You are on page 1of 9

Ecological Indicators 43 (2014) 288–296

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Transport sustainability index: Melbourne case study


Marzieh Reisi, Lu Aye ∗ , Abbas Rajabifard, Tuan Ngo
Renewable Energy and Energy Efficiency Group, Department of Infrastructure Engineering, The University of Melbourne, VIC 3010, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Quantifying the sustainability of urban transport is important as evidenced by a growing number of
Received 4 February 2013 studies to measure sustainability in transportation. This paper reviewed major initiatives reported in
Received in revised form 3 March 2014 the current literature, which dealt with the challenge of measuring transport sustainability using long
Accepted 10 March 2014
lists of indicators. To overcome the issue of using too many indicators for evaluation, this paper develops
a method for obtaining a composite transport sustainability index for Melbourne statistical local areas
Keywords:
(SLAs). Nine sustainability indicators relevant to urban transport which deal with environmental, social
Sustainable urban transport
and economic aspects were selected by assessing and reviewing past research and based on available
Sustainability indicators
Composite index
data for Melbourne. The indicators were integrated to environmental, social, and economic sub-indices
Principle component analysis and then to a composite index, in a way that overcomes the limitations on normalisation, weighting
Factor analysis and aggregation. Indices developed based on the applied method allow comparisons between different
statistical local areas (SLAs) and could help organisations for better understanding of the measures and
activities that influence the sustainability of urban transport, and for identifying appropriate responses
for addressing them in Melbourne.
Crown Copyright © 2014 Published by Elsevier Ltd. All rights reserved.

1. Introduction making because of their hard interpretation, integrating different


indicators into a single index is useful.
One-fifth of the carbon dioxide (CO2 ), one third of chlorofluoro- Aggregating individual indicators into a composite index is a
carbons (CFCs), and half of nitrogen oxides (NOx ) in the atmosphere practical approach for sustainability evaluation (Dur et al., 2010).
related to transport activities (OECD, 2008). Transportation has It measures multi-dimensional aspects of sustainability that can-
significant economic, social and environmental impacts and is an not be captured completely by individual indicators alone (Saisana,
important factor in sustainability. OECD defined sustainable trans- 2011; Zhou et al., 2007). Despite extensive use of composite
port as “transportation that does not endanger public health or indices, there are two conflicting views about them. The oppo-
ecosystems and meets the needs for access” (EA, 1999). nents of composite index believe that composite index is not
Sustainable indicators which was first brought up at the reliable because its construction is subjective (Cherchye et al.,
United Nations Conference on Environment and Development 2007). Moreover, no single index can answer all questions and
(UNCED) held in Rio de Janeiro in 1992, are used as important there is a need for multiple indicators (Jollands et al., 2003). On
tools for measuring different aspects of sustainable development the other hand, some researchers believe that composite indices
(Dobranskyte-Niskota et al., 2007). Indicators facilitate commu- are valuable communication tools because they limit the number
nication among scientists, policymakers, and public because they of presented information and allow for quick and easy comparisons
provide high volume of information in a simple form that are eas- (Freudenberg, 2003). These two ideas are two sides of the coin and
ier to interpret and understand (Alberti, 1996). They break down it can be concluded that indicators aggregation is successful if clear
the complex concept into small and manageable units of infor- assumptions and methodology are used and if the index can be
mation, so they can easily capture different aspects of transport disaggregated to its components (Jollands et al., 2003).
sustainability (Castillo and Pitfield, 2010). Current sustainability Melbourne is one of the Australian largest cities which has a
studies deal with the challenge of measuring transport sustaina- considerable amount of atmospheric emissions related to trans-
bility using long lists of sustainable indicators (Table 1). Since using port (Alford and Whiteman, 2008). It is beneficial to develop a
too many indicators is inappropriate and complex for decision composite transport sustainability index for Melbourne to com-
pare different planning scenarios. This paper aims to develop an
index for Melbourne which considers environmental, social and
∗ Corresponding author. Tel.: +61 3834 46879; fax: +61 3834 46215. economic performance. Although large numbers of attempts were
E-mail address: lua@unimelb.edu.au (L. Aye). made to identify sustainable transport indicators, limited number

http://dx.doi.org/10.1016/j.ecolind.2014.03.004
1470-160X/Crown Copyright © 2014 Published by Elsevier Ltd. All rights reserved.
M. Reisi et al. / Ecological Indicators 43 (2014) 288–296 289

Table 1
Sustainable transport indicators in previous studies.

Aspect Indicators Determinants

GHG emission • Vehicle kilometre travelled by car


• Passenger kilometre travelled by public transport
Environmental • Mode share
Other air pollutants • Vehicle kilometre travelled
• Passenger kilometre travelled by public transport
• Mode share
Energy use • Vehicle kilometre travelled
• Passenger kilometre travelled by public transport
• Mode share
Population exposed to noise • Traffic volume
Land consumption for transport • Land use mix
• Length of railways and main road
• Length of cycling and walking pass

Fatality and injuries of traffic accident • Vehicle ownership


per capita
Social
Accessibility to facilities and public • Railway and main road length
transport • Proportion of residents with public transit services within 500 m
Satisfaction of citizens and variety and • Quality of transport for disadvantaged, disabled, children, non driver
quality of transport options • Quality of pedestrian and bicycle environment
Fatality and injuries resulted from air –
pollution

Household expenditure allocated to • Cost of parking


Economic transport Fuel price
• Commute costs
Accident cost –
Transport emission costs • VKT
• Modal split

of studies aggregate different indicators into a single index. Zito 2.1. Melbourne – the study area
and Salvo (2011) and Haghshenas and Vaziri (2012) in develop-
ing transport sustainability index considered environmental, social, The city of Melbourne is located on the southeast coast of
and economic aspects. In these studies, various indicators were Australia, on the northern edge of Port Phillip Bay. The study area
given equal importance, which may not be the case. Moreover, contained 79 statistical local areas (SLAs) identified by the Aus-
these studies use statistical data and survey results for indicator tralian Bureau of Statistics (ABS). The topography and climate make
quantifications. There is a lack of studies which develop models Melbourne an area of high air pollution potential (EPA, 2000). As
for indicator quantifications and develop environmental, social, aviation is a small overall contributor to local air quality impacts
economic and composite indices for transport sustainability using (Waitz et al., 2004), therefore it was not considered in this study. In
different weights for various indicators. To fill these knowledge Melbourne, 51 billion kilometres are travelled annually by passen-
gaps and meet the research aim, this study attempts to develop gers, with 75.4% of those are by cars, 9.1% are by public transport,
some models to quantify environmental, social, and economic 13.1% are by walking and 2% are by cycling (Alford and Whiteman,
indicators, and develop transportation environmental index (IE ),
transportation social index (IS ), transportation economic index
(IC ) and finally composite transport sustainability index (ICST ) for
Select prominent transport indicators
Melbourne in 2006 using appropriate normalisation, weighting
and aggregation methods. Selected environmental, social, and eco-
nomic indicators in this study were the final impacts of transport,
which were quantified using transport determinants such as vehi-
Quantify transport indicators
cle kilometre travelled (VKT) and passenger kilometre travelled
(PKT). The most appropriate method was selected for weighting
which overcomes the limitations of the methods used in the previ-
ous sustainability studies. Explanations for choices of each method
Normalise the indicators
used in this study are provided in the following sections.

2. Method
Weight the indicators (using PCA/FA)
Composite indices are very essential for decision making since
they limit the number of presented information and allow for quick
and easy comparisons (Freudenberg, 2003). The process of calculat-
ing the composite index in this study was divided into several parts. Calculate sub-indices
At first, prominent indicators were selected in environmental,
social and economic aspects. Then indicators with different mea-
surement units were normalised. Weighting selected indicators,
relevant indicators were first aggregated into three sub-indices Combine sub-indices into one single index
and finally integrated to a composite transport sustainability index
(Fig. 1). Fig. 1. Procedure for calculating composite transport sustainability index (ICST ).
290 M. Reisi et al. / Ecological Indicators 43 (2014) 288–296

Table 2
Criteria for indicators selection.

Criteria Description Reference

Relevance Each indicator must properly hold the definition of transport Dur et al. (2010)
sustainability.
Maturity The indicator system should contain environmental, social and Dur et al. (2010), Li et al. (2009), Spiekermann and
economic aspects of transport. Wegener (2004), Zito and Salvo (2011)
Data availability Needed data must be available easily and at a reasonable cost. Haghshenas and Vaziri (2012); Zito and Salvo (2011)
Quantifiable Indicators must be quantifiable. Li et al. (2009)
Understandable To facilitate debates among experts and users, simplicity is important. Haghshenas and Vaziri (2012); Zito and Salvo (2011)
Independent Indicators should be independent of each other. Haghshenas and Vaziri (2012); Li et al. (2009)
Sensitivity Indicators should be sensitive to any changes in the system to reflect Haghshenas and Vaziri (2012); Li et al. (2009)
the changes of environmental, social, and economic development.
Predictability As indicators can be used to model future policy impacts, it is essential Dur et al. (2010); Spiekermann and Wegener (2004)
that indicators values can be forecasted for the future.

2008; DOT, 2007). With motor vehicle as a significant source of air economic). These are regarded as main aspects of sustainability
pollutants (83% of CO, 16% of PM10 , 63% of NOx , 13% of SO2 , 41% of in the field generally, and sustainable transport specifically (Black
VOC, and 97% of lead and compounds are related to motor vehicles et al., 2002; Haghshenas and Vaziri, 2012; Janic, 2006; Jeon and
(Brindle et al., 1999)) in Melbourne, transport sustainability issue Amekudzi, 2005; Levett, 1998; Sahely et al., 2005). Selected spatial
needs a special attention in this area. scale, SLA, for this study restricted the number of evaluated indi-
cators that can be quantified using available data. Some indicators
2.2. Indicators selection such as vehicle kilometres travelled (VKT), passenger kilometres
travelled (PKT), length of railways and main road, proportion
Sustainable indicators are variables which are defined to mea- of residents with public transit services within 500 m were not
sure progress towards sustainability. So indicators selection is the directly considered as an indicator, but rather they were used to
first step in all sustainability studies. This study started by evaluat- quantify selected indicators such as depletion of non-renewable
ing the list of indicators from past studies (Table 1). These indicators resources, emission, and accessibility. Fuel price and household
can be categorised in three main aspects: environmental, social, and expenditure on transport were evaluated as part of car owner-
economic. Indicators have been used to quantify both determinants ship cost indicator. Although energy consumption is one of the
of sustainability and final impacts. However, much of the literature major indicators considered in most of studies, it is highly corre-
does not distinguish between these two. lated with emission. As independency is an important criterion for
Selecting a set of indicators that provides a holistic picture of indicators selection, depletion of non-renewable resource is con-
the considered system is challenging (Castillo and Pitfield, 2010). sidered as an indicator instead of energy consumption. There was
Choosing indicators often involves tradeoffs. While using a smaller not enough data to quantify some indicators such as satisfaction of
set of indicators is more convenient, it may overlook the important citizens, variety and quality of transport options, quality of trans-
impacts. On the other hand, a large set of indicators is comprehen- port for disadvantaged, disabled, population exposed to noise, and
sive but its collection and analyses costs are extensive. So some cost of parking. So these indicators were excluded from evalua-
criteria are needed for indicators selection. Some selection criteria tion. Selected indicators for this study not only met the selection
were used to identify a list of indicators that can be used to quantify criteria, but also were a combination of environmental, social, and
transport sustainability in Melbourne (Table 2). economic aspects.
Screening by the selection criteria, only a small number of indi- It is worth mentioning that some of the selected environmen-
cators were found to be suitable for the study area (Table 3). As tal indicators are not the final outcomes of the transport system
the objective of sustainable urban transport changes with the con- and they are intermediate ones. For example, GHG emissions and
text, locality, time, and specific knowledge (Castillo and Pitfield, other air pollutants cause climate change, not only in national, but
2010), the selected set of indicators were flexible enough to repre- also in international scale. Climate change causes different dam-
sent Melbourne trend towards transport sustainability and covered ages such as changes in precipitation pattern, forest fires, flooding,
all aspects of sustainable development (environmental, social and drought events, sea level rise, and storms. These damages con-
tribute to people’s deaths and costs. So the final indicator might be
costs associated with pollutions and its social impacts. As quantify-
Table 3 ing these costs is impractical in this study, air pollutants emissions
Selected sustainable transport indicators for Melbourne. were selected as indicators.
Selected indicators for the study Unit
2.3. Indicators quantification
Transportation environmental indicators
Depletion of non-renewable resources Litres of crude oil per
After selecting indicators, values must be assigned to differ-
household annually
GHG emissions (CO2-e ) kg per household annually
ent indicators, so indicators must be quantified. For this study, the
Other air pollutants (CO, NO2 , PM10 ) kg per household annually databases applied were 2006 ABS database and Victorian Integrated
Land consumption for transport m2 per household Survey of Travel and Activity 2007 (VISTA07). Using these databases
Transportation social indicators selected transport indicators were quantified. As mentioned before,
Accessibility Score between 0 and 1 unlike other transport sustainability studies, this study attempt to
Fatalities and injuries related to traffic accidents Persons per household develop models for indicators quantification rather than directly
annually observed or measured indicators.
Mortality effects of air pollutants Persons per household
annually
2.3.1. Transport environmental indicators
Transportation economical indicators This group of indicators is about transportation resource
Car ownership costs $ per household annually depletion, transport related emissions, and land consumption for
Vehicle and general costs of accidents $ per household annually transport.
M. Reisi et al. / Ecological Indicators 43 (2014) 288–296 291

Table 4 Table 5
Emission factors for public and private transport (NPI, 2002, 2008). Appropriate travel distance to facilities (minimum and maximum values) (Pitot
et al., 2006).
Pollutant Emission factor for Emission factor for
public transport private transport Facilities Appropriate Appropriate
distances for distances for
CO 5.060 × 10−3 kg/VKT 4.440 × 10−3 kg/VKT
accessibility by accessibility by
PM10 0.569 × 10−3 kg/VKT 8.030 × 10−6 kg/VKT
walking public transport
NOx 10 × 10−3 kg/VKT 0.800 × 10−3 kg/VKT
Business centres 800–1600 m 20–50 min
Health centres 600–1200 m 20–40 min
• Transport fuel depletion: Transport fossil fuel usage is not sustain- Education centres 600–1200 m 20–40 min
Parks 500–1500 m 20–40 min
able in terms of depletion of a non-renewable resources and the 300–1000 for bus –
pollutions caused by combustion of fossil fuels (OECD, 1996). Public transport stations
600–1200 for train
In this study, litres of crude oil were used as a determinant
of resource depletion. To calculate the crude oil consumption,
first transport energy consumption was calculated using vehicle
kilometre travelled for private transports, passenger kilometre
gravity (all facilities divided by distance), travel cost (average dis-
travelled for public transport, and energy factor for private and
tance between origins and all facilities), minimum distance (the
public transport (Eq. (1)) (Rickwood, 2009).
distance from origins to the nearest facility), and coverage (the
Energy = VKTc EnFc + PKTp EnFp (1) number of facilities within specified distance) (Apparicio et al.,
2008; Lotfi and Koohsari, 2009). For example, Pitot et al. (2006)
where EnFc , energy factor for private car (4.6 MJ/VKT); EnFp , used both minimum distance and coverage to access accessibility
energy factor for public transport (1.4 MJ/PKT). by both walking and public transport to different range of facili-
To convert MJ of energy consumption in transport to litres of ties, while Wachs and Kumagai (1973) used coverage measures to
crude oil, petroleum refinery efficiency must be calculated. In estimate accessibility to employment by automobile. There is no
this study, petroleum refinery efficiency was estimated using single best approach for accessibility quantification because dif-
Eq. (2) (Wang, 2008) and volumes and energy contents of ferent situations and purposes need different approaches (Geurs
petroleum refinery inputs and products in Melbourne (Australian and Wee, 2004). As the objective of accessibility index in this
Government, 2010; BREE, 2010): study is to describe the proximity of SLAs to a series of facili-
Energy content of all petroleum products ties, distance measure and coverage were selected to quantify
Petroleum refinary efficiency = (2)
Energy content of crude oil inputs accessibility. Four types of distance can be used to calculate
the measure of accessibility: Euclidian distance (straight-line),
By knowing transport energy consumption and petroleum
Manhattan distance (distance along two sides of a right-angled
refinery efficiency in Melbourne (estimated to be 88%), crude oil
triangle), shortest network distance, and shortest network time.
consumptions (litres) can be calculated.
Network distance is a more accurate approximation of the travel
• GHG emissions: This indicator was included because motor vehi-
distance from an origin to a destination (Apparicio and Seguin,
cles produce about one-fifth of man-made carbon dioxide in the
2006) and so it is selected for the purpose of this study.
atmosphere. This indicator was calculated using emission factors
To quantify accessibility by walking in this study, points
for private car and passenger transport (Eq. (3)) (Rickwood, 2009).
of origin were Melbourne SLAs’ centres and points of des-
GHG emissions = VKTc EFc + PKTp EFp (3) tination included parks, education facilities, health services,
business zones, and public transport stations. In order to esti-
where EFc , emission factor for private car (0.260 kg CO2-e /VKT); mate accessibility, shortest network distance between origins
EFp , emission factor for public transport (0.095 kg CO2-e /PKT). and destinations were calculated, using ArcGIS 9.2 Network Ana-
It is worth mentioning that higher emission factor for Mel- lyst Extension, Melbourne’s land use map, and Melbourne’s roads
bourne compared to other countries in 2006 is that emissions network. Network Analyst finds distance (m) between origin
from passenger vehicles in Australia are 45% higher than the and destination. Then SLAs (points of origin) were weighted for
Europe (Beissmann, 2012). the level of accessibility according to the calculated distance,
• Other air pollutants: This indicator was also calculated using emis- using fuzzy linear function. The fuzzy linear function applies a
sions factors (Table 4) of selected air pollutants. Using 2002 linear function between the user-specified minimum and max-
emission factors for this study is justifiable because we cannot imum values. Anything below the minimum is assigned 0 and
expect extensive technological improvement to reduce air emis- anything above the maximum is assigned 1 (ESRI, 1999). This
sions in public transport between 2002 and 2006. Moreover, the minimum and maximum values show appropriate and inappro-
number of vehicle fleet replacement in four years time is gener- priate distance to facilities. For example, it was assumed that
ally small. places with less than 500 m from parks are appropriated (full
• Land consumption by transport: Areas of land which are devoted to fuzzy) and places with more than 1500 m from parks are inap-
transport infrastructures were estimated using Melbourne land propriate (fuzzy-less) and fuzzy linear function was assigned
use map. fuzziness between 500 m and 1500 m. Appropriate travel dis-
tance to facilities are shown in Table 5. To quantify accessibility
2.3.2. Transport social indicators by public transport, first public transport stops (PTx ) within a
This group includes accessibility, as well as all fatalities and specific walking distance (Table 5) from each destination were
injuries related to road transport. selected. Then public transport stops (PTy ) on the rail network
that are within a given travel time from PTx were selected. Then
• Accessibility: This indicator was added to consider the level of SLAs were weighted for the level of accessibility based on walking
walking and public transport accessibility in the city. Accessibil- distance to PTy , using fuzzy linear function.
ity, as the main goal of transport, is the ability to reach desired Therefore, each SLA is scored between 0 and 1 based on
goods, services, activities, and destinations. The most commonly its distance to different facilities. After that, the final score of
used accessibility measures in transport planning studies are accessibility for each SLA was obtained by using sum scores of
292 M. Reisi et al. / Ecological Indicators 43 (2014) 288–296

accessibility to different facilities by walking and public trans- and coroner costs. According to BITRE (2009), estimated human
port. losses are approximately $2.4 million per fatality and $214,000
• Fatalities and injuries related to traffic accidents: A sustainable per injury (BTE, 2000). Vehicle costs and general costs related
transport system should be designed and operated in a way to accidents which contain vehicle repairs, unavailability of
to minimise the number, severity, and risk of traffic accidents vehicles, towing, travel delay, insurance administration, police,
(Castillo and Pitfield, 2010). Based on ATSB statistics, number of non-vehicle property damage, fire and emergency services costs,
deaths and injuries due to accidents is 0.8 and 14.8 per 100 mil- are 49.02% and 29.64% of human costs respectively. As the num-
lion VKT respectively (ATSB, 2007). To our knowledge, there is no ber of fatalities and injuries related to transport was quantified
available published data for crash fatalities and injuries disaggre- as a social indicator, to avoid double counting, accidents’ costs
gated for Melbourne. Therefore the Australian average data was which were quantified in this study just contained vehicle and
used in this study. general costs related to transport accidents and human cost was
• Mortality effects of air pollutants: Health is linked to social, eco- excluded.
nomic and environmental aspects of life and so is a major factor Before integrating selected indicators into an index, they must
for sustainable development (Barton, 2009). To quantify the go through normalisation and weighting process.
health effects of air pollutants, three components must be esti-
mated: level of exposure to the air pollutant, exposure/response 2.4. Indicators normalisation
function (relative risk of considered air pollutant), and frequency
of the health outcome (Coffey, 2003). Transport related air pollut- Transport indicators contain different types of information so
ant is a mixture of many substances and many air pollutants are there might be some inconsistency in units among indicators.
highly correlated in the ambient air due to their common sources. Therefore, before indicators aggregation, it is necessary to trans-
The usual approach in air pollutant epidemiological studies is form them to numbers without any dimension. This process is
that at least one specific pollutant is measured as an indicator called normalisation (Nardo et al., 2005). Indicators whose increas-
of the complex mixture. Most available studies use PM10 as a ing values have positive impact on sustainability were normalised
useful indicator of the health risk related to transport. Although using Eq. (6a) and indicators whose increasing values have negative
PM10 may be a good indicator of air pollution, there is clear evi- impact on sustainability were normalised using Eq. (6b) (Krajnc and
dence that other pollutants which poorly correlated with PM10 Glavič, 2005):
may have independent health effects. One such example is Ozone +
I + − Imin
as an indicator of oxidant pollution (Kunzli et al., 1999). Moreover, IN+ = + + (6a)
according to the State of the Environment Report (2001), partic- Imax − Imin
ulate and Ozone concentrations remain a concern in Melbourne −
Imax − I−
with no clear downward trend (Coffey, 2003). So using relative IN− = − − (6b)
Imax − Imin
risk of death for PM10 , 1.0009 and Ozone, 1.0023 (EPA, 2000), the
expected number of deaths due to air pollution was estimated where IN , normalised indicator I; “+”, for indicator whose increas-
(Eq. (4)) (Ostro, 2004): ing value has positive impact on sustainability; “−”, for indicator
whose increasing value has negative impact on sustainability; min,
E = AF B P (4)
minimum value of indicator; max, maximum value of indicator.
where E, the expected number of deaths due to air pollution; AF,
the impact fraction of the health effect for exposed population; B, 2.5. Weighting and combining indicators
the population incident of mortality (i.e. death per 1000 people);
P, the relevant exposed population for the effect. This section compares different weighting methods to find the
The population attributable risk to air pollutants is expressed most appropriate technique for weighting in this study.
as a percentage of the total risk in the population. This is known
as the impact fraction of health effect for the exposed population. 2.5.1. Weighting methods comparison
For example if AF is 0.09% for PM10 , this means that, 0.09% of air Variables which are aggregated into a composite index have first
pollution related deaths in the population can be attributed to to be weighted using equal or different weights. Available weight-
PM10 . ing methods can be classified in three categories, equal weighting,
weighting based on opinions and weighting based on statistical
RR − 1
AF = (5) models (Saisana, 2011). In many composite indicators, all vari-
RR ables are given the same weight largely for reasons of simplicity
where RR, relative risk of air pollutant. (Dobranskyte-Niskota et al., 2007; Jeon et al., 2010; Lee and Huang,
2007). This implies that all indicators in the composite have equal
2.3.3. Transport economic indicators importance, which may not be the case. With the equal weighting
• Car ownership costs: Royal Automobile Club of Victoria (RACV) approach, there is a risk that certain aspects are double counted.
provides car ownership cost for 2010, 2011, 2012. The cost This is because two or more indicators may be measuring the same
includes standing costs (depreciation, interest on loan, registra- behaviour (Freudenberg, 2003).
tion, licence) and running costs (fuel, tires, and service/repairs The most common method of weighting based on opinions
Assuming the same growth trend in the past, average car owner- is the analytical hierarchy processes (AHP). It is a widely used
ship cost is estimated to be 72.18 cents per km in 2006 in technique for multi-attribute decision-making (Alexander, 2012;
Melbourne. It should be noted that fuel cost in Melbourne is about Lafleur, 2011; Tseng et al., 2008). It disaggregates a problem into
15% of the total cost of car ownership (RACQ, 2012). a hierarchical structure in which both qualitative and quantitative
• Cost of accidents fatalities and injuries: According to BTE (2000), aspects of a problem are evaluated using pair-wise comparisons.
road crash costs in Australia include human cost, vehicle cost, Pair-wise comparisons are made between pairs of indicators, asking
and general cost. Cost components of human losses are med- experts which of the two indicators is more important, and by how
ical, ambulance and rehabilitation, long-term care, labour in much. The preference is expressed on a scale of 1–9, with 1 indicates
workplace, labour in the household, quality of life, legal, correc- equal importance between two indicators, while 9 indicates that
tional services, workplace disruption and replacement, funeral one indicator is nine times more important than another. People’s
M. Reisi et al. / Ecological Indicators 43 (2014) 288–296 293

opinions are not always consistent. This method is recommended indicators, X1 , X2 ,. . ., Xp and find linear combinations of these to
for combining indicators less than 10, and is not transferable from produce principle components (or factors) Z1 , Z2 , Z3 ,. . ., Zp that are
one area to another (Saisana, 2011). Weightings based on opinions uncorrelated (Saisana and Tarantola, 2002).
are subjective and arbitrary according to experts’ judgments (OECD, Factor extraction using PCA requires two sets of values. Eigen-
2008; Saisana, 2011). So statistical models are more appropriate for vector which is simply a column or row of numbers in a correlation
weighting. matrix, and eigenvalue which is the sum of squares of factor
Data envelopment analysis (DEA), benefit of doubt (BoD), and loadings of each factor which reflects the proportion of variance
principle component analysis/factor analysis (PCA/FA) are the explained by each factor. The eigenvectors and eigenvalues are
major statistical methods used for weighting. Using mathematical derived by an iterative method. A vector is tried out and tested
programming techniques, DEA evaluates the efficiency of indi- against an indicator set of values. To the extent that it differs
vidual decision-making units (DMU) that use multiple inputs to from the indicator, the first trial vector is modified to produce
produce multiple outputs (Cherchye et al., 2007; Saisana, 2011). a second vector and so on until the solution converges, i.e. until
DEA allows DMUs to select weights that are the most advantageous additional iterations produce virtually identical results. In the iter-
for them in calculating their efficiency scores (Kao and Hung, 2005). ative approach eigenvectors are obtained one at a time. Once the
In other words, DEA will allow the DMU to assign very high weights iterative solution has converged, the eigenvalue can be calculated
to factors for which the DMU is particularly efficient and very low from the vector and the same iterative method is then used to
weights to all the other factors. As a result, the relative efficiency search for successive vectors (Kline, 1994). How this is actually
of a DMU may not really reflect its performance on the inputs and done was described in Kline (1994) in details. Usually, a few com-
outputs taken as a whole (Chaparro et al., 1997). Another limitation ponents (factors) will account for most of the variation, and these
of this method is that it classifies all DMUs into two groups, efficient components can be used to replace the original variables. Suitable
and inefficient, while in most cases ranking efficient and inefficient number of components is determined using eigenvalues. Normally
DMUs are essential. The third limitation of this method is that, an components that have eigenvalue larger than one are selected. The
inefficient DMU with a smaller efficiency score does not necessarily simplest justification for this rule is that it does not make sense
mean poorer performance than one with a larger efficiency score to add a component that explains less variance than is contained
because only the units under the same frontier facet are compa- in one individual indicator (OECD, 2008). SPSS software was used
rable (Kao and Hung, 2005). The application of DEA to the field of to carry out PCA/FA analysis in this study. As mentioned before,
composite indices is known as BoD. In this method, there is a need prior to PCA/FA analysis, indicators were separated to environmen-
for a benchmark to weight indicators. A composite index of an area tal, social, and economic aspects and PCA/FA analyses were run for
is not given by a weighted sum of its sub-indicators, but rather by each aspect independently.
the ratio of this sum to sum of the benchmark sub-indicators. This
method has the same limitation as DEA. Moreover, this method of
2.5.3. Determination of weighting factors
weighing deters the comparison among DMUs on a common base
Once principle components or factors are extracted, for weights
because the benchmark is endogenous and it may differ from one
calculation, it is needed to calculate the intermediate sustainability
DMU to another (Cherchye et al., 2007; Kao and Hung, 2005).
indicators (ISIj ), corresponding to each of the principle component
Principal Component Analysis/Factor analysis (PCA/FA) is a pop-
j. This is done by calculating a weighted aggregation of indicators
ular means for making comparisons between different indicators
(Gomez-Limon and Riesgo, 2008):
on several aspects. The method relies solely on the variation and
covariation of the data matrix to construct weights in the indexes 
n
(Zhang et al., 2011). So it is not based on experts’ opinions and ISIj = wkj Ik (7)
subjectivity problem of weighting would be solved by using this k=1
method. PCA is a better method than other statistical model, as
DEA or BoD just identified efficient and non-efficient DMUs (Kao where ISI, intermediate sustainability indicator; w, weight of indi-
and Hung, 2005), while as discuss further, PCA/FA has the ability to cator; I, normalised indicator; j, number of principle component; k,
rank efficient and non-efficient DMUs based on their sustainability. number of indicator.
Moreover, for creating useful index for decision makers, compar- The weights wkj are obtained from the factor loading:
ison between different DMUs are essential which do not exist in
2
BoD and DEA (Cherchye et al., 2007; Kao and Hung, 2005). (factor loadingkj )
wkj = (8)
eigenvaluej
2.5.2. PCA/FA principles
PCA/FA considers correlation among indicators to form a com- Finally sub-indices were calculated as a weighted aggregation
posite index that captures the information of individual indicators of the intermediate sustainability indicators:
as much as possible. Each factor or component reveals the set of

m
indicators with which it has the strongest association (OECD, 2008). I= ˛j ISIj (9)
Before applying PCA/FA, it must be checked that the correlation is
j=1
not due to redundancy of the information. To test whether PCA/FA
could be applied, Kaiser–Meyer–Olkin measure (KMO), which is where I, the value of sub-index; j, number of principle components;
a measure of sampling adequacy, is used. Reasonably large val- ISI, intermediate sustainability indicator; ˛ is the weight applied to
ues (between 0.6 and 1.0) are needed for a good combined PCA/FA the intermediate sustainability indicator. This weight is calculated
(Saisana, 2011). The first step in PCA/FA is to check the correlation as follows:
between indicators, so indicators correlation matrix is prepared
first. The second step is the identification of a certain number of eigenvaluej
˛j = m (10)
latent factors (fewer than the number of individual indicators) rep- eigenvaluej
j=1
resenting the data. Each factor depends on a set of coefficients
(loadings), each coefficient measures the correlation between In order to compare different Melbourne SLAs in terms of trans-
the individual indicator and the latent factor. Principal compo- port sustainability, it is necessary to build a composite index
nents analysis is usually used to extract factors. PCA considers ‘p’ to cover all environmental, social, and economic aspects. So
294 M. Reisi et al. / Ecological Indicators 43 (2014) 288–296

Table 6
Weights of indicators based on combined PCA/FA.

Aspect Kaiser–Meyer–Olkin Indicators Weights on factor 1 Weights on factor 2


measure

0.709 Depletion of non-renewable resources 0.3119 0.0272


GHG emissions (CO2-e ) 0.3119 0.0272
Environmental
Other air pollutants (CO, NO2 , PM10 ) 0.3113 0.0280
Land Consumption for transport 0.0076 1.1232

0.620 Accessibility 0.2946 0.2062


Social Fatalities and injuries related to traffic accidents 0.4924 0.0011
Mortality effects of air pollutants 0.0124 1.2355

0.500 Car ownership costs 0.5000 –


Economic
Vehicle and general costs of accidents 0.5000 –

0.669 IE 0.3465 0.2244


Final index Is 0.0440 1.8220
Ic 0.3573 0.1769

sustainability sub-indices were combined into a composite trans- sub-indices was high. IE , IS , IC , ICST are 0.75, 0.55, 0.49, and 0.61
port sustainability index (ICST ) (Gomez-Limon and Riesgo, 2008): respectively for Melbourne. The ICST value was the highest (0.93)
for Inner Melbourne, as one of inner SLAs and the lowest (0.16)

n
ISIj = wkj Ik (11) for Nimbulk-Bal, as one of the outer SLAs. Totally, sustainability in
inner SLAs is higher compared to other SLAs. The main reason is
k=1
that inner SLAs with highest population density, and better access
where ISI, intermediate sustainability sub-index; w, weight of sub- to facilities by walking and public transport, have lower VKT and
index; I, sub-index value; j, number of principle components; k, higher PKT compared to outer SLAs, which causes less transport
number of sub-indices. energy consumption and pollutions and its related social, and eco-
The weights wkj are obtained from the factor loading: nomic impacts.
2
(factor loadingkj )
wkj = (12)
eigenvaluej 3.2. Discussion
Finally composite transport sustainability index (ICST ) was calcu-
lated as a weighted aggregation of the intermediate sustainability This study aims to consider indicators affecting transport sus-
sub-indices: tainability. Although the number of indicators considered in this
study is limited compared to other transport sustainability stud-

m
ies (Haghshenas and Vaziri, 2012; Kane, 2010; Spiekermann and
ICST = ˛j ISIj (13)
Wegener, 2004), it is a unique transport sustainability study in SLA
j=1
level in Melbourne or even in Australia, which tries to develop sus-
where ICST , composite sustainable transport index; j, number of tainable index for transport. So it would be beneficial for urban
principle components; ISI, intermediate sustainability sub-indices; planning in Melbourne. Considering previous studies, this study
˛ is the weight applied to the intermediate sustainability sub- tried to differentiate between transport sustainability indicators
indices. This weight was calculated as follows: and transport sustainability determinants. Set of indicators in this
study, which were selected from previous studies using selection
eigenvaluej
˛j = m (14) criteria, are final indicators. Regardless of their limited numbers,
j=1
eigenvaluej they covered different aspects of sustainability, so they are appro-
priate to track transport sustainability in the study area.
3. Results and discussion This study is also different from other sustainability studies
because in other studies, the quantities of indicators were directly
3.1. Results observed or measured (Haghshenas and Vaziri, 2012; Zito and
Salvo, 2011), while in this study the indicators were modelled based
The results of index development and relationship between land on databases. VKT and PKT were obtained from the available statis-
use and transport sustainability index are considered in this sec- tics as determinants of transport sustainability. These were used
tion. In this study indicators in environmental, social and economic to quantify transport resource depletion, GHG emissions, other air
aspects were weighted separately using PCA/FA (Table 6). Accord- pollutants emission, and their related social and economic indica-
ing to Table 6, Kaiser–Meyer–Olkin (KMO) measure is high for all tors as final transport sustainability indicators.
aspects. Consequently PCA/FA can be used for weighting indica- Indicators in three aspects of sustainability were combined
tors in this study. In economic aspect, there was just one factor into environmental, social and economic sub-indices using weights
with eigenvalue more than 1, so weights of indicators in this aspect extracted by PCA/FA for each individual indicators. Conversely to
were calculated just for the first factor. While in environmental and the weighting approach proposed in the literature using analyt-
social aspect, there were two factors with eigenvalue greater than ical hierarchy process (AHP) (Castillo and Pitfield, 2010; Krajnc
one and weights of indicators were calculated for both factors. and Glavič, 2005) or similar weighting method (Haghshenas and
After assigning weights, indicators in each dimension were Vaziri, 2012; Ronchi et al., 2002), weighting approach in this study
aggregated into three sub-indices. In the later stage, these sub- is based on PCA/FA. The main reason of the choice is that PCA/FA
indices were weighted and aggregated into ICST (Fig. 2). The values considers correlation among indicators and eliminates the risk of
indices are between zero and one which show transport sustaina- double counting the effects. Moreover, the defined weights are
bility of each SLA compare to others. The value of ICST in each based on statistical methods and are not subjective. However, the
SLA would be high if the average of its individual sustainability possible disadvantage of PCA/FA may be interpretation of indicators
M. Reisi et al. / Ecological Indicators 43 (2014) 288–296 295

Fig. 2. Composite transport sustainability index for Melbourne, 2006.

weights. Weights of different indicators calculated by PCA/FA can- 4. Conclusions


not be compared with each other because the weights are not a
measure of theoretical importance of the indicators and they are Transport sustainability (environmental, social, and economic)
only used to correct overlapping information between correlated has became a central concern of urban design in the past decade.
indicators. This study provided a comparative analysis of different Melbourne
In the final step of the analysis the sub-indices were merged SLAs in terms of transport sustainability in environmental, social,
into composite transport sustainability index. The outcome of this and economic aspects. Three sub-indices were integrated into a
study, an index value, may be used by planning agencies and single composite index. Aggregating indicators into a single index
councils to diagnose the problem of SLAs with poor transport sus- in this study is a simplified and quantified expression of over-
tainability performances and to generate specific land use planning all transport sustainability in Melbourne. The index is derived by
for solving these problems. For example, the results suggested that applying normalisation, weighting and aggregation methods. Using
inner SLAs with high population density, mixed land use, and short PCA/FA for weighting indicators provides a proper way to resolve
distance to public transport and the CBD have higher sustainability co-linearity issue and consequently avoids double counting among
index. So, increasing population density and providing mixed land indicators and provides an unbiased measure of the transport sus-
use and an efficient public transport would result in more sus- tainability.
tainable transport and consequently lower adverse environmental, This study is the first attempt to quantify transport sustaina-
social and economic impacts. These results are consistent with Zito bility in SLA level for Melbourne which provides information about
and Salvo (2011) study which claimed that sprawl cities have lower current situation of urban transport in different parts of the city.
transport sustainability. The transport sustainability index can be used by policy makers to
As this paper investigated limited basic indicators for sustain- evaluate the effect of their policies on transport sustainability in dif-
able transport, there is a need for further research to consider wider ferent aspects. For example, evaluating a policy using the proposed
range of indicators in all three aspects for Melbourne in order to model can show whether the policy is expected to limit emissions
reflect sustainability better. Considering other spatial levels (such and resource consumption; or it is expected to provide access for
as local government area (LGA) or census collector district (CCD)), people, and enhance human health; or it is expected to help reduce
and other weighting methods (such as AHP) is also recommended economic costs to society.
for future studies. Another major limitation of this study is in the
quantification of the accessibility indicator. In the index developed
in this study centres of SLAs were considered as point of origins.
References
One point of origin may not be enough for considering accessibility
all around one SLA, and there is a need an index that considers more Alberti, M., 1996. Measuring urban sustainability. Environ. Impact Assess. Rev. 16,
origin points in each SLA. 381–424.
296 M. Reisi et al. / Ecological Indicators 43 (2014) 288–296

Alexander, M., 2012. Decision-making using the Analytic Hierarchy Process (AHP) Kane, L., 2010. Sustainable transport indicators for Cape town South Africa: advo-
and SAS/IML, SESUG. cacy, negotiation and partnership in transport planning practice. Nat. Resour.
Alford, G., Whiteman, J., 2008. Macro-urban form, transport energy use and green- Forum 34, 289–302.
house gas emissions: an investigation for Melbourne. In: 31st Australian Kao, C., Hung, H.T., 2005. Data envelopment analysis with common weights: the
Transport Research Forum. compromise solution approach. J. Oper. Res. Soc. 56, 1196–1203.
Apparicio, P., Abdelmajid, M., Riva, M., Shearmur, R., 2008. Comparing alternative Kline, P., 1994. An easy Guide to Factor Analysis. Routledge, London.
approaches to measuring the geographical accessibility of urban health services: Krajnc, D., Glavič, P., 2005. A model for integrated assessment of sustainable devel-
distance types and aggregation-error issues. Int. J. Health Geogr. 7, 7. opment. Resour. Conserv. Recycl. 43, 189–208.
Apparicio, P., Seguin, A., 2006. Measuring the accessibility of services and facilities Kunzli, N., Kaiser, R., Medina, S., Studnicka, M., Oberfeld, G., Horak, F., 1999. Health
for residents of public housing in Montreal. Urban Stud. 43, 187–211. costs due to road traffic-related air pollution. An impact assessment project of
ATSB, 2007. In: Bureau, A.T.S. (Ed.), Road Crash Casualties and Rates, Australia 1925 Austria, France and Switzerland. In: Third Ministerial Conference for Environ-
to 2005. Australian Transport Safety Bureau. ment and Health, London.
Australian Government, 2010. National Greenhouse Accounts (NGA) Factors. Lafleur, J., 2011. Probabilistic AHP and TOPSIS for multi-attribute decision-making
Department of Climate Change and Energy Efficiency. unider uncertainty. In: IEEE Aerospace Conference, Big Sky, Montana.
Barton, H., 2009. Land use planning and health and well-being. Land Use Policy 26S, Lee, Y., Huang, C., 2007. Sustainability index for Taipei. Environ. Impact Assess. Rev.
S115–S123. 27, 505–521.
Beissmann, T., 2012. Australia’s Vehicle Emissions Trail Europe by a Long Way. Levett, R., 1998. Sustainability indicators – integrating quality of life and environ-
BITRE, 2009. Road Crash Costs in Australia 2006. Bureau of Infrastructure, Transport mental protection. J. R. Stat. Soc. Ser. A: Stat. Soc. 161, 291–302.
and Regional Economics. Li, F., Liu, X., Hu, D., Wang, R., Yang, W., Li, D., Zhao, D., 2009. Measurement indicators
Black, J.A., Paez, A., Suthanaya, P.A., 2002. Sustainable urban transportation: per- and an evaluation approach for assessing urban sustainable development: a case
formance indicators and some analytical approaches. J. Urban Plann. Dev. 128, study for China’s Jining City. Landscape Urban Plann. 90, 134–142.
184–208. Lotfi, S., Koohsari, M., 2009. Measuring objective accessibility to neighbourhood
BREE, 2010. Australian Petroleum Statistics. Bureau of Resources and Energy Eco- facilities in the city (a case study: Zone 6 in Tehran, Iran). Cities 26, 133–140.
nomics. Nardo, M., Saisana, M., Saltelli, A., Tarantola, S., 2005. Tools for Composite Indicators
Brindle, R., Houghton, N., Sheriden, G., 1999. Transport Generated Air Pollution and Building. Institute for the Protection and Security of Citizen.
its Health Impacts – A Source Document for Local Government, Research Report NPI, 2002. Emission Estimation Technique Manual for Combustion Engine. Natinal
ARR 336. ARRB Transport Research, Victoria. Pollutant Inventory.
BTE, 2000. Road Crash Costs in Australia. Bureau of Transport Economics. NPI, 2008. Emission Estimation Technique Manual for Combustion Engine. Natinal
Castillo, H., Pitfield, D.E., 2010. ELASTIC – a methodological framework for identify- Pollutant Inventory.
ing and selecting sustainable transport indicators. Transport. Res. D: Transport OECD, 1996. Towards sustainable transportation. In: The Vancouver Conference,
Environ. 15, 179–188. Vancouver, BC.
Chaparro, F., Salinas-Jimenez, J., Smith, P., 1997. On the role of weight restrictions OECD, 2008. Handbook on Constructing Composite Indicators, Methodology and
in data envelopment analysis. J. Prod. Anal. 8, 215–230. User Guide. Organisation for Economic Co-operation and Development.
Cherchye, L., Moesen, W., Rogge, N., Puyenbroeck, T.V., 2007. An introduction to Ostro, B., 2004. Outdoor air pollution: assessing the environmental burden of disease
‘benefit of the doubt’composite indicators. Soc. Indic. Res. 82, 111–145. at national and local levels. Environ. Burden Dis. Ser., 5.
Coffey, 2003. Fuel Quality and Vehicle Emissions Standards – Cost Benefit Analysis. Pitot, M., Yigitcanlar, T., Sipe, N., Evans, R., 2006. Land Use & Public Transport Acces-
Coffey Geoscience Pty Ltd. sibility Index (LUPTAI) tool: The Development and Pilot Application of LUPTAI
Dobranskyte-Niskota, A., Perujo, A., Pregl, M., 2007. Indicators to assess sustaina- for the Gold Coast. Planning and Transport Research Centre (PATREC).
bility of transport activities. European Comission, Joint Research Centre. JRC RACQ, 2012. Vehicle Running Costs 2012.
European Commission. Rickwood, P., 2009. The Impact of Physical Planning Policy on Household Energy
DOT, 2007. Victorian Integrated Survey of Travel and Activity 2007 (VISTA07). Trans- Use and Greenhouse Emissions, Faculty of Design, Architecture, and Building.
port, D.o. University of Technology, Sydney.
Dur, F., Yigitcanlar, T., Bunker, J., 2010. Towars sustainable urban futures: evaluat- Ronchi, E., Federico, A., Musmeci, F., 2002. A system oriented integrated indicator
ing urban sustainability performance of the Gold Coast, Australia. In: 14th IPHS for sustianable development in Italy. Ecol. Indic. 2, 197–210.
Conference, Istanbul, Turkey. Sahely, H.R., Kennedy, C.A., Adams, B.J., 2005. Developing sustainability criteria for
EA, 1999. Sustainable Transport: Responding to the Challenges, Sustainable Energy urban infrastructure systems. Can. J. Civil Eng. 32, 72–85.
Transport Taskforce Report. The Institution of Engineers, Australia. Saisana, M., 2011. Weighting methods, Seminar on composite Indicators: From The-
EPA, 2000. Melbourne Mortality Study: Effects of Ambient Air Pollution on Daily ory to Practice, Ispra.
Mortality in Melbourne 1991–1996. Environmental Protection Authority. Saisana, M., Tarantola, S., 2002. State of the Art Report on Current Methodologies
ESRI, 1999. How Fuzzy Membership Works. and Practices for Composite Indicator Development. Institute for the Protection
Freudenberg, M., 2003. Composite Indicators of Country Performance: A Critical and Security of the Citizen.
Assessment. OECD Publishing, Paris. Spiekermann, K., Wegener, M., 2004. Evaluating urban sustainability using land-
Geurs, K., Wee, B., 2004. Accessibility evaluation of land-use and transport strategies: use transport interaction models. Eur. J. Transport Infrastruct. Res. 4, 251–
review and research directions. J. Transport Geogr. 12, 127–140. 272.
Gomez-Limon, J.A., Riesgo, L.,2008. Alternative approaches on constructing a com- Tseng, C., Hong, C., Chang, H., 2008. Multi attribute decision making model for med-
posite indicator to measure agricultural sustainability. In: 107th Seminar, ical service selection: an AHP approach. J. Qual. 15, 155–165.
January 30-February 1 2008, Sevilla, Spain. European Association of Agricultural Wachs, M., Kumagai, G., 1973. Physical accessibility as a social indicator. Socioecon.
Economists. Plann. Sci. 7, 437–456.
Haghshenas, H., Vaziri, M., 2012. Urban sustainable transportation indicators for Waitz, I., Townsend, J., Cutcher-Gershenfeld, J., Greizer, E., Kerrebrock, J., 2004. Avi-
global comparison. Ecol. Indic. 15, 115–121. ation and the Environment A National Vision Statement, Framework for Goals
Janic, M., 2006. Sustainable transport in the European Union: a review of the past and Recommanded Actions. Massachusetts Institute of Technology, Cambridge.
research and future ideas. Transport Rev. 26, 81–104. Wang, M., 2008. Estimation of energy efficiencies of U.S. petroleum refineries. Centre
Jeon, C., Chang, Y., Amekudzi, A., 2010. Incorporating uncertainty into transport deci- for Transportation Research, Argonne National Laboratory.
sion making: a sustianability-oriented approach. In: TRB 2010 Annual Meeting, Zhang, X., Wu, Y., Shen, L., 2011. An evaluation framework for the sustianability of
Washington. urban land use: a study of capital cities and municipalities in China. Habitat Int.
Jeon, C.M., Amekudzi, A., 2005. Addressing sustainability in transportation systems: 35, 141–149.
definitions, indicators, and metrics. J. Infrastruct. Syst. 11, 31–50. Zhou, P., Ang, B., Poh, K., 2007. A mathematical programming approach to construc-
Jollands, N., Lermit, J., Patterson, M., 2003. The Usefulness of Aggregate Indicators in ting composite indicators. Ecol. Econ. 62, 291–297.
Policy Making and Evaluation: A Discussion with Application to Eco-efficiency Zito, P., Salvo, G., 2011. Toward an urban transport sustainability index: an European
Indicators in New Zealand. Economics and Environment Network. Australian comparison. Eur. Transport Res. Rev. 3, 1–17.
National University.

You might also like