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.
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 8 CHAPTER 2 REVIEW OF WORK ZONE SAFETY LITERATURE The research studies reviewed were published from 1978 to 2004, and specific articles were selected based on their conformance to the scope of this project or for any original contribution to the state of knowledge on work zone crashes. They are summarized in the following sections. 2.1 DATA AND METHODOLOGIES Seventeen of the reviewed studies included a range of data, observational designs, and analysis methodologies used in work zone safety studies. Usually, data were obtained in one of two ways: 1) by using an electronic database or police accident report with work zone tags (7,8,9,16,17,18,19,20) or 2) by matching the date and location of accidents, either in electronic or hard copy form, with the dates and locations of work zones on certain roadways (1,10,11,12,13,14,15,21). A weakness of the former is that, in most cases, the work zone tag was marked only if the reporting officer considered the crash to be related to the presence of the work zone. A similar weakness is associated with the latter option if the researcher selects/de-selects accidents based on a perceived relationship to work zone presence, as in (21). As one research team pointed out, âone cannot ever be sure that an accident was or was not attributable to the presence of constructionâ (1). Nonetheless, the subjective identification of work zone-related crashes was a common technique in the safety literature. In addition to whether the crash occurred in a work zone, the databases or crash records usually included variables such as crash severity, crash type, date, time of day or light condition (day/night), vehicle types, major contributing factors (opinion of reporting officer), weather condition, roadway surface condition, and location within the work zone. Some databases, including the Fatality Analysis Reporting System (FARS) (22), have a generic variable for roadway alignment (presence of curve or presence of grade); however, specific work zone design features (e.g., lane width, shoulder width, radius of curve, etc.) at the location of the crash were not included in any of the data sets reviewed. This has been pointed to as a major cause of the wide range and sometimes conflicting results in the safety literature and as a significant weakness in most work zone safety research (20). The observational designs were also of two predominant types: 1) frequency observations of work zone crashes (7,8,9,16,17,18,19,20,21) or 2) before-during observational studies (1,10,11,12,13,14,15). Analyses of the former included simple observations of frequencies and proportions of different types of work zone crashes with no statistical testing; observations of frequencies and proportions of different types of work zone crashes with statistical testing (usually Ï2 tests of proportions); or comparisons of frequencies and proportions of work zone crashes with crashes outside of work zones, with and without statistical testing. Analyses of the latter included straight before-during
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 9 comparisons of frequencies, proportions, or rates without statistical testing; before-during comparisons of frequencies, proportions, or rates with statistical testing; or before-during comparisons of frequencies, proportions, or rates with statistical testing and use of control/comparison sites. This final type of analysis is recommended for before-during studies (23) and was conducted by (1,13,15). The accuracy and scope of the data sets and the different analysis types led to a wide range of conclusions regarding the magnitudes and characteristics of work zone crashes. As stated earlier, a major contributor to these different, sometimes conflicting, conclusions was the lack of data on specific work zone design features. 2.2 QUANTITATIVE SAFETY EFFECTS OF WORK ZONE DESIGN FEATURES Six studies addressed the safety effects of different design features in a controlled fashion (10,13,14,15,16,17). The features evaluated included work zone length, traffic diversion strategies (e.g., lane closures, median crossovers), and entrance ramps. The results are the focus of this section. 2.2.1 Work Zone Length Three studies developed negative binomial regression models to predict expected accident frequencies on work zone segments (13,14,16). Length as an independent variable was included in the final models of two (14,16). In one model (16), increasing the length of a work zone by 1 percent while keeping all other factors constant led to a 0.85 percent increase in the expected number of injury/fatality crashes and a 1 percent increase in the expected number of property-damage-only (PDO) accidents. The respective increases for the other model (14) were 0.75 and 0.61 percent. A model to predict expected accident frequencies on work zone approaches was also developed (16). A crash was assigned to a work zone approach if its location fell within the estimated congested segment upstream of the beginning of the work zone and the time of the crash coincided with the work zone presence. The model indicates that increasing work zone length will cause a decrease in the expected number of injury and PDO crashes on work zone approaches. The relationship is exponential, meaning that a 1 percent increase in the length of a long work zone results in a greater decrease in expected crashes than a 1 percent increase in the length of a shorter work zone.
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 10 2.2.2 Traffic Diversion Strategies Four studies quantified the safety effects of different traffic diversion strategies (10,13,17,24). In an analysis of 79 construction projects on mostly high-speed roadways in one of the studies (9), mean accident rates ranged from 0.77 accidents per million vehicle miles (MVM) for a crossover and detour to 5.29 accidents per MVM for a lane closure and temporary bypass. Table 1 provides a summary of these accident rates. It should be noted that the sample sizes for some lane closure strategies were quite small, which probably resulted in very large confidence intervals in predicting the population means. Table 1 Mean crash rates for different traffic diversion strategies (8) Traffic Diversion Strategy Number of Projects Mean Accident Rate (Accidents per MVM1) Lane closure 48 2.13 Crossover 4 2.24 Temporary bypass 0 -- Detour 0 -- Lane closure and crossover 5 1.50 Lane closure and temporary bypass 4 5.29 Lane closure and detour 10 2.99 Crossover and detour 3 0.77 Temporary bypass and detour 1 4.37 1 million vehicle miles. Another of the studies (24) reported on accident rates at 49 construction work zones on four-lane divided highways. The researchers compared long-term single lane closures installed in one direction to a crossover strategy, in which traffic in both directions of travel was reduced to one lane and a crossover provided two-lane, two-way operations on one set of travel lanes while work was completed on the other set of lanes. Overall, researchers found no statistically significant differences in the accident rates at both types of projects. Accident rates at the two types of projects averaged 1.96 and 2.62 accidents per MVM, respectively, before construction, as compared to 2.86 and 2.78 accidents per MVM during construction. Although the rate for the single lane closures did appear to increase more significantly than did the rate at the crossover sites, the high degree of variability in rates from site to site kept these differences in accident rates from being detected as statistically significant. An investigation of Indiana interstate highway work zone crashes also looked at before-during crash rates for lane closures and crossovers (13). The results are summarized in Table 2. Although the mean crash rates during the work zone period and the change in total and severe crash rates (during-before) are greater for crossover work zones, a comparison of the means showed no significant difference in both of these values (i.e., mean crash rate during work zone, mean change in crash rates from before to
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 11 during). This result is similar to that in the study (10) where work zones with crossovers had a slightly higher, but not much higher, crash rate than work zones with lane closures. Table 2 Mean crash rates for work zones with crossovers and lane closures (accidents per MVM1)(13) Sites Rate without Work Zone Rate with Work Zone Change in Crash Rate Sites using crossover (2 lanes in each direction) 0.62 0.83 0.21 Sites using partial lane closure (2 lanes in each direction) 0.58 0.77 0.19 Sites using crossover (3 lanes in each direction) 0.60 0.97 0.37 Sites using partial lane closure (3 lanes in each direction) 0.78 1.05 0.27 1 million vehicle miles. The same study (10) also looked at the effects on accident rates of degrading various road types. The results are shown in Table 3. In the final study reviewed on the subject of safety effects associated with different traffic diversion strategies (17), data were obtained for North Carolina work zone crashes through the Highway Safety Information System. The effects of work zone characteristics on the most seriously injured occupant (no injury, minor, moderate, severe, fatal) and the total harm (measured in economic cost) in truck-involved and non- truck-involved collisions were studied. The ordinal probit model was used to investigate the former, and linear regression was used to investigate the latter. Table 4 presents model coefficients for multi-vehicle collisions. For all crash types, severity and harm were highest in work zones located on two-way, undivided roadways. For truck-involved, multi-vehicle collisions, severity and harm were overwhelmingly the highest in work zones classified as âroadway closed, detour opposing side.â It is unclear what this category includes. Possible scenarios are median crossovers and one-way alternating traffic.
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 12 Table 3 Before-during accident rates by road degradation (accidents per MVM1) (10) 1 million vehicle miles. 2 two-way left turn lane. Roadway Number of Projects Before Construction During Construction Change (%) 6- or 8-lane interstate reduced to 2 lanes per direction 8 2.02 2.13 +5.3 6- or 8-lane interstate reduced to 1 lane per direction 3 2.37 5.10 +114.6 4-lane interstate reduced to 1 lane per direction 22 1.42 2.39 +68.6 4-lane interstate reduced to 2- lane, 2-way 2 0.42 1.05 +147.2 4-lane divided reduced to 1 lane per direction 5 3.28 3.77 +14.8 4-lane divided reduced to 2- lane, 2-way 5 1.84 2.14 +15.9 4-lane divided on new alignment 6 2.59 2.09 -19.5 4-lane undivided reduced to 2 lanes 3 8.35 7.94 -4.9 5-lane undivided with TWLTL2 reduced to 2 lanes 3 5.09 8.08 +59.0 2-lane roadway reduced to 1 lane 7 3.79 4.96 +30.7 2-lane roadway on new alignment 11 6.63 5.68 -14.3
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 13 Table 4 Coefficients for injury severity and total harm models (18) Injury (Ordered Probit) Harm (Semi-log) Variable All Truck Non-Truck All Truck Non- Truck One-way, not divided 0.192 0.305 0.168 0.108 0.211 0.076 Two-way, not divided 0.428 a 0.510 a 0.398 a 0.266 a 0.313 a 0.247 a Two-way, divided, no median barrier 0.137 c 0.362 b 0.085 0.104 b 0.260 b 0.072 R oa dw ay c on fig ur at io n Two-way, divided, median barrier 1 Lane closed -0.166 -0.096 Shoulder/median closed 0.278 0.230 Roadway closed, detour opposing side 1.011 a 0.889 a Lane shift/becomes narrow -0.689 -0.381 Other/unknown 0.122 0.136 C on st ru ct io n ef fe ct o n ro ad w ay None 1 a, b, and c: the coefficient is significantly different from 0 at the 1 %, 5 %, and 10 % level of significance (two-tailed test), respectively. 1 base category for coefficient comparison. 2.2.3 Ramps One study (15) investigated changes in accident occurrence during construction at urban freeway entrance-ramp areas and non-entrance-ramp areas to determine if accidents increased disproportionately in the entrance-ramp areas compared with the non- entrance-ramp areas within the construction work zone. The data were from two long- term urban freeway reconstruction projects in Texas, I-35 W in Fort Worth and I-45 in Houston. Comparison sites were also chosen and were located either upstream or downstream of the construction sites. The analyses were done separately for each construction site with inclusion of a G2 test for comparability of the work zone and comparison sites (22). The results are summarized in Table 5. A ratio greater than 1 indicated that accident frequencies increased more in entrance-ramp areas than non-
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 14 entrance-ramp areas during construction. For I-35 W, all accident types increased more at entrance-ramp areas than at non-entrance-ramp areas. The increase at entrance-ramp areas was significantly greater for total accidents, PDO accidents, severe accidents, daytime accidents, and other multi-vehicle accidents. For I-45, five of eight accident categories increased more and three of eight accident categories increased less at entrance-ramp areas than at non-entrance-ramp areas. However, none of these differences was statistically significant. Looking at the difference in results between the two locations, it should be noted that the ramp geometrics on I-35W were greatly altered during construction. This included having very short acceleration lane lengths (approximately 50 feet). Conversely, ramp geometrics on I-45 in Houston were not as greatly affected during most of the construction efforts. Table 5 Non-entrance-ramp areas versus entrance-ramp areas (15) I-35W Entrance-Ramp vs. Non- Entrance-Ramp Areas I-45 Entrance-Ramp vs. Non-Entrance- Ramp Areas Accident Category Percent Difference in Change in Accident Frequency Accident Category Percent Difference in Change in Accident Frequency Total accidents +30.4a Total accidents +3.6b Accident severity Accident severity PDO1 +26.1a PDO -2.0b Severe +45.6a Severe +19.0b Time-of-day Time-of-day Daytime +34.7a Daytime +10.7b Nighttime +22.5b Nighttime -7.3b Collision type Collision type Single vehicle +4.4b Single vehicle +9.8b Rear-end +15.4b Rear-end +1.4b Other multi- vehicle +49.2a Other multi-vehicle -1.7b a control and construction sections are comparable and differences are significant. b control and construction sections are comparable and differences are not significant. 1 property damage only. 2.2.4 Lane Widths One quantitative assessment of lane width was reviewed (10). Accident rate comparisons were between projects that had reduced lane widths and projects that maintained normal lane widths. The level of lane width reduction was not given. Six projects with reduced lane widths during construction experienced a 17.6 percent increase in accident rates during construction, whereas the other 69 projects with normal lane widths experienced a 6.6 percent increase.
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 15 2.3 WORK ZONE CRASH CHARACTERISTICS Many of the reviewed studies investigated work zone crash characteristics in some way. The many tests and conclusions do not provide consistent findings as to the magnitude of certain effects. For example, when looking at overall magnitude of work zone crashes, one group (10) found a 7.5 percent increase in accident frequency during the work zone period, while another investigation (11) found a 119 percent increase. Similarly, work zone accident rates ranged from 0.89 (12) to 8.63 (13) crashes per MVM. These large differences can be attributed to lack of control or data regarding possibly important explanatory variables (e.g., cross section, alignment, roadside). This section will present general findings with illustrative numerical results. It is important to note that the studies of crash characteristics were only able to report magnitudes and are plagued by a lack of exposure data. Therefore, it is difficult to draw meaningful conclusions about the actual safety consequences of work zones. A first attempt at estimating true work zone exposure has recently been made (25). 2.3.1 Crash Magnitude and Severity It was a common finding that the presence of a work zone on a specific roadway segment is likely to degrade its safety. Studies found crash frequencies and crash rates to be, on the average, higher during work zone periods than before (1,10,11,12,13,14). However, the results varied. Studies that used a sample of work zones often had a certain, sometimes significant, portion of the sample with lower crash rates during the work zone period (10,13). For example, in one study (10), 31 percent of the projects experienced decreases in accident rates during construction, 47 percent experienced increases between 0 and 50 percent (exclusive), and 24 percent experienced increases of 50 percent or more. There are probably factors, unaccounted for, that differentiate the sample sites. Similar variability was found in another study (26): 45 percent of the study sites experienced increases in accident rates of 40 percent or more, 8 percent experienced decreases in accident rates of 40 percent or more, and 47 percent of the study sites experienced less than 40 percent changes in accident rates. Conclusions regarding the severity of crashes in work zones were mixed. Some studies found work zone crashes to be less severe than crashes outside of work zones (6,9), while others found work zone crashes to be more severe (1,11). These determinations were made by either comparing the percent increases in PDO crashes to the percent increases in fatal and injury crashes in before-during studies (1,10,11), or by comparing work zone accidents to statewide accidents outside of work zones (7). The higher severity in one study (11) was due to a 300 percent increase in fatalities (from 2 fatal accidents to 8). The most severe work zone crash type appeared to be multi-vehicle collisions involving a heavy vehicle (17).
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 16 2.3.2 Crash Type and Location The most common work zone crash type in six studies (7,9,10,18,20,21) was a rear-end collision, accounting for anywhere between 35 and 52 percent of all work zone crashes. One investigation (11) found that fixed object collisions were the most predominant type, and another (12) found run-off-road and fixed object collisions to be most common. Yet another (8) found the most common fatal collision type to be a single vehicle crash. In a recent study (9), the predominant crash location was the work area, accounting for 70 percent of all work zone crashes. Although terminology is different, another team (21) found 39.1 percent of crashes to occur in the lane closure (perhaps equivalent to buffer area), 22.5 percent in the lane taper (transition area), and 16.6 percent in the construction area. Both of these studies also investigated crash type by location and are in pretty close agreement. One (21) found the most common location-type combination to be rear-end crashes in the lane closure, and the other (9) found it to be rear-end crashes in the activity area. A higher proportion of crashes occurred during daylight, but in three of four before-during studies, night crashes increased more than day crashes (1,10,11). One team (12) found no difference in the percent increase of night and day accidents. Another team (7) compared statewide work zone accidents to accidents outside of work zones and found that similar percentages occurred during day and night hours. In an analysis of fatal accidents, a team (8) found that 42 percent occurred during night conditions. The most common vehicle type involved in a work zone accident was a passenger car. However, in a comparison of work crashes to crashes outside work zones, a higher percentage of work zone crashes involved heavy vehicles than crashes outside work zones (7). 2.4 SUMMARY OF WORK ZONE RESEARCH This section focused on safety effects of specific work zone design features as well as prevalent characteristics of work zone crashes. It presented the following findings: ⢠In most cases, work zones with crossovers appear to have slightly higher accident rates than work zones with lane closures. In addition, multi-vehicle accidents in which a truck is involved are much more severe in work zones with crossovers than work zones with other types of roadway configurations. ⢠It is inconclusive whether accident magnitudes and characteristics are different at entrance-ramp locations from what they are at other areas in the work zone. However, removing or significantly shortening the length of entrance-ramp acceleration lanes may be associated with significant increases in accident rates in some cases.
Final Report for NCHRP Report 581: Design of Construction Work Zones on High-Speed Highways Copyright National Academy of Sciences. All rights reserved. 17 ⢠Work zones where lane widths are reduced from pre-work zone conditions experience a higher increase in accident rate than work zones with no lane width reductions. ⢠A wide, sometimes conflicting, range of results exists from work zone research investigating crash frequencies and characteristics. These large differences can be attributed to lack of control or data regarding possibly important explanatory variables (e.g., cross section, alignment, roadside). ⢠Many studies that have investigated crash frequencies and characteristics are plagued by a lack of exposure data. Therefore, it is difficult to draw meaningful conclusions about the actual safety consequences of work zones. ⢠In general, crash frequencies and crash rates appear to be, on the average, higher during work zone periods than before. ⢠Conclusions regarding the severity of crashes in work zones were mixed. Some studies found work zone crashes to be less severe than crashes outside work zones while others found work zone crashes to be more severe. ⢠The predominant crash type was a rear-end crash, and the predominant crash location was the activity area of the work zone. Other information relevant to construction work zone design and traffic control exists in current national guidance publications such as the MUTCD, Green Book, and Highway Capacity Manual. A review of the information from these publications is provided in Chapter 3.
