Sunteți pe pagina 1din 15

University of kirkuk – College of engineering

Civil department – Third stage (A)

Signalized intersection

Supervised by: Huda Nadhir


Prepared by: Mahmood Younis Mohammed
Abstract
Urban signalized intersections cater high volumes of traffic daily and are aimed at
separating conflicting traffic flows on a temporal basis. But preliminary analysis of
crash data obtained from the Statistical section of Kolkata Police shows a significant
share of crashes occurring at signalized intersections in Kolkata. This study intended
to find out the problems associated with signalized intersections from the
perspectives of geometric design and infrastructure present, and identify the risk
factors affecting the frequency and severity of crashes. Occurrence of crashes at
signalized intersections would be associated with reasons such as the infrastructure
planning and geometric design elements present at the intersection, the traffic
operation and management as well as the behavior of the road users. Assuming that
user behavior is influenced by the design and traffic operational characteristics of a
location, this study specifically aims to investigate the effects of geometric,
infrastructural, traffic control and land use parameters on number of crashes and their
severity across 52 signalized intersections of Kolkata. These data were associated
with crash records from 2011 through 2014, to extract important insights, using
statistical tests such as correlation analysis and independent samples t-test, as well as
multivariate regression analysis for crash prediction purposes. Based on the
assessments carried out in this study, there is evidence that factors like total approach
traffic volume, blocked carriageways, traffic configuration, presence of exclusive
turning phase, etc. significantly affects the frequency of crashes while pavement
condition, presence of all red time, presence of footpaths and zebra crossings, number
of schools affect the severity of crashes at signalized intersections of Kolkata. In the
absence of other road traffic safety studies involving rigorous statistical techniques,
these findings can significantly contribute in policy making decisions that are aimed
at improving the safety of signalized intersections in the city of Kolkata.

1.Introduction
1.1 Background

With economic prosperity, motorization is on the rise in developing countries like


India. Road traffic deaths in urban India have consistently been a major issue of
concern. NCRB 2014 reports that road traffic crashes has been the leading cause of
unnatural death among men and the second highest cause of unnatural death among
women in India [1]. Even though urban areas cover a small percentage of land area as
compared to rural India, the number of crashes is significantly higher in urban areas.
NCRB 2014 also reported that among all the metropolitan cities in India, Kolkata, one
of the metropolitan cities in India, experienced the highest spike in Road Traffic
accidents, an overwhelming 13.9 percent increase in 2014 as compared to 2013. Nearly
59% of these crashes occurred at intersections. Crashes at signalized intersections
accounted for around 30% of all crashes that took place in Kolkata, even though the
traffic signals offer a better separation of conflicting movements and help improve the
overall traffic management through intersections. Signalized intersections are major
blackspots in an urban road network [2, 3]. A better understanding of the crash
causative factors aids to develop more targeted countermeasures for improving the
safety of signalized intersections [4]. It is thus, essential to check the signal design
parameters as well as the infrastructure planning, land use and the overall traffic
condition which are critical for the safety performance of the signalized intersection. In
this background, 52 signalized intersections from the city of Kolkata were chosen for
detailed investigation.

1.2 Literature review

Safety at signalized intersections has been extensively studied by researchers in


different developed countries of the world like the USA [5-9] and in several European
countries [10] as well as in some emerging economies like Brazil [11], China [12],
Hong Kong [13] and Singapore [14]. These studies have pointed out that there may 2 be
a variety of factors responsible for crash occurrence at signalized intersections. These
are termed as risk factors. These may be geometric features, infrastructural features,
traffic control measures, driver factors and environmental factors. Although previous
studies have mentioned the importance of human factors influencing crashes, they have
not included it in their models. An earlier research suggested that roadway
characteristics, such as geometric design, infrastructural design, and environmental
conditions account for 28 percent of all traffic accidents, either alone or in combination
with road user and vehicle factors [15]. This indicates a common shortfall of the
previous studies. But, the primary focus of this study has been the removal of geometric
deficiencies and improvement of infrastructure. Road user behaviour has been assumed
to be influenced by infrastructural features and traffic characteristics and hence was not
incorporated separately in this study. Previous studies have shown that increase in
traffic volume, proportion of commercial vehicles in the traffic stream, number of signal
phases, protected left turn, number of bus bays, speed limit, number of lanes, road
surface conditions and land use pattern influence crashes, either in a negative or in a
positive manner. But in developing countries of south Asia like India, Pakistan, Sri
Lanka etc., safety studies are not well documented. Safety assessment using rigorous
statistical techniques is still in a nascent stage in India owing to the absence of
systematic crash data collection and challenges in obtaining proper site inventory and
traffic volume data. Apart from one study [16] that was conducted in three non-
metropolitan cities, existing studies regarding road traffic safety in urban India either
follow road safety audit technique or a descriptive approach [17, 18].
1.3 Research Objective
The principal objective of this study was to identify the factors related to traffic and
built environment such as road geometrics, infrastructure and land use which affects the
frequency and severity of crashes occurring at signalized intersections of Kolkata. A
secondary objective of this study was to associate crash frequency and severity at with
built environment and traffic characteristics, with the help of multivariate regression
models for the purpose of prediction of crash frequency and fatal crash occurrence.
Earlier studies performed in Kolkata were highly qualitative and descriptive in nature,
lacking any clear directives towards risk reduction through appropriate infrastructure
planning and design which this study aims to fill.

2. Methodology
2.1 Crash data collection
Crash data of the city of Kolkata, for 4 years (2011-2014), was obtained from Kolkata
Police. The data included details like severity of crash, year of occurrence, place of
occurrence, types of vehicles involved, number of persons involved in the crash and the
time of occurrence of the crash. The entire city has been divided by Kolkata police in to
25 traffic guards for more efficient administration. This study is concerned with only
selected signalized intersections located across nine traffic guards. Crash data pertinent
to the intersections included in the study were extracted. Summary statistics of the crash
data included in this study are shown in Table-1.

2.2 Site inventory data collection


In order to statistically associate geometric and infrastructural features with road
crashes, collection of site inventory data was necessary. Site visits were conducted at 52
signalized intersections to obtain data regarding road geometric features, infrastructural
details, land use and traffic control parameters by direct observation. The distribution of
the intersections according to traffic guards are shown in Table-2.
Site inventory data included geometric features like carriageway width, footpath width,
type of junction, infrastructural information like pavement condition, presence of
pavement markings, presence of medians and land use information like percentage share
of residential zone, commercial zone, school zone, etc. Data regarding traffic control
measures like presence of ‘All Red time’, number of cycles per signal phase, presence
of protected right turn, were also recorded. All the signals were ‘fixed time’ signals. The
descriptive statistics of the site inventory data are provided in Table-3.

2.3 Traffic volume data collection

Traffic volume, composition and characteristics have been found to be significant


variables in similar studies conducted in developed countries, which made it necessary
to collect traffic data as well. Traffic volume data was extracted from surveillance
camera footages placed at signalized intersections by Kolkata Police. Traffic volume
data included approach volumes from major and minor road, right turn volume,
composition of traffic by vehicle type, etc. Directional volume count was taken using 1
hour traffic volume data. For the study, 24-hour volume count was taken at one
representative location of the city in order to obtain Hourly Expansion Factors which
was used to convert 1 hour traffic volume into 24 hour traffic volume. The descriptive
statistics of extracted traffic volume data are provided in Table-4.

3. Analysis and Results


3.1 Preliminary crash data analysis

Preliminary analysis of crash data was performed in order to obtain a better


understanding of the crash data. During this period of 4 years, i.e. 2011-2014, a total of
8324 crashes took place in urban area of Kolkata. Out of these crashes, 1736 crashes
were fatal in nature, while 6588 crashes were non-fatal in nature. The percentage of
fatal crashes had been 18%, 23%, 20% and 21% which reflects that there is no
significant reduction in the fatalities from 2011 through 2014. Finally, from the location
of crashes provided in the crash data, it was found that 59% of the crashes took place at
intersections, while 30% of all crashes had occurred at signalized intersections which
clearly indicated that most intersections in Kolkata are not just potential, but existing
black spots.

3.2 Volume data analysis

The traffic of Kolkata is heterogeneous in nature, where different vehicles have


different impacts on a traffic stream, depending upon their nature, size and speed.
Hence, Passenger Car Equivalents for different vehicles were considered in order to
convert the heterogeneous traffic volume in to a homogeneous one [19]. Passenger Car
Equivalent values for different vehicle types as per Indian Roads Congress SP 41-1994
are given in Table-5.

As mentioned in section 2.3, 24 hour traffic volume was taken at one representative
location in the city of Kolkata. Hourly volumes were recorded at all other study
locations, and were multiplied with their respective Hourly Expansion Factors in order
to achieve 24 hour traffic volume.

3.3 Identification of risk factors

3.3.1 Correlation analysis

In order to understand the strength of relationship between two variables, correlation


analysis was performed first. While correlation analysis is often considered as a
prerequisite for the multivariate modeling process, it can be performed independently in
order to achieve important insight with respect to the overall nature of the data, as well
as identification of certain factors and their impact on safety. A strong or high
correlation means that two or more variables have a strong relationship with each other,
while a weak or low correlation means that the variables are hardly related. If the value
is positive, it indicates a positive correlation, which means that when the value of one
variable increases, the value of the other one increases as well, and vice versa. If the
correlation coefficient has a negative value, then it indicates that when the value of one
variable increases, the other one decreases and vice versa. Since the data obtained
contains variables that are continuous, ordinal and nominal in nature, it violated the
assumption of the Pearson product moment correlation. Hence the Spearman’s rank
order correlation was selected which is a non- parametric version of the Pearson’s
correlation. Independent variables that are highly correlated with the dependent variable
(crash frequency or fatal crash frequency) should be included in the final model. But,
two independent variables having high correlation should not be included in the model
at the same time. Results of correlation analysis are shown in Table-6.

3.3.2 Independent samples t-test

The independent samples t-test is typically performed in order to compare the means
between two unrelated groups, and determine whether the means of those two groups
are significantly different or not. In this study, the factors which were thought to be
posing risk (or risk factors), were checked against their presence or absence in the
dataset. The null hypothesis and the alternate hypothesis are stated as follows.

H0: µ1 = µ2 (The two population means are equal)


H1: µ1 ≠ µ2 (The two population means are not equal)
Or,
H0: µ1 - µ2 = 0 (The difference between the two population means is zero)
H1: µ1 - µ2 ≠ 0(The difference between the two population means is not zero)
Here µ1 and µ2 are the population means of two unrelated groups, i.e. in this case,
difference in means of crashes between the groups where the risk factor is present and
where the risk factor is absent

. In order to check for the equality of population variances, a Levene’s test is


conducted. The hypotheses for the Levene’s test are as follows.

H0: σ1 2 - σ2 2 = 0 (The difference between the two population variances is zero)


H1: σ1 2 - σ2 2 ≠ 0(The difference between the two population variances is not zero)
Here σ1 and σ2 are the population variances of two unrelated groups, ie. in this case,
difference in variance between the group where risk factor is present and the group
where the risk factor is absent.

The test statistic for the independent samples t-test is denoted as t. There are two
forms of the test statistic, one for equal and the other for unequal population variances.
When the two independent samples are assumed to be drawn from populations with
identical population variances (i.e., σ1 2 = σ2 2 ), the test statistic t is computed as:

Where,
x1 = Mean of first sample
x2 = Mean of second sample
n1 = Sample size (i.e., number of observations) of first sample
n2 = Sample size (i.e., number of observations) of second sample
S1 = Standard deviation of first sample S2 = Standard deviation of second sample
Sp = Pooled standard deviation

The calculated t value is then compared to the critical t value from the t distribution
table with degrees of freedom df = n1 + n2 - 2 and chosen confidence level. If the
calculated t value is greater than the critical t value, then the null hypothesis is rejected.
In this study, the statistical significance of the difference in mean crash frequency was
tested against two groups of intersections with certain exposure being present in one
group and absent in the other. A positive t value and a significance value of less than the
desired level of significance (α) indicates that the presence of the exposure is seen to
experience greater number of crashes than its absence. Such an exposure can be termed
as a risk factor. The results of the independent samples t-test are shown in Table-7. In
the table, only 95% confidence interval has been shown. Only variables which were
significant at least at a significance level of 0.10 were included in the table. N1
represents the number of locations where the exposure variable is present, while N2
denotes the number of locations where it is absent. As can be noticed in Table-7,
locations where the exposure variables belong under the headings of traffic volume and
infrastructural features were seen to have experienced significantly greater number of
crashes. On the other hand, the locations which saw the presence of protected right turn
signal phase were seen to have experienced significantly lesser number of crashes. It is
to be noted that, the t-test was not meant to be used as a screening method of variables
to be included in the multivariate models. By performing an additional test, we were
able to draw better insight regarding the causes of total and fatal crash occurrences in
the study area of Kolkata.

3.4 Model selection

In order to associate crash data with junction parameters, traditional count data
models like Poisson regression model and Negative Binomial regression model were
employed [20-24]. Owing to the presence of overdispersion in the crash data, Negative
Binomial model was chosen to predict crash frequency [25-27]. In addition to that, a
Binary logistic regression model was used to calculate the probability of occurrence of
fatal crashes [29-31].

3.5 Model development

Before starting with the modeling process, a correlation analysis was performed in
order to find the independent variables which were highly correlated with crashes.
Initially, the correlation matrix was developed using all the 69 independent variables
and then the chosen variables were checked to avoid multi-collinearity as well as wrong
signs or questionable magnitudes in the estimated coefficients. In order to achieve only
the uncorrelated variables from the total set of independent variables, only one variable
among the correlated variables was taken which had the most significant and plausible
correlation with the dependent variable. Afterwards, during successive testing, the
variables coming out to be insignificant or having implausible signs were omitted one
by one. After several trials, the final model was selected based on their goodness of fit,
sign of the coefficients and their statistical significance. For modeling all types of
crashes and fatal crashes, Poisson regression model and negative binomial regression
models were developed, which are also termed as Safety Performance Functions.
Further a binary logistic regression model was developed to identify elements of
planning, designs and operations associated with the probability of fatal crash
occurrence.

3.5.1 Safety Performance Function using negative binomial regression


model

In order to express the crash frequency in terms of geometric, infrastructural, traffic


signal and traffic volume variables, a Safety Performance Function was developed. As
mentioned earlier in section 3.4, since the variance of total crashes (11.61) was found to
be significantly greater than the mean (4.96), negative binomial regression was adopted
to build the SPF. The final model has been shown in Table-8.
3.5.2 Fatal crash prediction model using binary logistic regression
In order to develop an understanding about the variables significantly affecting the
probability of fatal crash occurrence at a particular signalized intersection, a Fatal Crash
Prediction model was developed. In this model, an attempt was made to associate the
geometric, infrastructural, traffic control and traffic volume variables with a
dichotomous variable as the dependent variable. The dependent variable assumed a
value of ‘1’ for locations where at least one fatal crash took place and ‘0’ where no fatal
crashes were recorded. Considering the nature of the dependent variable, binary logistic
regression was adopted to develop the final model. The final model has been shown in
Table-9.
4. Discussion

The broad objective of the studding Signalized intersection was to identify such
factors whose improvement, presence or absence would reduce the likelihood of crash
occurrence and consequently enhance the safety of signalized intersections in Kolkata
city. In this study, a host of different statistical tools were used to identify risk factors.
Correlation analysis and independent samples t-test helped us in the identification of
factors that significantly impact the safety of signalized intersections. The results of
these tests indicated that high approach traffic volume, high minor road traffic volume,
high volume of heavy commercial vehicles, blockage of carriageways by car parking
and encroachment, presence of tram lines and absence of protected right turn phase
were risk factors. But they were unable to quantitatively predict how much the increase,
decrease, presence and absence of certain elements of planning, design and operations
would affect the number of crashes. With the help of the Safety Performance Function
and The Fatal Crash Prediction model, it was possible to predict the number of crashes
and the probability of fatal crash occurrence respectively, at any signalized intersections
of Kolkata. Based on the Safety Performance Function, it can be deduced that variables
like total approach traffic volume, blocked major carriageway, two way traffic and car
parking on minor road are positively affecting the number of crashes while a protected
right turn phase have been negatively affecting the same. From the Fatal crash
prediction model developed, it can be inferred that among all the variables, presence of
a protected right turn phase was found to be significantly reducing the probability of
fatal crash occurrence. Also, an increase in the ratio of minor road traffic volume to
major road traffic volume was found to enhance the safety of an intersection by
reducing the likelihood of fatal crashes. On the other hand, blocked carriageway,
percentage of NMT and total approach volume were found to be significantly increasing
the probability of fatal crash occurrence. It is to be noted that some factors such as
carriageway width, traffic volume, land use pattern, etc. which were found to be
significantly affecting the crash frequency and severity, are not always possible to alter.
However the analysis provides a clear direction as to the type of risk factors based on
which new traffic management strategies and operations could be adopted. For example,
reducing approach traffic volume at any particular intersection requires enormous
mitigating policies involving large investments, and hence not possible for a developing
nation like India. So, locations with high approach traffic volume clearly require greater
attention and improved police enforcement so as to get immediate benefit, and call for
improved traffic planning and design for the long run.
5. Conclusions

Based on the results obtained from several statistical tests and models conducted in
this study, the following conclusions may be made. The infrastructure and traffic related
risk factors identified from the study are in similar tone with those obtained elsewhere.
For example, higher approach traffic volume, wider carriageway width, absence of
exclusive signal phase for turning traffic are all established in safety literature, and this
study confirms these findings. However, this study provides insights about several other
risk factors which may be unique to Kolkata, or other cities of India or cities of other
developing countries. Such variables are car parking on approach lanes, encroached
carriageways, percentage of non-motorized traffic, which are found to be major risk
factors with regards to crash occurrence. Since the law enforcement tends to be less
strict in developing countries and road users are not always aware of traffic rules,
presence of the above mentioned elements are consistently found to be associated with
higher risk and came out to be highly significant. Hence, safety improvement initiatives
in cities of developing countries should pay specific attention on removal of such
factors. Similar to other studies, this study is not without its limitations. Based on the
site visits, several issues were identified with respect to road geometric features,
infrastructure, traffic control and human behavior (both drivers and pedestrians). Some
of the observed issues were road and footpath encroachment, unauthorized onstreet
parking, absence of footpaths in minor roads, improper road markings, blockage of sight
distance, permissive right turn and absence of all red time or dedicated pedestrian phase.
But, the biggest issue that was observed was the non conformance with the traffic rules
by road users. But this study has not taken into account the perception and behavior of
the road users and hence, incorporation of human factors and conduction of a perception
behavior analysis of road users would certainly help in a better understanding of the
safety issues. Also, an increase in sample size may lead to a better understanding of
safety issues which are yet to be 10 unearthed. The geographical stability of the models
is yet to be validated, but has served to give us important insights regarding safety of
urban signalized intersections in a major city like Kolkata. These are recommended as
future work.
REFERENCES
El-Hakim, O., W. Clyma, and E. V. Richardson. 1988. Performance
functions of border irrigation systems. J. Irrig. and Drain. Eng.,
ASCE 114(1):118-129.
Hart, W. E., H. G. Collins, G. Woodard, and A. J. Humphereys.
1980. Chapter 13: Design and operation of surface irrigation
systems. In Design and Operation of Farm Irrigation Systems.
ASAE Monograph No. 3. M. E. Jensen, ed. St. Joseph, Mich.:
ASAE.
Holzapfel, E. A., M. A. Marino, and J. Chevez-Morales. 1986.
Surface irrigation optimization models. J. Irrig. and Drain.
Eng., ASCE 104(3): 275-281.
Holzapfel, E. A., and M. A. Marino. 1987. Surface-irrigation
nonlinear optimization models. J. Irrig. and Drain. Eng., ASCE
113(3): 379-391.
Reddy, J. M. 1980. Irrigation system improvement by simulation
and optimization. PhD diss. Fort Collins, Colo.: Colorado State
University.
Reddy, J. M., and W. Clyma. 1981. Optimal design of border
irrigation systems. J. Irrig. and Drain. Div., ASCE 107(3):
289-306.
Shatanawi, M. R., and T. Strelkoff. 1984. Management contours for
border irrigation. J. Irrig. and Drain. Eng., ASCE 110(4):
393-399.
Strelkoff, T., and M. R. Shatanawi. 1985. Normalized graphs of

S-ar putea să vă placă și