Journal of Transport Geography 35 (2014) 130–143

Contents lists available at ScienceDirect

Journal of Transport Geography

journal homepage: www.elsevier.com/locate/jtrangeo

Measuring transit oriented development: a spatial multi criteria

assessment approach for the City Region Arnhem and Nijmegen
Yamini Jain Singh a,⇑, Pedram Fard a, Mark Zuidgeest b, Mark Brussel a, Martin van Maarseveen a
a
Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, PO Box 217, 7500 AE Enschede, The Netherlands
b
Centre for Transport Studies, Department of Civil Engineering, Faculty of Engineering and the Built Environment, University of Cape Town, Private Bag X3, Rondebosch 7701,
Cape Town, South Africa

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

Keywords: Transit Oriented Development (TOD) is a planning approach that can stimulate sustainable development
TOD by encouraging better land use and transport integration. Arnhem Nijmegen City Region, a regional
TOD Index planning body, in the Netherlands, aims to promote sustainable development in their region and control
SMCA the current pattern of increased use of cars vis-à-vis transit for longer commutes. Planning for TOD can
Hot-spot analysis
help achieve this aim. It is believed that measuring the existing levels of TOD is a prerequisite for TOD
Arnhem Nijmegen City Region
planning and that it can be done using a TOD Index proposed in this research. A TOD Index measures mul-
tiple spatial indicators and aggregates them under the SMCA framework to arrive at a comprehensive
value depicting the existing levels of TOD at a location or an area. Using this TOD Index, TOD levels were
measured over the entire City Region covering approx. 1000 km2. High levels of TOD imply that the urban
development’s characteristics, at that location, are ripe for use of transit and these high levels, as
expected, were found in the urban areas of Arnhem and Nijmegen. From the results of TOD Index
measurement, using hot-spot analysis, those locations were identified that have high TOD levels but poor
transit connectivity. These locations are accordingly recommended for better transit connectivity. As a
part of our future work, it is intended to use the TOD Index to elevate TOD levels around existing transit
nodes.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction two-sided planning approach that ensures that necessary planning

Transit Oriented Development (TOD) as a planning tool
integrates the land use and transport system, thereby creating live-
ly, sustainable, pedestrian and cycling friendly areas and neigh-
bourhoods (CTOD, 2009), while encouraging people to choose
transit over cars for their long commutes. A more widely agreed
description of TOD is an urban environment with high densities,
mixed and diverse land uses, located within an easy walkable area
around a transit stop (Calthorpe, 1993; The City of Calgary, 2004;
Parker et al., 2002; Hale and Charles, 2006). But TOD planning is
not and should not only be about creating development that is ori-
ented towards transit use. It should also mean bringing transit to
those locations, where the development already possesses the
physical characteristics of that of a typical TOD, but without having
transit connectivity at that place. This actually signifies a
⇑ Corresponding author. Tel.: +31 (0) 619357011.
E-mail addresses: y.j.singh@utwente.nl (Y.J. Singh), fard28531@alumni.itc.nl
(P. Fard), mark.zuidgeest@uct.ac.za (M. Zuidgeest), m.j.g.brussel@utwente.nl
(M. Brussel), m.f.a.m.vanmaarseveen@utwente.nl (M.van Maarseveen).
interventions are made to either make the urban development
more transit oriented or to bring transit to places where the devel-
opment is already transit oriented, i.e. having high densities, etc.
We can also call this planning approach ‘TOD Planning’.
Public investments in infrastructure such as for TOD are too of-
ten made without understanding the existing situation and possi-
ble outcomes of the plans. Renne and Wells (2005) concluded in
their report on measuring TOD that there is a lack of comprehen-
sive tools of measurement that quantify the existing levels of
TOD. Measuring those levels helps in better TOD planning. This
has also been highlighted by Evans and Pratt (2007) as they point
out that there is a need to express the existing TOD-ness of an area
in terms of a ‘TOD Index’. They describe a TOD Index as a
‘‘. . .potential device for considering the degree to which a particu-
lar project is intrinsically oriented towards transit’’ (Evans and
Pratt, 2007, p. 17-3). Such a TOD Index measures all the criteria
of TOD and also makes different areas comparable, objectively.
TOD planning can also benefit from lessons learnt from previous
TOD projects. Renne and Wells (2005, p. 10), stress for example
that ‘‘Without measuring the outcomes of TODs, mistakes in

http://dx.doi.org/10.1016/j.jtrangeo.2014.01.014
0966-6923/Ó 2014 Elsevier Ltd. All rights reserved.
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
investment strategies will continue to be repeated’’. The literature
is rich with case studies, where TOD projects have been evaluated
qualitatively or quantified by measuring the changes in transit rid-
ership and development densities amongst others. However, a
comprehensive measurement of TOD levels in form of a single in-
dex that can be used for efficient TOD planning has been missing so
far. In addition to this, Schlossberg and Brown (2004, p. 2), point
out that it is the ‘‘fine grained, spatially explicit types of analyses
that have been lacking in TOD’’ efforts in spite of the fact that
two corner stones of TOD, land use and transport, are both inher-
ently spatial in nature. Hence, a number of indicators that relate
to TOD need to be measured and combined into an index that
can further help in making justified planning decisions, such as
for location decision-making.
This paper, thus, presents an approach and a tool capable of
measuring TOD levels, resulting in a TOD Index, which can be used
for TOD Planning. We have used spatial analytical tools, i.e. GIS and
Spatial Multiple Criteria Analysis (SMCA) to measure TOD levels.
For our case study, we have chosen the City Region of Arnhem
and Nijmegen, The Netherlands.
The Arnhem Nijmegen City Region in The Netherlands is the
third largest of eight City Regions in the country. These City Re-
gions have been formed to plan for urban growth at a regional le-
vel, around a few important cities. Arnhem and Nijmegen are two
such important cities. West of the City Region is the main eco-
nomic zone of the country, Randstad Metropolitan Region, that
covers a large area and includes four of the eight City Regions of
the Netherlands (Fig. 1).
Fig. 1. Arnhem Nijmegen City Region in the Netherlands.
131
Even as Randstad continues to be the main economic centre,
Arnhem Nijmegen City Region (called ‘City Region’ hereinafter) is
poised to become the second biggest economic zone in the Nether-
lands by 2020. Currently, the City Region consists of 20 municipal-
ities with around 735,000 inhabitants over an area of 1000 km2
and more than 40 percent of the population resides in its two
major cities (Fig. 2).
The City Region’s policy vision is centred on creating more
housing and employment and providing higher levels of mobility.
However, in the current scenario, cars are being increasingly used
for more and longer trips rather than public transport (Rijkswater-
staat, 2010). This issue coincides with the overall mobility situa-
tion in the country, as reported in the Mobility Report for the
Netherlands (KiM, 2013), where cars account for approximately
three-fourths of the total vehicle-kilometres travelled in the coun-
try in 2012. To stimulate a modal shift from cars to transit, the City
Region wishes to make public transport as competitive as private
transport and to ensure that more and more commuters use transit
for their regional travel demands. To do this, the City Region al-
ready offers an existing rail based transit service and is planning
for BRT. Hence, it would benefit from knowing which areas in
the region require transit connectivity. Realizing such modal shift
of trips from cars to transit is one of the most difficult tasks to
do anywhere in the world, however, ‘Transit Oriented Develop-
ment’ is one of those popular and effective planning tools that
can bring about the desired change as shown, for example, by Cer-
vero and Murakami (2009) in the case of Hong Kong. The City Re-
gion also sees TOD as a powerful tool that can help make people
132 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 2. The City Region Arnhem Nijmegen.
choose transit over their private modes of travel. Hence, the City
Region becomes a natural choice for our purpose of using the
TOD Index for TOD Planning.
The next section elaborates our methodology to build the tool
and arrive at the TOD Index results. In Section 3, our two-sided
planning approach has been discussed in detail while Section 4
gives the criteria and indicators used for TOD Index calculations.
Section 5 accordingly explains how the indicators were calcu-
lated and Section 6 explains how we used SMCA to assess all
indicators together to arrive at TOD Index values representing
the existing levels of TOD. As results are obtained at the end
of Section 6, we identify hot-spots based on the TOD Index val-
ues in Section 7. Our recommendations have been made in Sec-
tion 8 and results of the sensitivity analysis have been reported
in Section 9. The last section, Section 10, concludes the paper
while discussing the limitations faced and our plan of work in
the near future.
2. Measuring TOD – our methodology
Most literature on TOD deals with evaluation of completed TOD
projects. Some case studies present an example of planned TOD at
urban, regional or station level (Yang and Lew, 2009; Balz and
Schrijnen, 2009; Cascetta and Pagliara, 2009; Howe et al., 2009;
Curtis, 2009; Arrington, 2009), while some cases represent the suc-
cessful role of the private sector in TOD implementation (Chorus,
2009; Cervero and Murakami, 2009 and yet some cases are about
TODs centred around Bus Rapid Transit (BRT) (Hoffman, 2006;
McKone, 2010). It has been found that complete and quantitative
measurement of TOD-ness of an area has not been done. Only a
few studies measured one or more TOD criteria. In the work by
Schlossberg and Brown (2004), a ‘walkability’ indicator has been
assessed in ArcGIS for 11 TOD sites in Portland, Oregon. It is a good
example of quantitatively measuring TOD-ness, using a spatial
platform, i.e. GIS, however, for only one indicator.
As mentioned earlier, Evans and Pratt (2007) have supported
the need for characterizing the TOD-ness of an area not only for
evaluation purposes but also for planning new TODs. They even
identified and categorised indicators that ‘might likely’ make up
such a TOD Index. Our work is an extension from theirs. We took
the concept of TOD Index to fruition with an elaborate methodol-
ogy that calculates a number of spatial criteria and combines them
into this TOD Index, and apply this to the Arnhem Nijmegen City
Region. Our methodology to calculate and interpret TOD Index val-
ues has been depicted in Fig. 3 based on our conceptual framework
presented in Singh et al. (2012).
As per the methodology, first we identify the criteria, i.e. indica-
tors that make the TOD Index, and then we assess these individual
indicators spatially in ArcGIS. Using the Spatial Multiple Criteria
Assessment (SMCA) platform in ILWIS (ITC, 2007), these indicators
are then assessed again to arrive at comprehensive TOD Index val-
ues. The result of this SMCA analysis is a TOD Index map depicting
index values for various parts of the study area. The advantages of
SMCA include its capability of assessing multiple indicators and
allowing the decision maker(s) or stakeholders to weigh the crite-
ria (Beuthe, 2002; Van Wee, 2011), while being transparent and
back-traceable all the way. Some notable works on the use of
SMCA in the field of transport include those of Keshkamat et al.
(2009) and Beukes et al. (2011). Using TOD Index values, we can
recommend how TOD can be improved in certain areas and which
areas may be considered for better transit connectivity. Transit
agencies or civic bodies who are the real decision makers, will then
decide on detailed TOD planning proposals. If possible, an interac-
tive Spatial Decision Support System (SDSS) platform can be cre-
ated to help them. Such SDSS are explicitly designed to help
make decisions related to complex spatial problems such as plan-
ning for TOD.
Further to this section, in Sections 4 and 5, we will discuss the
various steps of SMCA where we will elaborate our chosen criteria
for the TOD Index, data acquisition and calculation of all criteria.
But before we do that, we want to revisit the two-sided planning
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143 133

Fig. 4. Schematic procedure of computing composite index map based on spatial

multi criteria analysis. Source: Abella and Westen (2007).

Table 1
Criteria for measuring Potential TOD Index.

Potential TOD Index

Criteria
What are the various densities?
How diverse is the land use?
Does the design of urban space
encourage walking and cycling?
What is the current level of economic
development?
Indicators
Residential density
Employment density
Commercial intensity/density
Land use diversity
Level of mixed-ness of land uses
w.r.t residential land use
Quality and suitability of
streetscape for walking
Quality and suitability of
streetscape for cycling
Density of controlled intersections/
street crossings
Private investment in the area
Number of business establishments
Tax earnings of municipality
Unemployment levels

We believe that transit systems can also benefit financially, by

Fig. 3. Research methodology.
approach for TOD planning that has been mentioned in Section 1.
We believe that to satisfy the requirements of our TOD approach,
a single TOD Index is not sufficient. Hence it was required to build
two types of TOD indices. These indices along with the motivation
behind them have been detailed out in the next section.
3. Actual and Potential TOD Indices
TOD is generally understood as a means of creating or changing
‘development’ so as to make it more oriented towards using tran-
sit. However, our TOD planning approach for the City Region has
two objectives as indicated below:
(a) Plan for higher TOD levels at various locations or areas, where
transit connectivity is available but TOD levels are low.
(b) Plan for transit connectivity at those locations or areas where
high levels of TOD exist but transit connection is absent or
poor.
way of fare-box revenues and value re-capture, if they plan and
locate their stops in areas with high TOD levels. Such locations,
as revealed by TOD Index values, can then be recommended to
the transit authorities and can be further put to detailed demand
assessment, technical feasibility analysis, financial – economic
feasibility analyses and others.
Since we wish to fulfil these two objectives focusing on two dif-
ferent types of areas, a single TOD Index is not sufficient. To fulfil
the first objective, an index must measure TOD levels within a
walking distance of each transit node. It should also measure the
characteristics of not just urban development surrounding the
node, but also those of the transit servicing that area. To fulfil
the second objective, an index must be able to measure urban
development characteristics in all areas of the region. It cannot
measure transit characteristics as we are finding those areas where
transit connectivity is absent and is desired. For both these objec-
tives, separate indices will be required as they will differ in scale
and in terms of the indicators to be measured. We propose to call
these two TOD Indices – Actual TOD Index and Potential TOD In-
dex. The Actual TOD Index measures existing TOD levels in areas
within walkable limits of an existing/actual transit node. For the
134 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 5. Residential density.
rest of the case study area, the existing TOD levels are measured
using an index called ‘Potential TOD Index’ as it helps in identifying
‘potential’ locations for transit connectivity at a regional scale.
It may be mentioned here that presence of transit is not a pre-
requisite for TOD qualification. As Calthorpe (1993, p. 42) said,
‘‘Transit-Oriented Developments can, and ironically, should devel-
op without transit – with justifiable focus on pedestrian and
healthier community structure’’. Further, he said that transit is
not the only goal of TOD but a ‘potential end result’ of TOD.
Actual TOD covers the areas around existing transit nodes only
and is hence at a smaller scale. This paper elaborates our work on
the Potential TOD Index since it is measured over an entire area
and can so help us to understand the levels of TOD at a regional le-
vel. The work on the Actual TOD Index is planned as one of the next
steps in our research. It is worthwhile to mention here that in this
study, the TOD Index has been measured within the context of
train system as the transit system. According to Newman (2009),
for transit systems to be successful, they must be able to compete
with cars/private transport in speed and capacity. The regular bus
service in the City Region is also too slow to compete with cars.
Since BRT is still being planned for the City Region, only the train
system was studied as transit system required for TOD in the City
Region. In the next section, we explain the criteria identification
and data procurement.
4. Criteria identification and data collection
Arriving at a comprehensive TOD Index score involves measure-
ment of individual indicators and criteria that make the index.
Many indicators have been used in various case studies in the past,
where TOD has been evaluated for its success. Renne and Wells
(2005), Evans and Pratt (2007) and Shastry et al. (2011), have listed
the most important indicators that can be used to evaluate TOD.
Dittmar and Poticha (2004) and Belzer and Autler (2002) have both
identified performance criteria that are necessary for evaluation of
TOD’s performance and according to Cervero and Murakami
(2008), TOD can be best assessed using the five Ds – density, land
use diversity, urban design, destination accessibility and distance
to transit. It is to be noted that most of the proposed indicators
and criteria have been proposed for the purpose of evaluation of
TOD’s performance while our TOD Index measures the existing lev-
els of TOD and does not evaluate them. Further, not all of these
indicators can be quantified and hence cannot be used for our in-
dex. Thus, in light of the earlier works and specific needs of our re-
search, quantifiable indicators were formulated.
In a typical SMCA process, there are a number of indicators,
clubbed under criteria that constitute a composite value or an in-
dex (Fig. 4). As a part of this process, indicators are standardised
and weighed. The following section elaborates the criteria that
we used for measuring the Potential TOD Index.
4.1. Criteria to measure Potential TOD Index
The four criteria that were used for measuring our Potential
TOD Index (Table 1) are urban densities, land use diversity, design
of urban space and economic development levels. The indicators
under these criteria have been primarily adapted from the indica-
tors proposed by Evans and Pratt (2007) for their TOD Index, from
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 6. Commercial density.
Fig. 7. Window of analysis.
the listing by Renne and Wells (2005) and from the 5Ds proposed
by Ewing and Cervero (2010). All these studies have discussed at
135
length which urban development characteristics can be typically
expected in transit oriented developments. They show that urban
densities and land use diversity are important for effective utilisa-
tion of transit, which is the eventual aim of TOD. Design of urban
space is very important for creating walkable and cyclable spaces
and economic development is necessary to attract or generate vol-
umes of travel demand. Together the first three criteria represent
the 3Ds of urban development that are also the corner stones of
TOD, while the fourth criterion, economic development level, has
a direct impact on travel demand and hence also affects use of
transit.
4.2. Data collection
Spatial data in GIS format and non-spatial data (mainly statisti-
cal data) at the regional level were collected from credible second-
ary sources including the City Region itself, ESRI Nederland,
Statistics Netherlands (CBS), and Open Street Map (OSM). Gather-
ing data from these fragmented sources involved issues like incom-
plete map coverage, inconsistent administrative boundaries,
conflicting land use classifications, different map projections and
so on, which in most cases were addressed by comparison between
sources and also with the help of a third party reference like Goo-
gle Maps. Data on quality of the streetscape for walking and cy-
cling, financial investments in an area and such was not available
as a result of which, some indicators from Table 1 had to be
dropped from our analysis. The final criteria tree that we used
has been put in bold in Table 1, and covers all four criteria classes.
We believe that these are sufficiently representative of the four
136 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 8. Land use diversity.
criteria, even though not most extensive and complete. Table A in
the Appendix A provides details on the formats and source of data
used for this work.
4.3. Spatial unit for Potential TOD Index calculations
An actual TOD area has been typically described as the area
within walkable limits of a transit node, i.e. an area within about
500–800 m of a transit node (Calthorpe, 1993; The City of Calgary,
2004). However, the Potential TOD Index has no such defined
boundary or recommended size of area, yet it has to be measured
for the entire City Region. Hence, using tessellation of space
(Mitchell, 2009), different grid tessellations of 100  100,
200  200, 300  300 and 500  500 m were created. Keeping in
mind the spatial scale of typical TOD area (which is about 80 ha
within a radius of 500 m from an existing station), the cell size of
grid, and computational performance factor, a cell size of
300  300 m was selected as the basic spatial unit for measuring
Potential TOD Index for the entire City Region. Overlaying 1000
km2 of Arnhem-Nijmegen region with such grid cells resulted in
more than 12,000 cells. In the data conversion process, neighbour-
hood polygon data were apportioned to the grid cells using a GIS
overlay process called ‘Data apportion for non-coterminous
polygons’.
It may be recalled here that in Section 3, we mentioned that the
Potential TOD Index is calculated for the entire study area except
for those areas that currently have transit connections. But it was
found to be practically less complicated to first calculate the index
for the entire City Region and then exclude the areas with existing
transit connectivity.
5. Computing indicators for Potential TOD Index
As given in Table 1, five indicators were calculated to measure
the TOD Index across the City Region. The following sub-sections
detail out the calculations for each indicator.
5.1. Density indicators
High densities are important to support higher frequency tran-
sit service and to foster lively, walkable communities (Bach et al.,
2006). They also optimize transit rider convenience (The City of
Calgary, 2004). Thus, high density development is an indication
of high travel demand, possibly high transit patronage and hence
high TOD levels in an area.
Under the hierarchy of administrative units, the City Region is
divided into 20 municipalities that are further divided into 78 dis-
tricts and 443 neighbourhoods. The data from CBS on the number
of houses and commercial establishments was available at neigh-
bourhood level and had to be disaggregated to the grid level. The
resulting maps representing the residential and commercial densi-
ties (Figs. 5 and 6) are as shown below.
The residential and commercial densities are concentrated
around the cities of Arnhem and Nijmegen and are quite low out-
side the urban areas, as expected.
5.2. Land use diversity indicator
Presence of diverse land uses creates more balanced and consis-
tent passenger flow at all times of the day, all days of the week and
throughout the year. The increased and varied passenger market
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 9. Mixed-ness of land use.
contributes to higher revenue and better utilisation of transit
capacity (Bach et al., 2006). If the areas are not dominated by a sin-
gle land use, we can assume that people will feel livelier, safer and
that the areas will be lighted for longer periods of time that would
encourage better interaction among people from all walks of life.
Land use diversity, LUd(i), is a measure of spatial heterogeneity
of land use types within an area i and measured here by use of en-
tropy. The established methods of computing land use diversity
use raster data where each pixel or portion of space represents
one dominant land use rather than a set of land uses, as is in vector
data. Since our study uses vector data, the research of Ritsema van
Eck and Koomen (2008) was considered and a geo-processing
model was developed capable of computing diversity over vector
data.
During the calculations of the land use diversity indicator, our
first approach was to calculate the index for each grid cell indepen-
dently. However, one cell of 300  300 m which covers an area as
small as 9 ha, only has a limited diversity since it cannot accommo-
date a variety of land uses (Cervero and Kockelman, 1997). Hence
we used a neighbourhood function and conducted a moving win-
dow analysis, as shown in Fig. 7, entropy was calculated for each
grid cell adapting the equation from Ritsema van Eck and Koomen
(2008) in Eq. (1) as shown below.
P Bach et al. (2006), a healthy mix of land uses w.r.t. residential land
 i Q lui  lnðQ lui Þ
LU d ðiÞ ¼
lnðnÞ
Slui
Q lui ¼
Si
137
lui is the land use class (1, 2, . . . , n) in window of analysis i, Q lui the
share of specific land use within the analysis window i, Slui the total
area of the specific land use within the analysis window i, Si is the
total area of the analysis window i.
The formula is only valid for areas with more than one land use
type otherwise it returns a zero, indicating no diversity at all. The
resulting land use diversity map for the City Region is shown in
Fig. 8.
5.3. Land use mixed-ness indicator
Calthorpe (1993, p. 17), has already mentioned that ‘‘Our
investments in transit must be supported by land use patterns
which put riders and jobs within an easy walk of stations.’’ Since
walkability is very important for high TOD levels, it is an important
indicator to be measured. Walkability is encouraged primarily by
urban design and land use mix (Molster and Schuit, 2013). Zhang
and Guindon (2006, p. 159), quote from (CST, 2002) that ‘‘. . .to sup-
port a high degree of walking and cycling, residential and employ-
ment uses must be mixed on a very small scale.’’ This notion of
mixed-ness is different from that of diversity and centres around
how the residential land use is supported by other land uses such
as commercial, industrial, institutional put together. According to
ð1Þ use can encourage people to carry out their daily chores on foot or
on cycle. To calculate the mixed-ness of residential land use w.r.t
commercial, industrial or other land uses, we adapted the formulae
ð2Þ used by Zhang and Guindon (2006) to our context and for our use
of vector land use data. Once again, a window of analysis as shown
138 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143

Fig. 10. Number of business establishments.

Table 2
Criterion weights and standardisation table.

Criteria Rank Resulting Indicator Contribution to Description Standardization

order weights criterion’s method a
weight
Level of density 1 0.35 Residential density 50% Higher residential and commercial densities are required Benefit/goal
Commercial density 50% for more efficient public transport Benefit/interval
Level of land use 1 0.35 Land use diversity 100% Higher diversity of land uses reduces vehicular trips Benefit/interval
diversity and enhances the liveliness and safety of a place where
people socialize
Level of mixed use 2 0.20 Mixed use 100% Higher mixed-ness of land uses (w.r.t residential Combination/U-
land use) encourages higher degree of walk and cycle for shape up
non-work trips
Level of economic 3 0.10 Number of business 100% The higher number of business establishments Benefit/interval
development establishments represents higher level of economic development
and hence higher TOD levels
a
There are various standardization methods in ILWIS depending on how the indicator’s value affects the final index value. For, e.g., a ‘Benefit’ method indicates that an
increase in indicator’s value will also increase the index’s value. This also means they are directly proportional to each other. A ‘Cost’ method indicates an inversely
proportional relationship between an indicator and the final index value. Refer ITC (2007) for explanation on other standardization methods.

in Fig. 7 was used to calculate the level of mixed-ness of land uses resulting indicator map for City Region is shown in Fig. 9 that shows
in a grid cell. The formula used by us is as follows in Eq. (3): that index values are low outside urban areas. The dark green patches
P showing null levels of land use mixed-ness also indicate that there
\i Sc
MI ¼ P ð3Þ are no residences found in those areas. It is interesting to see residen-
\i c þ Sr Þ
ðS tial development clusters around the train tracks and stations.
where MI is the ‘Mixed-ness Index’, Sc shows the sum of the total area
under commercial, industrial and institutional land uses and Sr rep- 5.4. Number of business establishments
resents the sum of the total land area under residential land use with-
in the analysis neighbourhood i. For grid cells, where residential land Economic activity has been proposed as an indicator of TOD lev-
use was absent, the formula was set to return a value of zero. The els by Renne and Wells (2005) and Evans and Pratt (2007) since a
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143 139

Fig. 11. Potential TOD Index scores in the City Region.

Fig. 12. Potential TOD Index hot spots (clusters of high index scores).

high economic activity indicates higher travel demand and hence
higher levels of TOD in the area. Number of business establish-
ments in the area is one of the indicators of economic activity
mapped as shown in Fig. 10 using the data from CBS. As expected,
it was found that the numbers are higher in urban areas implying
that urban areas have higher economic development and possibly
higher TOD levels.
6. Computing Potential TOD Index using SMCA
After computation of all indicators in ArcGIS, they were as-
sessed together to compute the Potential TOD Index value for each
grid cell. For spatial analysis of all the multiple criteria, ILWIS 3.3
(ITC, 2007) was used for which, indicator maps were imported in
raster format under the four criteria. The indicators were then
140 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 13. Hot spots and their proximity to a train station.
standardised and weights were assigned to them according to their
importance to the concept of TOD. Earlier works such as those by
Evans and Pratt (2007) and Bach et al. (2006) could be used to find
the threshold values or reference ranges for all indicators. The
weights are an indication of the strength of the relationship be-
tween the indicator and TOD. Weighing of indicators can be done
by one or many groups of stakeholders. Different stakeholders in
TOD, such as private developers, community, planners, researchers
and others, will rank the indicators differently based on their own
perceptions. We weighed the indicators as the researchers and to
base our views we referred to literature on the subject (Dittmar
and Poticha, 2004; Renne and Wells, 2005; Evans and Pratt,
2007; Bach et al., 2006). At first, we ranked the criteria and accord-
ingly translated these ranks to weights for rank k, in ILWIS, using a
rank sum method based on the following Eq. (4):
nþ1k
WðkÞ ¼ Pn ð4Þ
þ 1  1Þ
l¼1 ðn
where W(k) is the normalized weight for the criterion with rank ‘k’,
n is the total number of criteria in the set (52°North, 2012). The ‘l’ is
the index of summation that takes the value from 1 to n. The rank
sum method is also used to normalize the weights of the individual
indicators within the criteria. Table 2 shows the ranked order, cal-
culated weights of criteria and standardization method together
with the relevant indicators. For all indicators except for one, a
higher value would lead to a higher TOD Index value and hence they
have been standardised using the ‘Benefit’ method. ‘Benefit’ stan-
dardisation indicates that there is a positive linear relationship be-
tween the indicator value and the index value. ‘Cost’
standardisation means a negative linear relationship between the
indicator value and the index value. For standardisation of the
mixed-land use indicator, a median value of 0.5 (with values
ranging from 0 to 1) is considered the best since it implies a perfect
mixedness with equal share of residential and other land uses.
Hence, this indicator has been standardised in a way that as its va-
lue increases, it ‘benefits’ the TOD Index value. After it reaches 0.5,
any increase in indicator value will be a ‘cost’ to TOD Index value.
This combination of ‘benefit’ and ‘cost’ standardisation method
was thus adopted for this indicator. The composite Potential TOD
Index value for each and every grid cell was then computed by add-
ing up the results of multiplying the standardised indicator value
with its weight. The Potential TOD Index calculation results are dis-
cussed in the following section.
7. Potential TOD Index results and their interpretation
Fig. 11 shows the map of the computed Potential TOD Index
across the City Region. As a result of the SMCA process, the com-
posite index value can range from 0 to 100 (when all criteria score
highest in their range). The maximum score for the City Region
turned out to be 60, implying that overall TOD levels are not very
high in the City Region.
As can be seen from the map, most of the high index values are
in the urban agglomerations of Arnhem and Nijmegen and also
near the existing transit stations. A visual inspection of Fig. 11
shows that some form of clustering is visible. In the following sec-
tion this level of clustering is analysed using spatial statistical ap-
proaches. After the spatial statistical analysis, the clusters of high
index values are mapped and further analysed.
Firstly, the Potential TOD Index data were tested using spatial
autocorrelation statistics to determine the degree of spatial depen-
dency of the results. To this end, the Global Moran’s I statistic was
applied to the Potential TOD Index scores, where the null-hypoth-
esis indicated complete spatial randomness of results. We used a
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
Fig. 14. Hot spots under different scenarios.
distance band of 500 m as a threshold distance, where cells within
500 m of a central grid cell were considered as neighbours and
were compared with the average scores of the whole region. This
resulted in a Moran’s I score of 0.81 with a very high z-score value
(174), which indicates the presence of clustering in the region.
To further analyse spatial patterns and locations of the actual
clusters, we employed cluster mapping using selected spatial sta-
tistics. Three local statistical techniques of spatial association were
used that are capable of detecting spatial concentration of high or
low value locations or ‘hot’ or ‘cold’ spots as suggested by Páez and
Scott (2005). These statistical techniques are Getis and Ord’s Gi
statistic (Getis and Ord, 1992), Anselin’s Local Indicators of Spatial
Association (LISA) and Anselin Local Moran’s I index (Anselin,
1995). The Getis-Ord Gi statistic considers a feature as a hot spot
when it has a high value and is also surrounded by neighbours hav-
ing high values, whereby the sum of the values of the feature and
its neighbours compared to the total sum of all feature values
across the study area reflects a statistically significant difference.
Anselin’s LISA on the other hand is used to provide information
on spatial dissimilarities between neighbouring features. Both ap-
proaches can be used for comparison.
To map the location of clusters and to find their intensity, Getis-
Ord Gi and Anselin Local Moran’s I were computed for every grid
cell across the City Region. In addition to this, GeoDa software and
SPSS were used to construct the Anselin Local Moran’s I scatter plot
to help classify and interpret the related statistics.
Using the Getis Ord Gi⁄ and Anselin Local Moran’s I statistics,
the hot spots were accordingly identified. These statistically signif-
141
icant hot spots were identified at a confidence level of 95% as
shown in Fig. 12. It can be seen that there are vast areas of light yel-
low across the region which could either mean that Potential TOD
Index values are dispersed, randomly distributed or that these
areas are fairly homogeneous non-urbanized areas characterized
by low variation in TOD Index values among the cells. For the City
Region, the latter seems more likely even though, in either case
these areas have overall low TOD levels and are hence not of
interest. The areas in red are the hot spots that have a high
Potential TOD Index value. Each cluster or hot spot of grid cells
has a high average index value which is significantly higher than
the average of the entire region. It was also observed that the
Anselin Local Moran’s I recognized more cells as hot spots. Since
this method relies on similarity more than magnitude of the
values, it represents the clusters better and hence its results
were used for further analysis.
8. Recommending hot spots for transit connectivity
Now that the hot spots have been identified, it remains to iden-
tify them for transit connectivity, ensuring in the process that areas
with existing train connectivity are not included in our recommen-
dations. Thus, in this step, we remove those parts of hot spots that
fall within a walkable distance of 800 m from a train station and
hence are close to an existing transit node. The excluded areas will
be studied separately in Actual TOD Index calculations, when we
study the levels of TOD around existing train stations. Those calcu-
lations will also consider other indicators that were not a part of
142 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143
the Potential TOD Index calculations. The final clusters of high Po-
tential TOD Index value are shown in Fig. 13.
As can be seen in the figure, there are many clusters all across
the region. These hot spots or clusters are recommended to the City
Region for further detailed demand assessment, technical feasibil-
ity analysis, network studies, financial – economic feasibility anal-
yses and others. These analyses will be important to make
decisions about which location(s) within the clusters should be
connected by transit and how. The chosen locations/areas could
be either connected by train or by BRT since they provide better
transit connectivity than a regular bus as mentioned in Section 3.
9. Sensitivity analysis
We also conducted a sensitivity analysis to see how sensitive
our clusters are to the variation in weights. In addition to our initial
results, which we call our base scenario, we created three alternate
scenarios with different sets of weights to the criteria used in our
Potential TOD Index calculations. Those scenarios are given in
Table B in the Appendix B. Upon recalculation of our Potential
TOD Index values, we also analysed spatial statistical clustering
to find out the new hot-spots for each alternate scenario. The
comparison of resulting hot spots under all four scenarios is given
in Fig. 14.
From Fig. 14, we can infer that the red areas are least sensitive
to changed scenarios and we can attach a higher level of confi-
dence to our prediction of hot spots in those areas. In contrast, blue
areas are most sensitive to changes in scenarios since they do not
appear in any of the alternate scenarios. Thus, in further planning
and decision making processes more caution needs to be taken
with these highly sensitivity hotspots if these come out in the later
planning process as areas of political or other such priority.
10. Conclusions and discussion
The Arnhem-Nijmegen City Region in the Netherlands, seeks to
stimulate sustainable development in its territory and TOD is one
of the tools to achieve this goal. To plan for TOD, it is important
to be able to measure the existing levels of TOD, which can be done
using a TOD Index. We believe that two indices are required for
wholesome TOD planning for a region – Potential TOD Index and
Actual TOD Index. In this paper, however, we have presented our
work on the former only. Under the SMCA framework and using
multiple spatial criteria, Potential TOD Index values for the City Re-
gion were calculated. The index values were found to be higher in
the urban areas of Arnhem and Nijmegen, which implies that exist-
ing TOD levels are higher in the urban areas, while the rest of the
region generally scores low on TOD levels. The results were further
analysed and clusters of high index values, also known as hot spots,
were identified and mapped. These hot spots were further analysed
to identify those locations where the TOD levels are high but train
connections are further away than 800 m. It is recommended that
these locations are considered for better transit connectivity
through train or BRT systems. On conducting the sensitivity analy-
sis, we found that some areas of our hotspots (i.e. the blue areas in
Fig. 14) are highly sensitive to changed scenarios and hence need
Table A
Data used with their respective formats and sources.
Data used
Number of houses per neighbourhood
Number of commercial establishments
Land use data
Station locations
Administrative boundaries (municipal, district and neighborhood boundaries)
Rail network
to be considered with more caution. At the same time, we found
that some areas of our hotspots (i.e. the red areas in Fig. 14) are
more robust and hence we can attach higher confidence to them.
Further decisions and planning efforts require more research and
discussions with the City Region officials.
The novelty of our work lies in the development of a spatial tool
capable of assessing multiple criteria and arriving at a comprehen-
sive TOD Index score representing existing levels of TOD for an
area. Using the results of the TOD Index scores, clusters were iden-
tified and locations recommended for transit connectivity.
The Potential TOD Index, however, presents only one side of our
two-sided approach to the type of TOD Planning that we envisage as
explained in Sections 1 and 2. In the presented work, using our spa-
tial tools, we have been able to identify and map those particular
areas in the City Region, where urban development has high TOD
levels, but has poor transit connections. The next step is to use
the same methodological framework, as described in Section 2,
and develop a tool for measuring Actual TOD Index values for areas
around existing train stations. For Potential TOD Index calculations,
data paucity was an issue and it is hoped that for Actual TOD Index
calculations more data will be collectable. Statistical data at a more
detailed level would also lead to higher accuracy in results. Using
these two indices, one can plan for higher transit connectivity in
the region and higher TOD levels around existing transit nodes.
We have used criterion weights based on our understanding of
how strongly different indicators affect TOD. However, in order to
reflect different stakeholder groups’ priorities it would be ideal to
hold stakeholder consultations and record their priorities. Through
such a participatory workshop in SDSS environment, as discussed in
Section 2, our recommendations can also be discussed to find out
the stakeholders’ opinions and identify possible divergences, if
any. The framework of our work and the tool that we have devel-
oped is uncomplicated and repeatable for other case studies in
other countries too. None of the existing features of our case study,
except for availability of data, are pre-requisites for our tool to
quantify TOD levels in a region. Since local priorities can differ from
case to case, the stakeholder group, interested in measuring the
TOD in their region or city can weigh the criteria according to their
priorities. Depending on the urban structure, the TOD levels, once
measured, can guide proper planning and investment decisions.
Acknowledgements
The authors wish to express their gratitude to Mr. Jorn Mat-
thijsse, Ex-Project Leader/Advisor (Housing and Spatial Planning)
of the City Region Arnhem and Nijmegen, for sharing the available
data and assisting in further data collection. We also wish to thank
our reviewers for their insightful comments that helped us im-
prove this paper. This work was carried out as a part of the re-
search supported by the Erasmus Mundus 2 LOT 11: Asia
Regional program.
Appendix A
See Table A.
Format of data Source of data
Statistical data CBS (Central Bureau for Statistics, Netherlands)
Statistical data CBS (Central Bureau for Statistics, Netherlands)
GIS (vector) The City Region
GIS (vector) Open Street Map (OSM)
GIS (vector) CBS and the City Region
GIS (vector) ESRI Top 10 NL and Open Street Map (OSM)
 Y.J. Singh et al. / Journal of Transport Geography 35 (2014) 130–143 143

Appendix B

See Table B.

Table B
Base and alternate scenarios – sensitivity analysis.

Criteria Base scenario Scenario 1 Scenario 2 Scenario 3

Ranks Weights Ranks Weights Ranks Weights Ranks Weights
Level of density 1 0.35 1 0.4 2 0.3 1 0.25
Level of diversity 1 0.35 2 0.3 1 0.4 1 0.25
Level of mixed use 2 0.2 3 0.2 3 0.2 1 0.25
Level of economic Development 3 0.1 4 0.1 4 0.1 1 0.25

References Getis, A., Ord, J.K., 1992. The Analysis of Spatial Association by Use of Distance
Statistics. Geogr. Anal. 24, 189–206.
Hale, C., Charles, P., 2006. Making the Most of Transit Oriented Development
52°North, 2012. ILWIS 3.x User’s Guide.
Opportunities. In: Manager, A.C. (Ed.), 29th Australasian Transport Research
Abella, E.A.C., Westen, C.J., 2007. Generation of a landslide risk index map for Cuba
Forum The Australasian Transport Research Forum. Gold Coast, Queensland.
using spatial multi-criteria evaluation. Landslides 4, 311–325.
Hoffman, 2006. Transit: Can it sustain TOD? The Mission Group.
Anselin, L., 1995. Local Indicators of Spatial Association—LISA. Geogr. Anal. 27, 93–
Howe, A., Glass, Geoff., Curtis, Carey., 2009. Retrofitting TOD and managing the
115.
impacts: the case of Subi Centro. In: Curtis, C., Renne, J.L., Bertolini, L. (Eds.),
Arrington, G.B., 2009. Portland’s TOD evolution: from planning to lifestyle. In:
Transit Oriented Development: Making it Happen. Ashgate e-Book, pp. 65–74.
Curtis, C., Renne, J.L., Bertolini, L. (Eds.), Transit Oriented Development: Making
ITC, 2007. ILWIS 3.3 User manual. International Institute for Geo-Information
it Happen. Ashgate e-Book, pp. 109–124.
Science and Earth Observation, Enschede, The Netherlands, <http://www.itc.nl/
Bach, B., Hal, E.v., Jong, M.I.d., Jong, T.M.d., 2006. Urban Design and Traffic – A
ilwis>.
Selection from Bach’s Toolbox. CROW, The Netherlands.
Keshkamat, S.S., Looijen, L.M., Zuidgeest, M.H.P., 2009. The formulation and
Balz, V., Schrijnen, J., 2009. From Concept to Projects: Stedenbaan, The Netherlands.
evaluation of transport route planning alternatives: a spatial decision support
In: Curtis, C., Renne, J.L., Bertolini, L. (Eds.), Transit Oriented Development:
system for the Via Baltica project, Poland. J. Transp. Geogr. 17, 54–64.
Making it Happen. Ashgate e-Book, pp. 75–90.
KiM, 2013. Mobiliteitsbalans 2013. Kennisinstituut voor Mobiliteitsbeleid (KiM).
Belzer, D., Autler, G., 2002. Transit oriented development: moving from rhetoric to
The Netherlands.
reality. The Brookings Institution Center on Urban and Metropolitan Policy and
McKone, J., 2010. Cities in Focus: Curitiba.
The Great American Station Foundation.
Mitchell, A., 2009. ESRI Guide to GIS Analysis, vol. 2. Spatial measurements and
Beukes, E.A., Vanderschuren, M.J.W.A., Zuidgeest, M.H.P., 2011. Context sensitive
statistics, ESRI, Redlands.
multimodal road planning: a case study in Cape Town, South Africa. J. Transp.
Molster, A., Schuit, S., 2013. Voetsporen rond het station. Het Nationaal
Geogr. 19, 452–460.
Verkeerskunde Congres, Netherlands.
Beuthe, M., 2002. Transport Evaluation Methods: From Cost-Benefit Analysis to
Newman, P., 2009. Planning for transit oriented development: strategic principles.
Multicriteria Analysis and the Decision Framework. In: Giorgi, L., Pearman, A.
In: Curtis, C., Renne, J.L., Bertolini, L. (Eds.), Transit Oriented Development:
(Eds.), Project and Policy Evaluation in Transport. Ashgate Publishing Limited,
Making it Happen. Ashgate e-Book, pp. 13–22.
England, pp. 209–241.
Páez, A., Scott, D., 2005. Spatial statistics for urban analysis: a review of techniques
Calthorpe, P., 1993. The Next Americam Metropolis – Ecology. Community and the
with examples. GeoJournal 61, 53–67.
American Dream, Princeton Architectural Press, Canada.
Parker, T., McKeever, M., Arrington, G., Smith-Heimer, J., 2002. Statewide Transit-
Cascetta, E., Pagliara, F., 2009. Rail friendly transport and land-use policies: the case
Oriented Development Study: Factors for Success in California.
of the Regional Metro System of Naples and Campania. In: Curtis, C., Renne, J.L.,
Renne, J.L., Wells, J.S., 2005. Research Results Digest 294 – Transit Oriented
Bertolini, L. (Eds.), Transit Oriented Development: Making it Happen. Ashgate e-
Development: Developing a Strategy to Measure Success. Transport Research
Book, pp. 49–64.
Board.
Cervero, R., Kockelman, K., 1997. Travel demand and the 3Ds: density, diversity and
Rijkswaterstaat, 2010. Mobiliteitsonderzoek Nederland 2009 Tabellenboek.
design. Transp. Res. 2, 199–219.
Ministerie van Verkeer en Waterstaat/Rijkswaterstaat, The Netherlands.
Cervero, R., Murakami, J., 2008. Rail + property development: a model of sustainable
Ritsema van Eck, J., Koomen, E., 2008. Characterising urban concentration and land-
transit finance and urbanism. UC Berkely Centre for Future Urban transport.
use diversity in simulations of future land use. Ann Reg Sci 42, 123–140.
Cervero, R., Murakami, J., 2009. Rail and property development in Hong Kong:
Schlossberg, M., Brown, N., 2004. Comparing transit-oriented development sites by
experiences and extensions. Urban Stud 46, 2019–2043.
walkability indicators. Transit Plan. Dev. Manage. Perform. Mark. Fare Policy
Chorus, P., 2009. Transit oriented development in Tokyo: the public sector shapes
Capacity Qual. Serv., 34–42.
favourable conditions, the private sector makes it happen. In: Curtis, C., Renne,
Shastry, S., Zuidgeest, M.H.P., Munshi, T., 2011. Spatially assessing transit oriented
J.L., Bertolini, L. (Eds.), Transit Oriented Development: Making it Happen.
development (TOD) in Ahmadabad, India. The 1st Conference of Transportation
Ashgate e-Book, pp. 209–224.
Research Group India: CTRG 2011, Bangalore, India, p. 12.
CST, 2002. Sustainable Transportation Performance Indicators (STPI) Project (Phase
Singh, Y.J., Zuidgeest, M.H.P., Flacke, J., Maarseveen, M.F.A.M.v., 2012. A design
3). Centre for Sustainable Transportation (CST).
framework for measuring transit oriented development. In: Longhurst, J.W.S.,
CTOD, 2009. Why Transit – Oriented Development and Why Now? Reconnecting
Brebia, C.A. (Eds.), Urban Transport XVIII WIT Press, Southampton, UK, pp. 719–
America, US.
730.
Curtis, C., 2009. Implementing transit oriented development through regional
The City of Calgary, 2004. Transit-Oriented Development: Best Practices Handbook.
plans: a case study of Western Australia. In: Curtis, C., Renne, J.L., Bertolini, L.
The City of Calgary.
(Eds.), Transit Oriented Development: Making it Happen. Ashgate e-Book, pp.
Van Wee, B., 2011. Transport and Ethics: Ethics and the Evaluation of Transport
39–49.
Policies and Projects. Edward Elgar Publishing Ltd., UK.
Dittmar, H., Poticha, S., 2004. Defining transit-oriented development: the new
Yang, P.P., Lew, S.H., 2009. An asian model of TOD: the planning integration in
regional building block. In: Dittmar, H., Ohland, Gloria. (Eds.), The New Transit
Singapore. In: Curtis, C., Renne, J.L., Bertolini, L. (Eds.), Transit Oriented
Town – Best Practices in Transit Oriented Development. Island Press, pp. 19–39.
Development: Making it Happen. Ashgate e-Book, pp. 91–108.
Evans, J.E., Pratt, R.H., 2007. Transit Oriented Development, TCRP’s Traveler
Zhang, Y., Guindon, B., 2006. Using satellite remote sensing to survey transport-
Response to Transportation System Changes Handbook series. Transport
related urban sustainability. Part 1: Methodologies for indicator quantification.
Research Board, USA.
Int. J. Appl. Earth Obs. Geoinf. 8, 149–164.
Ewing, R., Cervero, R., 2010. Travel and the built environment. J. Am. Plan. Assoc. 76,
265–294.