Assessment Of Level Of Service Of Safety Of Two Lane Highways

ASSESSMENT OF LEVEL OF SERVICE OF SAFETY OF
TWO LANE HIGHWAYS USING ROAD SIDE LAND USE
Dr. RanjaBandyopadhyaya
Assistant Professor
NIT Patna
Mr. JiweshUjjal
M. Tech Student
Dr. Sanjeev Sinha
Professor
NIT Patna
ABSTRACT
Identifying high crash locations is essential for road safety improvement. Locations may be categorized based
on degree of hazard and safety improvement potential into level of service of safety (LOSS) groups.
Determining LOSS requires complex mathematical models and data of crash history, traffic exposure and road
inventory.
The present work develops a guideline for determining LOSS of two-lane rural highway with road geometry and
roadside features without systematic crash and traffic volume data or predication models. The study, done in
Patna, Bihar, identified unique road side and land use factors that govern the LOSS of similar road segments.
1.
Introduction
Growing trend in road crash occurrence, particularly in developing countries, calls for co-ordinated and
focussed effort to improve road safety scenario. Road safety improvement strategies include systematic
identification and improvement of high crash locations. It is difficult to demarcate a location as truly safe or
truly hazardous. In order to ascertain the degree of hazard on a road location, Kononov &
Allery(2003)introduced the concept of level of service of safety (LOSS) for categorising road sections into
different LOSS groups based on their safety performance and potential for safety improvement.
Determination of LOSS of a road location requires complex mathematical models (Safety Performance
Functions), crash history data, traffic exposure of the roads and road inventory data. Thus, it is imperative to
relate simple easy to assess factors with LOSS of roads so that determination of LOSS of roads become easy.
Identification of such factors may help in categorizing roads into different LOSS groups in absence of
systematic crash data and safety performance functions.
The present work aims to develop a guideline for determination of LOSS with road features of two lane rural
highways so that LOSS can be assessed without systematic collection of crash data and use of predication
models and traffic volume study.
Section 2 gives the background literature focussing on development of LOSS of road segments and road and
roadside features contributing to the safety performance of a road stretch. Section 3 discusses the methodology
adopted for the present work. Section 4 describes the data collected for the present work and section 5 details
the analysis results. Section 6 summarizes the conclusions.
2.
Background
The concept of level of service of safety (LOSS)was introduced in the framework of safety by Kononov &
Allery(2003).Similar to the concept of Level of Service, LOSS tries to measure and categorize safety
performance of a road section into pre-defined safety categories on the basis of Accident Reduction Potential of
a site. Four LOSS categories are proposed, where LOSS-I indicates a low potential for accident reduction and
LOSS-IV indicates a high potential for accident reduction. Pan et al. ,(2007) applied the concept
of safety level of service to highway intersections, analyzed factors that influence intersectional traffic safety,
and presented the methodologies used in developing the models of safety level of service at highway signalized
intersections. Based on conflict points, a basic model is initially built, and then based on intersection geometry
features, traffic signs, a modified model is built. Kononov & Allery(2009) implemented the LOSS concept at
Louisiana Department of Transportation and Development (LADOTD) and also showed that crash rate changes
with the Annual Average Daily Traffic (AADT). Between 1990 and 2004 AADT Increased from 36,010 to
77,682, Total Accident Rate Increased by 146%, Injury and Fatal Rate Increased by 60%. In Order to
understand how the Crash Rate changes, there is a need to develop a relationship between Safety and Traffic
Exposure. The Highway Safety Manual (AASHTO, 2013)uses Safety Performance Functions(SPF) to find the
predicted average crash frequency at base condition and comparing with the observed average frequency using
standard deviation and LOSS limit, a roadway segment is levelled as LOSS 1,2,3 or 4. The Highway Safety
Manual (HSM)gives a set of predictive tools for developing safety management programs and evaluating their
effects, estimating crash frequency and severity and as a catalogue of crash modification factors (CMF) for
many geometric and operational treatments. HSM gives predictive methods for estimating expected crash
frequency of a network or individual site. Predictive methods are available for segments and intersections on
rural two-lane two-way roads, rural multi-lane highways and urban and suburban arterials. The models proposed
in HSM can predict expected average crash frequency for given traffic volume and roadway characteristics.
Some research have focussed on development of SPFs and AMFs for Indian conditions. Srinivas et al.,
(2007)have employed Poisson and negative binomial modelling to develop road accident prediction models for
highways under mixed traffic conditions. Crash data from various highway sections in the state of Tamil Nadu
in India have been used in this study to develop the accident prediction models. The authors have opined that the
accident prediction models reported from the developed countries treat traffic flows to homogeneous and are not
appropriate for developing countries such as India, where the traffic is mixed. Models were proposed separately
for the various severity levels of accidents and it was shown that the traffic composition has significant effect on
occurrence of road accidents. The other variables which were shown to have significance in the models included
lane width, shoulder width and number of driveways. Vivian & Veeraragavan (2007) have proposed accident
prediction factors for rural highway segments in developing countries, based on statistical analysis of road
accident data from the state of Karnataka in southern India. The authors have proposed multi-regime Poisson
and negative binomial models for highway segments under different conditions of roadway width, traffic
volume and composition. Accident prediction factors, which when multiplied by average spot speeds giving the
expected number of road accidents, were proposed based on the models for different ranges of pedestrian
exposure. The authors have cited the unavailability of quality crash data, due to which disaggregate models for
accident type and severity could not be built. Dinu(2013) have developed the accident prediction model for rural
undivided segment. He proposed different models for single vehicle and multi vehicle accidents. Hedeveloped
models to evaluate the safety performance of two-lane undivided rural highways in India, which operate under
mixed traffic.Accident prediction models were built with the selected model specification, to predict the
expected number of single-vehicle and multi-vehicle accidents on two-lane undivided rural highways.To
account for the effect of the explanatory variables which were omitted from the base models, multiplicative
accident modification factors were developed for each of these variables.
Few researches focussed on the study of road and roadside features affecting safety features of a road location.
Naveen et al., (2013) proposed application of Bayesian modelling techniques for road traffic crash analysis on a
sample of Indian National Highways. Poisson-Gamma Hierarchical Bayes and Poisson-Weibull Bayesian
models were applied to the collected crash data. Explanatory variables were geometric characteristics like
median opening and access roads to main highways; traffic characteristics like Average Daily Traffic (ADT);
and roadside developments like Industrial, Commercial, Residential and School. These variables were analyzed
with the dependent variable as crash count per two hundred meter per year. The results of this study show that
Poisson-Gamma hierarchical model best predicts the crashes with higher accuracy compared to other models.
Traffic volume, Access Roads and Median opening emphasizes on increase in the probability of occurrence of
crashes.Roadway design is one of the most significant factors that affect driving behavior and perceived safety.
Ben-Bassat et al., (2011) studied combined effect of roadway design element such as shoulder width, guardrail
and roadway geometry (curvature) by taking both objective driving measures like speed and lane position and
subjective measures like perceived safe driving speed and estimated road safety into account. They found that
shoulder width had a significant effect on actual speed and lane position but guard rail had a significant effect on
perceived safe driving. Zegeer et al., (1991)studied the relationship between lane or shoulder widening and
accident reduction rate. He concluded that 21% reduction in accident can be achieved by widening the lane 4
feet per side.Somchainuek & Taneerananon, (2013) investigated road side safety on Thai NH. The result
showed that speeding vehicles accounted for about 70% of the total roadside crashes and 30% of road side
crashes were due to road side trees.Doohee & Jinsun(2003) established a relationship between accident severity
and roadside features. The result shows that run-off-roadway accidents can be reduced by avoiding cut side
slopes, decreasing the distance from outside shoulder edge to guard rail, decreasing the number of isolated tree
along a roadway section and increasing the distance from outside shoulder edge to luminaire poles. Run-off road
way accident is a complex interaction of road side factors such as presence of guardrails, miscellaneous fixed
objects, sign supports, tree groups and utility poles along the road way.
3.
Methodology
The overall methodology of the work is given in Figure 1. The work initially attempts to find the LOSS of twolane highways using conventional Highway Safety Manual (HSM) method. Each site is designated a LOSS based
on the difference between the observed average crash frequency and the predicted average crash frequency (which is
the accident reduction potential) for the study group. A road location or site can be placed into one of four predefined classification groups. The group LOSS-I indicates a low potential for accident reduction, group LOSS-II
indicates better than expected safety performance, group LOSS-III indicates less than expected safety
performance and group LOSS-IV indicates a high potential for accident reduction. The data required for
determination of LOSS of sites are Crash data by location (recommended period of 3 to 5 Years), calibrated Safety
Performance Function (SPF), over-dispersion parameter and traffic volume.
Then for each LOSS group the road features and roadside features are studied by surveying sites. The common
road features and roadside features observed in each LOSS group are identified and a guideline for
determination of LOSS of a site with road and roadside features is developed. With this guideline the LOSS of
two lane highway location can be determined using road and roadside features only, in absence of systematic
crash history and SPF for the road.
Determination of LOSS
with HSM method
Data
3 Year crash data
Traffic volume (total, Motorized two wheeler, heavy vehicle)
SPF & Accident Modification Factor (AMF)
For two lane highways for Indian conditions
Study of road features & roadside features for each site
Data
Survey of road & roadside features
Study common road & roadside features for
each LOSS group
Develop guideline for determination of LOSS
of a site with road & roadside features
Figure 1: Methodology for development of LOSS guideline
The HSM method of obtaining LOSS of a site is discussed in detail in sub-section 3.1 and the description of
road and roadside variables considered for determining the LOSS guideline is described in sub-section 3.2.
3.1. Determination of LOSS by HSM Method
The LOSS for each site or road location is determined using average observed crashes and expected no of
crashes in a year. The determination of expected crashes requires SPF for two-lane highways for Indian traffic
conditions. The SPFs for Single vehicle crashes and multi vehicle crashes for two lane rural highway with the
mixed traffic conditions are taken for calculating the expected crashes at each sites Dinu R R(2012). SPFs for
single vehicle crashes and multi vehicle crashes are given in the equations 1 & 2 respectively.
𝑁𝑆𝑉 = 𝑒π‘₯𝑝 (βˆ’11: 611 + 0.810𝑋1 + 0.859𝑋2 )
π‘π‘šπ‘‰ = 𝑒π‘₯𝑝 (βˆ’14: 752 + 0.883𝑋1 + 1.398𝑋2 )
(1)
(2)
Where,
Nsv is the expected number of single vehicle accidents per year
Nmvis the expected number of multi vehicle accidents per year
X1 is the length of highway segment in km and
X2 is logarithm of average daily traffic.
TheSPFsare developed considering base condition of traffic heterogeneity with 20% two-wheelers (Vtw) and
30% heavy goods vehicles (Vhv), 2m average distance of off-shoulder hazard from road edge (Doh), 7 driveways
per Km (Ndw) and 6.5m average carriageway width (CW). The conditions of the study stretch differ from the
base conditionsfor which SPFs are developed and thus the predicted crashes are modified using AMFs for the
deviations from base conditions.The AMFs for single vehicle crashes and multi vehicle crashesused for the
study are given in Table 1&2respectively.
Table 1: AMF for single vehicle crashes (Dinu, 2013)
Description
Expression for AMF
Percentage of motorised two-wheelers
exp [-3.404(Vtw– 0.2)]
Percentage of heavy goods vehicles
exp [-4:085 (Vhv – 0.3)]
Average distance to off shoulder hazards
Not significant
Number of driveways per K.M
exp [0.037 (Doh - 7)]
Average width of carriage way
exp [-0.958 (CW - 6.5)]
Table 2: AMF for multi vehicle crash (Dinu, 2013)
Description
Expression for AMF
Percentage of motorised two-wheelers
exp [1.105(Vtw – 0.2)]
Percentage of heavy goods vehicles
exp [0.822 (Vhv– 0.3)]
Average distance to off shoulder hazards
exp [-0.106 (Doh- 2)]
Number of driveways per K.M
exp [0.019 (Ndw- 7)]
Average width of carriage way
exp [-0.443 (CW - 6.5)]
The LOSS of a site can be predicted if the observed average crash frequency (K) lies in different ranges of
predicted crash frequency from SPF and standard deviation. The LOSS limit division isspecified in Highway
Safety Manual and is given in Table 3.
Table 3: Loss limits divisions by HSM(AASHTO, 2013)
LOSS
Condition
I
0 < 𝐾 < (𝑁 βˆ’ 1.5 × (𝜎 ))
II
III
IV
(𝑁 βˆ’ 1.5 × (𝜎 )) ≀ 𝐾 < 𝑁
𝑁 ≀ 𝐾 < (𝑁 + 1.5 × (𝜎 ))
𝐾 β‰₯ (𝑁 + 1.5 × (𝜎 ))
Description
Indicates a low potential for crash reduction
Indicates low to moderate potential for crash reduction
Indicates moderate to high potential for crash reduction
Indicates a high potential for crash reduction
Where;
𝜎=
π‘π‘ƒπ‘Ÿπ‘’π‘‘π‘–π‘π‘‘π‘’π‘‘ + π›Όπ‘π‘ƒπ‘Ÿπ‘’π‘‘π‘–π‘π‘‘π‘’π‘‘
2
 = over-dispersion parameter of SPF function
N = Npredicted = average expected crash frequency
3.2. Road characteristic variables
The road characteristics refer to its geometry and the roadside land use. Road geometryvariables include
elements like carriageway width, shoulder type, cross fall, super elevation, radius of curvature, gradient, off
shoulder hazards, road marking, traffic sign, road side land use etc. The roadside land use data include road side
land use and it refers to number of accesses, number of driveway, type of land use, parking, bus stops etc.
Environment data include weather, light conditions etc.Road characteristic variablesconsidered in this study are
shown in the Table 4.
Table 4: Road Characteristic Elements
Road Geometry
Road Side Land use
Carriageway width
Type of land use
Paved shoulder width
Number of driveways
Unpaved shoulder width
Number of accesses
Horizontal curvature
Gradient
Sight distance
Type of off-shoulder hazard
Distance to off-shoulder hazard
4.
Data
In this work21.3K.mstretch of two lane rural highway of NH98,in Bihar India shown in Figure 2 is studied. The
stretches include Phulwari Sharif to Janipur Chauk, K.m 13/K.m 17, Janipur to Naubatpur Market , K.m
18/K.m 24, Naubatpur Market to Bikram More , K.m25/K.m27 and Bikram More to Dariyapur , K.m28/K.m35.
Figure 2: Study stretch
The stretch is divided into segments of 200m (also referred to as site) and crash history data, traffic volume data
and road characteristic data is collected for each site. The crash data is secondary data and is collected from FIR
reports in the police stations (under IPC No. 279/337/338/304(A)) and is detailed in sub-section 4.1. Traffic
volume and road geometric and roadside features are primary data and collected by survey and is detailed in
sub-sections 4.2 and 4.3 respectively.
4.1Crashhistory
Crash records are mainly maintained by respective police stations. The crash records of the entire study stretch
is maintained in three police stations namely Phulwari Sharif, Janipur and Naubatpur. In this study crash history
of 2012, 2013 and 2014 are collected. The crash related FIRs contains date, time and location of crashes.
4.2Traffic volume
Traffic volume is collected by video-graphic techniqueat twelve points in the study stretch for peak hours. Here
peak hour is considered from 9.30 AM to 10.30 AM in the morning and 4.30 PM to 5.30 PM in the evening.
Data is collected during month of October 2014 for 7 days. Total traffic volume and volume of heavy vehicles
and two wheelers are extracted from the video. ADT is calculated considering that during peak time 15% of
total ADT volume move on rural highway. The traffic volumes of 2013 and 2012 are estimated using an annual
growth factor of 7.5%/
4.3Road Characteristics
The data of road geometry and road side characteristics mentioned earlier is collected from field survey
conducted during the months on November, December 2014 and January 2015. The type of land use is
categorized as agricultural, residential, commercial and hospital. Some stretches have mixed land use. The type
of off shoulder hazard (R1) taken is tree (T), boundary wall (B), Electric pole (E). Presence and number of
roadside Dhaba (R2) in the stretch is also noted. The effect of presence of Factory (R3) is also studied.
5. Analysis& Discussion
The Present study is an attempt to develop basis for accessing level of service of safety of two lane undivided
rural highway with road geometry and road side features. Using crash data and SPFs the sites are categorized
into four LOSS groups. Methodology adopted here is based on literature review, selection of study stretch,
selection of accident data and collection of road geometry data and road side features. It has been noticed that
different LOSS category have different relation with the road geometry and road side features and it has been
shown in Table 5.
Table 5: Relation between LOSS and road geometry and road side features
LOSS
I
Doh(m)
4–5
Ndw
3-6
CW(m)
6.5 –
7.5
SW(m)
1 - 1.5
II
3 –4
6-9
6.5 - 7
III
3.5-4
6-9
7
0.75 –
1.25
1
IV
3 – 3.5
6-9
5.6-6.5
0.75-1
R1
T
B
E
R2
R3
Land use
0
0
Agriculture.
B
E
E
0
0
Residential/Commercial
1
0
Agriculture
T
B
E
1-2
1
Residential/Commercial
Residential/Hospital
In LOSS1 it has been seen that the width of carriageway is above 6.5m, width of the shoulder is above 1m,
distance to off shoulder hazards is above 4m. It has also been experienced that there is no dhaba or factory
beside the roads of the segment which comes under it. No. of accesses vary from 3 to 6 per km. One of the most
remarkable features was that land use beside the road was mostly agricultural land.
In LOSS II, the width of carriage way is also above 6.5. Minimum shoulder width is 0.75m. Distance to off
shoulder hazard is above 3m. There is no dhaba or factory beside the roads of the segment which comes under
it. No. of the accesses is above 6 per km. Land use is both residential and commercial both.
In LOSS III, It has been observed that despite of having carriage way 7m and shoulder width above 1m, there
are two remarkable changes which make it to come under LOSS III. Usually there is a road side Dhabain these
stretches and number of accesses is more than 6.
In LOSS IV, there are mainly two features which bring change in the level of service of safety even others
features tally within the range of LOSS I or LOSS II. They are carriageway width and land use for residential. It
has been seen that in undivided two lane rural highway most of the segment with carriage way less than 6.5m
and residential land use falls under LOSS IV.
6.Conclusions
Based on the present study, the following conclusions may be drawn:
1. Level of service of safety has distinct relation with the road geometry and road side features mainly with the
carriageway width, shoulder width, distance to off shoulder hazards, land use and no. of accesses per K.M.
2. There is no significant role of type of off shoulder hazards in deciding level of service of safety. Type of off
shoulder hazards are generally trees, electric pole and boundary wall.
3. Semi residential area at end and start of the village experiences more accident than the dense area of the
residential zone.
References
AASHTO. (2013). Highway Safety Manual. Manual.
Ben-Bassat, Tamar, & Shinar, D. (2011). Effect of shoulder width, guardrail and roadway geometry on driver
perception and behavior. Accident Analysis & Prevention, 43 (6), 2142-2152.
Dinu, R. R. (2013). Studies on Safety Performance of Two-Lane Rural Highways under Mixed Traffic. Madras:
Department of Civil Engineering Indian Institute of Technology Madras.
Doohee, N. A., & Jinsun, L. E. (2003). Dedicated Lane Strategies for Urban Mobility. Journal of the Eastern
Asia Society for Transportation Studies, 5.
Kononov, J., & Allery, B. K. (2009). Implementation of Level of Service of Safety (LOSS). LADOTD
Transportation Conference. Louisiana.
Kononov, J., & Allery, B. (2003). Level of Service of Safety: Conceptual Blueprint and Analytical Framework.
Transportation Research Record, 1840, 57-66.
Kumar, C. N., Parida, M., & Jain, S. S. (2013). Poisson Family Regression Techniques for Prediction of Crash
Counts Using Bayesian Inference. Procedia - Social and Behavioral Sciences, 104, 982-991.
Pan, F., Lu, J., Xiang, Q., & Zhang, G. (2007). Safety Level of Service at Highway Signalized Intersections.
International Conference on Transportation Engineering, (pp. 1499-1504).
Somchainuek, O., & Taneerananon, P. (2013). nvestigation into roadside safety on Thai National Highways.
Indian Journal of Science and Technology, 6 (1), 3923–3927.
Srinivas, C., Dinu, R. R., & Veeraragavan, A. (2007). Application of Poisson and Negative Binomial
Regression for Modeling Road Accidents under Mixed Traffic Conditions. TRB 86th Annual Meeting.
Washington DC.
Vivian, R. R., & Veeraragavan, A. (2007). Accident Prediction Factors for Rural Highway Segments in
Developing Countries. TRB 86th Annual Meeting. Washington DC.
Zegeer, C., Stewart, R., Reinfurt, D., Council, F., Neuman, F., Hamilton, E., et al. (1991). Cost-effective
geometric improvements for safety upgrading of horizontal curves. TRB.