Access Management in the Vicinity of Interchanges, Volume 2: Research Overview (2022)

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.

J-1   Supplemental Safety Study Using HSIS Data A P P E N D I X J Introduction To conduct a comprehensive safety study on the crossroads in the vicinity of the diamond interchanges, the research team used Highway Safety Information System (HSIS) crash data for three states: California, Illinois, and Minnesota. The research team used Google Maps and Google Earth to manually extract roadway inventory data for the selected sites. Team members then combined this roadway inventory data with HSIS crash data to form the safety database required for crash severity model development. The research team used mixed effects logistic regression to model binary outcome (injury or no injury) by considering both fixed and random effects of the predictor variables. Data Collection The data collection process consisted of a series of activities to develop a comprehensive severity prediction methodology for the crossroads in the vicinity of diamond interchanges. It consisted of two major tasks. • Collect roadway inventory data for the selected states. • Merge road inventory data with HSIS crash data. Roadway Inventory Database The research team developed a comprehensive roadway inventory database for the diamond interchanges in the selected states. As HSIS databases contain crash information for state- maintained roadways (Interstate, State, and U.S. Highways), the final selected sites are the sites with state-maintained crossroads. Appendix Figure 168 illustrates the state-maintained roadway networks (with the location of the selected diamond interchanges) for California, Illinois, and Minnesota. Appendix Figure 169 illustrates elements of a single diamond interchange. The primary focus of the current study was to analyze the impact of crash severity on the crossroad segments in the vicinity of diamond interchanges.

J-2 Access Management in the Vicinity of Interchanges California Illinois Minnesota Appendix Figure 168. State-Maintained Roadway Networks. Appendix Figure 169. Diamond Interchange Illustrating Different Elements.

Supplemental Safety Study Using HSIS Data J-3  The goal of the data collection procedure was to construct a road inventory database with a wide range of variables. It is important to note that a significant number of variables are the characteristics of the upstream and downstream driveways within the first major access point or intersection. The research team collected driveway information for the first four downstream and upstream driveways. However, not all of the interchanges have four driveways within the first major access point or intersection. The respective fields were blank in values for the cases where the numbers of driveways were less than four. HSIS Crash Data To develop the model, the research team used HSIS crash data for three states. Each of the HSIS states has different variable names, and the data assembly procedure requires a generalization of the variable coding to enable symmetry among the databases. The research team used 3 years (2008–2010) of crash data. Merge Road Inventory and Crash Data Following assembly of the individual datasets, the team next conducted the data-merging task for the state specific HSIS crash data and the enhanced road inventory databases. The objective of this task was to accumulate the highway safety database that included all of the relevant data for model development. Appendix Figure 170 shows the framework of merging the road inventory data with the crash data. Appendix Figure 170. Merging of Road Inventory and Crash Data.

J-4 Access Management in the Vicinity of Interchanges The primary task in conducting a safety study for the crossroads in the vicinity of interchanges included an examination of the merged dataset. Developing models with a large number of variables does not serve the purpose in understanding the possible relationship between the variables. The research team conducted a regression subset selection with an exhaustive search method to reduce the number of variables through linear regression. Descriptive Statistics Ultimately, team members selected a subset of closely associated variables as the study variables for the descriptive statistics. Appendix Table 133 shows the list of important continuous variables. Appendix Figure 171 illustrates the correlation plots for the important continuous variables for California, Illinois, and Minnesota. The correlation coefficients are higher in values for Illinois compared to the other two states. A few common patterns observed from the correlation plots include: • The number of lanes is highly correlated with AADT, and • The number of driveways is correlated with distance to first major access point. The values in Appendix Table 133 show variation in three particular variables: speed limit, Annual Average Daily Traffic (AADT), and median width. To take a closer look at the variables, histograms for these factors are illustrated for the selected states in Appendix Figure 172. The trend of speed limit in California is significantly different from Illinois, and Minnesota. The AADT values for California are mostly centered towards 20,000 vpd, but for Illinois and Minnesota the values are centered toward 30,000 vpd. The median widths for Minnesota also show significant differences when contrasted to the other two states. Appendix Table 133. Significant Continuous Variables. Variable California Illinois Minnesota Mean St.Dev. Min. Max. Mean St. Dev. Min. Max. Mean St. Dev. Min. Max. Dist. to 1st major intersection (mi) 0.13 0.07 0.03 0.34 0.31 0.08 0.12 0.46 0.24 0.08 0.18 0.35 Total number of driveways 6.55 6.12 1.00 17.00 4.14 2.75 1.00 10.00 2.73 0.99 2.00 5.00 Dist. of 1st driveway to ramp (mi) 0.03 0.02 0.01 0.07 0.21 0.05 0.10 0.26 0.12 0.04 0.06 0.16 Median width (ft) 6.44 7.45 0.00 46.00 6.67 12.29 0.00 65.00 8.13 15.12 0.00 51.00 No. of lanes 2.83 1.06 2.00 6.00 3.67 0.74 2.00 4.00 4.69 0.96 4.00 6.00 Speed limit (mph) 47.18 13.01 25.00 65.00 36.56 6.78 30.00 55.00 39.65 10.77 0.00 70.00 AADT (vph) 20,294 9,254 4,900 51,000 27,002 11,555 5,300 39,800 26,347 11,731 6,200 55,220 Crash Severity Analysis

Supplemental Safety Study Using HSIS Data J-5  (a) California (b) Illinois

J-6 Access Management in the Vicinity of Interchanges (c) Minnesota Appendix Figure 171. Correlation Plots of the Important Continuous Variables.

Supplemental Safety Study Using HSIS Data J-7  Appendix Figure 172. Histograms of Speed Limit, AADT, Median Width, and Median Types.

J-8 Access Management in the Vicinity of Interchanges Appendix Table 134 and Appendix Table 135 list the percentage distribution of different categories for the selected states. The values clearly show that the percentage distributions of the categories vary significantly between the different states. Appendix Table 134. Summary Statistics of Driveway-Related Categorical Variables. Category Percentage Median type at downstream driveway 1 CA IL MN TWLTL 50.00% 0.00% 4.85% Painted 29.23% 67.91% 2.42% Raised 4.62% 32.09% 59.39% None 16.15% 0.00% 33.33% Total 100.00% 100.00% 100.00% Land use at downstream driveway 1 Minor commercial 100.00% 77.54% 50.91% Major commercial 0.00% 0.00% 46.67% Residential 0.00% 0.00% 2.42% Farms and Ranches 0.00% 22.46% 0.00% Total 100.00% 100.00% 100.00% Type of business access at downstream driveway 1 Single driveway serving single business 43.08% 32.09% 42.42% Two or more driveways serving single business 32.31% 67.91% 4.85% Single driveway serving multiple businesses 24.62% 0.00% 52.73% Total 100.00% 100.00% 100.00% Median type at upstream driveway 1 TWLTL 50.00% 0.00% 4.85% Painted 45.38% 0.00% 2.42% Raised 4.62% 100.00% 59.39% None 0.00% 0.00% 33.33% Total 100.00% 100.00% 100.00% Land use at upstream driveway 1 Minor commercial 60.77% 77.54% 47.27% Major commercial 23.08% 0.00% 47.27% Undeveloped sites 0.00% 22.46% 4.85% Government buildings 0.00% 0.00% 0.61% Others 16.15% 0.00% 0.00% Total 100.00% 100.00% 100.00% Type of business access at upstream driveway 1 Single driveway serving single business 67.69% 9.63% 45.45% Single driveway serving multiple businesses 21.54% 0.00% 47.27% Two or more driveways serving single business 10.77% 90.37% 7.27% Total 100.00% 100.00% 100.00%

Supplemental Safety Study Using HSIS Data J-9  Appendix Table 135. Summary Statistics of Crossroad Segment Categorical Variables. Category Percentage Major intersection type CA IL MN Signal 80.76% 32.08% 93.93% 2-way stop 8.09% 0.00% 0.60% 1-way stop 11.15% 67.92% 2.45% Roundabout 0.00% 0.00% 3.02% Total 100.00% 100.00% 100.00% Turn lanes at the intersection Single left 56.92% 22.45% 26.06% Single left & right 6.16% 9.64% 48.48% Dual left & single right 0.00% 0.00% 20.01% None 36.92% 67.91% 5.45% Total 100.00% 100.00% 100.00% Median type No median 46.91% 54.79% 72.63% Divided paved 9.47% 35.03% 2.51% Divided unpaved 11.52% 2.65% 13.13% Divided two-way left-turn lane (TWLTL) 24.69% 0.00% 0.00% Undivided striped 7.41% 7.54% 11.73% Total 100.00% 100.00% 100.00% Appendix Table 136 lists the summary of the selected sites with number of crossroad segments with crashes. Appendix Table 136. Summary of the Selected Sites. State No. of studied diamond interchanges No. of state-maintained crossroad segments No. of identi�ied state-maintained crossroad segments with crashes California 22 33 24 Illinois 37 68 17 Minnesota 18 33 22 Model Development To conduct a comprehensive safety study on the crossroads in the vicinity of the diamond interchanges, it is important to assess the crash severity analyses as they relate to the facility type and characteristics. To model binary outcome variables (injury or no injury), the research team used a mixed effect modeling for the model development as the data is considered to be clustered by having both fixed and random effects. The mixed logit method has been widely applied to

J-10 Access Management in the Vicinity of Interchanges analyze crash severities because it has an advantage on accommodating discrete unobserved heterogeneity by allowing parameters to differ across observations by providing more reliable parameter estimates. Except for cases with many observations at each level, the assumption of normally distributed is not accurate. There are varieties of alternatives to perform resampling, for example, Monte Carlo simulation, Bayesian estimation, and bootstrapping. The research team has considered the bootstrapping method as it is more straightforward to implement. By performing bootstrap, a sample size of 4,000 is considered for each state. The estimation values of the variables are shown in Appendix Table 137. The negative sign for the distance of the first major intersection variable can be interpreted that for each additional mile of the distance, there is a reduction in the odds of injury crashes by a certain amount. All of the continuous variables have significance in the level of 95% confidence interval. The estimation values of the variables for the combined model are shown in Appendix Table 138. The research team has applied bootstrapping for the combined model, and a sample size of 4,000 is considered for the model development. In this model, median type is not determined to be associated with the crash severities. AADT, speed limit, and distance to the first major intersection are found significant in the level of 95% confidence interval. For example, the negative sign for the distance of first major intersection variable can be interpreted that for each additional mile of the distance, there is a reduction in the odds of injury crashes by a factor of approximately −0.245 = 0.78 for the combined model. AADT is seen as closely associated with the increase of the injury crashes. With one unit (AADT value of 2,250) increase of AADT, there is an increase in the odds of injury crashes by a factor of approximately 0.314 = 1.37 for the combined model.

Appendix Table 137. Mixed Effect Model Estimation Results for Individual State. Variables and Statistical Measures California Illinois Minnesota Est. St.Err. z Pr (>|z|) Est. St. Err. z Pr(>|z|) Est. St. Err. z Pr(>|z|) Fixed Effect (Intercept) 2.187 1.011 1.075 0.282 0.979 0.457 3.451 0.001 -0.050 0.017 -0.785 0.434 Distance to First Major Intersection -0.581 0.187 -2.143 0.032 -0.178 0.022 -2.784 0.005 -0.103 0.041 -4.857 0.000 AADT (vph) 0.164 0.073 3.987 0.000 0.203 0.045 4.123 0.000 -0.003 0.003 6.421 0.000 Speed (mph) 0.086 0.122 -3.478 0.000 0.078 0.030 2.145 0.032 0.173 0.045 -3.744 0.000 Median Type: Divided Paved 1.342 0.213 4.145 0.000 - - - - - - - - Median Type: Divided Unpaved 1.556 0.443 -2.389 0.025 - - - - 1.789 0.351 4.445 0.000 Median Type: Undivided Striped 1.897 0.127 5.123 0.000 - - - - 2.004 0.575 1.974 0.048 Random Effect Site ID (Intercept): Variance 0.475 0.376 1.014 Site ID (Intercept): Std. Err. 0.143 0.114 0.765 Statistical Measures AIC 4632 2759 4453 BIC 4714 2809 4522 Log likelihood -2303 -1371 -2216 Deviance 4606 2743 4431 DF Residual 3987 3992 3989

J-12 Access Management in the Vicinity of Interchanges Appendix Table 138. Combined Mixed Effect Model Estimation Results. Variables and Statistical Measures Combined Est. St. Er. z Pr(>|z|) Fixed Effect (Intercept) 0.945 0.874 2.554 0.011 Distance to 1st Major Intersection (mi) -0.245 0.089 1.997 0.045 AADT (vph) 0.314 0.647 -2.742 0.006 Speed limit (mph) 0.091 0.102 2.415 0.016 Median Type: Strategic raised - - - - Median Type: Painted - - - - Median Type: Painted and TWLTL - - - - Random Effect Site ID (Intercept): Variance 0.845 Site ID (Intercept): Std. Err. 0.356 Statistical Measures AIC 4883 BIC 4935 Log likelihood -2169 Deviance 4714 DF. Residuals 4170 State-Specific Crash Severity Models California: log ( ) = 2.187 − 0.581 × + 0.164 × + 0.086 × + 1.342 × : + 1.556 × : + 1.897 × : + (1| ) where (1|Site ID) indicates the random effects of the sites. Illinois: log ( ) = 0.979 − 0.178 × + 0.203 × + 0.078 × + (1| )

Supplemental Safety Study Using HSIS Data J-13  Minnesota: log ( ) = −0.050 − 0.103 × − 0.003 × + 0.173 × + 1.789 × : + 2.004 × : + (1| ) Combined Model: log ( ) = −0.050 − 0.103 × − 0.003 × + 0.173 × + 1.789 × : + 2.004 × : + (1| ) California Example Problem To identify the impact of crash severity on crossroads in the vicinity of the interchanges, the research team considered using a mixed effect logistic model. The research team identified important variables from a wider range of variables associated with the crash severity. The research team estimated a mixed effects logistic model for California with distance from first major intersection, AADT, and speed limit as crash level continuous predictors, Median type as crash level categorical variable, and a random intercept by Site ID. The developed model is log ( ) = 2.187 − 0.581 × + 0.164 × + 0.086 × + 1.342 × : + 1.556 × : + 1.897 × : + (1| ) The interpretation of the modeling results are: - The negative sign for the distance of first major intersection variable indicates that for each additional mile of the distance, there is a reduction in the odds of injury crashes by a factor of approximately 0.56 [ −0.581 = 0.56]. - With one unit (AADT value of 2,250 vpd) increase of AADT, there is an increase in the odds of injury crashes by a factor of approximately 1.18 [ 0.164 = 1.18]. - The roadway inventory variables, like median type, are found to be associated with the crash severities. Crossroad segments with divided, paved medians show a reduction in the odds of injury crashes by a factor of 0.81 [ 1.342 1.556 = 0.81] compared to divided, unpaved medians.

J-14 Access Management in the Vicinity of Interchanges Example problem conclusions: The estimation results show that AADT and speed limit associate with the increase of injury severity. On the other hand, the distance of the first major intersection from the ramp associates with the decrease of injury severity. Divided, paved medians show higher safety impact compared to the other two alternatives: divided unpaved and undivided striped. Summary Comments The site characteristics and resulting severity models vary dramatically for the three HSIS states; however, some common conclusions can be drawn as follows: • As AADT increases, injury severity also increases. • As speed limit increases, injury severity increases. • As the distance to the first major intersection increases, the injury severity decreases. • Divided, paved medians generally have a lower number of severe crashes than other median types. It is notable that these characteristics were also identified as significant factors for the companion operational analysis, but driveway placement and density also tended to be critical factors for operational performance as well.

Abbreviations and acronyms used without denitions in TRB publications: A4A Airlines for America AAAE American Association of Airport Executives AASHO American Association of State Highway Officials AASHTO American Association of State Highway and Transportation Officials ACI–NA Airports Council International–North America ACRP Airport Cooperative Research Program ADA Americans with Disabilities Act APTA American Public Transportation Association ASCE American Society of Civil Engineers ASME American Society of Mechanical Engineers ASTM American Society for Testing and Materials ATA American Trucking Associations CTAA Community Transportation Association of America CTBSSP Commercial Truck and Bus Safety Synthesis Program DHS Department of Homeland Security DOE Department of Energy EPA Environmental Protection Agency FAA Federal Aviation Administration FAST Fixing America’s Surface Transportation Act (2015) FHWA Federal Highway Administration FMCSA Federal Motor Carrier Safety Administration FRA Federal Railroad Administration FTA Federal Transit Administration GHSA Governors Highway Safety Association HMCRP Hazardous Materials Cooperative Research Program IEEE Institute of Electrical and Electronics Engineers ISTEA Intermodal Surface Transportation Efficiency Act of 1991 ITE Institute of Transportation Engineers MAP-21 Moving Ahead for Progress in the 21st Century Act (2012) NASA National Aeronautics and Space Administration NASAO National Association of State Aviation Officials NCFRP National Cooperative Freight Research Program NCHRP National Cooperative Highway Research Program NHTSA National Highway Traffic Safety Administration NTSB National Transportation Safety Board PHMSA Pipeline and Hazardous Materials Safety Administration RITA Research and Innovative Technology Administration SAE Society of Automotive Engineers SAFETEA-LU Safe, Accountable, Flexible, Efficient Transportation Equity Act: A Legacy for Users (2005) TCRP Transit Cooperative Research Program TDC Transit Development Corporation TEA-21 Transportation Equity Act for the 21st Century (1998) TRB Transportation Research Board TSA Transportation Security Administration U.S. DOT United States Department of Transportation

Transportation Research Board 500 Fifth Street, NW Washington, DC 20001 ADDRESS SERVICE REQUESTED ISBN 978-0-309-67424-9 9 7 8 0 3 0 9 6 7 4 2 4 9 9 0 0 0 0