A Hybrid Approach for Evaluating

Environmental Impacts for Urban

Transportation Mode Sharing

H. Omrani*,**** — A. Awasthi** — L. Ion***— P. Trigano****

* CEPS/INSTEAD, Geode department, Luxemburg

** Concordia University, Montreal
*** EIGSI, 26 rue de Vaux de Foletier, 17041, La Rochelle cedex 1, France
**** UMR CNRS 6599 Heudiasyc, Université de Technologie de Compiègne,
BP 20529, 60205 Compiègne Cedex, France

ABSTRACT. The current paper presents an AHP based approach for evaluating sustainable
transport solution measures like car-sharing, park and ride, access control zones etc. In the
first stage, we identify the indicators (criteria) for evaluating the transportation solution
measure. These indicators (criteria) can be divided into several sub-indicators (sub-criteria).
In the second stage, we allot weights to the indicators and sub-indicators using AHP. The
values for the sub-indicators are measured using multiple sources and multiplied with their
weights in order to compute state change variable values that form part of the city state
equation. The respective values of the state change variables in the city state equation are
then used to evaluate the efficiency of the transportation solution measure. Finally, we
illustrate our approach by giving an example of car-sharing and measuring its impact on city
environmental conditions.
RÉSUMÉ. Cet article présente une approche hybride, basée sur la méthode AHP, pour
l’évaluation des projets de transport comme le parc relais, les voitures partagées, le chemin
d’accès contrôlé, etc. Dans un premier temps, nous identifions la liste des critères pertinents.
Ces critères peuvent être subdivisés en plusieurs sous-critères. Dans un second temps, nous
estimons le poids des critères et des sous-critères en utilisant la méthode AHP. Les valeurs
des sous-critères sont mesurées à l’aide de plusieurs sources d’information. Chaque valeur
est ensuite multipliée par son propre poids pour estimer l’état de changement des variables et
l’état de la ville. Les valeurs de l’état de changement des variables sont utilisées pour évaluer
la performance des projets. En conclusion, nous illustrons notre approche en donnant un
exemple de voitures-partagées et en mesurant son impact environnemental.
KEYWORDS: AHP, Environmental Impacts Assessment, Car-sharing.
MOTS-CLÉS: AHP, évaluation, impacts environnementaux, voitures partagées.

DOI:10.3166/JDS.18.185-201 © 2009 Lavoisier, Paris

Journal of Decision Systems. Volume 18 – No. 2/2009, pages 185 to 201

186 Journal of Decision Systems. Volume 18 – No. 2/2009

1. Introduction

Sustainable mobility solutions are the demand of modern times. The need has
arisen out of the ever-increasing air and noise pollution caused by continuously
increasing number of vehicles on roads. Emphasis is on finding sustainable mobility
solutions that minimize the number of private vehicle movements inside the cities
thereby reducing vehicle emissions and traffic congestion on urban network.
Examples of such kind of solutions are car-sharing, park-and-ride, multimodal
transport solutions and non-motorized use of transport.
The most commonly approaches reported for evaluation of sustainable mobility
solutions in literature are Life cycle analysis (Guine, 2002), cost-benefit analysis and
cost-effectiveness analysis (Kunreuther, 2003), environmental impact assessment,
environmental indicator, multi-criteria Decision analysis (Roy, 1985), AHP
(Delgado, 1992), etc.
The traditional approaches mentioned above lack the study of complexity of the
systems with their environmental, social, economic and cognitive dimensions. The
reason is the difficulty posed in treating partial and subjective judgments
(satisfaction degree of the users, spatial accessibility, security, quality of service
etc.) and heterogeneous data (cost, traffic and congestion levels, noise etc.).
Furthermore, uncertainty and vagueness common in human knowledge cannot be
properly catered for in such methods aforementioned.
Recently, others methods combining MCDA (Multi Criteria Decision Analysis)
and AI (Artificial Intelligence) have been explored to develop enhanced
methodologies for knowledge based decision support. By combining MCDA with
fuzzy logic theory, new methods have been developed like, the Fuzzy AHP
(Buckley, 1985) (Laarhoven, 1983). In addition, evidence theory with MCDA
methods has also been proposed. (Beynon, 2002) proposed AHP with Dempster-
Shafer (DS) Theory (Dempster, 1968; Shafer, 1976). Evidential reasoning (ER)
approaches for dealing with complex decision problems in management are reported
in (Beynon, 2002; Xu, 2005; Yang, 2005). The usage of evidential reasoning has
been reported in several applications (Denoeux, 2005), but not extensively in
evaluation of environmental impact of transport. In our knowledge, only one
research team has recently published papers for environmental analysis which use
the evidence theory and multi-criteria analysis (Xu, 2005; Yang, 2005). But in this
method, only experts’ opinions are taken into account. The final solution can
possibly be to couple or to adapt different approaches.
The current article proposes an AHP based approach for environmental impact
assessment of sustainable transportation solutions. The various steps of this
approach are:
1) Selecting the indicators for evaluating the transportation measure. AHP is
used to rate the various indicators and sub-indicators.
 An AHP/DS Based Approach 187

2) Data collection from different sources for measuring the indicator values. DS
theory is used for multi source data fusion.
3) Estimating the indicator values before and after implementation of the
measure. Observing the difference in terms of change variables for indicator values
that can increase, decrease or report no change.
4) Uniting change variables associated with indicator values (also called state
change variables) to generate “IF-THEN” rules for evaluating the effectiveness of
the measure for the city.
In the first step, we select the indicators and sub-indicators for evaluating the
environmental impacts. AHP will be used for computing weights for the various
criteria and sub-criteria using a 1-9 pair wise rating scale.
In the second step entitled ‘data fusion’, we collect information from multiple
and possibly heterogeneous sources and combine them in order to get a more
meaningful result. Dempster-Shafer Theory is used for data fusion. The main
processing stages in data fusion are:
– The “data collection phase” where various sensors monitor, detect and report
the values of the indicators/sub-indicators.
– The “data alignment & association phase” where the collected data is aligned
in time, space or measurement units.
– The “state estimation phase” where based on a model of the system behaviour
and the knowledge acquired by the sensors a data fusion algorithm estimates the
value of the indicators/sub-indicators.
In the third step, we evaluate the indicator/sub-indicator values before and after
implementation of the transportation solution measure. The difference between these
indicator values can be positive, negative or report no change. The indicators (also
called state change variables) are allotted a class (1, 2 or 3) depending upon their
sign (+, - or 0). These classes will be used for IF-THEN rule generation for city state
assessment.
In the last step, we unite different change variables using “IF THEN” rules to
assess the state of the city which can either ameliorate (very positive, positive),
deteriorate (very negative, negative) or report no change (indifferent) with respect to
the implemented transportation solution measure.
The rest of the paper is divided as follows. In Section 2, we present the
methodology for assessing the environmental impact of sustainable transportation
solution using AHP. In Section 3, we present a case study of car-sharing and finally
we present the conclusions in Section 4.
188 Journal of Decision Systems. Volume 18 – No. 2/2009

2. Methodology

Our approach for assessing the environmental impact of sustainable

transportation solution stations inside the city is based on AHP multi-criteria
decision making (Belton, 1986; Saaty, 1991; Saaty, 2000; French, 1988; Perez,
1995; Zeleny, 1990; Dyer, 1990; Watson and Freeling, 1989). The different steps of
AHP are:
1) Defining the problem and determining its goal.
2) Structuring the hierarchy from the top (the objectives) through the
intermediate levels (criteria) to the lowest level (alternatives).
3) Constructing a set of pair-wise comparison matrices (size n × n ) for each of
the lower levels with one matrix for each element in the level immediately above by
using the relative scale measurement shown in Table 1. The pair-wise comparisons
are done in terms of which element dominates the other.
4) There are n(n − 1) / 2 judgments required to develop the set of matrices in
step 3. Reciprocals are automatically assigned in each pair-wise comparison.
5) Hierarchical synthesis is now used to weight the eigenvectors by the weights
of the criteria and the sum is taken over all weighted eigenvector entries
corresponding to those in the next lower level of the hierarchy.
6) Having made all the pair-wise comparisons, the consistency is determined by
using the eigenvalue λmax to calculate the consistency index CI where CI =
(λ max − n) /( n − 1) where n is the matrix size. Judgment consistency can be checked
by seeing the value of consistency ratio CR for the appropriate matrix value in
Table 2. If CR ≤ 0.1, the judgement matrix is acceptable otherwise it is considered
inconsistent. To obtain a consistent matrix, judgments should be reviewed and
improved.
7) Steps 3-6 are performed for all levels in the hierarchy.
We apply the various steps of AHP for environmental impact assessment of
transportation solution measures. The strength of AHP is that it organizes tangible
and intangible factors in a systematic way, and provides a structured yet relatively
simple solution to the decision-making problems (Saaty, 1980; Saaty, 1982; Saaty,
1990; Saaty and Kearns, 1991; Saaty and Thomas, 2000). In addition, by breaking a
problem down in a logical fashion from the large, descending in gradual steps, to the
smaller and smaller, one is able to connect, through simple paired comparison
judgments, the small to the large.
The methodology for assessing the environmental impact of sustainable transport
solutions can be divided into following steps:
1) Identification of indicators for sustainable transport solution measure. These
indicators are also referred to as state change variables since they are used to judge
the state of the city.
 An AHP/DS Based Approach 189

2) Estimation of weights for the indicators and the sub-indicators. These weights
are allotted a scale of 1-9 by the experts. The indicators and sub-indicators values
are measured using sensors, traffic models, survey counts etc.
3) Development of a state change equation for the city. The measured state
change variables are used for computing the state of the city.
4) Development of rules for city state estimation. These rules are IF-THEN type
and obtained using different combination of classes for city state change variables.

Table 1. Pairwise comparison scale for AHP Preferences (Saaty, 1990)

Numerical Rating Verbal judgement of preferences

1 Equally preferred
3 Moderately preferred
5 Strongly preferred
7 Very strongly preferred
9 Extremely preferred
Intermediate values between the two adjacent
2,4,6,8
judgements
When activity i compared to j is assigned one of the
Reciprocals above numbers, then activity j compared to i is
assigned its reciprocal

Table 2. Average random consistency (RI) (Saaty, 1990)

Size of matrix 1 2 3 4 5 6 7 8 9 10

Random consistency 0 0 0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.49

2.1. Identification of indicators

The different indicators used for assessment of environmental impact were

identified by discussion with car-sharing experts, literature review and on-hand
experience with LISELEC car-sharing system in La Rochelle. Five categories of
indicators are proposed (Figure 1). These can be categorized into economy, energy,
environment, society and transport. The first indicator is the economy. This indicator
can be further divided into costs and benefits. The operating costs and revenues will
be used for measuring this indicator. These indicators are quantitative in nature and
will be obtained from reports.
190 Journal of Decision Systems. Volume 18 – No. 2/2009

Figure 1. Indicator categories used for impact assessment

The second indicator is the energy. This indicator can be represented in terms of
energy consumption. Fuel consumption will be used to measure this indicator. This
indicator is quantitative in nature.
The third indicator is environment. This indicator can be represented via the sub-
indicators Pollution and Noise. The sub-indicator pollution will be measured in
terms of Air Quality and Emissions. The sub-indicator air quality and emissions is
quantitative in nature and will be measured via models and sensors. The visual
pollution indicator is qualitative in nature and will be measured by questionnaire
surveys.
The fourth indicator is society. This indicator can be divided into acceptance,
accessibility, security, quality of service and safety. The acceptance indicator can be
measured in terms of awareness and acceptance. The accessibility indicator can be
 An AHP/DS Based Approach 191

measured in terms of spatial accessibility and acceptance. The security indicator can
be measured in terms of security. The quality of service can be measured in terms of
service reliability and quality of service. The safety indicator can be measured in
terms of transport safety. All these sub-indicators are qualitative in nature and
measured using experts opinions and questionnaire surveys.
The transport indicator is representative of the transport system. This can be
measured in terms of congestion levels, traffic levels, freight movements, modal
split and vehicle occupancy. This indicator is quantitative in nature and will be
obtained from traffic counts, models and experts opinions.

2.2. Data fusion of indicators and sub-indicators

The indicators and sub indicators used in our model are qualitative or
quantitative in nature. The qualitative variables are obtained using models, surveys
and experts opinions whereas the quantitative variables will be obtained using
sensors or traffic counts. The data collected from different sources is heterogeneous
in nature therefore; it has to brought on the same scale. This multi source data fusion
process can be described as follows:

Figure 2. Multi source data fusion process

The sources for data collection process are models, sensors and countings. Data
was collected for the indicator Air pollution (CO emissions, NOx emissions, CO2
emissions and Particulate emissions) and Noise. The data alignment and association
phase consists of the following steps:
– Collection of data from multiple sensors.
– Mass assignment to criteria values obtained from the sensors. The mass
function or bpa for the criteria value are the normalized form of the observed values
(on a scale of 0-1).
– Once bpa’s have been alloted to sensors, then the combination is calculated
from the aggregation of the bpa’s using DS theory of evidence.
192 Journal of Decision Systems. Volume 18 – No. 2/2009

The initialization methods for the mass function in Belief Theory are various and
depend on the field application like classification, identification, detection, decision
making etc. we present in (Omrani et al., 2006a), a mean for calculating mass
structure (theory defined in (Dyer, 1990)) for each information source.
In the proposed approach, various means are presented according to the nature of
the information source. In addition, each parameter can be evaluated by a specific
information source (experts, sensors, and questionnaire) and each one has its own
characteristic. The mass assignment is described in details in (Omrani et al., 2006a;
2006b) according to each information source.
The different methods of mass assignment will be applied later in the case study.
After the presentation of the proposed approach for decision making under
uncertainty, and its theoretical framework, we present next, more in detail, its
evaluation steps.
After the mass assignment according to each information source, in the belief
theory, when a decision must be taken, beliefs are transformed into a probability
measure denoted BetP (Smets and Kennes, 1994). The function building this
probability is called the “pignistic” transformation and it is defined as:

A∩ B m(B )
BetP ( A) = ∑ × , ∀A ⊆ Ω
B⊆Ω B 1 − m(∅ )

where |B| denotes the numbers of elements in the set B.

Then, we can determine a global utility related to each criterion by measure,
taking into account the utility given for each evaluation level schematically
represented by Figure 3.
It is important to note, that the utility (noted by u, Figure 3) given for the
assessment levels (H, e.g. small, medium and large) can be not only crisp as shown
in Figure 3, but also fuzzy or expressed by interval.
Then, a global utility related to each criterion, is computed. The crisp utility
based on evidential reasoning method is computed as:

p
ui = ∑ u (H k ) × BetP(H k )
k =1

where: u (H k ) represents the utility of an evaluation level H k , u (H k +1 ) ≥ u (H k )

if H k +1 is preferred to H k , and BetP(H k ) represents the “pignistic” probability
related to each evaluation level H k .
In order to compare several implementation phases (e.g. initial evaluation, post
evaluation etc.), a comparison by criteria is done which allow us to underline if there
is a benefit or decrease.
 An AHP/DS Based Approach 193

Figure 3. Utility Functions (linear and exponential)

2.3. Scoring and weighting the indicators and sub-indicators

This step performs the scoring and weighting of the different indicators (criteria)
and sub-indicators (sub-criteria). The indicators are also referred to as state change
variables. The pair-wise comparison for the indicators / sub-indicators is done using
a 1-9 ratio scale (Table 3). The ratio reflects the contribution of the indicators/ sub-
indicators towards the overall objective. The scores are obtained using a pairwise
comparison matrix.

Let us denote the indicators by C1 , C2 .. C N where N represents the total

number of indicators. Table 3 presents the N × N pairwise comparison matrix for
allotting the ratings to the different indicators. The ratings for the different elements
of the matrix are done by pairwise comparison of row vs column elements. Let us
denote the row elements by i and column elements by j where i =1,2,..N and
j =1,2,..N. The rating for element i vs j is denoted by rij . If the ranking of indicator i
with respect to indicator j is rij , then the ranking of indicator j with respect to i
is 1 / rij . The ranking for the diagonal elements rii = 1.

It can be seen in Table 3 that all the diagonal elements have a rating 1 because
each element (criteria) is being compared with itself. All the elements on the upper
half diagonal of the matrix are inverse of the elements in the lower half diagonal.
Likewise, ratings for the sub-indicators is done. These variables will be used
later in the state equation of the city for determining the impact of implemented
transportation solution measure. Once the individual weights for the indicators/sub-
indicators have been calculated, it is necessary to check if the ratings contained in
the matrices are consistent. This is done by measuring the Consistency Ratio
(Watson and Freeling, 1989). As a rule of thumb, if the consistency ratio CR ≤ 0.1 ,
then the ratings will be accepted otherwise a re-voting for the comparison-matrix
will be performed.
194 Journal of Decision Systems. Volume 18 – No. 2/2009

Table 3. Pair wise rating of indicators (criteria)

C1 C2 C3 - - CN
C1 1 r12 = 1 / r21 r13 = 1 / r31 - - r1N = 1 / rN 1
C2 r21 1 r23 = 1 / r32 - - r2 N = 1 / rN 2
C3 r31 r32 1 - - r3 N = 1 / rN 3
- - - - 1 - r4 N = 1 / rN 4
- - - - - 1 r5 N = 1 / rN 5
CN rN 1 rN 2 rN 3 - - 1

Table 4. Normalized pair wise rating of indicators (criteria)

Criteria Geometric Mean Normalized score

GM C1
C1 GM C1 = (1 × r11 × r12 × r13 × .... × r1N )1/ N NS C1 =
(GM C1 + GM C2 + GM C3 ...... + GM CN )

GM C2
C2 GM C2 = (r21 × 1 × r23 × r24 × .... × r2 N )1/ N NS C2 =
(GM C1 + GM C2 + GM C3 ...... + GM C N )

- - -
- - -
GM C N
CN GM CN = (rN 1 × rN 2 × rN 3 × ..... × 1)1 / N NS C N =
(GM C1 + GM C2 + GM C3 ...... + GM C N )

After verification of the consistency ratio, we will calculate the weights or the
normalized scores for the different indicators and the sub-indicators. The normalized
scores will be computed using geometric means because the geometric mean for a
series (For e.g. 1,2,…N ) is less affected by extreme values than the arithmetic
mean. Besides, it is useful as a measure of central tendency for some positively
skewed distributions. For a series containing n elements, the geometric mean is
given by the nth root of the product of n scores and the normalised criterion weight is
given by the ratio of the geometric mean divided by the sum of the geometric means
of all the elements of the series. For the indicator C p , p = 1,2,..N, the geometric
1/ N
mean ( GM C ) is given by  r 
N
and the normalized score ( NS C ) is given by
p ∏ pj   p

 j =1 
 An AHP/DS Based Approach 195

GM C p . Likewise, the normalized scores for the N sub-indicators will be

N

∑ GM
p =1
Cp

computed.

2.4. State estimation

Once the sub-indicator weights have been computed and their values obtained by
measurement, we compute the state of the city. The ‘State estimation’ phase consists
of the following steps:
– Computing indicator values.
– Computing indicator values using sub-indicator values, sub-indicator weights
and indicator weights.
– Creating the state equation of the city using indicator weights and indicator
values.
– Finding difference between the state equations at two instants of time to
compute the state change variables.
Let us denote the state change variables (indicators) at time instant tn by C1(tn),
C2(tn), C3(tn), …..CN(tn ) and their respective weights by wC1, wC2, wC3,….wCN. Let
us denote the state equation by S(tn ).
The state change equation at time tn can be represented as follows:

S(tn) = wC1* U1(tn) + wC2*U2(tn )+….+wCN* UN(tn)

with: wCi is the criterion weight and Ui(tn) is the utility given for the criterion Ci
according the equation above mentioned.
To determine the difference in state change variables, we consider the state
change equation at two instants of time say tn and tn-1. The difference between the
two state equations is represented as follows:

S(tn) - S(tn-1) = wC1* [U1(tn) - U1(tn-1) ]+ wC2*[U2(tn) – U2(tn-1) ]+….

+wCN*[UN(tn) – UN(tn-1) ]
Note that the state change variables of the state equation are not additive if they
are not in the same unit. For example, criteria C1 can be air pollution and C2 can be
noise. Their values are numeric but their units of measurement are different. Hence,
we cannot sum both of them together and their values will be considered
individually (piecewise) for consideration in the overall impact assessment.
Another consideration is that in few cases the indicator values may not be
measured directly but obtained by multiplication of sub-indicator values and
weights. Therefore, only criteria weights will be relevant.
196 Journal of Decision Systems. Volume 18 – No. 2/2009

In the above equation, each indicator will be associated with a change variable.
Let us denote the change variable for criteria C1 by ∆C1 , C2 by ∆C2, etc. According
to the sign of the state change variables, we will allot three classes: class 1, class 2
and class 3.

Table 5. State change variables (classes)

State Change variable Class

Sign of change variable > 0 1

Sign of change variables = 0 2

Sign of change variable < 0 3

Using these state change variables, we develop the rules for assessment of the
state of the city with respect to the sustainable transportation solution measure.
Table 6 shows the rules for environmental impact assessment.

Table 6. Evaluation rules

Evaluation Rule for the implemented transport Evaluation of the city

Rule No.
measure for the measure
IF (all change variables in class 1), THEN the change
--
1 brought by the implemented transportation measure in
(Very Negative)
the city is very negative.
IF (all change variables in class 2), THEN no change
0
2 has been brought by the implemented transportation
(Indifferent)
measure in the city.
IF (all change variables in class 3), THEN the change ++
3 brought by the implemented transportation measure in
the city is very positive. (Very Positive)

IF ((change variables in class 1, 2 or 3) AND (number

of change variables in class 2 and 3) <= (change
-
4 variables in class 1)), THEN the change brought by the
(Negative)
implemented transportation measure in the city is
Negative.
IF ((change variables in class 1, 2 or 3) AND (number
of change variables in class 2 and 3) > (change
+
5 variables in class 1)), THEN the change brought by the
(Positive)
implemented transportation measure in the city is
Negative.
 An AHP/DS Based Approach 197

It can be seen, that five evaluation classes for the city with respect to the
implemented transportation solution measure are possible.
Rule 1 states that if all the change variables are in class 1 i.e. report increase with
respect to the implemented transport solution measure, then the impact on the city is
very negative and such measure should be avoided for implementation.
Rule 2 states that if all the change variables are in class 2 i.e. report no increase
with respect to the implemented transport solution measure, then the impact on the
city is indifferent and such kind of transportation solution measure should be
avoided for implementation.
Rule 3 states that if all the change variables are in class 3 i.e. report decrease
with respect to the implemented transport solution measure, then the impact on the
city is very positive and such measure should definitely be considered for
implementation.
Rule 4 states that if the change variables belong to either class 1,2 or 3 and if the
number of change variables in class 2 and 3 (reporting decrease) are lesser than the
number of change variables in class 1 (reporting increase), then the impact on the
city is negative and such measure should not be favourably considered for
implementation.
Rule 5 states that if the change variables belong to either class 1,2 or 3 and if the
number of change variables in class 2 and 3 (reporting decrease) are higher than the
number of change variables in class 1 (reporting increase), then the impact on the
city is positive and such measure should be favourably considered for
implementation.
Note: The change variables may not report the same sense. For example,
lowering air pollution and noise are favourable with respect to the implemented
transportation solution measure whereas societal awareness should notice
improvement.

2.5. Sensitivity analysis

Sensitivity analysis addresses the question, `How sensitive is the overall decision
to small changes in the individual weights assigned during the pair-wise comparison
process?’ This question can be answered by varying slightly the values of the
weights and observing the effects on the decision. This is useful in situations where
uncertainties exist in the definition of the importance of different factors. In our
case, we will conduct sensitivity analysis in order to see the importance of change
variables in the city state equation implicating the high influence of few indicators
over others.
198 Journal of Decision Systems. Volume 18 – No. 2/2009

3. Case study

Car-sharing is an alternative system of car ownership, access and use. The car-
sharing vehicles are disposed for client’s use at different stations (24 hours a day, 7
days a week) just like a private car. The clients reserve the vehicle in advance and
get access to the vehicle at the nearest car-sharing station. The entry to the vehicle
and the car keys are obtained using a smart card and a PIN code. The billing is
affected according to the hours of car use and the mileage driven. The costs and
troubles of vehicle purchase, ownership and maintenance are transferred from the
individual users or clients to the car-sharing organisation (CSO). Figure 4 illustrates
the architecture of a typical CSO.

Figure 4. Elements of a carsharing organisation

The indicators considered for evaluating car-sharing are air pollution and noise.
The air pollution indicator can be further subdivided in to CO2, NOx, CO and
Pollutants. AHP is used to compute weights for the various indicators and sub-
indicators using a 1-9 pair wise rating scale. Let us consider that the weight for the
criteria “air pollution” is 0.5 and the criteria “noise” is 0.5. The criteria ‘air
pollution’ can be divided into four sub-criteria NOx, CO2, CO and particulates.
Assume that the sub-criteria values at time instant t1 for NOx is 0.5, CO2 is 0.23, CO
is 0.34, and particulates is 0.3. The sub-criteria values at time instant t2 for NOx is
0.6, CO2 is 0.3, CO is 0.25, and particulates is 0.25 and noise is 0.5. The criteria
weights for NOx is 0.3, CO2 is 0.4, CO is 0.2, particulates is 0.1 and Noise is 1.
The state equation at time t1 is:
S(t1) = 0.5*[0.3*0.5 + 0.4*0.23 + 0.2*0.34 + 0.1*0.3]+ [1*0.3]
Air pollution Noise
 An AHP/DS Based Approach 199

= 0.5[0.15+.092+0.068+.03] + [0.3]
Air pollution Noise
= [0.17] + [0.3]
Air pollution Noise

The state equation at time t2 is:

S(t2) = 0.5*[0.3*0.6 + 0.4*0.3 + 0.2*0.25 +0.1*0.25] + [1*0.5]
Air pollution Noise

S(t2) = 0.5*[0.18 + 0.12 + 0.05 +0.025] + [0.5]

Air pollution Noise
= [0.375] + [0.5]
Air pollution Noise
The difference between the two state equations is:
S(t2) – S(t1)= [0.205] + [0.2]
Air pollution Noise
Change variable for Air Pollution = 0.205 = +ive (Class: 1)
Change variable for Noise = 0.2 = +ive (Class: 1)
Rule 1 is applicable: IF (all change variables in class 1), THEN the change
brought by the implemented transportation measure in the city is very negative.
We notice a positive change in the state change variables for Air pollution and
Noise. Therefore, rule 1 of Table 6 will be applicable and on the basis of this rule we
can say that implementation of car-sharing in the city is very negative in terms of
reduction of air pollution and noise.
To observe the influence of change in weight of indicators/sub-indicators we
conduct the sensitivity analysis. Let us consider 4 experiments for the above
mentioned example. The details of these four experiments are:
– Experiment 1: Criteria weights obtained by pairwise comparison ratings.
– Experiment 2: Same weight: all criteria are equally important.
– Experiment 3: Criteria weights obtained by direct allocation of weights.
– Experiment 4: Criteria “Air pollution” allocated 60% of the weight.
We can see that for the above example of car-sharing, Rule 1 remains applicable
for all the above four experiments. The reason being that no change in the sign of
the indicator/subindicator values (or the state change variables) in the overall state
equation is observed even if we change the weights.
200 Journal of Decision Systems. Volume 18 – No. 2/2009

4. Conclusions

The current paper presents an Analytical Hierarchy Process (AHP) based

approach for environmental impact assessment of sustainable transport solutions in
the study. The approach was based on AHP and demonstrated using an example of
car-sharing. The indicators used to measure the environmental impacts are
quantitative and qualitative in nature and brought on a common numerical scale
before usage in the state change equation. The state change equation of the city
assesses the state of the city in terms of highly positive, positive, indifferent,
negative or highly negative. Transport solution measures that lead to highly positive,
positive and indifferent city state conditions are considered favourable for
implementation.
The current work considers the environmental impact assessment of a single
transportation measure. Future work will involve assessment of multiple
transportation solution measures at a time and we will show whether the
implementation of multiple solutions at a time leads to overall amelioration in the
city state conditions.

Acknowledgements
The authors gratefully acknowledge European Commission for the financial and
research support for the project SUCCESS (CIVITAS II), Contract number 513785
in La Rochelle, France.

5. References

Belton V., “A comparison of the analytic hierarchy process and a simple multi-attribute value
function”, European Journal of Operational Research, 26, 1986, p. 7-21.
Beynon M.J., “DS/AHP method: a mathematical analysis, including an understanding of
uncertainty”, European Journal of Operational Research, 140, 2002, p. 149-165.
Delgado M., Verdegay J.L, and Vila M.A., “Linguistic Decision Making Models”, Int. J. of
Intelligent System, 7, 1992, p. 479-492.
Buckley J.J., “Fuzzy hierarchical analysis”, Fuzzy Sets and Systems, Vol. 28, No. 1, 1985,
p. 233-247.
Dempster A.P., “A generalisation of Bayesian inference”, Journal of the Royal Statistical
Society, 1968, p. 205-247.
Denoeux T., Application du formalisme des Fonctions de Croyance en fusion d'informations
et en Classification, 2005.
Dyer J.S., “Remarks on the analytical hierarchy process”, Management Science, 3, 1990,
p. 249-58.
French S., Decision theory: an introduction to the mathematics of rationality, Chichester,
Ellis Horwood, 1988.
 An AHP/DS Based Approach 201

Guine J.B., Handbook on Life Cycle Assessment’. An operational guide to the ISO standard,
London, Kluwer Academic, 704, 2002.
Kunreuther H., Grossi P., Seeber N., Smyth A., “A Framework for Evaluating the Cost-
Effectiveness of Mitigation Measures”, Columbia University, Etats-Uni, 2003.
Laarhoven P.J.M., Pedrycz W., “A fuzzy extension of Saaty priority theory”, Fuzzy Sets and
Systems, Vol. 11, No. 3, 1983, p. 229-241.
Omrani H., Ion L., Trigano P., “An Approach for Environmental Impacts Assessment using
Belief Theory”, Proceedings of 3rd IEEE Conference On Intelligent Systems (IEEE
IS’06), London, UK, 4-6 September 2006a.
Omrani H., Ion L., Trigano P., “A New Approach for Environmental Impact Assessment”,
WSEAS (The World Scientific and Engineering Academy and Society), Venice, Italy,
November 20-22, 2006b.
Perez J., “Some comments on Saaty’s AHP”, Management Science, Vol. 41, No. 6, 1995,
p. 1091-1095.
Roy B., Méthodologie multicritre d’aide à la decision, Economica, Paris, or Multicriteria
methodology for decision aiding, Kluwer, Dordrecht, 1985.
Saaty T.L., The analytic hierarchy process, New York, McGraw-Hill, 1980.
Saaty T.L., Decision making for leaders, Belmont, California, Lifetime Learning
Publications, 1982.
Saaty T.L., “How to make a decision: the analytic hierarchy process”, European Journal of
Operational Research, North-Holland, 48, 1990, p. 9-26.
Saaty T.L, Kearns K.P., “Analytical planning: the organization of systems”, The analytic
hierarchy process series, Vol. 4, RWS Publications, Pittsburgh, USA, 1991.
Saaty A., Thomas L., Fundamentals of decision making and priority theory, 2nd ed.
Pittsburgh, PA, RWS Publications, 2000.
Shafer G., A Mathematical Theory of Evidence, Princeton University Press, Princeton, 1976.
Smets P. and Kennes R., “The Transferable belief model”, Artificial Intelligence, Vol. 66,
1990, p. 191-234.
Zeleny M., Multiple criteria decision making, New York, McGraw-Hill, 1994.
Watson S.R, Freeling A.N.S, “Assessing attribute weights”, Omega, 1982, Vol. 10, No. 6,
p. 582-583.
Xu, D.L., Yang J.B. and Wang Y.M., “The ER approach for multiattribute decision analysis
under interval uncertainties”, European Journal of Operational Research, 2005.
Yang J.B., Wang Y.M., Xu D.L. and Chin K.S., “The evidential reasoning approach for
MCDA under both probabilistic and fuzzy uncertainties”, European Journal of
Operational Research, 2005.
Zeleny M., “Multiple criteria decision making”, New York, McGraw-Hill, 1990.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.