Transportation Planning and Technology

ISSN: 0308-1060 (Print) 1029-0354 (Online) Journal homepage: www.tandfonline.com/journals/gtpt20

Analysis of Work Zone Traffic Behavior for

Planning Applications

Sujith Racha, Mashrur Chowdhury, Wayne Sarasua & Yongchang Ma

To cite this article: Sujith Racha, Mashrur Chowdhury, Wayne Sarasua & Yongchang Ma (2008)
Analysis of Work Zone Traffic Behavior for Planning Applications, Transportation Planning and
Technology, 31:2, 183-199, DOI: 10.1080/03081060801948175

To link to this article: https://doi.org/10.1080/03081060801948175

Published online: 13 Mar 2008.

Submit your article to this journal

Article views: 244

View related articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=gtpt20
Transportation Planning and Technology, April 2008
Vol. 31, No. 2, pp. 183
199

ARTICLE

Analysis of Work Zone Traffic

Behavior for Planning Applications

SUJITH RACHA, MASHRUR CHOWDHURY,

WAYNE SARASUA & YONGCHANG MA
Department of Civil Engineering, Clemson University, Clemson, SC, USA

(Received 31 May 2006; Revised 10 October 2007; In final form 17 January 2008)

ABSTRACT Understanding work zone traffic behavior is important for the

planning and operation of work zones. The objective of this paper is to develop
a mathematical model of work zone traffic flow elements by analyzing the
relationships between speed, flow, and density that can be used to estimate the
capacity of work zones. Traffic flow data were collected from 22 work zone sites
on South Carolina interstate highways. The scatter plots of the collected data
demonstrate that the relationship between speed and density does not follow
Greenshields’ linear model. A non-linear hyperbolic model was developed to
describe the relationship between speed and density. Using this model the
capacity of a work zone was estimated to be 1550 passenger cars per hour for 2-
lane to 1-lane closures. Adjustments to this capacity value to consider other types
of vehicle as well as the work zone intensity are provided. Highway agencies can
use this estimated capacity along with anticipated traffic demand to schedule
work zone operations to avoid long periods of over-saturation.
The tapered approach to work zone lane closures used by South Carolina is
similar to methods used in work zones throughout the world. The authors believe
that the methodology described in this paper for modeling work zone traffic as
well as estimating work zone capacity is transferable to other countries. The
conversion of actual volumes to passenger car equivalents may have to be
modified due to the significant differences in traffic makeup between the United
States and other countries.
KEY WORDS: Base capacity; non-linear model; speed
flow relationship;
speed
density relationship; work zone

Correspondence Address: Mashrur Chowdhury, Department of Civil Engineering, Clemson

University, 216 Lowry Hall, Clemson, SC 29634, USA. Email: mac@clemson.edu

ISSN 0308-1060 print: ISSN 1029-0354 online # 2008 Taylor & Francis
DOI: 10.1080/03081060801948175
184 S. Racha et al.

Introduction
Construction and maintenance activities take place in work zones along
highways, reducing capacity and resulting in traffic congestion and
delays. This is an international problem that transportation agencies
everywhere responsible for the construction and maintenance of
highways must consider in the planning of work zone operations. In
May, 2003, the South Carolina Department of Transportation
(SCDOT) under a funding agreement with the United States Federal
Highway Administration (FHWA) initiated a research study to
quantitatively examine the development and implementation of a
methodology for use in determining an updated lane closure policy
within work zones along South Carolina’s interstate highway system.
Specific emphasis of the study would focus on determination of the
number of vehicles per hour per lane (vphpl) that can pass through
open lanes in short-term interstate work zone when one or more lanes
have been closed with minimum or acceptable levels of delay. This
paper presents research results based on this project.
The literature on traffic behavior on highways is widespread. The
model developed by Greenshields in 1934 has served as a milestone in
describing the fundamental relationships between flow, density and
speed on highways (Greenshields, 1934). This single-regime model was
based on observing speed
density measurements obtained from an
aerial photographic study (May, 1990). Other researchers, such as
Greenberg (1959), developed multi-regime models based on non-linear
relationships between speed and density. Multi-regime models provide
a considerable improvement over the single-regime one proposed by
Greenshields in that they introduce the idea of generalizing traffic
stream models into families by considering different traffic parameters.
The multi-regime models attempt to more closely replicate observed/
measured relationships between major traffic flow parameters; how-
ever, none of these theoretical models perfectly estimate speed, density
and flow as observed under field conditions. In Greenberg’s model, as
traffic operations move from congested to uncongested situations, a
rapid reduction in the maximum flow occurs. Similarly, when queuing
conditions occur, there is a rapid decline in capacity due to the
turbulence caused in the traffic flow (Hall & Agyeman-Duah, 1991).
Most of the research of non-linear models has been conducted on
traffic flow patterns on freeways under normal daily conditions.
However, not many studies have involved traffic flow in work zones.
Investigations in this area are important because non-linear models
reflect real-world conditions more accurately in many situations than
linear models.
 Analysis of Work Zone Traffic Behavior 185

In the study of traffic flow, as in other areas, new analytical

techniques frequently produce findings that are not easily explained
using conventional theories. The latest investigations have introduced
the concept of three-dimensional models of traffic stream character-
istics, allowing for the three variables of speed, density, and flow to be
studied in three-dimensional space, rather than in two-dimensional
space. This method affords the possibility for identifying whether the
relationships follow two-dimensional traditional traffic flow theory
(Gilchrist & Hall, 1989).
Capacity as defined in the US Transportation Research Board’s
Highway Capacity Manual (HCM) is the ‘maximum rate of flow that
can be accommodated on a given traffic facility under prevailing
conditions’ (Transportation Research Board, 2000). This means that a
stated capacity for a roadway is the rate of flow that can be continually
attained during peak periods for which sufficient demand exists. This
definition does not fully explain how to estimate capacity and what
factors influence capacity under real-world conditions. Many studies
have been conducted in the past to estimate the capacity and the factors
influencing it and there have been several methods used to determine
traffic capacity in a work zone. The earliest work in this area was the
study of hourly volume under congested traffic conditions by the Texas
Transportation Institute (Dudek & Richards, 1982). Research in North
Carolina used the flow rate at which traffic behavior quickly changes
from uncongested to queued conditions to estimate highway capacity in
work zones (Dixon & Hummer, 1995). More recently, a study in
Indiana used the traffic flow rate just before a sharp speed drop
followed by a sustained period of low vehicle speed and fluctuating
traffic flow rate (Jiang, 1999). Researchers in Iowa estimated capacity
using the mean queue discharge flow rate from the bottleneck at the
end of the transition area (Maze et al., 2000).
One earlier study conducted by Chin and May (1991) examined the
speed
flow relationship at the Caldecott Tunnel, the freeway section of
California State Highway 24. Using these results, the study’s research-
ers proposed a set of speed
flow curves that differed from those in the
HCM. This research was conducted using only uncongested flows,
which are the upper segments of the speed
flow curves, and found that
work zone scenarios differ from these relationships in many aspects.
The mean passenger car speed was used as a speed variable instead of
the average travel speed. The capacity under ideal conditions,
independent of the design speed of the highway, was found to be
2200 passenger cars per hour per lane (pcphpl). Lane width, lateral
clearance, median type, and the frequency of access points did not
affect the capacity; rather it affected the operating speed of the
highway.
186 S. Racha et al.

Research conducted by Krammes and Lopez (1992) estimated the

capacity values for short-term freeway work zone lane closures using
the data collected at 33 work zone sites in Texas between 1987 and
1991. Data were collected for five different lane closure configurations
including (3 to 1), (2 to 1), (4 to 2), (4 to 3), and (5 to 3), noting that (3
to 1) represents that out of three lanes in one direction, one lane is
closed. Data collection consisted of more than 45 hours of capacity
counts at 33 different freeway work zones with short-term lane
closures. The counts were taken as vehicles entered the work zone
area through the transition area to minimize variability caused due by
differing numbers of lanes between sites. The average capacities for
five-lane closure configurations for the data collected were plotted to
find out the overall average capacity combining all types of lane closure
configurations. The research recommended that the base capacity value
represents conditions in which the impacts of ramps were negligible,
and the effect of ramps needs to be treated separately. A base capacity
value of 1600 pcphpl was recommended for all short-term freeway lane
closure configurations, which is compared to the standard accepted
capacities of 2200 pcphpl for freeways and 1900 passenger cars per
hour per green per lane (pcphgpl) for signalized intersections. These
values represent queue discharge rate for freeway facilities, and
saturation flow rate for signalized intersections. Several adjustments
were made to the base capacity value including adjustments for
intensity of work activity, effect of heavy vehicles, and the presence
of ramps when applying to specific work zone locations.
Unlike previous studies, Maze et al. (2000) evaluated the capacity of
lane closures and driver behavior within work zones on the rural
interstates of Iowa, focusing on the rate at which queues develop and
dissipate. This research determined that the approximate capacity of
the Iowa rural work zone lane closures varied from 1400 to 1600
passenger car equivalents (PCEs) per hour per lane. Similarly, Jiang
(1999) focused on the delays caused in work zones on freeway facilities
in Indiana. His study measured delays caused by the deceleration of
vehicles when approaching work zone areas, low speeds through the
work zone area, the time required for vehicles to return to freeway
speed and the queues formed at the work zones. Several delay
equations, which differentiate when the approaching traffic is below
and above the work zone capacity for all the above kinds of situations,
were developed for these studies.
Kim et al. (2001) developed a new method for estimating the
capacity of work zones. Their study investigated various factors
influencing work zones, leading to the creation of a multiple regression
model for estimating capacity. This model was a function of several
independent factors such as number of lanes closed, the proportion of
 Analysis of Work Zone Traffic Behavior 187

heavy vehicles, and the grade and intensity of the work zones. When
compared side by side, this model performed better than several other
capacity models in the HCM.
Adeli and Jiang (2003) proposed an innovative method using neuro-
fuzzy logic to estimate the capacity for freeway work zone. As many as
17 factors that possibly affect the work zone capacity was considered in
their case study. The authors applied back propagation neuro-network
to search for associated parameters used in fuzzy logic, which had a
Gaussian-shaped membership function. The case study indicated that
proposed method performed accurately, especially when only partial
parameters are available. They also declared the advantage of their
method in terms of incorporating large number of factors and no
requirement of prior knowledge about these factors.
Chitturi and Benekohal (2004, 2005) used selected software pro-
grams and analyzed the effect of lane width on work zone traffic flow.
Quewz, FRESIM and QuickZone models were used to analyze
capacity, queuing and delay in construction work zones and compared
results to field collected data. The findings indicated that none of the
programs produced results that were reasonably close to observed field
data.
As this review suggests, several factors play a role in speed
flow
relationships and in the capacity of work zones. While there are
numerous methods for modeling traffic flow and the capacity of work
zones, there is not a consensus on which method provides the best
estimate of actual traffic flow conditions. This research project focused
on developing models predicting relationships between speed, flow, and
density based on non-linear approaches in an effort to more accurately
model traffic and estimate capacity of work zones. Specifically, the
objectives of the research on which this paper is based were:
. to understand work zone traffic behavior using non-linear ap-
proaches to modeling the speed
flow
density relationships; and
. to develop capacity thresholds for work zone lane configurations,
where out of two lanes one lane is closed (a 2 to 1 lane closure).

Methodology and Data Collection

This paper is based on data collected at 22 work zones, which extended
over a period of approximately one year. These work zone projects
located throughout the state of South Carolina concentrated on
four-lane highways (two lanes in each direction) involving short-term
lane closures (B24 hours). Volume data were collected using a
camera combined with automated digital image processing or manual
188 S. Racha et al.

tabulation techniques. Because many of the projects were at night,

much of the video data had to be processed manually to determine
volumes and vehicle classification. The truck percentage varied
between 3 and 40% in the work zones. A database created in Microsoft
Excel included information on different traffic stream parameters based
on lane configurations and queues. Several scatter plots were generated
for speed
density, speed
flow and flow
density to understand the
fundamental relationships of traffic stream characteristics in work
zones. Statistical Analysis Software (SAS), which is a comprehensive
statistical and graphical package including modules for several types of
specialized analysis, was used to generate graphs of these fundamental
relationships (SAS, 2004).

Data Analysis
Data analysis included data preparation, developing relationship
between traffic flow parameters and capacity estimations.

Data Preparation
As the results from the project demonstrated, truck volume has a
significant effect on the traffic flow in the work zones. When
approaching these areas, trucks often occupy all available lanes,
blocking passenger cars and causing delays. Therefore, the impact of
trucks is a significant consideration in minimizing the effect of work
zones on traffic flow. Each truck is mathematically converted into
passenger cars with a conversion factor called PCE. By measuring the
distance from the rear end of a leading vehicle to the rear of a trailing
one, referred to as the headway, these PCEs were determined for trucks
and recreational vehicles. PCEs reflect the number of passenger cars
that will occupy the space of these larger vehicles in a specific traffic
stream. Comparing the headways for passenger cars and trucks, the
PCE value was determined. These PCEs were categorized based on
speeds of the vehicles traveling in the work zones. The following results
generated from earlier studies (Narapsetty, 2004) were used in
performing the analysis for this research, as shown in Table 1.
Narapsetty found that PCE values for trucks are a function of speed.
This traffic phenomenon is supported by the findings of Chitturi and
Benekohal (2004, 2005). Chitturi and Benekohal suggest that predict-
ing speed reductions should be a function of vehicle size and not be a
constant value for all vehicle classifications.
After this conversion, consecutive and discrete flows were calculated.
The consecutive flow for one-hour intervals was achieved by adding 12
consecutive five-minute flows, with the speed corresponding to this
 Analysis of Work Zone Traffic Behavior 189

Table 1. Average PCE values for trucks used for the analysis (Narapsetty, 2004)
Speed (mph) PCE for trucks

0
15 2.47
15
30 2.22
30
60 1.90

flow value being the average of the 12 five-minute speeds. The

calculated flow thus represented an actual hourly volume. The discrete
flow for a one-hour interval was calculated by projecting the five-
minute flow value over an hour, with the speeds being those calculated
for the five-minute intervals. This procedure for calculating consecutive
and discrete flows was carried out for 2 to 1 lane closures. Using the
basic definition of density, i.e. density equals flow divided by speed,
consecutive and discrete densities for one-hour intervals for all the
projects, including different lane configurations, were calculated. Once
the data for all parameters, i.e. speed, density, and flow, were
determined, scatter plots were generated considering two variables at
a time for both consecutive and discrete situations to analyze their
relationships.
Relationships between Traffic Flow Parameters
Flow
density relationships. The scatter plot of flow versus density for all
2 to 1 lane projects is shown in Figure 1, with the values in this graph
being the discrete values generated by projecting five-minute periods
over an hour for all of these projects. The flow here is in terms of
pcphpl, considering heavy vehicles. The overall relationship between
flow and density failed to follow the Greenshields’ parabolic path. The
maximum flow values were found to be approximately around 1800
pcphpl based on a visual interpretation of the graph.
Figure 2 shows the plot of flow versus density for 2 to 1 lane projects
using consecutive values generated by adding 12 successive five-minute
periods, which reflects actual hourly flow and average density. This
data follows the trends as described in Figure 1, with a slight variation
in the maximum flows observed. The flows remained steady above the
free-flow speeds and visually suggested the highest values to be
approximately 1700 pcphpl. There were fewer data points in this
graph than the number of discrete values in Figure 1, as one point
represents 12 records in the original field data.
Speed
flow relationships. To estimate the capacity, several graphs of the
relationship between speed and flow were plotted. Figure 3, which
represent speed versus flow for all 2 to 1 lane closures, was developed
using discrete five-minute flow values projected to hourly flows by
190 S. Racha et al.

2500

2000
Flow (pcph)

1500

1000

500

0
0 50 100 150 200 250 300
Density (pcpmpl)

Figure 1. Graph of flow versus density (discrete five-minute flows)

multiplying each by 12. Even though the data is significantly scattered,

it appears to follow Greenberg’s non-linear flow pattern, in that the
upper section of the graph is nearly stable under capacity, while the
lower section exhibits a rapid drop in speeds. This graph identifies
the capacity to be approximately 1700 pcphpl.
Similarly, Figure 4 shows the same trend for speed versus consecutive
flow for all 2 to 1 lane closures. Here the flow is based on actual hourly
volume calculated by adding 12 consecutive five-minute flow values
and using weighted-average speeds for these periods.
Two-dimensional model for speed and flow. To model speed versus flow,
this research began with a two-dimensional model using SAS, which
was developed directly from field data. SAS was successful in

2000
1800
1600
1400
Flow (pcphpl)

1200
1000
800
600
400
200
0
0 50 100 150 200 250 300
Density (pcpmpl)

Figure 2. Graph of flow versus density (12 consecutive five-minute periods)

 Analysis of Work Zone Traffic Behavior 191

70

60

50
Speed (mph)

40

30

20

10

0
0 500 1000 1500 2000 2500
Flow (pcphpl)

Figure 3. Graph of speed versus flow (discrete five-minute flows)

generating the model with necessary statistical calculations but failed in

providing statistically significant relationships. Figure 5 shows the SAS-
generated model for speed versus discrete flow for all 2 to 1 lane
projects. It was noted that the R2 value was very low because the data
was widely spread throughout the graph. The data failed to follow
Greenshields’ parabolic path because of the few data points between
the speed range of 30
45 mph and below 20 mph.
Three-dimensional model for speed, flow and density. The authors
attempted to find the relationship between traffic stream parameters
in three-dimensional space with speed, flow and density representing the
three different axes. Figure 6 shows the graph for speed, flow and density
in three dimensions using the discrete data collected for all 2 to 1

70

60

50
Speed (mph)

40

30

20

10

0
0 500 1000 1500 2000 2500
Flow (pcphpl)

Figure 4. Graph of speed versus flow (12 consecutive five-minute periods)

192 S. Racha et al.

Figure 5. Two-dimensional model of speed versus flow (discrete five-minute flows)

projects. However, the three-dimensional model failed to give an exact
profile of the field conditions because of the lack of data points with high
speed and high density, a situation that cannot occur in the real world.

Figure 6. Three-dimensional graph of speed, flow and density

(discrete five-minute flows)
 Analysis of Work Zone Traffic Behavior 193

Speed
density graphs. The graph of speed versus discrete density for

all 2 to 1 lane closures is shown in Figure 7. The field data plotted
on the graph clearly indicates that it is not following the linear
trend that Greenshields’ relationship defines for speed and density.
Alhough the increase in density did result in a decrease in speed,
the non-linear pattern relating speed and flow did not follow a straight
line.
The graph in Figure 8 shows a similar non-linear pattern for the data
plotted between speed and consecutive density. This graph is based on
actual hourly density rather than on five-minute density projected over
an hour, and the speed is the weighted average of these 12 periods. No
data points were found in the region for speeds between 30 and 35 mph
because none of the average speeds fall in that range.
Modeling speed and density. This research identified the non-linear
pattern between speed and density, modeling it using a non-linear
analysis procedure in SAS. Figure 9 presents the statistically significant
model between speed and discrete density for all 2 to 1 lane projects.
This model appeared to follow the shape of a hyperbola. Figure 9 also
shows the non-linear equation developed between speed and discrete
density. The non-linear model showed a statistically significant
improvement over a Greenshields’ linear model for the speed and
density data.
SAS was successful in generating a model as the best fit for the
field data. The partial output of the descriptive statistics provided in
Figure 10 indicates that the model generated is significant, as the
P-value was less than 0.05. The percentage error was determined using

70

60

50
Speed (mph)

40

30

20

10

0
0 50 100 150 200 250 300
Density (pcpmpl)

Figure 7. Graph of speed versus density (discrete five-minute flows)

194 S. Racha et al.

70

60

50
Speed (mph)

40

30

20

10

0
0 50 100 150 200 250 300
Density (pcpmpl)

Figure 8. Graph of speed versus density (12 consecutive five-minute periods)

the proportion of the residual to the corrected total as shown in Figure

11; the results explained more than 99% of the total data with a
marginal error of 0.2%.
Similarly, Figure 12 presents the model developed for 2 to 1 lane
closures using consecutive densities. The generated model reflected
hyperbolic structure. The model was observed to follow similar non-
linear trends to those in Figure 9. Since non-linear models did not
provide R2 values, the sum of the squares of the residual and the
P-value were used to determine the accuracy of the model.

Figure 9. Model of speed versus density (discrete five-minute flows)

 Analysis of Work Zone Traffic Behavior 195

Figure 10. Partial SAS output for speed
density model (discrete five-minute flows)

Capacity Estimation for 2 to 1 Lane Closure

The base capacity for 2 to 1 work zone lane closures was calculated
using the equations generated by the speed
density models. From the
general idea that flow equals velocity multiplied by density for a section
of freeway, speed was substituted for flow and density in all the
equations to generate capacities. The calculations for discrete conditions
are shown in Figure 13. It was assumed that at capacity, the flow
remains constant for a very short interval, which generates the optimum
density. The base capacity was then generated by calculating the flow at
optimum density.
Therefore, a value of 1550 pcphpl (rounding to the nearest 25) was
applied for estimating the base capacity of a freeway work zone for the 2
to 1 lane closure configuration, based on discrete conditions that project
the five-minute volumes over one hour. This estimate is similar to
capacity estimates in many earlier studies on work zone capacity, such as
Krammes and Lopez (1992), Jiang (1999), and Maze et al. (2000).
As discussed earlier, an adjustment is essential to the base PCE
capacity value in order to consider trucks. HCM 2000 provides
a procedure to determine the vehicular volume as a function of
PCEs. The formulas given below were applied to the calculated base
capacity, adjusting for heavy vehicles resulting in the first estimate for
capacity C?.
196 S. Racha et al.

Sum of squares of residuals

Percentage Error =
Corrected Total

10.4661
Percentage Error = = 0.2%
5286.4

Figure 11. Percentage error calculation for speed
density model

(discrete five-minute flows)

1
fHV 
1  [PT (ET  1)  PRV (ERV  1)]
where fHVheavy vehicle adjustment factor, PTproportion of trucks,
ET passenger car equivalents for trucks and buses, and ERV 
passenger car equivalents for recreational vehicles and cars with
trailers.
CB  1550fHV
where CB adjusted capacity for heavy vehicles with one lane open
(veh/h/lane).
It was recognized that the intensity of the work zone will also have
an effect on capacity. HCM 2000 suggests a 10% adjustment,
depending on whether or not the work zone activity is more or less

Figure 12. Model of speed versus density (12 consecutive five-minute periods)
 Analysis of Work Zone Traffic Behavior 197

  5.2369 
s = Ln −1 ( −0.8758 ) +  
 1 + (0.00456 * k )  
  5.2369 
Ln ( s ) = ( −0.8758 ) +  
 1 + (0.00456 * k )  
q
we know that s = ; substituting in above equation
k
q   5.2369 
Ln ( ) = ( −0.8758 ) +   →I
k   1 + ( 0 .00456 * k ) 
Differenci ating on both sides,
dq
Assuming at capacity, there is no change in flow, so →0
dk
 dq 
k× − q
1 dk  (5.2369 × 0.00456 ) 
  = − 2 
q k2   (1 + (0.00456 × k ))

k  

⇒ k 2 − 709 .85 k + 48091 .72 = 0
⇒ k = 75 .86 ≅ 76 pcpmpl

Substituti ng k in equation I to get q which is base capacity,

⇒ q = 1547 pcphpl

Figure 13. Base capacity calculations for discrete conditions for 2 to 1 projects

intense than normal. The final estimation for capacity is given below by
CWZ, which includes adjustments for intensity of work zone, heavy
vehicle factor and the number of lanes open through the work zone
area.
CWZ  (CB I)fHV N
where CWZ estimated capacity of a short-term work zone (veh/h) and
I adjustment factor for type, intensity, length and location of the
work activity.

Conclusions
This research collected five-minute traffic flow and velocity data in 22
work zone sites, which were maintained and operated by the South
Carolina Department of Transportation (SCDOT). The visual inspec-
tion of the scatter plot of the relationship for the traffic stream
parameters for 2 to 1 lane closure indicated that the relationships for
speed
density were non-linear and that the relationship between speed
and flow, and flow and density were not parabolic as indicated by
Greenshields. Therefore, non-linear models of speed and density for
198 S. Racha et al.

both discrete five-minute data projected over an hour and continuous

hourly data were developed using the Statistical Analysis Software
(SAS). The speed
density model was developed for 2 to 1 lane
configuration using SAS and took the form of a hyperbolic curve.
Descriptive statistics indicated that there was very little deviation
between the actual data and modeled estimates. It was not apparent
from the literature that previous researchers have attempted to model
traffic on freeways as a hyperbolic curve. While the data for this
research is based only on work zones, the results suggest that it would
be prudent to apply this approach to non-work zone highway traffic as
well.
Based on the developed model for speed and density, the work zone
capacity was determined to be 1550 pcphpl for 2 to 1 lane closure
configurations. This finding is similar to values found in many previous
studies on short-term work zone closures. Highway agencies can use this
estimated capacity along with anticipated traffic demand to schedule
work zone operations to avoid long periods of over-saturation.
The tapered approach to work zone lane closures used by SCDOT is
similar to methods used in work zones throughout the world. The
authors are confident that the methodology described in this paper for
modeling work zone traffic as well as estimating work zone capacity is
transferable to other countries. The conversion of actual volumes to
PCEs may have to be modified due to the significant differences in
traffic makeup between South Carolina and other countries.
The researchers postulated that factors such as terrain, work zone
activity, and weather might influence work zone capacity. It was
difficult, however, to examine critically the effects of these factors. The
work zones studied did not reveal major differences in terrain type, nor
were there sustained grades. There was insufficient data to conclude
that work zone activity, intensity, and length do not affect work zone
capacity. Similarly, no weather-related effects could be measured, since
short-term work is generally postponed during adverse weather
conditions. Additional data in a future research effort should focus
on a detailed examination of all of these factors as well as different
work zone lane closure configurations, such as 3 to 2 lanes and 3 to 1
lane closures.

References
Adeli, H. & Jiang, X. (2003) Neuro-fuzzy logic model for freeway work zone capacity estimation,
Journal of Transportation Engineering, 129(5), pp. 484
493.
Chin, H. C. & May, A. D. (1991) Examination of the Speed
Flow Relationship at the Caldecott
Tunnel, Transportation Research Record 1320, Washington, DC: National Research
Council.
 Analysis of Work Zone Traffic Behavior 199
Chitturi, M. & Benekohal, R. (2004) Comparison of Quewz, Fresim and QuickZone with field
data for work zones, in: Proceedings of the Transportation Research Board 83rd Annual
Meeting, January, Washington, DC.
Chitturi, M. & Benekohal, R. (2005) Lane width effect on speeds of cars and heavy vehicles in
work zones, in: Proceedings of the Transportation Research Board 84th Annual Meeting,
January, Washington, DC.
Dixon, K. K. & Hummer, J. E. (1995) Capacity for North Carolina Freeway Work Zones, Report
No. 23241-94-8 prepared for the North Carolina Department of Transportation by the
Center for Transportation Engineering Studies, North Carolina State University.
Dudek, C. L. & Richards, S. H. (1982) Traffic Capacity through Work Zones on Urban Freeways,
Transportation Research Record 869, Washington, DC: National Research Council.
Greenberg, H. (1959) An analysis of traffic flow, Operations Research, 7(1), pp. 79
85.
Greenshields, B. D. (1934) A study of traffic capacity, in: Highway Research Board Proceedings,
Vol. 14, Washington, DC, pp. 448
477.
Gilchrist, R. S. & Hall, F. L. (1989) Three-Dimensional Relationships among Traffic Flow Theory
Variables, Transportation Research Record 1225, Washington, DC: National Research
Council.
Hall, F. L. & Agyeman-Duah, K. (1991) Freeway Capacity Drop and Definition of Capacity,
Transportation Research Board 1320, Washington, DC: National Research Council.
Jiang, Y. (1999) Traffic capacity, speed and queue-discharge rate of Indiana’s four-lane freeway
work zones. Presented at the Transportation Research Board Annual Meeting, January,
Washington, DC.
Kim, T. Lovell D. J. & Paracha J. (2001) A New Methodology to Estimate Capacity for Freeway
Work Zones, Transportation Research Record 1320, Washington, DC: National Research
Council.
Krammes, R. A. & Lopez, G. O. (1992) Updated Capacity Values for Short-Term Freeway Work
Zone Lane Closures, Transportation Research Record 1442, Washington, DC: National
Research Council.
May, A. D. (1990) Traffic Flow Fundamentals (Berkeley, CA: University of California/Prentice-
Hall).
Maze, T. H. Schrock, S. D. & Kamyab, A. (2000) Capacity of freeway work zone lane closures, in:
Mid-Continent Transportation Symposium 2000 Proceedings, Ames, IA: Center for
Transportation Research and Education, Iowa State University.
Narapsetty, V. K. (2004) Passenger car equivalents for short-term interstate work zone lane
closures, Master’s thesis, Clemson University, December.
SAS (2004) Base SAS 9.1 Procedures Guide, Vol. 2 (Cary, NC: SAS Publishing).
Transportation Research Board (2000) Highway Capacity Manual
2000 (Washington, DC:
Transportation Research Board, National Research Council).