You are on page 1of 21

EFECT evaluation framework of environmental impacts and

costs of transport initiatives


Dimitrios Tsamboulas
*
, George Mikroudis
Department of Transportation Planning and Engineering, National Technical University of Athens, 5 Iroon Polytechniou,
15773 Zografou, Athens, Greece
Abstract
EFECT is a generalised methodological framework for evaluating the impacts resulting from trans-
portation projects with a specic orientation to environmental impacts. The innovative aspect of the
methodological framework is the combination of Multi-Criteria Analysis (MCA) with Cost-Benet
Analysis (CBA) methods to come up with an overall assessment of transport initiatives impacts over
dierent geographical regions and time periods. Thus, it addresses both spatial and time impacts of
transportation networks for all modes. The framework comprises four steps: structuring, weighting, rating,
and exploring. Uncertainty is explicitly treated in the framework through fuzzy sets or, indirectly, through
sensitivity testing. EFECT, using an additive function combining MCA and CBA methods, provides
simplicity and intuitive understanding of results. This allows the use of either a core approach with a basic
set of criteria and weights, or the application of a more detailed evaluation, when needed. The way EFECT
is applied and the results produced are presented through a case-study example. 2000 Elsevier Science
Ltd. All rights reserved.
Keywords: Transport investments; Environmental impacts; Multi-criteria analysis; Evaluation of transport projects;
Uncertainty in evaluation
1. Introduction
Evaluation (ex-post or ex-ante) of transportation plans, initiatives and projects have been
carried out in the past, using a variety of methodological frameworks. In terms of content, the
available methodologies can be classied into Cost-Benet Analysis (CBA) (Abelson, 1979;
Transportation Research Part D 5 (2000) 283303
www.elsevier.com/locate/trd
*
Corresponding author. Tel.: +301-7721367; fax: +301-7721327.
E-mail address: dtsamb@central.ntua.gr (D. Tsamboulas).
1361-9209/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S1361- 9209( 99) 00038- 3
Freeman, 1979; Danning and MacKnight, 1990), Multi-Criteria Analysis (MCA), (Nijkamp and
Blaas, 1993), Social-based Analysis (Tsamboulas and Mikroudis, 1996; Verhoef, 1994), Decision-
Analysis, other specic type-applications (land suitability analysis, rapid assessment methods,
resource management approaches) and simulation/mathematical modeling (Nickols and Heyman,
1982; Roson and Small, 1998). In terms of typology, they all use variations of checklists, matrices,
networks and overlay methods. In all cases, one of the key objectives is the inclusion of envi-
ronmental quality/impacts considerations in planning and decision-making. Recently, there is
research relating the environment impact assessments to policy considerations (Banister and
Button, 1993; Button, 1993; Banister, 1998).
At European level, it has been recognized that Environmental Assessment (especially at the
strategic level bearing the generic name Strategic Environmental Assessment (SEA)) should be
developed as an integral part of the decision-making process for policies, plans and programmes
(EC, 1993, 1996). The same importance is given in USA with the creation of Environment
Protection Agency (EPA). Especially for the planning or implementation of transportation
projects, environmental appraisals start with an identication of possible impacts to various
aspects of the natural and social environment, and result either in some form of tabular rep-
resentation of impacts or in an aggregate scoring of alternatives. The last part requires con-
verting environmental eects into monetary or non-monetary values, a task with serious
methodological and philosophical diculties. The nal balancing of values and ranking of al-
ternatives is either included in the method or left to the decision-maker (DM). The `easy' part of
an environmental evaluation is the analytical one, being the identication of the impacts. The
`hard' part is the evaluation of each impact and the synthesis of the results in order to facilitate
decision-making.
The diculties in arriving at a generalized framework for environmental evaluation are both
procedural and conceptual. In operational terms, such framework should be easy to be under-
stood and employed by DMs, while meeting a series of eciency criteria, such as generality,
independence, reliability, exibility, few data needs, etc. On the other hand the evaluation
framework should be philosophically coherent, able to deal with challenging issues, such as un-
certainty, time frame, network eects, modal changes, while integrating cost and judgement values
into the evaluation.
This paper presents a generic methodological framework, called Evaluation Framework of
Environmental impacts and Costs of Transport (EFECT) initiatives. Its scope is rather
broad since it aims to cover all kinds of transport environmental initiatives, namely policies,
plans and projects, either for small-scale (local) or large-scale (regional/national) assess-
ments.
In order to make a more simple presentation the method is focusing on environmental impacts.
However, the same principles and structure can be applied to evaluate other type of impacts as
well.
The requirements for such a generic methodology are presented in the following section. Next,
the detailed framework is formulated and the procedure used in applying the methodology are
explained. Then, a way to handle uncertainty is provided, as well as the manner to deal with the
time dimension of environmental evaluation. Finally, a case-study of EFECT application in the
evaluation of alternative road design schemes is presented, which demonstrates the advantages of
the methodological framework, when used in a GIS environment.
284 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
2. Requirements for the evaluation methodology
2.1. General requirements
The basic requirements for the development of a generic methodological framework for en-
vironmental evaluation of transport infrastructure projects are the following:
combine environmental eects with monetary values,
follow a network approach for the spatial variation of impacts,
consider variation of impacts over time,
handle uncertainty,
be practical and understandable.
The above requirements lead to certain fundamental choices in the mathematical structures that
are part of the framework. Thus, one basic choice is to employ an additive function for the ag-
gregation of individual values into a single score. This allows for an easy and robust combination
of MCA with CBA (Tsamboulas, 1998), and it facilitates application to networks on dierent
regions in space and at dierent periods of time. It also makes the method more practical and
easier to be understood by the DM, and eventually, more defendable.
2.2. Practical and methodological issues
End-users of the framework can be DMs. They might be either individuals such as transport
planners and designers or public entities (European, national, local) with institutional, adminis-
trative interests.
The methodological framework could be applied both to urban and inter-urban projects, re-
gardless of transport mode, as well as to the development of terminals. On the physical scale, it is
appropriate for projects referring to a section (e.g., 10 km), a corridor, or a network. In addition,
the project can be extended over several regions in dierent areas or countries. Alternatives can be
considered both at scheme (conception) level and at project level. Moreover, the methodology
handles environmental impacts at any level of detail in the natural and social environment. Lastly,
all of the above could be handled in a time framework and provisions are made to deal with
uncertainties.
2.3. Overall characteristics of the methodology
Based on the above requirements the main characteristics of EFECT are the following:
the use of a hierarchical approach,
the utilisation of an additive multi-criteria model,
the combined use of MCA and CBA methods,
the use of weights for criteria and spatial aggregation of impacts,
the inclusion of uncertainty at all stages of the evaluation,
the explicit handling of the time dimension of impacts.
For practical reasons a core method is used. For a more detailed evaluation, the treatment of
uncertainty and dynamic time eects can easily be added to.
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 285
3. Methodological framework
3.1. Overview
EFECT comprises four steps:
1
1. Structuring: Constructing a decision-tree from the environmental and monetary factors of the
problem:
(a) Objectives: Dening basic goal and/or objective(s).
(b) Criteria: Describing the appraisal parameters and their corresponding index functions.
(c) Hierarchy: Dening how the criteria are related hierarchically to the problem objectives.
2. Weighting: Dening how the criteria, geographical regions and time horizons are weighted
against each other in the evaluation:
(a) criteria weights,
(b) regional weights,
(c) time weights.
3. Rating of the alternative projects:
(a) applying the criteria, weights and structure to evaluate and rank the alternative projects,
(b) combining MCA with CBA results.
4. Exploring the results:
(a) performing sensitivity analysis on the criteria, weights and values in order to explore pos-
sible variations of end results,
(b) using fuzzy sets to incorporate uncertainty in the evaluation,
(c) considering the evolution of (1) and (2) of the above, over time.
In a given decision situation, alternative transport projects can be evaluated following the above
four steps and by receiving `scores' corresponding to the environmental criteria employed by the
methodology. For scoring, either a `quality index' can be used, where higher scores indicate better
quality for the environment, or on the contrary, an `impact/pollution index' can be used, where
higher scores mean higher impacts or levels of pollution, thus lower environmental quality.
The rst three steps comprise a `core' method, which can be applied in the majority of cases.
When, a more detailed evaluation is needed the last step (3) is followed, with the inclusion of
additional procedures for handling uncertainty and time dimension.
3.2. First step structuring
It denes a decision-tree, comprising the environmental and monetary issues consisting of (a)
the objectives, (b) the evaluation criteria and (c) how the criteria are combined to meet the ob-
jectives.
The `objectives' usually require specifying the goal and the options of the evaluation. Such
options can be alternative policies, plans, or projects that extend over a certain geographical
region (space) and a period of time. Thus, alternative `option', `space', and `time' combinations
1
The original idea of the four steps was developed within the EUNET project, referring to all types of impacts
(Beuthe et al., 1998).
286 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
can be dened. The main goal of the evaluation is either to rank the options directly through
MCA, or to rate the options for further combination with cost data or CBA results. In addition,
variation of options over space allows the user to follow a network approach: projects can be
broken down to nodes and links and evaluated according to the geographical regions they are
located in. On the other hand, variation of options over time allows the inclusion of time related
dynamic eects.
Depending on the objectives, a dierent set of `Criteria' may be used. The term criterion is used
instead of the more common term indicator, since the criteria are the elements forming a decision,
and each criterion requires one or more indicators for its description. The evaluation and ranking
of alternative transport options requires the construction of a hierarchy of several sub-criteria
which when combined together produce the score of a given option, according to the super-cri-
terion they are related to in the hierarchy. In this manner, any number of criteria groupings, which
in turn are grouped into other super-criteria can be formed.
In EFECT, the user, according to his/her objectives (Tsamboulas and Mikroudis, 1999) can
dene the form of the hierarchy. An initial set of environmental factors that can be used as
framework variants in the hierarchy includes impacts to the natural and the anthropogenic en-
vironment from the construction and operation of transport projects of all modes: land (road/
rail), ports for maritime transport and airports for air transport (Mikroudis, 1996). The formu-
lation of these environmental factors into a decision-tree is shown in Fig. 1.
The underlying mathematic structure of the model (additive function) allows for maximum
exibility in the selection (or even addition) of criteria (and subsequent indicators) that the DM
wants to include in the evaluation. Since weights are calculated for the selected criteria and
normalized at the end, the nal score represents the overall assessment for the set of weights
selected by the DM.
In forming the decision tree, balance is kept between CBA and MCA methods by keeping their
components separate all the way to the end, until making the nal decision. A two-stage approach
is adopted comprising:
Fig. 1. Proposed structure of decision-tree.
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 287
(a) a pre-screening stage using CBA and a budgetary threshold;
(b) an evaluation stage using a MCA method, where CBA output(s) is (are) one of the main
criteria used in the decision.
Note that, by keeping CBA and MCA results separate CBA results can be converted to a common
scale in order to come up with a total score for MCA. Other options consider: MCA results as
benets, using an environmental quality index, so they can be combined with costs; separate
monetary and non monetary terms, without converting one to the other, leaving the DMs to
balance the two at the end.
In constructing such a decision tree, the following are important:
Criteria should be able to cope with dierent modes, countries and types of investments (pri-
vate/public or both). If necessary, non-technical criteria, such as political and/or ethical should
be used.
The level of aggregation of impacts, at which the approach is typically required to work should
also be taken into account, with the possible introduction of sub-criteria, such as `public accep-
tance'. Such sub-criteria should assume a non-zero weight only when further disaggregation is
necessary.
Care to avoid double counting within both the CBA and MCA components of the tool and
between the CBA and MCA components. Impacts such as employment or accessibility almost
inevitably will double-count with standard CBA impacts in ways entirely impossible to disen-
tangle. The best proposed compromises seem to be:
to introduce criteria (especially environmental ones), being as orthogonal as possible to impacts
recorded within the CBA,
to measure the MCA criteria to represent only contributions (generally towards strategic
goals), which are over and above those recorded as direct project impacts.
An important feature of the EFECT is the capability to turn on or o, any given criterion, or
combination of option-spatial region or option-time dimension. Only `active' criteria that are
related to space and time options are used in the evaluation. In the second step of the evaluation
criteria weights are normalized based on the active options, space weights are normalized based
on the active options-space combinations, and time weights are normalized based on the active
options-time combinations. This means that certain criteria can be considered `inactive' either for
the purposes of the evaluation, or for a given region and period of time. This allows for changes in
basic features of the project, plan, or policy over space and time, which simply may not exist or be
irrelevant in dierent geographical regions or periods of time. The capability of the methodology
for the creation of some inactive criteria could be used to develop a simpler evaluation meth-
odology, if there is data availability problems or time constraints.
3.3. Second step weighting
A novel aspect of EFECT is that it incorporates the spatial variation of impacts by introducing
a network approach. This means that evaluation of transport infrastructures is not limited to a
single project or corridor but it can encompass an entire network extending over many dierent
regions or even countries. Lastly, impacts extending over dierent periods in time (e.g., envi-
ronmental ones) as well as combination of changes over space and time are handled by the
method. To accomplish this, EFECT employs three sets of weights:
288 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
(a) criteria related weights,
(b) regional (spatial) weights,
(c) time weights.
To determine the weights, one or a combination of the following methods can be used:
1. direct specication by the DMs,
2. specication by a panel of experts, (e.g., Delphi technique),
3. by pair-wise comparisons (Yager, 1977; Saaty, 1980).
For the measurement of criteria-related weights, EFECT employs the pair-wise comparison,
which is the most ecient weighting method, well-established, with a good record of practicality,
operational eciency and reliability, as well as acceptance by the DM (in terms of being short,
easy to understand).
The underlying principle is to establish relative impact weights and also, impact scores if re-
quired, using pair-wise comparison of the judged importance of impacts (or relative levels of
impact). A concern over using this approach for score estimation is that the relative scores
achieved are strongly inuenced by the particular set of alternatives considered. Therefore, if at a
later stage other transport projects are added, there could be rank reversals, particularly if they
have more extreme scores. This is the main reason behind the use of pair-wise comparisons for
weight estimation in the proposed methodology.
Pair-wise comparisons become impractical when the number of criteria is large, usually more
than 10. Thus, it is advisable to keep the number of criteria limited, or to group criteria together
and apply the method within and among groups of criteria. For the determination of regional
weights, EFECT applies a direct specication either by the DMs or by an expert panel.
Regional weights are used in EFECT to incorporate into the evaluation the relative impor-
tance of dierent regions traversed aected by the transport infrastructure. To accomplish it,
EFECT requires a subdivision of the specic project into segments as it traverses dierent re-
gions.
Alternatively, a network approach can be used, considering separate regions that correspond to
the various segments of a transport network, and consequently they can be assigned to both nodes
and links of the specic network. However, dierent weighting must be used for the two com-
ponents of the network since the links are longer and traverse larger geographical areas and the
nodes may conceptually represent entire cities or other anthropogenic activities (e.g., terminals).
Likewise, when comparing projects crossing two dierent geographical regions, all project seg-
ments in one region and all project segments in the other region need to be considered and their
results should be combined (see Fig. 2).
Before proceeding to the evaluation, once the dierent regions are dened, dierent sets of
criteria weights need to be established within each region. Each of these criteria weights is gen-
erally determined through pair-wise comparisons.
The regional weights can be determined through direct specication by experts. Nevertheless,
for a small number of criteria and regions, pair-wise comparisons may be employed to determine
the weights for all criteria simultaneously.
As an example to illustrate these applications, two regions a and b and two projects are con-
sidered: (I) an airport project in region a, (II) a motorway project in region b. Three super-criteria
C
1
, C
2
, C
3
are employed (e.g., C
1
social impacts, C
2
environmental impacts, C
3
construction
cost). The weights are determined with the following alternative approaches:
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 289
(i) By separate pair-wise comparison: For region a determine criteria weights C
1a
, C
2a
, C
3a
(3
comparisons) and for region b determine criteria weights C
1b
, C
2b
, C
3b
(3 comparisons). The
criteria determine the relative importance of construction costs, social and environmental im-
pacts in regions a and b. Initially the regional weights W
a
and W
b
(1 pair-wise comparison) are
specied. They determine the relative importance for the DM of regions a and b, respectively.
The combined regional-criteria weights are W
1a
W
a
W
3b
W
b
.
(ii) By combined pair-wise comparisons: For the two regions the combined regional-criteria
weights W
1a
, W
2a
, W
3a
, W
1b
, W
2b
and W
3b
(15 comparisons) are specied. This option allows
the modication of the weights obtained, by further multiplying them with the regional weights
W
a
and W
b
to account for the relative importance of regions a and b, in addition to the one
considered in criteria C
1
, C
2
, C
3
.
The DM can consider the weights obtained according to option (i) as a sucient approximation
of those obtained according to option (ii). It is evident that the second option, even for a rea-
sonably small number of criteria (e.g., 10) and regions, (e.g., 5), is practically impossible, requiring
the DM to carry out 1225 comparisons.
In the case of the evaluation of the new project (III) evaluation crossing regions a and b (e.g., a
railway line), EFECT requires the project to be divided into two segments: segment a for region a
and segment b for region b. EFECT employs the weights as dened with the two separate projects
I and II for the respective regions: W
1a
, W
2a
, W
3a
and W
1b
, W
2b
, W
3b
. It is interesting to note that
weights W
1a
, W
2a
, W
3a
are the same for both projects (I) and (III-segment a) of region a, as well
as the weights W
1b
, W
2b
, W
3b
are the same for both projects (II) and (III-segment b) of region b.
As for the regional weights, W
a
is applied to all projects in region a, and W
b
to all projects in
region b.
Lastly, a third set of weights can be handled by EFECT to consider the time dimension. En-
vironmental impacts can be considered as immediate (during construction), short-term (during the
rst years of operation) and long-term. Weights with same or dierent values can be applied,
depending on the length of time or other considerations.
The three types of weights are normalized and then multiplied to obtain an overall weight for
the three dimensions of the evaluation: criteria, space, and time. Normalization takes place by
using only the `active' criteria, the combinations of `options over space', and the combinations of
`options over time'.
Fig. 2. Schematic representation of regional and corresponding criteria-related weights.
290 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
3.4. Third step rating of the alternative projects
The third step requires the integration of steps 1 and 2 for the selection of the optimal
project option among competing alternative ones. At this stage, the methodology consists of a
core evaluation method with a series of alternative or additional procedures, which can enhance
the core process. The core approach is relatively straightforward and uncontroversial. A series
of optional enhancements and/or supplementary procedures are added, which represent im-
portant innovative steps in the environmental evaluation methodologies, the most important
being:
Fuzzy evaluation: An alternative way to represent impacts as fuzzy quantities instead of crisp
numbers, with associated membership functions. This allows for the treatment of uncertainties
both in values and weights used in the assessment.
Time-dependant evaluation: `Snap-shots' taken at dierent time-horizons, for example: current
situation, end of construction period, end of project life (e.g. 30 years). Each snap-shot may
require a re-denition of criteria-weights-regions before evaluating the alternatives. Employing
time integration functions corresponding to each criterion for the specied time intervals can
measure the environmental impacts and the accumulated costs over time. This is an extension
of the decision-making tree into a third dimension, time.
Thus, EFECT is designed in a modular fashion allowing the utilisation of either the core meth-
odology or the enhancements to the core method as presented above and possible future meth-
odological additions.
The core method follows the conventional framework of MCA. The impacts of CBA expressed
in money values can be converted into the MCA by EFECT. Thus `cost' can simply be considered
as an additional criterion in the MCA part. Environmental benets, if any, is incorporated
through the `quality index' derived by MCA, whereas a `pollution index' is assigned to costs.
The evaluation model in EFECT is expressed as:
Rank(A
i
Y O(C
1
Y F F F Y C
nc
)) for i = 1Y npY (1)
where A
i
are the np transport options under the aggregation function O(C
ij
) dened over the nc
criteria using the scores C
ij
.
The additive function chosen by EFECT, being a simple aggregation function and easy to
understand. However, it may exhibit the following problem: while at least one score is of for-
bidding low quality, the overall score may not present low quality and vice versa (due to com-
pensation between the criteria). On the contrary, the product functions do not exhibit this
problem, but they are more dicult to understand since they are highly non-linear. A combi-
nation of all types of functions is possible. The aggregation for the combination of options, space
and time function is used in EFECT in the following form:
O
i
=

j=ncY k=nrY l=nt


JYKYl=1
W
ji
W
ki
W
li
C
j
(X
ij
) (2)
or alternatively for fuzzy applications:
O
i
=

j=nc
j=1
W
jk
C
j
(X
ij
)Y (3)
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 291
S
i
=

k=nr
k=1
W
ik
o
i
Y (4)
T
i
=

l=nt
l=1
W
li
S
li

t=te
t=to
f
li
(t) dtY (5)
where O
i
or o
i
are the aggregation functions of a given segment of option i for all criteria c
j
, S
i
the
aggregation functions of option i, for all its segments k in dierent regions, T
i
the aggregation
function of a option i, over the time interval from t = to to t = te, w
jk
the criteria weights for
option i in region k and time l, w
ki
the regional weights for option i, w
li
the time weights for option
i, f
li
(t) the growth function for S
i
, the region in consideration (for a discrete function, the

is
replaced by R), nc the number of criteria, nr the number of geographical regions crossed by project
i, and nt is the number of time periods for evaluating the impacts, for project i.
The core approach comprises Eq. (2) which replaces Eqs. (3)(5) using crisp numbers. The
treatment of space (Eq. (4)), and the treatment of time (Eq. (5)) are optional to the core approach.
Eq. (2), when w
ki
= w
li
= 1 it is reduced to the classic case of MCA with just a set of criteria that
do not change over space or time. The treatment of uncertainty (presented in Section 4) using
fuzzy numbers is an enhancement to the core approach that can be used for a more detailed
evaluation of impacts.
The above presented model used in EFECT is additive in function, linear in the weights w
j
and physical parameters x
ij
of the criteria, but potentially non-linear in some or all of the
criteria scores c
j
and the growth functions f
ij
. To measure the performance of the specic project
with respect to the chosen criteria, measurement methods are chosen to produce values, which
are at least ordinal and, ideally, in interval/ratio scale measures. If the latter is not possible, a
way for the values to be converted is introduced. Verbal assessments should be avoided except
when used as a purely supplementary qualication to the formal evaluation model. Where
judgements are required, they need to be translated to a common scale with associated verbal
descriptions of typical situations corresponding to some or all of the scale gradations. To this
end, optionally, the fuzzy model can provide for conversions of verbal expressions to numerical
values.
Score levels may be estimated either directly/judgmentally, or by using a simple transformation
graph spanning over all plausible values, usually for criteria which have some direct interval/ratio
scale measurement. On the grounds of simplicity, it is recommended that this graph is linear, but it
can also be non-linear when needed.
The values converted to a scale or scores are measured independently of weights. Scores are
proposed to be on a 0 to 1 scale, where 0 is the lowest feasible and 1 is the highest feasible level to
be attained by the specic criterion. The scale could be increasing or decreasing, depending on the
objectives of the evaluation.
Finally, for the evaluation and ranking of the alternatives each alternative gets an overall score
as derived from the function (1) and thus, the projects under evaluation are ranked rst, second,
third and so on.
To be complete, the core approach should be used with a `default' set of criteria (e.g., those of
Fig. 1). Thus, several evaluations performed over dierent regions or dierent time periods could
292 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
easily be compared. Also, based on the standard set of criteria, reference could be made to
`standard' characterizations of environmental quality, such as `good', `average', `poor', etc.
The core approach, apart from being a reasonably common form of aggregation within the
multi-criteria literature, is also consistent with the idea that the relative inuence of dierent
dimensions of impact should be constant and predictable. Moreover, it is broadly transparent to
users and, as the specic impact level changes, it is consistent with the intuitive notion of non-
linear `returns' to some impacts. The spatial evaluation (Eq. (3)), time-frame evaluation (Eq. (4))
and handling of uncertainty using fuzzy numbers are add-ons to the core approach, for more
detailed evaluations. In addition, the option to measure the criteria performance in monetary
values is a standard option of EFECT.
The use of a non-linear or a step-wise growth function allows to assess the accumulation of
impacts over time considering that certain impacts (e.g., land take) may remain constant over
time, whereas others may have an exponential growth (e.g., pollution levels) or follow some other
type of growth (e.g., noise levels). The consistent use of an additive function over all regions,
project segments, criteria, and time periods of evaluation allows to compare dierent projects in
dierent regions and at dierent time horizons.
3.5. Fourth step exploring the results
Although this step is optional, it constitutes an essential part of the decision-making. It may
comprise any one of the following options:
(a) sensitivity analysis,
(b) treatment of uncertainty through fuzzy sets,
(c) evaluation over time.
The rst option, sensitivity analysis, is achieved through a software implementation of EFECT as
an electronic database of criteria, regions and time frames. This software has been applied to the
presented case study, dealing with the application of EFECT.
In order to explore the relative performance of projects when underlying assumptions change,
the method allows users to exchange one set of impacts and regional or time weights for another
set. This could involve user-specied changes in specic weights or, allow the opportunity for the
user to be able to derive his/her own weight set, if this has to do with MCA impacts. Evidently, in
MCA it is necessary to re-scale weights if any is changed. In EFECT the software model does it
automatically, as well as other necessary calculations, e.g., calculation of eigenvectors, aggrega-
tion, ranking, etc.
Since the user is allowed to add/delete criteria within the MCA, the model allows each time to
directly specify or derive the weights using pair-wise comparisons. For data base purposes, the
original assessment is retained with the initial weights and original score estimates to provide the
basic default projects scores and rankings. In addition, the software tool has the ability to handle
dynamic eects by considering a dierent array of weights corresponding to the considered time
period.
Various sets of criteria-weights can be determined according to dierent viewpoints expressed
by dierent groups (e.g., government, local authorities, interested groups). The dierent view-
points can be examined as part of the sensitivity analysis of the MCA results, taking into con-
sideration approaches to employ MCA in a fuzzy environment (Munda, 1995).
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 293
The sensitivity analysis, which should be used for treatment of uncertainty in the MCA com-
ponent, should be complemented by risk analysis in the CBA. This might include setting the IRR
(internal rate of return) or ROE (returns on equity) to reect levels of risk or changes to the
project design. The options related to the treatment of uncertainty with fuzzy sets and dierent
time horizons are explained in more detail in the following sections.
4. Treatment of uncertainty
4.1. Introduction
There are basically three ways for the treatment of uncertainty in evaluation:
(a) probability-based methods (e.g., Bayes theorem),
(b) fuzzy sets,
(c) informal methods (e.g., expert systems).
It is generally agreed that a fully rigorous treatment of uncertainty is impractical in cases of the
size and complexity tackled by the EFECT. On the other hand, the judgements on scores and
weights for impacts are likely to be particularly `uncertain'. Also, the denition of index functions
is in most cases a rough approximation introducing additional `imprecision' to the end results,
as shown in Fig. 3.
An initial uncertainty, in an observed or assigned (judgmental) value, results into an un-
certainty of the index value, which can be further magnied by an uncertainty in the index
function itself. The uncertainty in value and the resulting uncertainty in index are the most
common ones and can be dealt with in a variety of ways. In the EFECT, this type of un-
certainty can be described and incorporated into the assessment through the use of fuzzy
numbers. Such treatment of uncertainty is considered as an add-on to the core model of
evaluation.
Fig. 3. Uncertainty in evaluating the value of a given criterion, using an index function.
294 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
4.2. Uncertainty handled by EFECT
The method being used for fuzzy evaluation in the EFECT employs symmetric triangular fuzzy
numbers (STFNs). The basic equations (3)(5) of the core approach remain the same, but instead
of the crisp operators for addition, `+', and multiplication `', the fuzzy ones, `' and `', from
Eqs. (6)(8) are employed.
(L
1
Y M
1
Y U
1
) (L
2
Y M
2
Y U
2
) = (L
1
L
2
Y M
1
M
2
Y U
1
U
2
)Y (6)
(L
1
Y M
1
Y U
1
)(L
2
Y M
2
Y U
2
)(L
1
L
2
Y M
1
M
2
Y U
1
U
2
)Y (7)
k(LY MY U) = (k LY k MY k U)X (8)
They represent operations with TFN (triangular fuzzy number) and L is the lower bound, U the
upper bound and M is the modal value, estimated by (9)
M = (L U)a2X (9)
The rst three steps of the EFECT methodological framework essentially remain the same. The
only change is in the comparison of alternatives needed for ranking. This needs to introduce an
additional operator `<', for comparing the fuzzy numbers a and b, as follows:
(L
a
Y U
a
) ` (L
b
Y U
b
)[
c
= TRUE L
a
` L
b
(U
a
L
a
)(0X5 ca2)
0X5
Y (10)
where c is the cumulative membership function, and c >50%.
The above consideration is similar to a probabilistic approach considering a condence level
equal to c. If we substitute the membership functions with probability density functions, a similar
formulation can be achieved on stochastic terms.
On the other hand, in the core model, to achieve eective means for treatment of uncertainty in
the MCA both types of uncertainty (stochastic and fuzzy/imprecision) can be addressed solely
through simple sensitivity testing.
5. Treatment of time
Environmental impacts of transportation projects evolve over time. However, the general
practice is either to produce presentations of impacts (Moavenzadeh, 1994) or to carry assessment
separately, during construction and during operation, and at the end, the evaluator tries to come
up with an overall statement about the desirability of the project. However, a total evaluation,
considering all time horizons and all respective impacts, in rigorous mathematical terms, is not
fully included in EFECT.
Eq. (5) of EFECT introduces an extension to the core MCA so it can be applied over the time
dimension. If discrete time horizons are used the integral is replaced by summation. In the discrete
case, the following situations are possible:
transient impacts (e.g., construction noise) are counted as constant terms,
one-time reversible impacts (e.g., loss of vegetation cover what can be replanted) are only mul-
tiplied by the duration of impact, e.g., construction duration,
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 295
one-time permanent impacts (e.g., changes to the landscape) are multiplied by the respective
time, e.g., project life,
short-term impacts (e.g., noise before construction of noise barriers) are multiplied by the re-
spective duration of impact,
long-term permanent impacts (e.g., impacts on houses that exist or will be built in the future,
adjacent to a road) are multiplied by their estimate exposure times.
Furthermore, when both discrete and continuous criteria are used, a two-fold approach is em-
ployed: summations for discrete criteria and integration for criteria with continuous growth
function over time. An example of the latter is `air pollution emissions' which is a function ba-
sically of trac volumes growing over time. Fig. 4 illustrates an example with four criteria
evolving during the time periods, where the following impacts are presented: (a) discrete transient
impacts (C
1
), one-time permanent impacts (C
2
), and (b) continuous short-term and long-term
impacts (C
3
, C
4
).
Theoretically, in the time-frame evaluation, dierent sets of criteria-weights, in EFECT may
be used for dierent time horizons. For the sake of consistency and simplicity, only values
should change over time. Dierent weights could be used for construction and operation, in-
dicating the dierent attitudes of the public, who are more willing to accept nuisance during the
construction of a project due to its `temporary' nature and the anticipation of benets, once it is
completed.
6. Application of EFECT
The EFECT is used in a case study in Greece, for the appraisal of alternative road schemes in
Agios KonstantinosKamena Vourla area for the PATHE motorway project.
The evaluation concerns the environmental impacts of four alignment options in the coastal
area of Agios KonstantinosAsproneriKamena Vourla. A team of experts selected the most
important environmental criteria. The impacts on the study area from the construction and op-
eration of the motorway were examined both in the short-term and the long-term for each al-
ternative option.
The application is done with the implementation of software developed for EFECT. It is a
Decision Support System (DSS) comprising a MS Access Data Base application plus visual Basic
Fig. 4. Time-frame evaluation of impacts (example with 4 criteria).
296 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
and MapObjects enhancements for additional presentation tools in the form of graphs and GIS
maps.
6.1. Step 1: structuring problem denition
Options: Four alternative options were examined: (I) maximum use of existing corridor; (II)
bypass of Asproneri beach and Kamena Vourla; (3B) preliminary design option (with bridges and
tunnels); and (III) inland scheme through LoggaAgnanti connection. Option III is a full devi-
ation of the coastal zone of Agios KonstantinosKamena Vourla, and consequently its con-
struction has zero direct impacts on the natural and man-made environment of the coastal area.
On the long-term, this option works together with options I and II, whereas for option 3B there is
no long term alternative other than directing trac through the existing coastal National Road
(NR).
Objective: The general objective is the protection of the environment and minimization of
negative impacts on the coastal area.
The degree of environmental protection for each option is expressed on an articial scale of
impact scores, which corresponds to a verbal description as follows:
The scale is used for rating each criteria and the overall alternative options.
Criteria: The criteria used are shown in Table 1.
The above denitions of options, criteria and objectives are dened using the EFECT frame-
work as follows:
The user denes rst the project options (Table 2). The project options are further divided in
three geographical regions: Agios KonstantinosAsproneriKammena Vourla. The importance of
()4) strong negative impact
()3) large negative impact
()2) moderate negative impact
()1) small negative impact
(0) no impact
(+1) small positive impact
(+2) moderate positive impact
(+3) large positive impact
(+4) strong positive impact
Table 1
Criteria for evaluating the alternative solutions
Natural environment Operation nuisance Man-made environment
(1) Landscape (6) Air (10) Residential areas
(2) Soil (7) Noise (11) Land use
(3) Waters (8) Trac (12) City planning
(4) Ecosystems (9) Accidentshazards (13) Cultural heritage
(5) Natural resources (14) Public acceptance
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 297
these regions can also be specied, e.g., as low-moderate-high (or any other value the user may
select).
He also species the time frames for the evaluation, namely, short-term and long-term for this
example. The importance of these time frames can also be specied on the same user-dened scale.
He may then dene (or select from an available list) criteria (Table 3) and sub-criteria (Table 4)
for the evaluation. Then, he may specify the importance of criteria on a scale of his own denition,
for example, using gradations from 1 to 9 (Table 5).
These important values are the same ones applied throughout, namely, for geographical re-
gions, time frames, and criteria or sub-criteria. They are used to generate the weights of the
evaluation, which, once computed can be further edited for ne-tuning by the user.
Table 3
Criteria for evaluation
ID Name
1 Natural environment
2 Operational nuisance
3 Man-made environment
Table 4
Sub-criteria for evaluation
ID Criterion ID to which sub-criterion refers Name
1 1 Landscape
2 1 Soil
3 1 Waters
4 1 Ecosystems
5 1 Natural
6 2 Air
7 2 Noise
8 2 Trac
9 2 Hazards
10 3 Housing
11 3 Land use
12 3 Physical
13 3 Cultural
14 3 Public
Table 2
Project options
ID Name Description
1 Solution I Maximum use of existing corridor
2 Solution II AsproneriKammena Vourla bypass
3 Solution 3B Approved solution (with bridges and tunnels)
4 Solution III Midland solution through LogoAgnanti link
298 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
Also, a scale for rating the impacts can be specied (Table 6), using any range of numerical
values.
6.2. Step 2: criteria weights
The relative importance of criteria is applied as a weight vector during the rating stage of the
method. The required weights of criteria are determined with Yagers method (Yager, 1977).
According to this method, all criteria are compared to each other making a matrix of values for
each comparison pair. The value of a cell i, j (row i, column j) of the matrix is the ratio Wi/Wj,
namely the relative weight of criterion i over criterion j. The weight vector of all criteria results as
the main eigenvector of the comparison matrix. To apply the weights the vector is normalized in
the interval [0, 1].
For the case study, the relative weights of criteria were determined with the application of the
Delphi method. Weights were supplied independently by three `experts': the designer of the road
project, a geologisthydrogeologist, and a city plannerenvironmentalist. Thus, three dierent
Table 5
Importance of criteria
ID Verbal description Value
1 Least 1
2 Small 2
3 Weak 3
4 Less 4
5 Moderate 5
6 More 6
7 Strong 7
8 Great 8
9 Most 9
Table 6
Impacts rating
Verbal description Value
Strong negative impact )4
Large negative impact )3
Medium negative impact )2
Small negative impact )1
No impact 0
Small positive impact 1
Medium positive impact 2
Large positive impact 3
Strong positive impact 4
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 299
`Viewpoints' were introduced. The results shown in the following are based on the `average' view
of the above three experts.
6.3. Step 3: rating and ranking the alternatives
The MCA tool of EFECT generates all possible combinations of project optionscriteriare-
gionstimeframes. The user rates them accordingly. The criteria values are combined with a
summation function, which uses the score of each criterion multiplied by the respective weight of
the weight vector.
To derive an overall score, which includes the three sections of the project ((S
1
) Agios Kon-
stantinos, (S
2
) Asproneri, (S
3
) Kamena Vourla), the relative weights for the three sections were
calculated with Yagers method, using the following assumptions: (i) the natural and man-made
environment of each area was considered, (ii) equal weight was given in comparing the two (ratio
1), whilst, (iii) in comparing the man-made environment for each section weights were propor-
tional to the respective population of each area.
The total score of each alternative option results on the same articial scale [)4, 4]. Negative
scores correspond to negative impacts on the environment, whereas positive scores correspond to
positive impacts.
Hence, ranking of the alternative options is done considering as better solutions the ones with
higher scores. Fig. 5 presents in graphical form the results of the applications of the EFECT
software tool, for the long-term impacts of the alternative solutions.
6.4. Step 4: Exploring the results
For the purposes of the case studies, step 4 considered only the time dimension. The application
of EFECT has provided the following results:
The best solution for the overall project of Agios KonstantinosKamena Vourla Bypass is op-
tion II, namely the midland solution behind Knimida Mountain. This solution exceeds all oth-
er, both in the mid-term and long-term. It is the only solution with positive impacts in the mid-
term and even greater benets in the long-term.
As regards the remaining solutions, options I and II are better than option 3B, as they have
smaller negative impacts to the environment. Between options I and II, option II is preferable,
due to less impacts to the anthropogenic environment. In the long-term, both solutions have
important positive impacts to the area.
Fig. 5. Long-term results for the study.
300 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
Solution 3B, in the mid-term, has the largest negative impacts to the environment, which in the
long-term become even larger, as this option leaves no space for growth when trac volumes
approach the capacity of the road section. This results in reloading the existing NR with the
associated problems to the communities of the area.
Therefore it is concluded that option 3B, which was the preferred solution at the time, in reality is
a choice inferior to options I, II and III, considering spatial impacts and long-term eects from the
environmental point of view. This is due to the fact that these solutions have smaller environ-
mental impacts, and in the long-term resolve the trac and urban problems of the coastal zone.
The application of EFECT suggested the adoption of one of the options I, II, or II, and preferably
of option III, as being the most appropriate from the environmental point of view, especially in
the long-run.
7. Conclusions
The paper presented EFECT, a comprehensive, generalized methodological framework for
environmental assessment of transport initiatives, with the following features, some of which
exhibit unique aspects:
the combined use of MCA and CBA methods,
the simplicity in the use and understanding, based on an additive function, and a core ap-
proach,
the use of established techniques for determination of weights, index functions and ranking,
the use of a hierarchy of criteria, which can be expanded or collapsed accordingly, based on the
needs of the evaluation,
the explicit incorporation of the spatial distribution of impacts through the use of regional
weights,
the incorporation of the time dimension in the analysis allowing to consider dynamic eects,
the handling of uncertainty either through sensitivity analysis or with the use of fuzzy numbers.
Based on the EFECT methodological framework, software for the implementation of EFECT
was developed combining the above-presented tools with a GIS tool. The presented application in
a case study in Greece has demonstrated its advantages over other conventional approaches,
especially in the comparison of spatial impacts and in the assessment of impact evolution over
time.
It is an easy to use framework, friendly to the user and very exible. It allows to include any
number and type of criteria, and to compare any number of options over space and time. The
software tool presented covers the core approach of the EFECT, plus the sensitivity testing and
time assessment capabilities. It can be easily extended to include fuzzy set evaluation and other
forms of scoring functions. However, the summation function used in the core approach provides
better intuitive understanding and it is easier to explain to DMs.
The capability of EFECT to look into dierent geographical areas with a dierent perspective
and examine dierent time-horizons has been proven an additional advantage in the evaluation of
transport projects, and especially from the environmental point of view.
Last, but not least, EFECT is a modular system that can be used both for strategic and detailed
evaluation of transport initiatives for all modes and for all kinds of transportation networks. It
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 301
can provide an individual or the public authorities (national, European Commission) with the
appropriate methodological framework and tools regarding the Environmental Assessment. It is
noted that due to its exibility and open structure, the framework can be expanded to include
other aspects of the evaluation of transport projects or initiatives.
Acknowledgements
The present paper is based on research carried out partly with European Commission funding
under the IV Framework Programme for Research and Development, and in particular regarding
the EUNET and BRIDGES research projects.
References
Abelson, P., 1979. Cost-Benet Analysis and Environmental Problems. Saxon House, London.
Banister, D., Button, K. (Eds.), 1993. Transport Environment in Sustainable Development. Routledge, London.
Banister, D. (Ed.), 1998. Transport Policy and Environment. Routledge, London.
Beuthe, M., Pearman, A., Tsamboulas, D., Watson, S., 1998. Prioritising trans-European network transport initiatives.
In: Proceedings of the Eighth World Conference on Transport Research, Antwerp.
Button, K. (Ed.), 1993. Transport, Environment and Economic Policy. Edward Elgar, London.
Danning, D., MacKnight, S., 1990. Environmental considerations for ports and harbour developments. World Bank
Technical Paper in Transport and Environment Series no. 129, USA.
European Commission, 1993. Towards sustainability: a European community programme of policy and action in
relation to the environment and sustainable development. Report by the European Commission, Brussels.
European Commission, 1996. State of the art on strategic environmental assessment for transport infrastructure.
Report by the European Commission, Brussels.
Freeman, A.M., 1979. The Benets of Environmental Improvement. The John Hopkins University Press, Baltimore.
Mikroudis, G., 1996. Fuzzy indices in environmental decision-making. In: Third International Symposium on
Environmental Geotechnology, San Diego, USA.
Moavenzadeh, F., 1994. Global Construction and the Environment. Wiley, New York.
Munda, G., 1995. Multi-criteria Evaluation in a Fuzzy Environment. Physica-Verlag, Heidelberg, Germany.
Nickols, R., Heyman, E., 1982. Evaluation of environmental assessment methods. No. WR1, ASCE, vol. 108,
Washington, USA.
Nijkamp, P., Blaas, E., 1993. Impact Assessment and Evaluation in Transportation Planning. Kluwer Academic
Publishers, Dordrecht.
Roson, R., Small, K. (Eds.), 1998. Environment and Transport in Economic Modelling. Kluwer Academic Publishers,
Dordrecht.
Saaty, T.L., 1980. The Analytic Hierarchy Process. McGraw-Hill, New York.
Tsamboulas, D., Mikroudis, G., 1996. Environmental and social impacts of inter-urban transport projects. In:
Euroconference on Environment and Innovation, Vienna.
Tsamboulas, D. (Ed.), 1998. Innovations in multi-criteria decision analysis, Deliverable D1. EUNET Research Project
European Commission, DG Transport, Brussels.
Tsamboulas, D., Mikroudis, G., 1999. DSS, Deliverable D7 BRIDGES Research Project. European Commission, DG
Transport, Brussels.
302 D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303
Verhoef, E., 1994. External Eects and social costs of road transport. Transportation Research A 28, 273287.
Yager, R.R., 1977. Multiple objective decision-making using fuzzy sets. International Journal of ManMachine Studies
9, 375382.
D. Tsamboulas, G. Mikroudis / Transportation Research Part D 5 (2000) 283303 303

You might also like