WATER RESOURCES
RESEARCH, VOL. 21, NO. 2, PAGES 95-102, FEBRUARY
Interactive Water ResourcesModeling and Model Use'
DANIEL
1985
An Overview
P. LOUCKS
Schoolof Civil and Environmental
Engineering,
CornellUniversity,Ithaca,New York
JANUSZ KINDLER
Instituteof Environmental
Engineering,
WarsawTechnicalUniversity,Poland
KURT
FEDRA
InternationalInstitutefor AppliedSystemsAnalysis,Laxenburg,Austria
This servesas an introductionfor the followingsequence
of five paperson interactivewater resources
and environmental
management,
policymodeling,and modeluse.We reviewsomeimportantshortcomingsof manymanagement
and policymodelsand arguefor improvedhuman-computer-model
interaction and communication. This interaction can lead to more effective model use which in turn
shouldfacilitatethe exploration,analysis,and synthesis
of alternativedesigns,
plans,and policiesby
thosedirectlyinvolvedin the planning,management,
or policymakingprocess.
Potentialadvantages
of
interactive
modeling
andmodeluse,aswellassomeproblems
andresearch
needs,
arediscussed.
INTRODUCTION
able, when they need it, some easy-to-usetools to help them
identify, explore, analyze, and synthesizeeffective resource
managementplans and policies.
The developmentand applicationof managementor policyoriented models for helping water and environmental resource
managershave beentaking placefor severaldecadesthrough-
We recently conductedan informal survey. We asked many
of our acquaintancesin government agenciesconcernedwith
water resourcesand environmental protection what their biggest problems seemedto be. All of their responsescould be
summarized as keeping ahead of the increasing number of
water- and environmental-related"crises."Efforts spent in reducing each current crisis seemsonly to result in the emergenceof seeminglymore difficult ones.
In spite of all that has been done to reducethe ever reoccurring cycle of floods and droughts and their accompanying
damages,these events still occur, and their damages are increasing.On top of these and other reocurring problems it
seems,suddenly, we have a host of new problems. Increasingly, various toxic wastesdischargedon or under the ground
are contaminating groundwater supplies and occasionally
oozing up to the surface to create health hazards for those
living on it. Our urban water distribution systems,many that
have functionedwell beyondtheir designlife, have collectively,
it seems,decided to call for help. The control of nonpoint
source pollution is becoming a major challenge to surface
water quality managers.Increasingly,acid rain, in a variety of
wet and dry forms, is adversely altering aquatic and forest
ecosystemson regional scales.The potential for more Three
Mile Island nuclear accidentsis becomingincreasinglyapparent, and so on. Any reading of major daily newspaperscould
comprehensivestudy of the use of models for water resource
management,planning, and policy in the United States was
recently completed by the Office of Technology Assessment
(OTA) of the U.S. Congress [Friedman et al., 1982]. This
report emphasizesthe potential of currently existing models
for improving the accuracy and effectivenessof informaton
available to managers,decisionmakers, and scientists.It also
documentsthe need for modelsin meeting the requirementsof
a wide variety of Federal and State laws in the United States
concerningresource management and environmental protection. However, the primary focus of the report is on the constraints to effective model use. Among these constraints are
will.
However, there are other constraints that we as modelers and
out much of the world.
The extent to which this effort has
been as useful as it could have been. is in some doubt. A
the lack of information about available models, lack of train-
ing in model use and interpretation,lack of communication
between model users and developers,and lack of required
support services.
'Some constraintsto model use can only be reduced by actions taken by potential user agenciesor groups.Much of the
lead one to ,concludethat we are not yet very effectiveman- OTA report is directedtoward this issueand suggestspossible
agersof our water and other natural environmentalresources. coursesof action that resourcemanagement agenciescould
We ought to be able to do better, and we can if we have the take to increasethe benefitsthey can obtain from model use.
What is to be discussedin this and the following sequence analystscan help to reduce.It is this need and opportunity
of papers is not going to solve all of our current water and that motivatesthe discussionin this and the followingpapers.
Many of the constraints for effective model use that conenvironmental problems. However, we believe it can provide
front
us today are the same that Little [1970] identified well
some help toward that end. It seemsto us that managers,
planners,and policy makers could benefit from having avail- over a decade ago. Not every proposedmodel is appropriate
in all contexts.A very good model for one problem and instiCopyright 1985 by the American GeophysicalUnion.
tutional setting may not be as good, or oven applicable, in
another.
Paper number 4W0977.
0043-1397/85/004W-0977505.00
Unless
model
builders
are familiar
with
both
the
problem and the institutionalsettingin which the problem is
95
96
LOUCKSET AL..'INTERACTIVE
MODELINGOVERVIEW
to be addressed,it is unlikely any modelingwork will be very
effective.Good model parameterizationand validation is rare.
Models must be made credibleto the potential users.Regardlessof whether or not the usersfully understandthe model or
models, the outcome must be believable. It must conform to
the user'sperceptionof the world or at least that part of it
being modeled.Finally, the model must be complete.It must
include the capability of examining all of the issuesdeemed
important by the user,and the resultinginformationmust be
easyto understand.How to make our models more flexible or
adaptable to alternative paths of exploration and how to
designan interfacefor more effectivecommunicationbetween
the user and the model are continuing challengesto model
quencesof particular actions taken to help solve or reduce it
[Little, 1970]. It may also assistin the resolution of conflicts
by providing negotiators with a "face-saving" excuse for
changingtheir positions,i.e., for compromise[Sebenius,1981•
Straus and Clark, 1980; Lara and Sachs,1978].
Before discussingour proposed interactive approach to
modeling and model use it seemsappropriate to first briefly
discussin a general way how much of water resourcesand
environmental policy making seemsto take place. This discussionprovidesa context for comparingwhat we are proposing with somepast approachesto policy modeling that placed
little if any emphasison human-model-computerinteraction
and communication.
builders.
Our thesis,in short,is as follows.Modeling, if it attemptsto
describemore than somemicroaspects
of physicalsystems,
by
necessityinvolves a strong subjective and value-dominated
human element which defies a formal representationin any
generallyacceptableway. At the same time, any acceptable
formal model must be compatible with the set of mental
modelsand the cognitivestylesof their users.We argue that
widespreadmodel acceptanceand use can only stem from
direct user involvement at various phasesof the modeling
process,including definition of basicassumptionsand the control of output formats. Interactive methods which give the
user an appropriate role in controlling model calibration,
model use, and output display requires new ways of manmachineinteraction. The role of models thus changesfrom a
product-orientedfunctionwheremodelsare productsof a usually external modeling processand produce solutions which
the end-user does or does not accept, to a process-oriented
function.In a process-orientedfunction, modeling and model
use are the important elementsof a learning processwhich in
turn facilitatesthe policy and decision-makingprocess.
This paper, and those that follow, presentsome ideas and
general approachesfor model builders to consider as they
work toward improving the effectivenessof their models to
managers,planners,and policy makers. Given the increasing
number and complexity of today's water and environmental
managementproblems and those that are likely to emergein
the future, it follows that the potential utility of models designedto help studytheseproblemsis alsoincreasing.Clearly,
the demand
for information
that can and should be derived
from models exists;it is only a question of how well we as
model builders can meet it [Brown, 1982]. We will need, and
we will develop,better models,of course,as our knowledgeof
the systemswe model increases.However, equally important,
we must
devote
some attention
to the interface
between
the
model user and the models being used. More effective communication
is essential for increased effectiveness in model use.
Many gaps between model availability and model use can
be reducedby the developmentof a more user-friendlyinteraction between the model and its user. Facilitating humanmodel-computerinteraction will not necessarilyincreasethe
likelihoodthat any one or more modelsolutionswill be implemented. Rather, it should, we believe, increase the likelihood
of the modelsproducinginformation usefulin any debateover
what to do, who will be affected,how they will be affected,and
how much they will care.
Models, if and when usedin the planning or policy making
process,are used because they provide and organize information that together help in the understandingof the problemsbeingaddressed.Often this "additional"informationmay
simply better define the problem and the possible conse-
SOME CHARACTERISTICS OF PLANNING
AND POLICY MAKING
Water resourcesand environmental planning and policy
making are commonly characterized by their breadth of
impact, uncertainty, scarcity of causal evidenceon which to
basea plan or policy, and by their multiobjectiveand multiinstitutional involvement. Many individuals, interest groups,
and organizationsare affectedand concernedabout water and
the environment. We are not very good at predicting future
economicconditions,costs,and benefits,or estimatingfuture
social or political impactsof current decisions,plans, or policies. We must use considerablesubjectivejudgment to supplement what meagerevidencemight exist to help us predict
even the physical, biological, or chemical impacts of alternative decisions.Finally, water and environmental problems
tend to be heavily value-laden,and objectivesand institutions
are very much a part of the problem.
As a consequenceof all this apparent confusionand uncertainty, water resourcesand environmental problem-solving
tends to be an iterative, explorative,learning processthat redefines the problems as much as it seeks solutions to them.
Any review of the U.S. water pollution control policies over
the past several decades,for example, will provide ample evidence of this.
The very nature of this iterative, recursive,learning, and
decision-making process usually requires subjective evaluations and considerablecompromisesamong conflictinggoals.
The processis typically more political than analytical or scientific. The direct involvement of various interest groups is
common. Their goals and behavior are not easily modeledin
any acceptable way. What model builders may assume as
rational cannot be assumed.What is rational to one may not
be to another, and for reasonsthat may not ever be revealed.
Values and beliefs,and the resulting objectivesthat shape
policies, are the result of dynamic human and institutional
interactions.While one could suggest,therefore,that "planners
must design from the beginning for the complete range of
objectives"[Kahn, 1960], such a task may be difficult to accomplish. Objectives and values are usually conflicting and
changing.They are dependenton the choicesgiven, i.e., on the
status of the learning process.Values may very well be unknown to those who have them; i.e., one may not be able to
make them explicit when asked. Yet the same person will
usually be able to select preferred plans or policies from
among alternatives.
Values and priorities may also depend on the possibilities
and tradeoffs in a bargaining situation. Such situations are
typical in multiple objective decision making exercisesthat
characterizewater resourcesand environmental planning and
policy making. Goals may change and unknown constraints
may emergeduring suchprocesses.
LOUCKS ET AL.: INTERACTIVEMODELING OVERVIEW
To include these complex and uncertain behavioral characteristicsof planning and policy making within one or more
models becomes impossible,but to exclude them from consideration during the development of policy models invites a
high chance of model failure or rejection. So, what to do?
First, we suggest,is to include within models only those parts
of the problem under study which can be modeledwith credibility. That which cannot, i.e., most human or institutional
behavior,must be left outsidethe model. This then requires
that the modelsbe capableof directly and interactivelyinvolving humans,i.e., the managers,plannersor policy makers, or
their staffs,who must supply the crucial behavioral elements.
This capability requires, in turn, that the model or models
must fit into the managing,planning,and policy-makingprocesses.They must be appropriate for the institutions in which
suchprocessestake place and for the individuals within those
institutionsinvolved in managing,planning, or policy making.
Inadequate data of the kinds most needed is another
troublesome
characteristic
of most water
and environmental
planning or policy-making exercises.Each discipline involved
in such multidisciplinary exercisestends to collect data in its
own traditional way. Data compatibility can be a major constraint. Temporal and spatial resolution, and the resulting statistical quality as well as the appropriatenessof the data differ
widely when considering,for example, ecological, geographical, or economicand fiscaldata related to a given problem.
Some data in some casesmay be available in large quantities. Meteorological and hydrological records and data resulting for remote sensorsmay be available with considerable
temporal and spatial coverage.In suchsituations,data-related
problemsare often those of data reduction, aggregation,and
interpretation. Data on causalrelationships,however, are usually scarce. This is particularly so when the phenomena of
interestare the long-term and synergisticimpacts of a particular plan or policy pertaining to, for example, public health,
economic development, and ecological systems.Once again,
decisionsin the absenceof adequate evidence of causal relationshipswill be largely basedon subjectivejudgments.
97
that unspecified,perhapsunknownand unquantifiable,objectives and constraints also apply.
The difficultiesin identifyingappropriatepolicy objectivesis
increasinglybeing recognized. Many models found in the
literature assumethe existenceof one or more fixed objectives.
In the case of prespecifiedmultiple objectives,one can resort
to multiobjectiveor multicriteria modelingto generatesetsof
"efficient," "noninferior," or "Pareto-optimal" solutions
[Cohon,1978; Zeleny, 1981a,b; Goicoecheaet al., 1982]. However,solvinglarge multiobjectivemodelsmany times to generate even a small subsetof noninferior solutionshas proven to
be an expensiveand time-consumingeffort. Moreover, there is
increasingrealization that individuals may want to consider
an apparently inferior solution (with respectto those objectives being modeled) as well as noninferior ones. So-called
inferior solutions may not be inferior at all when considering
additional objectives.Rarely, if ever, can all relevant objectives
be explicitly consideredin multiobjectivemodels.Hence there
are legitimate needs to focus on more than only the set of
noninferior or efficient solutions generatedby multiobjective
models.
Perhapsone of the biggestreasonsfor model solution rejection, even as a basis for discussionin the managing, planning,
or policy-makingprocess,has been the lack of model understandingor confidence,and the lack of adequate communication betweenthe analystsand their clients.Throughout the
managing,planning, or policy-making process,analystsrarely
can addressthe right questionsthat managers,planners,and
policy makers want answeredunlessthey are aware of these
questions.Also, individualsoften do not know what questions
they want answersfor before some exploration and comprehension of the impacts of some of their ideas for plans and
policies.The outcome of this exploration will lead to new
questionsand the need for additional exploration.
Another tendency in the past, and to some extent even
today, has been to concentrate on the development of fairly
complex and comprehensive models. This is natural. Most
water and environmental problemsand their impacts are comHow can these and other troublesome
features of water
plex and involve a wide variety of individuals,interestgroups,
resourcesand environmental management and policy making and institutions. The trend toward bigger and more comprebe reduced by those who develop models? We believe there hensivemodels which simultaneouslytry to describeas many
relevant aspectsof the problem as can be put into the comneedsto be a change in the manner we approach such problems.Before discussingtheseproposedchangesa brief review puter has coincided with the growth in computer speed and
of our view of how most modelshave been developedin the capacity, and in the development of more efficient algorithms
for solving a variety of mathematical models. Large complex
past may serveas a basisfor comparison.
models,however, are not only more difficult to develop and
maintain; they tend to require more input data, and this can
TRADITIONAL APPROACHES TO POLICY MODELING
be expensiveto obtain. They also tend to produce large quanMany modelershave, perhapsuntil relatively recently,been tities of output, much of which can be a challenge to comfollowinga modelingapproachthat has focusedmainly on the prehend.
Figure 1 is an admittedly very simplified illustration of
analysisof alternativesand lessso on the generation,exploration, and synthesisof alternatives.Many of thesemodelshave many past and even some current attempts of comprehensive
been designedfor use outside of the planning, management, model building and application. The heart of the modeling
and policy-making process.They were designedfor use by processrevolvesaround the analysts(model builder and user)
policy model builders and analysts to provide quantitative and the comprehensivesimulation and/or optimization model.
solutions to specific problems. This information has oc- The potential client, the potential user of model information,
casionallyended up in one or more reports. Very often these is only weakly involved, usually through reports the analysts
quantitative solutionshave been based on the assumptionof may prepare basedon information derived from the model.
Typical of many resourcemanagementand policy-modeling
rational economicbehavior for lack of any better assumption.
Multiobjective modelshave been usedto attempt to eliminate applications,the analystshave someinitial contact with those
from further consideration solutions that were not efficient i.e., needing information to be obtained from models, e.g., manthat were inferior with respectto the objectivesconsideredin
agers,planners,and policy makers or their staffs.They also
the models.The trouble is that rational behavior of any kind
have some contact with them at the very end when some
is not likely to be the same among different individuals and
reasonableresultsfrom the models are available. During most
98
LOUCKSET AL.' INTERACTIVEMODELING OVERVIEW
Data
Base
J
JAnaly
sts
Comprehensive Model
I
I
I
Managers,
Policy Makers and Staffs
Fig. 1. Typical relationshipsamongmanagersand policy makers,
analysts,their model, and their data, indicating weak or nonexisting
links betweenpotential usersof model information and the models
and thosewho develop them.
of the intervening time, however, the analysts and potential
userstend to go their separateways. Once some model results
are available, the original resourceuse conflicts,management
questions,policy objectives, and sometimeseven the managers,planners,or policy makers are often quite different than
when the modeling work began. All too often, too little attention has been devoted to building the modelsinto the dynamic
management, planning, or policy-making process [Walker,
1982; Lara and Sachs, 1978]. This has made it difficult for
anyone but analyststo use models to explore their own ideas,
obtain information about the impacts of new policy options,
and indeed,to synthesizeas well as to analyze alternatives.
The fact that many of our past models were complex and
relatively comprehensivedid not necessarilyhave to mean
that they were also lesscomprehendableor lesscredible to a
potential user. That this has tended to be the rule rather than
the exception indicates only that in many casesnot enough
attention has been placed on model managementand model
use. The challenge for the future is to learn how to model
complex systemsso that the models, or more likely a system
of more comprehendablemodels, and the interface between
the models and the usersbecomesmore intelligible, manageable, useful, and reliable [-Quadeand Miser, 1983]. Too little
attention has been devoted to building models into the dynamic management, planning, and policy-making processes,
i.e.,model and implentation and communication.
AN APPROACH TO MODEL
IMPLEMENTATION
AND
COMMUNICATION
Any approach to overcoming many of the difficultiesand
limitations of many of our past models just discussedmust
recognizethe needto involve representativesfrom each appropriate level of decision making in the modeling process.Involvement of planners and the staffs of policy makers, rather
than analysts trying to guess and model their objectives,
values,and behavior, is the only way we see how models are
going to become more effective as tools for exploring and
understandingkey issuesand for identifying possiblemanagement or policy options and their probable impacts.This involvement of nonmodelers in model use, if not also model
building, will only happen if the interface between the computer, the models,and the model usersmakes modelseasyto
use and their results easy to obtain and understand.This
interfacemust permit easyinteraction betweenthe model user
and the model or modelsbeing used.It must make data input
and editing easy to perform and make data output easy to
manageand comprehend.
Increasing availability of interactive computer hardware
and softwaremakes this approach much more feasibletoday
than it was even severalyears ago. The current revolution in
the development of mini- and microcomputers,interactive
software,and color graphicsis making it possiblefor each of
us, as individuals or as members of an organization, to have
accessto fast, cheap, interactive (flexible) computing power,
supplementedincreasinglyby graphical or pictorial display
capabilitiesif desired [Weger, 1981; Whitted, 1982]. We are
convinced this technologicaldevelopment cannot but have a
profoundeffecton how we all approachpolicy modelbuilding
and use in the future. Other articles [Fedra and Loucks, this
issue; Loucks et al., this issue] discussthis subject in more
detail.
For management,planning,and policy studies,analystsare
increasingly beginning to develop systems of interrelated
smaller models rather than single large and more complex
models [Walker, 1982]. Current trends in management and
policymodelingseemto be directedtoward the interactiveuse
of relatively smallerinterrelatedmodelsdesignedto be adaptive and responsiveto a wide variety of questionsthat policy
makers and those affectedby a policy may want to ask. Even
if each of the smaller modelscan only addressa few questions,
together they can be used to study the more complexstructured aspectsof a resourcemanagement problem. The resulting information together with policy makers' experience
and judgment about the lessstructuredaspectsof the problem
can assistin the processof making more informed decisions.
This has to be done within the time and budget available and
without the need to identify and argue over explicit management or policy objectives[Quade and Miser, 1983; McClure,
1981; House and McLeod, 1977; Hadden, 1982].
Collections of interactively linked models for impact ex-
ploration, synthesis,and evaluationare often termeddecisionsupportor decision-aidingsystemsor model managementsystems [Spragueand Carlson,1982]. Figure 2 illustrateshow a
model managementor decision-supportsystemapproachdiffers from the approachillustratedin Figure 1. The focusis
broader and includesnot only modelsand policy analysts,but
reflectsalso the interactivepolicy making process.Admittedly,
Figure 2 is oversimplified,
as is Figure 1, but their comparison
demonstratesthe changein modelingphilosophynow becoming evident.
Figure 2 emphasizesthe interactionof managers,planners,
and policy makers(or their staffs)with a variety of impactpredictionmodelslinked togetherto their appropriatedata
bases.An essentialfeature is the human interaction through a
"command"programinterface.This givesmanagers,planners,
and policymake'rsor theirstaffsthe abilityto easilyexplore,
define,and synthesizea variety of managementor policy options at any time, includingduring the time when the negotiations over the relative merits of these options are taking
place.
This new modeling philosophy recognizes that policy
making typically involvesa number of actors and interests,
conflictingperceptionsof nature, contradictoryrationalities,
and divergentadvocacies.
Policy making and planningare not
a static phenomena,but dynamic historical processes.
Complexity, goal ambiguity,contradictoryuncertainties,conflict,
risk, institutional inertia, and temporal change are not rare
characteristicsof policy making or planning environments;
they are their essentialfeatures.
High-levelmanagers,policymakers,and seniorbureaucrats
are usuallywhere they are becauseof stronglyfelt, or at least
effectivelyarticulated, beliefs about certain issues.Existing
policies,basedon thosebeliefs,are identifiedwith thoseindi-
LOUCKS ET AL..' INTERACTIVE MODELING OVERVIEW
99
Data
Base
]
Decision/Impa ct-
Policy
Analy
sts
Prediction
Models
Interactive
Command Program
Managers,
Policy Makers and Staffs
Fig. 2.
Interactiverelationshipsamong policy makers,policy analysts,their models,and their data, showingstrong
link betweenmodel users,model builders,and thosebenefitingfrom the useof models.
viduals.Analystsgivinginformationor providingtools yielding resultsthat questionthosebeliefsand existingpoliciesare
not likely to have an audience.The problem is to resist the
urge to ignore thesefeaturesof political life and instead to
adapt our thinking and modelingto this fact. Political considerationsmay permit, at best, incrementalor transitional
policiesthat not only preservereputationsbut also provide
feedback prior to more significantmanagementor' policy
The appropriate procedurethus is also a heuristicone: starting from expert knowledgeand problemperception,guidedby
strict parsimony and whatever theoreticalframework may be
applicable,models are built to satisfytheir users'needsfor
information and enhancedunderstandingon the planning or
policy levels.They should not be built, necessarily,to reliably
predict, becausehow could that be establishedother than in
hindsight?
shifts.We believethat only when modelersinclude theseconModel credibility is also a very subjectivenotion. A model
siderations in their models will their contributions
be conis or is not credibleonly in the eye of the beholderor modeler.
sideredrelevantby thoseindividualsand institutionsinvolved Planning or policy modelingis as much an attempt to depict
reality as it is an attempt to formalizeand guideour subjective
in the processof managing,planning,or policymaking.
Any shift in an approach to modeling complex resource perceptionand understandingof reality. The only remaining
management problems brings with it new research op- criterion (given a certain professionaland scientificstandard)
portunities.Certainly this is true for the developmentof sys- is user satisfaction.
temsof interlinkedimpact predictionmodelsimplementedon
Each of the remaining five papers in this seriesaddresses
readily accessiblecomputerwork stations[Fedra and Loucks, this issue further. The conclusion to be drawn from this admitthis issue] possibly coupled to graphical and pictorial input tedly limited sample of experiencesand opinions is that the
and display devices [Loucks et al., this issue]. Some of the answerto the above questionis yes.However, it remainsto be
seen how this can best be done.
more challengingof theseresearchneedsare asfollows.
1. Can proceduresbe developedfor deriving systemsof
2. How can a systemof interlinked models be developed
simplified, yet sufficiently realistic and reliable predictive and used so that it contributes not only to increasing the
models for use in the planning or policy making process? efficiencyof any proposed plan or policy or decision (i.e.,
What is the appropriatelink betweenthesesimplifiedmodels doing things right) but also to the effectiveness
of the policy
and the more comprehensive
modelsthat are designedto en- making process(i.e., helping managers,planners,and policy
hance our scientificunderstandingof various natural, techno- makersdecidewhat are the appropriatequestionsto ask and
logical, or social systems?How can systemsof relatively the right things to do)? Related to this is the need to evaluate
simple predictive models representinga complex systembe a wide variety of impacts associatedwith a large number of
designedto be solvedrelatively quickly on a small computer objectivesor measuresof systemperformance,especiallywhen
and remain credible ?
some of those types of impacts and objectivesare not known
One way of looking at this problem is to realize that all
or thought of until well along in the planning or policy
planning or policy models, and especiallythe more sophisti- making process.
At the heart of this question is the problem of complexity,
cated ones, are largely determined by their underlying assumptions: what to include and what to leave out, how to
of everythingaffectingeverythingelse.It involvesthe problem
describea given process,which parameter value to choose, of not knowing what one wants to do, or knowing what will
how to interpret and prepare empirical data for inputs, etc. be consideredimportant, until one knows what can be done,
Models are thus at least partly based on subjectivefounda- or what impacts will result and how they will affect others.In
our view, the appropriate approach is a modular one made
tions. In other words, models rely heavily on expert knowledge.It is their heuristicvalue rather than any degreeof preci- from numerous more general building blocks in a problemsion that makesthem helpful in planning and policy making. specificframework. Also, the preconditionfor sucha mode of
100
LOUCKS ET AL.: INTERACTIVE MODELING OVERVIEW
modular model operation is a sufficientlyflexible and easy to
usesoftwaredevelopmentenvironmentthat makesthe cost of
modulardevelopmentvery smallversusthe costof not exploring one more possibility.In differentways,each of the other
five papersin this seriesexploresthis issuefurther.
3. How can one build into a model management or
decision-supportsystema means of updating the input data
and even the structure of the models, when appropriate, so
that the systemis adaptive to changingissues,problemsand
institutional interestsor objectives?How can one make the
systemadaptive to various levels of user sophisticationwithout the need of a full-time computer programmer? Can systems of predictive models be developed and used to provide
guidancefor, and yet not constrain,the identificationof new
policy alternatives? This is especiallyapplicablein an environment where not the analystsbut the managers,planners,and
policy makers (perhapsthrough their staffs)are asking analysts for information that may not have been considered
duringthe model building stage.
Again, the answeris flexible,easy to modify or reconfigure
models,and a development-orientedsoftwareenvironment.A
rich set of compatible and cooperatingsoftwaretools, which
can easily be combined and recombinedin an experimental
fashion,invites the continuousadaptive developmentof such
model systems.Independent but combinable tools, each of
them relatively small and easy to grasp, master, and, if necessary,modify or add to, givethe userthe necessary
flexibilityat
low cost. If it is easy to develop a special purpose ad hoc
solution to a given problem (like a special purpose editor
programto update or modify input values),the needfor necessarilycomplicatedgeneralpurposesystemsis reduced.
4. How can one most efficiently link various predictive
modelstogetherwhen eachof thesemodelsmay have differing
and varying spatial and temporal dimensions?The dimensionsmay dependon the spatial and temporal responseof the
systembeing modeled,on the policy or decisiontaken at each
spatial location and in each time period, and on external or
exogenousevents.Hence the spatial and temporal resolution
appropriateto a particular model or module may vary during
any simulation.
If individual elementsof a model systemhave different resolutions in space and time, then they can always be linked
togetherwith a setof specialpurposeinterfaceprograms,that,
in a purely formal or more heuristic way, transform, aggregate, and disaggregatethe flow of informationbetweenthese
respectivemodels. Each such interface can be very small,
simple,and inexpensiveto produce.However,therewill be the
needto assesshow suchsimplificationsdistort the information
derivedfrom each of the disaggregatedmodels.
5. How can one most effectively combine, contrast, and
presentinformationhavingboth spatial and temporaldimensionsin a manner that will improve the comprehensionof that
information by analysts,managers,planners, policy makers,
and the interestedpublic?
Interpreting, comparing, and evaluating complex information is a task for the human brain. However, the work of
the brain is enhanced if the information is presented in an
understandablemanner. This, however,may be quite different
for different persons.It is therefore necessaryto provide a
with the mental models and cognitivestyle of its userswill
they be understoodand thus contributeto enhancedunderstanding.
The problem clearly calls for the developmentof systems
designedfor usewith, and ultimatelyby, the usersin an interactivefashion.If the user can be involved in the designof the
interactivesystem,it can be designedto respondto his or her
needsin a more substantiveand comprehensibleway, using
his or her language,symbols,or modelsof thought.In this
respectthe papersby Kunreutherand Miller [this issue] and
Fedra [this issue] should be contrastedto those of Cosgriffet
al. [this issue]and Louckset al. [this issue].
These and other researchissuesare important. Many relate
to the acceptability of any quantitative modeling or information display systemwithin an essentiallypolitical setting.
How these problems are or are not overcome in the actual
processof buildingand usinginteractivemodelswill undoubtedly differ. The interface that links a system of models to
model users and the level or extent of interaction
needed will
be largely dictatedby the institutionalstructurein which the
modelsmust fit and adapt as well as by the particular managementproblemsrequiring policy.
Hence there are no single answersto these questions.Yet
these are some of the questionsaddressedin the following
seriesof papers on the developmentand implementationof
interactivemodeling.Some of our views are the same,some
quite different.Someof our work has been successfully
implemented; some has not. We leave it to others to critique our
ideasand the approacheswe have followed,and to add to the
experiencesreportedin thesefollowingpapers.
The next paper by Kunreutherand Miller [1984] reportson
one of the first attemptsto developand apply interactivemodeling to a pervasivewater problem: flood managementand
hazard assessment.
They began and completedtheir work in
the mid- to late 1970's, and long before anyone seriously
thought that thosewho would most likely be benefitingfrom
the use of such models would have their own personal computers,and in many caseswith graphical display capabilities
to make the output even more understandable.One can
wonder if the difficultiesKunreuther and his colleagueshad in
the late 1970'sin identifyingclientswho would be interestedin
usingthesevery innovativetools (a difficult task especiallyon
dry sunny days) will be diminishedin the near future. Advancesin computertechnologywill soon make it possiblefor
all of us to have an even more powerful modeling capability
linked to appropriate data basesfor assessingflood hazards
and options for their reduction.
Computer technologyis going to have such a profound
impact on how we developand use models that it seemsto
warrant a more detailedexamination.In the next paper, Fedra
and Loucks[this issue]speculateon what computertechology
will be like in the future and how we as model builders should
use this technologyfor the improvement of water resources
and environmentalplanning,management,and policymaking.
Any speculationon what this rapidly advancingtechnology
will be like and how it will impacton what we do is risky,but
there can be no doubt that it will provide even greater op-
portunitiesand potentialfor making what we do as modelers
more relevant and usefulto plannersand policy makers.We
large array of formats and styles,from the more traditional in researchmust make this happen.
The fourth paper is illustrative of what is taking place in
tablesof numbersto linguisticstatementsand color graphics,
so that individual userscan selectinteractivelywhat each likes Boston and a few other metropolitanareas; namely, the imbest.Only if the resultsof formal modelsare made compatible plementationof computermodelsfor the predictionand real-
LOUCKS ET AL.: INTERACTIVE MODELING OVERVIEW
101
accessto fast, inexpensive,interactivecomputingpower, supplementedincreasinglyby color displaycapabilitiesif desired.
Melbourne, Australia.
Thesedevelopmentswill have a profound effecton how we as
Cosgriffet al. [this issue]in the Melbourne and Metropoli- modelersapproachmanagementand policy model buildingin
tan Board of Works (MMBW) have clearly demonstratedthe the future, and how we can help managers,planners, and
potential advantages and economic benefits that can result policy makers better understand technical issues, identify
from the implementation of such a system of interactive variousoptionsthat are available for reducingconflict,and to
models and graphicsdisplay terminals for both analysisand assess
the impactsof thosealternativepolicies.
systemoperation. In our opinion, it is well worth the trip to
In this first of a seriesof articleson interactivemodelingfor
Melbournefor all havingdecision-making
responsibilities
for water resources and environmental planning and policy
managing and planning water and seweragesystemsto ob- making we argue why we think suchan approachmight lead
serveMMBW's integratedcomputer-modelingoperation and to more effectivemodeling and model use. We also focus on
telemetry capability. What is especiallyimpressiveis how a someof the problemsand researchtopicsrelated to the develfew modelersand computer programmerswere not only able opment and use or implementation of interactive models.
to successfullydevelop such an interactive computer-based Rather than providing a literature review of the subject[Polsystem,but also how they succeededin implementingit in a lack, 1976], we have attempted to provide some motivation
very complex setting,both physicallyand institutionally.It is and backgroundfor somemore detailed discussionsof several
judged by all within MMBW to be a success.It has already aspects,and several applications, of interactive systemsof
saved MMBW
six times its initial cost.
modelsdesignedto assistand be usedby water resourceand
Returning to more research-orientedwork, Loucks et al. environmentalmanagers,planners,and policy makers. These
[this issue]review someof the resultsthat they and others at more detailed discussionsare contained in the following five
Cornell University have obtained on integrating a variety of articles. By no means does the information contained in this
water resourceand environmentalplanningand policy models and the following papers representa review of the current
to interactive computer graphics.Of particular interest is the state of the art. We are only hoping to add a little to it and to
useof interactivecomputergraphicsfor assistingin the input, arguefor more researchand discussion
on this topic.
editing,management,and display of spatial and time-varying
REFERENCES
data. Someof the detailsassociatedwith the useof reasonably
high-resolutioncolor graphicsand video-digitizingtechniques Brown, G. E., Jr., Can systemsanalysisand operations researchhelp
Congress?, Interfaces,•2(6), 119-125, 1982.
are discussedand illustrated. This developing technologyis Cohon, J. L., Multiobjective Programmingand Planning, Academic,
becomingincreasinglyavailable and lessexpensive,seemingly
New York, 1978.
with each new issue of any of the numerousjournals that Cosgriff,G. O., P. E. Forte, and M. A. Kennedy,Interactive computer
modeling, monitoring, and control of Melbourne's water supply
describeand advertisethis technology.It will be interestingto
system,Water Resour.Res.,this issue.
observehow quickly our professioncan learn how best to use
Fedra, K., A modular interactive simulation system for eutrophicatheseand other similar researchresultsfor the improvementof
tion and regional development,Water Resour.Res.,this issue.
planningand policy making.
Fedra, K., and D. P. Loucks, Interactive computer technology for
planningand policy modeling,Water Resour.Res.,this issue.
The final paper by Fedra [this issue] describessome researchon the developmentof an interactive policy-oriented Friedman, R. M., C. Ansell, S. Diamond, and N. Ikeda, Use of models
for water resourcesmanagement, planning and policy, Office of
simulation and information managementand display system
Technology Assessment,U.S. Congress,Washington, D.C., August
for a lake region near Vienna, Austria. Fedra's modeling phil1982.
osophy is somewhat similar to that of Kunreuther and Miller
Goicoechea,A., D. R. Hansen, and L. Duckstein, Multiobjective Decision Analysis With Engineeringand BusinessApplications,John
[this issue],which can be contrastedto that reported by CosWiley, New York, 1982.
griff et al. [this issue] and by Loucks et al. [this issue], esHadden, S. G., Scientificdata and public policy, Science,216, 508pecially with respect to the involvement of the user in the
509, 1982.
modelingand programmingefforts.
House, P. W., and J. McLeod, Large Scale Models for Policy Evalutime monitoring and control of water supply systems.One of
the most impressiveexamplesof this is being carried out in
CONCLUSIONS
The interactiveapproachto modeling and model usefor the
exploration,study,and synthesisof resourcemanagmentplans
or policiesis motivated by both a need and by a technology.
There is clearly a continuingneedfor a better meansof generating and effectivelycommunicatingtechnicalinformation to
thoseinvolved in planning and policy making. If models are
to be usedto do this, we believesuchmodelsmust be easyto
use,easy to understand,and must be adaptive to a wide variety of possiblepolicy alternatives.They must also be able to
provide credibleinformation when suchinformation is needed.
Furthermore, they should be designedso as to provide an
unbiasedguide that helpspolicy makersdecidewhat it is they
should do as well as how best to do it.
In addition, the increasing availability and use of timesharingcomputer networks,microcomputers,and interactive
computerlanguages(software)make it possiblefor most of us,
as individualsor as membersof variousorganizations,to have
ation, John Wiley, New York, 1977.
Kahn, H., On ThermonuclearWar, Princeton University Press,New
Jersey,1960.
Kunreuther, H., and L. Miller, Interactive computer modeling for
policy analysis:The flood hazard problem, Water Resour.Res.,this
issue.
Lara, F., and W. M. Sachs,Modeling of the environment through a
systemof models,Appl. Math. Model., 2, 216-219, 1978.
Little, J. D.C., Models and managers: The concept of a decision
calculus,Manage. Sci., 16(8), 466-485, 1970.
Loucks, D. P., et al., Research in water resources and environmental
systemsmodeling,Nat. Res. Forum,8, 219-240, 1984.
Loucks, D. P., M. R. Taylor, and P. N. French, Interactive data
managementfor resourceplanning and analysis, Water Resour.
Res., this issue.
McClure, R. H., Educatingthe future usersof O.R., Interfaces,11(5),
108-112, 1981.
Pollack, M., Interactive models in operationsresearch--An introduction and some future research directions, Cornput. Oper. Res., 1,
305-312, 1976.
Quade, E. S, and H. J. Miser, Handbookof SystemsAnalyses,International Institute for Applied SystemsAnalysis,Laxenburg,Austria,
1983.
102
LoucIcs ET AL.: INTERACTIVEMODELING OVERVIEW
Sebenius,J. K., The computer as a mediator, J. Policy Anal. Manage.,
1(1), 77-95, 1981.
Sprague, R. H., Jr., and E. D. Carlson, Building Effective Decision
SupportSystems,EnglewoodCliffs, Prentice-Hall,New Jersey,1982.
Straus,D. B., and P. B. Clark, Computer-assisted
negotiations:Bigger
problemsneedbetter tools, Environ.Prof., 2, 75-87, 1980.
Walker, W. E., Models in the policy process:Past, present, and
future, Interfaces,12(5),91-100, 1982.
Weger,C., Graphicshelp policy makersutilize data, Cornput.Graphics
News, 2(3), 1-6, 1981.
Whitted, T., Somerecentadvancesin computergraphics,Science,215,
767-774, 1982.
Zeleny, M., Notes, ideasand techniques:New vistasof management
science,Cornput.Oper.Res.,2, 121-125, 1981a.
Zeleny, M., Multiple Criteria Decision Making, McGraw-Hill, New
York, 1981b.
D. P. Loucks,HollisterHall, Cornell University,Ithaca, NY 14853.
(Received March 17, 1984;
acceptedMarch 21, 1984.)
•
•
-'
o
0
0
ß
::T:l',,:'::, m
Plate 8. Lake Neusiedl watershed on the Austrian-Hungarian
border(redline)showing
datafromsimulation
of interaction
between
touristandagriculture
development
andlakewaterquality.
Plate 7. Color-coded predicted stream quality displayed over
streamon a darkeneddigitizedmap of river basin.
information
can be disPlate9. Rangesof nitrogenconcentrations
in the unsaturated Plate10. Graphicaland alphanumeric
rasterimagefor easein interactively
anasubsurface
zonesof a portionof LongIslandare shownin different playedovera darkened
lyzingvariousspatialandtimevaryingphenomena.
colors,asdefinedin menusectionof the display.
!
'
'.
m
Teik
j
Plate 11 A schematic
of a portionof Melbourne'swater supply
.'
Plate 12. A schematicof a portion of Melbourne'swater transler
from a majorreservoir
to the distribution
systemshowing
system
showing
actualstatusof flows,valvesettings,
andreservoir system
volumes.
modeled and actual operational data.