Application of Crash Modification Factors for Access Management, Volume 1: Practitioner's Guide (2021)

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.

50 Chapter 4 introduces the general methods for quantifying safety performance at the segment and intersection level. The segment- and intersection-level predictive method provides a struc- tured approach to estimate crash frequency, by crash type and severity, for urban and suburban segments and intersections. This method is consistent with the Highway Safety Manual (1st Edi- tion) (AASHTO 2010) but expands the Part C Predictive Method to incorporate the safety impacts of additional access management strategies. The segment- and intersection-level analysis methodologies for estimating the effects of access management variables are essentially the same as those in the Highway Safety Manual (1st Edition) Part C Predictive Method (and proposed 2nd Edition). Those methodologies already facilitate the consideration of a limited number of access management variables. For segment-level predictions, the existing method accounts for the number and type of driveways along the segment. For intersection-level predictions, the existing method accounts for the presence of left- and right-turn lanes, left-turn signal phasing (at signalized intersections), and right-turn-on-red restrictions (at signalized intersections). NCHRP Project 17-74 research suggests that the Highway Safety Manual (1st Edition) Part C Predictive Method performs relatively well across the range of other access management features not accounted for in the existing predictive method. In other words, the existing Part C Predictive Method (i.e., combi- nation of SPFs and CMFs) performs well for sites with similar geometry, but different access management features such as median opening spacing, number of median openings by type, and corner clearance along a segment. There are, however, a few scenarios where the existing models do not perform well across sites with different access management features. The remainder of this chapter presents the recommended predictive method, access man- agement features to be considered, and examples that demonstrate how to use the predictive method with different levels of data, as well as nuances and limitations. This chapter is only applicable to estimating the safety performance of individual segments and intersections, assuming independence among each unit of analysis. While the results can be aggregated from multiple segments and intersections to estimate the safety performance of a corridor, as suggested in the Highway Safety Manual (1st Edition), this method does not consider the poten- tial interactions among adjacent or nearby sites (e.g., access spacing and density). The existing Part C Predictive Method may even produce counterintuitive results (e.g., fewer estimated seg- ment crashes with an increase in the number of intersections along a corridor). As such, the corridor-level predictive method presented in Chapter 5 of this guide is more appropriate for considering interactions among access management features and estimating safety per- formance at the corridor level. Refer to Chapter 5 for further guidance on when to combine the results from segment- and intersection-level analysis (from Chapter 4) and when to use a corridor-level prediction model for corridor-level analysis. C H A P T E R   4 Predictive Method for Segment- and Intersection-Level Analysis

Predictive Method for Segment- and Intersection-Level Analysis 51   e Part C Predictive Method is an 18-step process, as dened in the Highway Safety Manual (1st Edition) and illustrated in Figure 16. is chapter focuses on the use of the predictive method to estimate the safety performance of urban and suburban segments and intersections under dierent access management scenarios. Specically, this chapter focuses on Step 9 (Select and apply SPF) and Step 10 (Apply CMFs) of the 18-step process. e remainder of the Part C Predictive Method remains the same. Source: Adapted from AASHTO 2010. Figure 16. 18-step process for Highway Safety Manual (1st Edition) Part C Predictive Method.

52 Application of Crash Modification Factors for Access Management In general, the equation shown in Figure 17 represents the Part C Predictive Method for urban and suburban arterials. This equation includes five main components: (1) SPF prediction for base conditions, (2) CMF(s) to adjust the SPF prediction to reflect conditions other than the base conditions, (3) predicted pedestrian crashes, (4) predicted bicycle crashes, and (5) calibra- tion factor. Source: AASHTO 2010. Figure 17. Part C Predictive Method for urban and suburban arterials. Variables for the equation shown in Figure 17 are defined as follows: • Npredicted = predicted average crash frequency for a specific year on site type x. • Nspf x = predicted average crash frequency determined for base conditions of the SPF devel- oped for site type x. • Npedx = predicted average number of vehicle-pedestrian collisions per year for site type x. • Nbikex = predicted average number of vehicle-bicycle collisions per year for site type x. • CMFyx = CMFs specific to site type x and specific geometric design and traffic control features y. • Cx = calibration factor to adjust SPF for local conditions for site type x. The Part C Predictive Method applies to individual segments and intersections, assuming independence (i.e., the safety performance of a segment or intersection does not impact the safety performance of adjacent or nearby segments or intersections). Figure 18 shows an example of how segments and intersections are defined along a corridor. Segments are measured from the center of intersection to the center of intersection, assuming the segment characteristics remain consistent along the entire length. If the geometric or traffic operational elements change, then a new segment is created, and the segment length is measured as the length of the homogeneous segment. Crashes are assigned to either a segment or intersection as follows: A. All crashes that occur in region A are coded as intersection crashes and assigned to the respec- tive intersection. As shown in Figure 18, region A is the physical area of the inter section, not the functional area. As discussed in Chapter 3 and illustrated in Figure 4, the functional area extends both upstream and downstream from the physical intersection area and includes the upstream approaches where deceleration, maneuvering, and queuing take place, as well as the downstream areas beyond the intersection where driveways could introduce conflict points and generate queues backing up through the intersection. B. Crashes that occur in region B may be coded as segment- or intersection-related crashes. Crashes that occur within the functional area of the intersection and are related to the pres- ence of the intersection are coded as “intersection-related” and assigned to the respective intersection. All other crashes (i.e., those occurring between intersections and not related to the intersections) are coded as “segment-related” and assigned to the respective segment. Source: Adapted from AASHTO 2010. Figure 18. Illustration of segments and intersections for Part C Predictive Method.

Predictive Method for Segment- and Intersection-Level Analysis 53   The next section of this chapter focuses on Step 9 of the Part C Predictive Method, which involves selecting and applying an applicable SPF. Select and Apply an Applicable SPF Prior to selecting SPFs, it is important to understand the applicability of SPFs available in the Part C Predictive Method. The SPFs for urban and suburban arterials in the Highway Safety Manual (1st Edition) apply to the following site types: • Two-lane undivided roadway segment (2U). • Four-lane undivided roadway segment (4U). • Four-lane divided roadway segment (4D). • Two-lane roadway segment with TWLTL (3T). • Four-lane roadway segment with TWLTL (5T). • Three-legged minor road stop-controlled intersection (3ST). • Four-legged minor road stop-controlled intersection (4ST). • Three-legged signalized intersection (3SG). • Four-legged signalized intersection (4SG). It is also important to understand the limitations of SPFs available in the Part C Predictive Method. The SPFs for urban and suburban arterials in the Highway Safety Manual (1st Edition) do not apply to the following site types: • One-way segments. • Segments with six or more lanes. • Segments with full access control. • Segments within the limits of an interchange that has free-flow ramp terminals on the arterial of interest. • All-way stop-controlled intersections. • Intersections with more than four approaches. • Roundabouts. While the Part C Predictive Method can be used to estimate safety performance by mode (i.e., vehicle, pedestrian, and bicycle), the information related to pedestrians and bicycles is relatively limited. In most cases, the prediction of pedestrian and bicycle crashes is based on a proportion of total crashes, with the exception of three-legged and four-legged signalized intersections. The method in this guide is applicable to vehicle, pedestrian, and bicycle safety; however, the CMFs apply only to vehicle-vehicle crashes and, similar to most cases of the existing Part C Predictive Method, pedestrian and bicycle crashes are based on a proportion of total crashes. Safety Performance Functions for Segments For urban and suburban arterial segments, the general equation shown in Figure 17 is modi- fied slightly to represent segment crashes. Specifically, the term representing the predicted average crash frequency for the base conditions (Nspf x) is replaced by Nspf rs, indicating the base prediction for a roadway segment. From the Highway Safety Manual (1st Edition) and as shown in the equation in Figure 19, Nspf rs includes three components: (1) predicted number of multivehicle segment crashes under base conditions (Nbrmv), (2) predicted number of single- vehicle segment crashes under base conditions (Nbrsv), and (3) predicted number of driveway- related crashes (Ndrwy). Note the Highway Safety Manual (2nd Edition—forthcoming), based on NCHRP Project 17-62, will provide the option to further disaggregate crashes by type (e.g., head-on, rear-end, etc.). There will be an option to use the higher-level aggregated predictions or sum the disaggregated predictions and then proceed with the current method.

54 Application of Crash Modification Factors for Access Management Figure 19. Predicted segment-related crash frequency for base conditions. Variables for the equation shown in Figure 19 are defined as follows: • Nspf rs = predicted average crash frequency of an individual segment for base conditions (excluding vehicle-pedestrian and vehicle-bicycle crashes). • Nbrmv = predicted average crash frequency of multivehicle (non-driveway) segment crashes for base conditions. • Nbrsv = predicted average crash frequency of single-vehicle segment crashes for base conditions. • Ndrwy = predicted average crash frequency of multivehicle driveway-related crashes. Figure 20 presents the SPF for multivehicle (non-driveway) segment crashes for base condi- tions from the Highway Safety Manual (1st Edition). Figure 20. SPF for multivehicle (non-driveway) segment crashes. Variables for the equation shown in Figure 20 are defined as follows: • Nbrmv = predicted average crash frequency of multivehicle (non-driveway) segment crashes for base conditions. • AADT = annual average daily traffic volume (vehicles per day) for the segment and year of interest. • L = segment length (miles). • a and b = regression coefficients estimated for each specific crash severity and site type (see Table 50 for coefficients by crash severity and site type). • k = dispersion parameter associated with the SPF. This value is used in Figure B-6 (see Appendix B). Crash Severity Site Type Intercept (a) AADT (b) Dispersion Parameter (k) Applicable AADT (vehicles per day) Total 2U −15.22 1.68 0.84 0 to 32,600 3T −12.40 1.41 0.66 0 to 32,900 4U −11.63 1.33 1.01 0 to 40,100 4D −12.34 1.36 1.32 0 to 66,000 5T −9.70 1.17 0.81 0 to 53,800 Fatal and Injury 2U −16.22 1.66 0.65 0 to 32,600 3T −16.45 1.69 0.59 0 to 32,900 4U −12.08 1.25 0.99 0 to 40,100 4D −12.76 1.28 1.31 0 to 66,000 5T −10.47 1.12 0.62 0 to 53,800 PDO 2U −15.62 1.69 0.87 0 to 32,600 3T −11.95 1.33 0.59 0 to 32,900 4U −12.53 1.38 1.08 0 to 40,100 4D −12.81 1.38 1.34 0 to 66,000 5T −9.97 1.17 0.88 0 to 53,800 Source: AASHTO 2010. Table 50. Coefficients for multivehicle (non-driveway) segment SPF.

Predictive Method for Segment- and Intersection-Level Analysis 55   Figure 21 presents the SPF for single-vehicle segment crashes for base conditions from the Highway Safety Manual (1st Edition). Figure 21. SPF for single- vehicle segment crashes. Variables for the equation shown in Figure 21 are defined as follows: • Nbrsv = predicted average crash frequency of single-vehicle segment crashes for base conditions. • AADT = annual average daily traffic volume (vehicles per day) for the segment and year of interest. • L = segment length (miles). • a and b = regression coefficients estimated for each specific crash severity and site type (see Table 51 for coefficients by crash severity and site type). • k = dispersion parameter associated with the SPF. Figure 22 presents the SPF for total multivehicle driveway-related crashes from the Highway Safety Manual (1st Edition). This equation produces the total predicted driveway-related crashes. To predict the number of fatal and injury driveway-related crashes, multiply the result from the equation in Figure 22 by the proportion of fatal and injury crashes (fdrwy) from Table 52. To pre- dict the number of PDO driveway-related crashes, subtract the predicted number of fatal and injury driveway-related crashes from the predicted number of total driveway-related crashes. Figure 22. SPF for total multi- vehicle driveway-related crashes. Variables for the equation shown in Figure 22 are defined as follows: • Ndrwy = predicted average total crash frequency of multivehicle driveway-related crashes. • AADT = annual average daily traffic volume (vehicles per day) for the segment and year of interest. Crash Severity Site Type Intercept (a) AADT (b) Dispersion Parameter (k) Applicable AADT (vehicles per day) Total 2U −5.47 0.56 0.81 0 to 32,600 3T −5.74 0.54 1.37 0 to 32,900 4U −7.99 0.81 0.91 0 to 40,100 4D −5.05 0.47 0.86 0 to 66,000 5T −4.82 0.54 0.52 0 to 53,800 Fatal and Injury 2U −3.96 0.23 0.50 0 to 32,600 3T −6.37 0.47 1.06 0 to 32,900 4U −7.37 0.61 0.54 0 to 40,100 4D −8.71 0.66 0.28 0 to 66,000 5T −4.43 0.35 0.36 0 to 53,800 PDO 2U −6.51 0.64 0.87 0 to 32,600 3T −6.29 0.56 1.93 0 to 32,900 4U −8.50 0.84 0.97 0 to 40,100 4D −5.04 0.45 1.06 0 to 66,000 5T −5.83 0.61 0.55 0 to 53,800 Source: AASHTO 2010. Table 51. Coefficients for single-vehicle segment SPF.

56 Application of Crash Modification Factors for Access Management • nj = total number of driveways within the segment (both sides of road) of driveway type j. • Nj = average number of driveway-related crashes per driveway per year for driveway type j (see Table 52 for average values). • t = coefficient for traffic volume adjustment estimated for each specific site type (see Table 52 for coefficients by site type). • k = dispersion parameter associated with the SPF. The results from the equations shown in Figures 20 through 22 are combined to predict the average crash frequency of an individual segment under base conditions (from the equation shown in Figure 19). As shown in the equation in Figure 17, this estimate of predicted segment crashes for base conditions is adjusted using CMFs. The CMFs reflect the safety effects of geo- metric and traffic operational characteristics that are different from the base conditions. CMFs applicable to the Part C Predictive Method for urban and suburban arterials are presented in a section titled Select and Apply Applicable CMFs. Safety Performance Functions for Intersections For urban and suburban arterial intersections, the general equation from Figure 17 is modi- fied slightly to represent intersection crashes. Specifically, the term representing the predicted average crash frequency for the base conditions (Nspf x) is replaced by Nspf int, indicating the base prediction for an intersection. From the Highway Safety Manual (1st Edition), and as shown in the equation in Figure 23, Nspf int includes two components: (1) predicted number of multivehicle intersection crashes under base conditions (Nbimv), and (2) predicted number of single-vehicle intersection crashes under base conditions (Nbisv). Note the Highway Safety Manual (2nd Edition— forthcoming), based on NCHRP Project 17-62, will separate multivehicle crashes into same- direction, opposite-direction, and intersecting-direction crashes. The remainder of the method remains the same. Variables for the equation shown in Figure 23 are defined as follows: • Nspf int = predicted average crash frequency of an individual intersection for base conditions (excluding vehicle-pedestrian and vehicle-bicycle crashes). • Nbimv = predicted average crash frequency of multivehicle intersection crashes for base conditions. • Nbisv = predicted average crash frequency of single-vehicle intersection crashes for base conditions. SPF Component Driveway Type 2U 3T 4U 4D 5T Number of Driveway-Related Crashes per Driveway per Year (Nj) Major Commercial 0.158 0.102 0.182 0.033 0.165 Minor Commercial 0.050 0.032 0.058 0.011 0.053 Major Industrial/Institutional 0.172 0.110 0.198 0.036 0.181 Minor Industrial/Institutional 0.023 0.015 0.026 0.005 0.024 Major Residential 0.083 0.053 0.096 0.018 0.087 Minor Residential 0.016 0.010 0.018 0.003 0.016 Other 0.025 0.016 0.029 0.005 0.027 Coefficient for AADT (t) All 1.000 1.000 1.172 1.106 1.172 Dispersion Parameter (k) All 0.81 1.10 0.81 1.39 0.10 Proportion of Fatal and Injury Crashes (fdrwy) All 0.323 0.243 0.342 0.284 0.269 Proportion of PDO Crashes All 0.677 0.757 0.658 0.716 0.731 Notes: Applies only to unsignalized driveways; signalized driveways are treated as signalized intersections. Major driveways serve 50 or more parking spaces; minor driveways serve less than 50 parking spaces. Source: AASHTO 2010. Table 52. Coefficients for total multivehicle driveway-related SPF. Figure 23. Predicted intersection- related crash frequency for base conditions.

Predictive Method for Segment- and Intersection-Level Analysis 57   Figure 24 presents the SPF for multivehicle intersection crashes for base conditions from the Highway Safety Manual (1st Edition). Figure 24. SPF for multivehicle intersection crashes. Variables for the equation shown in Figure 24 are defined as follows: • Nbimv = predicted average crash frequency of multivehicle intersection crashes for base conditions. • AADTmajor = annual average daily traffic volume (vehicles per day) for the major road and year of interest (both directions of travel combined). • AADTminor = annual average daily traffic volume (vehicles per day) for the minor road and year of interest (both directions of travel combined). • a, b, and c = regression coefficients estimated for each specific crash severity and site type (see Table 53 for coefficients by crash severity and site type). • k = dispersion parameter associated with the SPF. Figure 25 presents the SPF for single-vehicle intersection crashes for base conditions from the Highway Safety Manual (1st Edition). Figure 25. SPF for single-vehicle intersection crashes. Variables for the equation shown in Figure 25 are defined as follows: • Nbisv = predicted average crash frequency of single-vehicle intersection crashes for base conditions. • AADTmajor = annual average daily traffic volume (vehicles per day) for the major road and year of interest (both directions of travel combined). • AADTminor = annual average daily traffic volume (vehicles per day) for the minor road and year of interest (both directions of travel combined). • a, b, and c = regression coefficients estimated for each specific crash severity and site type (see Table 54 for coefficients by crash severity and site type). • k = dispersion parameter associated with the SPF. Crash Severity Site Type Intercept (a) AADTmajor (b) AADTminor (c) Dispersion Parameter (k) Applicable AADT (vehicles per day) AADTmajor AADTminor Total 3ST −13.36 1.11 0.41 0.80 0 to 45,700 0 to 9,300 4ST −12.13 1.11 0.26 0.33 0 to 46,800 0 to 5,900 3SG −8.90 0.82 0.25 0.40 0 to 58,100 0 to 16,400 4SG −10.99 1.07 0.23 0.39 0 to 67,700 0 to 33,400 Fatal and Injury 3ST −14.01 1.16 0.30 0.69 0 to 45,700 0 to 9,300 4ST −11.58 1.02 0.17 0.30 0 to 46,800 0 to 5,900 3SG −11.13 0.93 0.28 0.48 0 to 58,100 0 to 16,400 4SG −13.14 1.18 0.22 0.33 0 to 67,700 0 to 33,400 PDO 3ST −15.38 1.20 0.51 0.77 0 to 45,700 0 to 9,300 4ST −13.24 1.14 0.30 0.36 0 to 46,800 0 to 5,900 3SG −8.74 0.77 0.23 0.40 0 to 58,100 0 to 16,400 4SG −11.02 1.02 0.24 0.44 0 to 67,700 0 to 33,400 Source: AASHTO 2010. Table 53. Coefficients for multivehicle intersection SPF.

58 Application of Crash Modification Factors for Access Management The results from the equations in Figures 24 and 25 are combined to predict the average crash frequency of an individual intersection under base conditions (from the equation in Figure 23). As shown in the equation in Figure 17, this estimate of predicted intersection crashes for base conditions is adjusted using CMFs. The CMFs reflect the safety effects of geometric and traffic operational characteristics that are different from the base conditions. CMFs applicable to the Part C Predictive Method for urban and suburban arterials are presented in the next section, Select and Apply Applicable CMFs, which covers Step 10 of the Part C Predictive Method and involves selecting and applying applicable CMFs. Select and Apply Applicable CMFs Prior to selecting CMFs, it is important to understand the applicability and base conditions of CMFs available in the Part C Predictive Method. The CMFs for urban and suburban arterials in the Highway Safety Manual (1st Edition) apply to the same site types as the corresponding SPFs presented in the previous section. Further, the CMFs apply only to vehicle-vehicle crashes and not pedestrian and bicycle crashes. Table 55 summarizes the CMFs available in the Highway Safety Manual (1st Edition), indicating the base conditions and highlighting those CMFs related to access management features. In addition to the CMFs available in the Highway Safety Manual (1st Edition), Table 56 pro- vides a list of variables assessed in NCHRP Project 17-74, indicating the base condition and whether or not the analysis resulted in a recommended CMF to adjust crash predictions from the SPF for base conditions. The research did not recommend CMFs for many of these vari- ables, particularly those related to access spacing and density, which are more appropriately addressed through corridor-level models (see Chapter 5). The research also explored inter- actions among access management features. For example, the Part C Predictive Method for urban and suburban arterials accounts for the number and type of driveways in each segment; however, it does not account for potential interactions of driveways and nearby intersections. The research explored access density and spacing (i.e., the combination of unsignalized inter- sections and driveways) and segment-based corner clearance (i.e., minimum, maximum, and average corner clearance of driveways at signalized intersections). The results suggest that the Part C Predictive Method does not perform well for variations in these variables. Specifically, Table 54. Coefficients for single-vehicle intersection SPF. Crash Severity Site Type Intercept (a) AADTmajor (b) AADTminor (c) Dispersion Parameter (k) Applicable AADT (vehicles per day) AADTmajor AADTminor Total 3ST −6.81 0.16 0.51 1.14 0 to 45,700 0 to 9,300 4ST −9.02 0.42 0.40 0.36 0 to 46,800 0 to 5,900 3SG −5.33 0.33 0.12 0.65 0 to 58,100 0 to 16,400 4SG −10.21 0.68 0.27 0.36 0 to 67,700 0 to 33,400 Fatal and Injury 3ST -- -- -- -- 0 to 45,700 0 to 9,300 4ST −9.75 0.27 0.51 0.24 0 to 46,800 0 to 5,900 3SG -- -- -- -- 0 to 58,100 0 to 16,400 4SG −9.25 0.43 0.29 0.09 0 to 67,700 0 to 33,400 PDO 3ST −8.36 0.25 0.55 1.29 0 to 45,700 0 to 9,300 4ST −9.08 0.45 0.33 0.53 0 to 46,800 0 to 5,900 3SG −7.04 0.36 0.25 0.54 0 to 58,100 0 to 16,400 4SG −11.34 0.78 0.25 0.44 0 to 67,700 0 to 33,400 Note: -- indicates no SPF available; to estimate crashes by severity, multiply the predicted total crashes for the site type by the following proportions, as applicable: 0.31 for 3ST and 0.28 for 4ST. Source: AASHTO 2010.

Predictive Method for Segment- and Intersection-Level Analysis 59   the Part C Predictive Method assumes each site is independent, which does not capture these potential interactions, and the segmentation process is not conducive to quantifying the safety impacts of interactions among sites. Interactions among access management features, particu- larly those related to access spacing and density, are more appropriately addressed through the corridor-level models in Chapter 5. The following subsections describe the CMFs from Tables 55 and 56 that can be applied in the Part C Predictive Method from the Highway Safety Manual (1st Edition) to adjust crash Facility Type Features with CMFs Base Condition Resulted in Recommended CMF Segments Median Opening Density User-defined No Median Opening Spacing User-defined No Number of Median Openings by Type No median openings No Corner Clearance User-defined No Signalized Intersection Spacing User-defined No Number of Signalized Intersections and Density User-defined No Unsignalized Intersection Spacing User-defined No Number of Unsignalized Intersections and Density User-defined No Number of Unsignalized Access Points and Density User-defined No Intersections Channelized Right-Turn Lanes No channelized right-turn lanes Yes Distance to Ramp Terminal Ramp terminal > 1,500 ft from intersection Yes Note: CMFs with user-defined base conditions are based on predictive equations where the equation predicts the number of crashes for the scenario of interest. To develop a CMF for these variables, the user defines a base condition and proposed condition, uses the predictive equation to estimate the number of crashes for both scenarios, and divides the estimate for the proposed condition by the estimate for the base condition. Source: Gross et al. 2021. Table 56. CMFs for urban and suburban arterials from NCHRP Project 17-74. Facility Type Features with CMFs Base Condition Access Management- Related Segments On-Street Parking No on-street parking Yes Roadside Fixed Objects No roadside fixed objects No Median Width 15-ft median (on divided segments) See note Lighting No lighting No Automated Speed Enforcement No automated speed enforcement No Intersections Left-Turn Lanes No left-turn lanes Yes Left-Turn Signal Phasing Permissive left-turn signal phasing No Right-Turn Lanes No right-turn lanes Yes Right-Turn-on-Red No right-turn-on-red No Lighting No lighting No Red-Light Cameras No red-light cameras No Note: NCHRP Research Report 900 (Butorac et al. 2018a) discusses median width in relation to access management. Specifically, Chapter 5: Unsignalized Median Openings indicates that the wider the median the greater the storage for the left-turn egress movement and potential for a two-stage crossing. While median width may be considered related to access management, the Highway Safety Manual (1st Edition) only considers median width for segment-related crashes and not for access-related crashes. Source: AASHTO 2010. Table 55. CMFs for urban and suburban arterials from the Highway Safety Manual (1st Edition).

60 Application of Crash Modification Factors for Access Management predictions for features on urban and suburban arterials related to access management. Refer to Chapter 3 of this guide for additional CMFs that could potentially be applied in the Part C Predictive Method. Again, prior to applying any CMFs that were not developed specifically for use with the Part C Predictive Method, there is a need to consider the applicability of the CMFs with respect to crash type, crash severity, and base condition as well as the potential for double- counting crash reductions with each additional CMF. CMFs for On-Street Parking Figure 26 presents the CMF for on-street parking along urban and suburban arterials where the base condition is no on-street parking along the segment. Table 57 presents the values for fpk for use with the CMF. The applicable CMF is applied to the total predicted crashes from the base SPF using the general equation from Figure 17, excluding vehicle-pedestrian and vehicle- bicycle crashes. Figure 26. CMF for on-street parking. Variables for the equation shown in Figure 26 are defined as follows: • CMFparking = CMF for on-street parking. • ppk = proportion of curb length with on-street parking (= 0.5*Lpk/L). • Lpk = sum of curb length with on-street parking for both sides of the road combined (miles). • L = length of roadway segment (miles). • fpk = factor from Table 57. CMFs for Left-Turn Lanes Table 58 presents the CMFs for left-turn lanes at urban and suburban arterial intersections. The applicable CMF is applied to the total predicted crashes from the base SPF using the general equation from Figure 17, excluding vehicle-pedestrian and vehicle-bicycle crashes. According to the Highway Safety Manual (1st Edition), the supporting research did not indicate a change in safety performance for the presence of a left-turn lane on the stop-controlled approaches (AASHTO 2010). Site Type Type of Parking and Land Use Parallel Parking Residential/ Other Land Use Parallel Parking Industrial/Institutional Land Use Angle Parking Residential/Other Land Use Angle Parking Industrial/Institutional Land Use 2U 1.465 2.074 3.428 4.853 3T 1.465 2.074 3.428 4.853 4U 1.100 1.709 2.574 3.999 4D 1.100 1.709 2.574 3.999 5T 1.100 1.709 2.574 3.999 Note: The base condition is no on-street parking along the segment. Source: AASHTO 2010. Table 57. Values of fpk for on-street parking along urban and suburban arterials.

Predictive Method for Segment- and Intersection-Level Analysis 61   CMFs for Right-Turn Lanes Table 59 presents the CMFs for right-turn lanes at urban and suburban arterial inter sections. The applicable CMF is applied to the total predicted crashes from the base SPF using the general equation from Figure 17, excluding vehicle-pedestrian and vehicle-bicycle crashes. According to the Highway Safety Manual (1st Edition), the supporting research did not indicate a change in safety performance for the presence of a right-turn lane on the stop-controlled approaches. These CMFs are only applicable to right-turn lanes identified by pavement markings or signing. The CMFs are not applicable to locations used as informal right-turn lanes (e.g., tapers, flares, or paved shoulders). Channelized Right-Turn Lanes Table 60 presents the CMFs for channelizing right-turn lanes at urban and suburban arterial intersections. The applicable CMF is applied to the total predicted crashes for locations with a right-turn lane present. This would use the base SPF prediction from the general equation in Figure 17, excluding vehicle-pedestrian and vehicle-bicycle crashes, adjusted for the presence of right-turn lanes using an appropriate CMF from Table 59. The research did not indicate a change in safety performance for the channelization of a right-turn lane on four-legged stop- controlled approaches or three-legged and four-legged signalized approaches. Distance to a Ramp Terminal Table 61 presents the CMFs to adjust for the distance to a nearby ramp terminal at urban and suburban arterial intersections. The applicable CMF is applied to the total predicted crashes from the base SPF using the general equation from Figure 17, excluding vehicle-pedestrian and Site Type Number of Approaches with Left-Turn Lanes 1 2 3 4 3ST 0.67 -- -- -- 3SG 0.93 0.86 -- -- 4ST 0.73 0.53 -- -- 4SG 0.90 0.81 0.73 0.66 Note: The base condition is no left-turn lanes present. Minor road stop-controlled approaches are not considered in counting the number of approaches with left-turn lanes. Source: AASHTO 2010. Table 58. CMFs for left-turn lanes at urban and suburban arterial intersections. Site Type Number of Approaches with Right-Turn Lanes 1 2 3 4 3ST 0.86 -- -- -- 3SG 0.96 -- -- -- 4ST 0.86 0.74 -- -- 4SG 0.96 0.92 0.88 0.85 Note: The base condition is no right-turn lanes present. Minor road stop-controlled approaches are not considered in counting the number of approaches with right-turn lanes. Source: AASHTO 2010. Table 59. CMFs for right-turn lanes at urban and suburban arterial intersections.

62 Application of Crash Modification Factors for Access Management vehicle-bicycle crashes. The research indicates that predicted crashes increase at stop-controlled intersections where a ramp terminal is present within 1,500 feet; however, the results are not statistically significant, even at the 90-percent confidence level (i.e., CMF = 2.12 with standard error = 0.91). As such, the observed variability suggests that this strategy could result in an increase, decrease, or no change in crashes. For signalized intersections, there is no CMF required to adjust the base prediction for the distance to a ramp terminal because the research indicated that the base SPF prediction for urban and suburban signalized intersections is predicting well across the range of predicted values and distances to ramp terminals. Sample Problem 1 Estimate the predicted average crash frequency along a segment of an urban two-lane arte- rial with a TWLTL and then estimate the change in predicted average crash frequency if minor commercial driveways are converted to major commercial driveways. Problem Definition This scenario is based on Sample Problem 1 from the Highway Safety Manual (1st Edition). The first part of the problem calculates predicted average crash frequency for a segment of an urban arterial with a TWLTL. The second part of the problem shows how to estimate the change in predicted average crash frequency if the minor commercial driveways are converted to major commercial driveways. Note the calculations are based on the use of applicable equa- tions without rounding. While the values from interim calculations are shown as rounded to the nearest second or third decimal, these values should not be rounded during the calculation process until the final step. Table 62 shows the conditions and assumptions for existing and proposed conditions. Site Type Number of Approaches with Right-Turn Lanes 1 2 3 4 3ST 0.72 -- -- -- 3SG -- -- -- -- 4ST -- -- -- -- 4SG -- -- -- -- Note: The base condition is the presence of a right-turn lane with no channelization. Minor road stop-controlled approaches are not considered in counting the number of approaches with right-turn lanes or channelization. Source: NCHRP Project 17-74. Table 60. CMFs for channelizing right-turn lanes at urban and suburban arterial intersections. Table 61. CMFs for distance to a ramp terminal at urban and suburban intersections. Site Type Distance to Ramp Terminal ≤ 1,500 feet Distance to Ramp Terminal > 1,500 feet 3ST 2.12* 1.00 3SG 1.00 1.00 4ST 2.12* 1.00 4SG 1.00 1.00 Note: The base condition is no ramp terminal within 1,500 feet of the intersection. *Observed variability suggests this strategy could result in an increase, decrease, or no change in crashes. Source: NCHRP Project 17-74.

Predictive Method for Segment- and Intersection-Level Analysis 63   Step 1: Predict Multivehicle Non-Driveway Collisions Figures 27 through 29 show the calculations to predict multivehicle non-driveway collisions for total, fatal and injury, and PDO crashes, respectively, based on the SPF from Figure 20 and coefficients from Table 50. Variable Existing Condition Proposed Condition Road type 3T 3T Length of segment 1.5 miles 1.5 miles AADT 11,000 vehicles/day 11,000 vehicles/day Type of on-street parking Parallel-commercial Parallel-commercial Median width Not present Not present Lighting Present Present Auto speed enforcement Not present Not present Major commercial driveways 0 10 Minor commercial driveways 10 0 Major industrial/institutional driveways 0 0 Minor industrial/institutional driveways 3 3 Major residential driveways 2 2 Minor residential driveways 15 15 Other driveways 0 0 Posted speed limit 35 mph 35 mph Roadside fixed object density 10 fixed objects/mile 10 fixed objects/mile Offset to roadside fixed objects 6 feet 6 feet Calibration factor 1 1 Source: AASHTO 2010. Table 62. Sample Problem 1 conditions and assumptions. Figure 27. Calculation of predicted total multivehicle non-driveway crashes/year for existing conditions. Figure 28. Calculation of predicted fatal and injury multivehicle non-driveway crashes/year for existing conditions. Figure 29. Calculation of predicted PDO multivehicle non-driveway crashes/year for existing conditions. Figures 30 and 31 show the calculations to adjust the predicted fatal and injury multivehicle non-driveway crashes and PDO multivehicle non-driveway crashes to sum to the total multi- vehicle non-driveway crashes. Figure 30. Adjustment of predicted fatal and injury multivehicle non-driveway crashes/year for existing conditions. Figure 31. Adjustment of predicted PDO multivehicle non-driveway crashes/year for existing conditions.

64 Application of Crash Modification Factors for Access Management Step 2: Predict Single-Vehicle Crashes Figures 32 through 34 show the calculations to predict single-vehicle collisions for total, fatal and injury, and PDO crashes, respectively, based on the SPF from Figure 21 and coefficients from Table 51. Figure 34. Calculation of predicted PDO single-vehicle crashes/year for existing conditions. Figure 32. Calculation of predicted total single-vehicle crashes/year for existing conditions. Figure 33. Calculation of predicted fatal and injury single- vehicle crashes/year for existing conditions. Figures 35 and 36 show the calculations to adjust the predicted fatal and injury single-vehicle crashes and PDO single-vehicle crashes to sum to the total single-vehicle crashes. Figure 35. Adjustment of predicted fatal and injury single-vehicle crashes/year for existing conditions. Figure 36. Adjustment of predicted PDO single-vehicle crashes/year for existing conditions. Step 3: Predict Multivehicle Driveway-Related Collisions Figure 37 shows the calculations to predict total multivehicle driveway-related crashes (Nbrdwy (total)) based on the SPF from Figure 22 and coefficients from Table 52. Figure 37. Calculation of predicted total multivehicle driveway-related crashes/year for existing conditions.

Predictive Method for Segment- and Intersection-Level Analysis 65   Figures 38 and 39 show the calculations to separate the predicted total multivehicle driveway- related crashes by severity to obtain fatal and injury multivehicle driveway-related crashes (Nbrdwy(FI)) and PDO multivehicle driveway-related crashes (Nbrdwy(PDO)). (f indicates the propor- tion of fatal and injury crashes.) Figure 39. Adjustment of predicted PDO multivehicle driveway- related crashes/year for existing conditions. Figure 38. Adjustment of predicted fatal and injury multivehicle driveway-related crashes/year for existing conditions. Step 4: Calculate Adjustment Factors The next step is to calculate appropriate CMFs to adjust the base predicted crash frequency. The following sections show the calculations of appropriate CMFs. On-Street Parking CMF Figure 40 shows the calculation of the proportion of curb length that has on-street parking. Figure 40. Calculation of proportion of curb length with on-street parking. Figure 41 shows the calculation of the on-street parking CMF (CMF1r). Figure 41. Calculation of on-street parking CMF. Roadside Fixed Objects CMF Figure 42 shows the calculation of the CMF for roadside fixed objects (CMF2r). Figure 42. Calculation of roadside fixed objects CMF. Variables for the equation shown in Figure 42 are defined as follows: foffset = fixed object offset factor, Dfo = fixed object density (fixed objects per mile) for both sides of the road combined, and Pfp = fixed object collisions as a proportion of total crashes. Median Width CMF The roadway has a TWLTL, so there is no median, and the CMF = 1.0.

66 Application of Crash Modification Factors for Access Management Lighting CMF Figure 43 shows the calculation of the CMF for lighting (CMF4r). Figure 43. Calculation of lighting CMF. Variables for the equation shown in Figure 43 are defined as follows: Pnr = proportion of total crashes for unlighted roadway segments that occur at night, and Pinr = proportion of total nighttime crashes for unlighted roadway segments that involve a fatality or injury. Automated Speed Enforcement CMF There is no automated speed enforcement, so the CMF = 1.0. Combined CMF Once all the CMFs are calculated, Figure 44 shows the calculation of the combined CMF value (CMFcombined). Figure 44. Calculation of combined CMF. Step 5: Combine and Adjust Predicted Single-Vehicle and Multivehicle Crashes The previous steps calculated single-vehicle and multivehicle crashes. Figure 45 shows the calculation of the predicted total crash frequency for the road segment using those values. Figure 45. Calculation of predicted total crashes for existing base conditions. Figure 46 shows the calculation of the adjusted predicted total crash frequency for the existing road conditions (Nbr) by multiplying the combined CMF and the predicted total crash frequency for base conditions. Figure 46. Calculation of adjusted predicted total crashes for existing conditions. Step 6: Predict Vehicle-Pedestrian and Vehicle-Bicycle Collisions Figures 47 and 48 show the calculations of vehicle-pedestrian (Npedr) and vehicle-bicycle (Nbiker) crashes, respectively.

Predictive Method for Segment- and Intersection-Level Analysis 67   Step 7: Calculate Predicted Average Crash Frequency Figure 49 shows how to combine the previous calculations to estimate the predicted average crash frequency for existing conditions (Npredicted rs). The method predicts approximately seven crashes per year for existing conditions. Note that the predicted average crash frequency in this example differs slightly from the Highway Safety Manual (1st Edition) due to a rounding dis- crepancy in the proportion of curb length with on-street parking. Cr in the equation shown in Figure 49 indicates the calibration factor for roadway segments. Figure 48. Calculation of vehicle-bicycle crashes for existing conditions. Figure 47. Calculation of vehicle-pedestrian crashes for existing conditions. Step 8: Estimate the Predicted Average Crash Frequency for Proposed Conditions The next series of calculations is related to the proposed conversion of minor commercial drive- ways to major commercial driveways. Specifically, the existing conditions include zero major commercial driveways and 10 minor commercial driveways. The proposed conditions include 10 major commercial driveways and zero minor commercial driveways. The change in major and minor commercial driveways affects the previous multivehicle driveway-related collisions calculations, which changes the subsequent related numbers. Figure 50 shows the calculations to predict total multivehicle driveway-related crashes based on the proposed conditions. Figure 49. Calculation of predicted average crash frequency for existing conditions. Figure 50. Calculation of predicted total multivehicle driveway-related crashes/year for proposed conditions. Figures 51 and 52 show the calculations to separate the predicted total multivehicle driveway- related crashes by severity to obtain fatal and injury multivehicle driveway-related crashes and PDO multivehicle driveway-related crashes.

68 Application of Crash Modification Factors for Access Management Figure 53 shows the calculation of the predicted total crash frequency for the proposed base conditions using the multivehicle driveway-related crashes from Figure 50 and the multivehicle non-driveway crashes and single-vehicle crashes from Figure 27 and Figure 32, respectively. Figure 51. Adjustment of predicted fatal and injury multivehicle driveway-related crashes/year for proposed conditions. Figure 52. Adjustment of predicted PDO multivehicle driveway- related crashes/year for proposed conditions. Figure 53. Calculation of predicted total crashes for proposed base conditions. Figure 54 shows the calculation of the adjusted predicted total crash frequency for the pro- posed road conditions by multiplying the combined CMF and the predicted total crash frequency for proposed base conditions. Figure 54. Calculation of adjusted predicted total crashes for proposed conditions. Figures 55 and 56 show the calculations of vehicle-pedestrian and vehicle-bicycle crashes, respectively, for the proposed conditions. Figure 56. Calculation of vehicle-bicycle crashes for proposed conditions. Figure 55. Calculation of vehicle-pedestrian crashes for proposed conditions. Figure 57 shows how to combine the previous calculations to estimate the predicted average crash frequency for proposed conditions. The method predicts 7.89 crashes per year for pro- posed conditions. The method predicts approximately seven crashes per year for existing condi- tions. As such, the proposed conversion of minor driveways to major driveways would result in an expected increase in crashes. Table 63 presents a summary and comparison of the results for existing and proposed conditions.

Predictive Method for Segment- and Intersection-Level Analysis 69   Sample Problem 2 Estimate the expected average crash frequency per year for a three-legged stop-controlled intersection located on an urban arterial, employing the EB method, and then estimate the change in expected average crash frequency per year for two proposed conditions: (1) a pro- posed right-turn lane on the major road and (2) a proposed channelization of a right-turn lane on the major road. Problem Definition This scenario is based on Sample Problem 3 from the Highway Safety Manual (1st Edition) but uses a major road traffic volume of 11,000 vehicles per day instead of 14,000 vehicles per day. It is assumed that the site conditions and traffic volumes remain constant over the 3-year period. This problem is broken into three parts. The first part of the problem calculates the expected average crash frequency for a three-legged stop-controlled intersection located on an urban arterial, employing the EB method to combine the predicted crashes and observed crashes over a 3-year study period. The second part of the problem shows how to estimate the change in expected average crash frequency if a right-turn lane is proposed on the major road. The third part of the problem shows how to estimate the change in expected average crash frequency if the right-turn lane is channelized on the major road. Table 64 shows the conditions and assump- tions for existing and proposed conditions. Step 1: Predict Multivehicle Collisions for Existing Conditions Figures 58 through 60 show the calculations to predict multivehicle collisions for total, fatal and injury, and PDO crashes, respectively, based on the SPF from Figure 24 and coefficients from Table 53. Figure 57. Calculation of predicted average crash frequency for proposed conditions. Calculation Existing Condition Proposed Condition Multivehicle non-driveway crashes Total 3.085 3.085 Fatal and injury 0.742 0.742 PDO 2.343 2.343 Single-vehicle crashes Total 0.734 0.734 Fatal and injury 0.209 0.209 PDO 0.524 0.524 Multivehicle driveway-related crashes Total 0.455 0.969 Fatal and injury 0.111 0.235 PDO 0.345 0.733 Total single and multivehicle crashes 6.906 7.735 Total vehicle-pedestrian crashes 0.09 0.101 Total vehicle-bicycle crashes 0.048 0.054 Total predicted average crash frequency 7.044 7.89 Table 63. Summary of Sample Problem 1 calculations and results.

70 Application of Crash Modification Factors for Access Management Figures 61 and 62 show the calculations to adjust the predicted fatal and injury multivehicle crashes and PDO multivehicle crashes to sum to the total multivehicle crashes. Variable Existing Condition Proposed Condition A (right-turn lane on major road) Proposed Condition B (channelized right-turn lane on major road) Intersection type 3ST 3ST 3ST AADT major 11,000 veh/day 11,000 veh/day 11,000 veh/day AADT minor 4,000 veh/day 4,000 veh/day 4,000 veh/day Intersection lighting Not present Not present Not present Calibration factor 1 1 1 Data for unsignalized intersections only: 1 1 1 0 1 1 No No Yes Observed multivehicle crashes in last 3 years 12 -- -- Observed single-vehicle crashes in last 3 years 3 -- -- Source: AASHTO 2010. Number of major road approaches with left-turn lanes Number of major road approaches with right-turn lanes Presence of channelized right- turn lanes • • • Table 64. Sample Problem 2 conditions and assumptions. Figure 58. Calculation of predicted total multivehicle crashes/year for existing conditions. Figure 59. Calculation of predicted fatal and injury multivehicle crashes/year for existing conditions. Figure 60. Calculation of predicted PDO multivehicle crashes/year for existing conditions. Figure 61. Adjustment of predicted fatal and injury multivehicle crashes/year for existing conditions. Figure 62. Adjustment of predicted PDO multivehicle crashes/year for existing conditions. Step 2: Predict Single-Vehicle Crashes for Existing Conditions Figures 63 and 64 show the calculations to predict single-vehicle collisions for total and PDO crashes, respectively, based on the SPF from Figure 25 and coefficients from Table 54.

Predictive Method for Segment- and Intersection-Level Analysis 71   There are no models for fatal and injury crashes at three-legged stop-controlled intersections. Figure 65 shows the calculation of fatal and injury crashes, which is based on the proportion of total crashes using a default proportion of fatal and injury crashes. Figure 64. Calculation of predicted PDO single-vehicle crashes/year for existing conditions. Figure 63. Calculation of predicted total single-vehicle crashes/year for existing conditions. Figures 66 and 67 show the calculations to adjust the predicted fatal and injury single-vehicle crashes and PDO single-vehicle crashes to sum to the total single-vehicle crashes. Step 3: Calculate Adjustment Factors The next step is to calculate appropriate CMFs to adjust the base predicted crash frequency. The following sections show the calculations of appropriate CMFs. Intersection Left-Turn Lane CMF From Table 58, the CMF for a left-turn lane on the major road of a three-legged, minor road stop-controlled intersection is 0.67. Intersection Right-Turn Lane CMF From Table 59, the CMF for no right-turn lanes on the major road of a three-legged, minor road stop-controlled intersection is 1.0. Lighting CMF From Chapter 12 of the Highway Safety Manual (1st Edition), the CMF for no lighting is 1.0. Figure 65. Calculation of predicted fatal and injury single-vehicle crashes/year for existing conditions. Figure 66. Adjustment of predicted fatal and injury single-vehicle crashes/year for existing conditions. Figure 67. Adjustment of predicted PDO single- vehicle crashes/year for existing conditions.

72 Application of Crash Modification Factors for Access Management Combined CMF Figure 68 shows the calculation of the combined CMF value once all of the CMFs are calculated. Figure 68. Calculation of combined CMF. Step 4: Combine and Adjust Predicted Single-Vehicle and Multivehicle Crashes for Existing Conditions The previous steps calculated single-vehicle and multivehicle crashes. Figure 69 shows the calculation of the predicted total crash frequency for the intersection using those values. Figure 69. Calculation of predicted total crashes for existing base conditions. Figure 70 shows the calculation of the adjusted predicted total crash frequency (Nbi) for the existing conditions by multiplying the combined CMF and the predicted total crash frequency for base conditions. Figure 70. Calculation of adjusted predicted total crashes for existing conditions. Step 5: Predict Vehicle-Pedestrian and Vehicle-Bicycle Collisions for Existing Conditions Figures 71 and 72 show the calculations of vehicle-pedestrian and vehicle-bicycle crashes, respectively. fpedi indicates pedestrian crash adjustment factor, and fbiker indicates bicycle crash adjustment factor. Figure 71. Calculation of predicted vehicle- pedestrian crashes for existing conditions. Figure 72. Calculation of predicted vehicle- bicycle crashes for existing conditions. Step 6: Calculate Predicted Average Crash Frequency for Existing Conditions Figure 73 shows how to combine the previous calculations to estimate the predicted average crash frequency for existing conditions (Npredicted int). The method predicts approximately 1.2 crashes per year for existing conditions or 3.7 crashes over the 3-year study period. Note that the predicted average crash frequency in this example differs slightly from the Highway Safety

Predictive Method for Segment- and Intersection-Level Analysis 73   Manual (1st Edition) due to the difference in major road traffic volume. Also, recall that the site conditions and traffic volume remained constant over the 3-year period. If the traffic volumes change over time, then the previous calculations would be performed separately for each year and summed to estimate the total predicted crashes over the 3-year period. Ci in the equation shown in Figure 73 indicates calibration factor for intersections. Figure 73. Calculation of predicted average crash frequency for existing conditions. Step 7: Apply EB Method to Calculate Expected Average Crash Frequency for Existing Conditions The next series of calculations is related to the application of the EB method to calculate the expected average crash frequency based on a combination of observed and predicted crashes. The intersection has an observed crash history of 12 multivehicle crashes and three single-vehicle crashes in the most recent 3-year period (n = 3) for this sample problem. Note the EB method is applied separately for multivehicle and single-vehicle crashes and the results are combined to estimate the total expected average crash frequency. Also, it is important to note that the observed, predicted, and expected crashes should be in terms of the same time period, which in this case is 3 years. Multivehicle Crashes Figure 74 shows the calculation of the weight factor for multivehicle crashes (wmv), which is used in the EB method. Note the dispersion parameter (k) for multivehicle crashes at three- legged stop-controlled intersections is 0.80 from Table 53. While the dispersion param- eter applies to the SPF for the base condition, it approximates the dispersion parameter for the adjusted predicted crashes. The weight is relatively insensitive to the value of k, so this approximation is not a concern. Also, note that the number of adjusted predicted multivehicle crashes per year (Npredicted mv) is based on the product of the combined CMF (0.67) and the pre- dicted multivehicle collisions per year for existing conditions (1.447), which is 0.970 crashes per year. Figure 74. Calculation of weight (w) for multivehicle crashes for existing conditions. Figure 75 shows the application of the EB method to calculate the expected average crash frequency for multivehicle crashes (Nexpected mv) for existing conditions over the 3-year study period. Figure 75. Calculation of expected average crash frequency for multivehicle crashes for existing conditions.

74 Application of Crash Modification Factors for Access Management Single-Vehicle Crashes Figure 76 shows the calculation of the weight factor for single-vehicle crashes (wsv), which is used in the EB method. Note the dispersion parameter (k) for single-vehicle crashes at three- legged stop-controlled intersections is 1.14 from Table 54. Again, the dispersion parameter for the base condition approximates the dispersion parameter for the adjusted predicted crashes. Also, note the number of adjusted predicted single-vehicle crashes is based on the product of the combined CMF (0.67) and the predicted single-vehicle collisions per year for existing conditions (0.336), which is 0.225 crashes per year. Npredicted sv in the equation in Figure 76 indicates adjusted predicted average crash frequency for single-vehicle crashes. Figure 76. Calculation of weight (w) for single- vehicle crashes for existing conditions. Figure 77 shows the application of the EB method to calculate the expected average crash frequency for single-vehicle crashes (Nexpected sv) for existing conditions over the 3-year study period. Figure 77. Calculation of expected average crash frequency for single-vehicle crashes for existing conditions. Pedestrian and Bicycle Crashes Figures 78 and 79 show the calculations for expected vehicle-pedestrian (Nexpected ped) and vehicle- bicycle crashes (Nexpected bike), respectively, over the 3-year study period. Figure 78. Calculation of expected vehicle-pedestrian crashes for existing conditions. Figure 79. Calculation of expected vehicle-bicycle crashes for existing conditions. Total Expected Crashes Figure 80 shows how to combine the previous calculations to estimate the total expected average crash frequency for existing conditions over the 3-year study period. The result is 11.359 expected crashes for existing conditions over the 3-year study period, or 3.786 expected crashes per year for existing conditions. Figure 80. Calculation of expected average crash frequency for existing conditions.

Predictive Method for Segment- and Intersection-Level Analysis 75   Step 8: Estimate the Expected Average Crash Frequency for Proposed Condition A This step is related to Proposed Condition A (i.e., installation of a right-turn lane on a major road of a three-legged stop-controlled intersection). To estimate the expected average crash frequency for Proposed Condition A (Nexpected condition A), there is a need to apply the CMF for right- turn lanes to the expected average crash frequency for existing conditions (Nexpected int). From Table 59, the CMF for the presence of an intersection right-turn lane is 0.86. Figure 81 shows the calculations to estimate the expected average crash frequency for Proposed Condition A. Note that if the existing condition included a right-turn lane, then the CMF of 0.86 would be applied as an adjustment factor in the calculation of the predicted crashes (see Figure 70) before computing the expected crashes. Figure 81. Calculation of expected average crash frequency for Proposed Condition A. Step 9: Estimate the Expected Average Crash Frequency for Proposed Condition B This step is related to Proposed Condition B (i.e., channelize right-turn lane on major road of three-legged stop-controlled intersection). To estimate the expected average crash frequency for Proposed Condition B, there is a need to apply the CMF for channelizing right-turn lanes to the expected average crash frequency for Proposed Condition A. From Table 60, the CMF for channelizing a right-turn lane is 0.72. Figure 82 shows the calculations to estimate the expected average crash frequency for Proposed Condition B (Nexpected condition B). Again, if the existing con- dition included a channelized right-turn lane, then the CMF of 0.72 would be applied as an adjustment factor in the calculation of the predicted crashes (see Figure 70) before computing the expected crashes. Figure 82. Calculation of expected average crash frequency for Proposed Condition B. Table 65 presents a summary and comparison of the predicted and expected crashes for existing and proposed conditions. The predicted crash frequency is based on the SPF and related adjustment factors while the expected crash frequency incorporates the observed crash history. Whether focused on predicted or expected crashes, Proposed Condition A represents a 14-percent reduction in crashes compared to the existing condition, which is based on the respective CMF of 0.86. Similarly, Proposed Condition A represents a 28-percent reduction in crashes compared to Proposed Condition B, which is based on the respective CMF of 0.72. Performance Measure Existing Condition (no right-turn lane) Proposed Condition A (right-turn lane on major approach) Proposed Condition B (channelized right-turn lane on major approach) Predicted average crash frequency (per year) 1.239 1.066 0.767 Expected average crash frequency (per year) 3.786 3.256 2.344 Table 65. Summary of Sample Problem 2 calculations and results.

76 Application of Crash Modification Factors for Access Management Sample Problem 3 Estimate the predicted average crash frequency along a 1.5-mile urban two-lane arterial corridor with a TWLTL and 10 three-legged stop-controlled intersections. Problem Definition This scenario is based on the existing conditions in Sample Problem 1 and Sample Problem 2. The first part of the problem calculates the predicted average crash frequency for the urban arterial segment with a TWLTL (Sample Problem 1). The second part of the problem calcu- lates the predicted average crash frequency for the three-legged stop-controlled intersections along the urban arterial corridor (Sample Problem 2). Table 62 and Table 64 show the existing conditions and assumptions for the segment and intersections. It is assumed for this example that all 10 intersections are identical. Step 1: Predict Crash Frequency for Existing Segment From Sample Problem 1, the predicted average crash frequency for the segment is 7.044 crashes per year. Step 2: Predict Crash Frequency for Existing Intersections From Sample Problem 2, the predicted average crash frequency for one intersection is 1.239 crashes per year. For 10 identical three-legged minor road stop-controlled intersections along the corridor, the total is 12.39 predicted crashes per year. Step 3: Calculate Predicted Average Crash Frequency for the Corridor Figure 83 shows how to combine the previous calculations to estimate the predicted average crash frequency for existing conditions along the corridor (Npredicted corridor). The method predicts approximately 19 crashes per year for existing conditions along the corridor. For illustrative purposes, the results are also compared to the results from Sample Problem 4 in Chapter 5. Applying the corridor-level prediction models from Chapter 5, the predicted crash frequency for the same corridor is approximately 28 crashes per year. Note the two methods do not pro- duce the same results. The predictive method in Chapter 4 (this chapter) does not consider the potential interactions among adjacent or nearby sites (e.g., access spacing and density). As such, the corridor-level predictive method in Chapter 5 is more appropriate for considering inter actions among access management features and estimating the safety effect of variables related to access spacing and density. The segment-intersection predictive method in Chapter 4 could provide a reasonable estimate of safety performance when the analyst expects minimal interactions among adjacent or nearby driveways and intersections. The Chapter 4 method is also appropriate for considering or comparing the safety impacts of detailed segment and inter- section characteristics (e.g., presence of left- or right-turn lanes and corner clearance). Figure 83. Calculation of predicted average crash frequency for existing corridor conditions.