Estimating the Safety Effects of Work Zone Characteristics and Countermeasures: A Guidebook (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.

5 Rationale Sometimes, an agency or contractor will have the need to estimate how many crashes are expected to occur in a work zone, or how many additional crashes are expected above and beyond what would normally occur on that roadway segment over the duration of the work zone. Consider the following examples: • An agency might want to compare the cumulative number of crashes actually occurring at a work zone over time with a predicted rate of crashes expected. If more crashes begin to accrue in the work zone than what the estimate pre- dicted, that may be a cue to the agency or contractor that a safety issue exists within the work zone and that additional investigation into the reasons for those crashes may be warranted. • A planning-level estimate of the number of additional crashes expected at a work zone could also be useful to include as part of safety costs in road user cost computations when developing TMPs. • Finally, planning-level work zone crash estimates will often be needed when comparing alter- native design or operational strategies through a trade-off analysis because CMFs available for assessing the incremental effect of those alternatives require a baseline work zone crash estimate as a key input. In this guidebook, two approaches are presented for generating overall planning-level work zone crash estimates: 1. Applying an overall work zone CMF to a pre-work-zone baseline estimate of crashes expected on the roadway segment where the work zone will occur. The pre-work-zone baseline may come from existing agency safety performance functions (SPFs) or another estimate of crash frequency on the route when a work zone is not present. 2. Utilizing a general SPF that has been created using work-zone-specific data. Limitations Significant limitations and caveats exist with the use of each approach. Both the overall work zone CMF and the general work zone SPF developed as part of this research effort were based on long-term Interstate and freeway work zone projects with the following characteristics before and during the work zones themselves: • Pavement width = 40 ft in each direction for four-lane segments and 52 ft in each direction for six-lane segments (equal to 12-ft lanes, 6-ft inside shoulder, and 10-ft outer shoulder) • No lane shifts present C H A P T E R 2 Planning-Level Work Zone Crash Estimation Procedures Uses of Planning-Level Work Zone Crash Estimates • Real-time crash frequency monitoring • Road user cost computations • Baseline for CMF analyses

6 Estimating the Safety Effects of Work Zone Characteristics and Countermeasures: A Guidebook • No long-term lane closures present • Median width of 60 ft, which includes the inside shoulder width of 6 ft in both directions • No longitudinal barriers present The more that the pre-work-zone baseline conditions and work zone conditions differ from these ideal attributes, the more that the results of these planning-level analysis tech- niques will likely deviate from what actually occurs during the work zone. The practitioner must recognize the level of uncertainty inherent in the results obtained from these analyses. The results do not attempt to account for differences in work activities, work space opera- tions, frequency of temporary lane closures, or other features that are believed to have some influence on crashes at a location during a work zone. The results also do not attempt to differentiate between severe (injury and fatal) crashes and property-damage-only (PDO) crashes. Finally, the results reflect work zones that were in place for several weeks at a time, which may or may not accurately reflect the effects of short-term or short-duration work zones. Nevertheless, for those agencies that do not have locally calibrated SPFs or work zone CMFs, these procedures provide at least an order-of-magnitude approximation of work zone safety impacts. It is suggested that application of these procedures be limited to Interstate and freeway facili- ties, and potentially multilane divided highways with little or no driveway access and only infre- quent non-signalized intersections. There has not been enough research conducted to date of work zones on other types of roadways to yield meaningful work zone SPFs or generic work zone CMFs. It is possible to perform some simplistic trade-off analyses of certain temporary changes to the roadway during the work zone, relying on baseline pre-work-zone SPFs, but the practitioner will again need to recognize the level of uncertainty associated with such analyses on these types of facilities. Method 1: Using Pre-Work-Zone Crash Estimates and an Overall Work Zone CMF For developing planning-level work zone crash estimates, the preferred approach is to apply an overall generic work zone CMF to the pre-work-zone baseline crash estimate for the roadway segment of interest. Although using a work-zone-based SPF to predict the number of crashes expected in a work zone may be simpler, estimates generated in that way do not take site-specific factors (other than the annual average daily traffic [AADT] value) into consideration. Therefore, applying an overall work zone CMF to a good pre-work-zone crash estimate would be expected to yield a better work zone crash estimate. Ideally, the normal (pre-work-zone) crash estimates would come from an SPF calibrated to the particular roadway segment utilizing methods con- tained in the HSM or agency-calibrated crash-prediction models. However, if that data is not available, the analyst may have to use other estimation methods. For example, the last 3 years of crashes occurring along the roadway segment could be used along with AADT values each year to determine a weighted yearly average of crashes. An overall work zone CMF (WZCMF) for freeways and Interstate facilities has been developed from a multi-state dataset of four- and six-lane freeway and Interstate work zones. The CMF is based on a ratio of pre-work-zone and during-work-zone SPFs developed for those roadway segments. The ratios of those SPFs are given in the following equations: Four-Lane Facilities = ( ) ( ) − + − + 4- 10.036 1.164 ln 11.231 1.248ln WZCMF e e lanes AADT AADT

Planning-Level Work Zone Crash Estimation Procedures 7 Six-Lane Facilities = ( ) ( ) − + − + 6- 9.987 1.164 ln 12.318 1.344 ln WZCMF e e lanes AADT AADT These equations result in very high CMFs at very low volumes, and taper off as traffic volumes increase. This is illustrated graphically in Figure 4. Even though the CMFs are higher at lower volumes, the number of additional crashes per mile per year that occur due to a work zone are lowest at the lower volumes and increase non-linearly as volumes increase. To compute the total number of crashes expected during a work zone, the number of crashes normally occurring on the roadway segment each year is multiplied by the duration of the work zone and the overall work zone CMF calculated for the AADT of the roadway segment. If a crash rate is used, the rate is first multiplied by the length of the project, as shown following equation: ExpectedWZ crashes Non WZ crashes mile year Project length WZ duration mo WZCMF - 12 ( ) ( ) ( )=     Method 2: Using a Work-Zone-Based SPF If good data does not exist for the crash frequency that normally occurs on a section of freeway or Interstate where a work zone will be placed, a work-zone-based SPF can be used to develop a planning-level estimate of crashes expected during the work zone. The multi-state database of work zones performed on four- and six-lane Interstates and freeways was used to develop the following two predictive functions of the total number of work zone crashes expected to occur based on work zone length, work zone duration, and overall roadway AADT (5): Four-Lane Freeway and Interstate Work Zones Number of work zone crashes expected L n e AADT10.036 1.164ln= × × ( )− + 1 1.2 1.4 1.6 1.8 Ca lc ul at ed W or k Zo ne C M F AADT, veh/day 4-Lane Freeways/Interstates 6-Lane Freeways/Interstates 1800000 20000 40000 60000 80000 100000 120000 140000 160000 Figure 4. Planning-level work zone CMFs for freeways and Interstates.

8 Estimating the Safety Effects of Work Zone Characteristics and Countermeasures: A Guidebook Six-Lane Freeway and Interstate Work Zones Number of work zone crashes expected L n e AADT9.987 1.164ln= × × ( )− + where, L = length of work zone, miles n = number of years the work zone will require (or number of months/12) The functions were developed with the following work zone conditions: • Pavement width = 40 ft in each direction for four-lane segments and 52 ft in each direction for six-lane segments (equal to 12-ft lanes, 6-ft inside shoulder, and 10-ft outer shoulder) • No lane shifts present • No lane closures present • Median width of 60 ft, which includes the inside shoulder width of 6 ft in both directions • No longitudinal barriers present • AADTs ranging between 5,000 and 70,000 vehicles per day (vpd) on the four-lane segments • AADTs ranging between 50,000 and 150,000 vpd on the six-lane segments Additional details regarding the development of these models (such as the standard errors of the parameters and overdispersion factors) are available in NCHRP Web-Only Document 240. Figure 5 illustrates these functions on a per-mile, per-year basis. One sees that the two functions are non-linear and nearly identical within the 50,000 to 70,000 vpd range where they overlap; however, similar work-zone-based functions do not exist for other roadway types at the present time. Examples of Computing Planning-Level Work Zone Crash Estimates The following examples illustrate how to calculate planning-level work zone crash estimates using these methods for a variety of possible purposes. 0 20 40 60 Es ti m at ed W or k Zo ne C ra sh es /M ile /Y ea r AADT, veh/day 4-Lane Freeways/Interstates 6-Lane Freeways/Interstates 0 20000 40000 60000 80000 100000 120000 140000 Figure 5. Work zone SPFs for freeways and Interstates.

Planning-Level Work Zone Crash Estimation Procedures 9 Computing an Expected Crash Rate per Month During Construction A project engineer plans to monitor crashes occurring during a 2-year, 3-mile Interstate wid- ening construction project. The engineer will extract crashes from the statewide crash data- base for the roadway segment each month during the project, and wants to be able to detect if the number of crashes that occur becomes unusually high. The roadway has the following characteristics: • Rural four-lane Interstate facility with typical Interstate standards (12-ft lanes, 6-ft inside shoulder, 10-ft outside shoulder, wide median) • Expected traffic volume on the facility: 42,000 vpd during year 1 of the project and 45,000 vpd during year 2 • Based on a calibrated SPF developed by the agency, the normal non-work-zone crash rate on this facility is estimated to be 6.9 crashes per mile in year 1 and 7.4 crashes per mile in year 2 The project engineer decides to create a plot of cumulative crashes expected to occur over the duration of the project. Because good calibrated pre-work-zone crash models exist for the road- way segment, the project engineer chooses to use the overall WZCMF to estimate the number of crashes expected to occur during the project. A separate CMF is first computed for each of the 2 years of the project based on the following expected traffic volumes: = = ( ) ( ) − + − + 1.3514- , 1 10.036 1.164ln 42,000 11.231 1.248ln 42,000 WZCMF e e lanes year = = ( ) ( ) − + − + 1.3434- , 2 10.036 1.164ln 45,000 11.231 1.248ln 45,000 WZCMF e e lanes year The project engineer then multiplies the normal expected crash rate each year by the length of the project, the duration of 1 year, and the work zone CMF computed for that year: WZ crashes expected crashes mile year miles year crashesyear 6.9 3 1 1.351 28.01 ( )( ) ( )=   = WZ crashes expected crashes mile year miles year crashesyear 7.4 3 1 1.351 29.82 ( )( ) ( )=   = 28.0 29.8 57.8WZ crashes expected crashestotal = + = In this example, the project engineer chose to utilize the AADT value for the entire year, and then divide the annual number by 12 to get a month-by-month crash estimate. If a more refined analysis is desired, seasonal factors can be applied to the AADT factors and seasonal WZCMFs can be developed and applied to determine the WZ crashes expected month-by-month of each season of each year. Crashes actually occurring during the project could then be compared directly with the expected crashes month-by-month (or whatever time period was of interest to the agency) as well as cumulatively. Large deviations between the actual and expected crashes might be an indication that a safety issue exists at the site and additional investigation is needed. For example, if the number of crashes actually occurring during the first 9 months of the project were as plotted against the expected number of crashes in Figure 6, the agency might conclude that actual crashes had tracked fairly closely to what was expected in the early months. However, crashes during the latest 2 months could be interpreted as trending somewhat higher than what would be expected. Based on this assessment, the project engineer might initiate a

10 Estimating the Safety Effects of Work Zone Characteristics and Countermeasures: A Guidebook 0 5 10 15 20 25 30 35 Cu m ul ati ve N um be r of C ra sh es t o D at e Month of Project Expected Actual 0 9 10 11 121 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 N um be r of W or k Zo ne C ra sh es Ea ch M on th Expected Actual Month of Project 9 10 11 121 2 3 4 5 6 7 8 Figure 6. Expected versus actual work zone crashes using the work zone CMF method for this example. more in-depth review of the crashes to determine potential reasons for the uptick in crashes. Perhaps the project has had a major traffic switch in recent months, work activities have involved more frequent deliveries of materials to the job site, or weather conditions have been particularly poor during this time. Now, suppose that the project engineer does not have access to good non-work-zone crash data for the segment. In that case, the work zone SPF for four-lane facilities can have been applied instead. The computations would be as follows: 3 1 31.61 10.036 1.164ln 42,000WZ crashes expected miles year e crashesyear ( )( )( )= =( )− + 3 1 34.32 10.036 1.164ln 45,000WZ crashes expected miles year e crashesyear ( )( )( )= =( )− + 31.6 34.3 65.9WZ crashes expected crashestotal = + = Use of the work zone SPF would have yielded an estimate of 65.9 crashes expected over the 2-year project (2.75 crashes per month), approximately 14% higher than what was computed using the calibrated pre-work-zone crash rates and the overall work zone CMF. Plotting the

Planning-Level Work Zone Crash Estimation Procedures 11 actual crashes through 9 months of the project against this estimate (as shown in Figure 7) may or may not have led the engineer to conclude that crashes were becoming excessive relative to what was expected. The difference in results by the two methods is another reminder of the importance of engineering judgment when interpreting and using planning-level estimates such as these. Estimating the Effect of Accelerated Construction on the Expected Number of Work Zone Crashes An agency is contemplating including contract incentives in a bid package to reduce the overall duration of the project and is trying to come up with a rational, defensible value for the incentive. The work zone design team believes that by using traditional methods, the project would take 2 years to complete. If the winning contractor could reduce the duration of the project by 6 months, the agency would like to estimate how much that acceleration of the project could reduce crashes. The roadway has the following characteristics: • Project is 4 miles long. • It is an urban six-lane freeway facility. • The traffic volume on the facility is expected to be approximately 120,000 vpd for year 1 of the project and 130,000 vpd for year 2 of the project. • The freeway has 12-ft lanes, 6-ft inside shoulders, and 10-ft outside shoulders. • The total crash density on this section of freeway is 32.6 crashes per mile per year before construction based on 3 years of historical data. Traffic volumes during those years averaged 110,000 vpd (similar to what is expected for year 1 of the project). In this example, the agency is interested in the difference in crashes that would be expected to occur under the original 2-year project duration to the number that would be expected if the project were reduced to 18 months followed by 6 months of normal (non-work-zone) crash frequency. Unlike the previous example, the agency does not have a normal non-work-zone expected crash frequency for the roadway for each year of the project. Consequently, it will be necessary to estimate the normal pre-work-zone crash frequency for each of the 2 years of the 0 5 10 15 20 25 30 35 Cu m ul ati ve N um be r of C ra sh es t o D at e Month of Project Expected Actual 0 9 10 11 121 2 3 4 5 6 7 8 Figure 7. Expected versus actual crashes using the work zone SPF for this example.

12 Estimating the Safety Effects of Work Zone Characteristics and Countermeasures: A Guidebook project based on the data available. Since the only data available for use is the historical crash rate for the roadway segment associated with a lower AADT than that anticipated during the project, the agency would first factor the crash rate for the 2 years of the project using the following ratio of AADT numbers: Year non work zone crash rate crashes mile year crashes mile year 1 - - 32.6 120,000 110,000 35.6 =     = Year non work zone crash rate crashes mile year crashes mile year 2 - - 32.6 130,000 110,000 38.5 =     = This factoring process assumed a linear relationship between crashes and AADT, which is typically not true. However, in the absence of local safety performance functions, it is considered to be a plausible planning-level assumption. Once the agency has a predicted non-work-zone crash rate for each year, work zone CMFs are then computed for each of the 2 years of the project based on the following expected traffic volumes: 1.2536- , 1 9.987 1.164ln 120,000 12.318 1.344ln 120,000 WZCMF e e lanes year = = ( ) ( ) − + − + 1.2356- , 2 9.987 1.164ln 130,000 12.318 1.344ln 130,000 WZCMF e e lanes year = = ( ) ( ) − + − + The total number of crashes expected for each alternative can then be calculated. For Alternative 1: Year expectedWZ crashes crashes mile year miles year crashesAlt1 35.6 4 1 1.253 178.41 ( )( ) ( )=   = Year expectedWZ crashes crashes mile year miles year crashesAlt2 38.5 4 1 1.235 190.21 ( )( ) ( )=   = 178.4 190.2 368.61Total expectedWZ crashes crashesAlt = + = For Alternative 2, the expected number of crashes for year 1 of the project would remain the same. For the second year, the first 6 months would be at the expected work zone crash rate, and the second 6 months would be at the non-work-zone crash rate: Year expectedWZ crashes crashes mi yr mo mo yr crashes mi yr mo mo yr miles crashesAlt2 38.5 6 12 1.235 38.5 6 12 4 172.12 ( ) ( )=     +           = 178.4 172.1 350.52Total expectedWZ crashes crashesAlt = + = ExpectedWZ crashes difference crashesAlt Alt 368.6 350.5 18.11 2 = − =− Reducing the duration of the project by 6 months would be expected to result in 18.1 fewer crashes over the non-accelerated project schedule. If desired, the agency could apply

Planning-Level Work Zone Crash Estimation Procedures 13 comprehensive crash cost numbers to this value to estimate the road user safety cost savings that could be attributed to this reduction. For example, if the agency typically experiences a crash severity distribution on the facility similar to that shown in Table 1 and used typical crash cost values recommended in the HSM, reducing the project duration would be estimated to yield nearly $900,000 in crash cost savings. This would be in addition to any mobility-related savings that might also be achieved. Crash Severity Level Proportional Distribution of Crash Severities Proportion of the 18.1 Crashes Reduced Average Crash Cost1 Crash Costs Saved If Project Is Accelerated Fatality (K) 0.005 0.0905 $4.509,991 $408,154 Disabling injury (A) 0.018 0.3258 $242,999 $79,169 Evident injury (B) 0.088 1.5928 $88,875 $141,560 Possible injury (C) 0.136 2.4616 $50,512 $124,340 Property damage only (PDO) 0.753 13.6293 $8,325 $113,464 TOTAL 1.000 18.1 $866,987 1Crash Costs in the Highway Safety Manual, First Edition, updated to 2016 dollars. Table 1. Estimated crash cost savings if project duration can be reduced for this example.