Incorporating Travel Time Reliability into the Highway Capacity Manual (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

248 Introduction There is little research documented in the literature on the effect of work zone presence on urban street operation. Most of the research on the effect of work zone presence on operation has been conducted for freeways. Chapter 10 of the 2010 Highway Capacity Manual (HCM2010) (Transpor- tation Research Board of the National Academies 2010) pro- vides a synthesis of this research. It indicates that work zone presence tends to reduce the capacity of the freeway lanes that remain open during the work zone. A similar effect is likely to be found on urban streets. An examination of the nationwide impact of work zones on capacity and delay was conducted by Chin et al. (2004). They used data from Rand McNally Construction Informa- tion, the Federal Highway Administration’s Fiscal Manage- ment Information System, and the Highway Performance Monitoring System to obtain work zone location data and highway capacity data. They modeled the work zone effect on freeways by using capacity adjustment factors documented in the HCM. This same approach was extended to the modeling of work zones on urban streets. The results of the analysis by Chin et al. (2004) are shown in Table J.1. They estimate that work zones on principal arte- rials cause about 10 million vehicle hours of delay each year. Freeways are likely to experience more than seven times this amount of delay. The objective of this appendix is to document the research conducted to quantify the effect of work zone presence on signalized intersection operation. The approach taken in this research is to quantify this effect on intersection saturation flow rate. Data were collected at several intersections for this purpose. This appendix consists of three main sections that follow this introductory section. The next section summarizes the findings from a review of the literature on the topic of urban street work zones. The third section describes the site selection and data collected for the purpose of quantifying the effect of work zone presence on saturation flow rate. The fourth section describes the findings from an analysis of the field data and the recommended saturation flow rate adjustment factors. Literature Review A review of the literature on the topic of urban street work zones focused on work zone factors affecting intersection capacity. However, most of the work zone–related publica- tions found in the literature address freeway operations and safety. In some instances of this review, reference is made to this freeway research when the findings may also be applica- ble to urban streets. Work Zone Characteristics Urban street work zones have several characteristics that dif- ferentiate them from highway or freeway work zones. These characteristics are summarized in Table J.2. The focus of this summary is the characteristics that are likely to have a negative influence on urban street traffic operation. In most instances, the influence is likely to be more adverse for the urban street than for the freeway or highway. A typical intersection work zone is shown in Figure J.1. The work area is shown to be in the lower-left corner of the inter- section conflict area. Channelizing devices are used on the eastbound and westbound intersection approaches such that only one lane is open on each approach. This technique facili- tates safe intersection operation using flagger direction. The signal is set to a red flash operation. Work Zone Capacity Studies Hawkins et al. (1992) measured the capacity of one urban street midsignal work zone in Texas. The work zone was on a A p p e n d I x J HCM Urban Streets Methodology Enhancements: Saturation Flow Rate Adjustment Factor for Work Zone Presence

249 four-lane arterial street. It was a short-term work zone that closed one lane and left one lane open for the subject direc- tion of travel. Hawkins et al. measured the flow rate through the work zone during time periods when a queue was con- tinuously present. They estimated a work zone capacity of 760 vehicles per hour per lane (vphpl) for the open lane. Relative to a typical capacity of 1,800 vphpl for a traffic lane, the value estimated by Hawkins et al. (1992) suggests that the presence of a midsignal work zone reduces capacity by 1,040 vphpl (58%). This magnitude of reduction is significant and perhaps not typical of most urban street work zones. No other published reports could be found to corroborate the findings by Hawkins et al. The HCM2010 recommends 1,600 vphpl for a short-term freeway work zone, which is con- siderably larger than 760 vphpl. Elefteriadou et al. (2008) used a simulation model to estimate work zone capacity when the work zone was in the vicinity of a signalized intersection. They developed a set of regression equa- tions that could be used to predict the capacity based on a variety of factors that describe the signal timing, approach geometry, and distance between the work zone and inter- section. They used the calibrated models to determine that, for a single-lane closure on a three-lane approach, the capacity would range from 385 to 1,005 vphpl, depending on the fac- tors mentioned. An examination of the model’s regression coefficients indicated that the presence of the work zone at the intersection reduces approach capacity by about 218 vphpl. Factors Affecting Work Zone Operation Joseph et al. (1988) developed a simulation model for evalu- ating work zones on signalized arterial streets. Their research revealed that work zone effect on traffic operation was depen- dent on work zone location (relative to the signalized inter- section), signal timing, and the degree to which arrivals were concentrated in platoons. Elefteriadou et al. (2008) developed a series of equations for predicting intersection approach capacity when a work Table J.1. Nationwide Effect of Work Zones on Operation Highway Type Work Zone Type Capacity Reduction, Vehicles/Year (thousands) Delay, Vehicle Hours/Year (thousands) Urban freeways and expressways All 1,702,000 730,000 Urban other principal arterials All 1,329,000 10,000 Note: Based on 1999 data. Urban freeway and expressway use is 544,000 million vehicle miles; urban other princi- pal arterial use is 393,000 million vehicle miles. Table J.2. Urban Street Work Zone Characteristics Category Characteristic Relative to Highway and Freeway Work Zones, the Urban Street Work Zone Has . . . Geometry Midsignal access More frequent driveway access, which may disrupt platoon progression by vehicles turning into or out of the major-street work zone and introduce significant speed variation on the urban street. Cross section Undivided cross section in some cases, which reduces the lateral separation between opposing vehicles in many work zone configurations. Higher likelihood of right-of-way constraint, which may result in narrow traffic lanes and the need for barrier protection for work zone occupants. No shoulders, which may limit work zone configuration options that could otherwise minimize work zone impact on capacity. Traffic characteristics Pedestrians More frequent pedestrians, whose accommodation in the work zone can reduce the right-of-way available to serve vehicles in the work zone. Left turns A larger portion of left-turn vehicles, which could cause increased delay if left-turn capacity is reduced by work zone lane restrictions or queue spillback. Traffic control Signals More frequent signalized intersections, whose coordinated operation is often disrupted by work zone presence and whose detectors are often disabled by construction activities. Stop or yield control High-volume turn movements at unsignalized access points that may not have adequate capacity due to work zone–related queue spillback.

250 zone was present. Each equation addressed different approach lane configurations at the intersection. The variables in the equations indicated that the capacity of the work zone is a function of the percentage of left-turning vehicles, the dis- tance between the work zone and intersection, and the green– to–cycle length ratio of the lane group. Hawkins et al. (1992) observed several urban street work zones in Texas. They noted that the following factors had some influence on the operation of an urban street with a midsegment work zone: pedestrian presence, driveway access, barriers that block sight lines, narrow lanes that make it dif- ficult to turn into or out of driveways, and lateral clearance between open lanes and the work zone. The factors identified in the review of the literature that can affect the capacity of a signalized intersection are summarized in Table J.3. These factors could represent input variables in a model or procedure for predicting intersection capacity. In this regard, the model or procedure would be used to evaluate each intersection approach separately. Kianfar et al. (2011) conducted a state-of-the-practice survey that included all state departments of transportation in the United States. The survey focused on freeway work zones; however, some of the findings are relevant to the dis- cussion of urban street work zones. One of the questions related to the factors that influence work zone capacity. The factors selected by a majority of the respondents are identi- fied in Column 2 of Table J.3. They include work zone length, number of open lanes, lane width, and heavy-vehicle percentage. Freeway Work Zone Capacity Chapter 10 of the HCM2010 provides information for esti- mating the capacity of freeway work zones. It differentiates between short-term work zones and long-term work zones. Short-term work zones are noted to have standard channel- izing devices (e.g., cones, drums) to demarcate the work area and work activities that tend to last a few hours or weeks. Long-term work zones are noted to have portable concrete barriers to demarcate the work area and work activities that tend to last a few months or years. For short-term work zones, a procedure is provided in the HCM2010 to predict freeway capacity. A base capacity value of 1,600 passenger cars per hour per lane (pcphpl) is recom- mended. It can be adjusted for the level of work activity, the presence of heavy vehicles, and the presence of ramps. For long-term work zones, Exhibit 10-14 in the HCM2010 lists default freeway capacity values for selected lane reduc- tion combinations. These values are repeated in Table J.4. Two trends are suggested by these values. One trend is that a Figure J.1. Typical intersection work zone.

251 long-term work zone has a lower capacity than a short-term work zone. The second trend in the values shown in Table J.4 relates to the change in capacity with number of lanes. The capacity values shown in the table suggest that capacity per lane is higher for freeways with many normal lanes. It is possible Table J.3. Work Zone–Related Factors that May Affect Intersection Capacity Category Factora Work zone data Work zone lengtha Location of closed lane (outside, middle, or inside; parking) Work intensity (presence of equipment and workers) Work duration (number of days since work zone installed) Police presence Time of work activity (daytime, nighttime) Geometry Number of open lanes in the work zonea Approach grade Lane width in the work zonea Lateral clearance to the work zone and to opposing lanes Driveway presence Provision or closure of turn lanes at intersection Traffic characteristics Traffic demand volume Heavy-vehicle percentagea Lane utilization (or lane volume) on intersection approach Turn-movement percentages Pedestrians at intersection and along street, if sidewalk is closed Traffic control Speed limit prior to work zone and speed limit in work zone Use of flagger or signal control Type of devices used to delineate work zone (cones, barrier, other) Effective green duration and cycle length, if signalized aThese factors are most frequently considered by practitioners (Kianfar et al. 2011). Table J.4. HCM2010 Default Capacity Values for Long-Term Freeway Work Zones No. of Lanes Freeway Capacitya (vphpl)Normal Operation During Work Zone 2 1 1,400 2 2 N/A 3 1 1,450 3 2 1,450 3 3 N/A 4 2 1,450 4 3 1,500 a Values from Exhibit 10-14 in Chapter 10 of the HCM2010. N/A = not applicable, data not available. that this trend is confounded with area type (i.e., urban free- ways tend to have more lanes and more aggressive drivers than rural freeways). Freeway work zone capacity has been the subject of sev- eral research projects in the past 20 years. Data from the reports associated with several of these projects are listed in Table J.5. Collectively, these data represent a range in num- ber of lanes, proportion of heavy vehicles, and work zone duration. These data were statistically reexamined to determine if there was an underlying trend between the number of lanes, lanes reduced for work zone, proportion of heavy vehicles, size of lane closure (not shown), work zone duration, and capacity. The regression model described by Equations J.1 to J.4 was used for this evaluation. = × × ×fh fh fh (J.1)wz hv long reduceh bo with ( )= + −fh 1.0 1.0 (J.2)hv hv hvp b

252 Table J.5. Reported Capacity Values for Freeway Work Zones Source Lanes Open During Normal Operation Lanes Open During Work Zone Proportion Heavy Vehicles Work Zone Duration Measured Capacitya (vphpl) Benekohal et al. (2003) 2 1 0.294 Long 2,062 2 1 0.347 Long 1,710 2 1 0.382 Long 2,088 2 1 0.061 Long 1,981 2 1 0.426 Long 1,615 2 1 0.169 Long 2,167 2 1 0.189 Long 2,033 2 1 0.145 Long 2,004 Al-Kaisy and Hall (2002) 3 3 0.0b Long 2,252 4 4 0.0b Long 1,853 4 2 0.0b Long 1,989 4 2 0.0b Long 1,985 Kim et al. (2001) 4 3 0.082 Short 1,612 4 3 0.081 Short 1,627 4 3 0.090 Short 1,519 4 3 0.103 Short 1,790 4 3 0.080 Short 1,735 4 3 0.101 Short 1,692 4 2 0.143 Short 1,290 4 2 0.085 Short 1,228 4 2 0.110 Short 1,408 4 2 0.113 Short 1,265 4 2 0.046 Short 1,472 4 2 0.099 Short 1,298 Dixon et al. (1996) 2 1 0.072 Shortc 1,637 2 1 0.118 Shortc 1,644 2 1 0.045 Shortc 1,787 2 1 0.214 Shortc 1,692 2 1 0.193 Shortc 1,440 Jiang (1999) 2 1 0.250 Shortc 1,500 2 1 0.120 Shortc 1,572 2 1 0.110 Shortc 1,190 2 1 0.320 Shortc 1,308 2 1 0.310 Shortc 1,320 (continued on next page)

253 Table J.5. Reported Capacity Values for Freeway Work Zones Source Lanes Open During Normal Operation Lanes Open During Work Zone Proportion Heavy Vehicles Work Zone Duration Measured Capacitya (vphpl) Krammes and Lopez (1992) 3 1 0.121 Short 1,304 3 1 0.129 Short 1,387 3 1 0.151 Short 1,534 3 1 0.044 Short 1,665 3 1 0.105 Short 1,435 3 1 0.118 Short 1,311 3 1 0.031 Short 1,470 3 1 0.133 Short 1,405 3 1 0.150 Short 1,498 3 1 0.227 Short 1,502 3 1 0.174 Short 1,544 2 1 0.049 Short 1,447 2 1 0.071 Short 1,539 2 1 0.032 Short 1,641 2 1 0.034 Short 1,555 2 1 0.034 Short 1,478 2 1 0.028 Short 1,668 2 1 0.132 Short 1,522 2 1 0.049 Short 1,521 2 1 0.034 Short 1,615 2 1 0.040 Short 1,682 2 1 0.036 Short 1,661 4 2 0.167 Short 1,479 4 2 0.151 Short 1,430 4 2 0.041 Short 1,860 4 2 0.085 Short 1,402 4 2 0.045 Short 1,406 5 3 0.018 Short 1,681 5 3 0.021 Short 1,479 4 3 0.037 Short 1,668 4 3 0.039 Short 1,471 4 3 0.037 Short 1,681 4 3 0.057 Short 1,387 a The capacity values were measured, but the measurement technique varies among researchers. b Reported capacity is in terms of equivalent passenger cars per hour per lane. c Duration not stated by author. Short duration assumed from work zone description. (continued)

254 = +fh 1.0 (J.3)long long longb I ( )= + −fh 1.0 (J.4)reduce reduce wzb n no where hwz = saturation headway when a work zone is present, s/vehicle (s/veh); fhhv = adjustment factor for heavy vehicles; fhlong = adjustment factor for work zone duration; fhreduce = adjustment factor for reducing lanes during work zone presence; phv = proportion of heavy vehicles; Ilong = indicator variable for work zone duration (1.0 if long term, 0.0 if short term); no = number of lanes open during normal operation; nwz = number of lanes open during work zone presence; and bi = regression coefficient i. The regression model is developed to predict the satura- tion headway when a work zone is present. This headway is computed by dividing 3,600 by the freeway capacity provided in the far-right column of Table J.5. The regression coeffi- cient bo represents the equivalent through-vehicle saturation headway for short-term freeway work zones with no lane reduction. The adjustment factor for heavy vehicles is a variation of Equation 10-8 from the HCM2010. The regression coefficient in Equation J.2 represents the passenger car equivalent for heavy vehicles. The adjustment factor for reducing lanes was derived fol- lowing an examination of the trends in Tables J.4 and J.5. Alter- native forms of Equation J.4 were explored, but that shown was found to provide the best fit to the data. The statistics associated with the calibrated model are shown in Table J.6. The coefficient of determination R2 is .59. The coefficient bo suggests that the saturation headway for short-term work zones is 2.0739 s/pc. This value equates to a capacity of 1,736 pcphpl for short-term work zones. The coefficient bhv has a value of 1.4556. This value is similar in magnitude to the passenger car equivalent for trucks in level terrain of 1.5 that is provided in Exhibit 11-10 of the HCM2010. The regression coefficient blong represents the effect of work zone duration and demarcation devices. Its value is -0.2371. This value suggests that the saturation headway for long-term freeway work zones is 24% smaller than that for a short-term headway. Alternatively, it suggests that the capacity for the long-term work zone is 31% larger than for a short-term work zone. Al-Kaisy and Hall (2002) rationalize that this increase is likely due to drivers feeling more secure with con- crete barriers than plastic barrels, and their greater familiarity with long-term work zones than short-term work zones. However, it is noted that this trend is opposite to that in the capacity values provided in Chapter 10 of the HCM2010. The regression coefficient breduce represents the effect of lane reductions through the work zone. The positive value of this coefficient suggests that saturation headway is higher at work zones where there are many lanes closed relative to work zones where there are few lanes closed. This trend may reflect the amount of turbulence in the approaching traffic stream that is forced to merge before reaching the work zone. If there is one lane closed for a work zone, this factor has a value of 1.075. If it is a short-term work zone, then the saturation headway is 2.228 (2.0739 × 1.075), which equates to a capac- ity of 1,616 pcphpl. This latter value compares favorably with that recommended in Chapter 10 of the HCM2010 for short- term work zones. Methodological Issues This subsection describes the appropriate method for estimat- ing saturation flow rate using field data. The accuracy of the Table J.6. Model Statistical Description: Freeway Saturation Headway During Work Zone Model Statistic Value R2 0.59 Observations no 67 sites Calibrated Coefficient Values Variable Inferred Effect Value Std. Dev. t-statistic b0 Saturation headway for short-term work zones (s/pc) 2.0739 0.0801 25.9 bhv Passenger car equivalent for heavy vehicles 1.4556 0.1468 9.9 blong Adjustment factor for long-term work zone -0.2371 0.0281 -8.5 breduce Adjustment factor for lane reduction at work zone 0.0745 0.0261 2.9

255 saturation flow rate estimate for a specific lane (or lane group) is highly dependent on the method used to aggregate the data that are recorded in each signal cycle. The underlying issue is whether to base the computation of overall saturation flow rate either on individual measurements of average headway per cycle or on individual observations of saturation flow rate per cycle. The two methods yield estimates of overall saturation flow rate that differ by about 50 vphpl. The reason for the difference in the two methods is due to two factors: (1) the cycle-based statistics (i.e., average headway per cycle and average saturation flow rate per cycle) have a random component, and (2) one statistic is the reciprocal of the other. From a mathematical standpoint, a randomly distributed variable that is converted by reciprocal and averaged will not equal the reciprocal of the average value of the randomly distributed variable. The appropriate averaging method is the one that yields an unbiased estimate of cycle capacity. Bonneson et al. (2005) demonstrated that average saturation flow rate is accurately computed from individual measurements of average head- way per cycle. Site Selection and data Collection This section describes the criteria used to select study sites and the plan established for collecting the data needed to quantify the effect of work zone presence on saturation flow rate. Intersections in several states were considered for inclu- sion in the database assembled for this project. For this research, a study site is defined as one intersection approach. At each site, data were collected for the through-lane group. This lane group includes any combination of exclusive through lanes and shared through and right-turn lanes. The study design is described as an observational during– after study. Data were collected at each study site when the work zone was present, and then again after the work zone was removed. The study was observational because the local transportation agencies selected the intersections requiring maintenance or reconstruction. The next part of this section describes the site selection cri- teria and the process used to select the study sites. The third part of this section describes the data collection plan. This plan describes the data to be collected, data collection meth- ods, study duration, and sample size. The last part of this sec- tion describes the data reduction procedures. Site Selection The selection of suitable study sites was based on a range of criteria. The criteria used were based on the annual average daily traffic (AADT), work zone end date, work zone dura- tion, and number of lanes closed for the work zone. The volume criterion was established as a minimum AADT of 3,550 vehicles per day per lane. This volume was used to max- imize the potential for acquiring the desired minimum sam- ple size during a study during one peak traffic period. The work zone end-date criterion was used to ensure that the work zone would be removed in a timely manner, such that the after study could be completed within the time schedule of the research project. The other two criteria were used to guide site selection such that a range of values for each crite- rion were represented in the database. In addition to these criteria, the following desirable site characteristics were established to guide the selection process: • A left-turn bay must be present on any approaches where left-turn movements occur; • Approaches should not have sharp curves or other unusual horizontal or vertical geometry; • Approaches should have a grade in the range of -0.5% to +0.5%; and • Approaches should not experience queue spillback during the study period. It was determined that a minimum of eight study sites would need to be in the database to collectively represent the desired combinations of work zone duration and number of lanes. Table J.7 lists the study sites selected for field data collec- tion. The sites represent three states. The approach width in Column 6 describes the total width of open lanes when the work zone is present. It includes the width of the left-turn, through, and right-turn lanes and describes the lateral distance between the work zone channelizing devices (and curb, if the devices are only on one side of the approach). As stated previously, a long-term work zone typically includes portable concrete barriers to demarcate the work area and work activities that tend to last a few months or years. Only Site 3 had these characteristics. Sites 5 and 7 had the characteristics of a short-term work zone. The other sites had a combination of the characteristics of both categories. Data Collection For a given site, the data were collected using two camcorders. The location of these camcorders is shown in Figure J.2. One camera was mounted on a pole just behind the curb and facing the intersection. This camera was used to deter- mine whether there were at least 10 vehicles in queue at the start of green and the time that the signal indication changed. This camera was positioned such that its field of view included (1) at least one controlling signal head for the subject through and right-turn movements and (2) a view of each traffic lane serving the subject movements (up to three lanes). At most

256 sites, this camera was located in the range of 300 to 500 ft upstream of the stop line. The second camera was mounted on a pole just behind the curb at the stop line and facing in a direction perpendicular to the flow of traffic on the subject approach. This camera was used to determine the time that the front axle of each queued vehicle crossed the stop line. The clock in each video- tape recorder was synchronized to a master clock at the start of each study. The cameras recorded traffic events for a minimum of 4 h during each study (i.e., 4 h during work zone operation and 4 h after the work zone was removed). The objective for each site was to record a minimum of 270 vehicles that were in saturation flow (i.e., in Queue Position 5 and higher) during each study. At all sites, the cameras were maintained for longer periods of time to maximize the number of headway obser- vations. Resource constraints limited the camera deployments to a maximum of 40 h of recorded traffic operation at any given site. The technicians recorded for each site the data described in the following list: • Street names; • Approach lane assignments; • Approach lane width; • Left-turn bay length and lanes; • Right-turn bay length and lanes; • Posted speed limit; • Bus stop location, if present; • On-street parking, if present; • Driveway location, if present; • Adjacent land use (e.g., office, commercial, residential, industrial), if present; and • Camera location (i.e., distance from stop line). Figure J.2. Typical camera location on the intersection approach. Camera 1 Camera 2 Table J.7. Study Sites by Location State Site No. Intersection No. of Left and Through Lanes Approach Width During Work Zone (ft) Work Zone Duration (days) Traffic Control Devices Demarcating Work Area After Work Zone During Work Zone Arizona 1 E. Valencia Road and S. Alvernon Way 5 2 26.6 120 Cone Texas 2 Kirby Drive and US-59 westbound Frontage Road 3 2 31.0 290 Drum Florida 3 Brickell Avenue and S.E. 13th Street 3 2 20.0 290 Concrete barrier 4 Sample Road and N.W. 54th Avenue 5 4 50.0 150 Drum Texas 5 W. Holcombe Blvd. and Buffalo Speedway 4 2 22.0 2 Drum Arizona 6 N. Sabino Canyon Road and E. Tanque Verde Road 5 4 48.0 100 Cone 7 E. Thomas Road and N. 46th Street 4 2 22.3 10 Cone 8 Cave Creek Road and Greenway Pkwy. 4 2 23.9 100 Cone

257 In addition to the data in the previous list, the data described in the following list were collected at each site that had a work zone present: • Weather condition; • Distance from start of work zone to stop line; • Distance from end of work zone to stop line; • Type of traffic control devices used to demarcate the work area; • Number of open lanes on the study approach; and • Number of work zone–related workers present. For each work zone study site, the technicians obtained (when available) a copy of the traffic control plan for the work zone and the date the work zone was installed. Data were collected in the winter of 2011 and spring of 2012 during time periods that are reflective of typical peak traffic periods at each study site. These periods typically occurred during midweek days (i.e., Tuesday, Wednesday, and Thurs- day) in the morning, afternoon, and evening peak periods. Data were not collected during holidays, periods of inclement weather, or incidents. Data Reduction After each study, the videotape recordings were replayed in the office. Saturation headway data were extracted from these recordings using the technique described in Section 6 of Chapter 31 of the HCM2010. The data extracted for each sig- nal cycle included the following items: • Time of start of green; • Time of end of green; • Discharge time of the first queued vehicle; • Discharge time of the fourth queued vehicle; • Discharge time of the eighth queued vehicle; • Discharge time of the last queued vehicle (only used in spe- cial circumstances); • Number of heavy vehicles in Queue Positions 1 through 4; • Number of heavy vehicles in Queue Positions 1 through 8; • Number of heavy vehicles in queue (only used in special circumstances); • For shared-lane groups, number of right-turn vehicles in Queue Positions 1 through 4; • For shared-lane groups, number of right-turn vehicles in Queue Positions 1 through 8; • For shared-lane groups, number of right-turn vehicles in queue (only used in special circumstances); • Number of queued vehicles at the start of green; and • Number of vehicles served during the cycle. In those situations when volumes were unexpectedly light, it was necessary to record the times for the fourth and last queued vehicles. In this situation, the last queued vehicle was in Queue Position 6 or 7. data Analysis and Findings This section describes the development and evaluation of sev- eral saturation flow adjustment factors that explain the effect of work zone presence on saturation flow rate. The effect of each factor on saturation flow rate, as reported in the litera- ture, was described in a previous section. This section con- cludes with a description of the equations used to estimate the saturation flow rate for a signalized movement when a work zone is present. Database Summary This subsection summarizes the data collected at eight sig- nalized intersection approaches in three states. Each approach represents one field study site. Study site traffic characteristics are summarized in Table J.8. One observa- tion represents the average of the saturation headways measured during one signal cycle. The data represent mea- surements for 3,429 vehicles in saturation flow during the after study and for 3,772 vehicles during the work zone present study. Data were not collected at Site 3 for the after work zone condition. This outcome was unexpected because the work period was scheduled for completion during the data collec- tion phase of this project. However, the operating agency extended the work period for unknown reasons such that it was not possible to collect the after data in a time frame that would be useful to the project. The after and during saturation headway for each site can be compared to assess the effect of work zone presence. For six of the sites, the saturation headway during the work zone was found to be higher than that of the after condition. In contrast, Sites 2 and 4 were found to have slightly larger satu- ration headways during the after condition than when the work zone was present. It is not clear from the data why this result occurred, but it is contrary to expectation. Site 3 is the only site considered to have a long-term work zone. Its saturation headway when the work zone was pres- ent was 2.62 s/veh. This value corresponds to a saturation flow rate of 1,374 vphpl. This saturation headway is larger than that of any other site. This trend suggests that intersec- tions with long-term work zones have lower capacity than those with short-term work zones. This trend is consistent with that described in Chapter 10 of the HCM2010, but it is contrary to that found in the analysis of the freeway data in Table J.5. This finding suggests that characterizing work zone characterizations as short-term or long-term is insufficient

258 to describe systematic variation in work zone saturation flow rate or capacity. The average values in the last row can be used to quantify the approximate effect of work zone presence on saturation flow rate. Specifically, the ratio of the two saturation head- ways indicates that work zone presence decreases saturation flow rate by about 5.2% (100 × [1.0 - 2.18/2.30]). Model Development This subsection describes the development of a regression model that was used to estimate the adjusted saturation head- way for each site. The adjusted saturation headway is defined as the saturation headway for an equivalent through-car traf- fic steam served in a 12-ft traffic lane. This adjusted value is estimated separately for the after and during periods by using Equations J.5 to J.8: = × × × ×fh fh fh (J.5)site hv rtT b ns s w with = + + + + + + + (J.6)site 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8b b b I b I b I b I b I b I b I ( )= + −fh 1.0 1.0 (J.7)hv hv hvp b ( )= + −fh 1.0 1.0 (J.8)rt rt rtp b where Ts = saturated discharge time, s; ns = number of queued vehicles represented in saturated discharge time; fhhv = adjustment factor for heavy vehicles; fhrt = adjustment factor for right-turn vehicles; fhw = adjustment factor for lane width (1/0.96 if lw < 10.0 ft; 1/1.04 if lw > 12.9 ft; 1.0 otherwise); lw = average lane width, ft; Ii = indicator variable for site i (1.0 if site i, 0.0 otherwise); phv = proportion of heavy vehicles; prt = proportion of right-turn vehicles; bi = regression coefficient for site i (i = 1, 2, . . . , 8); bhv = regression coefficient for the effect of heavy vehicles; and brt = regression coefficient for the effect of right-turn vehicles. The model described by Equations J.5 to J.8 estimates the adjusted saturated headway by using a method that is consistent with the technique described in Chapter 31 of the HCM2010. The regression coefficients bi quantify the adjusted saturation headway for each site. The adjusted saturation flow rate is computed post hoc by dividing the adjusted saturation headway into 3,600. The regression coefficient b1 defines the adjusted saturation headway of Site 1. The regression coefficient b2 describes the incremental change in the saturation headway for Site 2 rela- tive to Site 1. Thus, the adjusted saturation headway for Site 2 is computed as the sum of b1 and b2. A similar approach is used to determine the adjusted saturation headway for Sites 3 to 8. The adjustment factor for lane width is based on the satu- ration flow rate adjustment factor in Chapter 18 of the HCM2010. The reciprocal of the values cited in the HCM are used as headway adjustment factors in Equation J.5. The regression coefficient bhv represents the passenger car equivalent for heavy vehicles. Similarly, the regression coef- ficient brt represents the through-vehicle equivalent for right- turning vehicles. Equivalency factors are cited in Chapter 18 Table J.8. Study Site Traffic Characteristics Site No. After Work Zone During Work Zone no ho (s/veh) phv prt no ho (s/veh) phv prt 1 98 2.20 0.023 0.000 98 2.43 0.105 0.084 2 199 2.18 0.008 0.000 53 2.16 0.009 0.000 3 N/A N/A N/A N/A 75 2.62 0.020 0.147 4 219 2.17 0.021 0.000 257 2.03 0.019 0.000 5 139 2.24 0.045 0.169 162 2.59 0.014 0.134 6 143 1.98 0.005 0.000 141 2.04 0.048 0.009 7 24 2.40 0.021 0.010 84 2.53 0.036 0.033 8 132 2.29 0.030 0.078 73 2.46 0.031 0.038 Summary 954 2.18 0.021 0.036 943 2.30 0.033 0.051 Note: no = number of observations of average saturation headway per cycle; ho = average saturation headway; phv = proportion of heavy vehicles; prt = proportion right-turn vehicles in shared through and right-turn lane; N/A = not applicable, data not available.

259 of the HCM2010 for heavy vehicles and right-turn vehicles as 2.0 and 1.18, respectively. These values can vary widely on a cycle-by-cycle basis depending on truck and turn vehicle presence. For this reason, it is appropriate to quantify repre- sentative values for these factors as part of the regression analysis. Unlike the heavy-vehicle and right-turn adjustment fac- tors, the lane width adjustment factor is not derived from the collected data. Rather, the lane width adjustment factor is obtained from the HCM. The reason for this approach is that lane width varies on a site-by-site basis and not on a cycle-by-cycle basis. It is rationalized that the HCM factor values represent the best-estimate lane width effect given that there are only eight sites represented in the database, and that lane width does not vary widely among these sites. The statistics associated with the calibrated model using after work zone data are shown in Table J.9. The coefficient of determination R2 is .61. The coefficient bhv has a value of 1.5115. This value is slightly smaller in magnitude than the passenger car equivalent for heavy vehicles provided in the HCM. It suggests that there is either a larger proportion of small trucks in the observed traf- fic streams, or that heavy-vehicle performance has improved since the HCM value was quantified. The coefficient brt has a value of 1.2076. This value is simi- lar in magnitude to the through-vehicle equivalent for right- turn vehicles provided in the HCM. The statistics associated with the calibrated model using during work zone data are shown in Table J.10. The coefficient of determination R2 is .29. This coefficient is about one-half as large as that shown in Table J.9. This trend suggests that there is more random variability in the headways measured at the sites when a work zone is present. This trend is plausible given the added uncertainty in a work zone driving environment. The adjusted saturation headway for each site was com- puted using the data in the two previous tables. These head- way estimates are shown in Table J.11. The two adjusted values for Site 1 were obtained directly from Tables J.9 and J.10. The value for Site 2 during the after condition was computed as 2.1768 (2.0230 + 0.1538). The values for the other sites and conditions were computed in a similar manner. The far-right column in Table J.11 compares the two rates using the ratio of after headway divided by during headway. A ratio that is less than 1.0 indicates that the work zone–related saturation headway is larger than the saturation headway for the same movement when there is no work zone. The ratios in Table J.11 represent an estimate of the average saturation flow rate adjustment factor for work zones. The overall average of 0.90 shown in the last row suggests that work zone presence decreases saturation flow rate by about 10%. This value is larger than that found when examining the unadjusted values shown in Table J.8. The saturation headways listed in Table J.11 were exam- ined to determine if there was a plausible systematic variation that could be related to the work zone characteristics. A regression analysis was used for this examination. The adjusted saturation headway for the after condition for Site 3 was not computed because the data were not collected. The overall average value of 2.0903 s/pc was substituted for this missing value for the regression analysis. For the analysis, the after headway values were compared with the during headway values on a site-by-site basis. This Table J.9. Model Statistical Description: Saturation Headway After Work Zone Model Statistic Value R2 0.61 Observations no 954 cycles (3,429 vehicles) Calibrated Coefficient Values Variable Inferred Effect Value SD t-statistic b1 Saturation headway of Site 1, s/veh 2.0230 0.0536 37.7 bhv Passenger car equivalent for heavy vehicles 1.5115 0.0889 17.0 brt Through-vehicle equivalent for right-turning vehicles 1.2076 0.0531 22.7 b2 Incremental saturation headway of Site 2, s/veh 0.1538 0.0599 2.6 b4 Incremental saturation headway of Site 4, s/veh 0.1285 0.0592 2.2 b5 Incremental saturation headway of Site 5, s/veh 0.0893 0.0650 1.4 b6 Incremental saturation headway of Site 6, s/veh -0.0448 0.0622 -0.7 b7 Incremental saturation headway of Site 7, s/veh 0.1160 0.1673 0.7 b8 Incremental saturation headway of Site 8, s/veh 0.0285 0.0705 0.4

260 approach was used to control for other, unmeasured differ- ences among sites. The work zone characteristics that were considered during model development included number of lanes after the work zone was removed, number of lanes when the work zone was present, approach width, work zone dura- tion, and traffic control devices used to demarcate the work area. The values for each characteristic are shown in Table J.7. Equations J.9 to J.11 were used to model the effect of vari- ous work zone characteristics on the adjusted saturation headway: = × × ×fh fh (J.9)wz wz wid reduceh b ho with ( )= + −fh 1.0 12 (J.10)wid widb aw ( )= + −fh 1.0 (J.11)reduce reduce wzb n no where hwz = adjusted saturation headway during work zone, s/pc; ho = adjusted saturation headway after work zone, s/pc; fhwid = adjustment factor for approach width; fhreduce = adjustment factor for reducing lanes during work zone presence; aw = approach lane width during work zone (total width of all open left-turn, through, and right- turn lanes), ft; no = number of left-turn and through lanes open during normal operation; nwz = number of left-turn and through lanes open during work zone presence; bwz = regression coefficient for the effect of work zone presence; bwid = regression coefficient for the effect of approach lane width; and breduce = regression coefficient for the effect of reducing lanes. The statistics associated with the calibrated model are shown in Table J.12. The coefficient of determination R2 is .81. The R2 adjusted for sample size is .73. It is recognized that there are only eight sites in the database and that the model Table J.11. Estimated Adjusted Saturation Headway Site No. Adjusted Saturation Headway After Work Zone (s/pc) Adjusted Saturation Headway During Work Zone (s/pc) Ratio (after/during) 1 2.0230 2.4248 0.83 2 2.1768 2.1522 1.01 3 N/A 2.5470 N/A 4 2.1515 2.0211 1.06 5 2.1123 2.5195 0.84 6 1.9782 2.0096 0.98 7 2.1390 2.4984 0.86 8 2.0515 2.4229 0.85 Average 2.0903 2.3244 0.90 N/A = not applicable, data not available. Table J.10. Model Statistical Description: Saturation Headway During Work Zone Model Statistic Value R2 0.29 Observations no 934 cycles (3,736 vehicles) Calibrated Coefficient Values Variable Inferred Effect Value SD t-statistic b1 Saturation headway of Site 1, s/veh 2.4248 0.0455 53.3 bhv Passenger car equivalent for heavy vehicles 1.2744 0.0599 21.3 brt Through-vehicle equivalent for right-turning vehicles 1.1641 0.0376 31.0 b2 Incremental saturation headway of Site 2, s/veh -0.2727 0.0722 -3.8 b3 Incremental saturation headway of Site 3, s/veh 0.1222 0.0639 1.9 b4 Incremental saturation headway of Site 4, s/veh -0.4037 0.0515 -7.8 b5 Incremental saturation headway of Site 5, s/veh 0.0946 0.0545 1.7 b6 Incremental saturation headway of Site 6, s/veh -0.4153 0.0554 -7.5 b7 Incremental saturation headway of Site 7, s/veh 0.0736 0.0640 1.2 b8 Incremental saturation headway of Site 8, s/veh -0.0018 0.0643 0.0

261 Table J.12. Model Statistical Description: Saturation Headway Adjustment Factors Model Statistic Value R2 0.81 Adjusted R2 0.73 Observations no Eight sites Calibrated Coefficient Values Variable Inferred Effect Value SD t-statistic bwz Adjustment factor for work zone presence 1.1654 0.0670 17.4 bwid Adjustment factor for work zone approach width -0.0057 0.0012 -4.9 breduce Adjustment factor for lane reduction at work zone 0.0402 0.0265 1.5 has three regression coefficients. Hence, the coefficient of determination is likely to be larger than would truly be obtained if there were more sites in the database. The adjusted R2 value accounts for the small sample size to some degree. The regression coefficient bwz represents the effect of work zone presence on saturation headway. Its value of 1.1654 indicates that saturation headway increases 16.54% when a work zone is present. Alternatively, saturation flow decreases by 14% when a work zone is present. The regression coefficient bwid represents the effect of approach width on saturation headway. It describes the lateral distance between the work zone channelizing devices (and curb, if the devices are only on one side of the approach). The coefficient value of -0.0057 indicates that headway decreases with increasing approach width. Thus, the adverse effect of work zone presence on performance is lessened if there are many open lanes (or a few wide lanes) on the approach. The regression coefficient breduce represents the effect of lane reductions through the work zone. This coefficient is slightly smaller than that found in the regression analysis of the freeway data listed in Table J.5. The positive value of this coefficient suggests that saturation headway is higher at work zones where there are many lanes closed relative to work zones where there are few lanes closed. This trend may reflect the amount of turbulence in the approaching traffic stream that is forced to merge before reaching the work zone. Figure J.3 compares the predicted headway values from Equation J.9 with those from Column 3 of Table J.11 (i.e., the dependent variable). The trend line shown is not the line of best fit. Rather, it is an x = y line such that each data point would lie on this line if the predicted value equaled the measured value. There are eight data points shown in the fig- ure, one data point for each site. They are shown to vary around the line, with no apparent bias over the range of pre- dicted values. 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 1.9 2.1 2.3 2.5 2.7 2.9 Predicted Saturation Headway, s/pc M ea su re d Sa tu ra tio n H ea dw ay , s/ pc 1 1 Figure J.3. Comparison of measured and predicted saturation headways. Saturation Flow Rate Adjustment Factors This subsection describes a series of equations that can be used to compute a saturation flow rate adjustment factor for work zone presence. This factor would be used with the pro- cedure in Chapter 18 of the HCM2010 to estimate the satura- tion flow rate of a lane group when there is a work zone on the associated intersection approach. The saturation flow rate adjustment factor can be com- puted using Equations J.12 to J.14: = × × ≤0.858 1.0 (J.12)wz wid reducef f f with ( )= − − 1.0 1.0 0.0057 12 (J.13)widf aw ( )= + − 1.0 1.0 0.0402 (J.14)reduce wz f n no

262 where fwz = saturation flow rate adjustment factor for work zone presence; fwid = saturation flow rate adjustment factor for approach width; freduce = saturation flow rate adjustment factor for reducing lanes during work zone presence; aw = approach lane width during work zone (total width of all open left-turn, through, and right-turn lanes), ft; no = number of left-turn and through lanes open during normal operation; and nwz = number of left-turn and through lanes open during work zone presence. Equation J.12 produces values less than 1.0 for a wide range of conditions. However, when the approach has many lanes open while the work zone is present (or a few wide lanes), then Equation J.12 can mathematically produce a value that exceeds 1.0. In these few instances, a value of 1.0 is recom- mended as an upper bound on the factor value. Table J.13 illustrates the value of the factor predicted by Equation J.12 for typical work zone conditions. The values obtained from Equation J.12 are shown in Column 7. The val- ues are shown to range from 0.790 to 1.000. If the saturation flow rate for a given intersection lane group is 1,800 vphpl when no work zone is present, then the saturation flow rate for this lane group when a work zone is present is shown in the far-right column of Table J.13. These saturation flow rate values are consistent with the freeway capacity values shown in Table J.4. The equations described in this section were calibrated using data for through-lane groups. However, it is suggested that the computed adjustment factor can also be used to esti- mate the saturation flow rate for left- and right-turn move- ments from exclusive lanes. The findings from the literature review were inconclusive regarding the effect of work zone duration (i.e., long term, short term) on saturation flow rate. The data collected for this project were not sufficient in number to shed further light on this issue. It appears that these two designations mask the many underlying factors that truly do have an effect on traffic operation when work zones are present. It is likely that future research on work zone capacity will be more fruit- ful if researchers abandon the use of descriptors such as long term and short term and instead focus on the individual characteristics of the work zone that may truly be influencing driver behavior. References Al-Kaisy, A., and F. Hall. Guidelines for Estimating Freeway Capacity at Long-Term Reconstruction Zones. Presented at 81st Annual Meet- ing of the Transportation Research Board, Washington, D.C., 2002. Benekohal, R., A.-Z. Kaja-Mohideen, and M. Chitturi. Evaluation of Con- struction Work Zone Operational Issues: Capacity, Queue, and Delay. Report No. ITRC FR 00/01-4. Department of Civil and Environmen- tal Engineering, University of Illinois at Urbana– Champaign, 2003. Table J.13. Illustrative Saturation Flow Rate Adjustment Factor Values No. of Left and Through Lanes Approach Width During Work Zonea (ft) Factor for Work Zone Presence Factor for Approach Width Factor for Lane Reduction Combined Factor Value Predicted Saturation Flow Rateb (vphpl) After Work Zone During Work Zone 2 1 11 0.858 0.994 0.961 0.820 1,476 2 2 22 0.858 1.060 1.000 0.910 1,638 3 1 11 0.858 0.994 0.926 0.790 1,421 3 2 22 0.858 1.060 0.961 0.875 1,575 3 3 33 0.858 1.136 1.000 0.975 1,755 4 2 22 0.858 1.060 0.926 0.842 1,516 4 3 33 0.858 1.136 0.961 0.937 1,687 4 4 44 0.858 1.223 1.000 1.000c 1,800 5 3 33 0.858 1.136 0.926 0.902 1,624 5 4 44 0.858 1.223 0.961 1.000c 1,800 5 5 55 0.858 1.325 1.000 1.000c 1,800 a Based on an average lane width of 11 ft per lane during work zone presence. b Based on a saturation flow rate of 1,800 vehicles per hour per lane without work zone, after adjustment for other conditions (e.g., grade). c Value rounded down to 1.00.

263 Bonneson, J., B. Nevers, J. Zegeer, T. Nguyen, and T. Fong. Guidelines for Quantifying the Influence of Area Type and Other Factors on Satura- tion Flow Rate. Project No. PR9385-V2. Florida Department of Transportation, Tallahassee, 2005. Chin, S., O. Franzese, D. Greene, H. Hwang, and R. Gibson. Temporary Losses of Highway Capacity and Impacts on Performance: Phase 2. Oak Ridge National Laboratory, Oak Ridge, Tenn., 2004. Dixon, K., J. Hummer, and A. Lorscheider. Capacity for North Carolina Freeway Work Zones. In Transportation Research Record 1529, TRB, National Research Council, Washington, D.C., 1996, pp. 27–34. Elefteriadou, L., M. Jain, and K. Heaslip. Impact of Lane Closures on Roadway Capacity, Part B: Arterial Work Zone Capacity. University of Florida, Gainesville, January 2008. Hawkins, H., K. Kacir, and M. Ogden. Traffic Control Guidelines for Urban Arterial Work Zones: Volume 2: Technical Report. FHWA/ TX-91/1161-5. Texas Transportation Institute, Texas A&M Univer- sity, College Station, February 1992. Jiang, Y. Traffic Capacity, Speed, and Queue-Discharge Rate of Indiana’s Four-Lane Freeway Work Zones. In Transportation Research Record: Journal of the Transportation Research Board, No. 1657, TRB, National Research Council, Washington, D.C., 1999, pp. 10–17. Joseph, C., E. Radwan, and N. Rouphail. Work Zone Analysis Model for the Signalized Arterial. In Transportation Research Record 1194, TRB, National Research Council, Washington, D.C., 1988, pp. 112–119. Kianfar, J., P. Edara, and C. Sun. Deriving Work Zone Capacities from Field Data: Case Studies of I-70 in Missouri. Presented at 90th Annual Meeting of the Transportation Research Board, Washington, D.C., 2011. Kim, T., D. Lovell, and J. Paracha. A New Methodology to Estimate Capacity for Freeway Work Zones. Presented at 80th Annual Meet- ing of the Transportation Research Board, Washington, D.C., 2001. Krammes, R., and G. Lopez. Updated Short-Term Freeway Work Zone Lane Closure Capacity Values: Interim Report. Report FHWA/TX-92/ 1108-5. Texas Transportation Institute, Texas A&M University, College Station, 1992. Transportation Research Board. Highway Capacity Manual 2010. TRB of the National Academies, Washington, D.C., 2010.