Analysis of Work Zone Crash Characteristics and Countermeasures (2018)

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.

NCHRP Project 17-61 6 CHAPTER 2 Causal Assessments of Work Zone Crashes Overview Much of the work zone safety literature is based around methodologies that develop statistical associations between work zone crashes and work zone features. Knowledge on crash causation, including the road environment as one of many factors in the causal chain, is limited. Unlike fundamental, causal relationships, the ‘signs’ and ‘magnitudes’ of statistical associations may be an artifact of the database used for analysis. Data accuracy and completeness vary from jurisdiction to jurisdiction and few existing roadway and work zone databases contain all measureable variables. Therefore, the predicted ‘outcomes’ of statistically estimated models do not always transfer from location to location. One approach to addressing this challenge is to explore the crash-generation process for individual crashes as opposed to looking only at descriptive statistics for commonly available crash variables. These methods were historically used in studies of pedestrian and bicyclist crashes (see, for example, Snyder and Knoblauch 1971), but have also been applied to understanding work zone crashes (Schrock et al. 2005). The basic hypothesis is that examining individual crash reports and photographs to better understand factors leading to a crash, instead of relying solely on computer-automated ‘slicing,’ modeling, and summarizing of pre-coded data also holds promise for identifying common crash “causal types.” For example, in the work zone context, crashes can be dissected and causality explored based on three factors as shown in Figure 1. Within a particular work zone crash type, one assesses the sequence of events pre-, during, and post-crash according to the following factors:  Precipitating events: the specific nature of the failure in the function/event sequence that led to the collision  Predisposing factors: specific environmental, human, or vehicle variables that influenced the function failure  Target groups: human populations and/or kinds of vehicle types involved in the crash type In this chapter, work zone crash information was extracted from three different datasets and examined using this approach to identify key characteristics that commonly found in work zone crashes. The first dataset was the VDOT RNS crash database, which is representative of databases available to most state DOTs. Analysis of VDOT RNS Crash Data The VDOT RNS crash database is constructed based on police crash report forms. Given that VDOT has responsibility for operating and maintaining all roadways outside of incorporated cities in the state, the work zone crashes extracted from this database included a range of roadway types.

NCHRP Project 17-61 7 Methodology Work zone crashes (defined as those occurring within a defined work zone area) were extracted from the database for calendar years 2011 and 2012. In total, 6,774 crashes were identified as occurring in work zones during this time frame. The crash causes were analyzed by crash type since causes varied significantly among crash types. Only those crash types that comprised at least 10% of work zone crashes were examined in the analysis. This resulted in the consideration of only rear-end, angle, sideswipes, and fixed object run–off-road crashes. These crash types represent 94.8% of all coded work zone crashes over the two-year study period and are assumed to be a good representation of the work zone crash population. Figure 1. Illustration of dissecting work zone crashes into causal factors. Crash report data fields and narrative descriptions provided by the responding officer, were used to assess the crash’s relationship to the work zone. Crashes were categorized as “directly related” if their crash description, in tandem with the other crash report fields, indicated that the work zone influenced the likelihood of the crash in some capacity. Factors considered when determining if a crash fell into the directly related category included whether:  A work zone vehicle or piece of equipment was struck  The crash narrative directly referenced a work zone feature  The crash narrative directly indicated that the work zone created changes in flow or speed  The narrative indicated a specific driver response to the work zone Work zone crash type: e.g., side-swipe crash in advanced warning area Precipitating events: The specific nature of the sequence of events that led to the collision Predisposing factors: Specific environmental, human, or vehicle variables that influenced the crash occurrence Target groups: Human populations and/or vehicle types involved

NCHRP Project 17-61 8 Even so, crashes determined to be directly related were not necessarily caused by the work zone and may have been only slightly influenced by the work zone’s presence. Conversely, crashes not determined to be directly related could still have been influenced by the work zone, but the officer failed to capture this influence in the crash report. In many cases, the crash report contained insufficient information to make a determination as to the role, if any, that the work zone played in the crash. Next, crashes were categorized based on their crash cause. The crash-causal categories were based on trends observed within the data while determining the crash’s relationship to the work zone. For instance, crashes which involved stopping or slowing due to congestion created by the work zone were broken out as a category within the directly related data. The most common crash-causing categories identified in the dataset were:  Stopping/ slowing due to work zone (including congestion and merging)  Stopping/ slowing due to flagman, police officer, or work zone sign  Changing lanes due to work zone lane closure or congestion  Confusion due to work zone traffic control  Limited sight distance due to work zone activities  Work zone vehicle entering/exiting work zone  Unauthorized work zone entry  Avoiding crash with another vehicle or object  Lost control, struck work zone device/barrier  Unknown, but work zone device/barrier was struck Other crash-cause categories that were identified but did not represent a significant percentage of the work zone crashes included: vehicle backing up, falling asleep/fatigued, improper work zone instruction (such as improper flagger operations), driver inattention, reckless driving, and uneven pavement. Of the 6774 total work zone crashes occurring in the two-year study period, 6424 crashes belonged to the four primary crash types (rear-end, angle, sideswipe and fixed object- off road). Of these, 1480 (23%) were found to be directly influenced by the presence of a work zone (directly-related crashes), and these were examined in greater detail via crash diagrams and narratives provided by the investigating officer in order to develop common work zone features contributing to the crashes. It is still possible that the work zones may have contributed to some aspect of the remaining 77% of the work zone crashes, but insufficient information was available to make a direct determination. Results of Analysis The three most common crash-cause categories for each of four major crash types are displayed in Table 1. Each of these primary categories are described in greater detail in the sections that follow. Table 1 shows the number of crashes in each crash type that were determined to be directly related to the work zone using the criteria specified earlier. The percentage of crashes within each crash type that were determined to be directly related to the work zone are also shown (e.g., 18.1% of all recorded work zone rear-end crashes were determined to be directly related to the work zone based on the available data).

NCHRP Project 17-61 9 Table 1. Summary of Crash Causes Directly Related to Work Zone Presence Crash Type # of Directly Related Crashes % of Coded Crash Typea 1st Most Common Cause 2nd Most Common Cause 3rd Most Common Cause Rear-end 679 18.1% Stopping/ slowing due to work zone presence (64.7%) Stopping/ slowing for flagman, police officer, or work zone TTC (11.8%) Changing lanes due to work zone TTC (8.5%) Angle 160 15.1% Changing lanes due to work zone TTC (22.5%) Confusion due to work zone traffic control (21.3%) Limited sight distance due to a work zone feature (11.9%) Sideswipe – Same Direction 170 18.8% Changing lanes due to the work zone (62.4%) Work zone vehicle entering/exiting work zone activity area (5.3%) Recovery from vehicles intruding into the work zone activity area (4.1%) Fixed Object – Off Road 471 66.2% Unknown, but work zone device/ barrier was struck (37.4%) Lost control, struck work zone device/ barrier (23.4%) Avoiding crash with another vehicle or object (14.6%) a The percentage of crashes within each crash type that were determined to be directly related to the work zone The following sections describe the specific circumstances and characteristics of the directly related crashes for each broad crash cause of each crash type shown in Table 1. Rear-end Crashes The most common crash-cause categories for directly-related rear-end crashes in Table 1 pertain to activities and events that require velocity and/or lane changes. It was determined that a large proportion of rear-end crashes occurred at lane closures, indicating that congestion created by vehicles merging at a work zone lane closure is a large contributor to increased rear-end crashes. Collectively, the three most common rear-end crash-cause categories displayed in Table 1 described 85% of the directly-related rear- end crashes in work zones. Definitions and further information about each of those common causes are provided below. Common Cause #1: Stopping/ slowing due to work zone presence This category included 439 crashes (64.7% of directly-related rear-end crashes) resulting from vehicles stopping/slowing due to some aspect of the work zone. This does not include crashes caused by stopping or slowing due to a flagman, police officer, or a work zone sign. Typical events or activities that were found to lead to this type of crash cause were the presence of work zone-induced congestion, merging situations, unexpected hazards in the travel lane, and confusion about the proper driving action to take. Congestion. For rear-end collisions, 296 of the 439 crashes in this crash-cause category (65.1%) were related to congestion at the work zone. Only crashes that were attributed to abnormal congestion were included in this category. Crashes attributed to recurring congestion, such as congestion occurring during

NCHRP Project 17-61 10 peak periods, were not included. The majority of these crashes (73.0%) occurred when a lane closure work zone was present, but were not directly attributable to merging actions. Merging. Fifty of the 439 rear-end crashes in this category occurred while a vehicle was stopping or slowing due to merging maneuvers. This included crashes involving either:  Congestion caused by merging at a lane closure or crossover (27 crashes, or 54% of the merging rear- end crashes)  A vehicle slowing to allow another vehicle to merge at a lane closure (12 crashes, or 24%)  A vehicle stopped in the closed lane waiting to merge at a lane closure (9 crashes, or 18%)  An existing highway acceleration lane shortened by the work zone (2 crashes, or 4%) Hazard. Twenty-six of the 439 rear-end crashes in this causal category (5.9%) were related to avoiding a work zone hazard in the travel lane. Typical hazards included equipment that had fallen into a travel lane, work zone channelizing devices that had fallen over or otherwise slid into an open lane, bumps, or roadway debris created by work zone activity. Other. Ten of the 439 rear-end crashes in this category (2.3%) were related to apparent driver confusion with the work zone TTC, according to the crash report narratives. Common Cause #2: Stopping/slowing for flagman, police officer, or work zone TTC This category included 80 crashes (11.8% of the directly-related rear-end crashes) occurring when vehicles stopped at the direction of a flagman, police officer, or work zone TTC. Traffic stopping for flaggers was responsible for 73 out of the 80 crashes (91.2%) in this category. It was not possible to ascertain whether driver distraction, poor sight distance to the advance warning devices or to the upstream end of the queue, or other factors contributed to these types of crashes. Common Cause #3: Changing lanes due to work zone TTC This category consisted of 58 crashes (8.5% of directly-related rear-end crashes) involving a vehicle actively in the process of, or just completing, a lane change maneuver due to the work zone TTC. Overall, 58.6% of these 58 crashes occurred in the advance warning area or transition area at a lane closure work zone. Angle Crashes As shown in Table 1, changing lanes is the most common cause for this type of crash. However, confusing work zone TTC or limited sight distance caused by a particular feature of the work zone also appeared to contribute substantially to angle crashes, particularly at intersections. Definitions and further information about the most common angle crash-causal categories are given below. Common Cause #1: Changing lanes due to the work zone This classification involved 36 crashes (22.5% of the directly related angle crashes) where a vehicle was actively in the process of changing lanes due to a work zone. This category includes crashes near the work zone taper where a vehicle must change lanes, discretionary lane changes in the advance warning area, often during congestion, and discretionary lane changes within the work zone activity area. The majority (61.1%) of crashes within this causal category occurred in the advance warning area or transition area of a lane closure work zone. While driver actions were not always documented consistently in the

NCHRP Project 17-61 11 crash reports, it appears that failure to yield on the part of the merging vehicle was a common cause of crashes. Common Cause #2: Confusion due to work zone traffic control A total of 34 crashes (21.3% of the directly-related angle crashes) reflect crash report narratives where driver confusion about some aspect of the work zone was noted. All of these crashes occurred on non- interstate routes. Twelve of these crashes (35% of the 34 crashes) were attributed to confusion over unmarked lanes as a result of repaving or other roadwork. Meanwhile, eight crashes (24%) involved improper work zone instruction by either a flagman or a member of the work crew. Seven crashes (21%) involved a change in traffic patterns, causing the driver to be unaware that a lane or street was closed. In cases when a turn lane was closed, the driver was typically confused about if and how to properly execute the turn. Common Cause #3: Limited sight distance due to a work zone feature A total of 19 crashes (11.9% of directly-related angle crashes) were attributed to a decrease in sight distance or visibility due to a work zone obstruction such as construction equipment, construction vehicles (both parked and not parked), and/or TTC signs. Nine of these crashes (47%) occurred at a minor road stop-controlled intersection. In all of these cases, the stopped minor road vehicle could not see around the work zone obstruction and was hit by a major road vehicle. Six of the crashes (32%) were left- turn maneuvers that occurred at a four-lane signalized intersection where both parties had a green ball on the opposite approach, but the turning vehicle did not yield right of way because the driver’s sight was obscured by some feature of the work zone. Sideswipe - Same Direction Similar to rear-end and angle crashes, 55.3% of directly-related same-direction sideswipe crashes occurred at a lane closure work zone, and another 20.6% of these types of crashes occurred in a crossover or short-duration or mobile work zone. In addition, work zone entry and exit of both authorized work zone vehicles and unauthorized vehicles were two significant causes for this crash type. Additional details regarding these crash-causal categories are given below. Changing lanes due to the work zone A total of 106 crashes (62.4% of the directly-related sideswipe crashes) involved a vehicle actively in the process of changing lanes within a work zone. This category includes sideswipe crashes that occurred in the advance warning and transition area of a lane closure, as well as sideswipe crashes within the work zone activity area where multiple lanes were available. About 54% of these crashes occurred in the advance warning area or transition area of a lane closure. Unfortunately, the narratives and diagrams were usually not sufficient to know for certain whether the sideswipe crashes occurred under uncongested or congested road conditions. It was also not possible to assess whether these crashes were due to a failure to detect adjacent vehicles in blind spots, or whether they reflected aggressive driving behaviors where one driver tried to force their way into a lane and the following vehicle attempted to keep the space in front of them too small for the merging vehicle. When multiple lanes were present in the activity area, a number of cases indicated that a vehicle traveling adjacent to barrier encroached on the adjacent lane Work zone vehicle entering/exiting the work space This category included nine crashes (5.3% of directly-related sideswipe crashes) involving work vehicles that were entering or exiting the travel lanes to or from work space. Eight of these nine sideswipe crashes occurred on interstates. This result suggests that work zone vehicles are finding it more difficult to merge into and out of interstate work zones than other facility types, possibly due to higher speeds.

NCHRP Project 17-61 12 Recovery from intrusions into the work zone activity area This casual category included seven crashes (4.1% of directly-related sideswipe crashes) involving non-work vehicles unintentionally entering a work zone activity area, and then sideswiping vehicles in the adjacent travel lane as they attempted to recover back to the travel lane. The vehicle also typically struck cones and/or barrels during the vehicle’s intrusion into or recovery from the activity area. Fixed Object Run-Off-Road Crashes This crash type exhibited the largest percentage of coded crashes determined to be directly-related work zone crashes in Virginia (66.2% as shown in Table 1). This is likely due to how directly-related work zone crashes were defined in this analysis. Any crash where a vehicle hit a work zone object was designated as a directly related crash. Unfortunately, many of the crash reports in this category only contained vague crash descriptions. Therefore, the most common crash causes for directly-related fixed object run-off-road crashes were “Unknown” or “Lost Control” (usually due to an unknown reason). The analysis showed that 74.3% of interstate fixed object run–off-road road crashes (equal to 66.8% of these crashes on all facility types) occurred during a shoulder or median work zone. This is contrary to the other crash types which experienced the largest amount of crashes in lane closure work zones. Further information about the most common fixed object run–off-road road crash-causal categories are given below. Unknown, but work zone device/ barrier was struck This category included 176 crashes (37.4% of directly-related fixed object run-off-road crashes) involving either a collision with a work zone object or crashes that were determined to have possible connections to the work zone. However, a specific crash cause could not be determined from the crash report’s narrative. Lost control, struck work zone device/ barrier This category included crashes attributed to the driver losing control of the vehicle and striking a device or barrier. Overall, this category represented 110 crashes (23.4% of the directly-related fixed object run-off-road crashes). Unfortunately, for 99 of the 110 crashes in this category, the reason for the driver’s loss of control was not included in the crash report narrative. With regards to the specific items struck, 64 crashes (58.2% of the 110 crashes) occurred with concrete barrier, 29 crashes ((26.5% of the 110 crashes) occurred with crash cushions, and seven crashes (5.9% of the 110 crashes) involved impacts with a traffic control device. Avoiding a crash with another vehicle or object This category included 69 crashes (14.6% of directly-related fixed object run-off-road crashes) that were attributed to the driver swerving off the road in an effort to avoid a collision with another vehicle or object. Of these 69 crashes, 37 (53.6%) appeared to be caused by the vehicle trying to avoid what would have been a sideswipe crash, while 26 (37.7%) were the result of the vehicle running off the road to avoid a rear-end collision. Collisions with Work Zone Vehicles Work zone vehicles were the second most common object hit in the directly-related rear-end, angle, and sideswipe crashes (the first most common object hit was another vehicle in transport). Work zone vehicles were the third most common object hit across all the crash types, involved in 10.1% (149 crashes) of all directly-related crashes. Therefore, crashes involving work zone vehicles were further investigated. Table 2 indicates the location or action of the work zone vehicle during those crashes.

NCHRP Project 17-61 13 Crash protection/shadow vehicles at 1) lane closures or 2) mobile work zones were the most common type of work vehicle struck. These 57 crashes (which represented 38.3% of the work vehicle crashes) involved a vehicle fulfilling its function by protecting the work crews as a shadow vehicle (with or without a truck-mounted attenuator). The second most common situation where a work vehicle was struck was when a non-work zone vehicle intruded into the work space and struck a work vehicle, accounting for 37 crashes (24.8% of the work vehicle crashes). The third most common case involved work vehicles entering or leaving the work space, resulting in 27 crashes (18.1%) of the work vehicle-involved crashes. Table 2. Location/Action of Work Zone Vehicles Involved in Crashes Location/Action of Work Zone Vehicle Number of Crashes Involving a Work Zone Vehicle Percentage of Crashes Involving a Work Zone Vehicle Work zone work vehicle struck by non-work zone vehicle intruding into the work zone activity area 37 24.8% Shadow vehicle in work zone transition area struck by non-work zone vehicle 35 23.5% Work zone vehicle struck while entering/exiting work zone activity area 27 18.1% Shadow vehicle struck in mobile work zone 22 14.8% Improper use/placement of vehicle or equipment 12 8.1% Work zone vehicle struck while in transit within work zone 8 5.4% Work zone vehicle hit another work zone vehicle or pedestrian within work zone activity area 8 5.4% Summary of Findings of VDOT Crash Analysis Table 3 highlights seven crash-causal trends that arose out of this analysis, and illustrate high-leverage opportunities for improving work zone safety nationally. These crash causes are summed across crash types in Table 3, and include crash causes that were not in the top 3 categories shown in Table 1. As a result some crash counts in Table 3 exceed what is shown in Table 1 if a characteristic was not in the top 3 presented earlier. Table 3. Summary of Significant Crash Causes across Crash Types Crash Cause Number of Crashes Directly Related to Work Zone Percentage of All Crashes Directly Related to Work Zone Stopping/slowing due to congestion 356 24.1% Changing lanes due to work zone 229 15.5% Involved work zone vehicle 149 10.1% Work zone vehicle entering/exiting work zone activity area 33 2.2% Involved a flagman 73 4.9% Confusion due to work zone traffic control 44 3.0% Limited sight distance due to work zone 19 1.3%

NCHRP Project 17-61 14  Overall, 24.1% of the directly related crashes occurred during work zone congestion. Of these, 73.3% (261 crashes) occurred on an interstate, with 57.0% (203 crashes) occurring during an interstate lane closure.  The analysis also indicates that 15.5% of all directly related crashes were associated with lane changes at a work zone. Logically, lane change and congestion related crashes are closely linked. Congestion significantly shrinks the number of gaps large enough to allow effective merging.  The analysis also found that 10.8% of directly related rear-end crashes occurred at work zones where flagger control existed. Unfortunately, the available data was not sufficient to assess whether improper advance signing, poor work zone location just beyond horizontal or vertical curvature, or improper flagging operations played a role in the crashes.  Work zone vehicles entering/exiting the work zone activity area were also found to be an issue in this assessment. The analysis found that 57.6% (19 crashes) of crashes involving work zone vehicles entering/exiting crashes involved a shoulder or median work zone. This suggests that the work zone vehicle was attempting to merge or exit from a high-speed roadway, and additional attention should be paid to developing traffic control plans for ingress and egress from these sites.  The narratives and diagrams pertaining to directly related crashes at work zones on primary/secondary roads suggest that driver confusion was more of an issue on these types of roadways than on interstates or freeways. This suggests that traffic control being utilized on these facilities may not be adequate in many cases to serve motorist positive guidance needs. It should be noted that these facilities typically have more access points than freeways, and navigation demands on drivers are often more difficult.  Limited sight distance due to poor placement of work vehicles and equipment at intersections was noted as an issue for angle crashes. Changes in agency traffic control requirements that currently require closing lanes may need to be revised when the lane is a turning lane to better avoid creating sight distance challenges at intersections and driveways. Analyses of the NMVCCS and LTCCS Databases The other datasets examined for insights into work zone crashes were the NMVCCS and the LTCCS databases. In contrast to the VDOT police crash database, these datasets contained data specifically focused on developing crash causation and often had much more detailed documentation about site conditions. The NMVCCS database includes detailed information for crashes involving at least one light vehicle (gross vehicle weight less than 10,000 pounds) that was towed due to damage. Data were collected on-scene by trained staff as quickly as possible after the crash occurred and through detailed follow-up investigations and interviews. The NMVCCS investigated a total of contains 6,949 crashes that occurred between January 1, 2005 and December 31, 2007 collected at locations covering wide geographic regions and urbanization types in the U.S. Similarly, the LTCCS database includes information for 1,070 crashes involving at least one large truck with a gross vehicle weight of more than 10,000 pounds and resulting in one or more fatalities or injuries. The crashes in the database occurred in 17 states from April 2001 through December 2003. Up to 1,000 crash-related variables are available for each crash. The data were collected on-scene and through follow- up efforts by trained staff from NHTSA’s National Automotive Sampling System (NASS) and State truck inspectors. The dataset includes variables indicating if the crash site was inside of a construction work zone, if the sequence of events involved work zone equipment, and if the crash involved a barrier.

NCHRP Project 17-61 15 NMVCCS Analysis Methodology NMVCCS records contain more than 600 variables that capture details related to the driver, vehicle, roadway, and the environment, including crash narratives, photographs, and diagrams collected at the crash scene. The NMVCCS database includes variables indicating whether there was a work zone present at the crash site and whether traffic barrier was present. For this analysis, there was particular interest in the events and factors that led to a crash, the roadway configuration in the work zone, and other significant factors, such as special weather conditions. Unfortunately, variables in the NMVCCS records do not include descriptions of the work zone signs, tapers, or special configurations at the time of the crash beyond what is observed from the photos and crash diagrams. Also, the survey does not contain information about the specific geographical location of crashes, making further investigation of details a difficult task. Exploration of the data identified 178 crash records of events that occurred at or near work zones. These crashes were initially identified by filtering for traffic flow restrictions or interruptions due to work zone presence at the time of the crash. These crashes were segregated according to roadway type on which they occurred, and then analyzed focusing on identifying common crash “scenarios” that occurred in work zones on these types of facilities, resulting on an initial classification. Then, crashes were re- classified taking into account if they were directly related to a work zone. Crashes were classified as directly related to a work zone if at least one vehicle interacted directly with work zone elements, such as temporary traffic control devices, modified geometry or physical roadway features, such as lane shifts or temporary lane drop-offs, or congested traffic conditions generated in part due to the presence of the work zone, as described in NMVCCS records. Therefore, for crashes classified as being directly related to a work zone, had the work zone not been in place, such interactions could not have occurred. It is noted that for each subsection, labels common to both initial and re-classified diagrams have been included for the reader to link or track crashes between the two classifications. Results of Analysis Freeway/Interstate Crashes Crash exploration and causal analysis began with crashes that occurred on segments with full access control (i.e., freeways and interstates), as this was expected to yield more uniform conditions than other roadway types and offer a context to further develop and test alternative in-depth crash analysis methodologies. A total of 65 crashes (36.5% of the 178 work zone crash records in the NMVCCS dataset) were coded as occurring at or near freeway work zones. The 65 crashes were examined in order to classify them into categories by crash type and main contributing factors, as shown in Figure 2. Out of the 65 freeway work zone crashes, 10 were not related to vehicle-roadway interactions (e.g., roadway design, surface conditions) or direct vehicle-vehicle interactions. Instead, these crashes were related to external factors such as the physical condition of one or more drivers, the vehicle mechanical condition, or vehicle-debris interactions. The remaining 55 crashes were classified into four main groups:

F igure 2. Initial classification of NMVCCS crashes in or near wo rk zones on freeway/interstate roadways.

NCHRP Project 17-61 17  Roadway departure  Late/forced lane change  Rear-end due to congestion  “Other” situations As expected, the classification with the highest frequency was the rear-end crashes due to congestion (25 crashes, or 45.5% of the 55 crashes). Further breakdown of these four groups is also given in Figure 3 with additional details of situations leading to the crashes. In the roadway departure group there were 12 crashes (21.8%), further divided into four sub-groups. Only one sub-group could be directly related to the presence of a work zone and involved drivers losing control of the vehicle due to a temporary edge drop-off (two crashes, or 3.6%). The remaining three sub- groups were as follows:  Vehicles traveling too fast for rainy/wet conditions (five crashes), with three of these crashes occurring while negotiating a curve  Driver inattention/fatigue (four crashes)  Losing control while traveling near an on-ramp in a work zone (one crash) This last case occurred when a vehicle on an entrance ramp made a sudden lane change, entering the right through lane and sideswiping a through vehicle. It is noted that out of the three cases where drivers lost control of the vehicle while negotiating a curve, two cases were on entrance ramps that were not modified due to the work zone. In the remaining case, a vehicle traveled over standing water on a curved freeway section within the work zone, hydro-planed, and then lost control crossing a gravel median. It was not possible from the available documentation to ascertain whether the work zone somehow contributed to the presence of the standing water. Therefore, the visualization of the relation of crashes and work zones only includes the two crashes associated with the drop-off, as shown in Figure 3 in the category “changes in surface conditions”. In the group of “late/forced lane change” crashes, most occurred at or near exit ramps (10 out of the 12 crashes) when drivers lost control of the vehicle as they attempted to exit the roadway but were too close and/or traveling too fast at or near the exit gore point. Out of the 10 crashes near exit ramps, five of them occurred at locations where work zones were not visible in the crash photos or had no influence on the roadway configuration. The five remaining exit-ramp crashes were considered to be directly related to the work zone presence with shifted lanes (three cases), one of the two ramp lanes temporarily closed (one case), or the exit lane pavement was grooved and uneven (one case). In addition to the 10 crashes at exit ramps, one crash at an entrance ramp involved a truck causing other vehicles to depart the roadway. This occurred as the truck changed from the right to the left lane in anticipation of an entrance ramp while traveling on a two-lane freeway section upstream of any influence from the work zone. Lastly, the twelfth crash involved a vehicle making a sudden lane change just ahead of a lane drop and after a curved segment, causing others to suddenly decelerate and resulting in a rear-end crash. This crash was considered work zone related, but based on the crash report the main critical reason was driver distraction/inattention even though the drivers from the two vehicles stated that advance signs were not placed at a sufficient distance from the taper for the driver to properly react.

F igure 3. NMVCCS crashes in or near work zones on freeway/in terstate roadways based on direct relation to the work zone.

NCHRP Project 17-61 19 Thus, a re-classification of the 12 crashes in this group, this time based on cases where work zones could have had an influence, results in six crashes being work zone related and another six grouped as having no clear work zone influence. Out of the six cases classified as work zone related, in four occasions the right lane leading to the off-ramp was shifted and had temporary concrete barriers (in one case the right lane was closed with cones), in one case the pavement on the right lane near an off-ramp was grooved, and in one case a late merge in advance of a lane drop caused a rear-end crash. These crashes are shown in Figure 3 in the categories “Near off ramp” and “Near Taper / Lane Drop.” For the group of rear-end crashes due to congestion shown in Figure 3, in at least 19 of the 25 cases (76%), the crash report explicitly mentions that congestion was influenced by the work zone presence. In the remaining 6 cases, congestion was present and a connection between congestion and work zone presence was not explicit it but could be inferred from the report (the work zone was in all cases cited as a traffic flow interruption factor). A further breakdown of these crashes indicated that 20 of the 25 rear-end crashes (80%) occurred on segments without access points, where the drivers were distracted or failed to recognize slowing down/stopping traffic. Out of the five remaining crashes, four of them were related to yield maneuvers and inattentive drivers during stop-and-go conditions at locations adjacent to on-ramps. The fifth on-ramp crash occurred as a vehicle forcedly changed lanes from the ramp into the right lane and caused a rear-end crash but it was not related to stop-and-go conditions. Out of the five on-ramp crashes, two of them were in a work zone where the ramp was modified, and two others occurred where one of the main lanes was closed for construction. However, from the report pictures it seemed that modifications to the ramps in these two cases were not significant enough to cause significant challenges for drivers, and sudden speed reduction combined with driver inattention had more important roles in the crashes. Lastly, six of the remaining crashes did not fit the previous classifications and were grouped in a classification called “other situations”, where only one of them may have been directly influenced by the nearby work zone. In this case, a vehicle became disabled in the right-most lane at a location without enough shoulder width for a driver to pull over (the shoulder was not available due to a work zone barrier). This crash was classified as related to “Changes in lane configuration” in Figure 3. In summary, the 65 freeway crashes were re-classified based on the direct influence of work zones, as shown in Figure 3. A total of 31 crashes (48%) were likely not influenced by the work zone and 34 (52%) could be related to work zone presence in one of three ways:  Rapid decrease in operating speeds or congestion due to the construction activities (74% of the 34 work zone-related crashes)  Changes in lane configuration (21%)  Changes in the roadway surface conditions (5%) Based on the diagrams and photos from the crash records, it is noted that even though the 34 work zone related crashes did not necessarily occur on segments where the work space was located, the sequence of events leading to the crash was always initiated within the area of influence of a work zone, including the advance warning area. Non-Freeway/Interstate Crashes Crashes in non-freeway segments were identified by subtracting the freeway crashes (65) from the total work zone crashes initially found (178) in the NMVCCS records, for a total of 113 crashes. After eliminating a small number of records because some crashes were not at/near a roadway work zone (e.g., a building construction or no indication of a work zone), a total of 106 crashes on non-freeway segments remained for further examination.

NCHRP Project 17-61 20 An initial classification of the crashes is shown in Figure 4. Similar to freeway crashes, those not related to vehicle-roadway interactions (e.g., roadway design, surface conditions) or direct vehicle-vehicle interactions were identified first, resulting in 21 out of the 106 records. Further classification of the 21 crashes showed that they were related to driver physical condition, vehicle mechanical condition, and other factors such as reckless/aggressive driving, and an animal-related evasive maneuver that resulted in a single-vehicle crash. Next, the 85 remaining crashes were examined to determine whether they were located at/near intersections. It was found that most non-freeway crashes were at intersections, with a total of 47 records. Even though the sample size in this evaluation is small, the ratio of intersection-related crashes to all sampled crashes was about 55% (47/55), which is not far from a national average of about 40% (Choi, Eun-Ha, 2010). Out of the 47 intersection-related crashes, 23 (49%) occurred at signalized and 24 (51%) at non-signalized intersections (stop-controlled and non-controlled). Further examination of the records showed whether at least one of the involved vehicles was turning left or right, or if all vehicles were traveling through the intersection. At signalized intersections, 10 crashes (43%) involved through vehicles only, eight of which had vehicles entering the intersection without the right of way (i.e., after the signal indication had changed to red) and the remaining two were rear-end crashes near the stop bar. All the 10 crashes with only through vehicles, however, were mainly attributed to driver performance errors such as inadequate surveillance and distraction, without clear influence of work zone elements. The remaining 13 crashes at signalized intersections (57%) involved turning vehicles. A total of 12 out of those 13 crashes (92%) involved left-turning vehicles, and eight of those were associated with driver performance errors similar to those mentioned above for crashes with through vehicles only. However, in the remaining four crashes with left-turning vehicles, the presence of the work zone could have influenced the crash. In two cases, the left-turn lanes were closed and vehicles had to turn from through lanes. In one other case, left turns were prohibited but a vehicle proceeded to turn anyway. In still one final case, the through lanes were closed and vehicles were required to turn left. In each of these four cases, drivers turned left without having the right of way. The reports mentioned driver confusion and/or misjudgment related to the modified intersection layout, implying that it was more difficult for drivers to adequately judge acceptable gaps in the opposing traffic stream. One other crash involved a right-turning vehicle, but appeared to result from a driver misjudging the gap with the vehicle in front while turning, and so was not considered to be work zone related. At unsignalized intersections and non-controlled entryways the proportion of crashes involving through and turning vehicles followed a similar trend. Out of 24 crashes, 12 (50%) involved a left-turn vehicle, two (8%) involved a right-turning vehicle, and remaining 10 (42%) involved only through vehicles. All 10 crashes with only through vehicles and the two crashes involving right-turning vehicles were not directly associated with work zone elements or characteristics, but rather to driver-related performance error elements. On the other hand, out of the 12 crashes involving left turns, four crashes (33%) were associated with driver performance errors but eight (67%) appear to have a more direct relation with work zone elements. These eight crashes include: five cases (63%) where the field of view for drivers turning left could have been reduced by construction equipment, temporary signs, or construction-related congestion; two cases (25%) where the work zone configuration and signs contributed to driver confusion resulting in vehicles turning left while oncoming traffic was approaching; and one case (13%) of a last- second decision to turn left and divert to avoid construction-related congestion that resulted in a crash with opposite-direction traffic. Crashes that did not occur at intersections (38) were also classified into three specific sub-groups and one general sub-group that contained unique or less common cases. The first sub-group included rear-end collisions (18 crashes, or 47% of the non-intersection crashes). Of these, most (12 crashes, or 67% of the non-intersection rear-end crashes) occurred at the back of queue due to congestion, two other crashes (11%) occurred at/near a flagger directing traffic, and a few cases (22%) that involved working trucks stopped in a traveled lane and participating in construction activities.

F igure 4. Initial classification of NMVCCS crashes in or near wo rk zones on non-freeway/interstate roadways.

NCHRP Project 17-61 22 The second sub-group included roadway departure crashes (eight cases), where driver performance errors including inattention, health issues (two cases), and driver age (one case involving a young driver and one case involving an older driver) were contributing factors along with the construction zone being present. The crash involving a younger driver (18 years old) was speed-related, and the crash involving an older driver (79 years old) occurred at a location where the driver intended to turn left to the northbound lanes (which were closed) and had to use one of the two southbound lanes that was opened for this purpose (a temporary two-lane two-way operation). The driver became confused and in an attempt to U- turn and head back, he crashed into a vehicle on the southbound lane. The third sub-group included four crashes with vehicles attempting late or forced lane changes at lane drops (three cases) and a construction vehicle re- entering the main travel lanes. These crashes were directly related to the presence of a work zone and also to driver distraction or inadequate action to change lanes. Similar to freeway crashes, the potential effect of construction zones in the occurrence of each crash was also examined and resulted in the diagram shown in Figure 5. Out of the 106 crashes, in most cases (63 crashes, or 59%) the construction zone did not appear to have a clear role in the sequence of events that led to the crashes. The remaining 43 crashes (41%) could be related to work zones and were sub- grouped into the following categories (from largest to smallest):  Crashes that occurred as a consequence of speed drop due to construction zone congestion, as indicated in the NMVCCS records (18 crashes, or 42% of the work zone-related crashes)  Crashes that involved left-turning maneuvers (12 crashes, or 29%)  Roadway departure due to poor driver performance and distraction (five crashes, or 12%)  Late/forced lane changes (four crashes, or 10%) at lane reductions or due to construction vehicles re- entering to the main lanes  Other situations (two crashes [5%]) LTCCS Analysis Methodology In addition to work zone crashes in the NMVCCS, work zone crashes in the LTCCS were also re- examined to determine if specific sequences of events were predominant in the crash descriptions. Initial screening of the records in the LTCCS showed a total of 77 crashes where a construction zone and/or a flow restriction associated with a work zone was present. Out of these 77 crashes, a total of 55 were identified on segments belonging to a rural or an urban freeway or interstate with full access control. The other 21 out of 22 remaining crashes occurred at facilities not classified as being part of a freeway section, and one crash was found not to be in close proximity or influenced by a work zone, and thus not considered further. The analysis of the freeway and non-freeway crashes from the LTCCS is described as follows.

F igure 5. NMVCCS crashes in or near work zones on non-freew ay roadways based on potential influence of the work zone.

NCHRP Project 17-61 24 Results of Analysis Freeway/Interstate Crashes The re-examination of the 55 crashes on urban and rural freeways resulted in the groupings by type of crash shown in Figure 6. Only one crash was not related to vehicle-roadway interactions (e.g., roadway design, surface conditions) or direct vehicle-vehicle interactions and it was attributed to brake failure. From the 54 remaining cases, 35 (65%) were rear-end crashes and the majority of those (26 cases) occurred at or near the back of queue when drivers did not react on time to avoid making contact with stopped or slow moving vehicles. All 26 crashes at the back of queue were considered work zone related, as the congested conditions were generated or intensified by the presence of construction. It is also noted that out of the remaining rear-end crashes, four cases involved vehicles re-entering the traveled lanes from the construction area. A closer examination of these records showed that three cases could be attributed to a driver decision error rather than a deficiency in the design of the construction area. However, one of these crashes could be related to the reduced field of view experienced by a truck driver entering the traveled lanes from a left-side (median side) work zone at the end of a section curved to the right. The truck driver expressed difficulty spotting oncoming traffic from that position, and made the decision to proceed resulting in the truck being impacted by another truck traveling in the left-most lane. Other rear-end crashes included a daytime crash involving a trailer truck that impacted a slow moving (five mph) attenuator truck part of a moving lane closure on lane one out of five lanes (considered work zone related), two cases of disabled vehicles, one on the left lane without emergency shoulder (considered work zone related) and one outside of the traveled lanes on the right shoulder (not work related), one case of a driver over-reacting to barrels misaligned on a bridge (considered work zone related), and one case of an illegal maneuver when a driver aborted using an off-ramp and suddenly returned to the traveled lanes using the gore area (considered work related because the left lane of the two-lane freeway was closed). One more category included seven crashes where vehicles performed a late/forced lane change. Three of these crashes occurred near the beginning of the taper closing a lane as drivers attempted to change lanes. However, even though the crashes were work zone related, from these crash records there was no indication of deficient work zone design or layout and crashes were mainly associated with driver performance or recognition errors. Similar conclusions were also reached from the remaining four late/forced lane change crashes, two of which occurred near the back of queue at congested construction areas, one near an off-ramp that was not modified by the work zone when a driver tried to change lanes from the right (exit) lane and hit an adjacent vehicle, and one rear-end crash as a driver initiated a late diverge maneuver to take an off-ramp. Regarding the five roadway departure crashes, only two cases were directly related to temporary roadway characteristics. In other instances, truck drivers traveling on the right lane lost control of the vehicle when their tires dropped off the edge of the travel lanes because the shoulder had been converted to a temporary travel lane. The other roadway departure crashes involved drivers traveling too fast to negotiate an off-ramp and a curved segment near an interchange, both of which were not modified by the construction.

F igure 6. Initial classification of LTCCS crashes in or near work zones on freeway/interstate roadways.

NCHRP Project 17-61 26 In a separate category there were seven sideswipe crashes involving at least one truck. In five cases, the lane configuration was modified by construction, and for this reason were considered related to the work zone presence. There was one crash involving a bicyclist traveling unlawfully on the freeway involved in a crash, and another crash that appeared to be due to driver error in detecting a vehicle in an adjacent lane during a lane change maneuver on a tangent section unmodified by the work zone, and so both of these were not considered to be directly related to the work zone. In four of the seven cases, the traffic lanes were shifted, with two of the four crashes on curved segments. In one case the driver failed to see an adjacent vehicle and attempted to change lanes, and in the other case the driver was checking his mirror and the truck drifted out of the lane while negotiating the curve, sideswiping a vehicle. The two other crashes on straight segments involved driver inability to control the truck and jackknifed or drifted onto the adjacent lane because of internal distraction (lighting up a cigarette). The last sideswipe crash occurred at the merge point (on ramp) between two interstates, where the right lane of one interstate and the left lane of another interstate merged into one lane. During construction, merging vehicles had to use the left lane (the right lane was closed with barrels) and the open lane was controlled by a temporary yield sign. The crash occurred as a truck traveling on the right lane and a vehicle merging from the only open lane sideswiped at the merge point. A summary of the crashes classified as directly related to the work zone or not is shown in Figure 7. Based on the crash descriptions, diagrams, and photos most work zone crashes from the LTCCS were likely affected by the presence of a construction zone itself, with 47 of the 55 crashes in this sub-group. It is noted that direct comparisons of the crash breakdown between the LTCCS and the NMVCCS should be avoided since they were not conducted following the same guidelines and sampling methods. Most of the crashes where construction could have affected the crash were related to sudden speed drops and congestion (28 crashes), and all of them except one occurred at or near the back of queue in congested conditions. Also, 17 crashes occurred at locations with changes in lane configurations, four of them near ramps, four involving construction vehicles re-entering the travel lanes, three near the beginning of a taper, four on segments with lanes shifted, one with a disabled vehicle on a section without shoulders, and one involving a driver overreaction to barrels near the traveled lane along a bridge. Lastly, two crashes were in a separate category where trucks lost control due to temporary lane drop-offs.

Figure 7. LTCCS crashes in or near work zones on freeways/interstates based on direct relation to the work zone.

NCHRP Project 17-61 28 Non-Freeway/Interstate Crashes A total of 21 work zone crashes from the LTCCS records were identified and analyzed at facilities without full-access control. Following a similar format used for the freeway/interstate crashes, non- freeway/interstate crashes were first classified into vehicle-roadway or vehicle-vehicle interaction categories, as shown in Figure 8. A total of eight crashes (38%) were not related to vehicle-roadway or vehicle-vehicle interaction but rather to the driver physical condition and/or the vehicle mechanical condition, and two of those were very specific situations where construction equipment hit workers inside the work space (and thus did not involve public road users). The remaining 13 crashes were classified as follows: rear-end, roadway departure, at intersections, and sideswipe. As expected, the most common category was rear-end crashes, with a total of six crashes (46%). Three of those cases involved trucks rear-ending a vehicle in congested conditions, two cases involving trucks rear-ending a vehicle at locations where a flagger was directing traffic (one-lane two- way operation), and one case where a vehicle rear-ended a truck stopped in traffic. One of the two cases with a flagger directing traffic was attributed to driver distraction, but in the other case the report explicitly notes the lack of warning signs in advance of a curve that preceded the work zone. Moreover, this crash was a secondary crash following another rear-end collision at that location. The report also mentions that advance warning signs were added after this crash occurred to help prevent future incidents. It is also noted that the four rear-end crashes not involving a flagger were directly related to driver performance factors such as distraction or poor directional control. Meanwhile, three roadway departure crashes were also identified, each with a distinctive characteristic. Whereas in two cases the driver lost control of the vehicle at a curved segment, in one occasion a driver overcompensated as he drifted to the right and almost hit a temporary concrete barrier next to the edge of the right lane, and in the other case the driver lost control of the vehicle at a section with uneven pavement due to resurfacing. The third case involved a head-on collision as an automobile departed the lane on a two-lane two-way roadway, crossed the double yellow line, and collided with an oncoming truck. It is noted that in normal conditions (no work zone) this was a four-lane divided roadway. The third category included three sideswipe crashes at intersections, two of the crashes at unsignalized locations and one at a signalized location. The crash at the signalized location occurred at a four-lane undivided roadway, was categorized in the report as a driver decision error, and involved a left-turning truck not yielding to oncoming traffic. This work zone was not considered to have influenced the crash. At the unsignalized intersections, one of the crashes involved a driving error by an older driver (82 years old) turning left from a four-lane divided road with channelizing drums onto an uncontrolled access road and being impacted by opposing through traffic. The second crash involved an illegal maneuver by the driver not stopping and yielding to crossing traffic. Even though the driver stated that barrels blocked the stop signs, from the diagram and photos it does appear that one of the stop signs was clear from any obstruction. Whether the driver’s eye height or location relative to the barrels actually obscured the stop sign could not be determined from the available crash data. Lastly, an additional crash was identified in the sideswipe category, as a distracted driver could not react on time to avoid hitting stopped vehicles, changed lanes to avoid a rear-end crash, entered opposite direction travel lanes, and was sideswiped by an oncoming truck. A re-classification of these non-freeway crashes based on the direct relation to the work zone is shown in Figure 9. As expected, the largest group was crashes related to speed drop/congestion (seven crashes), followed by changes in lane configuration (two crashes), changes in surface conditions (one crash), and at an intersection (one crash). It is also noted that a final category (grayed out in Figure 9) includes the two incidents where workers were hit by construction equipment inside the work space and did not involve public road users.

F igure 8. Initial classification of LTCCS crashes in or near work zones on non-freeway/interstate roadways.

Figure 9. LTCCS crashes in or near work zones on non-freeway/interstate roadways based on potential influence of the work zone.

NCHRP Project 17-61 31 Summary of Findings from the NMVCCS and LTCCS Analyses A breakdown of the freeway and non-freeway crashes grouped by main categories and subcategories for each of the two databases is shown in Table 4. Overall, back of queue rear-end crashes related to speed drop/congestion were by far the most common type and represented about 48% of all work zone related crashes (64 of the 133 crashes). Rear-end crashes at the back of queue were also the single largest contributor for both freeway and non-freeway categories in both NMVCCS and LTCCS databases. Congestion due to construction activities and/or roadway changes were in most cases explicitly mentioned as contributing factors for the rear-end crashes. Additional major contributing factors were mostly driver-related, in particular driver distraction, inattention, and failing to recognize stopped or decelerating traffic. The second largest category was vehicle crashes at or near freeway off-ramps (eight crashes). In these cases, access to off-ramps was modified within the construction zone in a variety of ways: reduced number of deceleration lanes, length of deceleration lane, or temporary barriers and other roadway hardware. In general, these crashes were coded as involving driver performance errors when trying to exit (or trying to avoid exiting) and while traveling too fast or attempting a late lane changing maneuver. Similar to rear-end crashes, the work zone presence was relevant because it modified the off-ramp, increased congestion, and narrowed the available width for drivers to maneuver. The next largest category included crashes at or near a taper or a lane drop (seven crashes). These crashes occurred on both freeway and non-freeway facilities, and involved both passenger vehicles as well as trucks. All of these cases were associated with the late recognition or a late maneuver in advance of the lane reduction, and most of these crashes resulted in rear-end crashes not due to congestion but to the late maneuver. Another significant category observed at non-freeway facilities involved six crashes that were associated with vehicles turning left onto minor roads that were not controlled from the main road. These maneuvers required drivers to judge available gaps in oncoming traffic before proceeding. Based on the analysis of available data, it is possible that these crashes were influenced by limited driver visibility due to work zone hardware or equipment. Examples of such hardware/equipment included construction equipment in the left-turn lane of opposing traffic, vertical panels dividing traffic on a two- lane, two-way road, and congestion in one lane blocking the view of traffic approaching in adjacent lanes. Five crashes involving construction vehicles re-entering the travel lanes comprised the fifth most common category. Speed differentials between main lane traffic and trucks entering from the work space were a significant factor in these cases, with vehicles rear-ending the construction trucks. In one instance in particular, a truck was rear-ended as it re-entered from the median lane on a segment curved to the right. From this angle, the truck driver stated that he had reduced visibility of oncoming traffic, which contributed to the crash occurrence.  

Table 4. Summary of NMVCCS and LTCC Crashes where the Work Zone Influence was Likely Category Sub-category NMVCCS LTCCS Total Freeway Non-freeway Freeway Non- freeway Speed drop/ congestion At back of queue (driver distraction / inattentive / failed to recognize stopping traffic) 20 12 27 5 64 At/near on-ramp 5 5 Moving lane closure 1 1 At/near flagger 2 2 4 Working vehicle stopped in lane 3 3 Disabled vehicle in lane - no shoulder 1 1 1 3 Changes in lane configuration Near off-ramp (modified ramp access) 5 3 8 Near on-ramp (on-ramps lanes reduced) 1 1 At lane drop / Near taper 1 3 3 7 Construction vehicle re-entering roadway 1 4 5 Lanes shifted - curved segment 2 2 Lanes shifted - straight segment 2 2 Drifted towards a temporary barrier on a curved segment, then overcompensated 1 1 Divided Highway temporarily operated as undivided (2-lane 2-way) 1 1 Bridge - Driver overreacted to barrels in shoulder 1 1 Changes in surface conditions Temporary lane drop-off 2 2 4 Uneven lanes 1 1 Left-turning vehicles Signalized Intersection 4 4 Minor access or stop controlled minor road view possibly blocked by WZ hardware 5 1 6 Minor access or stop controlled minor road view not blocked by WZ hardware 3 3 Roadway Departure Straight segment - Driver performance error 3 3 Curved segment - Distracted drivers 2 2 Other WZ hardware in open lane (cone, pipe) 2 2 Total 34 41 47 11 133

NCHRP Project 17-61 33 New Insights Gained Through the Analyses The combined results of the analysis of the three different databases illustrate the significance of work zone-induced congestion upon crash potential. Certainly, work zone congestion is a major contributor to rear-end collisions, as previous research has found or hypothesized. Depending on the database considered, work zone-induced congestion was found to contribute to between 24% and 70% of work zone crashes. The analysis also shows that the presence of congestion may also contribute to sideswipe crashes at entrance ramps and at upstream ends of queues as drivers swerve to avoid running in to the back of the vehicle in front of them. The analyses also suggest that work zone congestion also contributes substantially to roadway departure fixed object crashes with barriers or work zone hazards as drivers swerve to avoid rear-end collisions. Although this issue is most commonly associated with freeway/interstate work zones, it does show up on other roadway types as well. For instance, the analysis of the VDOT crash database found that 11% of the rear-end collisions occurred at work zones where flaggers were being used. In most locations, flaggers are used predominantly on two-lane highways for alternating one-way traffic control or for intermittent stop control to manage construction vehicle access points. Although, typically not considered a “congestion” issue, work zones that periodically require temporary stoppages in traffic flow were found to contribute substantially to work zone rear-end collisions. Analysis of the NMVCCS and LTCCS databases also suggest that ramp junctions at interstate/freeway work zones also create challenges for drivers and contribute to work zone crashes. Often, ramp geometrics must be temporarily degraded (reduced acceleration/deceleration lanes, reduced ramp widths, etc.) in order to accommodate construction activities. Unfortunately, this analysis could not associate the type or magnitude of the ramp degradation with the magnitude of increased crash risk. However, work zone designers should acknowledge in their decisions that temporarily degrading ramp designs will likely increase crashes. Work zone vehicles entering/exiting the work zone activity area were also found to be an issue in this assessment. Both rear-end collisions and sideswipe crashes in the three databases could be attributed to this issue. In the LTCCS database, 10% of the crashes on interstate/freeway facilities appeared to be the result of trucks attempting to enter or exit main lane traffic to or from the work space. Improving work zone ingress/egress should be made a greater priority for agencies and contractors. Finally, the analyses of the three databases highlighted that work zones on non-freeway/interstate facilities are creating challenges for drivers that were previously unknown or underestimated. Data from several crashes examined in these analyses indicated that drivers became confused when approaching and entering work zones on non-accessed controlled facilities at intersections and driveways, especially in urban areas. Work zones on divided facilities that were temporarily converted to two-way operation in one of the directional roadways, while the other direction was repaired or rehabilitated appear to be particularly problematic. Improved training on how to properly design and implement TTC for approaches to these types of work zones may be needed. In addition, sight distance challenges were noted for several crashes occurring at these types of work zones. Obstructions created by the presence of work equipment too close to an intersection or driveway were cited as a contributor in a number of crash narratives, as was the presence of certain temporary traffic control devices. In one instance, type 3 barricades placed in the opposing left-turn lane at a signalized intersection limited the ability left-turning traffic coming from the other direction to see approaching through lane traffic during the permissive green phase. Thus, improved training for designing and implementing TTC at both signalized and unsignalized intersections appear to be needed based on this analysis.

NCHRP Project 17-61 34 The fact that these trends were observed in all three databases, despite the fact that they were created for different purposes and cover very different levels of detail and geography, is particularly noteworthy. Such correlations across the databases adds credibility to the significance of these issues in work zones nationally.