You are on page 1of 19

Journal of the Eastern Asia Society for Transportation Studies, Vol.

13, 2019

Safety Assessment of Urban Un-Signalized Intersections Using Conflict


Analysis Technique

Sudeshna MITRA a, Dipanjan MUKHERJEEb*, Subham MITRAc

a, b,cDepartmentof Civil Engineering, Indian Institute of Technology Kharagpur, Kharagpur,


India
aE-mail: sudeshna@civil.iitkgp.ernet.in
bE-mail: ddiipp90@gmail.com
cE-mail: subhamm.micec@gmail.com

ABSTRACT: Road traffic crash data from Kolkatacity showsa high share of fatal crashes at
the un-signalized intersections. Based on the idea that crashes mainly evolve from conflicts,
this study developed a novel methodology to correlate conflicts at the un-signalized
intersection with factors of the built environment as well as traffic operational and regulatory
parameters; so that proactive measures could be made to enhance safety.A multinomial logit
model and an ordered probit model are developed to find how the conflict types and conflict
severities are linked with the design and operational elements of intersections and what can be
expected in terms of safety given certain conditions prevail at an un-signalized junction.
Finally, a conclusiverelationshipis established between the share of conflicts and police
reported fatal crashes, confirming the potential of conflict analysis as a promising technique
for safety assessment in the absence of reliable and precise crash data.

Keywords:Un-Signalized intersection; Fatal Crashes; Proactive active Approaches; Conflict


Analysis Techniques; Conflict Severity

1.INTRODUCTION

Scientific analysis of road traffic crashes have been gaining serious attention in developing
countries such as India quite recently, due to the ever-increasing problem of crashes, injuries
anddeathsowing to the increased urbanization, motorization as well as construction of high-
speed facilities. Since crash occurrence islinkedwith some undesirable problems in the traffic
system, scientific community has been researching the symptoms or the failure mechanism for
over three decades— primarily with the help of police recorded crash data. However, in
developing countries such as India, obtaining crash data is a major challenge, not due to the
fewer occurrences, but mainly due to the poor maintenance of the crash database as well as
access to such data. As a result, scientific studies on road traffic safety with crash data have
either been limited(Prajapati and Tiwari 2013;Gupta and Rokade 2016) or failed to link
plausible safety issues with actual crash counts (Deshpande et al. 2005; Vedagiri andKilli
2015).
Recent research has shown that traffic conflicts provide useful insight into the failure
mechanism thatleads to road collisions due to the fact that collisions are more frequent and of
minor social cost (Sayedet al. 2012). As defined by Glauz and Migletz (1980) traffic conflict
is a traffic event involving the interaction of two vehicles where one or both drivers have to
take evasive decision to avoid a collision. These conflicts are the situations where the

*Crosseponding author.

2163
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

probability of collision is relatively higher. Based on the idea that crashes mightoccurif there
is no temporal or spatial separation, it should provide very important insight about the safety
of a location if the predominant types of conflicts are identified properly. As rightly pointed
by El-Basyouny and Sayed (2012) for identification of conflicts, several studies have adopted
different methods of conflict data collection, such as (i) field observers (Perkins andHarris
1967; William 1972; Older and Spicer 1976) (ii) simulation models (Sayedet al. 1994; Huang
and Pant 1994; Persaudet al. 1995; Rao and Regaraju 1998)and (iii) video camera (Ismailet al.
2009a; Ismailet al. 2009b; Ismailet al. 2010a; Ismailet al. 2010b). While the video-based
technique is most reliable, the automatic extraction of conflict information is still under
development and is not widely used (Auteyet al. 2012; El-Basyounyand Sayed 2013).
Considering the potential of conflict based studies a group of researchers in Europe
have planned the InDEV study, with an aim to assess the suitability of predicting crashes
using conflicts (www.indev-project.eu). However, one of the major challenges faced was low
number of crash data for validation. As mentioned by El-Basyouny and Sayed (2012) traffic
conflicts can serve as an appropriate predictor to collisions, due to the fact that conflicts are
based on vehicle interactions; and it can be argued that they would be appropriate predictors
for collisions. Consequently, the researchers (El-Basyouny and Sayed 2012) used the traffic
conflict techniques to establish a relationship between predicted collisions and conflicts for 51
signalized junctions in British Columbia, and also predicted the conflicts in terms of various
exogenous design and operational factors. After studying extensive literature available related
to this field, it can be concluded that almost no study attempts to define and quantify the
predominant type of conflicts at the un-signalized intersections in the urban context in a
developing country such as India. Further, cities in developing countries such as India could
be ideal for validation of crash and conflict studies if police recorded crash data could be
accessed.Finally, considering the current state of the art for conflict data collection, it is
thought that for Indian condition a combination of video camera based recording with manual
extraction could be performed successfully without loss of information.
In this background, the principal objective of the study is to identify the major types of
conflicts at un-signalized intersections and identify factorsinfluencing such conflicts. For this
purpose, exogenous variables related to the traffic exposures such as volume and speed,
vehicle composition or categories, built environmental characteristics such as road width, and
traffic operational parameters such as the presence of traffic police, which may influence the
conflict types and the conflict severity at the un-signalized junctions are considered. A
secondary objective is to also explore the relationship, if any, between the counts of fatal
crashes recorded at various sites and the share of conflict at those sites.
In order to achieve the research objectives, firstly, the crash-prone locations have
been identified from the “AccidentReview Report (2011-16)” provided by “Kolkata Police”
from Kolkata, India. Subsequently, 10 un-signalized intersections have been selected from the
Central Business District (CBD) of Kolkata, where video recording was conducted for one-
hour duration. The necessary data is extracted from the video-records, and a thorough dataset
of 813 conflicts is developed. Using this dataset suitable statistical techniques are used to
develop models that could explain conflict types and severity with the exogenous variables
mentioned before. Finally, a relationship was established between the share of conflicts
(compared to the total entering traffic) and the fatal crash counts recorded by the Kolkata
police for that location.

2164
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

2. LITERATURE REVIEW

Traffic conflict analysis can provide a potentially reliable and inexpensive tool which can be
utilized as a replacement or supplement for assessing the safety and operational deficiencies
(Grayson 1984). Most of the researchers have attempted to investigate and study the traffic
characteristics that may be an alternative for the crash data. Hence, a number of extensive
research works have been performed on the attributes of conflict and how they are related to
crashes.An early research on this subject was conducted by Perkins and Harris (1967) who
gave the idea of traffic conflict analysis techniques as an alternative to crash data. They
identified potential crash situations as “traffic conflicts”. Amundsen and Hyden (1977)
defined “traffic conflict is an observable situation in which two or more road users approach
each other in space and time to such an extent that a collision is imminent if their movements
remain unchanged”. Spicer et al. (1980) and Al-Maita (1995) found that the total number of
traffic conflicts is proportional to the square root of the product of the conflicting volumes.
Parker and Zegeer (1988) defined various types of conflicts and provided a standard,
cost-effective method for accurately observing and recording traffic conflicts, as the authors
emphasized the fact that the most literature in surrogate measures is related to Traffic
Conflicts Technique (TCT) and that the accurate recording and extraction of conflict is
critical. Hayward (1972) defined Time to Collision (TTC) as the time for two vehicles to
collide if they continue at their present speed and on the same path. The value of the TTC is
infinite if the vehicles are not on a collision course. The minimum TTC reached as the
vehicles approach on the collision course is taken as the critical measurement in estimating
conflict severity (Sayed and Zein 1999).
Although traffic conflict techniques have been used widely, the actual association between
crash and conflict has been a concern among researchers. For instance, Migletzet al. (1988)
concluded that crashes and conflicts were significant only at the 90% confidence
interval.Salman and Almaita(1995) found a linear relationship between crashes and traffic
conflicts. Glauzet al.(1985) attempted to find out the relationship between conflicts and
predicted crashes, and concluded that certain types of traffic conflicts are good surrogates of
crashes. On the contrary, Engel (1985) found a poor correlation between conflicts and crashes,
although they concluded that therelationship could be improved if the types of conflicts and
their corresponding crashes are separately analyzed. Kulmala(1993) explored the role of
traffic conflict technique in road safety analysis and concluded that conflict analysis technique
act as a complementary method to support crash studies and provide a complete picture of the
safety at a site. Horst and Hogema(1993) established an association between time to collision
(TTC) and the severity of traffic conflicts. Fruhman (1993) undertook a study to see the
possible association between conflict data and traffic crash data. He also discussed the
assessment of the safety quality for particular road users. The study concluded that, the
possibility of crash prognostication based on traffic conflicts observed depends on focal
marginal conditions. On the other hand, it depends on the probability of the occurrence of
particular crash and conflict types. Sayed and Zein (1999) developed an Intersection Conflict
Index (ICI) to evaluate conflict risk. Additionally, the authors presented a regression analysis
to establish the relationship between traffic conflicts and crashes, and concluded that there is a
strong relationship between the conflict rate and crashes at a junction. Li et al. (2011) also
evaluated crash severity of unsignalized intersections using Conflict Index (CI).
Songchitruksa and Tarko(2006)presented a new approach by using surrogate safety
measures, which is based on proximal safety indicators representing the temporal and spatial
proximity characteristics of risky interactions and near-crashes. Zhang et al. (2017) developed
Poisson regression model to identify the key factors influencing TTC at three-legged un-

2165
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

signalized junctions in Nanjing City, China. The authors concluded that while volume of left-
turning traffic on minor road or speed of left-turning vehicles on minor road increases, the
number of traffic conflicts increases. Archer (2005) and De Mouzonet al. (2007)identified
safety measures such as Time to Collision (TTC), time-integrated TTC, time-extended TTC,
Post Encroachment Time (PET), deceleration rate and other indicators, but according to
Gettman and Head (2003), the most commonly used conflict analysis techniques are PET and
TTC.
Brilon and Wu (2001) determined capacity of unsignalized junctions using conflict
analysis. Workineh(2014) established the relationship between traffic conflicts and level of
service at four-legged, signalized intersections in SacramentoCity in California.Prasetijoet al.
(2016) measured the performance of non-priority intersections under mixed traffic conditions
based on conflict streams analysis. Nakamura and Mabuchi (2007) evaluated the safety
performance of roundabout in Japan, considering safety analysis techniques.
Further, Liet al. (2011) developed a soft computing technique to identify potential
conflicting points on Highway, Minnesota in US State.Ambroset al. (2014)explored the
significance of micro-simulation methods for conflict analysis techniques.
Several research studies havefocusedin the past on the safety assessment of un-signalized
intersections(Patil and Pawar 2014;Vedagiri and Killi2015; Priyadarshini and Mitra 2018) in
developing countries, such as India. However, very few studies on the safety of un-signalized
intersections focused on a combined crash and conflict approach to understand what
influences conflict at a site and how that is linked with the recorded crashes at those sites.
Further, in the Indian context, very few researchers have actually used conflict analysis
technique as a surrogate measure to assess safety. Tiwari et al. (1998) attempted to predict
fatal crashes in the mixed traffic environment using conflict analysis approach, and concluded
that the best use of conflict data may lie in reducing stress and improving traffic management
rather than reducing fatal crashes. Vedagiri and Killi(2015) evaluated traffic safety of
uncontrolled T-intersections using surrogate safety measure under mixed traffic conditions.
To determine the degree of severity of acrash, they used the concept of the conflict zone and
Post-Encroachment Time (PET).Using a negative PET and the concept of area occupancy,
they calculated PET and determined the grid with the highest possibility of collision, which
they validated in other locations. However, the effect of planning, design and operational
factors present at a location are not associated and thus not known which may affect safety.
Further, the authors also cited the inability to validate the PET with the crash data of a
location, which makes it difficult to get an insight about how PET and safety performance of a
location is linked.
The reviewthus shows limited work combining conflicts and crashes for safetyassessment of
un-signalized intersection in general, and specifically for developing countries such as India.
Hence, an attempt has been made in this study to fill that gap by identifying conflicts using
video-based techniques and studying their association with planning, design and operational
issues of un-signalized intersections. Further, these conflicts are associted to police recorded
fatal crashes of these locations to gain an insight as to how conflicts of un-signalized
intersections relate to the safety performance of those locations in a case study city, Kolkata.

3. METHODOLOGY

The methodological steps followed in this research study are presented in Figure 1, and each
step is explained in the following:

2166
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

3.1 Crash Data Collection and Selection of Study Sites

For eases of road traffic management, Kolkata police have divided the entire city is divided
into a total of twenty-five zones known as “traffic guards”. The crash data for the last six
years from 2011 to 2016 was collected for these traffic guards from the statistics section of
Kolkata traffic police department. Finally, 10 un-signalized intersections across four traffic
guards have been selected for the present study. The study locations are shown in a
Geographic Information System (GIS) map (Figure 2a).

3.2Collection of Exogenous Factors

Data collection and survey for this study involved three distinct approaches – (1) road
inventory survey, and (2) video-graphic survey (3) speed survey

3.2.1 Road inventory survey

The road inventory survey was conducted to collect information regarding road geometry,
existing infrastructure, land use, etc.

3.2.2 Video-graphic survey

To obtain the information regarding traffic volume, operational and conflict characteristics
video-graphic survey was carried out at each of the selected study locations. The video
recording was done during the peak period of 10:00 A.M. to 11:00 A.M. to capture the
greatest interaction between vehicles. The counts of crashes, vehicular volume, an average of
approaching speed, and built environmental characteristics of the selected intersections are
summarized in Table 1.

2167
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Crash Data Collection

Identification Study Location

Survey and Data Collection

Road Inventory Survey Video-graphic Survey

Identification and Estimation


of Conflicts

Descriptive Study

Identification of Factors Influencing


Conflict Type and Severity

Explore the Relationship between Crashes


and Conflicts

Discussion and Conclusion

Figure 1. Research design

2168
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Figure 3a. Study locations Figure 2c categorized conflict types from identified
conflicts - a. Crossing, b. Through and Turning, c.
Overtaking and d. Merging

1 - Crossing
20%
29%

2 - Turning &
Through
18%
3 - Overtaking
33%
4 - Merging

Figure 2b Grid laid intersection of C.R Avenue – Figure 2d. Share of different type of conflicts
MahajatiSadan Crossing

Figure 2. Selections of study intersections and conflict identification techniques

2169
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table1. Un-signalized intersections selected for the study

Fatal Average
Traffic Major
Crash Vehicular
Traffic Type of Volume Per Road
Name of the Intersection Statistics Speed
Guard Intersection Hour (10-11 Width
(2011-16) (Kilometer
AM) (meters)
Per Hour)

C. R. Avenue –Jorasanko Jorabagan 0 4-Legged 4262 23.16 40


M.G. Road - Syed Salley Street Jorabagan 1 3-Legged 1735 17.01 35
M.G Road – KalabaganBasti Jorabagan 1 3-Legged 905 16.00 35
M.M. Burman Street and C.R. 4-Legged
Jorabagan 2
(20111-16) 5085 18.29 45
Avenue
CR Avenue – MahajatiSadan Jorabagan 2 4-Legged 4884 22.86 45
Howrah 4-Legged
Strand Rd – JagannathGhat 3 1390 12.19 35
Bridge
4-Legged
Basanti Highway and EM Bypass Tiljala 4 1710 36.58 62

Strand Rd and Raja Woodmount Howrah 4-Legged


5 2735 16.98 42
Street Bridge
RG Kar Hospital Shyambazar 9 4-Legged 3172 12.19 55
Chowbaga Bus Stop Tiljala 12 4-Legged 1353 7.920 65

In order to define the conflict zones and for ease of identification and extraction of conflict
data, “AVS Video Editor” software was used to place grids on the recorded videos tapes (4).
Figure 2b shows a sample grid laid a snapshot of“C.R. Avenueand Mahajati Sadan”crossing.
From the videotapes following data were extracted:

After completing data extraction for ten locations, 11 types of conflict was observed, which
were finally narrowed down to four predominant types, to be used for further investigation.
The conflict types are as follows:

I. Crossing Conflict (CC)


II. Turning andThrough Conflict (TT)
III. Overtaking Conflict (OC)
IV. Merging Conflict (MC)

Figure 2c shows the different types of conflict that were finally categorized and Figure 2d
presents the share of these four predominant types of conflicts in study intersections.

• For identification and extraction of conflict cases from video data, the conflicting
vehicles were identified and denoted as “First Vehicle” and “Second Vehicle”, in the
order of their leaving and arriving at the particular conflicting zone, respectively
(Shown in Figure 3). The type of vehicles included in the present study is primarily of
five different categories, namely, (a) Two-wheeler, (b) Three-wheeler, (c) Four-
wheeler, (d) Bus and (e) Commercial vehicles.

2170
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Figure 3.Conflict zone

• While extracting data from the video-graphs, the speed changing (i.e., speed reducing)
characteristics of the conflicting vehicles are also noted and extracted. The vehicles
were coded with ‘0’ if the vehicle doesn’t change its speed and as ‘1 if the vehicles
changed their speed while approaching the junction.

• During the instance of the conflict, how many vehiclesare actually present at the
physical area of the intersection that data was also recorded.

• Finally, the proportion of conflicting vehicles as compared to the total entering traffic
volume of the junction is also estimated from the recorded videotapes.

3.2.3 Speed survey

To collect the operating speed at the crash sites, approaching speeds of the motorized vehicles
were measured using the speed radar gun.The ‘spot speed survey’ was conducted for morning
10 AM and 11 AM, same of the video-graphic survey. The variables used in the model
development and further analysis are explained in Table 2.

2171
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table2. Variable descriptions


Variable Variable
Variable Name Remarks
Characteristics Type
Conflict Type Conflict Type Categorical 1 – Crossing
2 – Through and Turning
3 – Overtaking
4 - Merging
Traffic Operational Vehicle Type Categorical 1 – Vehicle of the respective type
Characteristics 0 – Not
Presence of Traffic Police Categorical 1 – Traffic police is Present
0 – Traffic police is not present
Speed Changing condition of Categorical 1 – If the vehicle changes (i.e. reduces) its
Vehicle speed
0 – If the vehicle doesn’t change speed
Interaction between vehicles Categorical 1 - Both vehicles changes speed
0 - Either vehicle doesn’t change speed
Share of Conflict Continuous The ratio of the conflicting vehicle and the total
entering the vehicle at the intersection at a
certain time
Traffic Exposure Traffic Volume Continuous Total Number of Traffic volume per hour
Variables The speed of Motorized Continuous Average approach speed of the junction
Vehicles (Kilometer per hour)

Occupying vehicle Continuous Number of vehicles present at the physical area


of the intersection at the instance of conflict
Built Environment Major Road Width Continuous The width of a major road in meters
Minor Road Width Continuous The width of the minor road in meters
Safety Performance Fatal Crash Statistics Continuous Number of fatal crashes for the period 2011-
2016

3.3Modeling Methodology

The statistical assessment in this study initiated with a correlation analysis using ‘Spearman
Rank Correlation’ test to estimate correlation among several independent variables and
conflict types. Finally, conclusions are made based on the output. The choice of different
model type is primarily decided based on the type of the outcome/dependent variable. For
example,
• Multinomial Logistic Regression Model (MNL) was developed to identify the factors
affecting each conflict type, due to the fact that different conflict types are indicated
with different numbers or choice and there were more than two conflict types that
were modelled in this study, making MNL model to be most suitable for such study.
• On the other hand, Ordered Probit Model (OP) was developed to identify the factors
affecting the conflict severity with respect to their order of severity. For example, in
this study the three different conflict severities were modelled, and they were
“crossing”, “turning-and-through”, and “overtaking and merging”. While apparently it
may seem that these variables are categorical in nature, there is an inherent order of
severity with “crossing conflicts” being most severe, followed by “turning-and-
through” and “overtaking and merging”. To capture the ordered nature of the severity
of conflict, Ordered Probit model was thought to be most suitable.
• Finally, Multiple Linear Regression Model (MLR) was developed to identify the
factors affecting the share of conflicting vehicles. As the share of conflicting variable

2172
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

is a continuous number, and not an categorical or ordered variable, the variable is most
suitable to be modelled using MLR model.

A multinomial logistic regression model was introduced to identify the relationship and
betweenthe type of conflicts and built environment, traffic operational characteristics and
traffic regulatory parameters. The linear function Yin that determines the outcome i of
observation n may be written as shown in equation (1)
Yin =βiXin +εin (1)

Where βiis the vector of estimable parameters for discrete outcome i


Xinis a vector of observable characteristics that determine the outcome for discrete observation
n. and εinis the random error term (Washingtonet al. 2003)

The probability is a function of the linear utility equation(equation 1), which in turn is a
function of observable characteristics such as vehicle type, and road geometry. The discrete
outcome of observation nis i, if equation 2 is satisfied

Pn(i) = (Yin>YIn)∀I ≠i (2)

Where, Pn(i) is the probability of observation nhaving discrete outcome i(i∈ I), where I denote
all possible outcomes for observation n. The predicted conflict outcome for nth observation is
1 if the probability of ‘conflict type 1’ is greater than the probability of ‘conflict type 2’,
‘conflict type 3’ and ‘conflict type 4’, as shown in equation 3.
Similarly, predicted the type of conflict for nth observation is ‘conflict type 2’; if the
probability of ‘conflict type 2’ is greater than the probability of ‘conflict type 1’, ‘conflict type
3’ and ‘conflict type 4’.

Pn(1)>Pn(2 , Pn(3) and Pn(4) (3)


From the basic equation of discrete choice model, the multinomial logit model assumes
Gumbel distribution for the εijerror term. The conflict type equation for this model gives the
probability of the certain type of conflict being equal to any discrete outcome ‘conflict type
1’, ‘conflict type 2’, ‘conflict type 3’ and ‘conflict type 4’ can be written as in equation 4.

𝑒𝑥𝑝 (𝛽𝑗𝑋𝑛) (4)


𝑃𝑟(𝐶𝑜𝑛𝑓𝑙𝑖𝑐𝑡 𝑇𝑦𝑝𝑒 = 𝑗) = , 𝑗 = 1,2,3,4
∑∀𝑗 𝑒𝑥𝑝 (𝛽𝑗𝑋𝑛)

Further, conflict severity levels are modeled on a scale from 0 to 2, with 0 being not at all
severe and 2 being extremely severe. These conflict severity levels are discrete and ordered
(i.e., 2 is worse than1, 1is worse than 0). Researchers have established several discrete choice
models such as the multinomial, nested, and probit models may not be appropriate for the
present data set, which is ordered as well as discrete in nature (Washington et al. 2003). OP
model is the most suitable for the data, which is discrete as well as ordered in nature
(McKelveyand Zavoina 1975; Kadali and Vedagiri 2016). The basic form of the OP model is
specified as given in equation 5; with a normally distributed random error term εij.
(Washington et al. 2003)

2173
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

0 𝑖𝑓 − ∞ ≤ 𝑌𝑖∗ < 0
𝑌𝑖 = { 1 𝑖𝑓 0 ≤ 𝑌𝑖∗ < 𝜇1 (5)
2 𝑖𝑓 𝜇1 ≤ 𝑌𝑖∗ < +∞

Where, μ1is the threshold estimated based on the observations for severity levels 0, 1,
2.Finally, the goodness of fit of the ordered probit model can be estimated as follows
(Washington et al. 2003).

𝐿(𝛽) (6)
𝜌2 = 1 −
𝐿(0)

Where L(0) is the value of the log-likelihood function when all parameters are zero, L(β) is the
value of the log-likelihood function at its maximum.
Finally, the present study exploresthe suitablerelationship between the share of conflicts and
actual crash data of a site. To model continuous variables such as percentage share of conflict
Multiple Linear Regression (MLR) model has been explored. A multiple linear regression
model with k predictor variables X1,X2,.......Xkand a response Y can be written as:
𝑦 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 + · · · 𝛽𝑘 𝑥𝑘 + 𝜖 (7)

Where ϵ is the residual term of the model and the distribution assumption placed on the
residuals allowsdoing inference on the remaining model parameters. β0,β1,β2..............βkare the
regression coefficients in this model.The R2and adjusted R2 of Multiple Linear Regression
models are formulated as equation (8) and (9).

𝑆𝑆𝑅 𝑆𝑆𝑃 (8)


𝑅2 = = 1−
𝑆𝑆𝑇 𝑆𝑆𝑇

2 𝑆𝑆𝐸/(𝑛 − (𝑝 + 1))
𝑅𝑎𝑑𝑗 =1− 𝑆𝑆𝑇 (9)
−1
𝑛

Where SST is the total sum of squares, SSR is regression sum of squares; SSE is error sum of
squares, n is the number of observations and p is the number of predictors.

4. RESULTS AND DISCUSSION

Oncetheconflicttypesareextracted,theyarecorrelatedwiththebuiltenvironment,traffic
operationsandregulatoryfactorsoftheintersections.Twodifferentmodelsweredeveloped,to
checktheeffectofthesefactorsonconflicttypeandonconflictseverity.Keyfindingsobtained from
the statistical models are summarized in the following sub-sections.

To identify the factors influencing each type of conflictsan MNL model was formed
(shown in Table 4).

Variables Thorough and Turning Overtaking Merging


Coefficient p-Value Coefficient p-Value Coefficient p-Value

2174
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Constant -2.292 0.068* 0.562 0.673 3.863 0.001***


Vehicle Present during
0.153 0.022** 0.496 0.000*** 0.479 0.000***
instance of conflict
Major Road Width 0.102 0.000*** 0.017 0.559 -0.019 0.477
Presence of Traffic Police 0.361 0.517 -1.134 0.075* -2.949 0.000***
First Vehicle – Two Wheeler -1.320 0.000*** 0.733 0.109* 1.228 0.001***
Second Vehicle – Two
-0.949 0.000*** -3.355 0.000*** -1.804 0.000***
Wheeler
First Vehicle – Car -0.424 0.240 1.087 0.021** 0.091 0.804
First Vehicle – Speed
-0.335 0.203 -1.324 0.000*** -0.574 0.022**
Change
Second Vehicle – Speed
-0.723 0.007*** -0.920 0.000*** -1.253 0.000***
Change
Restricted log likelihood -1102.064
McFadden Pseudo R-
0.219
squared
Chi squared 482.361
Degrees of freedom 27
Number of Observations 813
*** 99% Confidence Interval **95% Confidence Interval *90% Confidence Interval

Table3.Factors influencing conflict type of vehicles (MNL model)

Sincethebaseconflicttypeforthemodeliscrossingconflict;theoutcomesoftheMNL model indicate


how a particular factor influence certain conflict types compared to the
basecase.Theeffectofvehiclepresentintheconflictingzonesuggeststhatwithanincreaseinthenumb
er ofvehicles at the physical area of the
intersection,allotherconflictswillincreasecomparedtothecrossingconflict.Thisfinding
indicatesthatwith increase in traffic congestion crossing conflict which generally leads to
severecrashes will reduce, but there may not be any reduction is number of conflicts in
general. The finding supports the established principles ofsafety, that with increase in entering
traffic volume, frequency of
crashincreases.Further,thereisalsoevidencethatwithanincreaseinthewidthoftheminorroad,there
willbe anincreasein“throughandturning”conflicts.Resultsalsohintthatthe
presenceoftrafficpolicepersonnelcouldonlyreducethemergingconflicts.Theeffectofany
changeinspeedoftheapproachingvehicleswastestedandinterestinglyitwasobservedthatin
allcasesreductioninspeedwillresultinfewerconflictsof any type
comparedtothecrossingconflicts,which indicates that random reduction of speed by the
approaching vehicles may notensure fewerconflictsattheun-
signalizedintersections.Thereisalsosomeevidencethatifthefirst vehicle, i.e., the vehicle leaving
the junction is a two-wheeler; there is a higher chance ofmerging
conflictcomparedtootherconflicts.Thisisduetothesmallsizeofthe two-wheelers,due
towhichothervehiclesgenerallyignoretheirpresence and enter the
intersection,increasingthechance of mergingconflicts.Onthe otherhand,whenthetwo-
wheeleristhesecondvehicleortheenteringvehicle in the junction,theirchanceof involving in
crossingconflictishighercomparedtootherconflicts.Thisfindingclearly point towards the risk-
taking behavior of the two-wheeler riders and their high propensity of involving in high risk
traffic maneuvers, such as crossing another vehicle at a right angle, which may lead tohigher
injuryseverities.Whencheckingthe nature of overtakingconflicts,theresultsshowthatchance
oftheovertakenvehiclebeingthecarishigher,indicatingthefact that highshareofovertakingconflict
are taking place at intersections with a high volume of cartraffic. From these results, it can be

2175
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

concluded that un-signalized junctions with high volume of motorcycles may witness higher
number of severe crashes, and the crash frequency will increase with increase in road widths
and volume.

Table 4.Factors influencing severity of conflicts

Characteristics Variables Coefficient t- p-Value


statistics
Constant -2.311 -2.856 0.004
Traffic Exposures Average Vehicular Speed 0.043 4.165 0.000
Occupying Vehicle -0.146 -7.977 0.000
Built
Environmental Width of Minor Road (meters) -0.090 -3.907 0.001
Characteristics
Traffic Operational Presence of Traffic Police 1.857 6.090 0.000
Characteristics Interaction Between First and Second
0.321 3.028 0.002
Vehicles
Percentage Share of Conflicting Vehicles 0.104 2.112 0.034
First Vehicle: Two Wheeler 0.303 3.480 0.000
Second Vehicle: Two Wheeler 0.542 5.031 0.000
Second Vehicle: Four-Wheeler (car) -0.161 -1.658 0.097
Model Summary μ1 0.988 18.929 0.000
McFadden Pseudo R-squared 0.100
Restricted log likelihood -887.834
Number of Observations 813

Further, to identify the factors influencing the severity of conflict an OP modelwas


formed.Afterreviewingtherelevantliterature,theconflicttypes, i.e., thecrossing,turning-and-
through,overtaking,andmerging are ordered asmostsevereto leastsevere(Alhajyaseen 2015)
respectively.Asthe dependentvariableisorderedinnature,
OPmodelcanbeusedtoidentifythefactorsinfluencing
theseverityofconflicts.ThefinalmodeloutcomesaresummarizedinTable4.TheOPmodel
outcomesshowhighervehicularspeedislikelytoresultinmoresevereconflict,thusimplying
thatcrossingconflictsar e exp ected to occuratahigherspeedandarealso
mostsevere.Whenthephysical area of the intersection is occupied by more number of vehicles,
the severity of conflicts
islikelytobelesser,becauseoftherestrictedmovementoftheconflictingvehicles.This is also
confirmed from the previous choice model, where it is observed that all other conflict except
crossing conflict will increase with an increase in vehicle presence at the junction.
Additionally, the severityof conflict is likely to be more when the width of theminor roadis
lesser, probably because theremight be several safety-related inadequacies at the smaller
junctions. This is probably due to the fact that locations withless important minor roadwith
narrower widthprobablyreceive less priority or are noticeable, resulting in more
severeconflictsthan at junctions of two comparable roads. From the model, it isalso evident
that the presence of traffic police is not beneficial toreduce or control the conflict severity.
Further, a very intuitive finding is that with an increase in speedof
vehiclesapproachingthejunction,conflicts resulting in high severityare likelytoincrease.Finally,
the severitywilltendtobehigherifeitheroftheconflicting vehiclesistwo-
wheeler,butlesseriftheentering vehicle or the secondvehicle is acar.

2176
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Characteristics Variables Coefficient t- p-


statistics Value
Constant 13.686 35.925 0.000
Traffic Operational Presence of Traffic Police (1/0) -4.688 -35.934 0.000
Speed Changing Condition of any of the Conflicting 0.162 2.892 0.004
Vehicle (1/0)
Traffic Exposure Speed of the Motorized Traffic (kmph) -0.173 -26.220 0.000
Fatal Crash Frequency (in the Year between 2011-16) 0.134 10.106 0.000
Built Environment Minor Road Width (meter) 0.132 8.920 0.000
Model Summary Adjusted R2 0.631
Number of Observations 813
Table 5. Factors influencing percentage share of conflicts (MLR model)

Lastly, to establish a relationship between crash frequency and the share of the conflicting
vehicles (compared to the total entering traffic of the un-signalized intersection), an MLR
model was developed (refer Table 6). As expected, it was observed that locations with higher
crash counts are linked with a higher share of conflicts, supporting the use of conflict as a
good surrogate measure for the crash. Further, it was also observed that the presence of traffic
police can lead to significant reduction in the total share of conflict, even though the severity
may not be reduced by their presence (refer Table 4). However, variables such as higher
approach speed in general and the speed reduction characteristics by the approaching vehicles
indicated reduction is the share of conflicts, which may apparently look counterintuitive.
However, it is important to note that speed has a significant role in crash severity and not as
much on the crash occurrence and that lesser approach speed probably indicating higher
traffic volume which generally increasesthe number of conflicts.

5. CONCLUSION

This study was conducted in order to identify the built environment, traffic operation and
traffic exposures related factors which significantly affect conflict type, and conflict severity
at the un-signalized intersections of Kolkata city. A secondary intention of the present study
was to form an apposite relationship between conflicts and police recorded crash data. Based
on the results obtained from this study, the following conclusions may be drawn.

• Based on the video-data it is evident that there are predominantly four types of
conflicts in a typical un-signalized junction in Kolkata. These are – i) Crossing, ii)
Turning and Through, iii) Overtaking, and iv) Merging
• Two-wheelers with higher speed are mostly responsible for the conflicts with higher
severities such as crossing conflicts.
• Study outcomes show that conflict severity significantly reduces with the increase in
traffic volume at the intersection.When the physical area of the intersection is
occupied by more number of vehicles, the severity of conflicts is likely to be lesser,
because of the restricted movement of the conflicting vehicles.
• Further, the study results indicate that random reduction of speed by theapproaching
vehiclesmaynotensurethe reductioninover all

2177
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

conflictseveritiesattheunsignalizedintersectionsinKolkata.Thussuitablespeedreduction
measuresneedtobeadoptedtoeffectivelybring down the approachspeeds.
• Minor road width of an un-signalized junction provides a very imperative aspect for
traffic safety. If minor carriageway width is wider, the number or share of conflict will
be higher, but interestingly conflicts are found to result in more severe types in case of
narrow minor road width. This is an interesting observation, which consistently
emerged in other studies performed with crash data (Bhowmic and Mitra 2017). This
is probably due to lack of attention in junction treatment and traffic operations in case
of intersections major and small minor roads.
• Presence of traffic police can be effective for reducing the total counts of conflicts or
the share of conflict, but when it comes to the conflict severity, it cannot be reduced or
controlled significantly by traffic police.
• Finally, there is clear evidence that conflict analysis techniques can be utilized as a
surrogate measure of road traffic safety, as results show that locations with a higher
share of conflicts also recorda highernumber of crashes.

Likeany other study, this study is also limited to certain aspects. Although the developed
models provide acceptable estimates, increased sample size would have resulted in a better
estimation of the variables with a better confidence interval. For a better estimation of the
temporal variation of the conflicts, the extraction duration should have been increased.
Though intersections from different parts of the city have been tried to incorporate in the
model, model validation for upcoming years needs to be done and transferability of the
models for other uncontrolled junctions of Kolkata is yet to be tested. The geographical
stability of the models is yet to be validated. Based on the models obtained and the factors
identified, appropriate measures and their effectiveness can also be checked in further study.
Also, a corridor level safety analysis can be conducted for better identification of safety-
related issues to intersections on a particular stretch of road. Although there is ahuge scope of
further research in this field, findings from the present study can significantly assist in policy-
making decisions that are aimed at improving the safety of un-signalized junctions in Kolkata.

2178
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

REFERENCES

1. Alhajyaseen, W. K. (2015)The integration of conflict probability and severity for the


safety assessment of intersections. Arabian Journal for Science and
Engineering, 40(2), 421-430.
2. Ambros, R. T., Paukrt, J., Ambros, J., Turek, R., &Paukrt, J. (2014, November). Road
safety evaluation using traffic conflicts: pilot comparison of micro-simulation and
observation-Jiří. In International Conference on Traffic and Transport Engineering-
Belgrade.
3. Amundsen, F. H., Hyden, C. (1977) Proceedings of first workshop on traffic
conflicts. Oslo, TTI, Oslo, Norway and LTH Lund, Sweden.
4. Archer, J. (2005) Indicators for traffic safety assessment and prediction and their
application in micro-simulation modelling: A study of urban and suburban
intersections (Doctoral dissertation, KTH).
5. Autey, J., Sayed, T., Zaki, M. H. (2012) Safety evaluation of right-turn smart
channels using automated traffic conflict analysis. Accident Analysis &
Prevention, 45, 120-130.
6. BhowmicK, K., S. Mitra. (2017) Status of signalized intersection safety- A case study
of Kolkata; Road Safety and Simulation International Conference; The Hague, NL,
17-19 Oct. 2017
7. Brilon, W., Wu, N. (2001).Capacity at unsignalized intersections derived by conflict
technique. Transportation Research Record, 1776(1), 82-90.
8. De Mouzon, O., Pham, M. H., El Faouzi, N. E., Chung, E. (2007) Road safety
indicators: Results in Vaud canton (No. CONF).
9. Deshpande, N., Chanda, I., Arkatkar, S. S. (2011) Accident mapping and analysis
using geographical information systems. International Journal of Earth Sciences and
Engineering, 4(6), 342-345.
10. El-Basyouny, K., Sayed, T. (2012) Measuring safety treatment effects using full
Bayes non-linear safety performance intervention functions. Accident Analysis &
Prevention, 45, 152-163.
11. El-Basyouny, K., Sayed, T. (2013) Safety performance functions using traffic
conflicts. Safety Science, 51(1), 160-164.
12. Engel, U. (1985). Validation of conflicts-studies: an international review.
13. Gettman, D., Head, L. (2003) Surrogate safety measures from traffic simulation
models. Transportation Research Record: Journal of the Transportation Research
Board, (1840), 104-115.
14. Glauz, W. D., Bauer, K. M., Migletz, D. J. (1985) Expected traffic conflict rates and
their use in predicting accidents. Transportation Research Record, 1026, 1-12.
15. Glauz, W. D., Migletz, D. J. (1980) Application of traffic conflict analysis at
intersections (No. HS-028 882).
16. Grayson, G. B., Hyden, C. K., JH, M., N. Oppe, S. (1984) The Malmö study: a
calibration of traffic conflict techniques. A study organised by ICTCT (The
International Committee on Traffic Conflict Techniques).
17. Gupta, H., Rokade, S. (2016) Development of Crash Prediction Model Using
Multiple Regression Analysis.
18. Hayward, J. C. (1972). Near miss determination through use of a scale of danger.

2179
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

19. Huang, P., Pant, P. (1994) Simulation neural-network model for evaluating dilemma
zone problems at high-speed signalized intersections. Transportation Research
Record. 1456: 34–42.
20. Ismail, K., Sayed, T., Saunier, N. (2010a) Automated safety analysis using video
sensors: technology and case studies. In Canadian Multidisciplinary Road Safety
Conference, Niagara Falls, Ontario.
21. Ismail, K., Sayed, T., Saunier, N. (2010b) Automated analysis of pedestrian-vehicle:
conflicts context for before-and-after studies. Transportation Research Record:
Journal of the Transportation Research Board, (2198), 52-64.
22. Ismail, K., Sayed, T., Saunier, N., Lim, C. (2009a) Automated analysis of pedestrian–
vehicle conflicts using video data. Transportation research record, 2140(1), 44-54.
23. Ismail, K., T. Sayed, N. Saunier. (2009b) Automated pedestrian safety analysis using
video data in the context of scramble phase intersections. In: Annual Conference of
the Transportation Association of Canada, Vancouver, BC.
24. Kadali, B. R., Vedagiri, P. (2016) Proactive pedestrian safety evaluation at
unprotected mid-block crosswalk locations under mixed traffic conditions. Safety
science, 89, 94-105.
25. Kulmala, R. (1993) Safety evaluation of traffic systems: traffic conflicts and other
measures.Proceedings of the 6th ICTCT Workshop. Austrian Road Safety Board,
Salzburg. 2699271.
26. Li, R., Li, W., Li, L., Zheng, C., Ran, B., Cheng, Y. (2011) Crash severity evaluation
for unsignalized intersection using conflict data. International Journal of
Computational Intelligence Systems, 4(6), 1325-1333.
27. McKelvey, R. D., Zavoina, W. (1975) A statistical model for the analysis of ordinal
level dependent variables. Journal of mathematical sociology, 4(1), 103-120.
28. Migletz, D.J., W.D. Glauz, K.M. Bauer. (1985) Relationships between Traffic
Conflicts and Accidents.US Department of Transportation, Federal Highway
Administration.FHWA/RD-84/042.
29. Nakamura, H., MABUCHI, T. (2007).Performance evaluation of roundabouts
considering traffic conflicts. In Proceedings of the Eastern Asia Society for
Transportation Studies Vol. 6 (The 7th International Conference of Eastern Asia
Society for Transportation Studies, 2007) (pp. 316-316). Eastern Asia Society for
Transportation Studies.
30. Older, S. J., Spicer, B. R. (1976) Traffic conflicts—a development in accident
research. Human Factors, 18(4), 335-350.
31. Parker, M. R., Zegeer, C. V. (1989) Traffic conflict techniques for safety and
operations: Engineers guide.
32. Patil, G., Pawar, D. (2014) Temporal and spatial gap acceptance for minor road at
uncontrolled intersections in India. Transportation Research Record: Journal of the
Transportation Research Board, (2461), 129-136.
33. Perkins, S. R., Harris, J. I. (1967) Criteria for Traffic Conflict Characteristics,
Signalized Intersections. Research Laboratories, General Motors Corporation.
34. Persaud, B. N., Mucsi, K. (1995) Microscopic accident potential models for two-lane
rural roads. Transportation Research Record, (1485).
35. Prajapati, P., Tiwari, G. (2013) Evaluating safety of urban arterial roads of medium
sized Indian city. In Proceedings of the Eastern Asia Society for Transportation
Studies (Vol. 9).

2180
Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

36. Prasetijo, J., Wu, N., Ambak, K., Sanik, M. E., Daniel, B. D., Hadipramana, J. (2016).
Performance of Non-priority Intersections Under Mixed Traffic Conditions Based on
Conflict Streams Analysis. Transportation in Developing Economies, 2(1), 2.
37. Priyadarshini, P., Mitra, S. (2018) Investigating pedestrian risk factors leading to
pedestrian fatalities in Kolkata city roads. Transportation in developing
economies, 4(1), 1.
38. Rao, V., Regaraju., (1998) Modeling conflicts of heterogeneous traffic at urban
uncontrolled intersections. Journal of Transportation Engineering.124(1), 23-34.
39. Salman, N. K., Al-Maita, K. J. (1995). Safety evaluation at three-leg, unsignalized
intersections by traffic conflict technique. Transportation Research Record, 1485(28),
177-185.
40. Sayed, T., Brown, G., Navin, F. (1994) Simulation of traffic conflicts at unsignalized
intersections with TSC-Sim. Accident Analysis & Prevention, 26(5), 593-607.
41. Sayed, T., Ismail, K., Zaki, M. H., Autey, J. (2012) Feasibility of computer vision-
based safety evaluations: case study of a signalized right-turn safety
treatment. Transportation Research Record, 2280(1), 18-27.
42. Sayed, T., Zein, S. (1999).Traffic conflict standards for intersections. Transportation
Planning and Technology, 22(4), 309-323.
43. Songchitruksa, P., Tarko, A. P. (2006)The extreme value theory approach to safety
estimation. Accident Analysis & Prevention, 38(4), 811-822.
44. Spicer, B. R., Wheeler, A. H., Older, S. J. (1980). Variation in vehicle conflicts at a
T-junction and comparison with recorded collisions (No. TRRL SR545 Monograph).
45. Tiwari, G., Mohan, D., Fazio, J. (1998) Conflict analysis for prediction of fatal crash
locations in mixed traffic streams.
46. Van der Horst, R., Hogema, J. (1993) Time-to-collision and collision avoidance
systems (pp. 109-121).na.
47. Vedagiri, P., Killi, D. V. (2015) Traffic safety evaluation of uncontrolled intersections
using surrogate safety measures under mixed traffic conditions. Transportation
Research Record: Journal of the Transportation Research Board, (2512), 81-89.
48. Washington, S. P., Karlaftis, M. G., Mannering, F. (2010) Statistical and econometric
methods for transportation data analysis.Chapman and Hall/CRC.
49. William, T.M. (1972)An evaluation of traffic conflict technique. Highway
Research.Record, pp. 384: 1–8.
50. Workineh, A. A. (2014). Analysis of the relationship between traffic conflicts and
level of service at four-legged, signalized intersections in Sacramento.
51. www.indev-project.euAccessed July. 29, (2018)
52. Zhang, G., Chen, J., Zhao, J. (2017). Safety Performance Evaluation of a Three-Leg
Unsignalized Intersection Using Traffic Conflict Analysis. Mathematical Problems in
Engineering, 2017.

2181

You might also like