The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 77... ...
61 5 MODELS FOR URBAN AND SUBURBAN ARTERIALS 5.1 ROADWAY SEGMENTS Estimation Data The process of developing models for urban and suburban arterial road segments involved developing an initial set of models and then validating them with a second dataset obtained later. Following the successful validation, we combined the two datasets into a larger dataset to re‐estimate the models. We attempted three sets of models: 1. Base Condition SPFs -- using only those sites meeting the HSM base conditions 2.
|
From page 78... ...
62 Single‐vehicle run‐off‐road (SV ROR) Single‐vehicle fixed object (SV FIXEDOBJ)
|
From page 79... ...
63 Table 5‐1: Ohio Segment Length and Crash Type Totals for Five‐Year Period for Base Condition Sites (Urban/Suburban Arterial Segments) Site Type Length (mi.)
|
From page 80... ...
64 Table 5‐2: Ohio Segment Crash Type Statistics for Five‐Year Period for Base Condition Sites (Urban/Suburban Arterial Segments) Site Type Stat. PED BIKE MVD RE HO ANG SSD SOD MVN OTHER SV SV ROR ANIM AL FO MO SV OTHER Night 2U N 760 760 760 760 760 760 760 760 760 760 760 760 760 760 760 760 2U MIN 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2U MAX 2 1 13 66 2 7 7 8 8 109 36 81 32 3 8 78 2U MEAN 0.03 0.01 0.49 1.65 0.06 0.10 0.18 0.32 0.41 4.71 1.79 2.78 1.70 0.05 0.19 3.31 2U STD 0.19 0.11 1.30 4.21 0.25 0.47 0.62 0.88 1.01 8.75 3.84 5.79 3.64 0.25 0.60 6.28 3T N 182 182 182 182 182 182 182 182 182 182 182 182 182 182 182 182 3T MIN 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3T MAX 1 1 10 34 1 6 2 4 7 56 13 42 12 1 2 36 3T MEAN 0.02 0.01 0.59 1.42 0.02 0.20 0.12 0.13 0.24 2 0.52 1.39 0.52 0.03 0.06 1.64 3T STD 0.13 0.10 1.54 3.70 0.15 0.63 0.37 0.52 0.70 5.23 1.42 3.92 1.41 0.16 0.26 3.73 4D N 358 358 358 358 358 358 358 358 358 358 358 358 358 358 358 358 4D MIN 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4D MAX 2 2 19 172 1 7 28 4 19 98 37 61 36 4 8 85 4D MEAN 0.05 0.01 0.37 3.22 0.01 0.16 0.89 0.11 0.49 4.50 1.36 2.92 1.24 0.12 0.22 3.82 4D STD 0.24 0.14 1.64 13.98 0.11 0.67 2.64 0.44 1.62 10.76 3.51 7.32 3.19 0.45 0.86 9.13 4U N 348 348 348 348 348 348 348 348 348 348 348 348 348 348 348 348 4U MIN 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4U MAX 2 3 34 78 2 21 32 7 13 35 15 25 15 1 2 28 4U MEAN 0.05 0.02 0.67 1.89 0.05 0.31 0.71 0.18 0.32 1.33 0.45 0.80 0.46 0.03 0.04 1.52 4U STD 0.26 0.20 2.55 5.91 0.23 1.34 2.38 0.68 1.14 3.26 1.31 2.25 1.32 0.18 0.22 3.61 5T N 180 180 180 180 180 180 180 180 180 180 180 180 180 180 180 180 5T MIN 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5T MAX 2 2 29 115 2 11 24 4 18 46 16 29 16 2 4 43 5T MEAN 0.07 0.04 2.20 6.69 0.07 0.81 1.79 0.33 0.92 3.74 1.13 2.33 1.14 0.10 0.17 4.41 5T STD 0.27 0.23 4.37 15.99 0.27 1.75 3.48 0.76 2.16 6.75 2.34 4.66 2.39 0.35 0.50 7.54
|
From page 81... ...
65 Table 5‐3: OH Segment Continuous Variable Statistics for Base Condition Sites (Urban/Suburban Arterial Segments) Site Type Stat. Length AADT Med Width Parking Prop FODensity Offset FO Maj Comm Min Comm Maj Ind Min Ind Maj Res Min Res Other Dwy 2U N 760 760 760 760 760 760 760 760 760 760 760 760 760 2U MIN 0.01 100 0 0 25 2 0 0 0 0 0 0 0 2U MAX 6.29 23,028 0 0 75 20 10 41 6 28 4 193 5 2U MEAN 0.59 6975 0 0 37.76 8.95 0.19 2.34 0.08 0.97 0.03 10.96 0.03 2U STD 0.72 3978 0 0 12.95 3.80 0.77 4.70 0.48 2.63 0.26 20.03 0.24 3T N 182 182 182 182 182 182 182 182 182 182 182 182 182 3T MIN 0.02 1356 0 0 25 2 0 0 0 0 0 0 0 3T MAX 3.29 23780 0 0 75 20 11 49 12 10 2 65 1 3T MEAN 0.34 1022 0 0 41.87 8.11 0.98 5.24 0.29 0.41 0.05 4.82 0.02 3T STD 0.44 4034 0 0 13.75 4.09 1.89 9.12 1.17 1.27 0.27 8.99 0.15 4D N 358 358 358 358 358 358 358 358 358 358 358 358 358 4D MIN 0.01 256 10 0 25 10 0 0 0 0 0 0 0 4D MAX 4.81 45,874 100 0 75 30 33 47 8 5 2 64 2 4D MEAN 0.45 14,384 33.27 0 34.32 21.63 0.42 1.11 0.14 0.13 0.01 0.87 0.02 4D STD 0.67 8758 29.14 0 11.67 4.11 2.11 4.60 0.69 0.61 0.14 4.24 0.18 4U N 348 348 348 348 348 348 348 348 348 348 348 348 348 4U MIN 0.01 1150 0 0 25 2 0 0 0 0 0 0 0 4U MAX 5.96 41,418 0 0 75 25 11 57 11 16 3 78 4 4U MEAN 0.28 14,281 0 0 43.09 7.84 0.48 3.56 0.27 0.52 0.04 3.45 0.05 4U STD 0.47 7350 0 0 13.84 4.54 1.32 7.05 0.93 1.93 0.25 9.06 0.34 5T N 180 180 180 180 180 180 180 180 180 180 180 180 180 5T MIN 0.01 5356 0 0 25 2 0 0 0 0 0 0 0 5T MAX 2.91 50,553 0 0 75 20 23 75 9 16 2 46 2 5T MEAN 0.42 19,422 0 0 38.97 8.47 2.23 8.09 0.42 0.52 0.07 3.05 0.08 5T STD 0.51 83456 0 0 10.74 4.37 4.06 12.22 1.20 1.92 0.30 7.02 0.31
|
From page 82... ...
66 Table 5‐4: OH Segment Categorical Variable Total Mileage (mi.) for Base Condition Sites (Urban/Suburban Arterial Segments)
|
From page 83... ...
67 Estimated Models Documented in this section are the base condition models intended for use in the HSM predictive chapter for urban and suburban arterials. Models developed for all sites representing the average site conditions are found in Appendix A. These could be applied for network screening or other safety management tasks where models for average site conditions are desired. The model development process involved using the Ohio data to estimate a set of initial models, which we subsequently validated where possible using a dataset that later became available from Minnesota. Following the validation, we combined the two datasets to re‐estimate the final models. We developed the initial models using the same base conditions as those in the current HSM chapter for urban and suburban arterials: No on‐street parking No roadside fixed objects A 15‐foot median width for divided roads No lighting No automated speed enforcement The model predictions do not include intersection‐related or animal crashes. The initial base condition models were estimated for the following crash types: Total (TOT) Multiple‐vehicle driveway related (MVD)
|
From page 84... ...
68 For TOT, MVD, and NIGHT crashes, parameter estimates for driveway count variables were inconsistent in their levels of statistical significance and whether one driveway type was associated with fewer or more crashes. In light of these findings, we considered two options. Option 1 used the same driveway definitions and model form for considering driveways as in the current HSM chapter. Option 2 used the total driveway density (driveways per mile) as an alternate variable. For TOT and NIGHT crashes, the model form for Option 1 did not include length, as the inclusion of this variable created poor parameter estimates for the relationship between average annual daily traffic (AADT)
|
From page 85... ...
69 Table 5‐5: Total (TOT) for Base Conditions Option 1 (Urban/Suburban Arterial Segments)
|
From page 86... ...
70 Table 5‐7: Multiple‐Vehicle Non‐Driveway (MVN) for Base Conditions (Urban/Suburban Arterial Segments)
|
From page 87... ...
71 Table 5‐9: Sideswipe‐Same‐Direction (SSD) for Base Conditions (Urban/Suburban Arterial Segments)
|
From page 88... ...
72 Table 5‐11: Multiple‐Vehicle Non‐Driveway Other (MVN OTHER) for Base Conditions (Urban/Suburban Arterial Segments)
|
From page 89... ...
73 Table 5‐13: Nighttime (NIGHT) for Base Conditions (Option 1)
|
From page 90... ...
74 Table 5‐15: Multi‐Vehicle Driveway (MVD) for Base Conditions (Option 1)
|
From page 91... ...
75 Table 5‐17: MN Segment Length and Crash Type Totals for 5 Year Period for Base Condition Sites (Urban/Suburban Arterial Segments) Facility Type No. of segments No. of miles Average AADT TOT Multi‐ Veh Driveway Rear End Head‐On + SOD SSD Multi‐ Veh Other Single‐ Vehicle Night 2U 236 33.86 9,511 320 21 118 25 16 36 115 69 3T 63 7.23 10,841 76 4 39 6 7 17 7 18 4D 92 14.95 22,150 308 4 182 15 31 25 57 63 4U 113 11.72 10,386 160 15 53 9 24 41 22 29 5T 15 1.60 15,753 20 2 6 2 2 6 2 3
|
From page 92... ...
76 Table 5‐18 to Table 5‐21 provide goodness‐of‐fit measures for the initial base condition models validated with the Minnesota data. The cumulative residuals (CURE) plot measures are for CURE plots using the predicted number of crashes on the x‐axis. Only site/crash types with a reasonable number of crashes (approximately 100)
|
From page 93... ...
77 Table 5‐21: Validation of 4U Base Condition Models 5.1.3.2 Final Models Given the reasonable results of the validation exercise and the paucity of the initial model estimation data, we re‐estimated all base condition models using the combined Ohio and Minnesota base condition sites. We re‐estimated all initial crash type models, as well as estimating models by crash severity (all crash types combined) . As with the initial base model development, we used only sites with no lighting, parking, or automated enforcement to develop the base condition SPFs. Because no sites had zero roadside fixed objects and few divided roadways had a median width of exactly 15 feet, we included these variables in the models only if we considered them appropriate for the crash type, and if the variable was statistically significant in the model and with the expected direction of effect. If we included a variable, we would set it to the base condition for application. We attempted to include driveway density in all models developed, acknowledging that driveway presence might affect different crash types in different ways. We entered the number of driveways in a segment in those models in which it was included -- that is, where there was no base condition for the number of driveways in a segment. Final base condition models were calibrated for the following crash types (as defined previously)
|
From page 94... ...
78 Table 5‐23: Total for Base Conditions Combined Data (Urban/Suburban Arterial Segments) Site Type Alpha1 Ohio Beta1 Alpha2 Beta2 Beta3 Beta4 Beta5 2U ‐6.2667 (0.5588)
|
From page 95... ...
79 For KA crashes, no model was calibrated for four‐lane divided (4D) or four‐lane plus two‐way left‐turn‐ lane (5T)
|
From page 96... ...
80 For SSD crashes, a base condition model was not successfully calibrated for 3T sites. We calibrated the recommended model for 3T sites by using all sites and representing average conditions. Table 5‐29: SSD for Base Conditions Combined Data (Urban/Suburban Arterial Segments) Site Type Alpha1 Ohio Beta1 Alpha2 Beta2 Beta3 Beta4 Beta5 2U ‐14.1461 (1.9684)
|
From page 97... ...
81 Table 5‐31: MVN OTHER for Base Conditions Combined Data (Urban/Suburban Arterial Segments) Site Type Alpha1 Ohio Beta1 Alpha2 Beta2 Beta3 Beta4 Beta5 2U ‐11.8140 (1.2722)
|
From page 98... ...
82 Table 5‐33: Nighttime for Base Conditions Combined Data (Urban/Suburban Arterial Segments) Site Type Alpha1 Ohio Beta1 Alpha2 Beta2 Beta3 Beta4 Beta5 2U ‐3.5624 (0.8332)
|
From page 99... ...
83 5.2 INTERSECTIONS Estimation Data Models have been estimated for four‐leg signal‐controlled (4SG) intersections, three‐leg signal‐controlled (3SG)
|
From page 100... ...
84 No right‐turn‐on‐red prohibition (that is, right‐turn‐on‐red is allowed on all legs) No red light cameras Lighting is present. (Note: This is different from what is currently in the HSM. As shown in Table 5‐35, most of the signal‐controlled intersections had lighting, which left us with an insufficient sample of intersections without lighting.)
|
From page 101... ...
85 Table 5‐35: Distribution of categorical variables by intersection type (urban/suburban arterials) Variable 3SG 3ST 4SG 4ST Number of legs with left‐turn lanes 0 485 7214 803 2342 1 301 315 210 74 2 189 48 692 106 3 0 0 323 11 4 0 0 734 2 Number of legs with right‐turn lanes 0 721 7470 1985 2466 1 204 101 430 59 2 50 6 243 8 3 0 0 68 2 4 0 0 36 0 Number of legs with left‐turn lanes on major road 0 619 7282 998 2374 1 323 282 331 69 2 33 13 1433 92 Number of legs with right‐turn lanes on major road 0 865 7523 2286 2496 1 105 54 359 38 2 5 0 117 1 Number of legs with left‐turn lanes on minor road 0 703 7481 1396 2474 1 254 89 430 48 2 18 7 936 13 Number of legs with right‐turn lanes on minor road 0 792 7518 2221 2498 1 177 59 411 33 2 6 0 130 4 Lighting Not Present 91 2407 278 680 Present 884 5170 2484 1855 Number of approaches prohibiting right‐turn‐on‐ red 0 852 7574 2454 2532 1 84 0 98 0 2 39 0 79 0 3 0 0 35 0 4 0 0 96 0 Red light camera Not Present 963 7576 2708 2535 Present 12 0 54 0 Schools within 1000 feet Not Present 849 6961 2420 2289 Present 126 616 342 246 Number of liquor stores within 1000 feet 0 937 7341 2559 2437 1 to 8 38 236 203 98 Number of bus stops within 1000 feet 0 707 6615 2322 2318 1 or 2 32 179 101 50 3 or more 236 783 339 167
|
From page 102... ...
86 Table 5‐36: Descriptive statistics for base condition SPFs (3SG: 345 intersections) Variable Number of crashes Mean Std. Dev Minimum Maximum Major road AADT 12,363 4949 3050 32,109 Minor road AADT 4077 3026 110 18,415 Total Intersection AADT 16,440 5989 4440 44,345 Ratio of Minor to Total Intersection AADT 0.25 0.13 0.02 0.5 KA 62 0.18 0.42 0 2 KAB 375 1.09 1.37 0 9 KABC 854 2.48 2.47 0 13 KABCO 4026 11.67 9.14 0 52 SV_KA 13 0.04 0.19 0 1 SV_KAB 47 0.14 0.38 0 3 SV_KABC 67 0.19 0.47 0 4 SV_KABCO 253 0.73 1.21 0 15 SD_KA 22 0.06 0.24 0 1 SD_KAB 158 0.46 0.75 0 4 SD_KABC 424 1.23 1.38 0 6 SD_KABCO 2302 6.67 5.8 0 39 OD_KA 10 0.03 0.17 0 1 OD_KAB 60 0.17 0.53 0 6 OD_KABC 99 0.29 0.68 0 7 OD_KABCO 369 1.07 1.47 0 11 ID_KA 17 0.05 0.24 0 2 ID_KAB 108 0.31 0.73 0 6 ID_KABC 253 0.73 1.36 0 10 ID_KABCO 974 2.82 3.52 0 24
|
From page 103... ...
87 Table 5‐37: Descriptive statistics for base condition SPFs (3ST: 2082 intersections) Variable Number of crashes Mean Std Dev Minimum Maximum Major road AADT 8187 5221 270 38,460 Minor road AADT 2137 1400 33 18,460 Total Intersection AADT 10,324 5810 540 56,920 Ratio of Minor to Total Intersection AADT 0.23 0.12 0 0.5 KA 198 0.1 0.32 0 3 KAB 840 0.4 0.78 0 7 KABC 1422 0.68 1.15 0 11 KABCO 4756 2.28 3.47 0 49 SV_KA 59 0.03 0.18 0 2 SV_KAB 222 0.11 0.36 0 5 SV_KABC 297 0.14 0.43 0 6 SV_KABCO 952 0.46 0.87 0 13 SD_KA 52 0.02 0.16 0 2 SD_KAB 323 0.16 0.48 0 6 SD_KABC 661 0.32 0.75 0 9 SD_KABCO 2390 1.15 2.37 0 31 OD_KA 43 0.02 0.14 0 1 OD_KAB 128 0.06 0.25 0 2 OD_KABC 184 0.09 0.3 0 3 OD_KABCO 453 0.22 0.54 0 4 ID_KA 43 0.02 0.16 0 3 ID_KAB 163 0.08 0.33 0 4 ID_KABC 272 0.13 0.49 0 9 ID_KABCO 885 0.43 1.08 0 14
|
From page 104... ...
88 Table 5‐38: Descriptive statistics for base condition SPFs (4SG: 589 intersections) Variable Number of crashes Mean Std. Dev Minimum Maximum Major road AADT 11,067 5650 1810 34,960 Minor road AADT 3803 3167 72 27,228 Total Intersection AADT 14,870 7344 2061 56,488 Ratio of Minor to Total Intersection AADT 0.25 0.13 0.01 0.5 KA 148 0.25 0.56 0 4 KAB 767 1.3 1.79 0 14 KABC 1798 3.05 3.67 0 35 KABCO 7253 12.31 12.7 0 109 SV_KA 16 0.03 0.16 0 1 SV_KAB 73 0.12 0.37 0 3 SV_KABC 112 0.19 0.48 0 3 SV_KABCO 409 0.69 1.05 0 9 SD_KA 53 0.09 0.33 0 3 SD_KAB 283 0.48 0.92 0 8 SD_KABC 868 1.47 2.15 0 18 SD_KABCO 3964 6.73 8.32 0 76 OD_KA 27 0.05 0.23 0 2 OD_KAB 167 0.28 0.74 0 6 OD_KABC 309 0.52 1.14 0 10 OD_KABCO 1021 1.73 2.81 0 25 ID_KA 51 0.09 0.29 0 2 ID_KAB 239 0.41 0.83 0 8 ID_KABC 483 0.82 1.34 0 8 ID_KABCO 1671 2.84 3.36 0 26
|
From page 105... ...
89 Table 5‐39: Descriptive statistics for base condition SPFs (4ST: 551 intersections) Variable Number of crashes Mean Std Dev Minimum Maximum Major road AADT 8251 6179 450 37,301 Minor road AADT 2088 1459 50 13,773 Total Intersection AADT 10,339 6658 810 40,111 Ratio of Minor to Total Intersection AADT 0.23 0.13 0 0.5 KA 120 0.22 0.56 0 4 KAB 432 0.78 1.39 0 9 KABC 706 1.28 1.99 0 16 KABCO 1931 3.5 4.58 0 51 SV_KA 20 0.04 0.2 0 2 SV_KAB 61 0.11 0.37 0 2 SV_KABC 72 0.13 0.39 0 2 SV_KABCO 265 0.48 0.81 0 5 SD_KA 15 0.03 0.18 0 2 SD_KAB 84 0.15 0.46 0 4 SD_KABC 219 0.4 1.05 0 14 SD_KABCO 720 1.31 2.91 0 46 OD_KA 21 0.04 0.2 0 2 OD_KAB 60 0.11 0.38 0 3 OD_KABC 83 0.15 0.44 0 3 OD_KABCO 214 0.39 0.82 0 6 ID_KA 64 0.12 0.41 0 4 ID_KAB 225 0.41 0.99 0 7 ID_KABC 328 0.6 1.31 0 9 ID_KABCO 705 1.28 2.24 0 15 Estimated Models We estimated all the models using negative binomial regression with a constant overdispersion parameter and the traditional log‐linear framework. Most previous studies on this topic have used a power function, which provides limited flexibility in the functional form. In this section, we used the Hoerl function to provide more flexibility in the functional form (Hauer, 2015)
|
From page 106... ...
90 Model A included as the starting point the following independent variables in the following form: 𝑌 e e 𝐴𝐴𝐷𝑇 e 𝐴𝐴𝐷𝑇 (5‐2) Model B included as the starting point the following independent variables in the following form: 𝑌 e e 𝐴𝐴𝐷𝑇 e (5‐3)
|
From page 107... ...
91 Table 5‐40: Prediction Models for 3SG Intersections Crash Type Severity Model Form a (S.E.)
|
From page 108... ...
92 Table 5‐41: Prediction models for 3ST intersections Crash Type Severity Model Form a (S.E.)
|
From page 109... ...
93 Table 5‐42: Prediction models for 4SG intersections Crash Type Severity Model Form a (S.E.)
|
From page 110... ...
94 Table 5‐43: Prediction models for 4ST crashes Crash Type Severity Model Form a (S.E.)
|
From page 111... ...
95 Validation of Models To calibrate and validate the models estimated using the data from Ohio, we used six years of data (2010– 15) from North Carolina. Some of the data for calibration were compiled as part of project funded by the North Carolina Department of Transportation (Smith et al. 2016)
|
From page 112... ...
96 Table 5‐45: Summary of Calibration/Validation data from North Carolina for 3ST intersections (52 intersections: 2010 to 2015 data) Variable Sum Mean Standard deviation Minimum Maximum KABCO 304 5.85 8.10 0 36 SV_KABCO 55 1.06 1.55 0 7 SD_KABCO 124 2.38 4.43 0 19 OD_KABCO 48 0.92 1.59 0 8 ID_KABCO 87 1.67 2.38 0 10 Major AADT 7,682 7,232 67 45,733 Minor AADT 1,764 2,227 18 9,500 Total AADT 9,446 7,832 117 45,803 Table 5‐46: Summary of Calibration/Validation data from North Carolina for 4SG intersections (102 intersections: 2010 to 2015)
|
From page 113... ...
97 We considered two basic options for calibration and validation. The first was to estimate a calibration factor following the approach outline in the HSM. The second was to estimate a calibration function (Srinivasan et al. 2016) . As discussed in Srinivasan et al. (2016)
|
From page 114... ...
98 Table 5‐48: Calibration/Validation Results for 3SG intersections Crash Type Option Total Observed Crashes Total Predicted Crashes ln(a)
|
From page 115... ...
99 Table 5‐49: Calibration/Validation results for 3ST intersections Crash Type Option Total Observed Crashes Total Predicted Crashes ln(a)
|
From page 116... ...
100 Table 5‐50: Calibration/Validation results for 4SG intersections Crash type Option Total Observed Crashes Total Predicted Crashes ln(a)
|
From page 117... ...
101 Table 5‐51: Calibration/Validation Results for 4ST Intersections Crash Type Option Total Observed Crashes Total Predicted Crashes ln(a)
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.