This article appeared in a journal published by Elsevier. The attached copy is furnished to the ...

147 Pages·2013·11.11 MB·English

(Thisisasamplecoverimageforthisissue.Theactualcoverisnotyetavailableatthistime.)
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
JournalofEnvironmentalManagement99(2012)61e75
ContentslistsavailableatSciVerseScienceDirect
Journal of Environmental Management
journal homepage: www.elsevier.com/locate/jenvman
A new spatial multi-criteria decision support tool for site selection for
implementation of managed aquifer recharge
M. Azizur Rahmana,*, Bernd Rusteberga, R.C. Gogub,d, J.P. Lobo Ferreirac, Martin Sautera
aGeoscienceCenter,Georg-AugustUniversityofGöttingen,Goldschmidtstr.3,D-37077Göttingen,Germany
bTechnicalUniversityofCivilEngineering,B-dulLaculTei,Nr.124,Sector2,RO-020296Bucharest,Romania
cNationalLaboratoryforCivilEngineering(LNEC),AVDOBRASIL101,1700-066Lisbon,Portugal
dResearchInstituteforArtificialIntelligence,RomanianAcademy,Calea13Septembrie,13,RO-050711,Bucharest,Romania
a r t i c l e i n f o a b s t r a c t
Articlehistory: Thisstudyreportsthedevelopmentofanewspatialmulti-criteriadecisionanalysis(SMCDA)software
Received26October2010 toolforselectingsuitablesitesforManagedAquiferRecharge(MAR)systems.ThenewSMCDAsoftware
Receivedinrevisedform tool functions based on the combination of existing multi-criteria evaluation methods with modern
14October2011 decision analysis techniques. More specifically, non-compensatory screening, criteria standardization
Accepted5January2012
and weighting, and Analytical Hierarchy Process (AHP) have been combined with Weighted Linear
Availableonlinexxx
Combination (WLC) and Ordered Weighted Averaging (OWA). This SMCDAtool may be implemented
with a wide range of decision maker’s preferences. The tool’s user-friendly interface helps guide the
Keywords:
Artificialrecharge decision maker through the sequential steps for site selection, those steps namely being constraint
Siteselection mapping,criteriahierarchy,criteriastandardizationandweighting,andcriteriaoverlay.Thetooloffers
Spatialmulticriteriaanalysis somepredetermineddefaultcriteriaandstandardmethodstoincreasethetrade-offbetweenease-of-
Decisionsupport use and efficiency. Integrated into ArcGIS, the tool has the advantage of using GIS tools for spatial
Querença-Silvesaquifer analysis,andhereindatamaybeprocessedanddisplayed.Thetoolisnon-sitespecific,adaptive,and
comprehensive, and may be applied to any type of site-selection problem. For demonstrating the
robustness of the new tool, a case study was planned and executed at Algarve Region, Portugal. The
efficiencyoftheSMCDAtoolinthedecisionmakingprocessforselectingsuitablesitesforMARwasalso
demonstrated.Specificaspectsof thetoolsuchasbuilt-in defaultcriteria,explicit decisionsteps,and
flexibilityinchoosingdifferentoptionswerekeyfeatures,whichbenefitedthestudy.ThenewSMCDA
toolcanbeaugmentedbygroundwaterflowandtransportmodelingsoastoachieveamorecompre-
hensiveapproachtotheselectionprocessforthebestlocationsoftheMARinfiltrationbasins,aswellas
the locations of recovery wells and areas of groundwater protection. The new spatial multicriteria
analysistoolhasalreadybeenimplementedwithintheGISbasedGabardinedecisionsupportsystemas
aninnovativeMARplanningtool.
(cid:1)2012ElsevierLtd.Allrightsreserved.
1. Introduction groundwaterquality,storageofsurfacewaterinthesub-surface,
and as a barrier to salinity intrusion. Depending on the water
In the field of Water Resources Planning and Management, source, water quality, geology, surface conditions, soils, and
managed aquifer recharge (MAR) is becoming an important hydrogeology, a variety of methods have been developed to
solutionformitigatingwaterscarcityrelatedproblemsinaridand recharge groundwater (Bouwer, 2002). The spreading basin
semi-arid areas. MAR has been practiced throughout the world technique (infiltration) iswidelypracticed andisusefulinareas
for the recovery of groundwater levels, improvement of withhighlandavailability,highlypermeablesoil,andwherethe
hydrogeology allows for infiltration to an unconfined aquifer
(Ghayoumian et al., 2005). Other MAR techniques employing
* Correspondingauthor.Tel.:þ49551399397;fax:þ49551399379. injection wells require less area but a better quality of source
E-mail addresses: [email protected] water due tothe fact that the water is directly injected intothe
(M.A.Rahman),[email protected](B.Rusteberg),radu.gogu@
aquiferwithouttakingadvantageofnaturalattenuationprocesses
utcb.ro (R.C. Gogu), [email protected] (J.P. Lobo Ferreira), [email protected]
withinthevadosezone.Theinterdependencyofthewaterquality,
(M.Sauter).
0301-4797/$eseefrontmatter(cid:1)2012ElsevierLtd.Allrightsreserved.
doi:10.1016/j.jenvman.2012.01.003
Author's personal copy
62 M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75
MAR location, and technology makes project planning multifac- Richard,1994; Boroushaki and Malczewski, 2008). The combina-
etedandcomplex. tionofAHPwithWLCand/orOWAcanprovideamoreeffectiveand
Many factors need to be considered during the site selection robust MCDA tool for spatial decision problems. Boroushaki and
process for MAR projects. Complex regional characteristics, Malczewski(2008)implementedAHP-OWAoperatorsusingfuzzy
heterogeneities in surface and/or subsurface characteristics, and linguistic quantifiers in the GIS environment, which has been
variablegroundwaterqualitiesmakesiteselectionforMARdifficult proventobeeffective.
(Anbazhagan et al., 2005). Apart from these hydrogeological An intensive review of the respective literature has indicated
considerations,otherfactorssuchaspoliticalandsocialfactorsare thatthemodernandupdatedanalysistechniquesasGISandMCDA
important in the decision-making process. National and interna- havenotbeenwellinvestigatedandcompiledinthefieldofMAR
tional water policies, natural conservation regulations, environ- siteselection(seethefollowingsectionfordetails).Inthisrespect,
mental impact assessments, and socio-economic considerations astructured,non-sitespecificandflexibledecisionanalysistoolhas
makethesiteselectionprocedure complex.Complexity increases beendeveloped.Inthisstudy,amethodologyhasbeensettledupto
whenMARprojectmanagersarefromdifferentdisciplinaryback- support the identification of suitable sites bycombining modern
grounds; thismayoftenleadtodisagreementsconcerningwhich spatial multi-criteria analysis techniques with decision analysis
criteriatogivemoreweightinthedecision-makingprocess.These methods.Asconsequence,anewtoolhasbeendevelopedtooffer
conflictsalwaysneedtobedealtbeforetheMARprojectisimple- thefollowing:
mented. GIS and the traditional Decision Support Systems (DSS)
alone do not effectively facilitate the implementation of MAR (cid:2) AcomprehensiveframeworkconsistingofAHP,WLC,andOWA
projectparameters,whichareequallybasedoncomplexdecision analysistechniquesforspatialmulticriteriaanalysisforMAR
criteria and spatial information (Jun, 2000). GIS based analysis siteselection.
methodsarepoorindealingwithuncertainty,risks,andpotential (cid:2) Awiderangeofflexibilityandpreferencesforcriteriaselection,
conflicts;therefore,thereisalargepossibilityoflosingimportant standardization,andweighting.
information,whichinturnmayleadtoapoordecision(Baileyetal., (cid:2) Aninteractive userinterface,whichoffersthestandardtech-
2003).Multi-CriteriaDecisionAnalysis(MCDA)integratedintoGIS niquesandleadstheusersystematicallytocompletethesite
(SMCDA) provide adequate solution procedures to this problem selectionprocess.
becausetheanalysisofpotentialMARprojectsmaybedonemore
comprehensivelyandatalowercost.Variableprojectsites,risks, ThispaperincludesareviewonMARsiteselectiontechniques
MARtechniques,policies,andlimitsingeologicalaswellassocial, and may be read at any time as needed for reference purposes
environmental,andpoliticalrealmscaneasilybeconsideredbythe (section2).Section3isadescriptionofSMCDAforsitesuitability
SpatialMulti-CriteriaDecisionAnalysis(SMCDA)approach(Calijuri analysisandalsoinformationonhowAHP,suitabilitymapping,and
etal.,2004). weighting are involved in the analysis. A brief description on
MCDAishelpfulinidentifyingprioritiesforagivenMARproject possiblesensitivityanalysisisdonethepresentationofaGISBased
(Gomes and Lins, 2002). The integration of MCDA techniques SiteSuitabilityAnalysisToolinSection4.Thissectiontogetherwith
with GIS has considerably advanced the traditional map overlay Section 3 provides distinctive information to MAR site selection.
approaches for site suitability analysis (e.g. Malczewski, 1996; The two sections together are considered to embody the core
Eastman, 1997). MCDA procedures utilize geographical data, objective of this paper, which is to explain the development and
consider the user’s preferences, manipulate data, and set prefer- functionalityofanewSMCDAtoolforMARsiteselection.Section5
encesaccordingtospecifieddecisionrules(Malczewski,2004).The presentstheconceptsdescribedinSections3and4asappliedin
advantage of integrating GIS with MCDA has been elaborated by thefield.ThecasestudypresentedinSection5referstoaMARsite
manyauthors (e.g. Malczewski,1996; Jun, 2000; Gomes andLins, selection. It has been developed on the Querenca-Silves aquifer
2002;SharifiandRetsios,2004).AccordingtoMalczewski(2004), systemin Portugal.Section 6 provides a summaryof conclusions
thetwocriticalconsiderationsforSMCDAare:(i)theGIScapabilities and recommendations in the scope of future applications and
of data acquisition, storage, retrieval, manipulation, and analysis; furtherdevelopments.
and(ii)theMCDAcapabilitiesforcombiningthegeographicaldata
andthemanager’spreferenceintounidimensionalvaluesofalter- 2. Thestate-of-the-art:MARsiteselectiontechniques
native decisions. A number of methodologies have already been
developedforSMCDAindifferentfieldsofscienceandengineering Only very few studies exist which focus on site selection
toselectthebestalternativesfromasetofcompetingoptions(e.g. proceduresforManagedAquiferRecharge(MAR).Respectively,the
Sharifietal.,2006;Zuccaetal.,2007). following three sections differentiate data types (section 2.1),
TheoverlayMCDAplaysanimportantroleinmanyGISappli- presentdataprocessingviaGIS(section2.2),andgivereferenceto
cations. Boolean logic and Weighted Linear Combination (WLC) thestepsinvolvedinsitesuitabilityanalysismethods(section2.3)
are the most popular decision rules in GIS (e.g. Eastman, 1997; forexample,screeningofsites,criteriahierarchyandstandardiza-
Malczewski and Rinner, 2005) and both can be generalized tion,criteriaweighting,overlay,andsensitivityanalysis.Thethree
within the scope of Ordered Weighted Averaging (OWA) (e.g. parts of this chapter are intended to serve as a reference for the
Malczewski and Rinner, 2005; Malczewski, 2006). In OWA, basicmethodswhichhavebeenintegratedintotheSMCDAtoolfor
anumberofdecisionstrategymapscanbegeneratedbychanging sitesuitabilityanalysisofMAR.
theordered weights.Several OWA applicationshave been imple-
mentedalready(e.g.RinnerandMalczewski,2002;Calijurietal., 2.1. Datatypes
2004;Malczewskietal.,2003;Malczewski,2006).TheAnalytical
HierarchyProcess(AHP),proposedbySaaty(1980),isanotherwell- For MAR site selection, different types of data are required.
knownprocedure.Thisprocedureisimportantforspatialdecision Considerationsfordatatypeselectionderivefromdataavailability
problemswithalargenumberofcriteria(Eastmanetal.,1993).AHP andtheobjectiveoftheanalysisasdependentoneachdatatype.
can be used to combine the priorities for all levels of a “criteria Geologicalmaps,geomorphologicmaps,lineamentmaps(e.g.Saraf
tree,”includingthelevelrepresentingcriteria.Inthiscase,arela- andChoudhury,1998;Jothiprakashetal.,2003;ReddyandPratap,
tively small number of criteria can be evaluated (Jankowski and 2006),slope,infiltrationrate(e.g.Ghayoumianetal.,2005;Werz
Author's personal copy
M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75 63
etal.,2009),lineamentdensity,structure,fluvialanddenudational Brown et al., 2008). The Analytical Hierarchy Process (AHP),
geomorphology(e.g.Anbazhaganetal.,2005;ShankarandMohan, introducedbySaaty(1980),hasbeenrecentlyusedbyChowdhury
2005; Chowdhury et al., 2010,), soil texture (e.g. Kalantari et al., et al. (2010). The advantage of AHP in site selection or spatial
2010; Jothiprakash et al., 2003), and land use (e.g. Brown et al., multi-criteria analysis is well established (e.g. Jun, 2000; Sharifi
2008; Reddy and Pratap, 2006; Ghayoumian et al., 2007) have et al., 2006) and is therefore implemented in the new tool pre-
beenusedtoprovidedetailedquantificationofsurfacecharacter- sentedinthisreport(seesection3.3B).
istics.Infiltrationrate,transmissivity(e.g.Ghayoumianetal.,2005; The most important step for site selection is map overlay.
Brownetal.,2008),boreholerechargecapacity,boreholeabstrac- Conventionaloverlaymethods(e.g.WeightedLinearCombination
tioncapacity,andrechargeretentiontime(Andersonetal.,2005), (WLC)) have been practiced in most of the studies for MAR
havebeenusedtoquantifysubsurfacecharacteristics.Groundwater site selection (e.g. Saraf and Choudhury, 1998). Kallali et al.
qualitydataisusuallynotpaidanyattentionasbeingimportantfor (2007) used Boolean logic for combining maps. Normal succes-
site selection (Brown et al., 2008; Ghayoumian et al., 2007; sive intersection of the thematic maps has been used, too (e.g.
Chowdhuryet al., 2010), although it should be due tothe nature Jothiprakash et al., 2003). The Ordered Weighted Averaging
ofMARasbeingofteninvolvedwithwaterstorageandrecovery.In (OWA) method has not been implemented in any study. It is
addition to surface and sub-surface characteristics, Brown et al. importanttonotethatcombinationofAHP-WLCorAHP-OWAhas
(2008) considered ecological status, road density, power lines, yettobeimplementedandisthuspresentedforthefirsttimein
proximity to a water source, and groundwater pollution among this report (see Chapter 3). Ghayoumian et al. (2005) briefly
other parameters. Wood (1980) considered the possibility of mentionedtheintegrationofDSSandGISforMARsiteselection,
aquiferplugging.Legalaspectstogetherwithcost-benefitanalyses but no special considerations for DSS decision rules or techno-
shouldalsorankamongimportantconsiderations(O’Hare,1986).A logical descriptions were provided. A spatial knowledge-based
comprehensive combination of all of these input considerations, decision analysis system for pond site selection has been re-
however, is absent from the literature of MAR project site ported by Shrier et al. (2008). The authors coupled spreadsheet
characterization. softwarethatfacilitatesruleprocessing,withGIStodisplayspatial
data.
2.2. Dataprocessing Site suitability analysis for MAR is like most other site suit-
abilityanalyses,butthemethodswhicharebestappliedarebeing
The quantity of spatial data needed to collect, integrate and applied in a unique combination. The advantage of SMCDA has
analyze for MAR project site evaluation is very large and the not been properly and fully utilized in this field. This report
applicationoftraditionaldataprocessingmethodsforsiteselection presentsallthedatatypesanddataprocessingtechniquesneeded
can be very complex and tedious (Anbazhagan et al., 2005). In for MAR site assessment as well as the familiar yet unique
groundwater management studies, land use suitability mapping procedure to do so. Also, MAR is a developing field but also has
and other geographical research, GIS and remote sensing tech- alongandfragmented history.Thisreportisabreakthroughfor
nology have been used separately or in combination to process, the field of MAR in the respect that state-of-the art analysis
integrate,andanalyzespatialdata(e.g.Krishnamurthyetal.,1996). techniques are being compiled and applied in this field and
UseofGISandremotesensingisalsocommonlyusedforMARsite finally more work is being done to ensure the continued imple-
selection studies (e.g. Saraf and Choudhury, 1998; Brown et al., mentation of MAR with decreased uncertainty, as demonstrated
2008; Ghayoumian et al., 2007; Werz et al., 2009). Anderson in the entirety of this paper. This report presents for the first
et al. (2005) used the mathematical functions implemented time an interactive non-site specific decision tool for MAR site
inGIStogetherwithspatialanalysisoperationstocalculatereten- selection.
tion time and rechargecapacityof an aquiferoverawide spatial
distribution. 3. Thespatialmulti-criteriadecisionsupportmethodforsite
suitabilityanalysis
2.3. Sitesuitabilityanalysismethods
Theoverallmethodologyofthenewsiteselectiontoolisshown
Ingeneral,thesitesuitabilityanalysisfollowsthepath:screening inFig.1.Thisflowchartshowsthemaindecisionstepswhichare
offeasibleareas/classificationofthematiclayers/weightingof implemented for spatial analysis. In general, the entire process
thecriteria/overlaying. involves three main steps: (a) constraint mapping, (b) suitability
Onlyfewstudieshaveconcentratedonscreening-outtheareas mapping,and(c)sensitivityanalysis.Afterpreparingtheconstraint
whereMARisactuallynon-feasible(e.g.Ananeetal.,2008;Brown map, AHP is combined with WLC and OWA for the suitability
et al., 2008; Ghayoumian et al., 2007). Boolean logic is usually mapping,whichisbasedonstandardizedsubcriteria.Thefunction
usedtodemarcatefeasibleandnon-feasibleareas.Studiesmostly of AHP is threefold: (1) developing the hierarchy after the
concentrateonclassifyingmapsaccordingtorelativeimportance. selection of criteria (2) doing a pair-wise comparison to assess
Each thematic map is classified according to importance of criteria importance and (3) undergoing construction of the
therespectivelyrepresentedparameters.Linguisticclassifiers,such overallcompositeweight(globalweight).Afterward,WLCorOWA
as very good, good, suitable, etc. (e.g. Jothiprakash et al., 2003; operators are used for the final suitability map. These steps,
Ghayoumianetal.,2005)andvaluetypeclassifierssuchasclass1 following the MAR problem statement, are described in greater
toclass4areimplementedinthesestudies(e.g.Ghayoumianetal., depthbelow:
2007).Step-wisefunctionsareusedindifferentstudiesinorderto
standardize thematic maps for aggregation. Ghayoumian et al. 3.1. Problemstatement
(2007) uses membership functions for map standardization. No
linearorpiece-wise linearfunction isused inanystudy forMAR In water resources management, MAR has proven to be an
siteselection. effective response to water scarcity problems. MAR is helpful for
Weighting of each criterion is an important factor for spatial the recoveryof groundwater levels, the improvement of ground-
multi-criteria analysis. Direct weighting after consultation with water quality, and for storage of water and as a barrier against
experts has mostly been used (e.g. Saraf and Choudhury, 1998; salinity intrusion. The selection of suitable locations for MAR
Author's personal copy
64 M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75
Fig.2. Flowchartforconstraintmapping.
consistingofsub-criteriaofthegoal.Thetopofthehierarchyisthe
goal of the analysis/problem. The middle level contains more
specificcriteriawithregardstotheobjectiveandthebottomlevel
refers tothe most specific criteria. The sub-criteria in the lowest
level are related to the main criteria in the middle level, while
the top level relates to the “suitability map” (see Fig. 12). The
sub-criteria are represented by thematic maps or attributes. The
Fig.1. TheprocedureforMARsitesuitabilitymapping.
tool’suser-interface allowsthe user toconstructthe hierarchyor
“criteriatree.”
implementationbasedonpropertechnologiesisoneoftheprimary (C) Standardizationofsub-criteriamaps
requirements.
Eachsub-criterioninthecriteriatreeisrepresentedbyamapof
different types such as a classified map (e.g. land use) or avalue
3.2. Constraintmapping
map(e.g.slope,infiltration).Fordecisionanalysis,thevaluesand
classesofallthemapsshouldbeconvertedtoacommonscaleto
The main objective of constraint mapping is to screen out
reducethedimensionality.Suchconversioniscalledstandardiza-
alargenumberofalternativeswhichhavebeendeemedasbeing
tion(SharifiandRetsios,2004).Differentstandardizationmethods
non-feasible. This step helps the user to avoid conflicts in
maybeappliedtodifferentmaps.Thismodelofferslinear,piece-
decision-making. The sites which are of prime interest to other
wise linear, and step functions for standardization. The outcome
planningprojectsorwhicharesimplynotavailableorcompletely
ofthefunctionisalwaysavaluebetween0and1.Thefunctionis
non-feasible for MAR implementation are screened out in this
choseninsuchawaythatcellsinamapthatarehighlysuitablefor
step.Aconjunctivescreeningapproachwaschosenforconstraint
achievingthegoalobtainhighstandardizedvaluesandlesssuitable
mapping.Underconjunctivescreening,analternativeisaccepted
gridsobtainlowvalues.
if it meets specified thresholds for all evaluation criteria.
Fig. 2 shows the general procedure for constraint mapping. The
(D) Relativeweightsofcriteriaandsub-criteria
developed constraint map serves as a mask for suitability
mapping.
Thenextstepinthesiteselectionprocedureisassigningvalues
of importance for all criteria and sub-criteria, which is done by
3.3. Suitabilitymapping assigningaweighttoeachcriterion.Differentweightingmethods
areavailable.Pair-wisecomparisonanddirectweightingareused
(A) Choiceofcriteriaandsub-criteria here. The sub-criteria under each main criterion are compared
amongstthemselvesandaweightisassignedtoeachone.Themain
Inthisstep,allrelevantsurface,subsurface,andregionalchar- criteriaarealsoevaluatedinthisway.
acteristics are selected. Each characteristic is defined as a sub-
criterion. The sub-criteria are grouped under the main criteria. (E) Combinationofcriteriaandsub-criteriamaps
Thecombinedmaincriteriaarethe“suitabilitymap,”whichisthe
goaloftheSMCDA. Afterstandardizationandweighting,thenextstepistoobtain
theoverallsuitabilityindexofeachalternative.Theindexvalueis
(B) Hierarchyofcriteriaandsub-criteria giventothecellsofthemap.OverlaymethodsavailableareWLC
andOWAwithfuzzylinguisticquantifiers.WLCisthemostsimple
The role of AHP begins at this step. This step involves the and the most commonly used aggregation method in spatial
decomposition of the ultimate goal into a three-level hierarchy analysis(Eastmanetal.,1993).
Author's personal copy
M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75 65
3.4. Sensitivityanalysis
Asensitivityanalysismaybeundertakenbytheuserinorderto
study the robustness of the suitability map with respect to the
linguisticquantifier(a).ThenewSMCDAtoolalsopermitsassess-
mentofsitesuitabilityasrespectivetotheinfluenceoftheappli-
cationofdifferentweightingschemesandstandardization.Inthis
respect,sensitivityanalysesareusefulwhereuncertaintyexistsin
the construction of hierarchy and in the assignment of relative
importance(StoreandKangas,2001).
4. GISbasedsitesuitabilityanalysistool
4.1. Overallsystemframework
Thesitesuitabilityanalysistoolextensionistightly integrated
in the ArcMap environment. This instrument is developed as an
Fig.3. Thedecisionstrategyspaceshowingrelationbetweentrade-offandrisk,nis ArcMap extension, using ArcObjects and VB.Net. ArcObjects is
thenumberofcriteria(modifiedafterEastman,2000;Malczewski,2006).
a developer kit for ArcGIS based on Component Object Model
X (COM). This solution considerably extends the functionalities of
WLC; SðxiÞ ¼ wi$siðxiÞ (1) ArcMapbyimplementingtheMCDAwithintheGISenvironmentby
allowing the developer to combine the advantages given by the
wi¼normalised weight; Swi¼1; si (xi)¼standardized criteria userinterfacecontrolsavailableinthe.NetframeworkwiththeGIS
function/map.
functionality included with ArcGIS (ESRI). The advantages of
OWAisaclassofmulticriteriacombinationoperators,involving
customizedcomponentsbyusingaCOM-Compliantenvironment
twosetsofcriteriaweights,whichare“criteriaimportanceweight”
suchasVisualStudio2005are:(1)awiderrangeoffunctionalities
and“orderedweight”(Yager,1988).Theconceptoffuzzylinguistic
canbeintegratedintocustomization,(2)codesarenotaccessibleby
quantifiers,introducedbyZadeh(1983),allowstheconversionof
theuser,(3)allaspectsofArcGISapplicationcanbeusedfurther,
naturallanguagestatementsintopropermathematicalformulation
extended, and customized, (4) the customization can be easily
(Munda, 1995). In this study, the regular increasing monotone
supplied to the client machines (ESRI, 2004; Boroushaki and
quantifierclasswasconsidered.Giventhecriteriaweights,wi,the Malczewski,2008).Fig.4showstheoverallsystemdevelopment.
quantifier-guidedOWAcanbedefinedasfollows(Boroushakiand
Themodelhasbeenincorporatedintothetableofcontentsof
Malczewski,2008): ArcGIS as the “Site Selection” option. By activating the tab, the
!a !a! user can access the main steps of the site selection instruments:
OWAðiÞ ¼ Xn Xj uk (cid:3) Xj(cid:3)1 uk zij (2) “RCaonnksintrga.i”nFturMthaeprpoipntgi,o”ns“SreitleateSdutiotaebailcihtymMainapspteinpg(,”Figa.n5d)de“rSiivtee
j¼1 k¼1 k¼1
fromthisone.
zij¼weightedattributevalue;a¼parameterforlinguisticquanti- The supporting database structure is an ArcGIS Personal Geo-
fier;uk¼criteriaweight reordered according tozij; j¼numberof database(ESRI).Thegeodatabasecanstore,besidethegeographical
criteria. data, data behavior rules such as domains, relationship classes,
OWA allows for a high degree of input variability and for the and custom behavior. The geodatabase management module is
trade-off of importance among input variables (Fig. 3). When composed of two sections: (i) Data Input/Output and (ii) Spatial/
a¼0 (linguistic quantifier categorized as “at least one criterion Time dependent query and visualization. The geodatabase
satisfies”), the result yields no trade-off and full ORness; when management module focuses on designing the user screens, so
a¼N(linguisticquantifiercategorizedas“allcriteriasatisfy”),the thesematchthedifferentsectionsofthedatamodel.Thiscompo-
resultyieldsnotrade-offandfullANDness.Usingavaluebetween nentincludesthefollowingsubcomponents:
0 to N, yields a range of MCE operators in the decision strategy
space.Whena¼1(linguisticquantifieriscategorizedas“halfofthe (cid:2) Data access subcomponent, which contains functions for
criteriasatisfy”),theresultsyieldsthefulltrade-off(WLC)(Fig.3). databaseconnection,datareading,anddatabaseupdate.
ThedetaileddescriptionofAHPcombinationwithOWAisgiven (cid:2) Data model objects, which are used for storing the data in
byBoroushakiandMalczewski(2008). memory while the application is running. These data model
Fig.4. StructureofthesiteselectiontooldevelopedintheArcGISenvironment.
Author's personal copy
66 M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75
easily upgraded for further developments to an ArcSDE (ESRI)
geodatabase. The ArcSDE allows connecting ArcGIS and the Site
Suitability Analysis Tool interface to future database versions
developed using other Spatial Relational Database Management
System (RDBMS) software like Oracle, SQLServer, IBM-DB2, and
others.
4.2. Sitesuitabilitymapping
Thisfirststepoffersdefaultcriteriaforchoosingandselecting
the corresponding raster map to generate a constraint map. The
defaultconstraintcriteriahavebeenselectedafteraclosediscus-
sionwithin a consortium consistingof a number of international
experts from different organisations (e.g. LNEC e National Labo-
ratoryforCivilEngineering,Portugal;UniversityofLiege,Belgium;
EWRE e Environmental & Water Resources Engineering Limited,
Israel; Universityof Nottingham, UK; PHG e PalestineHydrology
Group, Palestine; GeohidroConsult, Romania; University of Goet-
tingen, Germany, etc.). Moreover, new constraint criteria may be
addedbytheuser(Fig.6).Bothvaluetypeandclasstypemapcan
behandledbythesystem.Theuserdefinesthethresholdvaluefor
valuetypecriteriaandtoeachclassoftheclasstypemap;theuser
mayassignazeroforanon-potentialareaoraoneforapotential
area.Thesystemthencreatesaconstraintmapofeachsub-criteria
separately. Afterwards, the maps may be overlain and one
constraintmapmaybepreparedwithBooleanlogic.Theconstraint
mapsareaddedtotheArcGISdocumentandcanbeusedforfurther
analysis.
Sitesuitabilitymappingstartswiththepreparationofahierar-
chicalstructure,whichisperformedbyselectingcriteriaandsub-
criteria for each level. The user selects the criteria from the
Fig.5. ExemplarytableofcontentsinArcGISDisplayforthesitesuitabilityanalysis, defaultlist.Thedefaultcriteriaareprepared,consideringallrele-
incorporatedtoGabardineDSS. vantcharacteristicsthatshouldbeincludedforthespatialanalysis.
Specialcarehasbeengiventoavoidanyduplicationofthecriteria/
sub-criteria.Newcriteriaorsub-criteriacanalsobeeasilyaddedvia
objectsabstractthefeatureclassesandthetablesinthegeo- theuser-interface.Theusercanvisualizethehierarchicalstructure
databaseandmimictherelationshipsbetweenthem. and edit for presentation and reporting purposes. The standardi-
(cid:2) Interface components, which include the user screens that zationprocessfollowsthebuildingofhierarchy.Theuserselectsthe
provide user access to the data stored in the data model criteria,theconstraintmap,thethresholdvalues,andthepreferred
objects.Theusercaninputdatathroughalistofstandarduser standardizationfunction.Forabettervisualization,theconverted
interfacecontrols,suchastextboxes,comboboxes,datagrids, functionisdrawngraphicallyintheinterface(Fig.7).Theoverlay
etc. commandofthecriteriatreeproceedstothestepofweightingand
overlay.Thesystemoffersthepair-wisecomparisonandthedirect
The personal geodatabase format was considered suitable for weightingmethods.Theweightsofeachcriterionineachlevelcan
the scale of the current application; however, the format can be begivendirectlyorcanbegeneratedbythepair-wisecomparison
Fig.6. Interfacetoselectconstraintcriteriaandassignthethresholdvalue.
Author's personal copy
M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75 67
oftheAHPfunctionistheconstructionofacriteriatreeaswellas
to calculate the relative weights of the criteria and of the sub-
criteriabypair-wisecomparison.AfterapplyingtheAHP,theWLC
or OWA are used. WLC computes the overall suitability for
eachalternativeorcellsusingthestandardizedmap,weights,and
constraintmap.OWAproducesthesuitabilitymapsbyspecifying
the linguistic quantifier (a set of ordered weights are generated,
whicharerelatedtoa;thegeneratedvaluesforeachalternativeare
combined).
By changing the weights of each overlay method and of the
linguistic quantifier associated with the objectives and attributes
forOWA,awiderangeofdecisionscenarioscanbegeneratedand
thecorrespondingmaplayersareaddedtothemapdocument.This
helpstocheckthesensitivityofthesystemwithchangingweights
andlinguisticquantifiers.
Areasonthesuitabilitymapcanbeclassifiedasverygood,good,
moderate,poor,andbad.Thesystemoffersfivedifferentcoloursfor
the five classes (Fig. 8), taking into account the colour code for
ecologicalstatusclassificationproposedbytheWaterFramework
Directive(WFD)(WaterFrameworkDirective,2003).Theuserhas
theopportunitytochangetherangeofclassmanually.
ThethirdstepisaspatialanalysisoftheoptimalMARlocations
with respect to water source locations. In a user-defined buffer
zone spatial query, the most favorable MAR locations based on
proximity towater sourceare chosen. The result is a raster map,
which shows the optimal MAR locations that satisfy the user
chosendistancetoproximalpotentialsourcesofwater.
5. Acasestudy
5.1. Problemdescription
Fig.7. Standardizationprocedure.
Duetothegeographicallocation,theAlgarveregioninsouthern
Portugal is prone for experiencing droughts, and the region has
been affected by many droughts over the last few decades. The
method. In thepair-wise comparison method, the usercan input hydrological year of 2004/2005 was extremely dry in the entire
preferredvaluesusingascalebar.Theweightsaregeneratedusing Portuguese mainland and especially in the Algarve region. The
thespecifiedformulabySaaty(1980). drought caused severe problems, considering the availability of
After finishing the weighting procedure, the user reaches the water resources. Surface water reservoirs reached volumes that
finalstepsofthesitesuitabilitymapping(Fig.8).Theuserchooses were below acceptable levels, and the Querença-Silves aquifer
anoverlayprocedure,eitherWLCorOWA.IntheOWAprocedure, system was over-exploited (Fig. 9). The Querença -Silves aquifer
thelinguisticquantifiersareassignedtoeachlevelofoverlay.The systemwas a major source of drinking water to the urban areas
resultingmapisthencreatedandshowninArcGISformat.Therole withintheAlgarveregion,duringthe90’s.TodaytheFunchoDamis
Fig.8. Theoverlayforthesuitabilityanalysis(left)andthereclassificationstepofthesuitabilitymap(right).
Author's personal copy
68 M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75
Fig.9. Studyarea(Querença-SilvesAquifer)mapshowingtheelevation(inmASL).
consideredtobethemostimportantdrinkingwatersourceinthe waterdemandofSilves,Lagoa,AlbufeiraandLoulé.Thisvaluewas
westernpartoftheAlgarve.Anotherdam,theAradeDamislocated higherduringthedroughtyearsof2004e2005.
downstreamofFunchoDamintheAraderiver.Morethan50hm3 Inthisstudy,onlythesouthwesternpartoftheQuerença-Silves
ofriverwaterwaslosttotheseaduringthewethydrologicalyearof Aquifer is being taken into account due to geology and aquifer
2000/2001,andindryhydrologicalyearof2004/2005therewasan properties.Thegroundwatercatchmentareais114km2.Foranal-
equivalent shortage of water resources (54hm3). (Lobo-Ferreira ysis purposes, the study area has been divided into four zones
and Oliveira, 2007). MAR is considered as a potential strategy to (Fig. 9), according to the residence time of groundwater in the
store water during the wet years and use it during dry years. aquifer. These are:Zone I(residence time is lessthan 6 months),
TheoverallplanningandmanagementofMARconsistsof:selection ZoneII(residencetimeis6monthsto1year),ZoneIII(residence
of water source, location of infiltration basin, and location for timeis1yearto3years)andZoneIV(residencetimeisgreaterthan
recoveryof theinfiltratedwater.Thisstudyfocusesonsuitability 3years).Thesezonesareoverlainineachconstraintandsuitability
mapping for the implementation of infiltration ponds for aquifer mapsoastoassesssuitableMARsitesaccordingtotheresidence
recharge. timezonation.ResultsofGISanalysisandofgroundwatermodeling
havebeenusedasspatialinputinformationforMARsiteselection
procedure.
5.2. Generalcharacteristicsofthetestsite
TheQuerença-SilvesAquiferSystemisa318km2aquifersystem, 5.3. Selectionofcriteriaforspatialanalysis
the largest of the Algarve, located in the municipalities of Silves,
Loulé,LagoaandAlbufeira(CentralAlgarve).Theaquiferismainly Afterdiscussionwithlocalandinternationalexpertsandinsti-
composed by karstified Lower Jurassic (Lias-Dogger) dolomite tutions and under the prevailing site characteristics and study
structures.Thesouthwesternpartoftheaquiferismainlyuncon- objectives,twodifferentsetsofcriteriawereselected:a)criteriafor
fined.ThegeneralgroundwaterflowdirectionisfromNortheastto constraint mapping and b) criteria for suitability mapping. Some
Southwest. According to the characterization of the Querença- importantcriteriawereselectedforbothcasesafteranalyzingtheir
Silvesaquifersystem(Almeidaetal.,2000),thehydraulicparam- importance and relevance. Table 1 lists the selected criteria for
etersareheterogeneousandaquiferproductivityvaluesarehigh. constraintandsuitabilitymapping,showingtherelevanceandthe
The transmissivity values range between 83 and 30,000m2/day usefulnessofeachcriterionforMARsitesuitabilitymapping.
andthestoragecoefficientrangesfrom5(cid:4)10(cid:3)3to3(cid:4)10(cid:3)2.INAG
(2001) presents the recharge value as being 220(cid:5)54mm/year. 5.4. Constraintmapping
This represents a percentage of precipitation of around 40(cid:5)10%.
Monteiro(2005)obtainedanaverage rechargeof292.5mm/year. In order to screen out the non-feasible areas, constraint
Theseareaveragevaluesusingtheaverageprecipitationvaluesin mappingwasundertakenatanearlystage.Table2showsthelist
thearea,thereforewhentheprecipitationismuchsmaller(e.g.the of criteria and their threshold values for screening. For the land
hydrologicalyearof2004/2005,whenprecipitationwasmorethan use map, land class feasibility was defined separately (Table 3).
halftheaverage)therechargeisalsomuchlower. The threshold values for each constraint criteria are chosen so
Analyzing69wellsoftheaquiferfortheyear2002,awithdrawal that criteria values or classes should satisfy the minimum
rateof19.5mm/yearwascomputedasbeingpossibletomeetthe requirementofMARimplementationsuchastheinfiltrationbasin
Author's personal copy
M.A.Rahmanetal./JournalofEnvironmentalManagement99(2012)61e75 69
Table1
ListofcriteriachosenforconstraintmappingandsuitabilitymappingandtheirrelevancetoMARsiteselection.
Criteria Inthemodel,usedfor Description
Landuse Constraintmapping TheexistinglanduseprovidesinformationaboutthelandavailabilityforMAR.
Forexample,areasthatareundercommercialandindustrialuse,arenon-feasible
areasforMARimplementation.
Slope(topography) Constraintmappingand Steeperslopesdonotpermittheimplementationofinfiltrationbasins.
Suitabilitymapping Furthermore,waterrunoffisdirectlyrelatedtoslopeangle.Flatareas
allowhighinfiltrationratesandissuitableforaquiferrecharge.
Thelowerthevalue,thehigherthepriority.
Infiltrationrate(soil) Constraintmappingand Infiltrationrateofsoilcontrolsthepenetrationofsurfacewaterintoan
suitabilitymapping aquifersystem.Soilswithhighinfiltrationcapacityaremoresuitablethan
thoseoflowinfiltrationcapacity.
Sub-surfaceimpermeablelayerthickness Constraintmappingand Thethicknessofimpermeablelayershouldnotbehigh,otherwisetheexcavation
suitabilitymapping costswouldbehigh.Thelowerthevalue,themoresuitabletheplace.
Groundwaterdepth Constraintmappingand Intermsofwaterqualityimprovementbynaturalattenuationprocesses,
suitabilitymapping considerableunsaturatedzonethicknessispreferred.Adeepergroundwater
levelbenefitsofthenaturalattenuationcapacityatthestudiedlocation.
Distancetogroundwaterpollutionsource Constraintmapping TheplaceofMARshouldhaveasufficientdistancefromgroundwaterpollutionsources.
Aquiferthickness Suitabilitymapping Suitablesitesshouldhavehighthicknessvalues.Transmissivityandaquiferstorage
volumedependsontheaquiferthickness.Thehigherthevalue,thehigherthepriority.
GWquality(chlorideandnitrate) Suitabilitymapping Thegroundwaterqualityshouldbeadequateattheplaceofrecharge,exceptthe
objectiveoftheMARistoimprovethegroundwaterquality.Theparameter
hastobeconsideredfunctionofthegroundwaterqualityofthearea.
Residencetime Constraintmappingand Theresidencetimeoftheinfiltratedwaterintheaquifershouldbesufficient
suitabilitymapping tobeabletousetheaquiferaswatertransferandrecoverysystem.
construction, the water quality improvement by using unsatu- as linear, piece-wise linear, and step-wise linear functions were
ratedzone,theaquiferstoragecapacityandothers.Forexample, usedforthisapproach(Fig.13).
the used threshold value for the residence time of 6 months, as Theimportanceofeachsub-criterionhasbeencalculatedusing
mentioned by most of the international standard guidelines for pair-wisecomparisonsandthisisshowninFig.12.Infiltrationrate
MAR(CDPH,2008;NRMMC,EPHC,NHMRC,2009)suggesttokeep of the soil, residence time of groundwater, and depth toground-
the water in the aquifer at least 6 months for water quality water were given highest priority in the analysis. Groundwater
improvement. quality was a low priority criterion because of low variability of
After defining the threshold values for each criterion, the groundwater quality over the entire area. The weighted criteria
thematic map of each constraint criterion has been converted to were then overlaid by two state-of-the-art overlay procedures:
aconstraintmap.Fig.10showsthethematicmapofslopeandthe WLCandOWA.
convertedconstraintmap.Alltheconvertedthematicmapswere Fig.15showsthesuitablesitesforMARintheregionusingthe
overlain by conjunctive screening to achieve the final constraint WLC method. Fig. 16 shows suitability maps using the OWA
map (Fig.11). This constraint map was used later as a mask for procedure.Themapshowsthesuitableplacesunderthefollowing
suitabilitymapping. decisioncondition:“half”oftheimportantcriteriaaresatisfiedby
anacceptablealternative.AccordingtothedefinitionoftheOWA,
5.5. Suitabilitymapping whena¼1,theoutputshouldcomplywiththeWLCoutput.
After analyzing all available data and site characteristics, sub- 5.6. Sensitivityanalysis
criteria were selected according to their characteristics, and the
mainhierarchicalstructurewasprepared(Fig.12).Thesub-criteria, A sensitivity analysis was done which indicates a significant
orthematiclayers,werestandardized.Threevaluefunctions,such change insitesuitability based on changes inriskacceptance of
Table2
Definedthresholdvalues(discardingconditions)oftheselectedcriteriaforMARconstraintmapping.
Criterianame Thresholdvalue Explanation
Landuse e SeeTable3.
Infiltrationrate(soil) 25cm/day Theareaswhereinfiltrationrateisgreaterthan25cm/dayareconsideredaspotentialarea.
Groundwaterdepth 5m Theplaceswheregroundwaterdepthisgreaterthan5mareconsideredaspotentialsites.
GWpollutionsources 500m Theplaceswhicharewithintheradiusof500mofgroundwaterpollutionsourcesarerejected.
Residencetime 6months Aresidencetimeofatleast6monthsshouldbeguaranteed.
Slope(topography) 5% MARisfeasibleforareaswithlessthan5%slope.
Table3
CategorizationofthelandusetypesatthestudyareaforMARconstraintmapping.
Landusetype Threshold
Agriculturalsystems,agriculturalareasoutsideirrigationperimeters,irrigatedareas,quarries/stonepits, Non-feasible(valueis0)
marshyplaces,salt-pits,isolatedurbanareas
Permanentcrops,orchards,poorpastures/grasslands,naturalvegetation,underwood,rivers Feasible(valueis1)
(waterinlinestobuildcheckdams,andinfiltrate)