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.
14 Study Methodology Researchers designed the experimental plan for this com- ponent of the research study to best answer the previously identified question, âHow does doing a project task at night affect traffic safety relative to doing this same task during the day?â Implied in this question is the recognition that the comparison should be based on the same work task being performed (at least from a traffic control perspective) and on doing the work at the same location. Also, implied in this question is the assumption that the comparison is being made on the basis of total additional crash costs being incurred over the duration of the project task that needs to be completed (since the severity of the additional crashes occurring during the day and at night may be different). Because directly comparing traffic crash experiences for similar temporary traffic control setups used during night- time and daytime work operations at a given location would be too limited to allow proper statistical analyses, researchers developed an experimental plan to make use of a large num- ber of projects on freeway facilities that involved frequent or intermittent temporary lane closures during either daytime or nighttime work hours. Researchers expected that daytime temporary lane closures would exist at project sites on lower- volume facilities where the closure of travel lanes would not generate significant congestion and delay. Similarly, researchers anticipated that projects where temporary lane closures are done at night would occur predominantly on higher-volume facilities because of desires to avoid creating significant con- gestion and delays during daytime hours. There would be some overlap of these ranges, depending on the specific criteria used by the highway agency having jurisdiction over each project, roadway and work task characteristics, etc. From this dataset, statistical techniques would be used to establish separate relationships of the additional crash costs per period of work activity involving temporary lane closures during daytime work periods and during nighttime work periods. For the daytime work periods, the effect of working during the day at higher-volume locations would have to be extrap- olated beyond the limits of the data; for nighttime work peri- ods, an extrapolation to lower-volume freeway types would be required. This analysis approach is depicted graphically in Figure 1. The relationships are depicted as nonlinear, re- flecting expectations that the effects of congestion and queues significantly add to the crash risk as volumes increase. The potential for the increased crash costs at night work operations to exceed those at daytime work operations is also implied at very high traffic volume levels since temporary lane closures on very high-volume facilities can still create traffic and con- gestion even at night. Similar relationships could likewise be developed for other comparable work conditions. For example, daytime versus nighttime crash cost increase comparisons during periods of work activity without temporary lane closures would be of in- terest to assess the relative effects of work distractions, vehicle and equipment access impacts, and other non-lane closure influences on traffic safety. In addition, daytime versus night- time comparisons of crash cost increases during periods of work inactivity could provide an indication of the relative effects of work zone design features (reduced lane and shoul- der widths, lane shifts, etc.) on traffic safety. Researchers initially contemplated stratifying the data on the basis of roadway type. One would expect the decision of whether or not to work at night to be influenced by different reasons for an urban arterial work zone versus a work zone on a freeway or expressway segment, for example. Differences in driver demographics, traffic volumes and speeds, and crash rate increases between daytime and nighttime conditions are likely not the same for different roadway types. Unfortunately, the sample sizes required to properly evaluate nighttime and daytime work operations on multiple roadway types were beyond the budget limitations of this study. Consequently, researchers focused their efforts in this study on freeway and expressway facilities. Although the results of the study do C H A P T E R 3 Analysis of Traffic Crashes during Nighttime and Daytime Work Zone Operations
provide important insights into this issue, it should be re- membered that the findings may not represent all possible types of work zones where a decision of whether to work at night must be made. Data Collection The experimental plan for this portion of the study called for the collection of crash and project activity data across a range of geographically dispersed highway work zones nation- ally, each of which involved occasional to frequent temporary lane closures to complete the work. For some of the projects, these temporary lane closures occur primarily during daytime hours; for the other projects, these temporary lane closures occur almost exclusively at night. Researchers targeted states that participate in the FHWA HSIS so that multiple years of crash data, annual average daily traffic (AADT), and roadway characteristic data would be more easily accessible. Ultimately, projects were identified from four HSIS states: ⢠California, ⢠North Carolina, ⢠Ohio, and ⢠Washington. Originally, the intent was to obtain data from several proj- ects in Texas as well. However, the lack of available crash data prompted researchers to drop that state from the analysis. Researchers contacted department of transportation (DOT) officials in each of the other states to request assistance in identifying suitable candidate projects to use in this study. An initial list of 92 projects was generated through this effort. In order to accomplish the analysis approach described in the previous section, researchers required details about the daily (and nightly) work activity performed by the highway contractor at each project such as: hours of work, hours and locations of temporary lane closures set up and removed, and the number of travel lanes closed each work period. Project details such as these must be extracted manually from the daily diaries of the project inspectors who were onsite each day or night. A few states have construction management databases (such as the Trnsâ¢port SiteManager software available through the American Association of State Highway and Transportation Officials [AASHTO]) where this information may be entered electronically (49). Even so, it was necessary for research staff to sit down with either the diaries or the SiteManager pro- gram itself to extract the pertinent information for each work period on each project of interest. In addition to work activ- ity information, researchers also required information about the traffic control plan used, construction phasing, etc., for each project. Two- to three-person data collection teams traveled to each state, except Ohio, to gather the necessary project data for analysis. Ohio uses a construction management database that they were willing to download and send electronically, negating the need to travel to that state. In the other three instances, the state DOT staff in each state provided key assistance in gaining the data collection access to the neces- sary project records. In most cases, the projects themselves had been closed out and the records archived, and so the DOT staff had to request that the files be pulled and trans- ported to a location where the research team could use them. Once the records were in hand, the data collection team had to verify that all of the necessary information was available 15 Nighttime Operations AADT Most temporary lane closures performed during daytime operations Most temporary lane closures performed during nighttime operations Daytime Operations In cr ea se in C ra sh C os ts p er W or k Pe rio d Figure 1. Theorized relationships between increased traffic crash risk and roadway traffic demand at nighttime and daytime work operations that require temporary lane closures.
and that the project was useable. In a few cases, project diary information or traffic control plans were misplaced and not with the rest of the project documents. For those situations, no project data could be collected. From the initial list, data were located and retrieved by research staff for 84 of the projects. Once the data collection team returned to the office, re- searchers requested crash and roadway inventory data for each project segment length for the duration that the project itself was active and for several years preceding. Additional details regarding the before periods for each project can be found in Appendix A. Researchers actually requested data for an additional upstream distance around each project to permit a check for indicators that traffic queues or other work-zone- related effects were contributing to crashes occurring in those adjacent segments. A high percentage of crashes coded as âwork zone involvedâ in adjacent segments over the dura- tion of the project was the key indicator that the effects of the work zone were extending beyond the limits of the project. In these cases, the limits of the project were then expanded to incorporate those segments. When the project limits were expanded, the added segment generally totaled less than 0.5 mi per direction. The number of usable projects was further reduced due to the following issues that unfortunately were not discovered until after the data collection effort: ⢠Based on the roadway inventory data, three projects con- tained sections of roadway that were not limited access facilities (i.e., freeway). This occurs when the work activity is conducted in the area where a roadway changes to or from a limited access facility. Researchers decided not to in- clude these projects in the dataset since the accident trends on these roadways may differ from the rest of the dataset. ⢠Ten of the Ohio projects could not be used since the elec- tronic diary data did not include the exact work times (e.g., midnight to 5 am), information concerning lane closures, or both. Researchers contacted the Ohio DOT to obtain more detailed hard-copy diaries but found that the hard-copy diaries did not contain any additional information. ⢠During the data collection and reduction stages, it was an- ticipated that the 2005 Ohio crash data would become available; however, this did not come to fruition. Thus, two Ohio projects conducted during this time period could not be used. ⢠One of the Ohio projects could not be used since the mile points where the project occurred were missing from the HSIS. ⢠Washington did not provide 1997 and 1998 crash data to FHWA for inclusion in the HSIS. Thus, four Washington projects conducted during this time period could not be used. This reduced the final dataset to a total of 64 projects. Even though somewhat smaller than the initial list of proj- ects targeted, this dataset is substantial. An overall sum- mary of project characteristics and crash statistics is presented in Table 9. Overall, the projects encompass approximately 16 Statistic State California North Carolina Ohio Washington Overall Duration of Projects, Days: Total of All Projects in Sample Average per Project Standard Deviation Minimum per Project Maximum per Project 6,719 419.9 215.6 74 862 11,329 566.5 549.6 44 2,114 5,710 571.0 363.5 81 1,033 6,048 336.0 321.3 40 1,236 29,806 466.0 399.0 40 2,114 Lengths of Projects, Miles: Total of All Projects in Sample Average per Project Standard Deviation Minimum per Project Maximum per Project 110.9 6.9 4.4 1.4 17.0 155.9 7.8 7.2 2.0 30.2 44.0 4.4 2.8 0.3 9.4 154.0 8.6 18.2 0.7 80.5 464.8 7.3 10.6 0.3 80.5 Traffic Exposure of Projects, mvm: Total of All Projects in Sample Average per Project Standard Deviation Minimum per Project Maximum per Project 4,369.5 273.1 359.6 27.6 1,425.8 4,742.0 237.1 315.0 3.4 1,234.5 1,371.7 137.2 159.1 25.7 544.1 2,430.4 135.0 193.4 0.7 716.1 12,913.6 201.8 279.4 0.7 1,425.8 Traffic Crashes Occurring during Projects: Total of All Projects in Sample Average per Project Standard Deviation Minimum per Project Maximum per Project Average per Mile per Year 6,613 413.3 607.3 27 2,292 289.0 4,831 241.6 325.0 0 1,294 106.96 2,776 277.6 412.1 12 1,382 200.8 3,008 167.1 272.6 0 1,105 139.6 17,228 269.2 415.3 0 2,292 245.9 Table 9. Summary of project characteristics and crash statistics.
465 centerline-mi of roadway and over 82 years of work, for which researchers had to manually determine days and hours of work activity and whether temporary lane closures were present. Both project length and duration were highly variable, with an average length of slightly more than 7 mi and an aver- age duration of about 16 months. Actual lengths ranged from 0.3 to 80.5 mi, and durations ranged from 40 days to 5.8 years. Also summarized in Table 9 are the work zone crashes occur- ring on these projects. More than 17,000 crashes were reported during the performance of these 64 projects. Additional proj- ect details and crash statistics can be found in Appendix A. Next, Table 10 summarizes the daytime and nighttime crash rates per 100 mvm that would normally be expected for the sample project locations in each state if a work zone were not present. As shown, these non-work zone crash rates tended to be higher at night than during the day. Furthermore, the difference between the nighttime and daytime rates tended to be greater for the severe crashes. These numbers indicate that, even in the absence of a work zone, driving at night is normally more risky for drivers than driving during the day on a per-vehicle-mile traveled basis. Although the per-mvm rates are usually higher on roadway facilities nationally at night than during the day, the much lower traffic volumes using the facilities at night means that the actual number of crashes occurring on a per-night, per-mile basis is still usually less than for a per-day, per-mile basis on the same facility. Researchers developed exposure estimates and stratified the crashes occurring during each project in the database into one of six categories: ⢠Daytime and nighttime periods when the project was inac- tive and no temporary lane closures were in place in the work zone; ⢠Daytime and nighttime periods when work activity was occurring somewhere within the project but temporary lane closures were not in place (i.e., no work was occurring in the way of travel); and ⢠Daytime and nighttime periods when work activity was oc- curring somewhere within the project and temporary lane closures were in place that reduced the available capacity of the roadway. A fourth possible category, daytime and nighttime periods when the project was inactive and temporary lane closures were in place, was very limited in the dataset and so was not considered in this analysis. Researchers attempted to ensure that the projects obtained from each state were somewhat balanced between those that had work activity and temporary lane closures during the day, and those that had work activity and temporary lane clo- sures at night. Ultimately, however, very few projects with daytime work activities and temporary lane closures were available from California and Ohio. Therefore, this category is overrepresented by North Carolina and Washington proj- ects. Also, the projects obtained from California and Ohio tended to be on higher AADT facilities than those from North Carolina and Washington. Researchers hypothesized that increases in crash risk dur- ing the inactive periods of the project (relative to the crash risk normally expected on that roadway segment) reflected the influences that temporary geometric changes and other work zone design decisions had upon safety. Similarly, crash risk increases during periods of work activity but with no tempo- rary lane closures was assumed to reflect the combined effects of the geometric changes/work zone design decisions and dis- tractions and turbulence caused by work activities adjacent to the travel lanes. Finally, the increase in crash risk during peri- ods of work activity with temporary lane closures represented the combined effect of geometric changes/work zone design decisions, work activity distractions and turbulence, and additional traffic turbulence caused by the temporary roadway capacity restrictions. Table 11 illustrates this concept. The projects used in this study varied widely in terms of the relative amount of work activity performed during the day and night, as well as the frequency with which these active work periods required one or more travel lanes to be tem- porarily closed. This is illustrated in Table 12. Averaged across each state and over the entire study sample, the projects tended to be active more often during the day than at night in North Carolina, Ohio, and Washington (the California projects were approximately equally active day or night). However, much of the activity during the day at these projects occurred outside the roadway, whereas most of the work activity at night involved temporary lane closures. Specifically, when work occurred at night on the sample projects, 88 percent of the time it involved a temporary lane closure. In contrast, temporary lane closures were utilized only 26 percent of the time that work activity occurred during the day. Of course, these statistics may not be indicative of all freeway projects in these states because projects involving temporary lane closures were specifically targeted in this analysis. 17 Crashes per 100 mvm Crash Severity State Nighttime Daytime California 97.1 78.7 North Carolina 91.6 72.1 Ohio 92.7 77.1 Washington 87.5 89.7 Severe (Injury or Fatality) Overall 92.8 78.1 California 150.7 151.8 North Carolina 153.5 121.4 Ohio 248.4 231.2 Washington 112.6 108.9 PDO Overall 155.8 140.5 Table 10. Expected (non-work zone) average crash rates in the project dataset.
As expected, the majority of projects where temporary lane closures were performed during the day occurred at locations where AADTs were relatively low, and those performed where AADTs were relatively high involved predominantly night- time temporary lane closures. This is illustrated graphically in Figure 2, which shows the percentage of hours involving a temporary lane closure at each project that was performed during nighttime hours (nighttime was defined as beginning at 7:00 pm, after the evening peak period, and ending at 6:00 am, prior to the start of the morning peak period). As shown, temporary lane closures at projects on roadways with AADTs less than about 40,000 vehicles per day (vpd) were mostly performed during daytime hours, whereas those on freeways with AADTs in excess of 100,000 vpd were almost all performed at night. Between these ranges, the results were mixed. Night work was used extensively on some projects as low as 35,000 vpd, while a few projects on freeways with AADTs of up to 75,000 vpd still had about 60 percent or more of temporary lane closures occur during daytime hours. Data Analysis For each of the work zone analysis periods of interest, re- searchers used empirical Bayesian (EB) statistical techniques to estimate the incremental increase in crash risk that occurred relative to what would have been expected to have occurred if the work zone were not present at that location. EB techniques increase the precision of estimation and correct regression-to- the-mean bias (50). Often, the limited duration of a particular work zone project means that the sample size of crashes avail- able for use in the analysis is quite small. Regression-to-the- mean biases may also exist at some work zone locations if the selection of roadway segments being targeted for repair and improvement is based partially on the recent crash experi- ences of that roadway segment. Consequently, EB techniques provide better estimates of the safety impacts of highway work zones than traditional before-during crash comparisons. The EB procedure required researchers to develop safety performance functions (SPFs), using data from a reference group, of freeway facilities under daytime and nighttime con- ditions in each of the states where work zone projects were taken. The estimates from the SPFs were then combined with crash data occurring within the project limits for several years preceding the work at that location. The combination of the SPF and pre-work zone crash data provided a more precise estimate of the crashes that would be expected to occur over a given period of time at that location if the work zone had not been present. The ratio of the actual number of crashes occurring during the operation of the work zone to the EB es- timate is then used to estimate the incremental effect of the 18 Work Zone Analysis Period Ef fe ct o f T em po ra ry G eo m et ri c a nd W or k Zo ne D es ig n In flu en ce s o n Sa fe ty Ef fe ct o f O ff- Tr av el La ne W or k A ct iv ity D ist ra ct io ns a nd D isr up tio ns o n Sa fe ty Ef fe ct o f T ra ffi c D isr up tio ns d ue to Te m po ra ry R oa dw ay C ap ac ity R ed uc tio ns o n S af et y When Work Zone Inactive and No Temporary Lane Closures X When Work Zone Active and No Temporary Lane Closures X X When Work Zone Active and Temporary Lane Closures Present X X X Table 11. Relationship between work zone analysis periods and influences on work zone safety. State Statistic California North Carolina Ohio Washington Overall % of Time Active, Daytime 20.9 33.5 40.2 26.9 30.6 % of Time Active, Nighttime 20.8 13.6 15.2 15.7 16.0 % of Active Time with Temporary Lane Closures, Daytime 21.6 37.0 5.2 34.4 26.2 % of Active Time with Temporary Lane Closures, Nighttime 89.8 84.9 88.3 88.9 87.8 Table 12. Amount of work activity and temporary lane closures during daytime and nighttime periods in sample projects.
work zone upon safety and is referred to herein as the âindex of changeâ observed. Researchers analyzed fatal and injury crashes separate from PDO so that possible daytime and nighttime differences in crash severity could be estimated as well (reference to âinjuryâ crashes herein implies the combi- nation of both fatal and injury crashes). Additional details of the EB procedure employed for this study can be found in Appendix A. After the differences in crashes were estimated at each proj- ect location for each time period of interest, the crash costs associated with these differences were computed. Recent cost values for freeway crashes (51) were used: ⢠Injury crash (fatality or injury)â$206,015, and ⢠PDO crashâ$7,800. The differential crash costs per unit duration of work activ- ity or inactivity (with and without temporary lane closures) per mile of work zone were computed and modeled as a func- tion of AADT. Consideration was given to developing incremental crash increase models for selected collision types (rear-end, sideswipe, run-off-road, etc.), but this level of dissection of the data was determined to be too fine to permit statistically significant conclusions to be drawn from the data. Therefore, a simple comparison of the percentage involvement of these factors in the crashes before and during construction, aggregated across each state, was used to determine whether significant differ- ences existed between the before project conditions and each of the work activity periods of interest in this study. Results Increases in Traffic Crashes Occurring during Nighttime and Daytime Work Activities Increases in Crash Risk Appendix B provides the number of injury and PDO crashes occurring at each project during each nighttime and daytime work period type (work activity or no work activity, with or without lane closures) and those expected to have occurred during those same periods if the work zone were not present. In Figures 3 through 5, the index of change estimated by the EB procedure is plotted against AADT for each project for each of the following scenarios: ⢠Project work was occurring (the work area was active), and temporary lane closures were in place (Figure 3); ⢠Project work was occurring, but no temporary lane clo- sures were in place (Figure 4); and ⢠The project was inactive, and no temporary lane closures were in place (Figure 5). An index of change of 1.0 indicates that the number of crashes actually occurring is equal to the number of crashes that were expected to have occurred based on the EB analysis. Values greater than 1.0 reflect an increase in actual crashes during construction relative to the number of crashes that would be expected if the work zone was not 19 0% 20% 40% 60% 80% 100% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f T em po ra ry L an e Cl os ur e Ho ur s Pe rfo rm ed a t N ig ht Figure 2. Percentage of temporary lane closures performed at night at each project.
present. A higher ratio indicates a greater increase in actual crashes. Across the three figures, considerable variability is evident from project to project. The index of change is as high as eight for some projects (i.e., the actual number of crashes that oc- curred is eight times greater than the crashes expected at that project location based on the EB analysis). In other instances, the actual number of crashes was less than the expected number (i.e., index of change is less than one). In still other cases, no crashes occurred during the project work period of interest, so the index of change is zero. Although it was initially hypothesized that the effects of work activity and temporary lane closures would be higher (i.e., the index of change would be higher) at higher AADT levels, the analysis results did not bear this out. The projects were stratified into three AADT regions (less than 50,000 vpd; 20 0. 0 1. 0 2. 0 3. 0 4. 0 5. 0 6. 0 7. 0 8. 0 9. 0 0 50000 100000 150000 200000 Roadway AADT In d e x o f C h a n ge Da yt im e Ni gh tti me (a) Injury Crashes 0. 0 1. 0 2. 0 3. 0 4. 0 5. 0 6. 0 7. 0 8. 0 9. 0 0 50000 100000 150000 200000 Roadway AADT I n d e x o f C h a n ge Da yt im e Ni ght ti me (b) PDO Crashes Figure 3. Index of change for injury and PDO crashes during periods of work activity with temporary lane closures in place.
50,000â100,000 vpd; and greater than 100,000 vpd) and ana- lyzed to determine the average index of change across the projects in each region. Table 13 presents the results of the analysis for periods when the work area was active and tem- porary lane closures were present. A weak trend of increasing ratios at higher AADT levels is evident for the nighttime work periods, but this is not replicated for daytime work periods. Furthermore, the fairly sizeable standard errors of these esti- mates indicates that there are no statistically significant dif- ferences in the crash ratios between any of the AADT regions for either injury or PDO crashes during either nighttime or daytime work periods. Consolidated across the entire AADT range, the index of change was essentially the same for both nighttime and daytime periods. Subtracting one from the index of change, expressed as a percent, defines the percent increase in crashes that occurred overall across the projects for the work condition and time period of interest. Injury crashes increased by 42.3 percent at night and by 45.5 percent during the day. For PDO crashes, the increase was 74.8 per- cent at night and 80.8 percent during the day. Researchers 21 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 0 50000 100000 1500000 200000 Roadway AADT In de x of C ha ng e Daytime Nighttime (a) Injury Crashes (b) PDO Crashes 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 0 50000 100000 1500000 200000 Roadway AADT In de x of C ha ng e Daytime Nighttime Figure 4. Index of change for injury and PDO crashes during periods of work activity but no temporary lane closures in place.
also combined all crash severities together and computed an index of change. Total crashes increased 60.9 percent during daytime work activity with temporary lane closures and 66.3 percent at comparable work operations at night. In Table 14, analysis results are presented for periods when work activity was occurring but no temporary lane closures were in place. Overall, this was a fairly infrequent event dur- ing night operations. Consequently, the crash ratio estimates obtained were not extremely reliable (as indicated by the large standard errors associated with the estimates). Again, no statistically significant differences were detected across the different AADT levels during either nighttime or daytime periods for either the injury crash or PDO crash indices of change. Although the overall injury index of change for the nighttime period appears to be higher than it is for the day- time period (indicating a 41.4 percent increase at night versus 17.4 percent increase during the day), they are, in fact, not statistically different from each other. Similarly, the index of change of PDO crashes in the nighttime period (indicating a 66.6 percent increase) is not statistically different than it is 22 Daytime Nighttime 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 0 50000 100000 150000 200000 Roadway AADT In de x of C ha ng e (a) Injury Crashes Daytime Nighttime (b) PDO Crashes 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 0 50000 100000 150000 200000 Roadway AADT In de x of C ha ng e Figure 5. Index of change for injury and PDO crashes when work area was inactive (no temporary lane closures in place).
for the daytime period (indicating a 39.8 percent increase). Finally, total crashes during this condition increased 57.7 per- cent during daytime periods and 31.4 percent during night- time periods. Interestingly, the only category in which statistically sig- nificant differences were found between nighttime and day- time conditions was when the work area was inactive and no temporary lane closures were present. In Table 15, the differ- ences in nighttime and daytime injury and PDO crash ratios across the AADT regions are not statistically significant. Overall, a slightly higher increase in injury crashes is seen for nighttime conditions (11.4 percent increase) than for daytime conditions (2.0 percent increase), but these are not statisti- cally different from each other. For PDO crashes, the average increase at night (33.0 percent) was actually significantly higher than during the day (19.6 percent). For total crashes, the average increases were 23.7 percent and 12.7 percent dur- ing nighttime and daytime periods, respectively. The greater increases at night presumably reflect degraded geometric con- ditions in the work zone (relative to a no-work zone condition) thatâcoupled with nighttime-specific issues such as limited visibility, less attentive drivers, and so forthâraise nighttime crash risk more substantially than during inactive times in daytime periods. It is important to note that the indices of change in each work condition category are higher for the PDO crashes than for the injury crashes, indicating that the additional crashes that do occur while the work zone is present tend to be less severe in nature. This trend exists regardless of whether the work is performed during the day or at night and is consistent with previous studies that indicated that crash rates may increase in work zones but that crash sever- ity often decreases. The magnitude of the change indices when work activity was occurring but no travel lanes were closed was a rather surprising finding from this analysis, especially for the night- time period. These indices and those from when work was occurring and lanes were closed are compared directly in Table 16. Theoretically, the lack of temporary lane closures when work is occurring means that motorists do not have an obstacle (the lane closure) in their travel path that requires a driving reaction, and they do not have to deal with signifi- cant reductions in speed because of traffic congestion up- stream of the closure. This would imply that the increase in crash risk when the work zone is active but lane closures are not present should be lower than when work activity is occur- ring and temporary lane closures are present. Although the crash ratios for daytime operations with and without tem- porary lane closures are consistent with this hypothesis, the results of this analysis indicate that working at night outside the travel lanes may have more substantial impacts on motorist safety than was known previously. Unfortunately, it is not clear from the data whether the increase in nighttime 23 Index of Change (S.E.) Crash Level AADT Range Nighttime Daytime <50k 1.318 (0.227) 1.596 (0.149) 50-100k 1.335 (0.151) 1.166 (0.244) >100k 1.491 (0.116) 1.261 (0.224) Injury Overall 1.423 (0.085) 1.455 (0.112) <50k 1.630 (0.188) 1.899 (0.126) 50-100k 1.712 (0.137) 1.338 (0.213) >100k 1.798 (0.103) 1.870 (0.199) PDO Overall 1.748 (0.076) 1.808 (0.096) <50k 1.527 (0.147) 1.770 (0.096) 50-100k 1.569 (0.103) 1.262 (0.161) >100k 1.649 (0.076) 1.645 (0.150) All Crash Types Combined Overall 1.609 (0.057) 1.663 (0.073) S.E. = Standard Error Indices in italics are not significantly different than 1. Table 13. Index of change for injury and PDO crashes by AADT range during periods of work activity and temporary lane closures. Index of Change (S.E.) Crash Level AADT Range Nighttime Daytime <50k 2.256 (1.302) 1.452 (0.216) 50-100k 1.341 (0.338) 1.189 (0.062) >100k 1.395 (0.318) 1.132 (0.057) Injury Overall 1.414 (0.229) 1.174 (0.042) <50k 1.359 (0.680) 1.371 (0.147) 50-100k 1.227 (0.253) 1.410 (0.056) >100k 2.037 (0.293) 1.388 (0.044) PDO Overall 1.666 (0.191) 1.398 (0.034) <50k 1.642 (0.622) 1.386 (0.121) 50-100k 1.285 (0.205) 1.323 (0.042) >100k 1.797 (0.215) 1.299 (0.035) All Crash Types Combined Overall 1.577 (0.148) 1.314 (0.027) Indices in italics are not significantly different than 1. Table 14. Index of change for injury and PDO crashes by AADT range during periods of work activity but no temporary lane closures.
crash risk during these work conditions is the result of the following factors: ⢠Work area lighting glare that work crews do not mitigate well when they are not located in travel lanes; ⢠More frequent construction equipment and material de- liveries into and out of the work area at night that create large speed differentials and subsequent crashes; or ⢠Other differences between daytime and nighttime work ac- tivity behaviors when travel lanes are not closed, such as higher speeds, reduced driver expectancy of encountering a work zone, and more impaired and drowsy drivers. Comparison of Daytime and Nighttime Work Based on Increased Crash Costs Associated with Work Zone Figures 3 through 5 and Tables 13 through 16 provide es- timates of the increased crash risk resulting from the presence of a work zone under each of the different work period cate- gories examined. The ratios (reflecting a percentage increase in actual crashes from what would have been expected had the work zone not been present) identify how the crash risk of individual drivers encountering these work zones is affected. As noted above, drivers approaching a work opera- tion at night where travel lanes are closed have a 42.3 percent greater risk (on average) of being in an injury crash and a 74.8 percent greater risk of being in a PDO crash than they would if the work zone were not there. Similarly, those same drivers traveling through that location at night when work is not occurring and no temporary lane closures are present have an 11.4 percent greater risk of being in an injury crash and a 33.0 percent greater risk of being in a PDO crash. In both instances, the increase in risk to individual drivers does not appear to depend upon the amount of traffic that the roadway handles on a daily basis. From the perspective of the practi- tioner who has to decide whether or not to work at night, though, the issue is not simply of the effects upon individual drivers, but on the entire driving population as a whole. Specif- ically, the question is whether completing a particular project or project task at night results in more or less additional crash consequences to motorists in total than doing the same project or task during the day. Given that the severity of crashes nor- mally differs between nighttime and daytime conditions, cal- culation of the effects of a standardized project task duration and length under a nighttime and daytime work scenario at a given location is the appropriate basis of comparison. Theoretically, computation of the additional crash costs expected for a particular project task duration and length could be accomplished uniquely for any project location, as long as the analyst had the following data: ⢠AADT of the roadway segment and how that AADT is dis- tributed between the nighttime and daytime work periods of interest, ⢠Estimated duration of the project task to be completed, ⢠Length of the work zone or work area, and ⢠Normal or typical crash rates for the nighttime and daytime periods being analyzed (or models that allow the analyst to estimate the number of crashes normally expected on the facility). If the more traditional (but less accurate) crash rate per mvm of the roadway segment is available, the analyst esti- mates the total vehicular exposure that would be experienced if the work were done during the day (number of days of 24 Index of Change (S.E.) Crash Level AADT Range Nighttime Daytime <50k 1.054 (0.087) 1.106 (0.061) 50-100k 1.141 (0.071) 0.936 (0.038) >100k 1.106 (0.063) 1.051 (0.030) Injury Overall 1.114 (0.042) 1.020 (0.022) <50k 1.133 (0.068) 1.271 (0.050) 50-100k 1.309 (0.067) 1.102 (0.033) >100k 1.455 (0.059) 1.234 (0.025) PDO Overall 1.330 (0.039) 1.196 (0.018) <50k 1.094 (0.054) 1.208 (0.039) 50-100k 1.240 (0.051) 1.042 (0.025) >100k 1.303 (0.043) 1.159 (0.019) All Crash Types Combined Overall 1.237 (0.029) 1.127 (0.014) Indices in italics are not significantly different than 1. Table 15. Index of change for injury and PDO crashes by AADT range during periods of no work activity and no temporary lane closures. Index of Change (S.E.) Crash Severity AADT Range Nighttime Daytime With lane closures 1.423 (0.085) 1.455 (0.112) Injury Without lane closures 1.414 (0.229) 1.174 (0.042) With lane closures 1.748 (0.076) 1.808 (0.096) PDO Without lane closures 1.666 (0.191) 1.398 (0.034) With lane closures 1.609 (0.057) 1.663 (0.073) All Crash Types Combined Without lane closures 1.577 (0.148) 1.314 (0.027) Table 16. Index of change comparisons with and without temporary lane closures during periods of work activity.
work required multiplied by the amount of traffic passing through the work area each day period multiplied by the length of the work zone) versus what it would be if the work were performed at night. The appropriate percentage increases in crashes from Table 13 through Table 15 are then applied to the crash rate and multiplied by the estimated vehicle exposure for each period to estimate the additional crashes anticipated to occur because of the work task during that time period. Finally, multiplying these crashes by the appropriate unit crash costs and summing across all severity levels (if severe and PDO crashes are estimated separately) would allow an equivalent comparison of increased costs between the two time periods. To illustrate the differences in daytime and nighttime work activity crash costs, SPF models for six-lane urban freeways in California (see Appendix A) were used to demonstrate how crash costs would be expected to increase for daytime and nighttime work zones as a function of the AADT level of the roadway segment. Similar trends would be obtained if the SPF models from the other states were used, although the ab- solute numbers would be different. The SPF model for free- way segments within large interchanges (where crashes tend to be somewhat higher) was averaged with those segments between interchange areas. A comparison of the estimated in- creased crash costs for a project task that requires a temporary lane closure to be used when work activity is occurring is pro- vided in Figure 6. The values are computed assuming that a project task requires 100 work hours to be completed regard- less of whether it is done at night or during the day. The data are also normalized to a per-work-zone-mile basis. Computed in terms of additional crash costs, it is apparent from Figure 6 that working at night when work activities require travel lanes to be temporarily closed results in lower crash costs than the same work performed during the day over the entire range of AADT levels shown. On higher AADT roadways, there is actually a very sizeable overall economic benefit to the motoring public of doing this work at night from a safety standpoint; on lower AADT roadways, the ben- efit may not be particularly large but still exists. For example, the reduction in crash costs for 100 hours of work per mile of work zone at night versus doing the work during the day exceeds $40,000 at a roadway AADT of 250,000 vpd. The differences between working at night versus working during the day on a project task that does not require tempo- rary lane closures are less clear. Using the same California SPFs as before, researchers applied the appropriate percentage crash rate increases for this condition from Table 14 to estimate the total increased crash costs on a per 100 work hours per mile of work zone basis. Figure 7 provides the results of that analysis. Overall, the increased crash costs per 100 hours of work activity per mile at night are very close to what they were in Figure 6. However, the increased crash costs for this par- ticular work condition are much lower for the daytime con- dition than they were in Figure 6. Ultimately, there is little or no benefit for working at night when a lane closure is not present. While there is a small advantage for day work at lower AADTs and a slight advantage for night work at higher AADTs, these differences are too small to significantly impact a decision of whether or not to work at night. Finally, Figure 8 presents the estimate of increased crash costs during the day and at night when the work zone is in- active and no temporary lane closures are required. For this particular case, the increased crash costs at night are slightly higher than during the day across the entire range of AADTs 25 Daytime Nighttime $0 $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000 0 50000 100000 150000 200000 250000 Freeway AADT In cr ea se d Cr as h Co st s pe r 1 00 W or k H ou rs p er M ile o f W or k Zo ne $40,000+ Figure 6. Increased crash costs with active work and lane closure present.
examined. Of course, the increased crash costs are much lower across the entire range of AADTs for both daytime and night- time conditions when compared to the previous figures when work is occurring (either with or without lane closures present). Types of Crashes Occurring during Nighttime and Daytime Work The preceding section examined the differences in work zone crash risk and crash costs between daytime and nighttime work for comparable periods of work activity and inactivity with and without lane closures present. The analysis consid- ered both individual drivers and the driving population as a whole by including differences in traffic volume during night and day periods. Another relevant question is whether the types of work zone crashes differ significantly between night- time and daytime periods. If so, such differences could lend insight into improved work zone safety policies, procedures, and practices that may produce an overall reduction in work zone crash risk. In this section, an analysis of the distribution of different crash types/manners of collision is presented. Specifically, crashes were subdivided into one of four collision types: ⢠Rear-end collisions, ⢠Sideswipe collisions, 26 $0 $5,000 $10,000 $15,000 $20,000 $25,000 $30,000 0 50000 100000 150000 200000 250000 Freeway AADT In cr ea se d Cr as h Co st s pe r 1 00 W or k H ou rs p er M ile o f W or k Zo ne Daytime Nighttime Figure 7. Increased crash costs with active work and no lane closure present. $0 $5,000 $10,000 0 50000 100000 150000 200000 250000 Freeway AADT In cr ea se d Cr as h Co st s pe r 1 00 H ou rs o f W or k Zo ne In ac tiv ity p er M ile o f W or k Zo ne Daytime Nighttime Figure 8. Increased crash costs with inactive work zone and no temporary lane closures present.
⢠Fixed-object (generally single-vehicle) collisions, and ⢠other remaining crash types. The results of the analysis of collision types are provided in the following section. Rear-End Collisions As noted in the background section, several studies have indicated that rear-end crashes tend to be overrepresented in work zones. This was previously confirmed in the NYSDOT crash data analysis in Chapter 2, especially when lane closures are present in the work zone. Typically, it is hypothesized that the overrepresentation of such crashes occurs because of in- creased traffic congestion and queues associated with the re- duction in roadway capacity in the work zone. The HSIS data collected in this project allow for a more thorough investiga- tion of this hypothesis. Table 17 presents the percentage of crashes that involved a rear-end collision by work condition and time of day across the entire range of projects contained in the four-state dataset. The second column (active work with lane closures) represents the greatest work zone capacity reduction and thus the highest potential for congestion and queuing. It would be expected to have the highest percentage of rear-end crashes associated with it. The third column (active work with no lane closures) represents the next most significant capacity reduction, due to driver rubber-necking, work vehicle inter- ference, and lesser geometric restrictions. The fourth column (no work activity and no lane closures) would be expected to produce the least capacity reduction and thus the lowest per- centage of rear-end crashes. Finally, the percentage of rear- end crashes across all project locations prior to the start of work is presented in the fifth column as an indication of non- work conditions for comparison purposes. The above expectations are generally confirmed in Table 17 for night work crashes. Active night work with lane closures re- sulted in 38.4 percent rear-end crashes, compared to 33.6 per- cent during active work without lane closures and 26.0 percent with no active work or lane closures. The percentage of rear- end crashes during periods of no active work at night was iden- tical to the corresponding pre-construction percentage. However, this trend is not exhibited by crashes during day- time periods. Instead, the percentage of rear-end crashes is nearly consistent and is actually lowest when active work ac- tivity and lane closures are present. In other words, there does not appear to be a strong association between the capacity re- ductions associated with lane closures and the likelihood of a rear-end crash. In fact, the percentage of rear-end crashes in the work zone is actually a little lower than in the before work zone condition at these sites. The effects of work zone conditions on rear-end crashes were further examined by stratifying the crash data by AADT. In Figure 9, rear-end crash percentages are provided by AADT level before the work zone was present and in the work zone during periods of inactive work with no lane closures. For these conditions, rear-end crashes typically increase as a function of AADT in both daytime and nighttime periods and are lower at night than during the day across the entire range of AADTs. Also, the difference between daytime and nighttime rear-end crash percentages increases with AADT. Furthermore, for both day and night, the percentage of rear-end crashes in the work zone is very similar to the before- construction condition. At higher AADTs, a small increase (approximately 5 percent) in rear-end crashes is seen when the work zone is present. The rear-end crash percentages during periods of active work with and without temporary lane closures are provided in Figure 10 and Figure 11 for daytime and nighttime, re- spectively. During the day, a significant increase in rear-end crashes is evident at AADT levels below 100,000 vpd when the work zone is active regardless of whether or not a lane closure is present. At higher AADTs, however, rear-end crashes when the work zone is active are about the same or even lower than when work is inactive. This may be partly attributable to small sample sizes at higher AADTs for the active work with lane closure condition. These results may indicate that there is an upper limit in terms of how much of the total crash experience at a location will be rear-end crashes. These high-volume locations may already experience so much congestion and stop-and-go traffic (which lead to rear-end crashes) that further degradation in operating conditions associated with the work zone simply results in the same distribution of crash types that normally exist on that facility. At lower AADT levels, rear-end colli- sions do not normally comprise the majority of crashes that occur, and so the introduction of capacity reductions and other turbulence on the roadway leads to more congestion 27 Time of Day Active Work with Lane Closures Active Work without Lane Closures No Active Work, No Lane Closures No Work Zone Present Daytime Periods 46.9% 54.4% 48.7% 52.8% Nighttime Periods 38.4% 33.6% 26.0% 26.0% Table 17. Percent of rear-end crashes.
and unexpected traffic, resulting in a greater proportion of rear-end collisions. Authors of past studies have commonly assumed that the increase in rear-end crashes in work zones is primarily asso- ciated with unexpected congestion and traffic queues created by lane closures. As was shown in Figure 10, however, rear- end crash percentages during work activities without lane closures are almost identical to when a temporary lane closure is present. While it is possible that there are a few in- stances in which the project diaries failed to note a temporary lane closure, resulting in incorrectly coding the work zone as having no lane closure present, these instances are not believed to be frequent enough to explain the close agreement between the lane closure and no lane closure conditions. Rather, this close similarity in rear-end crashes between lane closure and no lane closure conditions appears to indicate that other work zone features and conditions also affect traf- fic flow during periods of work activity. These features and conditions may include construction traffic entering or exit- ing the workspace, actions by workers or equipment near the travel lanes that cause nearby motorists to brake unexpect- edly, overall changes to roadway geometry, etc. These condi- tions and work zone features appear to contribute as much to the increased crash risk of the work zone as a temporary lane closure. The rear-end crash trends by AADT for nighttime work are more consistent with expectations. Active work lane closures result in more rear-end crashes than during periods of inactive 28 0% 10% 20% 30% 40% 50% 60% 70% 80% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a R ea r- En d Co lli si on Daytime No Work Zone (Before) Daytime Work Zone Inactive Nighttime No Work Zone (Before) Nighttime Work Zone Inactive Figure 9. Comparison of roadway AADT to rear-end collision percentages, no work zone versus inactive work zone conditions. 0% 10% 20% 30% 40% 50% 60% 70% 80% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a R ea r- En d Co lli si on Work Zone Active with Temporary Lane Closures Work Zone Active without Temporary Lane Closures Work Zone Inactive Figure 10. Comparison of work activity and roadway AADT to rear-end collision percentages, daytime work periods.
work over the entire AADT range. Further, the percentage of rear-end crashes during active work but without a lane closure is also somewhat higher than during periods of work inactivity over the same AADT range. Finally, rear-end crashes during active work without lane closures are less than active work with lane closures, except at the lower range of AADT. Simi- lar to the discussion for daytime conditions, there appear to be many sources of traffic disruptions during periods of work activity with and without temporary lane closures at night that contribute to an increase in rear-end crashes. While other effectsâsuch as construction vehicle and equipment access and egress, distractions due to workers or equipment near the travel lanes, and so forthâcannot be determined with the current dataset, it is reasonable to expect that driver inatten- tion, which is believed to be a greater concern at night, also is a factor. Overall, it seems reasonable to believe that capacity reductions associated with lane closures contribute some to the increase in rear-end crashes in active nighttime work zones, but other factors may also reasonably be believed to contribute. All of these contributing factors should be con- sidered when work zone designers look for opportunities to improve work zone traffic safety. Sideswipe Collisions Sideswipe crashes are summarized in Table 18, which shows that daytime versus nighttime work, work activity, and lane closure presence did not dramatically influence the per- centage of sideswipe crashes, especially during daytime. Side- swipe collisions comprised between 13 and 16 percent of crashes at the project locations under all conditions except for nighttime active work without lane closures. For that group, sideswipe crashes comprised 21 percent of the total. How- ever, none of these differences are statistically significant. Fixed-Object Collisions Table 19 illustrates that fixed-object collisions consis- tently comprise a greater proportion of nighttime crashes than daytime crashes. Also, fixed-object crashes were almost identical for the no work zone present (before) condition and the inactive work zone condition both during the day and at night. Figure 12 presents fixed-object crashes com- pared to AADT, and it is apparent that fixed-object crashes for both the before and inactive periods decrease markedly 29 0% 10% 20% 30% 40% 50% 60% 70% 80% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a R ea r- En d Co lli si on Work Zone Active with Temporary Lane Closures Work Zone Active without Temporary Lane Closures Work Zone Inactive Figure 11. Comparison of work activity and roadway AADT to rear-end collision percentages, nighttime work periods. Time of Day Active Work with Lane Closures Active Work without Lane Closures No Active Work, No Lane Closures No Work Zone Present Daytime Periods 13.6% 14.8% 14.8% 14.1% Nighttime Periods 15.8% 21.0% 15.0% 13.3% Table 18. Percent of total crashes that involve sideswipe collisions.
as a function of roadway AADT in daytime, and to a lesser degree at nighttime. The change in fixed-object crashes with work activity is ex- amined in Figure 13 for daytime conditions and Figure 14 for nighttime conditions. Generally speaking, no clear trends are evident with regards to the percentage of fixed-object crashes that occurred at different AADT levels. As Figure 13 indicates, the percentage of such collisions during periods of daytime work activity ranged between 10 and 25 percent across most of the AADT levels shown, with no clear trend evident. At night, the percentage of fixed-object collisions and AADT levels during work activity also did not demonstrate any clear trends (see Figure 14). Other Vehicle Collision Types The remaining crashes not previously categorized were con- solidated into an âother crashesâ category and examined for trends across the same time periods and work conditions as previously performed for the other categories. The results of this examination are shown in Table 20. Overall, the percentage of crashes that fall into this remaining category is slightly less during daytime conditions than during nighttime conditions. However, no clear trends exist in the percentages across the work conditions examined in either time period. During the day, the percentages range between 14 and 21 percent; at night, the percentages range between 23 and 28 percent. Neither of these ranges includes a statistically significant dif- ference on the basis of work condition. Summary The following is a summary of key findings from the analy- sis of traffic crashes from 64 work zone projects across four states: ⢠Overall, when work activity is occurring and travel lanes are temporarily closed, the risk of a crash to a motorist traveling through the work zone increased by about 66 per- cent during daytime conditions and by 61 percent during nighttime conditions, compared to the expected crash risk that would normally exist at a particular location. ⢠The actual change in crash risk in these work zones varied substantially from project to project, even when stratified on the basis of time period (daytime or nighttime) and work condition (no work activity, active work without lane 30 Time of Day Active Work with Lane Closures Active Work without Lane Closures No Active Work, No Lane Closures No Work Zone Present Daytime Periods 20.3% 10.3% 15.9% 15.3% Nighttime Periods 22.8% 21.0% 31.9% 32.4% Table 19. Percent of total crashes that involve fixed-object collisions. 0% 10% 20% 30% 40% 50% 60% 70% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a F ix ed - O bje ct Co llis ion Daytime No Work Zone (Before) Daytime Work Zone Inactive Nighttime No Work Zone (Before) Nighttime Work Zone Inactive Figure 12. Comparison of roadway AADT to fixed-object collision percentages, no work zone versus inactive work zone conditions.
31 0% 10% 20% 30% 40% 50% 60% 70% 0 500000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a Fi xe d- O bje ct Co llis ion Work Zone Active with Temporary Lane Closures Work Zone Active without Temporary Lane Closures Work Zone Inactive Figure 13. Comparison of work activity and roadway AADT to fixed-object collision percentages, daytime work periods. 0% 10% 20% 30% 40% 50% 60% 70% 80% 0 50000 100000 150000 200000 Roadway AADT Pe rc en t o f C ra sh es T ha t I nv ol ve a F ix ed - O bje ct Co llis ion Work Zone Active with Temporary Lane Closures Work Zone Active Without Temporary Lane Closures Work Zone Inactive Figure 14. Comparison of work activity and roadway AADT to fixed-object collision percentages, nighttime work periods. Time of Day Active Work with Lane Closures Active Work without Lane Closures No Active Work, No Lane Closures No Work Zone Present Daytime Periods 19.2% 20.6% 14.1% 17.7% Nighttime Periods 23.1% 24.4% 25.2% 28.3% Table 20. Percent of total crashes that involve all other collision types combined.
closures, or active work with lane closures). Crash risks increased on some projects and decreased on others, com- pared to the expected values. Furthermore, no relationship appears to exist between the change in crash risk and road- way AADT. ⢠When work was active and lane closures were in place, severe crashes increased by 42.3 percent when the work was done at night and by 45.5 percent during the day. For PDO crashes, the increase was 74.8 percent at night and 80.8 percent during the day. ⢠When work activity was occurring but no temporary lane closures were used, the increase in severe crashes com- puted at night was higher (41.4 percent) than during the day (17.4 percent). Similarly, the increase computed in PDO crashes during work activity at night (66.6 percent) is greater than that during the day (39.8 percent increase). However, neither of these differences was found to be sta- tistically significant due to the high project-to-project vari- ability in the results. Still, it appears that working at night with no lane closures in place may be affecting crash risks more than was previously known. It is not clear from the data whether the greater nighttime crash risk during these work conditions is the result of the following factors: â Work area lighting glare that work crews might not be mitigating well when they are not located in travel lanes; â More frequent construction equipment and material deliveries into and out of the work area at night that cre- ate large speed differentials and subsequent crashes; or â Issues that continue to plague drivers encountering a work zone at night (lack of expectancy, poorer visibil- ity, increased levels of impairment, etc.) regardless of whether or not a travel lane is closed. ⢠When the work was inactive and no lane closures were present, the increase in injury crashes was slightly higher for nighttime conditions (11.4 percent) than for daytime conditions (2.0 percent), but this difference is not statisti- cally significant. For PDO crashes, the increase at night (33.0 percent) was significantly higher than that during the day (19.6 percent). The slightly greater increases at night presumably reflect somewhat degraded geometric condi- tions in the work zone relative to a pre-work zone condi- tion thatâcoupled with nighttime-specific issues such as limited visibility, less attentive drivers and so forthâraise nighttime crash risk more than daytime conditions do when work is inactive. ⢠For each of the work conditions examined, the increases in crash risk are higher for the PDO crashes than for the in- jury crashes, indicating that the additional crashes that do occur due to the work zone tend to be less severe in nature. This trend exists regardless of whether the work is per- formed during the day or at night. This is consistent with previous studies that found similar results. ⢠The increased costs of work zone crashes, compared to expected crash costs based on the pre-construction crash history, were consistently lower for nighttime work than daytime work when the work was active and a lane closure was in place. This is true for the entire range of AADTs examined, and the difference between day and night was substantial for higher AADTs. This means that the overall safety impacts to the motoring public of work activities that involve temporary lane closures tend to always be less at night, and the benefit of working at night increases as AADTs increase. ⢠For work activities that do not involve a temporary lane closure, there appears to be little difference in working dur- ing the day or at night in terms of increased crash costs generated. The increased crash costs at night are actually slightly higher than during the day at lower AADT levels but slightly lower at higher AADT levels. ⢠The increase in crash costs when the work is inactive and no temporary lane closures are required is slightly higher at night than during the day across the range of AADTs examined, although these differences are not statistically significant. For both daytime and nighttime periods, the increased crash costs when work zones are present but with no work activity are much less, at any AADT, than when work is active, whether or not a lane closure is present. ⢠In terms of work zone crash characteristics, the percent of crashes involving rear-end collisions typically increases as a function of AADT in both daytime and nighttime periods, although the percentages remain substantially lower in the nighttime periods for the higher AADT regions. Further- more, for both time periods, the percentage is very similar between the before (no work zone) and work zone inactive conditions. ⢠The effect of active work during the day with or without lane closures on rear-end collisions is not consistent across all AADT ranges. Rear-end collisions increase markedly during work activity on low- to moderate-volume road- ways, but not on higher-volume roadways. There may exist an upper limit in terms of how much of the total crash experience at a location will be the result of rear-end collisions. ⢠At night, work activity resulted in an increase in the per- centage of crashes that are the result of rear-end collisions across all roadway AADTs. The effect is somewhat greater when temporary lane closures are in place than when they are not, consistent with expectations. ⢠Overall, the percent of sideswipe collisions was not affected by time period, work activity, or lane closure presence. Sideswipe collisions accounted for between 13 and 21 per- cent of crashes occurring in the work zone. ⢠Fixed-object collisions consistently comprise a greater pro- portion of nighttime crashes than daytime crashes. Also, 32
fixed-object collisions comprise an almost identical per- centage of crashes between the before (no work zone) con- dition and the work zone inactive condition. Fixed-object collision involvement in crashes for both of those condi- tions decreases significantly as a function of roadway AADT in the daytime period and to a lesser degree in the nighttime period. ⢠Overall, the percentage of all remaining crash types is slightly less during daytime conditions than during night- time conditions. No clear trends exist in the percentages across the work conditions examined in either time period. During the day, the percentages range between 14 and 21 percent; at night, the percentages range between 23 and 28 percent. 33
