, it was concluded that Area 18 experienced unacceptable high pressures which were probably contributing to high leakages, explaining the high RL in the DMA of 86%. McKenzie (2001) recommends that pressure management should be applied to the DMA with MNF of 20 m3/h or above if its application is to yield optimum potential savings. In the section ‘Partitioning of UFW into real and apparent losses’, it was found that MNF for Area 18 was 102 m3/h. Therefore, high pressures coupled with high MNF makes active pressure management to be the best option of reducing RL in Area 18. Active pressure management for Area 18 would involve installation of PRV.

Pressure analysis in Area 28

Figure 14 shows 24 hour pressure variations for all the five logged points in the DMA and Figure 15 shows the spatial variation of pressures for Area 28.
Figure 14

Twenty-four hour pressure variations for all logged points – Area 28 (27/01/2015).

Figure 14

Twenty-four hour pressure variations for all logged points – Area 28 (27/01/2015).

Figure 15

Spatial variations of pressures at two critical periods (27/01/2015): (a) night and (b) peak hour.

Figure 15

Spatial variations of pressures at two critical periods (27/01/2015): (a) night and (b) peak hour.

The 24 hour pressure variations showed that 60% of the logged points for Area 28 had their pressures above the minimum allowable of 10 m throughout the day while 40% of the logged points experienced pressures lower than the minimum allowable in some hours of the day. The average pressure at a critical point was 10 m corresponding to the minimum allowable pressure as recommended by LWB. The average zone pressure was 40 m. The spatial pressure variations showed that, unlike in Areas 15 and 18, night pressures and peak hour pressures were predominantly within the same pressure range of 30–40 m. This pressure analysis concluded that Area 28 generally experienced acceptable pressures and hence it was unlikely that high pressures were a major cause of leakages in Area 28. Pressure management was therefore not applicable for Area 28.

CONCLUSIONS AND RECOMMENDATIONS

Conclusions

The study drew the following conclusions:

  • (1) The average UFW for Lilongwe City (37.5%) was found to be higher than that recommended for developing countries (23%). In the specific study areas, UFW was found to be higher than recommended in Area 15 (33%) and Area 18 (44%) but within the acceptable limit in Area 28 (20%).

  • (2) RL formed a major component of UFW for Areas 15 and 18 (81% and 86%, respectively) while APL were a major component of UFW for Area 28 (80%).

  • (3) High system pressures were the main drivers of high RL in Area 18 while aged infrastructure (41 years) was the main driver for RL in Area 15. The main driver for APL in all the three areas was unauthorized consumption.

Recommendations

The study made the following recommendations:

  • (1) LWB should implement active pressure management in areas of high pressures. For ease of monitoring, the entire supply area should be demarcated into DMA installed with functioning bulk meters.

  • (2) LWB should establish an active leak detection programme on the main and service lines, joints and fittings and repair all the leakages discovered. This should be an ongoing programme prioritizing the old sections of the distribution system.

  • (3) LWB should consider ‘modelling of water losses’ to form an integral part in its water loss management plans and strategies. There is a need for continued MNF assessment to come up with real loss patterns and further partitioning of UFW in the remaining areas.

REFERENCES

REFERENCES
Arregui
F. J.
Cabrera
E.
Cobacho
R.
García-Serra
J.
2006
Reducing Apparent Losses Caused by Meters Inaccuracies
.
Polytechnic University of Valencia
.
46022
Valencia
,
Spain
.
(accessed 11 June 2015)
.
Balkaran
C.
Wyke
G.
2007
Managing water loss: strategies for the assessment, reduction and control of non-revenue water (NRW) in Trinidad and Tobago. Available at: www.naruc.org/international/Documents/Wyke1.pdf
(accessed 23 June 2015)
.
Brunson
L.
Busenitz
L.
Sabatini
D.
Spicer
P.
2013
In pursuit of sustainable water solutions in emerging regions
.
J. Water Sanit. Hyg. Dev.
3
(
4
),
489
499
.
Chipwaila
J. A.
2009
Partitioning of Unaccounted for Water for Zomba City Supply System in Malawi
.
MSc. Thesis
,
Integrated Water Resources Management, University of Zimbabwe
,
Harare
, p.
131
.
(accessed 11 May 2015)
.
DWSD (Detroit Water and Sewerage Department Final report)
2004
An Evaluation of Unaccounted-for Water Comprehensive Water Master Plan DWSD Contract No. Cs-1278
.
Detroit Water and Sewerage Department
,
New York
,
USA
.
EPA
2010
Control and mitigation of drinking water losses in distribution systems
.
United States Environmental Protection Agency (EPA)
.
(accessed 14 June 2015)
.
Fanner
P.
2004
Assessing Real Losses, Including Component Analysis and Economic Considerations: a Practical Approach
.
Water 21 Article No 6
,
IWA Publishing
,
London
,
UK
.
Farley
M.
2001
Leakage Management and Control. A Best Practice Training Manual
.
World Health Organization
,
Geneva
,
Switzerland
.
(accessed 1 January 2015)
.
IWA
2000
Losses from water supply systems: standard terminology and recommended performance measures. IWA Blue Pages. International Water Association
.
(accessed 17 June 2015)
.
Kafodya
V. J.
2010
Water Loss Management in Blantyre City, Malawi
.
MSc Thesis
,
UNESCO-IHE Institute of Water Education
,
Delft
,
the Netherlands
.
Kingdom
B.
Liemberger
R.
Marin
P.
2006
The challenge of reducing non-revenue water (NRW) in developing countries. How the private sector can help: a look at performance-based service contracting
.
Water Supply and Sanitation Sector Board Discussion Paper Series
8
,
1
52
.
Lambert
A. O.
2003
Assessing non-revenue water and its components: a practical approach
.
Water
21
,
50
51
.
LWB (Lilongwe Water Board)
2011
Five Year Corporate Plan (2012–2017)
.
Lilongwe Water Board
,
Lilongwe
,
Malawi
.
LWB (Lilongwe Water Board)
2014
Sales Report for 2014
.
Lilongwe Water Board
,
Malawi
.
LWB (Lilongwe Water Board)
2014a
Waterworks (Lilongwe Water Board) (Amendment) By-laws, 2014
.
Lilongwe Water Board
,
Lilongwe
,
Malawi
.
McKenzie
R.
1999
SANFLOW User Guide
.
South Africa Water Research Commission
,
WRC Report No. TT 109/99
,
Pretoria
,
South Africa
.
McKenzie
R.
2001
Pressure Management Programme (PRESMAC) Modelling User Guide Manual
,
Report No. TT152-01
,
South African Water Research Commission
,
Pretoria
,
South Africa
.
Mulwafu
W.
Chavula
G.
Chipeta
C.
Ferguson
A.
Chilima
G.
Nkhoma
B.
2002
The Status of Water Demand Management in Malawi and Strategies for Promoting IT
.
IUCN Country Report
.
Nkhoma
B.
Hoko
Z.
Gumbo
B.
2005
Assessment of leakage control as a tool of infrastructure water system demand management: A case of area 49 in the City of Lilongwe, Malawi. Water & wastewater management for developing countries (WAMDEC 2004)
. In:
Selected Proceedings of the Water & Wastewater Management for Developing Countries Conference
, pp.
28
30
.
NSO (National Statistical Office)
2008
Population and Housing Census
.
Preliminary Report
.
Government of Malawi
,
Zomba
,
Malawi
.
Pack
G.
1997
Water Meter Classification. Available at: www.rsfor.narod.ru/datasheets
(accessed 11 June 2015)
.
SOGREAH
2010
SOGREAH Consultants, 2010. Feasibility Studies and Preliminary Design for Lilongwe's new Water Source
.
Water Demand Assessment Report
.
Lilongwe Water Board
,
Lilongwe
,
Malawi
.
SP (Studio Pietrangeli)
2015
Studio Pietrangeli (SP) Consulting Engineers. Detailed Design of Diamphwe Multipurpose Dam and Associated Structures
.
National Water Development Programme, Ministry of Irrigation and Water Development, Government of Malawi
.
Thornton
J.
Sturm
R.
Kunkel
G.
2008
Water Loss Control
.
2nd edn
.
McGraw-Hill
,
New York
,
USA
.
Trifunovic
N.
2006
Introduction to Urban Water Distribution, UNESCO-IHE Lecture Note Series
.
Institute for Water Education
,
Delft
,
the Netherlands
.
Tyanan
N.
Kingdom
B.
2002
A Water Scorecard – Setting Performance Targets for Water Utilities
.
World Bank Private Sector Note 242
.
WHO/UNICEF JMP
2014
World Health Organization and United Nations Childrens Fund Progress Report on Sanitation and Drinking Water, 2014 update
.
(accessed 11 June 2015)
.