You are on page 1of 12

Quantification and Analysis of Land-Use Effects on Travel

Behavior in Smaller Indian Cities: Case Study of Agartala


Partha Pratim Sarkar 1 and Mallikarjuna Chunchu 2
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

Abstract: The land-use mix observed in the smaller Indian cities is peculiar, and in this context there were no past studies on how the land-
use mix influences the travel pattern. The existing indices used for quantifying the land-use mix were found to have limitations in capturing
the characteristics of land-use mix observed in the smaller Indian cities. The present study analyzed the drawbacks and limitations of the
existing indices and modified the dissimilarity and entropy indices, as well as formulated new indices suitable to quantify the mixed land use.
The objectives behind the mix quantification were to capture the land-use balance, land-use mix, and land-use complementarity. The modified
dissimilarity and entropy indices were found to be significant in explaining the variation in the trip lengths as well as preference toward
nonmotorized transport (NMT). Proposed new parameters such as the area index and mix-type index were found to be significant in explain-
ing the variability observed in the trip length and the nonmotorized mode choice. From the elasticity analysis it has been observed that a slight
change in the land-use mix significantly affects the travel patterns. Further, from the estimated multinomial logit (MNL) models it has been
observed that the mixed land-use parameter was found to be significant in explaining the choice of public transport and nonmotorized modes.
DOI: 10.1061/(ASCE)UP.1943-5444.0000322. © 2016 American Society of Civil Engineers.
Author keywords: Land-use mix; Travel behavior; Intersection density; Entropy index; Dissimilarity index (DI); Nonmotorized trips;
Area index.

Introduction (Cervero 1991; Frank and Pivo 1994; Cervero and Radish 1996;
Kockelman 1997; Rajamani et al. 2003; Chapman and Frank
Land-use mix is one of the important measures of land-use devel- 2004; Frank et al. 2009; Vance and Hedal 2007). The other major
opment patterns and it refers to the diversity of land uses within an factors influencing travel patterns are the socioeconomic and road
area. When diverse land uses exist in a given area (generally a cen- network characteristics (Cervero and Kockelman 1997; Lee et al.
sus tract or municipal ward) it is expected that many trips originat- 2014; Wang et al. 2013). The travel-pattern changes observed in
ing from that area may have destinations in the same area. Land-use smaller Indian cities may be attributed to both changing socioeco-
mix is generally characterized using various indices such as en- nomic factors and land-use characteristics. In this context there
tropy, dissimilarity index, Gini coefficient, Herfindahl index, etc. were no past studies, hence it is not clear whether the changing
The relationship between the land-use mix and travel behavior land-use characteristics have a significant effect on the changes
has been widely studied and evaluated by researchers. Land-use in travel patterns.
planning also has significant impact on many travel-related param- Ewing et al. (2010) have pointed out that smaller mixed-use
eters. Appropriate land-use planning results in the overall reduction developments in walkable areas with good transit access generate
of travel (expressed in kilometers traveled by vehicles or persons) significantly high shares of walking and transit trips. In smaller
and a significant modal shift from the personalized modes to public Indian cities the smaller mixed-use developments are prevalent.
and nonmotorized transport modes. Land-use mix in these cities can be observed at the building level,
A report by the Ministry of Urban Development, Government of street level, and zonal level. Many buildings contain shops, residen-
India (2008) states that there will be a significant increase in the ces, and offices, which is a common feature in smaller Indian cities.
share of private transport modes such as cars and motorized Similarly, offices, schools, residences, and shops can be seen on
two-wheelers in smaller Indian cities. The report predicts the share the same street. These two features automatically lend the mixed-
of private transport to go up from 57% in 2007 to 72% by 2031. land-use nature to a zone. Another important feature is the size of
This report has also mentioned the decreasing share of public trans- a particular land-use type. Shops with a 5 m2 floor area can be
port modes in most of these smaller cities. Also, smaller Indian observed at many locations spread across the city. Trips generated
cities are rapidly growing, and significant changes have also been by the smaller establishments are mostly nonmotorized, hence their
observed in travel patterns. Several past studies have found that existence needs to be captured while quantifying the land-use mix.
the land-use characteristics significantly influence travel patterns City administration in the smaller cities used to be flexible in
allowing the mixed land use, though different zones are demarcated
1
Assistant Professor, Dept. of Civil Engineering, National Institute of for different land uses (Toutain and Gopiprasad 2006). Because
Technology Agartala, Agartala, Tripura 799046. India (corresponding of recent changes in socioeconomic character and the associated
author). E-mail: ps_partha@yahoo.com issues, the city administrators have started enforcing the land
2
Associate Professor, Dept. of Civil Engineering, Indian Institute of
use, if not completely.
Technology Guwahati, Guwahati, Assam 781039, India.
Note. This manuscript was submitted on December 19, 2014; approved Entropy (Cervero 1989) and dissimilarity index (Cervero and
on November 3, 2015; published online on March 21, 2016. Discussion Kockelman 1997) are widely used indices for quantifying mixed
period open until August 21, 2016; separate discussions must be submitted land use. Both these indices characterize the zones based on the
for individual papers. This paper is part of the Journal of Urban Planning extent of land-use mix. Many of the past studies attributed the zonal
and Development, © ASCE, ISSN 0733-9488. land use the mix index to the households located in that zone and

© ASCE 04016009-1 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


studied the correlation between the land-use parameters and the
households’ travel parameters (Cervero and Kockelman 1997;
Kockelman 1997). When the dissimilarity index has been used
for quantifying the land-use mix, a cell size of one hectare has been
adopted. This cell size is suitable, perhaps, when the mixed land use
is not observed at the building level and street level. Bigger cells
also obscure the relevance of smaller commercial establishments.
Some of researchers have considered the household-specific en-
tropy while relating the travel and land-use parameters (Cervero
and Duncan 2003; Pushkar et al. 2000; Sun et al. 1998). Ewing
and Cervero (2010) have reported that the land-use mix measured
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

in terms of distance to store has higher elasticity value than the


other diversity measures when used for analyzing walking trips.
This result also indicates the need to consider the household-centric
land-use-mix parameters.
From the review it can be seen that land-use parameters have
significant impact on travel parameters. Some researchers have sug-
gested using the household-specific land-use parameters, whereas
the zonal-specific land-use parameters are widely used. This study
made an attempt to understand the effect of mixed land use on
travel parameters in the context of smaller Indian cities. To under-
stand the relationship between land use and travel parameters, the
land-use mix must be quantified using appropriate land-use param-
eters. Various indices that quantify the mixed land use have been
formulated, and some commonly used land-use indices have been
modified suitably. Travel parameters related to work and shopping
trips have been used to analyze the effects of land-use mix. Detailed
analysis of land-use mix quantification using various existing Fig. 1. Land-use map of the study area
parameters and the proposed parameters is provided in the sub-
sequent section.

land use. Buildings meant only for office use have been considered
Study Area Details and Data Collection as service areas. Land-use mix at the building level is negligible
in the present study area and has not been considered. Wherever
In this study, Agartala, capital of the state of Tripura, located a particular building is used for both commercial and residential
in northeastern India, has been chosen as the study area. Agartala purposes (for example ground-floor shops and other floors for res-
municipality consists of 35 municipal wards, divided mostly idential), that building area was considered as commercial. If more
for administrative purposes. This city is the second-largest in than one floor was used for commercial purpose, only the area
northeastern India, after Guwahati, in terms of municipal area corresponding to the ground floor was taken into consideration.
(58.84 km2 ). According to census data from 2011, the population In collecting the land use data, global positioning system (GPS)
of Agartala city was 399,688 with a population density of 6,793 device (Juno SB, Trimble, California) was used for digitizing
persons per km2. The region falls under subtropical and temperate big markets, offices, and shopping complexes. For isolated small
climatic zones. The average annual rainfall of Agartala city is shops, the location was digitized using the GPS and the floor area
220 cm. A major part of the municipal area consists of flat terrain, was calculated using a measuring tape. All the commercial, serv-
and with respect to land-use pattern, industrial land use is relatively ices, educational, industrial, and other developed land uses were
low. Most of the commercial areas are located on either side of the subtracted from the total built-up area to get the residential land
major roads. Small retail shops are present all over the study area use. ArcGis 10.1 software was used for storing and analyzing
beside the roads. Fig. 1 is a land-use map of Agartala city. There are the land use and travel data.
small retail shops on the side of the road that cannot be seen in the Travel data have been collected through a household survey
figure due to the scale of the map. conducted in the study area during March–September 2012. Sam-
Availability of the digitized land-use map is a major problem ple size, in terms of households, is about 1% of the total number
in studying the effect of land-use mix on travel behavior in smaller of households in the study area, taken randomly from different
Indian cities. In this study an effort has been made to collect and municipal wards. Information related to the trips and the travel
digitize the land-use details of Agartala city. Different land-use data modes such as origin, destination, purpose of the trip, mode of
have been collected, including residential, commercial, industrial, travel, and length of the trip have been collected. Also, the socio-
service, educational, forest, agricultural, river, ponds, social welfare economic characteristics including age, gender, and years of edu-
centers, playgrounds, and vacant land. Except for forest, agricul- cation, household size, household income, vehicle ownership, and
tural, vacant, and water bodies, all the remaining the land uses the license status of the trip makers have been collected. The sam-
have been categorized into five different types by suitably merging ple was found to be representative of the overall travel pattern of
different land uses. Residential, commercial, educational, service, Agartala residents. In the collected sample, 37.2% of the trips were
and others are the five land-use types considered in the present nonmotorized and more or less a similar figure was reported in a
study. All the retail shops, including shopping complexes and report (2008) by the Ministry of Urban Development, Government
the buildings with retail shops and industries in the study area of India. Statistics of the sample data are given in Table 1. The
(Agartala municipality area) have been considered as commercial modal split of the collected data is shown in Fig. 2.

© ASCE 04016009-2 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Table 1. Summary of Socioeconomic Data Obtained from the Sample entropy for quantifying information. The concept of entropy,
Value in widely used in thermodynamics, information theory, and image
Socioeconomic characteristic percentage processing (Pun 1981; Kapur et al. 1985; Pal and Pal 1993),
provides a useful basis for quantifying the land-use mix. Detailed
Gender
Male 73.38
analysis of land-use mix quantification using various existing
Female 26.62 parameters and proposed parameters is given in the following
Percentage of individuals in the age category sections.
Up to 20 3.95
20–30 17.69
30–40 18.53 Entropy Index
40–50 24.08 The entropy index is the most widely accepted and commonly
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

50–60 22.77 used index for quantifying the land-use mix. Entropy measure
>60 12.98
was first used to quantify land-use mix by Cervero (1989). This
Percentage of households having a driving license
Having 49.5 index quantifies the homogeneity of land use in a given area.
Not having 50.5 Entropy is expressed as
Individuals’ education levels by percentage X lnðPj Þ
Illiterate 0.19 Entropy ¼ Pj × ð1Þ
Primary level 13.18 j
lnðJÞ
Secondary level 20.51
Higher secondary level 15.62 where J = total number of developed land uses considered; and
Undergraduate level 33.02 Pj = proportion of the area of jth land-use category found in
Postgraduate level 17.31 the tract being analyzed.
Above postgraduate level 0.19
Because the entropy is normalized using natural logarithm of
Car ownership 13.83
MTW ownership 44.21
the number of land uses, its value lies between 0 and 1, where
Monthly household income (in Indian rupees) 0 represents homogenous land use, and one indicates the tract
0–2,000 0.28 of land is equally distributed across all the land-use types.
2,001–10,000 33.96
10,000–20,000 30.48
20,000–50,000 24.46 Limitations of Entropy Index
>50,000 10.82 Main drawback of the entropy index is that it could not represent
the intensity of land-use mixing properly. Entropy index takes the
same value for two different scenarios having different land-use
patterns if the proportion of land-use mix is the same. Also, when
Land-Use Mix in the Study Area the previously discussed approach is used (in calculating the en-
tropy for census tract or municipal ward as shown in Fig. 5) in
Fig. 3 shows the land-use composition details observed in Agartala quantifying the mixed land use observed in smaller Indian cities,
city. From this figure it can be observed that the study area includes many households with varying travel behavior were characterized
a significant quantity of vacant land, agricultural land, and water with similar entropy value. To overcome this problem to some ex-
bodies. Once the vacant land and the land related to the water tent, the entropy index is measured for each sampled household
bodies and agriculture are disregarded, residential land use is pre- instead of the census tract. For this purpose, a buffer of 1 km radius
dominant and the remaining land uses are relatively small (Fig. 4). (Fig. 6) was created around each of the sampled households.
Diversity indices were used for quantifying the dissimilarities Entropy index values were computed for the buffer area created
in different disciplines of study. Shannon (1948) has proposed around the sampled households. Even after this modification,

Fig. 2. Percentage of chosen mode: (a) work trips; (b) shopping trips

© ASCE 04016009-3 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Land-use distribution observed in the study area

Fig. 4. Land-use distribution of the study area after excluding the va-
cant land and the land related to agriculture and water bodies

the entropy index may take the same numerical value if the sampled
households are very close to each other.

Dissimilarity Index Fig. 5. Entropy index measured for some of the census tracts of the
The dissimilarity index (DI) was used to compute the dissimilarity study area
among the grid cells, constituting a tract (Cervero and Kockelman
1997). According to Cervero and Kockelman (1997), the dissimi-
larity index is calculated based on the points awarded to each ac- X1X
8
X ik
tively developed hectare cell on the basis of the dissimilarity of its Dissimilarity index ¼ ð2Þ
land use from those of eight adjacent hectare cells (Fig. 7). The k
K i
8
average of these point accumulations, across all the active hectares
in a tract, is the dissimilarity index for that tract. It is calculated where K = number of actively developed grid-cells in a
using the following equation: census tract or municipal ward; X ik ¼ 1 if land-use category of

© ASCE 04016009-4 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Points alloted to 100 × 100 m cells for estimating DI at tract


level (Ward 22, AMC) and estimated DI for the census tract

Similarly, for all the developed cells (i.e., cells without vacant
lands) the DI was determined and averaged for the entire census
tract. Similar to the entropy index, DI approaches the value of 1
as the land-use mix increases.
Fig. 6. A 1,000-m buffer created around a sampled household for
calculating the entropy index
Limitations of DI
As explained earlier, the points awarded to each actively developed
neighboring grid-cell differs from the central grid-cell; and X ik ¼ cell play a major role in the calculation of DI for a given tract of
0, otherwise. land. Points were awarded based on the comparisons of the land
In calculating the DI, a tract of land is divided into actively uses of the eight neighboring cells with that of the subject/central
developed land parcels of uniform size known as cells. Thus, cell. As shown in Fig. 7 two different land-use configurations
the DI presents more information about the type or intensity of mix- around a residential land use: (1) having three commercial, three
ing compared to the entropy index. The first step in calculating the service, and two residential land uses out of eight neighboring cells;
DI is to find the points to be awarded to each cell based on the and (2) having six commercial land uses and two residential land
comparisons of the land use of the subject cell with that of the eight uses out of eight neighboring cells, resulting in 6=8 points being
neighboring cells. Fig. 7(a) represents a tract of land divided into awarded to the central cell.
cells with each cell representing a land use (R = residential, S = From this it can be seen that DI does not consider the type of
service, C = commercial). The central cell is surrounded by three land use of the adjacent cell, thus neglecting one of the important
commercial and three service land uses, which are different from determinants of travel behavior, namely, land-use interaction be-
its own land use. Hence, the central cell is awarded 6=8 points. tween adjacent cells. Also, DI does not consider the mix of the land
uses. Thus, the DI represents the dissimilarity of the adjoining cells
but does not incorporate the information about number of land-use
types around the central cell. To overcome this drawback it is nec-
essary to incorporate the information of number of land-use types
around the cell as well as the interaction between the adjacent cells.
A new measure, the mix-type index, is proposed in this study to
consider the land-use types around the central cell.
Another drawback of DI is the cell size, using which the census
tract is divided. In many of the past studies a cell size of 100 ×
100 m was used in the calculation of DI. Each cell of the census
tracts is allotted a land-use type based on the dominant land use
observed in that cell. Fig. 8 shows the DI value measured for a
census tract using 100 × 100 m cell size for Ward 22 (Melarmath).
With reference to Fig. 9(a) the cell (100 m) in the top right corner
would be considered as residential land use as the residential area
is dominant in that cell. In smaller Indian cities like Agartala,
Fig. 7. Hypothetical land uses for awarding points to the central cell
within the residential zones many commercial and educational

© ASCE 04016009-5 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. (a) Points alloted to 100 × 100 m cell for estimating DI; (b) points alloted to 10 × 10 m cell for estimating DI

establishments can be seen in Fig. 9(a). Hence, if a bigger cell such for 1,000 × 1,000 m tract of land instead of the census tract. This
as 100 × 100 m is considered there is a possibility that the infor- may solve the problem of averaging effect related to the census
mation related to such land uses may be neglected. To represent tracts, which are relatively bigger compared to the 1,000 ×
these land uses the cell size must be reduced so that small area fea- 1,000 m tract.
tures would also get due importance, as shown in Fig 9(b).
In this study 100 × 100 m, 50 × 50 m, 20 × 20 m, and 10 ×
10 m cell sizes were used, and the frequency of cells, attributed Mix-Type Index
with different land uses, has been computed. It has been observed As discussed earlier, DI has limitations in quantifying the land-use
that a cell of 10 × 10 m could capture the effect of minor land uses mix of a census tract. To overcome this, the mix-type index param-
in a reasonable manner. When a 10 × 10 m cell size was consid- eter quantifying land use is proposed in this study. This measure
ered, there was a twofold increase in the cells attributed with com- allots points to each of the actively developed cells based on the
mercial land use. By keenly observing Fig. 8 (specifically the top mix of the land uses in the surrounding cells
right corner) it can be seen that many cells of Ward 22 are awarded
with zero points. Figs. 9(a and b) show points allotted to a part Mixed type index
of this ward, one with 100 m cell size and other at 10 m cell size.
The points attributed to a 100 × 100 m cell are zero (since all the 1X Xk
¼ ð3Þ
neighboring cells are of residential land use), but when the same K k ðNo: of distinct land uses in the study areaÞ
cell is further divided into 10 × 10 m cells [Fig. 9(b)], some of
the cells are attributed with nonzero points. Table 2 shows the fre- where xk = number of distinct land uses observed in the surround-
quency of dominant land uses corresponding to different cell sizes. ing cells k including it; and K = number of actively developed cells
In addition to the use of smaller cells, the DI values were calculated in a tract.
With reference to Fig. 7, the intensity of mixing is more in the
case of Fig. 7(a) as there are three types of land uses present in the
Table 2. Frequency of Dominant Land Uses Corresponding to Different eight adjoining cells. On the other hand, in the case of Fig. 7(b),
Cell Sizes only two distinct land uses can be seen in the eight adjoining cells.
Frequency of dominant land uses in the study area Using the proposed index for Case 1, the points allotted to the cen-
Cell sizes
(m) Commercial Residential Others Educational Service
tral cell are 3=5 (three types of land uses present and five types of
major land uses considered in this study) and the points allotted in
100 × 100 113 5,180 22 115 37 the case; and for Case 2, it is 2=5. It can be said that the new index is
20 × 20 1,880 80,787 279 1,991 721 able to incorporate the land-use mix and also information about the
10 × 10 7,425 243,124 2,593 6,274 40,039
differences in the land use around the central cell.

© ASCE 04016009-6 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Even though the previously mentioned modified indices have case) to the outside of the buffer zone—affects the mode choice. To
used smaller cells, when these indices were calculated as the aver- adequately represent the spatial and functional complementarity a
age points allocated to various cells of a 1,000 × 1,000 m tract, the new index, termed area index, has been proposed in the study.
averaged index may not represent the actual mix especially when a The area index for work trip is defined as the ratio of the work
part of the tract has a high mix and the other part is has a mix with areas in the buffer zone created around the household to the work
less land use. areas in the whole study area, including that of the buffer zone.
Fig. 5 shows a buffer zone created around the household and area
index for work trips considers the proportion of work areas in the
Area Index buffer zone. Commercial and service areas were considered as
Hess et al. (2001) have explained the need to develop an index work areas. It was considered that the residential land uses are
based on land-use functional and spatial complementarity. Land- linked by travel to the work areas, thereby incorporating the land
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

use functional complementarity ensures the consideration of use functional complementarity. The buffer created around the
origins and destinations that are likely to be linked by travel. household plays major role in ensuring that the area index considers
Land-use spatial complementarity ensures that the land uses linked the functional and spatial complementarity. The ratio when close
with travel are within adequate proximity. to 1 indicates that most of the workplaces lie in the buffer zone,
It was known that the purpose of the trip affects the mode thus more nonmotorized trips would be taken. The ratio when close
choice. Similarly, it can be said that for a trip’s purpose—spatial to 0 indicates that most of the workplaces are outside the buffer
concentration of the corresponding land use inside the buffer zone zone and more motorized trips may be taken. Area index can be
of a 1 km radius created around the trip origin (household in this expressed as

Land area; related to a trip purpose; in a buffer of 1,000 m radius around the origin
Area index ¼ ð4Þ
Total area of the land related to a trip purpose in the entire study area

Area index explains the relationship between mode choice parameters are correlated with each other. Entropy measured using
behavior and the amount of particular land-use area available in conventional approach, and using a buffer zone of 1,000 m radius,
the vicinity of the household. When area index was used for the DI measured using conventional approach, using 100 × 100 m, and
analysis of shopping trips, the index value was computed based 10 × 10 m cell sizes with 1,000 × 1,000 m tract size, mix-type in-
on the shopping space available in the buffer zone and the total dex measured using 10 × 10 m cell with 1,000 × 1,000 m tract,
shopping space in the study area. area index measured at origin of the trip using 1,000 m buffer, were
considered in this analysis. Further a variety of land-use develop-
ment pattern measures such as the distance to transit (Euclidian
Results and Analysis distance from home to bus stop), intersection density (number
of intersections, both three-way and four-way, inside 1 km grid
The effect of socioeconomic and land-use variables on travel around the origin, centeredness (distance between the origin and
parameters is analyzed in this section. Only work and shopping city center), and ratio of total road length inside the 500-m grid
trips have been considered for this analysis. A work trip indicates
were also tested but all the variables could not be incorporated in
travel from home to work, and a shopping trip indicates travel from
the same model because of cross correlation. When only socioeco-
home to a shop. Data used for the present analysis contain 546
nomic variables were included in the model, the model could
work-related trip data and 370 shopping-related data. Return trips
explain only 4.7% of the variability of trip length for a work trip.
were not considered in the present analysis. Two types of analysis
But when land use related variables were included in the model, the
have been carried out, namely, studying the effect of land-use
final model could explain 19.6% variability in the trip length.
parameter in explaining the variability of trip length (for work trips,
Though the model (Table 4) presented may have lower adjusted
mean trip length = 2.89 km and standard deviation = 2.08 km;
for shopping trips mean trip length = 1.32 km and standard (adj) (R2 ) value, all the land-use-mix parameters were found to
deviation = 1.30 km), and to analyze the effect of land use on mode be significant in explaining the variability observed in trip length
choice for work-related trips. In analyzing the former, linear regres- data. From the related literature the value of adj (R2 ) for modeling
sion models were prepared to understand the ability of land-use trip distance/vehicle mile traveled (VMT) was also found to be on
parameters in explaining the variability of the trip length. In the the lower side. This may be due to fact that all the factors that ex-
second analysis, effect of land-use parameters on the utility of plain the variability of VMT were not considered in the model.
the various modes has been analyzed using a multinomial logit Entropy measured with 1,000 m radius buffer, and DI measured
model. Table 3 describes the variables used in modeling the travel using 10 × 10 m cell with 1 × 1 km tract, were found to be more
behavior. significant than the conventional entropy index and dissimilarity
index measured for the census tract. Further, when the network
parameter of road length in the 500 m grid was considered, there
Effect of Mixed Land Use on Trip Length was significant improvement in the model. The coefficients of land-
A base linear regression model was prepared considering the socio- use mix and length of roads in the 500 m grid at trip origin are neg-
economic characteristics that have a strong influence on trip length ative, which means an increase in length of road and land-use mix
for shopping and work purposes. In the subsequent models, only near household tends to reduce the trip length. From the model, it
one land-use parameter entered the model at a time, as the land-use can be seen that older people generally have shorter trip lengths.

© ASCE 04016009-7 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Final model includes
Mix-type index area index at origin,
Table 3. Variables Used in Modeling Travel Behavior

(10 × 10 m cell in (10 × 10 m cell in and road length in

−57.786 (−6.843)
500 × 500 m tract

−0.016 (−2.342)

−0.898 (−2.636)
4.445 (10.407)

0.093 (3.690)
0.308 (2.367)
Serial

336.17
0.205
number Variables Description
1 Age Age in years
2 Education Years of education
3 Gender Dummy variable
1, for male respondent

−2.317 (−6.987)
1 × 1 km tract)

−0.02 (−2.682)
0, for female respondent

4.272 (12.466)

0.499 (3.754)
4 License 1, for respondent having license

163.83
0.124
0, for respondent not having


license
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

5 Family size Number of members in the


family
6 Gross monthly 0–2,000 (1)

−2.290 (−7.086)
1 × 1 km tract)

−0.02 (−2.682)
4.228 (11.466)
income in rupees 2,001–5,000 (2)

0.499 (3.754)
(relevant codes 5,001–10,000 (3)

168.09
0.126
used is shown in 10,001–15,000 (4)

DI


the parenthesis) 15,001–20,000 (5)
20,001–30,000 (6)
30,001–40,000 (7)
40,001–50,000 (8)

−0.026 (−3.575)

−2.038 (−6.619)
(1,000 m buffer)
Base model with land-use parameters

4.784 (13.356)

0.417 (3.113)
50,001–70,000 (9)

Entropy
70,001–90,000 (10)

148.94
0.117


90,001–150,000 (11)
>150,001 (12)
7 Bicycle ownership Dummy variable
1, for households having a

(cell size of 100 m Area index at origin


bicycle

−0.024 (−3.128)

−2.224 (−7.618)
(1,000 m buffer)
0, for households having no

3.816 (8.783)

0.047 (1.874)
0.383 (2.834)
bicycle

191.49
0.137
8 Motorized two-wheeler Dummy variable


(MTW) ownership 1, for households having
motorized two-wheeler
0, for households having no
motorized two-wheeler
−0.030 (−3.984)

−0.893 (−2.589)
9 Time In vehicle travel time
in census tract)
4.201 (11.768)

0.469 (3.396)
10 Cost Fare or fuel cost

21.277
0.057
11 Area index Land-use parameters obtained
DI


Entropy with buffer from GIS analysis
Entropy index
(census tract)

Note: Parameters are tested at 5% significance level; number of observations = 546.


DI (census tract)
DI (10 m cell over 1 km2 )
−0.716 (−1.956)
−0.031 (−4.167) −0.03 (−4.063)
4.220 (11.576)

0.457 (3.300)
(census tract)

Mix-type index
Entropy

10.638

(10 m cell over 1 km2 )


0.052

12 ASCCar Alternate specific constants for


ASCBus different modes
ASCMotorizedThreeWhleerðMThWÞ
Table 4. Model for Trip Length per Individual for Work Trips

ASCMTW
4.036 (11.430)

ASCBicycle
0.471 (3.396)
Base model

ASCRickshaw
0.047
0.000


ASCWalk
Road length in 500 × 500 m grid (origin)

Motorized two-wheeler (MTW) vehicle ownership has significant


effect on explaining the variability of trip length. With increasing
MTW vehicle ownership in a household, the trip length increases.
Motorized two-wheeler ownership

Percentage increase in adj (R2 )

As shown in Table 5, for shopping trips also, land-use parameters


Land-use parameter (origin)

were found to have significant effect on trip length. All the land-
Socioeconomic parameters

use-mix parameters were found to be significant when entered into


normalized by grid size

the model separately. In the cases of both shopping and work trips,
older persons seem to go to closer destinations (shorter trip lengths).
from base model

Both the coefficient of land-use mix and intersection density at the


origin of the trip is negative, thereby reducing the trip length. Similar
Education

to a work trip, adj (R2 ) value of the model is slightly low but
Adj (R2 )
Constant

comparable with the values from the previous studies (Cervero


Age

and Kockelman 1997; Vance and Hedal 2007; Frank et al. 2009).

© ASCE 04016009-8 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


and intersection

−0.02 (−4.601) −0.018 (−4.137) −0.018 (−4.193)


−0.523 (−2.161)
−0.023 (−8.174) −0.019 (−5.671)
3.315 (13.851) 3.283 (13.762)
index at origin
including area
Kockleman (1997) could explain 14.52% of the variability of VMT

Final model

264.912
density
in personal vehicle for all trips using entropy and DI to quantify the

0.208
land-use mix. Both entropy and dissimilarity index had moderate
effects on vehicle miles traveled. Vance and Hedal (2007) from
the German Mobility Panel could explain 19% of the variability
of the VMT data and used commercial diversity and walk time

1 × 1 km tract)
intersections in
(numbers of
to public transit as land-use parameters. Frank et al. (2009) have
Intersection

250.877
density

0.2000
explained 10% variability of VMT data. Apart from the studies men-


tioned previously, Wang et al. (2013) and Lee et al. (2014) have ex-
plained the variability of VMT by more than 44% by using land use
and network variables along with the sociodemographic data. In this
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

study, the best model could explain 19.6% of the variability ob-
in 1 × 1 km tract)
Mix-type index

−1.442 (−5.962)
2.816 (12.223)
(10 × 10 m cell

served in trip length.


140.351
0.137

Effect of Land Use on Mode Choice


From the literature review it has been observed that the mixed land
use has significant impact on the probabilities of choosing nonmo-
in 1 × 1 km tract)

−1.441 (−5.952)
2.781 (12.176)
(10 × 10 m cell

torized and public transit modes. Finally, an MNL model has also
−0.02 (−4.6)

been estimated to understand the significance of various land-use


140.351
0.137
DI

parameters on the mode choice for work trips. To study the effect of
Base model with land-use parameters

mixed land use on mode choice, multinomial logit models were


prepared for work trips only. The software BIOGEME (Bierlaire
2003, 2008) was used in the estimation of choice models. The
(1,000 m buffer)

MNL model has been estimated to understand the significance


−0.021 (−4.689) −0.022 (−4.839)
−1.279 (−6.072) −1.043 (−4.284)
2.894 (11.42)

of land-use parameters in explaining the observed component of


Entropy

73.684
0.099

the preference heterogeneity, which otherwise is explained using


the socioeconomic characteristics.


A motorized three-wheeler (MThW), which carries passengers
on a shared basis and is generally available to all the passengers,
has been considered as the base mode for model estimation. Deter-
(1,000 m buffer)
2.698 (−12.043)
Area index

mination of the choice set for an individual is not straightforward


at origin

145.614
0.14

and very difficult in cases of smaller cities where no network and


level of service-related data are available. The choice set has been
prepared using a rule-based approach where availability of bus and
motorized three-wheeler is available to all, car and motorized two-
wheeler are available to the head of the family only if only one such
(cell size of 100 m

−0.022 (−4.891)
−0.668 (−2.843)
in census tract)
2.571 (10.897)

vehicle is present in the family. Bicycle ownership and rickshaws


Note: Parameters are tested at 5% significance level; number of observations = 370.

are available to all passengers with in the Central Business District


31.579
0.075
DI

(CBD) area for trips shorter than 3 km.


Results from the multinomial model estimated using the re-
vealed preference (RP) data on seven modes used for the work trips
are shown in Table 6. Out of various land-use-mix parameters, area
Table 5. Model for Trip Length per Individual for Shopping Trips

index, entropy with 1 km buffer, and DI with 10 × 10 m cell were


−0.022 (−4.814) −0.022 (−4.853)
−0.628 (−2.512)
2.377 (10.424) 2.576 (10.739)
(census tract)

found to be significant and positive for public and nonmotorized


Entropy

22.807
0.07

transport. The conventional DI and entropy index were not able


to capture the effect of land-use mix on the choice of public trans-


port. However, conventional indices were significant in capturing
the effect of mixed land use on nonmotorized modes.
Base model

0.057
0.000

Elasticity Analysis

Elasticities between the trip lengths and various land-use indices


are shown in Table 7. Aggregate elasticities, shown in Table 8
for the MNL model, were calculated using the probability weighted
Intersection density (at origin)

sample enumeration (PWSE) technique, given as


Land-use parameter (origin)
Socioeconomic parameters

X
Q  X
Q
% increase in adj (R2 )

P
EXP̄iikq ¼ P̄iq EXiqikq P̄iq ð5Þ
q¼1 q¼1
from base model

where P̄i refers to the aggregate probability of choosing alternative


P
i, EXiqikq ¼ β ik X ikq ð1 − Piq Þ, interpreted as the elasticity of the prob-
Adj (R2 )
Constant

ability of alternate I for decision maker q with respect to marginal


Age

change in the kth attribute of the ith alternative (i.e., X ikq ) as

© ASCE 04016009-9 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Table 6. Multinomial Logit Model for Work Trips with Different Mixed Land-Use Parameters
Entropy Entropy index DI DI Mix-type index
Area index with buffer (census tract) (census tract) (10 m cell over 1 km2 ) (10 m cell over 1 km2 )
Variable description Param. (t-value) Param. (t-value) Param. (t-value) Param. (t-value) Param. (t-value) Param. (t-value)
ASCCar −0.134 (−0.08) 0.320 (0.19) −0.358 (−0.22) −0.494 (−0.31) −0.208 (−0.13) −0.139 (−0.08)
ASCBus −4.420 (−6.64) −4.440 (−6.63a) −4.360 (−6.60a) −4.360 (−6.61a) −4.430 (−6.65a) −4.430 (−6.65a)
ASCMThW 0.000 (0.00) 0.000 (0.00) 0.000 (0.00) 0.000 (0.00) 0.000 (0.00) 0.000 (0.00)
ASCMTW −1.280 (−1.52) −0.713 (−0.80) −1.500 (−1.75) −1.600 (−1.91) −1.200 (−1.42) −1.140 (−1.34)
ASCBicycle −0.131 (−0.10) −0.201 (−0.15) −0.021 (−0.02) 0.030 (0.02) 0.199 (0.15) 0.188 (0.15)
ASCCycleRickshaw −1.880 (−2.06)a −1.920 (−2.03a) −1.670 (−1.84) −1.700 (−1.90) −1.540 (−1.67) −1.550 (−1.68)
ASCWalk −0.536 (−0.68) −0.485 (−0.58) −0.253 (−0.33) −0.268 (−0.35) −0.010 (−0.01) −0.028 (−0.04)
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

Land-use indexPublic 1.390 (1.81b) 2.420 (2.75a) 1.320 (0.88) 2.810 (0.82) 2.000 (2.52a) 2.060 (2.56a)
Land-use indexNMT 2.720 (3.24a) 3.270 (3.39a) 3.470 (2.34a) 8.580 (2.55a) 2.300 (2.71a) 2.370 (2.76a)
CostAuto;MTW −0.134 (−3.84a ) −0.137 (−3.85a) −0.123 (−3.74a) −0.125 (−3.79a) −0.136 (−3.90a) −0.136 (−3.89a)
CostRickshaw −0.040 (−2.18a) −0.042 (−2.25a) −0.046 (−2.52a) −0.044 (−2.41a) −0.047 (−2.55a) −0.047 (−2.54a)
Bicycle ownCar −1.820 (−1.71 b) −1.880 (−1.71b) −1.760 (−1.68b) −1.780 (−1.70b) −1.790 (−1.69b) −1.800 (−1.69b)
Bicycle ownBicycle 4.840 (4.05a) 4.740 (3.99a) 4.780 (4.05a) 4.730 (4.00a) 4.850 (4.07a) 4.860 (4.07a)
EducationRickshaw 0.115 (1.97a) 0.124 (2.13a) 0.121 (2.10a) 0.123 (2.14a) 0.131 (2.28a) 0.131 (2.28a)
Family sizeCar −0.419 (−2.14a) −0.411 (−2.06a) −0.405 (−2.10a) −0.401 (−2.07a) −0.398 (−2.02a) −0.400 (−2.03a)
Family sizeBicycle −0.557 (−2.66a) −0.537 (−2.60a) −0.544 (−2.65a) −0.534 (−2.60a) −0.548 (−2.61a) −0.549 (−2.62a)
GenderWalk 1.250 (2.69a) 1.200 (2.59a) 1.160 (2.51a) 1.170 (2.53a) 1.180 (2.54a) 1.180 (2.54a)
IncomeCar 0.591 (3.85a) 0.618 (3.95a) 0.583 (3.89a) 0.593 (3.94a) 0.621 (4.02a) 0.622 (4.02a)
IncomeBus 0.157 (1.68b) 0.155 (1.65b) 0.159 (1.71b) 0.158 (1.70b) 0.155 (1.66b) 0.155 (1.59)
LicenseMTW 3.360 (5.81a) 3.460 (5.78a) 3.300 (5.83a) 3.330 (5.87a) 3.510 (5.96a) 3.510 (5.96a)
IncomeWalk −0.196 (−1.94) −0.187 (−1.85b) −0.190 (−1.88b) −0.186 (−1.85b) −0.188 (−1.87b) −0.188 (−1.86b)
TimeAuto;MTW −0.024 (−1.69b) −0.024 (−1.66b) −0.025 (−1.78b) −0.025 (−1.75b) −0.024 (−1.67b) −0.024 −1.66b)
TimeCycle −0.069 (−3.26a) −0.070 (−3.22a) −0.075 (−3.39a) −0.075 (−3.46a) −0.077 (−3.41a) −0.077 (−3.40a)
TimeWalk −0.035 (−2.00a) −0.037 (−2.11a) −0.040 (−2.24a) −0.039 (−2.24a) −0.043 (−2.38a) −0.042 (−2.37a)
MTW ownershipCar −1.750 (−2.50a) −1.690 (−2.40a) −1.690 (−2.42a) −1.720 (−2.45a) −1.800 (−2.59a) −1.800 (−2.59a)
MTW ownershipMTW 1.630 (2.15a) 1.750 (2.23a) 1.640 (2.21a) 1.680 (2.26a) 1.660 (2.19a) 1.660 (2.19a)
Initial log- likelihood −679.526 −679.526 −679.526 −679.526 −679.526 −679.526
Final log-likelihood −315.303 −314.429 −317.873 −317.061 −316.907 −316.743
Adjusted rho-square 0.499 0.500 0.495 0.497 0.497 0.497
Note: Number of observations = 567.
a
Statistical significance at 5% level.
b
Statistical significant at 10% level.

observed by decision maker q and P̄iq is an estimated choice prob- the buffer zones and mix-type index are strongly associated. Area
ability. But in the case of variables that appear as alternate specific index, which was used to measure the land-use area complemen-
forms in all alternatives, it becomes a combination of direct tarity, also has strong influence on travel parameters. A 1% increase
response and multiple cross response. Elasticity in this case is in the area index can lead to a reduction of trip length by 0.235 and
given by 0.326% for work and shopping trips. Elasticity values estimated
X based on the MNL models are shown in Table 8. From the table
X
Q X Q
it can be seen that the probability of choosing walk and cycle rick-
EXP̄iikq ¼ P̄iq ½β ik X ikq ð1 − Piq Þ − β jk X jkq ðPjq Þ P̄iq
q¼1 j≠i q¼1
shaw increases with the increasing land-use mix. In the case of mo-
torized modes, elasticities between the land-use indices and trip
ð6Þ lengths are weak. The elasticities of probabilities for nonmotorized
modes are higher than for public transport modes. From the values
From the elasticities between the trip lengths and land-use in-
shown in Table 8, it can be said that the modified land-use indices
dices, it can be said that the trip length for shopping and work trips
have significant impact on the mode choice, including the choice of
and the land-use parameters such as entropy index measured using
nonmotorized mode of transport. To show the effect of mixed land
use on mode choice, probability of choosing different modes was
Table 7. Elasticities of the Travel Parameters With Respect To Land-Use
plotted with area index.
Variables When Single Land-Use Parameter Entered in the Trip Length From Fig. 10, it can be seen that the probability of choosing
Model MThW and MTW reduces with increasing area index. The prob-
ability of choosing bus travel does not increase when mixed land
Trip length
use increases, whereas the probability of choosing a private travel
Land-use parameters Work Shopping mode reduces. In the case of nonmotorized modes, it can be seen
Area index (1,000 m buffer) −0.235 −0.326 that with an increase in area index, the probability of choosing
Entropy (1,000 m buffer) −0.330 −0.241 nonmotorized mode increases. A higher area index represents
Dissimilarity index (10 × 10 m cell) −0.250 −0.371 a scenario where the work areas are nearer to the trip maker’s
Mix-type index −0.270 −0.397 household, thereby reducing the trip length. Because of this, with
Road length in 500 × 500 m grid −0.646 — increasing area index, there will be an increase in the share of non-
Intersection density — −0.857
motorized modes and decrease in the share of motorized modes.

© ASCE 04016009-10 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


Table 8. Elasticities of Mode Choice with Land-Use Parameters from MNL Model
Elasticity
Elasticity Elasticity Elasticity Elasticity Elasticity (mix-type index
Modes (area index) (entropy 1,000 m buffer) (entropy ward) (DI for census tract) (DI 10 × 10 m grid) 10 × 10 m grid)
Cycle 0.029 0.050 0.037 0.030 0.011 0.02
Cycle rickshaw 0.265 0.290 0.190 0.168 0.02 0.119
Walk 0.206 0.242 0.150 0.140 0.004 0.105
Bus 0.0003 0.063 — — 0.065 0.053
MThW −0.06 −0.011 — — 0.06 0.025
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

in the model’s ability to explain the variability of trip length com-


pared to the base model or base model with DI measured for 1 ×
1 km tract. Coefficients corresponding to land-use-mix parameter
and network parameters such as road length and intersection den-
sity were negative, indicating that an increase in network variable
and land-use mix tends to reduce the trip length.
From the MNL mode choice models estimated for work trips,
area index and entropy with 1 km buffer, DI and mix-type index
with 10 m × 10 m cell, were found to be statistically significant.
Consistent with the findings of Frank and Pivo (1994) and Frank
et al. (2009), in the case of the present study also land-use mix was
found to be strongly associated with the choice preference of public
transport. The coefficients were positive for public and nonmotor-
ized modes, which imply that the trip makers residing in the areas
with mixed land use prefer public and nonmotorized travel modes.
Conventional DI and entropy indices were not able to capture the
Fig. 10. Variation in the choice probability corresponding to various effect of land-use mix on the choice of public transport but they
modes with area index were significant in capturing the effect of mixed land use on the
choice of nonmotorized modes.
Elasticity between trip length and road network length was
found to be −0.645 for work trips. Elasticity of trip length with
When the performance of different indices is compared in terms intersection density was found to be −0.857 in the case of shopping
of the model improvements it can be said that area index quantifies trips. When the land-use mix was quantified in terms of area index,
the mixed land use effectively. Also, area index is comparatively the trip length elasticity values were found to be −0.235 and
easy to calculate and interpret as it is only the ratio of areas asso- −0.326 for work and shopping trips, respectively. Elasticities of
ciated to the nature of the trip. In the case of shopping trips, the area mode choice, explained in terms of area index, were found to
index simply becomes the ratio of the area of commercial space in be higher for cycle rickshaw (0.265) and walk modes (0.206) than
the buffer zone to the total extent of the commercial space in the cycle mode (0.029). From this it can be said that the probability of
study area. With the data on the extent of shopping space in the choosing walking and cycle rickshaw modes increases with an in-
study area one can quantify the effect of mixed land use in a better crease in land-use mix. It has also been observed that the elasticity
way than generally used mixed land-use variables. of probability of choosing nonmotorized mode (0.265 for cycle
rickshaw, 0.206 for walking) is higher than that of the public trans-
port mode (0.0003). Most of these parameters suggest that there is a
Conclusions need to systematically control the land-use mix to maintain the pa-
tronage for nonmotorized modes.
In this study the impact of land-use variables on travel behavior has The findings of the present study are similar to the findings and
been analyzed in the context of smaller Indian cities. Land-use mix claims of new urbanism and other research that indicates mixed
has been quantified using entropy, area index, dissimilarity index, land use as one of the major factors degenerating vehicle trips,
and mix-type index. The entropy index has been obtained for each reducing VMT per capita, and encouraging nonmotorized travel.
of the origins in two ways: for the census tract where the origin is The result of the present study suggests that the effect of mixed
located, and for a buffer of 1 km radius created around the origin. land use on travel demand was moderate and it can be found
The dissimilarity index has also been calculated in two ways: for a out from the reported elasticity values. Elasticities fall in the range
census tract with a cell size of 100 × 100 m, and for a 1 × 1 km of 0.23 to 0.397 for trip length with respect to mixed land use. In the
tract with a cell size of 10 × 10 m. Because of strong correlation case of mode choice model, elasticity value for mode choice with
between entropy and DI only one of these variables was signifi- land-use parameter was less than 0.265. Consistent with the find-
cantly entering the model. Entropy measured with 1,000 m radius ings of Frank and Pivo (1994) and Cervero (1996), having retail
buffer, and DI measured using 10 × 10 m cell for a 1 × 1 km tract activities within the neighborhood was closely associated with
were found to be relatively more significant than the same indices the mode choice for work trip.
corresponding to the census tract. When compared with entropy, DI From the discussion here it can be observed that to achieve the
was found to be better for quantifying the land-use mix, either mea- desired travel patterns the land-use development needs to be con-
sured for census tract or for the modified measurement area. When trolled in a systematic manner. The evolution of land use in the
the area index was considered there was a significant improvement smaller cities used to be need based, and city administration used

© ASCE 04016009-11 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009


to be flexible while enforcing the land-use plans. Given the recent Ewing, R., and Cervero, R. (2010). “Travel and the built environment:
changes in the socioeconomic characteristics and associated travel A meta-analysis.” J. Am. Plan. Assoc., 76(3), 265–294.
changes, the city administration has started controlling land use. Frank, L., and Pivo, G. (1994). “The impacts of mixed use and density on
This is evident from the fact that most of the smaller cities have the utilization of three modes of travel: The single occupant vehicle,
transit, and walking.” Transp. Res. Rec., 1466, 44–52.
started developing master plans with strict control on land use.
Frank, L. D., Kavage, S., Greenwald, M., Chapman, J., and Bradley, M.
While doing this, it is necessary to consider the implications of such (2009). “I-PLACE3S health & climate enhancements and their applica-
enforcements. Efforts should be made toward sustaining the land- tion in King County.” 〈http://your.kingcounty.gov/kcdot/planning/ortp/
use mix, which was allowed inadvertently. Sustaining the existing HealthScape/I-PLACE3S-FINALREPORT%2006-01-09.pdf〉 (May 29,
mix will not only encourage shorter trips but would also lead to 2014).
increased use of nonmotorized modes. Hess, P. M., Moudon, A. V., and Logsdon, M. G. (2001). “Measuring
land use patterns for transportation research.” Transp. Res. Rec.,
Downloaded from ascelibrary.org by National Institute of Technology Agartala on 10/13/19. Copyright ASCE. For personal use only; all rights reserved.

1780, 17–24.
Acknowledgments Kapur, J. N., Sahoo, P. K., and Wong, A. K. C. (1985). “A new method for
gray level picture thresholding using the entropy of the histogram.”
Authors acknowledge the financial support provided for travel data Comput. Graphics Vision Image Process., 29(3), 140–285.
collection as a part of a sponsored project, numbered 8023/BOR/ Kockelman, K. M. (1997). “Travel behavior as function of accessibility,
land use mixing, and land use balance: Evidence from San Francisco
RID/RPS-250 by All India Council for Technical Education
Bay Area.” Transp. Res. Rec., 1607(1), 116–125.
(AICTE), statutory body and a national-level council for technical Lee, J. S., Lee, S. K., and Jun, J. (2014). “Connections between land
education, under Department of Higher Education, Ministry of use and driving distance: Causal investigation using directed acyclic
Human Resource Development, Government of India. graphs.” J. Urban Plann. Dev., 10.1061/(ASCE)UP.1943-5444
.0000195, 06014001.
MoUD (Ministry of Urban Development). (2008). “Study on traffic and
References transportation policies and strategies in urban areas in India.” 〈https://
casi.sas.upenn.edu/sites/casi.sas.upenn.edu/files/iit/GOI%202008%20
ArcGIS version 10.1 [Computer software]. ESRI, Redlands, CA. Traffic%20Study.pdf〉 (Jul. 25, 2010).
Bierlaire, M. (2003). “BIOGEME: A free package for estimation of discrete Pal, N. R., and Pal, S. K. (1993). “A review on image segmentation tech-
choice models.” Proc., 3rd Swiss Transportation Research Conf., niques.” Pattern Recog., 26(9), 1277–1294.
Ascona, Switzerland. Pun, T. (1981). “Entropic thresholding: A new approach.” Signal Process.,
Bierlaire, M. (2008). “Estimation of discrete choice models with 2, 210–239.
BIOGEME 1.8.” 〈biogeme.epfl.ch〉 (Jun. 11, 2013). Pushkar, A. O., Hollingworth, B. J., and Miller, E. J. (2000). “A multivari-
Cervero, R. (1989). America’s suburban centres: The land use transpor- ate regression model for estimating greenhouse gas emissions from
tation link, Unwin Hyman, London. alternative neighborhood designs.” Proc., Transportation Research
Cervero, R. (1991). “Land uses and travel at suburban activity centers.” Board, 79th Annual Meeting, Washington, DC.
〈http://escholarship.org/uc/item/0d08h1bz〉 (Sep. 10, 2013). Rajamani, J., Bhat, C. R., and Handy, S. L. (2003). “Assessing impact of
Cervero, R. (1996). “Mixed land-uses and commuting: Evidence from the urban form measures on non-work trip mode choice after controlling for
American housing survey.” Transp. Res. Part A: Policy Pract., 30(5), demographic and level-of-service effects.” Transp. Res. Rec., 1831,
361–377. 158–165.
Cervero, R., and Duncan, M. (2003). “Walking, bicycling and urban land- Shannon, C. E. (1948). “A mathematical theory of communication.” Bell
scapes: Evidence from the San Francisco Bay area.” Am. J. Public Syst. Tech. J., 27(3), 379–423.
Health, 93(9), 1478–1483. Sun, X., Chester, G. W., and Kasturi, T. (1998). “Household travel, house-
Cervero, R., and Kockelman, K. (1997). “Travel demand and the 3Ds: hold characteristics, and land use: An empirical study from the 1994
Density, diversity, and design.” Transp. Res. Part D: Transp. Environ., Portland activity-based travel survey.” Transp. Res. Rec., 1617, 10–17.
2(3), 199–219. Toutain, O., and Gopiprasad, S. (2006). “Planning for urban infrastructure.”
Cervero, R., and Radisch, C. (1996). “Travel choices in pedestrian versus India Infrastructure Rep. 2006, Oxford University Press, Oxford, U.K.,
automobile oriented neighborhoods.” Transp. Policy, 3(3), 127–141. 59–81.
Chapman, J., and Frank, L. (2004). “Integrating travel behavior and urban Vance, C., and Hedal, R. (2007). “The impact of urban form on automobile
form data to address transportation and air quality problems in Atlanta, travel: Disentangling causation from correlation.” Transportation,
Georgia.” Research Project No. 9819, Task Order 97–13, U.S. Dept. of 34(5), 575–588.
Transportation, Washington, DC. Wang, X., Grengs, J., and Kostyniuk, L. (2013). “Using a GPS data set to
Ewing, R., et al. (2010). “Traffic generated by mixed-use developments— examine the effects of the built environment along commuting routes on
Six-region study using consistent built environmental measures.” J. Ur- travel outcomes.” J. Urban Plann. Dev., 10.1061/(ASCE)UP.1943-
ban Plann. Dev., 10.1061/(ASCE)UP.1943-5444.0000068, 248–261. 5444.0000181, 04014009.

© ASCE 04016009-12 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2016, 142(4): 04016009

You might also like