10 Paper No. 99-0066 TRANSPORTATION RESEARCH RECORD 1657

In addition to traffic capacity, vehicle speeds and queue-discharge rates atwork zones are essential in assessment of work zone traffic delays anduser costs. The traffic flow characteristics of freeway work zones wereanalyzed based on the traffic data collected from Indiana four-lane free-ways. It was found that traffic congestion at work zones was character-ized by sustained low vehicle speeds and fluctuating traffic flow rates.Therefore, work zone capacity was defined as the traffic flow rate justbefore a sharp speed drop, followed by a sustained period of low vehiclespeed and fluctuating traffic flow rate. Study results indicate that the meanqueue-discharge rates in Indiana freeway work zones were lower than thework zone capacities, even though, at times, individual values of queue-discharge rates could be higher than capacities. Therefore, it is not justi-fied to use work zone capacity values, instead of queue-discharge rates, inestimating traffic delays and user costs under congested conditions. Vehi-cle speeds at work zones under uncongested conditions remained stableand close to the work zone speed limit of 88.5 km/h (55 mph), while theydropped 31.6 to 56.1 percent from the normal work zone speeds duringcongestion. The values of work zone capacity, queue-discharge rate, and vehicle speed provide valuable input for predicting traffic congestion,estimating traffic delays, and analyzing user costs at work zones.

It is common knowledge that, each year, the highway constructionseason starts with establishment of work zones on roadways and thatwork zones cause traffic delays. To plan and schedule work zoneoperations efficiently, it is essential to know the traffic capacity val-ues of work zones. A work zone reduces the number of availablelanes for traffic and therefore causes vehicle deceleration and merg-ing. When traffic flow is below the capacity of a work zone, traffic isdelayed primarily by the reduced vehicle speed through the workzone. When traffic flow exceeds the work zone capacity, vehiclequeues form in the work zone and result in additional traffic delays.The 1994 Highway Capacity Manual(1) provides typical capacityvalues of freeway work zones. As Dixon et al. (2) indicated, thesevalues were obtained using the traffic data on the roadways in Texasand may not represent the work zone capacities of other statesbecause of different freeway characteristics and driving behaviors. Aprevious study (3) assumed that, under congested traffic conditions,vehicles travel through a work zone at a flow rate equal to the workzone capacity. This assumption is not accurate because, during con-gestion at work zones, traffic flow rates are mostly lower than thework zone capacities. Therefore, characteristics of work zone trafficflows and speeds during congestion are as essential as work zonecapacity values in the assessment of work zone traffic delays and usercosts. This study was conducted to analyze the traffic flow charac-teristics of freeway work zones based on the traffic data collectedfrom Indiana four-lane freeways. In addition to work zone capacity

values, the patterns of traffic flow and speed at work zones before andduring traffic congestion were analyzed to provide valuable input forestimation of work zone traffic delays and user costs.

DATA COLLECTION

Work zone is defined in the 1994 Highway Capacity Manualas “anarea of highway in which maintenance and construction operationsare taking place that impinge on the number of lanes available tomoving traffic or affect the operational characteristics of trafficflowing through the area.” Two types of work zones in Indiana onfour-lane divided highways are shown in Figures 1 and 2 and aredefined as follows (4):

1. Partial closure (or single-lane closure): one lane in one direc-tion is closed, resulting in little or no disruption to traffic in theopposite direction.

2. Crossover (or two-lane, two-way traffic operations): one road-way is closed, the traffic that normally uses that roadway is crossedover the median, and two-way traffic is maintained on the otherroadway.

Traffic data were collected at select work zones on Interstatehighway sections between October 1995 and April 1997. Trafficcounters with road tubes were used for data collection. Traffic vol-ume, vehicle speed, and classification were recorded at 5-minintervals during high-traffic-volume hours and at 1-h intervals dur-ing low-traffic-volume hours. The vehicle counters were set up toclassify the detected vehicles into three groups: (a) passenger cars,(b) heavy trucks, and (c) buses. In each work zone, traffic counterswere placed before the work zone transition area, within the tran-sition area, and within the activity area. Figure 3 illustrates a typi-cal layout of traffic counters in a work zone. Thus, the recordedtraffic data include free-flow traffic (uninterrupted by work zone),merging traffic, and work zone traffic. Eight work zones on Inter-state highways were selected randomly for traffic data collection.In each of the work zones, traffic data were recorded for 2 to 4 d.The traffic data showed that four of the eight work zones experiencedtraffic congestion during data collection.

WORK ZONE CAPACITY

As shown in Figure 1, the partial-closure work zone disrupts trafficin only one direction, and the crossover work zone affects traffic inboth directions (the median crossover direction and the opposite

Traffic Capacity, Speed, and Queue-Discharge Rate of Indiana’s Four-Lane Freeway Work Zones

YI JIANG

Indiana Department of Transportation, Division of Research, 1205 Mont-gomery Street, P.O. Box 2279, West Lafayette, IN 47906.

Jiang Paper No. 99-0066 11

direction). However, the crossover work zone allows the construc-tion crew to work on two lanes and also provides a safer work areabecause the work area is separated from traffic.

As defined in the 1994 Highway Capacity Manual,the capacity ofa highway facility is “the maximum hourly rate at which persons orvehicles can reasonably be expected to traverse a point or uniformsection of a lane or roadway during a given time period under pre-vailing roadway, traffic, and control conditions.” Previous studiesapplied different methods to identify capacities of freeway workzones. The Texas Transportation Institute (TTI) identified work zonecapacity as the hourly traffic volume under congested traffic condi-tions (5). A Pennsylvania study defined the hourly traffic volumeconverted from the maximum recorded 5-min flow rate as the workzone capacity. A North Carolina study (2) defined work zone capac-ity as “the flow rate at which traffic behavior quickly changes fromuncongested conditions to queue conditions” and used speed-flowcurve to identify the capacity value. The North Carolina definitionseems to be closest to the general definition of capacity given by the1994 Highway Capacity Manual.This is because it used the flow rateunder “prevailing conditions” (before traffic was congested) as thework zone capacity, while other methods used or included congestedtraffic flow rates (queue-discharge rates) in work zone capacity.

It was observed that traffic flows in Indiana freeway work zonesalways changed from uncongested to congested conditions with asharp speed drop. Therefore, work zone capacity is defined in thisstudy as “the traffic flow rate just before a sharp speed drop followedby a sustained period of low vehicle speed and fluctuating trafficflow rate.” To express work zone capacity in terms of passenger carsper hour (pcph), the traffic flow rate was converted to hourly vol-ume, and the adjustment factors from the 1994 Highway CapacityManual were used to convert trucks and buses to passenger car

equivalents. To identify capacity values, traffic flow and speed datapoints were plotted in order of time into one graph. For example,Figure 4 shows such a graph that plotted 1 d of traffic flow and speeddata in a work zone on I-69, where traffic flow values were dividedby 10 for easy comparison. As indicated in the figure, the capacityvalue is identified as the traffic flow rate (1,590 pcph) just before the sharp speed drop, from 87 km/h (54 mph) to 45 km/h (29 mph),and a long period of traffic congestion (low vehicle speed and fluctuating flow rate).

With this method, the traffic capacity values in the four freewaywork zones were identified as presented in Table 1. As shown in thetable, congestion at work zones could start within a work zone aswell as in the transition areas. To compare the capacity values of dif-ferent work zones, an analysis of variance (ANOVA) (6) was con-ducted on the work zone capacity data. The ANOVA tests whetheror not mean capacity values are the same in the four work zones:

where µi are the mean traffic capacity values at work zone i. Sup-pose a Type I error is controlled at α = 0.05, then F(0.95, 3, 8) = 4.07with 3 and 8 as the degrees of freedom associated with the factorlevel and the error term of the given data in Table 1. The decisionrule is thus

As part of the ANOVA test, the Bartlett test (6) of variance homo-geneity was performed. The purpose of the Bartlett test was to deter-mine whether the work zone capacity values had statistically equalvariances as assumed by the ANOVA model. If the test showed thatthe variances were statistically unequal, the data would have to betransformed in some way to improve the variance homogeneitybefore conducting ANOVA. Anderson and McLean (7) proposedthat, if the homogeneity test is accepted at the α = 0.01 level, thenthe data do not need to be transformed for the ANOVA test. TheBartlett test on the capacity values resulted in a p-value of 0.0166,which is greater than α = 0.01. Therefore, the homogeneity test wasaccepted at the α = 0.01 level, and a data transformation was notneeded for the ANOVA test.

Using the data from Table 1 and the Statistix (8) statistical program,the ANOVA test statistic was calculated:

Since F* = 0.81 < 4.07, it is concluded that the mean capacity val-ues of the four work zones are statistically equal. The mean capacityvalues and standard deviations for the four work zones are as follows:

Work Zone Mean Capacity (pcph) Standard Deviation

Zone 1 1,537 242.21Zone 2 1,745 268.69Zone 3 1,612 28.54Zone 4 1,521 5.66

The data in Table 1 show that the work intensities were differentin these work zones. In the partial-closure work zones (Zones 1and 4), construction work was performed in the lane adjacent to the

F* , / , .= = =MSTR/MSE 34 085 41 965 0 81

If conclude H

If conclude Ha

0F

F

* . ,

* . ,

≤

>

4 07

4 07

H

Ha: not all are equal

0: µ µ µ µ

µ

1 2 3 4= = =

i

FIGURE 1 Partial-closure work zone.

FIGURE 2 Crossover work zone.

FIGURE 3 Layout of traffic counters for data collection.

TABLE 1 Work Zone Capacity Data

FIGURE 4 Traffic flow and speed curves for capacity identification.

Jiang Paper No. 99-0066 13

traffic lane. However, in the crossover work zones (Zones 2 and 3),construction work was performed in the areas separated from thetraffic lanes. Three levels of work intensity were observed in the fourwork zones: (a) medium intensity (Zone 1), (b) work not adjacent totraffic (Zones 2 and 3), and (c) high intensity (Zone 4). The Bartletttest of variance homogeneity was also performed on the capacitydata, which were grouped according to the three levels of work inten-sity. The Bartlett test yielded a p-value of 0.0819, which is greaterthan α = 0.01. According to Anderson and McLean (7), the homo-geneity test was accepted at the α = 0.01 level, and a data transfor-mation was not needed for the ANOVA test. An ANOVA was thenconducted to test whether the mean work zone capacities were thesame for different work intensities. The ANOVA yielded: F* = 0.88< F(0.95, 2, 9)= 4.26, which indicates that the mean work zonecapacities are statistically equal for the three levels of work intensity.The mean capacity values and standard deviations are as follows:

Work Intensity Mean Capacity (pcph) Standard Deviation

Medium 1,537 242.21Nonadjacent 1,688 203.50High 1,521 5.66

Although the ANOVA tests indicated that the mean capacity val-ues are statistically equal for different work zone types and workintensities, some differences indeed exist in the individual mean values and confidence intervals. The lower mean value of the partial-closure work zones might be attributed to the influences of the workactivities in the work area adjacent to the traffic lane. Because of thestatistics equality, the mean capacity values could be combined intoone single value based on the principles of statistics. However, toreflect the minor differences in the capacities of the four types of workzones, the individual values (capacity means and confidence intervals)of the four work zone types are presented in Table 2 as the typicalcapacities of the work zones on Indiana four-lane freeways. It shouldalso be pointed out that these capacity values are one-directionalcapacities. Because a crossover work zone affects traffic flows in bothdirections, it might cause greater traffic disruptions and delays than apartial-closure work zone would.

WORK ZONE TRAFFIC FLOW AND SPEED

Of interest are the values of vehicle speed and flow rate at workzones before and during traffic congestion. Under uncongested traf-fic conditions, vehicle speed in a work zone remains relatively sta-

ble with minor fluctuations, and vehicles pass through the work zonesmoothly, without formation of vehicle queues. Figure 5 shows aplot of a consecutive 48-h uncongested traffic flow and speed at awork zone on I-65.

The figure clearly shows that vehicle speed was consistently sta-ble throughout the 48 h as traffic flow changed cyclically from day-time high to nighttime low. The mean speed for the 48 h is 90 km/h(56 mph), with a standard deviation of 2.44. This small standarddeviation indicates that vehicle speeds remained within a closerange from the mean speed of 90 km/h (56 mph) during the 48-htime period.

When traffic is congested in a work zone, vehicle speed remainslow and inconsistent, and traffic flow rate changes irregularly. Fig-ures 6 and 7 are two examples of work zone traffic flow and speedpatterns during congestion. The two figures show that, during con-gestion, the vehicle speeds in both work zones were considerablylower than the work zone speed limit of 88.5 km/h (55 mph). How-ever, the behavior of the traffic flows was quite different in the twowork zones. Traffic flow in the I-65 work zone remained consis-tently below capacity with relatively small fluctuations, while thatin the I-70 work zone had values both above and below capacity,with significant fluctuations. Traffic flow rate in a work zone duringcongestion is actually the rate of the queued vehicles being dis-charged from the work zone. Therefore, it is called the queue-dis-charge rate of the work zone.

The average queue-discharge rate and the average vehicle speedduring congestion are important input for estimating traffic delaysand user costs. Based on the traffic data collected in the Indiana free-way work zones, several statistical values of traffic characteristicswere calculated and are presented in Table 3. As shown in the table,the means of vehicle speeds under uncongested traffic conditionswere about the same for the four types of work zones, with meanspeed values of 90.6 km/h (56.27 mph), 91.6 km/h (56.93 mph),94.2 km/h (58.51 mph), and 92.3 km/h (57.34 mph). Under con-gested traffic conditions, the means of vehicle speeds were 41 km/h(25.45 mph) for crossover (in the opposite direction), 40.6 km/h(25.24 mph) for crossover (in the crossover direction), 50.6 km/h(31.46 mph) for partial closure (with the right lane closed), and62.1 km/h (38.58 mph) for partial closure (with the left laneclosed). Compared with vehicle speeds under uncongested condi-tions, the vehicle speeds under congested conditions had largerstandard deviations, which indicates that vehicle speeds vary moreunder congested traffic conditions. Under congested traffic condi-tions, the mean traffic flow rates (queue-discharge rates) were,respectively, 1,393, 1,587, 1,216, and 1,374 pcph for the four types

TABLE 2 ANOVA Results of Work Zone Capacities (pcph)

FIGURE 5 Uncongested traffic flow and speed (I-65 over SR-46, crossover work zone).

FIGURE 6 Work zone traffic flow and speed during congestion (I-65 N. of SR-32).

FIGURE 7 Work zone traffic flow and speed during congestion (I-70 E. of SR-9).

TABLE 3 Summarized Traffic Flow and Speed Data at Work Zones

of work zones. All of these queue-discharge rates are lower thantheir corresponding work zone capacity values. This indicates that,although traffic flow rate could be occasionally higher than the workzone capacity during congestion, the average flow rate remainedbelow the work zone capacity. For easy reference, the mean valuesof work zone capacity, queue-discharge flow rate, and congestedvehicle speed are listed in one table (Table 4), with the valuesrounded to whole numbers.

As shown in Table 4, the crossover (in the crossover direction) hasthe largest value of mean queue-discharge rate among the four typesof work zones. In addition, as shown in Table 3, the queue-dischargerates (congested traffic flow) for the crossover (in the crossover direc-tion) also have the smallest standard deviation, or the least traffic fluc-tuations. Therefore, the crossover (in the crossover direction) hasrelatively smoother traffic flow under congested traffic conditionsthan the other three types of work zones do. Compared with the twopartial-closure work zones, the two crossover work zones have highercapacities and queue-discharge rates; however, they also have lowermean speeds during congestion. The differences between the valuesof mean capacity and mean queue-discharge rate are 352 for cross-over (opposite direction), 25 for crossover (crossover direction), 321for partial closure (right lane closed), and 147 for partial closure (leftlane closed). The values of these differences correspond to drops intraffic flow rates of 20.2, 1.6, 20.9, and 9.7 percent. These percent-ages indicate that traffic congestion at work zones could result inminor (1.6 percent) as well as considerable (20.9 percent) reductionin traffic flow rates. The drops of mean vehicle speeds caused by con-gestion are 49.9 km/h (31 mph) (55.4 percent), 51.5 km/h (32 mph)(56.1 percent), 45.1 km/h (28 mph) (47.5 percent), and 29 km/h(18 mph) (31.6 percent), respectively, for the four work zones. It isapparent that traffic congestion exerts a more significant impact onvehicle speeds than it does on traffic flow rates.

CONCLUSION

This study indicates that traffic congestion at work zones is char-acterized by sustained low vehicle speeds and fluctuating trafficflow rates. Therefore, work zone capacity is defined in this study asthe traffic flow rate just before a sharp speed drop, followed by a

16 Paper No. 99-0066 TRANSPORTATION RESEARCH RECORD 1657

sustained period of low vehicle speed and fluctuating traffic flowrate. Based on this definition, the capacity values can be identifiedon the graph of traffic flow and speed data in order of time series.In addition to traffic capacities, this paper discussed and providedthe mean queue-discharge rates and vehicle speeds for both uncon-gested and congested traffic conditions. These values can be usedas a basis for predicting traffic congestion, estimating traffic delays,and analyzing user costs at work zones (9). The study results indi-cate that the mean queue-discharge rates in Indiana freeway workzones were lower than the work zone capacities, even though, attimes, individual queue-discharge rates could be higher than capac-ities. Therefore, the use of work zone capacity values, instead ofqueue-discharge rates, is not justified in estimating traffic delaysand user costs under congested conditions. Vehicle speeds at workzones under uncongested conditions remained stable and close tothe given work zone speed limit of 88.5 km/h (55 mph). The dropin traffic flow rates caused by traffic congestion ranged from 1.6 to20.9 percent, while the drop in vehicle speeds ranged from 31.6 to56.1 percent.

This study focused on the work zones on Indiana’s four-lane free-ways. Work zones on Indiana freeways with more than four travellanes were not analyzed because of a lack of traffic data. The inclu-sion of work zones on freeways with more than four lanes in a futurestudy is recommended.

ACKNOWLEDGMENTS

This study was supported by the Indiana Department of Trans-portation and the Federal Highway Administration through the StatePlanning and Research Program.

REFERENCES

1. Special Report 209: Highway Capacity Manual.3rd ed. TRB, NationalResearch Council, Washington, D.C., 1994.

2. Dixon, K. K., J. E. Hummer, and A. R. Lorscheider. Capacity for NorthCarolina Freeway Work Zones. In Transportation Research Record

TABLE 4 Mean Values of Work Zone Capacities, Queue-Discharge Rates, and Vehicle Speeds

1529, TRB, National Research Council, Washington, D.C., 1996, pp. 27–34.

3. Memmott, J. L., and C. L. Dudek. QUEWZ–85, A Model to Calculate theRoad User Costs at Work Zone.Texas Transportation Institute, TexasA&M University, Sept. 1982.

4. Construction Costs and Safety Impacts of Work Zone Traffic ControlStrategies, Volume II: Informational Guide.FHWA-RD-89-210. FederalHighway Administration, 1989.

5. Dudek, C. L., and S. H. Richards. Traffic Capacity Through Urban Free-way Work Zones in Texas. In Transportation Research Record 869,TRB, National Research Council, Washington, D.C., 1982, pp. 14–18.

Jiang Paper No. 99-0066 17

6. Neter, J. N., W. Wasserman, and M. H. Kutner. Applied Linear StatisticalModels.Richard D. Irwin, Inc., 1985.

7. Anderson, V. L., and R. A. McLean. Design of Experiments—A RealisticApproach.Marcel Dekker, Inc., 1974.

8. Statistix for Windows User’s Manual.Analytical Software, 1996.9. Jiang, Y. Traffic Capacities and Methods to Estimate Traffic Delays and

User Costs at Indiana Freeway Work Zones.Report FHWA/INDOT/SPR-2121. Indiana Department of Transportation, Dec. 1998.

Publication of this paper sponsored by Committee on Traffic Safety inMaintenance and Construction Operations.