Chapter 35 Appendix A: Multi-Criteria Decision

Analysis

Linkov I, Steevens J

A detailed analysis of the theoretical foundations of different MCDA

methods and their comparative strengths and weaknesses is presented in
Belton and Stewart (2002). MCDA methods utilize a decision matrix to
provide a systematic analytical approach for integrating risk levels,
uncertainty, and valuation, which enables evaluation and ranking of many
alternatives. MCDA overcomes the limitations of less structured methods
such as comparative risk assessment (CRA), which suffers from the
unclear way in which it combines performance on criteria (see Bridges et
al. 2005 for more information on CRA). Within MCDA, almost all
methodologies share similar steps of organization and decision matrix
construction, but each methodology synthesizes information differently
(Yoe 2002). Different methods require diverse types of value information
and follow various optimization algorithms. Some techniques rank op-
tions, some identify a single optimal alternative, some provide an income-
plete ranking, and others differentiate between acceptable and unaccept-
able alternatives.
Elementary MCDA methods can be used to reduce complex problems to
a singular basis for selection of a preferred alternative. However, these
methods do not necessarily weight the relative importance of criteria and
combine the criteria to produce an aggregate score for each alternative.
While elementary approaches are simple and can, in most cases, be
executed without the help of computer software, these methods are best
suited for single-decision maker problems with few alternatives and
criteria, a condition that is rarely characteristic of environmental projects.
Table A1 summarizes a number of more sophisticated MCDA methods.
Multi-attribute utility theory (MAUT), multi-attribute value theory
(MAVT), and the analytical hierarchy process (AHP) are more complex
methods that use optimization algorithms, whereas outranking eschews
optimization in favor of a dominance approach. The optimization ap-
proaches employ numerical scores to communicate the merit of each
816 I. Linkov and J. Steevens

option on a single scale. Scores are developed from the performance of

alternatives with respect to individual criteria and then aggregated into an
overall score. Individual scores may be simply summed or averaged, or a
weighting mechanism can be used to favor some criteria more heavily than
others. The goal of MAUT is to find a simple expression for the net
benefits of a decision. Through the use of utility or value functions, the
MAUT method transforms diverse criteria into one common scale of utility
or value. MAUT relies on the assumptions that the decision-maker is
rational (preferring more utility to less utility, for example), that the
decision-maker has perfect knowledge, and that the decision-maker is
consistent in his judgments. The goal of decision-makers in this process is
to maximize utility or value. Because poor scores on criteria can be
compensated for by high scores on other criteria, MAUT is part of a group
of MCDA techniques known as “compensatory” methods.
Similar to MAUT, AHP (Saaty 1994) aggregates various facets of the
decision problem using a single optimization function known as the
objective function. The goal of AHP is to select the alternative that results
in the greatest value of the objective function. Like MAUT, AHP is a
compensatory optimization approach. However, AHP uses a quantitative
comparison method that is based on pair-wise comparisons of decision
criteria, rather than utility and weighting functions. All individual criteria
must be paired against all others and the results compiled in matrix form.
For example, in examining the choices in the selection of a non-lethal
weapon, the AHP method would require the decision-maker to answer
questions such as, “With respect to the selection of a weapon alternative,
which is more important, the efficiency or the reduction of undesired
effects (e.g., health impacts)?” The user uses a numerical scale to compare
the choices and the AHP method moves systematically through all pair-
wise comparisons of criteria and alternatives. The AHP technique thus
relies on the supposition that humans are more capable of making relative
judgments than absolute judgments. Consequently, the rationality assump-
tion in AHP is more relaxed than in MAUT.
Unlike MAUT and AHP, outranking is based on the principle that one
alternative may have a degree of dominance over another (Kangas et al.
2001). Dominance occurs when one option performs better than another
on at least one criterion and no worse than the other on all criteria (ODPM
2004). However, outranking techniques do not presuppose that a single
best alternative can be identified. Outranking models compare the
performance of two (or more) alternatives at a time, initially in terms of
each criterion, to identify the extent to which a preference for one over the
other can be asserted. Outranking techniques then aggregate the
preference information across all relevant criteria and seek to establish the
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 817

strength of evidence favoring selection of one alternative over another.

For example, an outranking technique may entail favoring the alternative
that performs the best on the greatest number of criteria. Thus, outranking
techniques allow inferior performance on some criteria to be compensated
for by superior performance on others. They do not necessarily, however,
take into account the magnitude of relative underperformance in a criterion
versus the magnitude of over-performance in another criterion. Therefore,
outranking models are known as “partially compensatory.” Outranking
techniques are most appropriate when criteria metrics are not easily
aggregated, measurement scales vary over wide ranges, and units are
incommensurate or incomparable (Seager 2004).
 818

Table A1. Comparison of Critical Elements, Strengths and Weaknesses of Several Advanced MCDA Methods: MAUT, AHP, and
Outranking (after [19]).
Method Important elements Strengths Weaknesses
Multi- • Expression of overall performance • Easier to compare alternatives whose • Maximization of utility may
attribute of an alternative in a single, non- overall scores are expressed as single not be important to decision
utility theory monetary number representing the numbers makers
utility of that alternative • Choice of an alternative can be transparent • Criteria weights obtained
• Criteria weights often obtained by if highest scoring alternative is chosen through less rigorous
I. Linkov and J. Steevens

directly surveying stakeholders • Theoretically sound — based on utilitarian stakeholder surveys may not
philosophy accurately reflect stakeholders’
• Many people prefer to express net utility true preferences
in non-monetary terms • Rigorous stakeholder
preference elicitations are
expensive

Analytical • Criteria weights and scores are • Surveying pairwise comparisons is easy to • The weights obtained from
hierarchy based on pairwise comparisons of implement pairwise comparison are
process criteria and alternatives, strongly criticized for not
respectively reflecting people’s true
preferences
• Mathematical procedures can
yield illogical results. For
example, rankings developed
through AHP are sometimes
not transitive
Method Important elements Strengths Weaknesses
Outranking • One option outranks another if : • Does not require the reduction of all • Does not always take into
1. “it outperforms the other on criteria to a single unit account whether over-
enough criteria of sufficient • Explicit consideration of possibility that performance on one criterion
importance (as reflected by the very poor performance on a single can make up for under-
sum of criteria weights)” and criterion may eliminate an alternative from performance on another
2. it “is not outperformed by the consideration, even if that criterion’s • The algorithms used in
other in the sense of recording a performance is compensated for by very outranking are often relatively
significantly inferior good performance on other criteria complex and not well
performance on any one understood by decision makers
criterion”
• Allows options to be classified as
“incomparable”
Chapter 35 Appendix A: Multi-Criteria Decision Analysis
819
820 I. Linkov and J. Steevens

Example Ahp Application Framework

As an illustrative example of the analytical hierarchy process, consider the

selection of a harmful algal bloom management strategy. Three options
are available to the hypothetical managers:
• Algaecides
• Flushing
• Detoxification

The first step is to decide upon the objectives or criteria by which the
alternative management techniques will be measured. As an example, we
select the following criteria: (1) the strategy’s human health impacts, (2) its
environmental impacts, and (3) its social impacts.
The second step is to weight the importances of these criteria for the
decision maker. Although in this simple scenario it would be possible to
assign weights directly, in many practical applications it may be difficult
because of the multitude of criteria and subcriteria that the decision maker
may face. Therefore, in AHP, the decision-maker does not give
importance weightings directly; rather, the category weightings are derived
from a series of relative judgments. In this scenario, the decision-maker
has input three relative judgments, in the form of weightings ratios. He
has, for example, weighted human health impacts as four times more
important than social impacts (see Table A2). From these relative
weightings, AHP derives normalized weightings for the three criteria (see
Table A3).

Table A2. Relative importance weightings, in the ratio form of row element /
column element.
Main criteria table Human Health Environmental Social Impacts
Impacts Impacts
Human Health Impacts 4.0 4.0
Environmental Impacts 1.0
Social Impacts
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 821

Table A3. Importance weightings for

main criteria categories.
Main criteria weightings
Human Health Impacts 0.667
Environmental Impacts 0.167
Social Impacts 0.167

Additionally, even in this simple case, because the main criteria

categories are too broad to be used directly in evaluating management
alternatives, sub-criteria within each of these categories should be
developed. Within the Human Health Impacts category, for instance, one
might consider drinking water quality, dermal effects, and inhalation
effects. Similarly, sub-criteria may be developed for the other two criteria
categories – such as the strategy’s effects on fish, its birds, and mammals,
or its cost and public acceptability (see Table A4). Sub-criteria are
compared and weighted in a pairwise manner similar to that for the main
criteria (see Table A5, Table A6, and Table A7).

Table A4. Sub-criteria for each main

criteria category.
Goal: Identify best management
techniques for harmful algal blooms

Main criteria Sub-criteria

category
Human Health • Drinking water
Impacts quality
• Dermal effects
• Inhalation effects
Environmental • Effects on fish
Impacts • Effects on birds
• Effects on mammals
Social Impacts • Cost
• Public acceptability
822 I. Linkov and J. Steevens

Table A5. Importance weightings for Human Health Impacts sub-criteria.

Human Health Drinking water Dermal effects Inhalation effects
Impacts sub-table quality
Drinking water quality 7.0 5.0
Dermal effects 1.0

Inhalation effects

Table A6. Importance weightings for Environmental Impacts sub-criteria.

Environmental Impacts Effects on fish Effects on birds Effects on
sub-table mammals

Effects on fish 1.0 7.0

Effects on birds 8.0

Effects on mammals

Table A7. Importance weightings for Social Impacts sub-criteria.

Social Impacts sub-table Cost Public acceptability
Cost 6.0

Public acceptability

Once relative weightings have been given for each of the sub-criteria,
normalized weightings may be calculated for use in scoring different
harmful algal bloom management alternatives (see breakdown in Table
A8).
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 823

Table A8. Importance weightings for both main criteria categories and embedded
sub-criteria.
Goal: Select harmful algal bloom management Weighting Sub-weighting
response

Human Health Impacts 0.667

• Drinking water quality 0.747
• Dermal effects 0.119
0.134
• Inhalation effects
Environmental Impacts 0.167
• Effects on fish 0.458
• Effects on birds 0.479
0.063
• Effects on mammals
Social Impacts 0.167
• Cost 0.857
• Public acceptability 0.143

The third step is to measure relative performance of each management

option on each criteria. Again, the decision-maker inputs a relative ranking
– only now it is a preference ranking between alternatives rather than an
importance ranking among criteria. If a quantitative answer is not given, a
qualitative statement may be transformed into a numerical value through a
standardized system (i.e. the numbers 1, 3, 5, 7, and 9 correspond to the
judgments “equally important,” “moderately more,” “strongly more,”
“very strongly more,” and “extremely more,” respectively). Once the
decision-maker gives inputs for each alternative under each sub-criteria, he
may use the previously obtained weightings to calculate scores for each
main criteria, followed by an overall score for each alternative (see Table
A9). The highest scoring alternative is, according to the rankings and
preferences given by the decision-maker throughout the analytic hierarchy
process, the best strategy for the situation.
824 I. Linkov and J. Steevens

Table A9. Score breakdown for example decision.

Goal: Select harmful algal bloom Algaecides Flushing Detoxification
management response

Human Health Impacts 0.061 0.332 0.607

• Drinking water quality 0.061 0.353 0.586
• Dermal effects 0.060 0.249 0.691
• Inhalation effects 0.062 0.285 0.653
Environmental Impacts 0.779 0.112 0.109
• Effects on fish 0.783 0.174 0.043
• Effects on birds 0.778 0.042 0.180
• Effects on mammals 0.761 0.191 0.048
Social Impacts 0.100 0.320 0.581
• Cost 0.089 0.323 0.588
• Public acceptability 0.163 0.297 0.540
OVERALL SCORE 0.187 0.293 0.520

Many software packages exist to assist the decision-maker with

implementation of the above process.

Framework Effectiveness

Effective decision-making requires an explicit structure for jointly

considering the environmental, ecological, technological, economic, and
socio-political factors relevant to evaluating alternatives and making a
decision. Integrating this heterogeneous information with respect to
human aspirations and technical applications demands a systematic and
understandable framework to organize the people, processes, and tools for
making a structured and defensible decision. Based on our review of
MCDA, we have synthesized our understanding into a systematic decision
framework (Fig. A1). This framework is intended to provide a generalized
road map to the decision-making process.
Having the right combination of people is the first essential element in
the decision process. The activity and involvement levels of two basic
groups of people (decision-makers and scientists & engineers) are
symbolized in Fig A1 by dark lines for direct involvement and dashed lines
for less direct involvement. While the actual membership and the function
of these groups may overlap or vary, the roles of each are essential in
maximizing the utility of human input into the decision process. Each
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 825

group has its own way of viewing the world, its own method of
envisioning solutions, and its own societal responsibility. Policy- and
decision-makers spend most of their effort defining the problem context
and the overall constraints on the decision. In addition, they may have
responsibility for the selection of the final decision and its implementation.
Scientists and engineers have the most focused role in that they provide the
measurements or estimations of the desired criteria that determine the
success of various alternatives. While they may take a secondary role as
decision-makers, their primary role is to provide the technical input as
necessary in the decision process.
The framework places process in the center (Fig. A1). While it is
reasonable to expect that the decision-making process may vary in specific
details among regulatory programs and project types, emphasis should be
given to designing an adaptable structure so that participants can modify
aspects of the project to suit local concerns, while still producing a
structure that provides the required outputs. The process depicted follows
two basic themes: 1) generating alternatives, success criteria, and value
judgments and 2) ranking the alternatives by applying the value weights.
The first part of the process generates and defines choices, performance
levels, and preferences. The latter section methodically prunes non-
feasible alternatives by first applying screening mechanisms (for example,
overall cost, technical feasibility, possible undesired consequences, or
general societal acceptance) followed by a more detailed ranking of the
remaining options by decision analytical techniques (AHP, MAUT,
outranking) that utilize the various criteria levels generated by tools such
as modeling, monitoring, or stakeholder surveys.
 826
I. Linkov and J. Steevens

Fig. A1. General MCDA framework. Solid lines symbolize direct group involvement; dashed lines symbolize less direct
involvement.
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 827

As shown in Fig. A1, the tools used within group decision-making and
scientific research are essential elements of the overall decision process.
As with people, the applicability of the tools is symbolized by solid lines
(direct or high utility) and dotted lines (indirect or lower utility). Decision
analysis tools help to generate and map value judgments into organized
structures that can be linked with the other technical tools from risk
analysis, modeling and monitoring, and cost estimations. Decision
analysis software can also provide useful graphical techniques and
visualization methods to express the gathered information in
understandable formats. When changes occur in the requirements or
decision process, decision analysis tools can respond efficiently to
reprocess and iterate with the new inputs. The framework depicted in Fig.
A1 provides a focused role for the detailed scientific and engineering
efforts invested in experimentation, monitoring, and modeling that provide
the rigorous and defendable details for evaluating criteria performance
under various alternatives. This integration of decision and scientific and
engineering tools allows each to have a unique and valuable role in the
decision process without attempting to apply either type of tool beyond its
intended scope.
As with most other decision processes, it is assumed that the framework
in Fig. A1 is iterative at each phase and can be cycled through many times
in the course of complex decision-making. A first-pass effort may
efficiently point out challenges that may occur or modeling studies that
should be initiated. As these challenges become more apparent, one
iterates again through the framework to explore and adapt the process to
address the more subtle aspects of the decision, with each iteration giving
an indication of additional details that would benefit the overall decision.

Conclusions

The end result of the application of multi-criteria decision analysis is a

comprehensive, structured process for selecting the optimal alternative in
any given situation, drawing from stakeholder preferences and value
judgments as well as scientific modeling and risk analysis. This structured
process would be of great benefit to decision-making for homeland
security, where there is currently no structured approach for making
justifiable and transparent decisions with explicit trade-offs between social
and technical factors. The MCDA framework links technological
performance information with decision criteria and weightings elicited
from decision-makers, allowing visualization and quantification of the
trade-offs involved in the decision-making process. As demonstrated
828 I. Linkov and J. Steevens

above, it is of great utility in applications such as management techniques

for HABs.

Chapter 35 Appendix A References

Ame MV, Diaz MD, Wunderlin DA (2003) Occurrence of toxic cyanobacterial

blooms in San Roque Reservoir (Cordoba, Argentina): A field and
chemometric study. Environmental Toxicology 18(3):192-201
Bridges T, Kiker G, Cura J, Apul D, Linkov I (2005) Towards Using Comparative
Risk Assessment To Manage Contaminated Sediments. In: Levner E, Linkov
I, Proth JM (eds) Strategic Management of Marine Ecosystems. Springer,
Amsterdam
Chorus I, Bartram J (1999) Toxic cyanobacteria in water: A guide to their public
health consequences. World Health Organization, Geneva
http://www.who.int/docstore/water_sanitation_health/toxicyanobact/begin.ht
m. Accessed 24 October 2005
Cole RA, Scodari PF, Franklin MA (2005) Impacts and management costs caused
by freshwater aquatic nuisance species at civil works projects operated by the
United States Army Corps of Engineers. Alexandria VA: USACE Institute
for Water Resources
Kangas J, Kangas A, Leskinen P, Pykalainen J (2001) MCDM methods in
strategic planning of forestry on state-owned lands in Finland: applications
and experiences. Journal of Multi-Criteria Decision Analysis 10:257-271
Landsberg JH (2002) The effects of harmful algal blooms on aquatic organisms.
Reviews in Fisheries Science 10(2):113-390
Nelson TA, Nelson AV, Tjoelker M (2003) Seasonal and spatial patterns of
“Green tides” (Ulvoid algal blooms) and related water quality parameters in
the coastal waters of Washington State USA. Botanica Marina 46(3):263-275
Office of the Deputy Prime Minister (ODPM) DLTR Multi-Criteria Decision
Analysis Manual. Downloaded June 16 2004
http://www.odpm.gov.uk/stellent/groups/odpm_about/documents/page/odpm_
about_608524-02.hcsp
Paerl HW (1996) A comparison of cyanobacterial bloom dynamics in freshwater,
estuarine and marine environments. Phycologia 35(Supplement):25-35
Pertola S, Kuosa H, Olsonen R (2005) Is the invasion of Prorocentrum minimum
(Dinophyceae) related to the nitrogen enrichment of the Baltic Sea? Harmful
Algae 4(3):481-492
Rhudy KB, Sharma VK, Lehman RL, et al. (1999) Seasonal variability of the
Texas ‘brown tide’ (Aureoumbra lagunensis) in relation to environmental
parameters. Estuarine Coastal and Shelf Science 48(5):565-574
Saaty TL (1994) Fundamentals of Decision Making and Priority Theory with the
Analytic Hierarch Process. The Analytic Hierarch Process Series: vol VI
Pittsburgh PA: RWS
 Chapter 35 Appendix A: Multi-Criteria Decision Analysis 829

Seager TP (2004) Understanding industrial ecology and the multiple dimensions

of sustainability. In Strategic Environmental Management by O’Brien and
Gere Engineers. New York: John Wiley & Sons
Woods Hole Oceanographic Institution (2005) Red tide and harmful algal
blooms. http://www.whoi.edu/redtide/. Accessed 24 October 2005
Yoe C (2002) Trade-Off Analysis Planning and Procedures Guidebook. Prepared
for Institute for Water Resources, US Army Corps of Engineers April 2002