Aw

LWCa

b

c

d

e

f

a

ARRAA

KGRW

1

tcHcbwmisn

h2

Sustainable Cities and Society 32 (2017) 67–77

Contents lists available at ScienceDirect

Sustainable Cities and Society

jou rna l h om epage: www.elsev ier .com/ locate /scs

GIS based spatial decision support system for analysing residentialater demand: A case study in Australia

asinidu Jayarathnaa, Darshana Rajapaksab, Shunsuke Managib,c,∗,asantha Athukoralad, Benno Torglerc, Maria A. Garcia-Valinase, Robert Giffordf,

levo Wilsonc

Land Use Policy Planning Department, 31, Prathiba Road, Narahenpita, Colombo 05, Sri LankaUrban Institute, Department of Urban and Environmental Engineering, Faculty of Engineering, Kyushu University, 744, Motooka Nishi-ku, Fukuoka, JapanQUT Business School, Queensland University of Technology, Level 8, Z Block, Gardens Point, 2 George St, Brisbane 4000, QLD, AustraliaDepartment of Economics, Faculty of Arts, University of Peradeniya, Peradeniya, Sri LankaSchool of Economics and Business, University of Oviedo, Oviedo, SpainSchool of Psychology, University of Victoria, Victoria, BC, Canada

r t i c l e i n f o

rticle history:eceived 6 September 2016eceived in revised form 21 February 2017ccepted 16 March 2017vailable online 24 March 2017

eywords:ISesidential water demand

a b s t r a c t

Managing water resources and the need to adapt both supply and demand side policies to a changingenvironment has become a priority in both developed and developing countries. This research demon-strates the application of the geographic information system (GIS) in modelling residential water demandin order to develop a spatial decision support system (SDSS). Household level survey data covering 90suburbs within the Brisbane City Council (BCC), Queensland, Australia, are used for the analysis. First,residential water demand was estimated and the most significant variables found to predict high wateruse at the suburban level. These variables included household size, presence of a swimming pool, incomeand people over 65 years of age. By integrating this model with an SDSS, a spatial decision support system

ater demand index for residential water demand (SDSS-RWD) is developed. By producing maps which clearly display thedifferent factors affecting residential water demand, the benefit of the SDSS-RWD is found in its use as apolicy making tool for manipulating and evaluating effective water management strategies. In particular,the flexibility of the SDSS-RWD offers in evaluating changing determinants of residential water demandcreates the capacity for local government bodies to analyse a range of alternative policies.

© 2017 Elsevier Ltd. All rights reserved.

. Introduction

Water has become an increasingly scarce resource due to fac-ors such as population and economic growth, urbanisation andhanging climatic patterns (Haque, Rahman, Hagare, & Kibria, 2014;ouse-Peters & Chang, 2011). Moreover, conflicts over water havereated a growing competition among alternative users. This hasecome a source of growing controversy, as increasingly scarceater supplies fail to meet demand in many countries. Importantly,anaging water demand is one of the main factors to be considered

n sustainable city development planning. Careful analysis of deci-

ions pertaining to the allocation of water resources are, therefore,eeded (Arbués, Garcıa-Valinas, & Martınez-Espineira, 2003).

∗ Corresponding author.E-mail address: managi.s@gmail.com (S. Managi).

ttp://dx.doi.org/10.1016/j.scs.2017.03.012210-6707/© 2017 Elsevier Ltd. All rights reserved.

Generally, there are a range of options for handling the problemof water scarcity such as supply side and demand side managementpolicies. For instance, last century most water management issuesin western countries centred on the pursuit of federally funded (andconstructed) projects serving agricultural water demands throughincreased storage and conveyance facilities (Kenney, Goemans,Klein, Lowrey, & Reidy, 2008). More recently, attention has turnedto water demand management. Demand-side policies generallyhave had lower environmental and social impacts and lower coststhan supply-side policies. Studies have shown that if well plannedand implemented as part of a targeted suite of measures, they canbe highly effective in reducing household water use (Kolokytha,Mylopoulos, & Mentes, 2002; Turner, White, Beatty, & Gregory,2005). Consequently, one of the most important management

needs is to better understand and predict how household demandis likely to respond to both management interventions (such asprice increases and outdoor water use restrictions) and exogenousfactors (such as climate and socio-economic factors).

6 e Citie

oGeMmSaiesitniraisSusatsBditeeiiof

rtd

2

wfmu2dTam

s(2oiabaiJm(

8 L. Jayarathna et al. / Sustainabl

There has been a number of studies estimating the influencef factors affecting residential water demand management (Beal,urung, & Stewart, 2016; Eslamian, Li, & Haghighat, 2016; Haquet al., 2014; Jorgensen, Graymore, & O’Toole, 2009; Shearer, 2011).ost have modelled the impact of socio-economic and environ-ental factors on residential water demand (Chang, Parandvash, &

handas, 2010; Panagopoulos, 2014; Shearer, 2011). However, theyre not analytically comprehensive given they do not quantify thempacts on water demand of the interactions between the influ-ncing factors. Most of these models are limited to discussing thetudy area specifics and they can rarely be used for decision makingn other locations. Thus, they are generally not adaptable nor arehey an effective decision making tool for investigating policy alter-atives at a regional level. Given these restrictions, the use of GIS

n creating a SDSS, becomes a particularly useful tool for managingesidential water demand. Importantly, SDSS allows us to considerll household, socio-economic and environmental characteristicsn which the interaction effects can be observed. Therefore, thistudy is carried out with the objective of developing a GIS-basedDSS-RWD the robustness and sensitivity of which make it partic-larly appropriate for analysing residential water demand withinelected suburbs of the BCC. In particular the SDSS-RWD provides

tool for investigating property related and socio-economic fac-ors that determined residential water demand within the BCC andimilar urban regions. While the water supplying authorities in theCC focus on pricing strategies, many other determinants of wateremand receive less attention in management policies. Our study

s different from existing studies because it provides a frameworko incorporate all water demand determinants; household, socio-conomic and environmental characteristics where the interactiveffects can be observed. The GIS-based SDSS is particularly appeal-ng given it can include all determinants in policy simulation. Fornstance, the SDSS-RWD enables changes to water policies basedn the consideration of property characteristics or demographicactors of the city.

The remainder of this paper is as follows; the next sectioneviews the relevant literature and is followed by a description ofhe methodological in Section 3. Section 4 presents the results andiscussion while conclusions are set out in Section 5.

. Literature review

There has been a growing global interest in the sustainability ofater resources in major metropolitan areas. This interest stems

rom ongoing population growth and the growing influence of cli-ate change—both of which are posing multiple challenges for

rban water resource managers (Morehouse, Carter, & Tschakert,002). Until recently there has been a tendency to apply wateremand modelling to seek solutions to water scarcity problems.here are, however, several researchers who have used newpproaches to data management and analysis of water demandodelling.Researchers have investigated different pricing mechanism

uch as increasing block price, marginal price and average priceHoffmann, Worthington, & Higgs, 2006; Nataraj & Hanemann,011; Olmstead, Hanemann, & Stavins, 2007). Analysing North Car-lina residential water demand, Wichman (2014) studies the wayn which households responded to a water supply system based onverage pricing. Olmstead et al. (2007) showed the way in whichlock pricing affects demand elasticity. However policy managersre constrained in developing policy solutions based only on pric-

ng policies, given they have only a short term impact (Campbell,ohnson, & Larson, 2004). Consequently, the socio-economic deter-

inants of water use behaviour are now receiving more attentionPanagopoulos, 2014; Chang et al., 2010; Shearer, 2011). In a recent

s and Society 32 (2017) 67–77

paper, Wichman, Taylor, and von Haefen (2016) show the policyresponsive heterogeneity with respect to the socio-economic fac-tors. The research found low-income groups are more sensitive topricing policies than high-income groups. Also revealed is that dif-ferences in city water use can be sourced to socio-economic factorsas well as climate factors and time dynamics (Eslamian et al., 2016).

Chang et al. (2010) present such a differentiated methodologi-cal approach in assessing the role of urban development patternson water demand in Portland, USA. Using GIS and statisticaltechniques, the objective was to explain the variation in water con-sumption by block groups of single-family residential householdsin terms of density and physical and socio-economic characteristicsthat differ across the census blocks. Their findings show that single-family residential water consumption is mostly explained by keybuilding structural variables, namely building size, building densityand building age. Also, revealed was a correlation between buildingsize and socio-economic variables (i.e. income and education). Theresults allowed the development of a water demand frameworkthat incorporated existing factors into urban development policiesto effectively manage limited water and land resources.

Responding to the Queensland Government demand-side pol-icy measures imposed during the drought of 2006–2008, Shearer(2011) explored the determinants of household water consump-tion. The study area included the local government areas of Brisbaneand the Sunshine Coast and used GIS, Principal ComponentsAnalysis and other statistical methods to explore the spatial, socio-demographic and structural determinants of household water usein the 2006–2008 period. The first phase of the study used GIS forspatial analysis and mapping providing a broad indication of wateruse at a variety of spatial scales. It also showed important insightsinto the relationship between various datasets such as aggregatesocio-demographic census data and structural variables. However,the paper’s findings are not sufficiently integrated to support thecreation of alternative policies and hence there are no tools createdwhich can be used to help assess water demand at a regional level.Moreover, the aggregation methods used are generally subjectivein the way they combine indicators, and tend to look at the indi-cators’ impact on water demand in isolation from other indicatorsthus ignoring any interactions. Accordingly, decision makers wouldbenefit from the adoption of a decision support framework which iscapable of developing an objective aggregation method that moreaccurately assesses water demand (Graymore, Wallis, & Richards,2009).

Basically, utility managers use monetary and non-monetarymechanisms to manage residential water demand. However, theanalysis of such policies does not reveal conclusively a supe-rior policy package. For instance, Olmstead and Stavins (2009)do not observe significant differences between monetary andnon-monetary strategies in managing residential water demand.Research implies that many factors influencing residential waterdemand management may differ spatially as well as temporarily,resulting in different impacts on pricing and non-pricing policies.Importantly, the GIS which has been developed as a means fororganising and analysing spatial data, can be used to investigate theeffects on decision-maker performance through SDSS (see, Lieske,2015). However, what decision makers most need is an SDSS whichis in the form of “canned” software that is intuitively obvious touse, can solve specific problems efficiently, and deliver immediateresults (Crossland, Wynne, & Perkins, 1995). In spite of the requiredadvances in know-how in developing SDSS, its main advantagelies in its ability to adjust policy according to spatial and tempo-ral differences. Satti and Jacobs (2004) develop a GIS based model

to analyse agricultural water demand in Florida and demonstrateits ability to capture the heterogeneous nature of soil and climaticfactors.

e Citie

ptooaTbtestrauo

tptaaipedhefi

3

mtaBQwgsahaatsac

QtwGpc

cbttrc

L. Jayarathna et al. / Sustainabl

Graymore et al. (2009) have applied the GIS based decision sup-ort tool as a means for enhancing the regional sustainability ofhe Glenelg Hopkins Catchment region in Victoria, Australia. Thebjective of this study was to integrate a multiple criteria analysisf locally selected sustainability indicators into the GIS to produce

spatial decision support system for sustainability assessment.he robustness and sensitivity of the developed tool was testedy applying it to the sub-catchments of south west Victoria and inhis way its usefulness for regional sustainability monitoring andvaluation. In their study, the multiple criteria analysis based onocio-economic and environmental indicators have been used ashe basis to build a model in ArcGIS. This GIS-based multiple crite-ia analysis produced maps showing sub-catchment sustainability,s well as environmental and socio-economic conditions. The eval-ation of the tool proved that it was a robust and sensitive methodf sustainability assessment.

Given the importance of managing residential water consump-ion, the BCC has introduced several policy options in the recentast. Until the third quarter of 2006, there was a flat pricing struc-ure within the BCC. In the third quarter of 2006 the BCC introduced

three-tiered pricing structure for water users (Shearer, 2011). Inddition, the Council has imposed outdoor water-use restrictionsn the form of alternate fixed sprinkling days (which has been inlace for more than 20 years), as well as running a ‘Water Wise’ducation campaign (Hoffmann et al., 2006). However, as a rapidlyeveloping city, water demand has been increasing in Brisbane at aigher rate compared to other Australian state Capitals (Hoffmannt al., 2006). Therefore, demand side management (DSM) policiesor managing residential water demand in the BCC have becomemportant in generating a sustainable city plan.

. Method

Our study aims to develop an SDSS for residential water demandanagement in BCC using both primary and secondary informa-

ion. This study is based on cross sectional data collected from large mail based survey of householders conducted within theCC covering 90 suburbs1 (see, Fig. 1). Brisbane, the capital city ofueensland, is a moderately sized city covering an area of 1367 km2

ith approximately 950,000 residents. In line with high populationrowth in the rest of south-eastern Queensland, the city has grownteadily since the mid-1990s, with population increasing by 9.4%nd residential dwellings by 12.3% annually. As a result, the averageousehold size, currently 2.57 persons, has fallen by 2.6%. The aver-ge household size of the selected samples is 2.59 (minimum 1.9nd maximum 3.3) and weekly household income varies from $727o $2774 which captures the heterogeneity of the households. Withtrong economic growth, Brisbane city has an $85 billion economy,lmost half of the state economy. As for many other fast growingities in the world water demand is an important issue for Brisbane.

Brisbane generally has a lower rainfall than other parts ofueensland averaging around 1200 mm a year in comparison with

he Sunshine Coast’s average of 1650 mm (BOM, 2007). High quality

ater services are provided by the South East Queensland Waterrid (SEQld) to all suburbs. The Brisbane River is the water sup-ly source, being collected from higher altitudes and distributed atonstant pressure to suburban areas.

1 We employ a systematic sampling procedure in order to select a random sampleovering residents in the BCC. In the first stage, we ranked the suburbs within the BCCy taking the percentage of households belonging to the highest income group andhen each other suburb was selected for the study. We select every 2nd suburb fromhe list making a total of 90 suburbs. Within each of these suburbs households wereandomly selected for the survey. Hence, our sample contains a well-constructedross section of households within the BCC.

s and Society 32 (2017) 67–77 69

The survey questionnaire mailed to householders includedquantitative and qualitative questions on household water useand behaviour. In the first stage we received householders’ con-sent for the participation of the survey (9% responded). Of 2142selected households, 1214 were included in this analysis based onthe responses and after data cleaning2. Additional socio-economicdata was obtained from the BCC community profile which is basedon the Australian Bureau of Statistics (ABS) household survey data(BCC, 2012). The billing records (quarterly from 2009 to 2011) ofsingle-family residential water consumption were used to deter-mine factors affecting the total consumption of all selected suburbsin the BCC. Individual households were geocoded in ArcGIS 10.1using addresses linked to survey data and data obtained from theBCC community profile. Aggregating the data into suburbs has someadvantages. First, the inclusion of several socio-economic variablesenables us to test the extent to which social and structural vari-ables contribute to water consumption. Also, from an analyticalperspective, aggregating households into suburbs provides a meansfor assessing the spatial variations of water consumption acrossthe whole BCC region (Shandas & Parandvash, 2010). All Brisbanesuburbs span around the lower catchment of Brisbane River. Somesuburbs are in flood-plain locations and in terms of amenities anddisamenities such suburbs are heterogeneous. However, all house-holds receive similar utilities. Furthermore, due to the relativelyhomogeneous topography, there is less than a 5 ◦C temperature dif-ference across the study area. As a result, water rates and weatherare likely to affect all households similarly. However, selected vari-ables, such as average number of households, vary across differentsuburbs (see, Appendix). The appendix figures display the distri-bution of descriptive variables which shows three categories; high,medium and low, based on each variable distribution.

3.1. Estimation of an index of residential water demand

The estimation procedure is summarised in Fig. 2. Structuraland socio-economic variables shown in Table 1 are used for theempirical analysis of residential water demand index. In the firststage a residential water demand index is developed using regres-sion analysis results of residential water consumption data and itsdeterminants as shown in Table 1. The following functional formwas specified to estimate the coefficients:

AWCi = ∝i +∑

ˇixi + εi

where AWCi is average water consumption, is a constant term,‘ˇ’ is the vector of coefficient that reflect the influence of each char-acteristics (e.g. structural, socio-economic and climatic factor) and‘ε’ is the error term.

An index of residential water demand describes residentialwater demand with respect to factors affecting household wateruse and behaviour. The index is developed considering weightedaverage of all factors affecting residential water demand. The majorfactors can be categorised into three groups such as structural,socio-economic and behavioural (psychological) factors. Structuralfactors (lot area, garden type, presence of swimming pool and otherwater features) are generally drivers of outdoor water use andlargely related to lot size (Askew & McGuirk, 2004). They are inter-related with each other and with socio-economic factors (i.e. higherincome households tend to have larger gardens with more waterfeatures). However, socio-economic drivers (age, gender, income,

education, household size) are mostly related to indoor water use.Of these, household size is often the major socio-economic factorthat influences total water consumption.

2 Total number of observations—4759.

70 L. Jayarathna et al. / Sustainable Cities and Society 32 (2017) 67–77

Fig. 1. Study

Table 1Selected variables.

Code Definition

AWC Quarterly water consumption (kL/quarter)Dq1 First quarterDq2 Second quarterDq3 Third quarterAgHH Average household sizeAge18 Average number of people below 18 years oldAge1864 Average number of people between 18 and 64Age65 Average number of people over 65Showers Average number of showers in the houseToilets Average number of toilets in the houseTap outdoors Average number of outdoor taps in the houseDish wash Average number of dish-washers in the houseD gard1 Garden type 1: Less than 50 m2

D gard2 Garden type 2: Between 51 and 250 m2: omitted group isgreater than 250 m2

D gardC A dummy variable: garden covered with grassPool Presence of swimming pool: 1 if yes, otherwise 0Rainwater Presence of rain water tank: 1 if yes, otherwise 0D edu1 Education: 1 if completed grade 10 to 12, otherwise 0D edu3 Education: 1 if completed a degree or postgraduate

qualification, otherwise 0Income Average weekly income in the suburb (AUS $)D born Born in Australia: 1 if yes, otherwise 0Avgprice Average price per water unit

Note: Definition of the variables used in the regression analysis are shown in thetable.Except for the income variable, data for other variables were obtained from thesurvey.Average suburb income was obtained from the BCC community profile which isbased on the ABS household survey data (BCC, 2012).

area.

It should be noted that per capita water use is inversely pro-portional to household size because smaller households can easilyreduce personal use (i.e. showering and flushing toilets) but notgeneral use (i.e. washing clothes and dishes). Income can be both anegative and positive influence: higher income households tend tohave larger houses and gardens, but have a better capacity to reducewater use (Birrell, Rapson, & Smith, 2005). Finally, psychologicalfactors can interact with structural and socio-economic factors toinfluence water use behaviour. Attitudes can influence intentionto conserve water, although not necessarily behaviour (Allon &Sofoulis, 2006).

After selecting the explanatory variables (see, Table 1), a suit-able statistical model is selected to explain the variations in waterconsumption at the suburb level. We used the natural logarithmof quarterly average water consumption (AWC) as the dependantvariable and regressed it with indoor factors (i.e. showers, toilets),outdoor factors (i.e. tap outdoor, pool), socio-economic factors (i.e.AgHH, Age18) and dummy variables (quarter). While the exist-ing literature on the estimation of water demand models showsnumerous econometric techniques, the OLS methods used in theanalysis is still the dominant methodology.

In this study, the relationships between the indictors and theirimpact on water demand of the selected suburbs in the BCC weredetermined by using regression coefficients (Table 3). All the indi-cators were analysed together to form a holistic systems analysis.After producing the relative priorities for each factor separately,

the total water demand index becomes a weighted sum of all thefactors and is called an index of residential water demand (AIRWD).

L. Jayarathna et al. / Sustainable Cities and Society 32 (2017) 67–77 71

ArcGIS 10 Data format tra nsfor m

Coor dinate transfor mati on Resample

Raster da ta

ArcGIS Model builder Weighted su mmati on

Regression analysis

Indicator weights determination

Residential water demand index

Raster cla ssi fication

Residential water demand index map

Spatial decision suppo rt system for resid ential water demand (SDSS-R WD)

Socio-economic factor s Water consumption

Hous eho ld s ize Age, education

Income

Structural Factors No of sh owers in the house No of toilets in the house

No of outdoor taps No of dishwasher s in the house

Type of gard ens, pres ence of pool Presence of ra inwat er tank

t of th

3

twtodatmasep2

cdddiaPBfi

average water price. Specifically, the results show household size,presence of pool, age over 653 and income have strong positive

Fig. 2. Flow char

.2. Integration of an index of residential water demand with GIS

Problem-oriented planning at the regional level requires a holis-ic approach in which the different spatial levels are combinedithin a framework. The modified methods have to be used at

he regional and local levels in order to visualise the consequencesf different planning measures. Therefore, AIRWD is required toirectly produce images showing the variation of water demandnd other factors. In this way, the addition of visual images makehe index tool a valuable aid for regional managers to target imple-

entation of management actions in areas most in need. The GISpproaches are found to be a highly efficient means for creatinguch an index. That is, the GIS provides a consistent visualisationnvironment for displaying the input data and model results—aarticularly useful aid in a decision-making process (Xu & Coors,012).

In this study, the ArcGIS Model Builder and ArcMap data pro-essing tools were used to produce an effective and user-friendlyecision support tool – an SDSS-RWD – for assessment of resi-ential water demand in the BCC. In Model Builder, the user canesign and use the GIS based model to automate data process-

ng by interactively dragging process tools and data objects into

visual diagram of the model (Graymore et al., 2009; Jiménez-erálvarez, Irigaray, El Hamdouni, & Chacón, 2009). The Modeluilder is a programming tool developed by ESRI. First, the shapeles were converted in to grid format, followed by several spa-

Policy implementation

e methodology.

tial processes—buffer, classification, reclassification and overlay.Finally, assigning weights influence to each factor, the weightingsummation was used to combine the factors in the form of a weight-ing overlay process.

4. Results and discussion

Summary statistics of selected variables are presented in Table 2.The average quarterly water consumption is considered as thedependant variable which is 44 kL per quarter (with minimum0.33 and maximum 1018). The average number of household is2.59 which is comparable with BCC statistics. More than 50% ofhouseholds have a small garden—less than 50 m2. Nearly 27% ofproperties owned a swimming pool and 66% a rainwater harvestingtank.

Table 3 provides the estimated coefficient, standard errors andp-values. The estimated coefficients on many of the variables aresignificant, including household size, age, presence of rainwatertank, number of toilets, secondary education, presence of pool,presence of dishwashers, household income, born in Australia and

relationships with water demand whereas, age between 18 and

3 Household composition: please see Table 1 for detail.

72 L. Jayarathna et al. / Sustainable Citie

Table 2Descriptive statistics.

Variable Mean Std. dev. Min Max

AWC 43.960 74.375 0.333 1018AgHH 2.595 1.252 1 9Age18 0.475 0.884 0 5Age1864 1.297 1.270 0 6Age65 0.476 0.774 0 3Showers 1.852 0.748 1 5Toilets 1.998 0.761 1 5Tap outdoors 2.692 1.220 0 18Dishwash 0.684 0.465 0 1d gard1 0.505 0.500 0 1d gard2 0.361 0.480 0 1d grdc 0.208 0.406 0 1Pool 0.273 0.446 0 1Rainwater 0.680 0.466 0 1Age 58.39 13.76 22 95d edu2 0.198 0.399 0 1d edu3 0.662 0.473 0 1Income 1771 1101 140 9615d born 0.789 0.408 0 1Avgprice 2.781 7.890 0 522

Table 3Regression results.

Coef. Std. err.

Cons 2.450645*** (0.08543)AgHH 0.204039*** (0.01525)Age18 −0.0411** (0.017976)Age1864 0.032001** (0.012645)Age65 0.050073*** (0.01638)Showers 0.037248 (0.024392)Toilets 0.044002* (0.024486)Tap outdoors 0.005374 (0.008372)Dishwasher 0.200432*** (0.02365)d gard1 0.008216 (0.030301)d gard2 0.053213* (0.031855)d grdc −0.03423 (0.024564)Pool 0.102261*** (0.023712)Rainwater −0.07462*** (0.021773)Age −0.00097 (0.001003)d edu2 0.09125** (0.034501)d edu3 −0.01477 (0.030186)Income 6.04E − 05*** (1.03E − 05)d born 0.059211** (0.02419)Avgprice −0.02283*** (0.001244)N 4759Adj-R2 0.29

Ni

6sotsbib

4

rnsilta

ote: Standard errors are in the parenthesis, ***, ** and * denote variables are signif-cant at the 1%, 5% and 10% level, respectively.

4, number of toilets, secondary education and born in Australiahow a weak positive relationship. On the other hand, presencef a rainwater tank and average price has a strong negative rela-ionship with residential water demand. The only other variable tohow a (weak) negative relationship was age below 18. The distri-ution of determinants of residential water demand is illustrated

n the Appendix. All variables are shown in 3 categories with colourands as is for instance the spatial differentiation of households.

.1. Residential water demand index map

In using GIS, weighted summation was used to produce anesidential water demand index (RWDI) map which highlights sig-ificant variation in residential water demand across the selecteduburbs in the BCC (see Fig. 3). The average water consumption

s divided into three groups; comparatively high, medium andow consumption. Three categories are based on the all observa-ions we used for the analysis. Suburbs with high water demandre easily distinguishable from those with low water demand.

s and Society 32 (2017) 67–77

High demand suburbs are Karana Downs, Anstead, Bellbowrie,Pullenvale, Westlake, Brookfield, Graceville, Yeerongpilly, Bardon,Ashgrove, Enoggera, McDowall, Aspley, Geebung, Bracken Ridge,Deagon, Sandgate, Hamilton, Ascot, Hendra, Carindale, Gumdale,Manly West, and Rochedale.

On the other hand, Inala, Doolandella, Calamvale, SunnybankHills, Sunnybank, Macgregor, Up Mt Gravatt, Greenslopes, Wool-loongabba, Highgate Hill, East Brisbane, South Brisbane, City, NewFarm, Fortitude Valley, Taringa, Toowong, Morningside, Albion,Paddington, Red Hill, Kelvin Grove and Herston have low waterdemand. In particular, in those suburbs exhibiting high waterdemand, there are a greater preponderance of large household sizesand swimming pools. More generally, the RWDI of each suburbreflects the relationship of all structural and socio-demographicalfactors. Thus, the RWDI map is able to identify the suburbs whichneed immediate policy implementation.

4.2. Spatial decision support system for residential water demand

The SDSS-RWD provides information on both residential waterdemand and other characteristics across the study area. ModelBuilder, which is used to integrate AIRS with the GIS, has sev-eral advantages in the making of a useful spatial decision supportsystem. First, this tool takes less time to run the large numberof geo-processing tasks after the model is built. Second, it allowsthe inclusion of new factors affecting residential water demandby changing the weights of factors. Third, the system can directlyproduce both intermediate maps and output maps. Such interac-tive colour graphical displays of information provide greater utilitythan static, black-and white information. The efficiency of the SDSSmodel is therefore augmented through enhanced visualisation ofthe problem to be solved. The tool also provides transparency indecision making as it provides information about residential waterdemand and structural and socio-economic factors that are easilyaccessible to a variety of audiences including the general public(Graymore et al., 2009). Moreover, the GIS provides facilities forstorage of intermediate files and the many methods of analysis. Inthis way, it reduces transfer errors and the number of steps requiredto carry out an analysis (Pettit & Pullar, 1999).

Despite these advantages, the model has some limitations. Inparticular it requires knowledge of GISs to fully interpret theresults. Therefore, the water resource managers need to acquirebasic GIS knowledge to simulate effective policies on time and alsoconsidering location. Also, those using this model need to have anoverall knowledge of the system in cases where there are a limitednumber of indicators with which to analyse the results.

5. Conclusions

This study provides a new approach in developing an SDSS-RWD for the selected suburbs in the BCC. Overall, our findings onthe determinants of residential water demand show it is largely afunction of household size, age, presence of a rainwater tank, num-ber of toilets, secondary education, presence of a pool, presence ofdishwashers, income, whether born in Australia and average waterprice. The RWDI was developed using household level survey infor-mation. The household level data provides wider variations andaccurate estimates for weights in developing the index. Moreover,the RWDI map shows a significant variation across the study area.The use of SDSS allows for the consideration of all residential waterdemand determinants in the decision making process. In particu-

lar, in policy simulation, SDSS captures the spatial and temporalheterogeneity (Satti & Jacobs, 2004).

An RWDI was integrated with GIS to form an integrated GISbased decision support tool for residential water demand using

L. Jayarathna et al. / Sustainable Cities and Society 32 (2017) 67–77 73

ater d

Mtpaffts

aipmoemfI

opinions, findings, and conclusions expressed in this material are

Fig. 3. Residential w

odel Builder. It has several favourable advantages as an effec-ive decision support tool. It is particularly useful in evaluation andrioritisation of suburbs according to their water demand given itsbility to directly produce maps of residential water use and impactactors. In addition, other factors such as its adaptability and userriendly operational characteristics, illustrate the capacity of GISo become a highly effective means of creating a spatial decisionupport system.

As noted, given changing global and local environments, there is clear need for expanded water demand research which can betternform and empower water managers in the task of more efficientlyredicting and managing residential water demand. This studyakes a contribution to this need by demonstrating the benefits

f taking into account a wider range of factors such as socio-

conomic factors and temporal variations in water use decisionaking. As indicated by our model, a number of socio-economic

actors can be significant determinants of high or low water use.ndeed, socio-economic factors are shown to have contributed to

emand index map.

positive attitudes towards water conservation thereby supportingthe application of demand-side policies. From a modelling perspec-tive, its predictive accuracy will clearly be enhanced by inclusionof such indicators.

Acknowledgements

This work was supported by the Australian Research Council(ARC) Discovery Project DP0776795 entitled ‘Determining urbanwater conservation and management strategies: a novel approachusing field and survey data’ and a grant from the Ministry of Edu-cation, Culture, Sports, Science and Technology (MEXT) of Japanunder a specially promoted research project (ID 26000001). Any

those of the authors and do not necessarily reflect the views ofthe ARC and MEXT. Authors also acknowledge the editor and tworeviewers for their valuable comments.

7 e Cities and Society 32 (2017) 67–77

A

4 L. Jayarathna et al. / Sustainabl

ppendix A. Structural and Socio-economic indicators map

e Cities and Society 32 (2017) 67–77 75

L. Jayarathna et al. / Sustainabl

7 e Cities and Society 32 (2017) 67–77

R

A

A

A

B

B

B

B

C

6 L. Jayarathna et al. / Sustainabl

eferences

llon, F., & Sofoulis, Z. (2006). Everyday water: Cultures in transition. AustralianGeographer, 37(1), 45–55.

rbués, F., Garcıa-Valinas, M. Á., & Martınez-Espineira, R. (2003). Estimation ofresidential water demand: A state-of-the-art review. The Journal ofSocio-Economics, 32(1), 81–102.

skew, L. E., & McGuirk, P. M. (2004). Watering the suburbs: Distinction,conformity and the suburban garden. Australian Geographer, 35(1), 17–37.

risbane City Council (BCC). (2012). Brisbane community profile. http://profile.id.com.au/brisbane/population-types?DataType=en

eal, C. D., Gurung, T. R., & Stewart, R. A. (2016). Demand-side management forsupply-side efficiency: Modeling tailored strategies for reducing peakresidential water demand. Sustainable Production and Consumption, 6, 1–11.

irrell, B., Rapson, V., & Smith, T. F. (2005). Impact of demographic change and urbanconsolidation of domestic water use. Water Services Association of AustraliaMelbourne.

ureau of Meteorology (BOM), Climate change and variability, http://www.bom.gov.au/climate/change/#tabs=Tracker&tracker=timeseries.

ampbell, H. E., Johnson, R. M., & Larson, E. H. (2004). Prices, devices, people, orrules: The relative effectiveness of policy instruments in water conservation.Review of Policy Research, 21(5), 637–662.

Chang, H., Parandvash, G. H., & Shandas, V. (2010). Spatial variations ofsingle-family residential water consumption in Portland, Oregon. UrbanGeography, 31(7), 953–972.

Crossland, M. D., Wynne, B. E., & Perkins, W. C. (1995). Spatial decision supportsystems: An overview of technology and a test of efficacy. Decision SupportSystems, 14(3), 219–235.

Eslamian, S. A., Li, S. S., & Haghighat, F. (2016). A new multiple regression model forpredictions of urban water use. Sustainable Cities and Society, 27, 419–429.

Graymore, M. L., Wallis, A. M., & Richards, A. J. (2009). An index of regionalsustainability: A GIS-based multiple criteria analysis decision support systemfor progressing sustainability. Ecological Complexity, 6(4), 453–462.

Haque, M. M., Rahman, A., Hagare, D., & Kibria, G. (2014). Probabilistic waterdemand forecasting using projected climatic data for Blue Mountains WaterSupply System in Australia. Water Resources Management, 28(7), 1959–1971.

Hoffmann, M., Worthington, A., & Higgs, H. (2006). Urban water demand with fixedvolumetric charging in a large municipality: The case of Brisbane, Australia.Australian Journal of Agricultural and Resource Economics, 50(3), 347–359.

House-Peters, L. A., & Chang, H. (2011). Urban water demand modeling: Review ofconcepts, methods, and organizing principles. Water Resources Research, 47(5).W05401. http://dx.doi.org/10.1029/2010WR009624

e Citie

J

J

K

K

L

M

N

O

O

P

responds to price and who responds to prescription? Journal of Environmental

L. Jayarathna et al. / Sustainabl

iménez-Perálvarez, J. D., Irigaray, C., El Hamdouni, R., & Chacón, J. (2009). Buildingmodels for automatic landslide-susceptibility analysis, mapping and validationin ArcGIS. Natural Hazards, 50(3), 571–590.

orgensen, B., Graymore, M., & O’Toole, K. (2009). Household water use behavior:An integrated model. Journal of Environmental Management, 91(1), 227–236.

enney, D. S., Goemans, C., Klein, R., Lowrey, J., & Reidy, K. (2008). Residential waterdemand management: Lessons from aurora. Colorado: Wiley Online Library.

olokytha, E. G., Mylopoulos, Y. A., & Mentes, A. K. (2002). Evaluating demandmanagement aspects of urban water policy—A field survey in the city ofThessaloniki, Greece. Urban Water, 4(4), 391–400.

ieske, D. J. (2015). Coping with climate change: The role of spatial decisionsupport tools in facilitating community adaptation. Environmental Modelling &Software, 68, 98–109.

orehouse, B. J., Carter, R. H., & Tschakert, P. (2002). Sensitivity of urban waterresources in Phoenix, Tucson, and Sierra Vista, Arizona, to severe drought.Climate Research, 21(3), 283–297.

ataraj, S., & Hanemann, W. M. (2011). Does marginal price matter? A regressiondiscontinuity approach to estimating water demand. Journal of EnvironmentalEconomics and Management, 61(2), 198–212.

lmstead, S. M., Hanemann, W. M., & Stavins, R. N. (2007). Water demand underalternative price structures. Journal of Environmental Economics andManagement, 54(2), 181–198.

lmstead, S. M., & Stavins, R. N. (2009). Comparing price and nonprice approachesto urban water conservation. Water Resources Research, 45(4)

anagopoulos, G. P. (2014). Assessing the impacts of socio-economic andhydrological factors on urban water demand: A multivariate statisticalapproach. Journal of Hydrology, 518, 42–48.

s and Society 32 (2017) 67–77 77

Pettit, C., & Pullar, D. (1999). An integrated planning tool based upon multiplecriteria evaluation of spatial information. Computers, Environment and UrbanSystems, 23(5), 339–357.

Satti, S. R., & Jacobs, J. M. (2004). A GIS-based model to estimate the regionallydistributed drought water demand. Agricultural Water Management, 66(1),1–13.

Shandas, V., & Parandvash, G. H. (2010). Integrating urban form and demographicsin water-demand management: An empirical case study of Portland, Oregon.Environment and Planning B: Planning and Design, 37(1), 112–128.

Shearer, H. (2011). Using geographic information systems to explore thedeterminants of household water consumption and response to theQueensland Government demand-side policy measures imposed during thedrought of 2006–2008. In Paper presented at the State of Australian Citiesnational conference.

Turner, A., White, S., Beatty, K., & Gregory, A. (2005). Results of the largestresidential demand management program in Australia. Water Science andTechnology: Water Supply, 5(3–4), 249–256.

Wichman, C. J. (2014). Perceived price in residential water demand: Evidence froma natural experiment. Journal of Economic Behavior & Organization, 107,308–323.

Wichman, C. J., Taylor, L. O., & von Haefen, R. H. (2016). Conservation policies: Who

Economics and Management, 79, 114–134.Xu, Z., & Coors, V. (2012). Combining system dynamics model, GIS and 3D

visualization in sustainability assessment of urban residential development.Building and Environment, 47, 272–287.