Journal of Hydrology 381 (2010) 181–188

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Optimal sizing of rain water tanks for domestic water conservation

Anirban Khastagir *, Niranjali Jayasuriya
School of Civil, Environmental and Chemical Engineering, RMIT University, Melbourne, Australia

a r t i c l e i n f o s u m m a r y

Article history: Melbourne is facing a severe drought having its 12th consecutive below average rainfall year. Water
Received 26 May 2009 authorities have been forced to impose rigorous water restrictions including voluntary per capita water
Received in revised form 15 November 2009 use targets after more than 20 years of unrestricted water supply. The current severe drought and dwin-
Accepted 22 November 2009
dling water resources have accelerated the use of alternative water sources including domestic rainwater.
There is a large variation in average annual rainfall in the Greater Melbourne area ranging from 1050 mm
This manuscript was handled by in the east to 450 mm in the west. Hence, there is a significant difference in the tank size required in the
K. Georgakakos, Editor-in-Chief west and the east of Melbourne to meet a similar demand and to provide the same supply reliability. The
paper presents a novel methodology and a relationship for optimal sizing of rainwater tanks considering
Keywords: the annual rainfall at the geographic location, the demand for rainwater, the roof area (catchment area)
Rainwater tank and the desired supply reliability. The characteristic of the developed dimensionless curve reflects these
Reliability variables and paves the way for developing a web based interactive tool for selecting the optimum rain-
Spillage water tank size.
Usage Ó 2009 Elsevier B.V. All rights reserved.
Melbourne

Introduction lished and relatively common practice, particularly in rural and re-
mote areas. Between 1994 and 2001, 13% of rural Victoria
Melbourne is world renowned for having reliable high quality households used rainwater tanks, with 11% of them using rain
drinking water. Similar to many other developed cities in the water as their prime drinking source. However, rainwater tanks
world, it has to confront a growing water demand due to ever are still used primarily in non-metropolitan regional cities (36%)
increasing population and continuous economic development in when compared with capital cities (3%) (ABS, 2001).
a diminishing water resources paradigm due to climate change. The Victorian Government introduced town planning legisla-
On top of this, Melbourne is facing a severe drought having its tion throughout Victoria for the installation of rainwater tanks
twelfth consecutive below average rainfall year. The Victorian Gov- up to 4500 L capacity except in heritage zones in September
ernment released its water strategy ‘Our Water Our Future’ in 2006 2001 (Urban Rainwater Systems, 2007). Amendments to the Water
setting a target of reducing per capita water consumption by 15% Act (1989) passed in April 2002 explicitly provided the property
by 2010 (Department of Sustainability and Environment, 2006). owners continued rights to the unrestricted use of rainwater for
The Government has further stretched the water conservation tar- domestic purposes on their property free of charge (Urban Rainwa-
get to 25% and 30% by 2015 and 2020 respectively. ter Systems, 2007). Victorian Government policy supporting the
Stormwater is a valuable resource that is currently under used. use of rainwater tanks for potable substitution whilst meeting
With the persistent drought the use of rainwater tanks is increas- the Department of Human Services public health guidelines and
ingly becoming popular. Water from the tanks can be used for non- the Plumbing Industry regulations is reflected in ‘‘Water Smart
potable purposes such as garden use, toilet flushing, washing Gardens and Homes Rebate Scheme” which allows every house-
clothes and in the hot water systems. This constitutes about 80% hold an opportunity to conserve their water resources and save
of the water consumed within a residential property. In Australia, money (Department of Sustainability and Environment, 2007).
the majority of water consumption is for outdoor use, hot water The Government has committed $10 million over 4 years to pro-
and toilet flushing (Coombes, 2002). Many households have shown vide incentives for Victorians connected to a reticulated water sup-
their willingness to use rainwater as an alternative source of water ply to conserve water resources. It is evident that the ‘one size fits
even in areas that are serviced with reticulated water supply. In the all’ approach to rainwater tank selection is flawed when one con-
state of Victoria, the use of domestic rainwater tanks is an estab- siders the rainfall characteristics within the Greater Melbourne
Area as the mean annual rainfall varies significantly with location.
For example, the rainfall is around 450 mm/year in the west,
* Corresponding author. Tel.: +61 (3) 99253228.
E-mail address: [email protected] (A. Khastagir). 750 mm/year in the east and 1050 mm/year in the north east.

0022-1694/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2009.11.040
182 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188

As far as rainfall is considered, currently the whole of the Greater balance equation used for the study is given in Eq. (1) as discussed
Melbourne area is treated as one homogeneous entity in rainwater in Khastagir and Jayasuriya (2007).
tank design. A house owner who is a potential customer for a rainwa-
St ¼ Stþ1 þ Q t  Dt 0 6 Stþ1 6 C ð1Þ
ter tank has very little guidance when selecting the rainwater tank
other than the area available to locate it, the aesthetic issues and where St+1 is the storage volume in the tank at the end of tth day, St,
the money he is willing to spend on it. However, there are a number the storage value at the beginning of tth day, Qt, the runoff from the
of other variables that directly influence the decision. These include: roof into the tank on the tth day, Dt, the total demand for water on
the rainfall in the area, the extent of the catchment (connected roof the tth day and C is the active tank capacity.
area) and the type of end use planned for the water. There are no set As the first step, the tank capacity C was assumed. The daily
guidelines to select the optimum size of the rainwater tank for the runoff from the roof (Qt) depends on the daily precipitation.
Greater Melbourne area depending on the above mentioned vari- The daily water demand (Dt) depends on outdoor and indoor
ables. The Water Sensitive Urban Design Manual (2005) provides use of rainwater on number of factors. Eq. (1) was applied at
some guidance when selecting tank sizes for three different hydro- the end of each daily time step to obtain the water storage level.
logical regions in Victoria. However, the guidance remains very On a particular day if the water storage level (St+1) was greater
coarse as the whole Victorian state is divided into three regions than the tank capacity (C) the excess water will spill over and
ignoring the significance of rainfall variability as highlighted earlier the tank storage level at the end of the day will be reset equal
for Metropolitan Melbourne. to C. The amount of water spilled is calculated using Eq. (2).
The primary objective of this study is to develop a relationship to The probability of tank having sufficient water to meet the de-
select the optimal rainwater tank size depending on the annual rain- mand is given as reliability Eq. (3). If the required reliability
fall at the location, the demand for rainwater and the catchment size was not achieved with the assumed tank capacity, a new tank
based on the supply and reliability desired. The decision related to size (C) was assumed and the above procedure repeated until
determining the size of the rainwater tank to be installed in a prop- the required reliability level was achieved. Usage of water can
erty is ultimately a compromise between maximizing the use of the be calculated from Eq. (4).
rainwater, the reliability of meeting the demand throughout the cal-
endar year, the investment required and the space available within Spill on the tth day ¼ Stþ1  C ð2Þ
the property to establish the tank. All of these factors influence the
final property owner decision. The paper presents the development P
Re ¼  100 ð3Þ
of a set of generalized reliability curves based on information gath- N
ered from rain gauges spread across Melbourne. The characteristics
of this curve will include all decision making variables (demand and Usage ¼ Demand  P ð4Þ
reliability), features and geographic locations of individual house- where Re is the probability of the tank being not empty as a per-
holds (roof area and average annual rainfall) and tank size in a single centage (reliability), P, the number of days the tank is not empty,
dimensionless curve. The development of a curve that represents the N, the total number of days and demand is the total water de-
complex interdependent variables mathematically enables automa- mand/year.
tion of the optimal tank size selection process, significantly benefit-
ing the consumer by maximizing the return on investment.
Determination of roof runoff (Q)

Overview of study areas The volume of water that could be collected is important to
optimize the size of the rainwater tank for sustainable water use.
The data base for the study consists of 20 stations distributed Eq. (5) was used to calculate the roof runoff volume.
across the Greater Melbourne metropolitan area (Fig. 1). The vari-
ation of annual rainfall is shown in Fig. 2. The daily rainfall data for Q ¼ Ieff  C R  A ð5Þ
the last 30 years were obtained from Melbourne Water and used
for the analysis. It can be observed from Figs. 1 and 2 that there where Q is the daily runoff (L), Ieff, the daily effective rainfall (mm)
is a significant variation in rainfall when one moves from the east calculated as given in Eq. (6), CR, the coefficient of runoff, and A is
to the west of Melbourne. the roof area connected to the tank (m2).
The stormwater quality of the initial discharge from the roof
surface is of poor quality due to accumulation of dust, sediments,
Estimation of rainwater tank size bird and animal droppings, and leaves and debris from the sur-
rounding areas. It is necessary to initially waste a fixed portion
Development of the rainwater tank selection model of rainfall from a storm event to improve water quality. This is
called the first flush when calculating the discharge into the rain-
A simple daily water balance model was developed to calculate water tank. Yaziz et al. (1989) and Jenkins and Pearson (1978)
the rainwater tank size (Eq. (1)). The volume of rainwater in the noted that the first flush contained large amounts of dust, animal
tank depends on the volume of water flowing into the tank and droppings and debris. Coombes (2002) verified this observation
the demand met from the rainwater as an alternative water source while monitoring the roof water in Figtree place. However, there
to conventional supply. It is important to ensure that there is en- are differences of opinion regarding the amount of first flush to
ough water in the tank to supply the demand with minimum risk be kept out of reuse to ensure acceptable water quality. A runoff
of the tank being empty (i.e. maximize reliability). The daily stor- coefficient value of 0.8 was used in the study to account for loss
age level of the water tank would depend on the frequency and of water due to evaporation and minor infiltration from the roof
the amount of precipitated rainfall and the end use. A daily time surface (Lancaster, 2006; Department of Environmental and
period was considered for the study as it is important to ensure Aboriginal affairs Government of South Australia, 1999). Further-
that there is sufficient water to meet the domestic demand con- more in some instances there will be overflow from roof gutters.
nected to rainwater use as far as possible. Evaporation of water Yaziz et al. (1989) reported that subtracting the first 0.33 mm of
from the tank was not considered as the tank is closed. The water rainfall from the total daily rainfall as the first flush would
 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188 183

Fig. 1. Locations of the rainfall stations used in this study.

1400
Mean annual rainfall (mm)

1200

1000

800

600

400

200
Werribee

Rockbank

Altona

Sunshine

St. Albans

Arthurs C reek

Caufield

Hampton

Mountview

Kew

Sandringham

Berwick

Caulfield North

Surrey Hills

Eastern Golf Club

Cranbourne

Notting hill

MItcham

Kinglake
Doncaster

Rainfall stations

Fig. 2. Variation in annual rainfall values for the 20 rainfall stations used in the study.

significantly improve roof water quality and his recommendation Daily effective rainfall ðIeff Þ ¼ Daily rainfall  first flush ð6Þ
was adopted for this study. Eq. (6) was used to calculate the daily
effective rainfall. where first flush = 0.33 mm.
184 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188

Determination of demand for water (Dt) the optimum tank size depending on the location, the demand
for rainwater use, the roof area and the reliability of supply.
The daily water demand (Dt) from the tank will depend on a
number of factors. It is legislated in Melbourne that rainwater Derivation of dimensionless numbers
could be used for toilet flushing, laundry use, hot water systems
and for garden watering (all non-potable use). As a result, the Buckingham developed Buckingham p theorem in 1915 to de-
water demand will depend on the number of factors such as num- rive dimensionless variables from a set of independent variables
ber of occupants in the house, the garden size and the prevailing (Crowe et al., 2007). Eq. (8) gives the number of dimensionless
weather. Only three demand types (toilet, garden and laundry) numbers that could be obtained from a set of independent
were considered in the study. The water used in the hot water sys- variables.
tem was not considered in the analysis considering the fact that
percentage of users connected with the hot water system at pres- Number of dimensionless variables; p¼nm ð8Þ
ent is very small. Water demand depends on the number of occu-
where n is the the number of independent parameters, and m is the
pants in the house especially for toilet and laundry use. Based on
basic dimensions in the independent parameters.
Gato et al. (2004) recommendations, 16 L/person/day (Lpcd) and
Crowe et al. (2007) stated that there are two methods to carry
39.7 Lpcd were taken as the typical demand for toilet and laundry
out the dimensionless analysis: step by step method and the expo-
use respectively in this Melbourne study. However, garden water-
nent method.
ing depends on the season whether it is summer or winter as well
In this study the exponent method was used to carry out the
as garden size. Roberts (2004) reported that on average for all gar-
dimensionless analysis. The characteristics of this method require
den types the water use was 69 kL/year. This reported value was
solving a number of simultaneous equations after identifying the
obtained based on data collected prior to water restrictions where
significant variables. These equations were derived from the
people could water the lawns and plants all 7 days of the week.
requirement of dimensional homogeneity of equations describing
However, under Stage 3a water restrictions the residential water
the physical systems.
consumption has reduced significantly. As a result, 69 kL/year
The following parameters were considered when developing
water usage for garden watering is insignificant in this present
dimensionless numbers for the study:
water restriction period. In this study it was decided that the de-
mand garden watering will be 382 L/day for 6 months and used
 Tank capacity (C), L3.
over 2 day within a week.
 Annual water demand (D), L/T.
Total indoor demand was calculated using:
 Roof area (A), L2.
 Mean annual rainfall (R), L/T.
Total indoor demand ðDÞ ¼ Demand ðL=p=dayÞ  No of persons
 Supply reliability (Re) % (dimensionless).
ð7Þ
For current study the household size was assumed to be be- The derived dimensionless numbers are:
tween 1 and 3 persons. This range was taken based on data col- C D
lected by Gato et al. (2004). p1 ¼ 3
; p2 ¼ and p3 ¼ Reliability ð9Þ
A 2 AR

The relationship between the tank size and reliability for Relationship between the dimensionless numbers
different rainfall stations in the study area
Rainwater tank sizes were calculated using the water balance
The primary objective of using rainwater harvested in tanks for model for each rain gauge station with different demand types
domestic use is to conserve the potable water supply. In addition to and from different roof sizes whilst varying the supply reliability
the rainfall in the area, the tank size varies with the roof area, the between 85% and 95%. The mean annual rainfall at selected sta-
demand for water considered and the reliability of supply desired tions across Melbourne varied between 450 mm and 1050 mm.
by the user. The supply reliability of the rainwater tank is extre- The demand for rainwater was calculated using the following
mely important for domestic water conservation as it reflects combinations:
whether the tank has sufficient water to meet the demand placed
on the tank. The supply reliability of rainwater tanks in different  Toilet only.
parts of Melbourne was considered as a variable when optimally  Garden only.
sizing the tanks. Figs. 3–5 show the variation in tank sizes for  Laundry only.
90% reliability when rainwater is harvested for toilet and garden  Toilet and laundry.
use from typical roof sizes of 100 m2, 150 m2 and 200 m2  Toilet and garden.
respectively.  Laundry and garden.
From the above figures, it is evident that in addition to the dis-  Toilet, garden and laundry.
tribution of rainfall, the roof area is also an important factor to be
considered in determining the optimal tank size. Similar results The demand was also calculated by varying the number of peo-
were obtained by varying the demand and supply reliability. As ple in the dwelling between 1 and 3 using per capita water use
such, a number of charts similar to above figures (Figs. 3–5) would data published by Gato et al. (2004). The roof size of an individual
be necessary to cover the study area to select the optimal rainwa- household was varied between 100 m2 and 250 m2 in establishing
ter tank size to meet the customer’s selected demand and expected relationships between tank capacity (C), mean annual rainfall of
reliability. To avoid having to develop vast numbers of nomograms, the location (R), consumptive demand (D) and roof area (A) and
the researchers decided to explore the use of dimensionless curve supply reliability (Re). Dimensionless numbers (Eq. (9)) were cal-
analysis with all above variables with a view to reducing the num- culated for each separate station with different combinations of
ber of independent variables in the analysis. The derived dimen- D, A and tank capacity obtained from the analysis. The tank sizes
sionless numbers were used to develop a set of curves to obtain were limited to a maximum of 20 kL as the study concentrates
 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188 185

Kinglake

Arthurs Creek

5.8
5.4
Rockbank 5
4.6
St. Albans
4.2
Sunshine Doncaster east 3.8
Eastern golf club
3.4
Northing

Kew Mitcham 3
Surrey Hills 2.6
Altona Mountview 2.2
Caulfield
Caulfield North Notting Hill 1.8
Werribee 1.55
Hampton
1.45
Sandringham Berwick
1.2
0.8

Cranbourne

Easting

Fig. 3. Variation of optimum tank sizes for a water supply reliability of 90% from a roof size of 100 m2 with a water demand for toilet flushing and garden watering.

Kingllake
Arthurs creek

3.4

3.1
Rockbank 2.8
St. Albans

Sunshine 2.5
Northing

Kew Doncaster east

Eastern golf club
Mitcham 2.2
Surrey Hills
1.9
Altona
Mountview
Caulfield 1.6
Caufield north Notting hill
Hampton
Werribee 1.3
Sandringham Berwick 1

0.8

Cranbourne

Easting

Fig. 4. Variation of optimum tank sizes for a water supply reliability of 90% from a roof size of 150 m2 with a water demand for toilet flushing and garden watering.

on determining the optimum tank size for domestic use. However, The large capacities resulted from the need for storage in low
a tank size larger than 5 kL is considered large for an average house rainfall areas to meet the reliability desired by the user to meet a
as it occupies considerable space (1.5 m diameter) adversely high demand. However, the threshold was fixed at 20 kL. Regres-
impacting aesthetics. However, it was appropriate to consider tank sion relationships were developed between p1 and p2 using expo-
sizes up to 20 kL as rainwater demand in commercial buildings nential, linear and power equations.
could be similar to domestic use, up scaled to fit the appropriate
number of people.
Rainwater tanks larger than 20 kL were obtained under follow- Development of a generalized curve to obtain the optimum
ing conditions: tank size

 Small roof areas (100 –150 m2) in low rainfall areas (necessi- To obtain a single dimensionless curve for selecting the opti-
tating carryover storage). mum rainwater tank size the data points from all the stations were
 High demand (laundry and above) in low rainfall areas combined into a single set. Data from 16 randomly selected
(necessitating carryover storage). stations were used to develop the combined curve. The other four
186 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188

Kinglake

Arthurs Creek

2.5
2.4
2.3
Rockbank 2.2
2.1
2
St. Albans 1.9
Doncaster east
1.8
Northing

Sunshine Easten Golf Club 1.7

Kew 1.65
Mitcham 1.6
Surrey Hills 1.5
Altona 1.4
Mountview 1.3
Caulfield
1.2
Caulfield North Notting Hill 1.1
Werribee
Hampton 1
0.9
0.8
Sandringham Berwick 0.7
0.6

Cranbourne

Easting

Fig. 5. Variation of optimum tank sizes for a water supply reliability of 90% from a roof size of 200 m2 with a water demand for toilet flushing and garden watering.

stations were selected as independent stations to verify the accu- Fig. 6 represents the best fit exponential regression curves in log
racy of the developed equations. The coefficient of determination scale. The advantage of deriving a straight line instead of a curve is
(R2) values of exponential regression equations Eqs. (10)–(12) were that the values could be accurately extrapolated if necessary. By
selected as the best form of the dimensionless curve. Fig. 8 depicts using the above exponential equations for different reliabilities,
the relationship between the tank size and the dimensionless num- tank sizes were calculated for each dependent and independent
bers p1 and p2 for three different reliabilities (95%, 90% and 85%). station with different roof sizes and demand types and checked
Exponential equation: for the model’s predictive ability by comparing the results with
the single site simulation tank capacity derived. A potential tank
95% reliability : Y ¼ 0:29e4:85X R2 ¼ 0:77 ð10Þ user will be able to select the appropriate tank size by using this
curve if the value of the roof area, consumptive demand and the
90% reliability : Y ¼ 0:15e5:16X R2 ¼ 0:84 ð11Þ mean annual rainfall of that particular area is known. The reliabil-
ity required could be pre-defined by the user. As an example, as-
85% reliability : Y ¼ 0:08e5:58X R2 ¼ 0:89 ð12Þ sume a potential customer for a rainwater tank living in the
southern side of the Greater Melbourne area where the mean an-
The standard form of the dimensionless curve is given by:
nual rainfall is around 700 mm. The user fixes the desired reliabil-
Y ¼ aebX ð13Þ ity to be at 95%. For the above customer, based on Fig. 6, for a roof
area of 200 m2 and a household of three with a designed demand
where a and b are empirical constants, Y is the dimensionless
to meet toilet flushing, garden watering and laundry use, the opti-
number p1 ¼ C3 , X, the dimensionless number p2 ¼ AR
D
, C, D , A and
A2 mum tank size is 9 kL (with 95% supply reliability). This may be too
R are as defined earlier.

100.00

5.16x
[Tank Capacity/(Roof Area ^1.5)]*10^(3)

4.85x
y = 0.29e y = 0.15e
2 2
R = 0.77 R = 0.842
10.00
95%
5.58x
y = 0.08e
2
R = 0.89
1.00

85%
90%
0.10

0.01
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90
Demand/(Roof Area*Rainfall)

Fig. 6. Exponential regression relationships between dimensionless numbers for 95%, 90% and 85% supply reliabilities.
 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188 187

large from an aesthetic point of view. However, for the same

Coeffecient of Determination (R2)

0.95
household, if the desired demand is limited only for toilet flushing
and garden use, the tank size computed will only be 2 kL. For the
same household and the desired demand (toilet + garden), the tank
0.90
size would be only 1.1 kL if the user drops the desired reliability to
90%. From the above information, it can be clearly seen that from
the single generalized curve, the customer has the opportunity 0.85
and flexibility to fix the optimum tank size after considering the
reliability and demand and matching the resulting tank sizes with
the investment required and the acceptable aesthetic impact. 0.80

Surrey Hills
Mountview
Mitcham
Kew
Verifications of the developed curve

The exponential regression equations developed for 95%, 90%

Rainfall Stations (Independent)
and 85% supply reliabilities Eqs. (10)–(12) were used to calculate
tank sizes at each location for different roof sizes and demand Fig. 9. Range in coefficient of determination (R2) obtained for tank sizes calculated
types. These values were compared with the tank sizes calculated from the water balance equation and generalized curve for different reliabilities,
from the water balance equation for that particular location with roof sizes and demand types for independent stations.

the same roof size and demand as used with the generalized curve.
Figs. 7–10 depict the range of coefficients of determination and
0.95
efficiency (R2 and E) between the tank sizes calculated from the
water balance equation and the generalized curve for 95%, 90%

Coeffecient of Efficiency (E)

and 85% reliability for different roof sizes and demand types for 0.90

each station used to develop the curve, both for dependent stations
as well as independent stations. 0.85

0.80
Coeffeicnt of Determination (R2)

0.95
0.75
0.90
0.70
0.85

Surrey Hills
Kew

Mitcham

Mountview
0.80

0.75
Rainfall stations (Independent)

0.70
Fig. 10. Range in coefficient of efficiency (E) obtained for tank sizes calculated from
the water balance equation and generalized curve for different reliabilities, roof
Rockbank
Eastern golf Club

Kinglake
Doncaster

Sunshine
Notting Hill
Cranbourne

Sandringham
Hampton
Caulfield

Caulfield north

St. Albans
Berwick
Arthur Creek
Altona

Werribee

sizes and demand types for independent stations.

From the above graphs it can be observed that the R2 and E val-
Rainfall Stations (Dependent) ues for both the dependent and independent stations are more
than 60% (except E for Werribee) when compared to the tank
Fig. 7. Range in coefficients of determination (R2) obtained for tank sizes calculated capacities calculated with the developed equation with the values
from the water balance equation and generalized curve for different reliabilities,
from the water balance study. When the generalized curve equa-
roof sizes and demand types for dependent stations.
tion was applied to data from Werribee, the E values with all three
reliabilities were low. This is mainly due to the large tank sizes ob-
tained from the water balance model in the very low rainfall
Coeffecient of efficiency (E)

1.0
Werribee area (mean annual rainfall = 450 mm). As a result, when
0.9 the generalized curve was used to calculate the tank sizes from the
whole data set and compared with the tank sizes from the water
0.8
balance equation, the R2 and E values are lower than from individ-
0.7 ual stations. This is expected as there will be some loss of accuracy
when using the generalized fitted curve instead of the individual
0.6
site data.
0.5

Conclusions
Eastern golf Club

Rockbank
Caulfield north

Sandringham
Cranbourne

Doncaster
Arthur Creek

Notting Hill
Kinglake

Sunshine
St. Albans
Hampton
Caulfield
Berwick

Werribee
Altona

There is a considerable variation in rainwater tank size required

to meet a similar demand in different parts of metropolitan Mel-
bourne due to spatial variability of rainfall across the whole region.
Rainfall stations (Dependent) To the consumer, the variation in tank size translates to different
Fig. 8. Range in coefficient of efficiency (E) obtained for tank sizes calculated from
amount of capital investment required to purchase the tank. As a
the water balance equation and generalized curve for different reliabilities, roof result, the variation in rainfall around metropolitan Melbourne
sizes and demand types for dependent stations. has confirmed that customised sizing of the rainwater tanks is
188 A. Khastagir, N. Jayasuriya / Journal of Hydrology 381 (2010) 181–188

required to optimize the return on investment. A set of generalized References

reliability centred curves for selecting the optimum tank size
across the Greater Melbourne Region has been developed. Bucking- ABS, 2001. Environmental Issues, People’s Views and Practices. Australian Bureau of
Statistics. <http://www.abs.gov.au> (last accessed 10.12.07).
ham’s p theorem (dimensionless analysis) was used to reduce the Coombes, P., 2002. Rainwater Tanks Revisited: New Opportunities for Urban Water
number of independent variables used in the study to develop the Cycle Management. PhD thesis, University of Newcastle, NSW, Australia.
curves. <http://www.eng.newcastle.edu.au/~cegak/Coombes/> (last accessed 10.07.07).
Crowe, C., Elger, D., Roberson, J., 2007. Engineering Fluid Mechanics, eighth ed.. pp.
The inputs to tank sizing are the demand to be serviced by the 286–295.
tank, the area of the roof draining to the tank and the mean annual Department of Environmental and Aboriginal affairs Government of South
rainfall of that particular area. The potential tank user has to pre- Australia, 1999. Rainwater Tanks: Their Selection and Maintenance.
Department of Sustainability and Environment, 2006. Sustainable Water Strategy
determine the planned demand on the tank: garden (external) or
Central Region, Action to 2055, pp. 38–40.
internal use (toilets only, laundry only or the combination of the Department of Sustainability and Environment, 2007. Water Smart Rebate Scheme.
above) and the desired level of reliability prior to selecting the <http://www.dse.vic.gov.au/DSE/wcmn202.nsf/LinkView> (last accessed
appropriate size by using the derived curves. Whilst the first two 20.09.07).
Gato, S., Jayasuriuya, N., Roberts, P., Hadgraft, R., 2004. Understanding residential
inputs are provided by the consumer planning to install the tank, water use. In: Proceedings of the Enviro 04 Conference, Sydney, 2004 (CD).
the rainfall precipitating in areas around Melbourne can be ob- Jenkins, D., Pearson, F., 1978. Feasibility of Rainwater Collection Systems in
tained by information published by the Bureau of Meteorology California. Water Resources Centre, University of California. Contribution No.
173.
(www.bom.gov.au) on the web. The developed curves have pro- Khastagir, A., Jayasuriya, N., 2007. Parameters influencing the selection of an
vided the user with an opportunity to either vary the demand or optimal rainwater tank size: a case study for Melbourne. In: Proceedings of the
the reliability to determine the optimal rainwater tank size whilst Rain Water and Urban Design Conference 2007, Sydney, August 21–23 (in CD).
Lancaster, B., 2006. Rainwater Harvesting for Dry Lands, vol. 1, Appendix 3. <http://
considering other issues impacting the decision such as aesthetic www.harvestingrainwater.com/wp-content/uploads/Appendix3Calculations.
acceptability and the maximum amount of money the user is will- pdf> (last accessed 12.09.07).
ing to invest. The capacity of the consumer to use the curves iter- Roberts, P., 2004. Yarra Valley Water, 2003, Appliance Stock and Usage Patterns
Survey. November 2004.
atively to arrive at the desired outcome satisfying all the needs is a Urban Rainwater Systems, 2007. Securing Our Water Supply <http://www.
significant advancement to the current tank size selecting method- urbanrainwater.com.au/vic> (last accessed 10.04.07).
ology. Future research should consider at least 20 years of rainfall Water Sensitive Urban Design, 2005. Water Sensitive Urban Design Guidelines.
<http://www.melbourne.vic.gov.au/rsrc/PDFs/Water/WSUDGuidelines.rtf>.
data along with 20 rainfall stations with significant spatial distri-
Yaziz, M., Gunting, H., Sapari, N., Ghazali, W., 1989. Variations in rainwater quality
bution of mean annual rainfall for selecting optimum rainwater from roof catchments. Water Resources 23 (6), 761–765.
tank size in other cities of the world.