Dynamic adaptive policy pathways: a new method for crafting robust decisions for a deeply uncertain world

Dynamic

adaptive

policy

pathways:

A

method

for

crafting

robust

decisions

for

a

deeply

uncertain

world

Marjolijn

Haasnoot

*

,

Jan

H.

Kwakkel

,

Warren

E.

Walker

,

Judith

ter

Maat

a_{UtrechtUniversity,DepartmentofGeosciences,P.O.Box80115,3508TCUtrecht,TheNetherlands} b

TwenteUniversity,DepartmentofWaterEngineering&Management,P.O.Box217,7500AEEnschede,TheNetherlands c

DelftUniversityofTechnology,FacultyofTechnology,PolicyandManagement,P.O.Box5015,2600GADelft,TheNetherlands d

Deltares,P.O.Box177,2600MHDelft,TheNetherlands

1. Introduction

Nowadays, decisionmakers face deep uncertainties about a

myriad of external factors, such as climate change, population

growth, new technologies, economic developments, and their

impacts.Moreover, notonlyenvironmental conditions,but also

societal perspectives and preferences may change over time,

including stakeholders’ interests and their evaluation of plans

(Offermans, 2010; van der Brugge et al., 2005). Traditionally,

decisionmakersinmanypolicydomains,includingwater

manage-ment,assumethat thefuturecan bepredicted.Theydevelop a

static‘optimal’planusingasingle‘mostlikely’future(oftenbased

ontheextrapolationoftrends)orastatic‘robust’planthatwill

produce acceptable outcomes in most plausible future worlds

(Dessai and Hulme, 2007; Dessai and Van der Sluijs, 2007; Hallegatteet al., 2012). However, if thefuture turns out to be

differentfromthehypothesizedfuture(s),theplanislikelytofail.

McInerney et al. (2012)liken this to‘‘dancing on thetop of a

needle’’.But,asthefutureunfoldspolicymakerslearnandusually

respondtothenewsituationbyadaptingtheirplans(adhoc)tothe

new reality. Adaptation over the course of time is not only

determinedbywhatisknownoranticipatedatpresent,butalsoby

whatisexperiencedandlearnedasthefutureunfolds(Yohe,1990)

andbythepolicyresponsestoevents(Haasnootetal.,2012).Thus,

policymaking becomes part of the storyline, and thereby an

essentialcomponentofthetotaluncertainty–infact,Hallegatte

etal.(2012)includetheadaptationofdecisionsovertimeinan

updateddefinitionof‘deepuncertainty’.

Toaddressthesedeepuncertainties,anewplanningparadigm

has emerged. This paradigm holds that, in light of the deep

uncertainties, one needs to design dynamic adaptive plans

(Albrechts,2004;deNeufvilleandOdoni,2003;Haasnootetal., 2011;Hallegatte,2009;Hallegatteetal.,2012;Rangeretal.,2010; SchwartzandTrigeorgis,2004;Swansonetal.,2010).Suchplans

contain a strategic vision of the future, commit to short-term

actions, and establish a framework to guide future actions

(Albrechts,2004;Rangeretal.,2010).Theseedsforthisplanning

paradigmwereplantedalmostacenturyago.Dewey(1927)argued

ARTICLE INFO

Articlehistory: Received15June2012

Receivedinrevisedform3December2012 Accepted18December2012 Keywords: Uncertainty Policymaking Adaptationpathways Adaptivepolicies Watermanagement Rhinedelta ABSTRACT

Anew paradigmforplanningunderconditionsofdeepuncertainty hasemergedin theliterature. Accordingtothisparadigm,aplannershouldcreateastrategicvisionofthefuture,committoshort-term actions,andestablishaframeworktoguidefutureactions.Aplanthatembodiestheseideasallowsforits dynamic adaptation over time to meet changing circumstances. We propose a method for decisionmaking under uncertain global and regional changes called ‘Dynamic Adaptive Policy Pathways’.Webaseourapproachontwocomplementaryapproachesfordesigningadaptiveplans: ‘AdaptivePolicymaking’and‘AdaptationPathways’.AdaptivePolicymakingisatheoreticalapproach describingaplanningprocesswithdifferenttypesofactions(e.g.‘mitigatingactions’and ‘hedging actions’)andsignpoststo monitorto seeifadaptationisneeded.In contrast,AdaptationPathways provides an analytical approachfor exploring and sequencing a set ofpossible actions based on alternativeexternaldevelopmentsovertime.We illustratetheDynamicAdaptivePolicyPathways approachbyproducinganadaptiveplanforlong-termwatermanagementoftheRhineDeltainthe Netherlandsthattakesintoaccountthedeepuncertaintiesaboutthefuturearisingfromsocial,political, technological,economic,andclimatechanges.Theresultssuggestthatitisworthwhiletofurthertest andusetheapproach.

ß2012ElsevierLtd.Allrightsreserved.

*Corresponding author at: Deltares, P.O. Box 177, 2600 MH Delft, The Netherlands.Tel.:+31883358175.

E-mailaddresses:Marjolijn.Haasnoot@deltares.nl(M.Haasnoot),

J.H.Kwakkel@tudelft.nl(J.H.Kwakkel),W.E.Walker@tudelft.nl

(W.E.Walker),Judith.TerMaat@deltares.nl(J.terMaat).

ContentslistsavailableatSciVerseScienceDirect

Global

Environmental

Change

j ou rna l hom e pa ge : w w w. e l s e v i e r. c om/ l o ca t e / gl oe n v cha

0959-3780/$–seefrontmatterß2012ElsevierLtd.Allrightsreserved.

thatpolicies shouldbe treatedasexperiments,with the aimof

promoting continual learning and adaptation in response to

experienceovertime.Earlyapplicationsofadaptiveplanscanbe

foundinthefieldofenvironmentalmanagement(Holling,1978;Lee,

1993;McLainandLee,1996),andinvolvetheabilitytochangeplans

basedonnewexperience andinsights (Pahl-Wostletal.,2007).

Collingridge(1980)arguesthat,givenignoranceaboutthepossible

sideeffectsoftechnologiesunderdevelopment,oneshouldstrivefor

correctability of decisions, extensive monitoring of effects, and

flexibility.Rosenhead(1990)andRosenheadetal.(1972)presented

flexibility,in terms of keepingoptions open,as an indicator to

evaluatetherobustnessofstrategiesunderuncertainty.

This planning paradigm, in one form or another, has been

receivingincreasingattentioninvariouspolicydomains.Dynamic

adaptiveplansarebeingdevelopedforwatermanagementofNew

York(Rosenzweiget al.,2011;YoheandLeichenko,2010),New

Zealand(LawrenceandManning,2012),andtheRhineDelta(Delta

Programme,2011,2012;JeukenandReeder,2011;Roosjenetal.,

2012), and havebeendeveloped fortheThamesEstuary(Lowe

etal.,2009;McGaheyandSayers,2008;ReederandRanger,online; Sayersetal.,2012;WilbyandKeenan,2012).Suchapplicationsare

also arising in other fields (see Swanson and Bhadwal, 2009;

Walkeretal.,2010forexamples).

Alarge numberofapproachesandcomputationaltechniques

existtosupportdecisionmakingunderdeepuncertainty(seee.g.

DessaiandVanderSluijs,2007;Hallegatteetal.,2012;IISD,2006; Metzetal.,2001;Swansonetal.,2010;Walkeretal.,acceptedfor

an overview of a strand of approaches). With respect to

approaches,theThames2100projectuseddecisiontreestoanalyze

sequentialdecisionsforpreparingtheThamesEstuaryforfuture

sealevelrise.IntheNetherlands,RealOptionsAnalysishasbeen

usedtoassess optimalcostsand benefitsof pathwaysforfresh

watersupplyoftheSouthwesternDelta(vanRhee,2011)andfor

studyinghowflexibilitycanbebuiltintofloodriskinfrastructure

(Gersonius et al., 2013). To show dependencies of choices for

shipping, a decision tree has been used in the Dutch Delta

Programme(DeltaProgramme,2011).Roadmapshavebeenused

toillustrateasequenceofactionsinwatermanagementstudies

(e.g.forthelakesIJsselmeer(unpublished)andVolkerak

Zoomm-eer (Projectteam Verkenning oplossingsrichtingen Volkerak-Zoommeer,2003). TheBackcasting approachaimsat describing

adesirablefuture,andthenlookingbackwardsfromthatfutureto

thepresenttodevelopapathwayofactionsneededtorealizethis

future (Ho¨jer and Mattsson, 2000; Lovins, 1976; Quist and

Vergragt, 2006). Assumption-Based Planning begins with an

existingplanand analyzes thecritical assumptions inthis plan

(Dewaret al., 1993). Ituses signposts tomonitor theneed for

changes.RobustDecisionMakingisanapproachthatusesmany

computationalexperiments tocreate an ensemble of scenarios

againstwhichcandidateactionsareevaluatedinordertodevelop

robustactions(GrovesandLempert,2007;Lempertetal.,2006).

Severalplanningapproachesconsiderreassessmentandtheability

to change policies based on new insights in a planning circle

(LoucksandVanBeek,2005;Pahl-Wostl,2007;Rangeretal.,2010; Swansonetal.,2010;WillowsandConnell,2003).ThePanelon

America’sClimateChoices(2010)referstothis as‘iterativerisk

management’ that ‘is a system for assessing risks, identifying

optionsthat are robust acrossa range of possiblefutures, and

assessingandrevisingthosechoicesasnewinformationemerges.’

Among the computational techniques are Scenario Discovery

(BryantandLempert,2010;LempertandGroves,2010),

Explor-atoryModelingandAnalysis(Bankes,1993;Bankesetal.,2013),

andInfo-Gapdecision theory(Hall andHarvey,2009;Korteling

etal.,2012).

These approaches and computational techniques, although

developedfor differentpurposes,have beenfoundvaluable for

designing adaptive policies (Bankes, 2002; Hall et al., 2012;

Hallegatteetal.,2012;Hamaratetal.,2012;Lempertetal.,2000,

2002).Theydifferintermsoftheconceptsemployed,andprovide

differentkinds decision supportinformation(Hall et al.,2012).

Consequently,theyhavedifferentstrengthsandlimitations.This

situationcallsforresearchintocomparingthevariousapproaches

and techniques, providing an understanding of their relative

strengths and weaknesses, and identifyingthe contexts within

whicheachoftheapproachesandtechniquesis most

appropri-atelyemployed(Halletal.,2012;Hallegatteetal.,2012;Ranger

etal.,2010).Inaddition,wearguethatitisworthwhiletoassess

theextenttowhichthedifferentterminologiesusedsignifyreal

differencesintheunderlyingconcepts,forthiscancontributeto

harmonizingthefield.

In this article, we analyze two existing adaptive planning

approaches andshowhowtheemployed conceptsare partially

overlappingandpartiallycomplementary,resultinginan

integra-tion of the two approaches. We look at Adaptive Policymaking

(Kwakkel et al., 2010a; Walker et al., 2001) and Adaptation

Pathways(Haasnootetal.,2012).AdaptivePolicymakingprovides

astepwiseapproachfordevelopingabasicplan,andcontingency

planningtoadapt thebasicplantonewinformationovertime.

Adaptation Pathways provide insight into the sequencing of

actionsovertime,potentiallock-ins,andpathdependencies.An

exampleofafamilyresemblancebetweenconceptsusedbythese

two approaches is the concept of an adaptation tipping point

(Kwadijketal.,2010)usedinAdaptationPathwaysandthenotion

of a trigger fromAdaptive Policymaking. Anadaptation tipping

pointisthepointatwhichaparticularactionisnolongeradequate

for meeting the plan’s objectives. A new action is therefore

necessary.Atriggerspecifiestheconditionsunderwhicha

pre-specifiedactiontochangetheplanistobetaken.

Afundamentalchallengeinplanningresearchistheassessment

oftheefficacyofnewplanningmethodsandconcepts.Theproblem

ispointedlysummarizedbyDewaretal.(1993,p.58)‘‘nothingdone

intheshorttermcan‘prove’theefficacyofaplanningmethodology,

norcanthemonitoring,overtime,ofasingleinstance ofa plan

generatedbythatmethodology,unlessthereisacompetingparallel

plan’’.Withrespecttohowaplanningconceptistested,theplanning

research literature tends to look toward controlled real world

application(Dewaretal.,1993;Hansmanetal.,2006;Straatemeier

etal.,2010).However,analogoustootherdesignsciences(Freyand

Dym,2006),theevaluationofaplanningconceptcanalsoutilize

othersourcesofevidence(KwakkelandVanDerPas,2011;Kwakkel

etal.,2012).Evidencecancomefromplanningpractice,fromvirtual

worldsthatrepresenttheworldofpracticebutarenottheworldof

practice(Scho¨n,1983),andfromtheoreticalconsiderations.Inthis

paper,toassesstheefficacyoftheoutlinedintegrationofAdaptive

PolicymakingandAdaptationPathways,weusesuchavirtualworld

intheformofapplyingthepresentedplanningconceptstoareal

world decision problem currently faced by the Dutch National

Government. This application serves to illustrate the concept,

describeshowitcouldbeusedtodevelopadynamicadaptiveplan,

andoffersafirstsourceofevidenceofitsefficacythroughacritical

reflectionontheapplication.

Thepaperultimately proposes amethodfor decisionmaking

underdeepuncertaintycalledDynamicAdaptivePolicyPathways,

whichisacombinationofAdaptivePolicymakingandAdaptation

Pathways. We first provide short introductions to each of the

underlyingapproaches,andthenexplorehowthetwoapproaches

can be integrated into a single approach based on the strong

elements of both to produce a dynamic adaptive plan. We

demonstratetheapproachbyproducingadynamicadaptiveplan

forwatermanagementoftheRhineDeltaregionoftheNetherlands

that takes into account the deep uncertainties associated with

2. Thetwounderlyingapproaches

2.1. AdaptationPathways

TheAdaptationPathways approachis summarizedin Figs.1

and2(Haasnootetal.,2011,2012).Centraltoadaptationpathways

areadaptiontippingpoints(Kwadijketal.,2010),whicharethe

conditions under which an action no longer meets the clearly

specifiedobjectives.Thetimingoftheadaptationpointforagiven

action, its sell-bydate, is scenariodependent. After reaching a

tippingpoint,additionalactionsareneeded.Asaresult,apathway

emerges.TheAdaptationPathwaysapproachpresentsasequence

ofpossibleactionsafteratippingpointintheformofadaptation

trees (e.g. like a decision tree or a roadmap). Any given route

throughthetreeisanadaptationpathway.Typically,thisapproach

usescomputationalscenarioapproachestoassessthedistribution

ofthesell-bydateofseveralactionsacrossalargeensembleof

transientscenarios.Thisdistributioncanbesummarizedin

box-whisker plots, and the median or quartile values are used in

generatinganadaptationmap.Theexactdateofatippingpointis

notimportant;themomentshouldberoughlyright—forexample,

‘‘onaveragethetippingpointwillbereachedwithin50years,at

earliestwithin40years,andatlatestwithin60years’’.Theeffects

of sequences of actions can be assessed in the same way as

individual actions. To cope with the presence of different

stakeholders,values, and worldviews, culturalperspectives can

beusedtomaptheseout(Hoekstra,1998;Middelkoopetal.,2004;

Offermansetal.,2011;VanAsseltandRotmans,1997).

The Adaptation Pathways map, manually drawn based on

modelresultsorexpertjudgment,presentsanoverviewofrelevant

pathways(seeFig.2foranexample).SimilartoaMetromap(see,

for example, http://www.wmata.com/rail/maps/map.cfm), the

Adaptation Pathwaysmap presents alternativeroutes toget to

thesamedesiredpointinthefuture.Allroutespresentedsatisfya

pre-specifiedminimumperformancelevel,suchasasafetynorm(a

thresholdthatdetermineswhetherresultsareacceptableornot).

Theycan,thus,beconsideredas‘differentwaysleadingtoRome’

(asis true of differentroutes toa specifieddestinationon the

Metro).Also,themomentofanadaptationtippingpoint(terminal

station),andtheavailableactionsafterthispoint,areshown(via

transfer stations). Due to unacceptable performance of some

actionsin a selection of scenarios,someroutes are not always

available(dashedlines).Decisionmakersorstakeholdersmayhave

apreferenceforcertainpathways, sincecostsandbenefitsmay

differ.Anoverviewofsuchcostsandbenefitsforeachpathwaycan

be presented in a scorecard (e.g. Walker, 2000). With the

adaptationmap, decisionmakers canidentifyopportunities,

no-regret actions,lock-ins, andthetimingofanaction,inorderto

supportdecisionmakinginachangingenvironment.Thatis,the

adaptationmapcanbeusedtopreparea planforactionstobe

takenimmediately,andforpreparationsthatneedtobemadein

order to beable to implement an action in the future in case

conditionschange.TheexampleofFig.2showsthatactionsare

neededintheshort-term.ChoosingactionBmaybeineffectiveas

soon additionalactionsare needed.ChoosingoptionCinvolves

takingarisk,asadditionalactionsmaybeneededincasescenario

Xbecomesreality.Incombinationwithascorecardofthecostsand

benefits for the pathways, a decisionmaker could make an

informeddecision.

2.2. AdaptivePolicymaking

Adaptive Policymaking is a generic structured approach for

designingdynamicrobustplans(Kwakkeletal.,2010a;Marchau

etal.,2009; Rangeret al.,2010). Conceptually,Adaptive

Policy-making is rootedin Assumption-Based Planning (Dewar et al.,

1993). Fig. 3 shows the steps of the Adaptive Policymaking

approachfordesigningadynamicadaptiveplan(Kwakkeletal.,

2010a).InStepI,theexistingconditionsofasystemareanalyzed

andtheobjectivesforfuturedevelopmentarespecified.InStepII,

thewayinwhichtheseobjectivesaretobeachievedisspecifiedby

assembling a basic plan. This basic plan is made more robust

throughfourtypesofactions(StepIII):mitigatingactions(actions

to reduce the likely adverse effects of a plan); hedging actions

(actionstospreadorreducetheuncertainadverseeffectsofaplan);

seizing actions (actions taken to seize likely available

opportu-nities); and shaping actions (actions taken to reduce failure or

enhancesuccess).EvenwiththeactionstakeninStepIII,thereis

stilltheneedtomonitortheplan’sperformanceandtotakeaction

ifnecessary.Thisiscalledcontingencyplanning(StepIV).Signposts

specifyinformationthatshouldbetrackedinordertodetermine

whether the plan is meeting the conditions for its success.In

addition, critical values of signpost variables (triggers) beyond

which additionalactions should be implementedare specified.

Therearefourdifferenttypesofactionsthatcanbetriggeredbya

signpost,whicharespecifiedinStepV:defensiveactions(actions

takentoclarifythebasicplan,preserveitsbenefits,ormeetoutside

challengesinresponsetospecifictriggersthatleavethebasicplan

unchanged); corrective actions (adjustments to the basic plan);

capitalizingactions(actionstotakeadvantageofopportunitiesthat

canimprovetheperformanceofthebasicplan);andareassessment

oftheplan(initiatedwhentheanalysisandassumptionscriticalto

theplan’ssuccesshaveclearlylostvalidity).

Oncethecompleteplanhasbeendesigned,theactionstobe

takenimmediately(fromStepIIandStepIII)areimplemented,and

amonitoringsystem(fromStepIV)isestablished.Thentimestarts

running,signpostinformationrelatedtothetriggersiscollected,

andactionsarestarted,altered,stopped,orexpandedinresponse

tothisinformation.Afterimplementationoftheinitialactions,the

implementationofotheractions(fromStepV)issuspendeduntila

triggereventoccurs.

2.3. Comparisonoftheapproaches

Table1compares thefeaturesof AdaptivePolicymaking and

Adaptation Pathways. Bothapproaches aimat supporting

deci-sionmakersinhandlinguncertaintyinlong-termdecisionmaking

andemphasizetheneedforadaptivityinplansinordertocope

withdeepuncertainty.Morespecifically,theybothoffersupportin

choosingnear-termactions,whilekeepingopenthepossibilityto

modify,extend,orotherwisealtertheplansinresponsetohowthe

futureunfolds.

Evaluateactions&develop pathways

Policyanalysis

Describecurrent&future situations,objectives

Problemanalysis

Determineactions

Analyseensemblesof transientscenarios Determinesell-bydate

ofactions

Currentpolicy ActionA ActionB ActionC ActionD 0 10 70 80 90_years100 1 2 3 4 5 6 7 8 9 Path actions Side effects Relative Costs Target effects +++ ++ 0 -- -0 0 0 --- -+ 0 +++ +++ +++++ ++++ +++ + + 0 0 0 0 0 0 + Scorecardpathways AdaptationPathwaysMap

Transferstationtonewaction

AdaptationTippingPointofanaction(Terminal) Actioneffectiveinallscenarios

ActionnoteffectiveinscenarioX

Fig.2.AnexampleofanAdaptationPathwaysmap(left)andascorecardpresentingthecostsandbenefitsofthe9possiblepathwayspresentedinthemap.Inthemap, startingfromthecurrentsituation,targetsbegintobemissedafterfouryears.Followingthegraylinesofthecurrentpolicy,onecanseethattherearefouroptions.ActionsA andDshouldbeabletoachievethetargetsforthenext100yearsinallclimatescenarios.IfActionBischosenafterthefirstfouryears,atippingpointisreachedwithinabout fiveyears;ashifttooneoftheotherthreeactionswillthenbeneededtoachievethetargets(followtheorangelines).IfActionCischosenafterthefirstfouryears,ashiftto ActionA,B,orDwillbeneededinthecaseofScenarioX(followthesolidgreenlines).Inallotherscenarios,thetargetswillbeachievedforthenext100years(thedashed greenline).ThecolorsinthescorecardrefertheactionsA(red),B(orange),C(green),andD(blue).

Necessary Conditions for Success Objectives Constraints Definitions of Success Options Set Policy Actions

II. Assembling the Basic Plan

Signposts Mitigating Actions (M) Hedging Actions (H) Triggers Likely Vulnerabilities Uncertain Vulnerabilities

III. Increasing the Robustness of the Basic Plan

V. Preparing the Trigger Responses

Defensive Actions (DA)

Corrective Actions (CR)

Reassessment (RE)

Other’s Actions

Unforeseen Events

Changing Preferences I. Setting the Stage

Vulnerabilities and

Opportunities

Likely Opportunities

Seizing Actions (SZ)

IV. Setting up the Monitoring System

Capitalizing Actions (CP)

Shaping Actions (SH)

Thewaysinwhichthetwoapproachesofferdecisionsupport

arequitedifferent.AdaptationPathwaysprovidesinsightintothe

sequencing of actions over time, taking into account a large

ensembleoftransientscenarios.Thetransientscenariosallowfora

wide variety of uncertainties about future developments to be

takenintoaccountintheplanningprocess.Notonlytrendsand

systemchangesareincluded,butalsouncertaintyduetonatural

variability.Theuseofafastandsimplemodelallowsforexploring

awidevarietyofpathwaysovertheensemble.Theseresultscanbe

usedtosketchanAdaptationPathwaysmap.Dynamicrobustness

oftheresultingplanisindirectlyhandledthroughthe

identifica-tionofanadaptationtippingpoint,thesell-bydate,andtheshiftto

other actions. The pathways map provides information to the

decisionmaker,butgivesnoguidanceonhowthedecisionmaker

cantranslatethisintoanactualplan.

Adaptive Policymaking supports the decisionmaker in a

different way. It specifies a stepwise approach to designing a

plan.First abasiccourseofaction isdevelopedin light ofwell

specifiedobjectives.Then,thevulnerabilitiesandopportunitiesof

thiscourseofactionareidentified,anddifferenttypesofactionsto

betakennoworinthefuturetoeithercopewiththevulnerabilities

or capitalize on the opportunities are specified. Through the

identificationofopportunitiesandvulnerabilities,awidevarietyof

uncertainties can be accounted for. The specification of a

monitoringsystemandassociatedactionsresultsinadynamically

robust plan. However, Adaptive Policymaking offers no clear

guidancebeyondtheseconcepts.Thatis,questions,suchashow

can one identify vulnerabilities, how should the actions be

sequenced,orhowdoesonedecide whethertohedgeagainsta

vulnerabilityortospecify a monitoringsystemwithactions to

handlethevulnerabilityinthefutureifandwhenitarises,arenot

addressedexplicitly.

3. Anewapproach:dynamicadaptivepolicypathways

The combination of Adaptive Policymaking and Adaptation

Pathways, which we call Dynamic Adaptive Policy Pathways,

resultsfromusingthestrengthsofbothapproaches.Inshort,this

integratedapproach includes: transient scenarios representinga

varietyofrelevantuncertaintiesandtheirdevelopmentovertime;

differenttypesofactionstohandlevulnerabilitiesand

opportu-nities; Adaptation Pathways describing sequences of promising

actions;andamonitoringsystemwithrelatedcontingencyactionsto

keeptheplanonthetrackofapreferredpathway.Thestepsinthe

approacharepresentedinFig.4.

Thefirststepistodescribethestudyarea,includingthesystem’s

characteristics, the objectives, the constraints in the current

situation,andpotentialconstraintsinfuturesituations.Theresult

is adefinitionof success,whichis aspecification ofthedesired

outcomes in terms of indicators and targets that are used in

subsequent steps to evaluate the performance of actions and

pathways, and to assess the ‘sell-by dates’ of the actions. The

descriptionofthestudyareaincludesaspecificationofthemajor

uncertaintiesthatplayaroleinthedecisionmakingproblem.These

uncertaintiesarenotrestrictedtouncertaintiesaboutthefuture,

Table1

Comparisonoftheapproaches.

Aspect AdaptivePolicymaking AdaptationPathways

Focus Startsfromavisionofthedecisionmakerandcreates

aplanforrealizingthisvisionandprotectingitfrom failure.

Exploresactionsforachievingobjectivesovertime byincludingdynamicinteractionbetweenthe systemandsociety.

Considerationofthemultiplicityoffutures Indirectlyviavulnerabilitiesandopportunities. Explicitlyviatransientscenarios. Planningprocess Comprehensivestepwiseapproachfordesigninga

plan.

ShortstepwiseapproachfordesigningAdaptation Pathways.

Clarityonhowtodesignaplan Limited;ahighlevelframeworkthatcanbe translatedintoaspecificplaninmanydifferent ways.

Applicationoriented,withaclearlinktotheuseof modelstodevelopaspecificplan.

Typesofactionsthatcanbetaken Distinguishesmanydifferenttypesofactionsthat canbetaken(e.g.hedging,mitigating,andshaping).

Nospecificcategorizationofactionsisused.Several actionsandpathwaysarepresented.Avarietyof actionsareidentifiedbasedondifferentsocietal perspectives.

Desirableplan Onebasicplanisdeveloped.Noclearguidelineon howdevelopthebasicplan.

Severalpathwaysarepresented.Different perspectivesresultindifferentpreferredpathways. Nofocusonhowtoidentifypromisingpathways whenconfrontedwithalargenumberofpossible actions.

Considerationoftypesofuncertainties Inprinciple,anyuncertaintycanbeaccountedfor. Inprinciple,anyuncertaintycanbeaccountedfor. Explicitattentionisgiventosocialuncertainty. Flexibilityofresultingplan Flexibilityisestablishedthroughthemonitoring

systemandassociatedactions.

TheAdaptationPathwaysmapclearlyspecifieswhen apolicyshouldbechanged,andwhatthenextaction shouldbe.

Dynamicrobustnessofresultingplan Dynamicrobustnessresultsfromthemonitoringset upinStepIVandtheactionstakeninStepV.

Dynamicrobustnessisproducedindirectlyviathe ideaofa‘sell-bydate’andtheshifttoanotheraction.

butcanalsocoveruncertaintiesrelatedtothedataormodelsthat

arebeingused(Kwakkeletal.,2010b).

Thesecondstepistheproblemanalysis.Inthisstep,thecurrent

situation and possible future situations are compared to the

specifiedobjectivestoidentifywhetherthereareanygaps.The

possiblefuturesituationsare‘referencecases’assumingno new

policiesareimplemented,andconsistof(transient)scenariosthat

spantheuncertaintiesidentifiedinstepone.Agapindicatesthat

actionsareneeded.Bothopportunitiesandvulnerabilitiesshould

beconsidered. Opportunitiesaredevelopmentsthatcanhelpin

achievingtheobjectives,whilevulnerabilitiesaredevelopments

thatcanharmtheextenttowhichtheobjectivescanbeachieved.

Theidentificationofopportunitiesandvulnerabilitiescanbebased

on the analysis of the reference cases, which can best be

accomplishedusingacomputationalmodel.

Inthethirdstep,oneidentifiespossibleactionsthatcanbetaken

to meet the definition for success. These actions can thus be

specifiedinlight oftheopportunitiesand vulnerabilities

previ-ouslyidentifiedandcanbecategorizedaccordingtothetypesof

actions specifiedin the Adaptive Policymaking framework (i.e.

shaping,mitigating,hedging,andcapitalizingactions).Theaimof

this step is to assemble a rich set of possible actions. An

identificationof actionsfordifferentperspectives couldenforce

this(e.g.donebyOffermansetal.,2011).

The fourth step is toevaluate the actions. Theeffects of the

individualactionsontheoutcomeindicatorsareassessedforeach

ofthescenariosandcanbepresentedusingscorecards.Theresults

are used to identify the sell-by date for each of the actions.

Furthermore, the vulnerabilities and opportunities need to be

reassessed.Wastheactionabletoreduceorremoveaspecified

vulnerability?Wastheactionabletoutilizeaspecified

opportu-nity?Does the action create new opportunitiesand/or

vulner-abilities?Ineffectiveactionsarescreenedout(Walker,1988),and

onlythepromisingactionsareusedinthenextstepsasthebasic

buildingblocksfortheassemblyofAdaptationPathways.

Thefifthstepistheassemblyofpathwaysusingtheinformation

generated in the previous steps. It is conceivable that the

reassessment of the vulnerabilities and opportunities in the

previoussteptriggersaniterativeprocess(backtostep3)wherein

neworadditionalactionsareidentified.Oncethesetofactionsis

deemedadequate,pathwayscanbedesigned.Apathwayconsists

ofaconcatenationofactions,whereanewactionisactivatedonce

itspredecessorisnolongerabletomeetthedefinitionofsuccess.

Pathways can be assembled in different ways. For example,

analystscouldexploreallpossiblerouteswithallavailableactions.

Eachoftheseroutescanthenbeevaluatedon itsperformance.

However,someactionsmayexcludeothers,andsomesequencesof

actionsmaybeillogical.Inaddition,fundamentalcriteria,suchas

theurgencyofactions,theseverityoftheimpacts,theuncertainty

involved,and thedesiretokeepoptionsopen,couldbeusedto

developa setofpromisingpathways.Theresultisanadaptation

map,whichsummarizesalllogicalpotentialpathwaysinwhich

‘success’(asdefinedinstep1)isachieved.Notethatactionsneed

notbeasingleaction,butcanbeaportfolioofactions,constructed

afteriterationofsteps3–5.

Thesixthstepistodevelopamanageablenumberofpreferred

pathways.Preferredpathwaysarepathwaysthatfitwellwithina

specified perspective. It can be useful to specify two to four

pathwaysthatreflect differentperspectives.Thiswillresultnot

onlyintheidentificationofphysicallyrobustpathways,butalso

‘sociallyrobust’pathways(Offermansetal.,2011).Thepreferred

pathwayswillformthebasicstructureofadynamicadaptiveplan

(likethebasicplanintheAdaptivePolicymakingframework).

Theseventhstepistoimprovetherobustnessofthepreferred

pathwaysthroughcontingencyplanning–inotherwords,todefine

actionstogetandkeepeachofthepathwaysontrackforsuccess.

Ingeneral,theseareactionstoanticipateandprepareforoneor

morepreferredpathway(e.g.keepoptionsopen),andcorrective

actionstostayontrackincasethefutureturnsoutdifferentlythan

expected.Wedistinguishthreetypesofcontingencyactionsfrom

Adaptive Policymaking: corrective, defensive, and capitalizing

actions,whichareassociatedwithamonitoringsystemandtrigger

values.Themonitoringsystemspecifieswhattomonitor,andthe

triggersspecifywhenacontingencyactionshouldbeactivated.

Theeighthstepistotranslatetheresultsfromalloftheprevious

stepsintoa dynamicadaptive plan.Thisplanshouldanswerthe

followingquestion:Giventhesetofpathwaysandthe

uncertain-tiesaboutthefuture,whatactions/decisionsshouldwetakenow

(and which actions/decisions can be postponed)? The plan

summarizestheresultsfromtheprevioussteps,suchastargets,

problems,andpotentialandpreferredpathways.Thechallengeis

todraftaplanthatkeepsthepreferredpathwaysopenforaslong

as possible. Thus, the plan specifies actions to be taken

immediately, actions to be taken now to keep open future

adaptations,andthemonitoringsystem.

Finally,theactionstobetakenimmediatelyareimplemented

andthemonitoringsystemisestablished.Then,timestartsrunning,

signpostinformationrelatedtothetriggersiscollected,andactions

are started, altered, stopped, or expanded in response to this

information.Afterimplementationoftheinitialactions,activationof

otheractionsissuspendeduntilatriggereventoccurs.

4. Casestudy:RhineDeltaintheNetherlands

WeillustrateandtesttheapproachofDynamicAdaptivePolicy

PathwaysforthelowerRhineDeltaintheNetherlands,andfocus

ontheIJsselmeerarea.In2007,theGovernmentestablishedthe

SecondDeltaCommissionforidentifyingactionstopreventfuture

disasters (Deltacommissie, 2008; Kabat et al., 2009), since the

expectedfutureclimatechangeandsealevelrise‘cannolongerbe

ignored’ (Deltacommissie,2008, p. 5).The Commission’s advice

resultedintheenactmentofa DeltaAct,andispresentlybeing

elaborated in a Delta Programme. The chair of the Delta

Programme summarized their main challengeas follows: ‘‘One

ofthe biggestchallengesisdealingwithuncertaintiesinthefuture

climate,butalsoinpopulation,economyandsociety.Thisrequiresa

newwayofplanning,whichwecalladaptivedeltaplanning.Itseeksto

maximize flexibility; keeping options openand avoiding ‘lock-in’’’

(Kuijken, 2010). This corresponds well with our integrated

approach, and thus provides an appropriate case to useas an

illustration.However,wehavemademanysimplifying

assump-tions.So,whatfollowscanbeusedonlyforillustrativepurposes

andafirsttentativetestofourapproach.Thestepswemention

refertothestepsinFig.4.

4.1. Steps1and2:currentsituationandproblemanalysis

TheNetherlandsisadenselypopulatedcountry,two-thirdsof

whichisvulnerabletobeingfloodedbytheseaorlargerivers.A

sophisticated and comprehensive water management system

satisfies the water system requirements for living in a delta.

But,forcopingwithfuturechangessuchasglobalclimatechange,

adaptationmaybeneeded.Havingtherightamountofwaterfor

users,attherighttime,intherightplace,andatsociallyacceptable

costsisakeytargetfortheMinistryofTransport,PublicWorksand

WaterManagement(Rijkswaterstaat,2011).Theobjectiveofthe

DeltaProgrammeis‘‘toprotecttheNetherlandsfromfloodingandto

ensureadequatesuppliesoffreshwaterforgenerationsahead.’’(Delta

Programme,2011). Accordingly,we define‘success’ asfollows:

‘Theplanwillbesuccessfulifnofloodsoccur,andifthereisenough

freshwaterduringthenext100years.Thefrequencyofwatershortage

watershortagemayoccur).’Constraintswouldincludethevarious

EU Directives that the Dutch Government must follow. For

example,theWaterFrameworkDirectiveimpliesthatecological

and water quality objectives have to be met. These Directives

imply that we need to add anothertarget to our definitionof

success:‘theplanwillbesuccessfulifitdoesnotresultinnegative

impactsonnature’.

4.1.1. Thewatersystemanditsfunctionsinthecurrentsituation

Thereareseveral keywatercharacteristicsthatneedfurther

explanationforourcase(see Fig.5).After theRhineentersthe

country,thewaterisdistributedoverthreebranches–theWaal,

Nederrijn, and IJssel – by means of a weir at Driel. The IJssel

suppliestheIJsselmeerandMarkermeerlakeswithfreshwater.

TheAfsluitdijkdamprotectstheadjacentareasfromfloodingand

enableswaterstorageinthelakes.ThelevelsoftheIJsselmeerand

Markermeerarecarefullymaintainedwithsluices,toensuresafety

inthewinterandenoughfreshwaterinthesummer.Safetyfrom

floodingisexpressedinstandardsofaprobabilityperyearthata

criticalwaterlevelwilloccur–e.g.1:1250years(Rijkswaterstaat,

2011). Thesestandards(also called‘norm frequencies’) arelaid

downbylawforeverydikeringarea,anddependlargelyonthe

economic activities, the number of inhabitants, and flood

characteristics associated with the dike ring. The Haringvliet

sluicegatesandtheMaeslantkeringprotecttheRhineestuaryfrom

(mainlycoastal) flooding.TheHaringvliet sluicesalsolimit salt

intrusionintotheriver.

TheIJsselmeerandMarkermeerarethemainwaterreservoirs

in thelowerRhineDelta.During dryperiods, waterfromthese

lakesisusedtosupplylargepartsoftheNetherlands.Despitethe

extensivenetworkofditchesandcanalsandthelargeamountof

waterstorage,thewatersupplyisinsufficienttofulfillthefresh

waterdemandsduringdryperiods.Duringsuchperiods,apriority

listisusedtodistributefreshwaterfordifferentuses.Themajor

usesofwater areforagriculture(forirrigation),forflushing (to

mitigateadverseimpactsforagricultureanddrinkingwaterfrom

the upward seepage of salt water and salt intrusion in the

waterwaysnearRotterdam),andforwatermanagementitself(to

maintainwaterlevelsinthelakesandcanals).Drinkingwaterand

industryarealsoimportantuses,althoughthequantityusedfor

theseisnegligiblecomparedtotheotheruses.

4.1.2. Thewatersystemanditsfunctionsinthefuture

Futuresocio-economicdevelopments,climatechange,andsea

levelrise,mayrequirechangestothewatermanagementsystem.

Recently, four water-related scenarios were developed for the

Netherlands(Bruggemanetal.,2011;TeLindeetal.,submitted).

These‘Deltascenarios’covertworepresentationsoffutureclimate

(based on Van den Hurk et al., 2007) and two sets of

socio-economicdevelopmentsintheNetherlands.Theclimatescenarios

cover a range from moderate increases in temperature and

precipitation(18C,3.6%precipitationinthewinter,and2.8% in

thesummer;usedinthescenario‘Crowd’)toalargetemperature

increase (28C in 2100; used in the scenario ‘Warm’), a large

precipitationincreaseinwinter(14.2%),andalargeprecipitation

decreasein thesummer(19%). Thesea level canincrease (35–

85cmin2100).Thesocio-economicscenariosdescribea

popula-tionchangefromthecurrent16millionto12millionor24million

in 2100,together withmajor changes in agricultural land use.

Thesescenarioswouldresultinanincreaseinwaterdemandsfrom

theregionalareastothenationalwatersystemduetolessrainand

lower river discharges, more salt intrusion, and/or agricultural

changes;andanincreaseinfloodriskduetosealevelrise,higher

riverdischarges,andpopulationandeconomicgrowth.

4.2. Step3:determineactions

Forillustrativepurposes,wefocusontheIJsselmeerarea,and

considerinouranalysisonlythemainalternativeactions,whereas

inrealitytheentireRhineDeltaandallkindsofcombinationsof

actionsarepossible.Asaresultofourproblemanalysis,itisclear

thattheIJsselmeerareawillbecomeevenmoreimportant asa

storagebasinforprovidingfreshwaterintimesofdrought.Either

thewaterstoragecapacityneedstobeincreased,orthe(growthin)

waterdemandneedstobereduced.Toincreasethewaterstorage,

thewaterleveloflakeIJsselmeercanbeeitherincreasedinthe

spring,and then used during dry periods, or decreased in dry

periods. Water demands can be reduced by increasing the

efficiencyof water usein theregional system, by changing to

saltand/ordroughttolerantcrops,and/orbydecreasing

agricul-tureormovingagriculturetoareaswithappropriate

environmen-tal conditions. Some of these actions can be taken without

changingthecurrentinfrastructure;thesecanbeconsidered as

improvements of the current system. For other actions, the

infrastructurewouldhavetobechangedconsiderably.Toensure

safetyfromfloodingincaseofsealevelriseandincreasedriver

dischargesinthewinter,floodmanagementactionswouldneedto

betakenaswell.SafetyfortheareasadjacenttotheIJsselmeercan

beachievedbyeitherraisingthewaterlevelin correspondence

withthesealevel,sotheexcesswatercanbedrainedundergravity

intotheWaddensea(ofcourse,dikesneedtoberaisedaccordingly

aswell),orbybuildinglargepumpsfordischargingwaterintothe

Waddensea.Ifthefirstactionischosen,theextraamountofwater

canbeusedin timesofdrought.Ifthesecondaction ischosen,

waterinletsandshippingsluicesneedtobeadaptedforenabling

wateruseduringdrought.Table2providesanoverviewofthisset

ofactions.

4.3. Step4:assessefficacy,sell-bydateofactions,andreassess

vulnerabilitiesandopportunities

Table2presentsanassessmentoftheefficacyofeachindividual

actionanditssell-bydatebaseduponexpertknowledge,previous

studiesonpossibleactions,andpreliminarymodelingresultsfor

2050and2100indicatinghowmuchwater(incmIJsselmeerlake

level)isneededtosupplytheamountofwaterdemandedforan

average,dry, andextremelydry yearforthedifferentscenarios

(Klijnetal.,2011).Fordeterminingthesell-bydate,weassumea

linearchangeofclimateand socio-economicdevelopments.For

theactions focusing on reducing the water demand,no model

resultswereavailable.Togetherwithstakeholders(waterboards)

theimpact of theseactions wastranslated into theamount of

IJsselmeerwaterneeded.Table2showsthatthecurrentplanis

likelytobesufficientforachievingobjectivesforapproximately30

years.Afterthispoint,changesarelikelytobeneeded.

Improve-mentsthatcanbemadetothecurrentsystemshouldenablethe

sell-bydatetobeextendedbyapproximately10years.

Thefloodmanagementactionsandtheactionsforfreshwater

supplyinfluenceeach other.Ahigherwater levelforincreasing

storage capacity will, at the same time, allow the system to

dischargeunder gravity(depending on thesealevel). If

policy-makers were to decide to ensure safety against flooding by

increasingthepumpcapacityandkeepingthesametargetwater

level,freshwatersupplyactionswithanincreaseofthewaterlevel

wouldbescreenedout.Thereisalsoarelationbetweentheactions

intheIJsselmeerareaandotherregionsinthelowerRhineDelta.

Forexample,aspartoftheactionstoensuresafetyalongtheWaal

andNederrijn,moreRhinewatercouldbedistributedtotheIJssel.

Inthiscase,enoughcapacityshouldbeavailableintheIJsselmeer,

implying that the water level can be raised at earliest in the

Table2

Actionsandassessmentoftheirrelativeperformanceintermsofimpactsonsafety,freshwatercapacity,sideimpactsonnatureareasandshippingintheIJsselmeerandIJssel region,andsell-bydateofactionsbasedonpreliminaryexpertknowledgeandmodelingresults.a

Action Impact Sell-bydate(years) Costs

Safety Freshwater Nature Shipping Floodmanagementactions

Increasetargetwaterlevelandthedikescorrespondinglyforenabling dischargingundergravitytosea.

+++ ++ >2100 +++

Keepthesametargetwaterlevelbyincreasingpumpcapacitylargely. +++ 0 0 0 2100 ++

Freshwatersupplyactions

Increasewaterlevelto+1.1minspring,andadaptregionalwater systeminfrastructure.MorewatertotheIJsselRiverinspring.

+++b

++ >2100 ++

Increasewaterlevelto+0.6minspring,andadaptregionalwater systeminfrastructure.MorewatertotheIJsselRiverinspring.

++b

+ 2070–2090 +

Increasewaterlevelto+0.1m,usingcurrentinfrastructure +b

+ /+ 0 2050–2060 0

Decreasewaterlevelto 0.8mindryperiods,andadaptinfrastructure. 0 +++ + 2100 ++

Decreasewaterlevelto 0.6mindryperiods,andusecurrentinfrastructure. Acceptnavigationobstructionsduringextremedroughts

0 ++ + 2060–2070 +

AdaptwaterdistributionRhinebranches:morewaterto IJsselRiverduringdroughts

0 + 0 + 2040 0

Improvingcurrentplanwithflexiblewaterlevels 0 + 0 0 2030–2040 0

Reducewaterdemandtothenationalwaternetwork,byimprovingthe managementoftheregionalnetwork

0 + 0 0 2050–2070 +

Reducewaterdemandanddamagebychangingtosaltand/or droughttolerantcrops

0 +++ 0 0 >2100 ++

Reducewaterdemandbychangelandusetonatureand/or urbanareas

0 +++ ++ 0 >2100 +

a _{largenegativeimpact, negativeimpact,0noorminorimpact,+positiveimpact,++moderatepositiveimpact,+++largepositiveimpact.} b_{TheseimpactsareconsideredaspositiveasthisfacilitatesthepreferreddrainageofexcesswaterfromtheIJsselmeertotheWaddenseaundergravity.}

beginningofspring.Insomeyears,therewillnotbeenoughwater

todothis.Startingearlierwithraisingthewaterlevelwouldbe

possibleonlyifthedikeswereraisedsufficiently.Ifmorewateris

transported tothe IJssel, there will beless water for theriver

branchestothewesternpartofthecountry(WaalandNederrijn),

andthuslesswaterforholdingbackthesaltintrusionfromthesea,

makingthewater inletat Goudaless reliable.In that case, the

MidwestareamightbesuppliedbyIJsselmeerwater.If,however,

policymakersweretodecidetoclosetheRhineestuary,thiswould

notbenecessary.

Withtheimpactsoftheactionsinmind,thevulnerabilitiesand

opportunitiesneedtobereassessed.Forexample,iftheIJsselmeer

level is raised, achieving the EU Directives (Water Framework

Directive,HabitatDirective,BirdsDirective)maybeendangered,

due to the disappearance of shallow waters that provide an

importanthabitatforspecies.

4.4. Step5:developpathways

Fig.6showstheAdaptationPathwaymapforthe10actionsfor

fresh water supply from Table 2. For flood management, two

actionsareavailable. Theyare notpresentedin theAdaptation

Pathways map, but they influence the preferences for certain

pathways,asexplainedabove.

Toconstructthepathways,theactionsaregroupedintoactions

influencingwaterdemandandactionsinfluencingwatersupply.

Actionswithlongsell-bydatesareshownonthetoporbottomof

themap,whileactionswithshortsell-bydatesareshowncloseto

thecurrentplan.Thenextstepistoaddthesell-bydatesandallthe

possibletransferstootheractionsthatwouldextendthesell-by

date.Sometimesactionsaffecteachother.Ifthesell-bydateforan

actionwillincreaseconsiderably,thisisshownbyanadditional

line in the same color. Next, illogical actions are eliminated

(backgroundcolorincontrasttobrightcoloredlogicalactions).For

example,implementingoneofthelargeactionsfirstisillogical,as

this may not be necessary to achieve success, and it can be

implementedlateraswell.Itisalsolesslogical,oncepolicymakers

havechosentosignificantlyadjustthewaterlevel,toswitchto

changingthecroptypeorlanduse.Thesell-bydateofanaction

dependsonthescenarioandtheobjectives.Thisisshownwiththe

twox-axes,oneforeachscenario.

4.5. Step6:selectpreferredpathways

FromtheAdaptationPathwaysmap,preferredpathwayscanbe

selected. Different decisionmakers and stakeholders can have

different preferred pathways, depending on their values and

beliefs.Fig.7presentsanexampleofthepreferredpathwaysfor

Raiselevel+1.1minspring

Changetodrought/salt tolerantcrops RaiseIJsselLakelevel withincurrentinfra+0.1m Decreaselevelwithin currentinfra(-0.6m) Decreaselevelandadapt infrastructure(-0.8m)

MorewaterthroughIJssel Raiselevel+0.6m

Changelanduse Moreefficientwateruse Optimisingcurrentpolicy

Water

de

mand

actions

Water

supply

actions

Transferstationtonewaction AdaptationTippingPointofanaction(Terminal) AdaptationPathways

archetypes of three perspectives: Hierarchist, Egalitarian, and

Individualist(seee.g.Hoekstra,1998;Middelkoopetal.,2004on

these perspectives related to water). For example, Hierarchist

believes in controlling water and nature, assigning major

responsibilitiestothegovernment.Thismeansa preferencefor

actions related to managing water levels and water use. The

Egalitarianfocuses on theenvironmentand equity, resultingin

strategiesfordecreasingwaterdemandsbyadaptingfunctionsto

theirenvironment(othercropsortheirrelocation).The

Individu-alistadherestoaliberalmarketandahightrustintechnologyand

innovation.Thismeansapreferenceforfacilitatingtechnological

developments for more efficient with water use and drought

tolerantcroptypes.Portionsofthepreferredpathwaysaresimilar.

Thepointatwhichthepathsstarttodivergecanbeconsideredasa

decisionpoint.Inourcase,therearethreedecisionpoints:(1)after

‘currentplan’,(2)after‘raisetheIJsselmeerlevelwithincurrent

infrastructure’, and (3) after ‘more efficient water use’. The

preferredpathwayscouldbeastartofadiscussiononanadaptive

plan.Inaddition,combinationsofthesepathwayscouldbedrawn

aspathsthathavesupportfrommorethanoneperspective.For

example,starting with‘moreefficientwateruseintheregional

areas’couldbefollowedbyasmallraisingoftheIJsselmeerwater

level(+0.1m),and,ifneeded,thatwaterlevelcanberaisedmore,

orthewaterdemandcouldbereducedbychangingcroptypes.The

short-termactionisonethatallperspectivescouldagreeupon,and

canthusbeconsideredasociallyrobustaction(Offermansetal.,

2011).

4.6. Step7:determinecontingencyactions,signposts,andtriggers

Togetorstayonthetrackofapathway,contingencyactionscan

bespecified. Forexample, theGovernmentcouldstimulate the

growthofsaltand/ordroughttolerantcropswithsubsidies,orby

limiting water availability and holding farmers responsiblefor

finding‘enough’water.Keepingtheoptionopenforanincreaseof

theIJsselmeerlevelwillrequirespatialplanningrules(e.g.allow

adaptivebuildingonlyoutsidethedikerings).Ifstructuresneedto

bereplaced,theycanbebuiltsuchthattheyarealreadyableto

copewithfutureactions.Correctiveactionsneedtobetakento

achieve objectives for nature. Constructing shallow zones and

islandscanmitigatethenegativeimpactsofraisingthewaterlevel.

Thiscanbringopportunitiesfordredgingcompanies.

Wedistinguishthreedifferentgroupsofsignpostsandtriggers:

(1) trends and events in the natural environment (the water

system);(2)human-drivenimpactsonthewatersystem,suchas

theautonomousadaptationoffarmersorachangeinupstream

Raiselevel+1.1minspring

Changetodrought/salt tolerantcrops RaiseIJsselLakelevel withincurrentinfra+0.1m Decreaselevelwithin currentinfra(-0.6m) Decreaselevelandadapt infrastructure(-0.8m)

MorewaterthroughIJssel Raiselevel+0.6m

Changelanduse Moreefficientwateruse Optimisingcurrentpolicy

Water d emand a ctions Water supply actions

PrefferedpathHierarchistPerspective:largerolegovernment,controllingthesystem

PrefferedpathEgalitarianperspective:protectenvironment,equity

PrefferedpathIndividualistPerspective:marketdrivensociety,smallroleforgovernment

Currentpolicy

2050 2100

2100 2050

Transferstationtonewaction AdaptationTippingPointofanaction(Terminal) AdaptationPathways

ScenarioWarm ScenarioCrowd

wateruse;and(3)societalperspectivesaboutthefuture,suchas

expectations about climate change and population growth,

knowledgeabout(orbeliefin)theeffectivenessofcertainpolicies,

andsocietalvalues, suchas thewish toprotectnatureand the

amountofacceptedflood/droughtrisk.Theamountofagricultural

areaandthecropsusedcouldbeanappropriatetriggerforchanges

inwaterdemand,sincetheycanbewellmonitoredandchange

slowlyovertime.

4.7. Step8:specifyadynamicadaptiveplan

Based on the problem, objectives, and pathways from the

previous steps, a dynamic adaptive plan can be specified.

Consideringthescenarios, theamountofwaterstorage needed

inthefuturerequiresuptoa1.5mwaterlevelintheIJsselmeer.

Raisingthewaterlevelisthepreferredactionfromasafetypointof

view, because in that case water can be discharged to the

Waddenseaundergravity.However,intheshort-andmid-term

(<2080)thisactionisnotneeded.Tokeepthisoptionopen,spatial

planningrulescouldbeimplemented.Initialactionscanfocuson

improvingtheperformanceofthecurrentplanbyintroducinga

flexiblewater level (e.g.outside thegrowingseason, thewater

levelmaydrop) and makingmoreefficientuseof waterin the

regionalareas (e.g.have a separatearea for brackish and salty

groundwater,inordertodecreasetheamountofwaterneededfor

flushing).Tokeepotheroptionsopen,theGovernmentcouldinvest

inresearchanddevelopmentofdroughtand/orsalttolerantcrops.

Theplanforfutureactionsneedstobeready,incaseawindowof

opportunity arises for adapting the water system to potential

future conditions. Anexample of suchan opportunityis when

infrastructure(sluices, dams,etc.)requiresmaintenance. Atthe

sametime asmaintenance is beingcarriedout,newstructures

couldbeaddedthat wouldbeabletocopewithanincreaseor

decreaseofthewaterlevelintheIJsselmeer.HuqandReid(2004)

assign the label ‘mainstreaming’ to actions that incorporate

‘‘potential climate change impacts into ongoing strategies and

plans’’.Anotherwindowforopportunityarisesinthecaseofadry

year.Insuchayear,societalsupportforimplementingsuchactions

islikelytobehigher.

4.8. Steps9and10:implementationofdynamicadaptiveplanand

monitoring

The first actions of the plan are implemented, and the

Government continues monitoring sea level rise and climate

changes.Furthermore,theGovernmentmonitorschangesinwater

demands through land usechanges and determines additional

signposts together with water boards (water managers of the

regionalsystem)andrepresentativesoftheagriculturalsector.

5. Evaluationofthemethod

Inthispaper,wehavepresentedanapproachforsupporting

decisionmaking under uncertain global and regional changes,

calledDynamicAdaptivePolicyPathways.Thisapproachassistsin

designing dynamic adaptive plans, and is built upon the best

featuresoftwoexistingadaptationmethods.Fromtheconceptof

Adaptive Policymaking we used the ideas of (1) thinking

beforehand of waysa plan might fail and designing actions to

guardagainstsuchfailures,(2)preparingforactionsthatmightbe

triggered later,in order tokeepa plan on track tomeetingits

objectives,and(3)implementingamonitoringsystemtoidentify

whensuchactionsshouldbetriggered.FromAdaptationPathways,

weusedtheideaofanAdaptationPathwaysmap,whichvisualizes

sequences of possible actions through time, and includes

uncertainties concerning societal values through perspectives.

Themap is enrichedwithtriggersfromAdaptivePolicymaking,

whichindicatewheneachnewactionshouldcomeintoforce.

Weillustratedtheintegratedapproachbyapplyingittoacase

inspired bya realstrategy development project toprepare the

Dutchwatersystemforfutureclimatechangetakingintoaccount

socio-economicdevelopments.Byapplyingourapproachtoareal

worldcase,wehavelearnedaboutthestrengthsandweaknessesof

theapproach,whichweelaborateinthissection.

A strength of the method is that it stimulates planners to

includeadaptationovertime intheirplans–toexplicitly think

aboutactionsthatmayneedtobetakennowtokeepoptionsopen,

anddecisionsthatcanbepostponed.Thus,theinevitablechanges

becomepartofalarger,recognizedprocessandarenotforcedtobe

maderepeatedlyonanadhocbasis.Planners,throughmonitoring

and corrective actions, would try to keep the system headed

towardtheoriginalgoals.

The concept of Dynamic Adaptive Policy Pathways may be

difficulttounderstand.But,thetenclearlydefinedstepsdescribed

inSection3provideasetofcleartasksthat,iffollowed,resultina

dynamicadaptiveplan.Wehavediscussedthemethodwithwater

and spatial planning policy advisors and policymakers in the

Netherlandsatboththenationalandregional/locallevels.Onthe

onehand,theapproachiscomprehensiveandmorecomplexthan

a traditional scenario-strategy impact analysis for one or two

pointsinthefuture.Ontheotherhand,plannershaveexperienced

that plans change over time, and an adaptive strategy is an

attractiveideaforplannersfacingdeepuncertainty.Moreover,if

political conditionsareunsuitable,theapproach helpsto

deter-mineforhowlongadecisioncanbepostponed.Thus,despitethe

complexity,bothpolicyadvisorsandpolicymakershaveshownan

interest inthemethod(see e.g.EEA,forthcomingin2013). The

adaptationpathwayspresentedinthe‘metromap’andthetriggers

and signposts are considered particularly valuable, as these

components of the method are the main new characteristics

comparedtoclassicalpolicyplanningapproaches.Foradiscussion

withhighleveldecisionmakersasimplifiedpathwaysmap,based

onpreferredpathways,couldbeusedincombinationwithamore

comprehensive map as background information. The case

pre-sented here has served as an inspiration for the Dutch Delta

Programme, and is included in their implementation guide for

‘adaptivedeltamanagement’(vanRhee,2012).Currently,

adapta-tion pathwaysare beingdeveloped for freshwater supply and

flood riskmanagement. Newmodel resultsshowthatwiththe

pathwayspresentedhere,anacceptablewatershortagemayoccur

oncein100years,andthatforatargetofoncein10yearsthe

sell-bydatesarefurtheraway(e.g.currentplanmaybesufficientfor

achieving objectivesforapproximately 50years if thetarget is

sufficientwaterforoncein10years).

Themomentofanadaptationtippingpoint(thesell-bydate)

helpsinidentifyingpossiblepaths.However,mostactionscannot

beimplementedimmediatelyattheirsell-bydate.Forthose,we

needtoincludealeadtime.Thethinkingbehindtriggershelpsin

identifyingrequiredleadtimes.However,climatechangemaybe

difficult todetect, especially changes in extremes,due tolarge

naturalvariabilitycomparedtothemagnitudeofchange(seee.g.

Diermanse et al., 2010; Hallegatte, 2009; Pielke, 2012). For

example,watermanagerswouldliketoknowifclimatechange

is happening because of the potential increase of floods and

droughts.However,measuring(forexample)peakdischargesasa

signthatclimatechangeishappeningisverydifficult,becauseof

highnaturalvariabilityandtheshorttimeperiodofmeasurements

(Diermanseetal.,2010).Still,landuse,populationchanges,andsea

levelrisearegradualdevelopmentsthatareeasiertodetect.

Withrespecttodecisionmaking,AdaptationPathwaysprovide

insights intooptions,lock-ins,and pathdependencies.Thus, an

decisionmakingonshort-termactions,whilekeepingoptionsopen

andavoidinglock-ins.Allpathwayssatisfyaminimumperformance

levelregardingthemain targets.Still,some pathwaysaremore

attractivethanothersduetocostsornegative/positivesideeffects.

Thiscanbeusedtoselect asetofpreferredpathways.Potential

futuredecisivemomentscanbeidentifiedbasedontheleadtimeof

actionsandthepointswherepreferredpathwaysstarttodiffer.

Todeterminethesuccessofactionsandpathways,quantitative

targetsareneeded.However,inreality,policymakerssometimes

choosetokeepthesetargetsvague,makingitdifficulttodetermine

the efficacy of an action and pathway. Exploring different

quantificationsofthetargetscanshowtheeffectsofthedifferent

targets,whichmaysupportadiscussionaboutappropriatetargets.

Aworthwhileelaborationontheapproachpresentedherewould

betheevaluationofpathwayswith,e.g.acost–benefitanalysisora

multi-criteriaanalysis.

The visualization of the pathways is seen as attractive by

policymakers.Thiswayofvisualizingworksbestiftheobjectives

canbesummarizedinasinglemainobjective,suchas‘freshwater

supplyfordifferentsectors’or‘safetyagainstflooding’.Inourcase,

weconsidered two main objectivesthat influenced eachother.

Because theflood management actions didnot vary a lot, the

relationbetweenthetwosetsofactionscouldbeeasilydescribed.

IntheDutchDeltaProgrammethesituationismorecomplexdue

toplanningfordifferentareasthathavedifferentpathwaysthat

influenceeachother.

The use of perspectives is an element that has previously

received little attention in the planning literature. We used

differentperspectives(orvisions)ofthedifferentstakeholdersto

identifyalternativepreferredpathwaysandsociallyrobustactions

(Offermansetal.,2008,2011).Differentstakeholdersmaysupport

differentplans,buttheycanalsohavedifferentreasonstosupport

thesameplan.Forexample,allocating‘roomforariver’maybe

preferredbysomebecauseitenhancesnatureandlowerswater

levelsinthecaseofpeakdischarges,whileothersmaypreferthis

action solely becauseit lowers theflood risk. Development of

pathways using stakeholder participation (decisionmakers and

stakeholders)hasbeenexploredinagamesetting(Valkeringetal.,

2012).Inthisway,uncertaintiesarisingfromdecisionmaking,and

preferencesamongplansarisingfromdifferentperspectives,can

befurtherexplored.

Theanalyticalbasisoftheapproach(e.g.fordeterminingsell-by

datesanddevelopingpathways)canbesupportedwith

computa-tionalscenario-basedapproaches.Makingthenecessaryrunsina

reasonableamountoftimerequiresapolicymodelthatisfastand

simple, but accurate enough tosimulate the relevanttransient

scenariosand assessthe relativeeffects froma wide varietyof

actions for the full set of performance indicators over time.

Currently, there is no such model of the lower Rhine Delta.

Therefore,weassessed theeffectivenessandsell-bydatesofthe

possibleactions using expert judgment andmodel results from

previous studies. We were able to assess the relative impacts

qualitatively. McDaniels et al. (2012) used expert judgment to

explorerobustalternatives.But,forabetterdeterminationofthe

sell-bydates,acomputationalexplorationiscrucial.Thereisaneed

forfastsimplemodelsthataresuitableforexploringactionsover

time in order to develop adaptation pathways. More complex

modelscanthenbeusedtoobtainmoredetailedinformationabout

theperformanceofthemostpromisingactionsresultingfromthe

initialexploration.

Furtherworkisalsoneededoncomputationaltechniquesthat

can help in identifying opportunities and vulnerabilities and

developingpromisingpathways.Inarealcase,thecombinationof

actionsandconsequentlythenumberpathwayscanbehuge.To

supportthe identification of themost promising sequences of

actions, we are working on an improved computer-assisted

approachfordesigninganadaptivepolicytoevaluatecandidate

pathwaysover anensembleofpossiblefuturesandassesstheir

robustness(KwakkelandHaasnoot,2012).Lempertetal.(2006),

LempertandGroves,2010presentacomputerassistedapproachto

developrobuststrategies acrossa varietyofdeepuncertainties,

grounded in Exploratory Modeling and Analysis (Agusdinata,

2008; Bankes, 1993; Bankeset al., 2013). We aredeveloping a

‘workbench’ to support such computational scenario-based

techniques.Earlyexperienceswiththeworkbenchindicatethat

usingafastandsimplemodel,exploringuncertaintiesinaddition

toclimatechange,andaccountingforthejointimpactofallthe

uncertainties, in support of the development of adaptation

pathwaysisusefulandfeasible(KwakkelandHaasnoot,2012).

6. Concludingremarks

In light of the deepuncertainties decisionmakers are facing

nowadays,anewplanningapproachisneededthatresultsinplans

thatperformsatisfactorilyunderawidevarietyoffuturesandcan

beadaptedovertimeto(unforeseen)futureconditions.Various

techniques are available (e.g. Robust Decision Making, Real

Options Analysis, decision trees, roadmaps, and several policy

planning approaches) that have been or are being appliedfor

supportingplanningunderdeepuncertainty(e.g.intheThames

EstuaryintheUK,theRhine-MeusedeltaintheNetherlands,and

NewYorkCityandthePortofLosAngelesintheUSA).Wehave

used two complementaryapproaches for planning under deep

uncertainty—AdaptivePolicymakingandAdaptationPathways—

todevelopanintegratedapproachbasedonthestrongfeaturesof

each of them. This approach, called Dynamic Adaptive Policy

Pathways, resultsin anadaptive plan that is abletodeal with

changing(unforeseen)conditions.

Key principles of the Dynamic Adaptive Policy Pathways

approachare:theuseoftransientscenariosrepresentingavariety

of relevant uncertainties and their development over time;

anticipatingand correctiveactionstohandlevulnerabilitiesand

opportunities;severalAdaptationPathwaysdescribingsequences

ofpromisingactions;andamonitoringsystemwithrelatedactions

tokeeptheplanonthetrackofapreferredpathway.Theapproach

supportstheexplorationofawidevarietyofrelevantuncertainties

in a dynamic way, connectsshort-term targets and long-term

goals, and identifies short-term actions while keeping options

open for the future. There is evidence that such policies are

efficacious(Kwakkeletal.,2012)andcost-beneficial(Yzeretal.,

submitted).Intheend,allthishastofitintoapoliticalprocess,

whichhasalwaysbeenarealsourceof‘deepuncertainty’.Political

circumstances can give a window of opportunity (or not) to

implement thedesigned adaptive plan. Also, theadaptive plan

could be used to create the right political circumstances, for

examplebyshowingpotentiallock-ins,potentialadverseimpacts,

andforhowlongadecisioncanbepostponed.ThePerspectives

method couldbe used to frame the plan for different societal

perspectives(asillustratedbyOffermansetal.,2008).

Inthispaper,wehaveillustratedandtestedtheapproachusing

avirtualworldinspiredbyarealworlddecisionproblemcurrently

facedbytheDutchNationalGovernmentintheDeltaProgramme.

Wewereabletoapplythemethod,andthisresultwasreceived

withgreatinterestbypolicymakersoftheDutchDeltaProgramme.

Theresultssuggestthatitisworthwhiletofurtheruseandtestthe

approachforarealquantitativecasestudy,otherpolicydomains,

andothercountries.

Acknowledgments

ThisresearchhasbeenfundedbytheDeltaresresearchproject