Safety performance of rural four-lane undivided roadways and rural four-lane roadways with a two-way left-turn lane

Georgia DOT Research Project 19-15 Final Report
SAFETY PERFORMANCE OF RURAL FOUR-LANE UNDIVIDED ROADWAYS AND RURAL FOUR-LANE ROADWAYS WITH A TWO-WAY LEFT-TURN LANE
Office of Performance-based Management and Research
600 W Peachtree NW Atlanta, GA 30308

TECHNICAL REPORT DOCUMENTATION PAGE

1. Report No.

2. Government Accession No.

FHWA-GA-21-1915

N/A

4. Title and Subtitle

Safety Performance of Rural Four-Lane Undivided Roadways and Rural Four-Lane

Roadways with a Two-Way Left-Turn Lane

3. Recipient's Catalog No. N/A 5. Report Date December 2020 6. Performing Organization Code

N/A

7. Author(s) Jun Liu, Ph.D. https://orcid.org/0000-0002-6336-4931 Praveena Penmetsa, Ph.D. https://orcid.org/0000-0001-9003-7568 Xiaobing Li, Ph.D. https://orcid.org/0000-0003-3425-743X Timothy E. Barnett, P.E., PTOE, RSP. https://orcid.org/0000-0002-7073-8751

8. Performing Organization Report No. N/A

9. Performing Organization Name and Address Alabama Transportation Institute (University Transportation Center for Alabama and Alabama Transportation Policy Research Center) The University of Alabama 248 Kirkbride Lane, 3rd floor Cyber hall, Tuscaloosa, AL 35487

10. Work Unit No. N/A 11. Contract or Grant No. RP 19-15

12. Sponsoring Agency Name and Address Georgia Department of Transportation Office of Performance-based Management and Research 600 W Peachtree St NW | Atlanta, GA 30308

13. Type of Report and Period Covered Final Report (August 2019 December 2020)
14. Sponsoring Agency Code N/A

15. Supplementary Notes

Conducted in cooperation with the U.S. Department of Transportation, Federal Highway Administration.

16. Abstract

Non-traversable medians consistently yielded improved safety performance compared to other median types such as undivided, 4feet flush medians, and two-way left-turn lane cross-sections. However, constructing non-traversable medians can be costly. The goal of this study is to 1) examine the safety performance of existing rural four-lane roadways with above-mentioned four median types in Georgia by using Safety Performance Functions (SPFs) and Crash Modification Factors (CMFs), and 2) develop criteria to determine under what conditions these four median types yield maximum safety benefits while considering construction costs. Data were refined and integrated from multiple sources such as from the Georgia Department of Transportation (GDOT), Federal Highway Administration (FHWA) and Google Maps. The Annual Average Daily Traffic (AADT), truck percentage, and access point density were considered as key independent variables in SPFs. The CMFs were estimated to show the effectiveness of a cross-section compared to the base- four-lane undivided roadway. Note, the SPFs and CMFs developed in this study did not consider the speed limit. The key results show that the estimated CMFs vary across different values of variables, indicating that the safety effectiveness of a treatment is likely to vary across different roadway and traffic conditions. Specifically, the segments with non-traversable medians outperformed the other three segment types across all AADTs, truck percentages, and access point densities, except at very low AADT under 5,000 where the 4-feet flush medians appear to have improved safety. The research team estimated average annual crash reductions (compared with undivided roadways), which were converted into monetary values using the average crash costs by severities. The safety benefits and project construction costs were used to estimate benefit-cost ratios (BCRs). Simulations were suggested in situations where safety solely did not help in the decision-making process of identifying cost-effective median type for rural four-lane roadways. It is highly recommended that decision-makers or practitioners use the estimated safety benefits from this study and the construction costs of a specific highway project to estimate BCRs for recommendations of the cross-section type.

17. Key Words Rural four-lane roadway; Undivided roadway; 4-feet flush median; Two-way left-turn lane; Non-traversable median; Safety Performance Functions, Crash Modification Factors

18. Distribution Statement No restrictions. This document is available through the National Technical Information Service, Springfield, VA 22161.

19. Security Classif. (of this report) Unclassified

20. Security Classif. (of this page) 21. No. of Pages 22. Price

Unclassified

140

Free

Form DOT F 1700.7 (8-72)

Reproduction of completed page authorized

i

GDOT Research Project 19-15 Final Report
SAFETY PERFORMANCE OF RURAL FOUR-LANE UNDIVIDED ROADWAYS AND RURAL FOUR-LANE ROADWAYS WITH A TWO-WAY LEFT-TURN LANE By Jun Liu, Ph.D., Principle Investigator Assistant Professor, Department of Civil, Construction, and Environmental Engineering College of Engineering, The University of Alabama jliu@eng.ua.edu
Praveena Penmetsa, Ph.D., Co-Principle Investigator Associate Research Engineer, Alabama Transportation Institute
The University of Alabama ppenmetsa@ua.edu
Xiaobing Li, Ph.D., Co-Principle Investigator Associate Research Engineer, Alabama Transportation Institute
The University of Alabama xli158@ua.edu
Timothy E. Barnett, P.E., PTOE, RSP, Co-Principle Investigator Operations & Safety Engineer, Alabama Transportation Institute
The University of Alabama tebarnett1@ua.edu
Contract with Georgia Department of Transportation
In cooperation with U.S. Department of Transportation Federal Highway Administration
December 2020 The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Georgia Department of Transportation or the Federal Highway Administration. This report does not constitute a standard, specification, or regulation.
ii

ACKNOWLEDGEMENTS The authors acknowledge and appreciate the Georgia Department of Transportation (GDOT) for providing financial support for this project. Further, a special thanks to Daniel Pass, Sunil Thapa, Scott Zehngraff, David Adams, Michelle Pate, Frank Flanders, Supriya Kamatkar of GDOT for providing excellent support, guidance and valuable inputs for the successful completion of this project. In addition, the efforts put by the graduate students Chenxuan Yang and Zihe Zhang and administrative staff Connie Harris, MargaretAnn Corbett, and Sherri Mink at the University of Alabama (UA) are invaluable and recognized.
iii

EXECUTIVE SUMMARY Non-traversable medians (NTMs) consistently yielded improved safety performance compared to other median types such as undivided, 4-feet flush medians (4FMs), and two-way left-turn lane (TWLTL) cross-sections. However, constructing NTMs can be costly because of added right-ofway and additional construction costs. The goal of this study is (1) to examine the safety performance of existing rural four-lane (4L) roadways with NTM, TWLTL, 4FM, and UR (undivided roadways) with speed limit 50 mph or higher in Georgia, and (2) to develop criteria to determine under what conditions these four median types yield maximum safety benefits while considering the cost of construction.
Extensive data cleaning and manipulation were adopted to refine the data. Data were pulled from various sources, including the Georgia Department of Transportation (GDOT), Federal Highway Administration (FHWA), and Google Maps. Manual data extractions were performed for uncompiled data such as the number of access points, along with manual verification of median types. Traffic volume data, crash data and roadway data were integrated into datasets for safety performance analysis.
Safety Performance Functions (SPFs) and Crash Modification Factors (CMFs) were used to create the criteria for median recommendation. SPFs were developed for all the four median types. The Annual Average Daily Traffic (AADT), truck percentage, and access point density were considered as independent variables in the SPFs. The SPFs were developed to predict the crash frequencies for different severities, which were used in the criteria. The CMFs were calculated to show the effectiveness of a cross-section compared to the base cross-section, which is a 4L-UR.
iv

The CMFs vary across different values of AADT, segment length, truck percent, and access point density, indicating that the safety effectiveness of a treatment on roadways is also likely to vary across different situations. The key findings based on estimated CMFs are as follows:
The segments with NTMs outperformed the other three segment types across all AADTs, truck percentages, and access point densities, except at very low AADT around or lower than 5,000 where the 4FMs appear to be associated with an improved condition.
AADT Is a principal factor to show the effectiveness of a cross-section compared to the base. In general, the 4FMs appear to have an improved performance than UR segments when the AADT is around or lower than 10,000; when the AADT increases (around or higher than 10,000), the TWLTLs have an improved performance than UR sections and 4FMs; when the AADT reaches 20,000, the NTMs may be considered.
Truck percentage Is positively related to the safety effectiveness of 4FM, TWLTLs, and NTMs compared to UR sections. In other words, converting a 4L-UR segment into one of the other three cross-section types can result in an even further improved condition for traffic with a higher truck percentage.
Access point density Is negatively related to the safety effectiveness of 4FMs, TWLTLs and NTMs compared to the base segment type. In other words, the safety effectiveness of these three types of medians decreases with the increase in the number of access points along a segment.
Note that, the safety performance on roadways may be associated with the posted speed limits. The estimation of SPFs/CMFs in this study did not consider the impact of speed limits. According
v

to a preliminary study in Appendix A, a higher speed limit is likely associated with a greater rate for fatal and serious injury crashes.

Besides estimating CMFs, to facilitate criteria development and make recommendations on the median type, the research team estimated average annual crash reductions (compared with URs). These crash reductions were converted into monetary values using the average crash costs by severities. The safety benefits (shown below) and the project costs could be used to estimate benefit-cost ratios.

Safety benefits (in $million) for 4FMs, TWLTLs, and NTMs

4-ft Flush Median

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 to >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 30

<=5%

$7 $7 $8 $6 $16 $17 $17 $16 $17 $15 $13 $18 $8 $2 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% $8 $9 $9 $8 $18 $18 $18 $17 $17 $16 $14 $18 $8 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>10% to <=15% $10 $10 $11 $9 $19 $20 $20 $18 $18 $16 $15 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% $12 $12 $13 $11 $20 $21 $22 $20 $18 $17 $15 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

$14 $15 $16 $13 $22 $23 $24 $21 $18 $17 $16 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5%

$3 $2 $2 $3 $14 $14 $13 $15 $30 $30 $29 $30 $49 $49 $48 $49 $71 $71 $70 $69 $93 $95

>5% to <=10% $5 $5 $5 $5 $19 $19 $19 $18 $35 $35 $35 $34 $51 $53 $53 $50 $69 $71 $72 $67 $87 $90

>10% to <=15% $8 $8 $8 $7 $23 $23 $24 $22 $37 $39 $40 $36 $52 $54 $56 $49 $66 $69 $71 $63 $80 $84

>15% to <=20% $10 $11 $11 $10 $26 $27 $28 $24 $39 $41 $43 $37 $51 $53 $56 $48 $62 $65 $68 $59 $73 $77

>20%

$13 $14 $15 $12 $28 $30 $32 $26 $39 $42 $44 $37 $49 $52 $55 $46 $57 $61 $64 $54 $65 $69

$95 $91 $92 $84 $87 $77 $80 $69 $73 $61

<=5% >5% to <=10% Nontraversable >10% to <=15% >15% to <=20%
>20%

N/A N/A N/A N/A $3 $2 $1 $3 $19 $20 $20 $19 $41 $43 $44 $39 $66 $70 $73 $62 $93 $99 $106 $87 N/A N/A N/A N/A $9 $9 $9 $9 $25 $26 $26 $23 $43 $45 $48 $40 $62 $66 $71 $59 $83 $89 $95 $78 $2 $2 $2 $2 $14 $15 $15 $13 $28 $30 $31 $27 $43 $46 $49 $40 $59 $63 $67 $55 $74 $80 $85 $69 $6 $6 $6 $5 $18 $19 $20 $17 $31 $33 $35 $29 $43 $46 $49 $40 $54 $58 $62 $51 $66 $71 $76 $61 $9 $10 $10 $9 $22 $24 $25 $20 $32 $35 $37 $30 $42 $45 $48 $39 $50 $54 $58 $47 $58 $62 $67 $54

Notes: The base cross-section type is undivided roadways. The safety benefits cover reductions in all crash

types. The safety benefits of 20 years are estimated. Cells with "N/A" are estimated with negative safety

benefits. The results are applicable to roadways with posted speed limits of 50 mph and higher.

The research team gathered project costs from GDOT and several openly available sources and estimated Benefit-Cost Ratios (BCRs) considering the upper bound of the project costs and developed the following criteria for cross-section recommendation.
vi

Preliminary criteria for recommending cross-section on four-lane rural roadways

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

A A A A B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>5% to <=10% A/B A/B A/B A/B B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>10% to <=15% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>15% to <=20% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

>20%

B B B B B/C B/C B/C B/C B/C B/C/D B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

Notes: A: Four-Lane Undivided

B: Four-Lane Divided w/4-ft Flush Median

C: Four-Lane Divided w/TWLTL

D: Four-Lane Divided w/Non-Traversable Median or Barrier

When multiple letters (e.g., A/B, or B/C) are given in a cell, simulations were suggested to further examine and compare the performance of alternative cross-section designs. The performance may be examined from safety, mobility, and environmental aspects. It is important to note that, the criteria table provides rough cross-section recommendations merely based on the estimated crash reductions, empirical crash costs, and project costs from the previous projects. The project cost is project-specific, and it can vary substantially from location to location in a state due to various factors, including the prior conditions of a project site, local labor cost, and material cost, etc. The findings are limited to the scope of this project (rural 4L roadways with NTM, TWLTL, 4FM, and UR with speed limit 50 mph or higher in Georgia). It is highly recommended that decision-makers or practitioners use the safety benefits table to compare with the cost of a specific proposed project and make a more realistic recommendation of the cross-section type.

vii

TABLE OF CONTENTS Contents Chapter 1: INTRODUCTION .........................................................................................................1
Background ..................................................................................................................................1 Project Objectives........................................................................................................................3 Chapter 2: LITERATURE REVIEW ..............................................................................................4 State DOTs...................................................................................................................................4 Researchers ..................................................................................................................................7 Chapter 3: DATA & METHODS ..................................................................................................14 Traffic Crash Data .....................................................................................................................14 Traffic Count Data.....................................................................................................................14 Roadway Data............................................................................................................................15
Undivided roadway segments................................................................................................17 4-ft Flush Medians.................................................................................................................18 TWLTLs ................................................................................................................................20 Non-traversable......................................................................................................................21 Access Points .........................................................................................................................23 Data Linkages to Segments .......................................................................................................23 Modeling Background ...............................................................................................................27 Methodology to Estimate CMFs................................................................................................30 Chapter 4: DESCRIPTIVE ANALYSIS .......................................................................................32 Traffic Crashes...........................................................................................................................32 Access Points .............................................................................................................................39
viii

Traffic Volume ..........................................................................................................................42 Chapter 5: CRASH RATES...........................................................................................................45
All Crashes.................................................................................................................................45 Crash Severity............................................................................................................................46 Crash Type.................................................................................................................................48 Single- and Multi-vehicle Crashes.............................................................................................49 Chapter 6: SAFETY PERFORMANCE FUNCTIONS ................................................................51 Chapter 7: CRASH MODIFICATION FACTORS.......................................................................53 Chapter 8: CRITERIA DEVELOPMENT.....................................................................................60 Chapter 9: GUIDELINES for MICROSIMULATION .................................................................71 Chapter 10: SUMMARY & CONCLUSIONS..............................................................................78 REFERENCES ..............................................................................................................................82 APPENDIX A. Speed and Safety Performance on Rural Four-lane Roadways. ..........................89 APPENDIX B. CMF Plots along AADT for different Truck Percentages and Access Point Densities. .......................................................................................................................................93 APPENDIX C. Estimated average CMFs for KAB and CO crashes. .........................................124
ix

LIST OF FIGURES Figure 1. Critical AADT values for 4l roads for the higher speed category (Pernia et al., 2004)...5 Figure 2. Collision difference for varying ADT and approach density (Phillips, 2004). ................8 Figure 3. Cost-effectiveness of TWLTL based on operation cost savings (Graham et al., 2014). .9 Figure 4. Cost-effectiveness of TWLTL based on accident cost savings (Graham et al., 2014). .10 Figure 5. Traffic count data example.............................................................................................15 Figure 6. Traffic count data example.............................................................................................16 Figure 7. HPMS data. ....................................................................................................................17 Figure 8. Examples of rural four-lane undivided roadway and identified locations of rural fourlane undivided roadways segment. ................................................................................................18 Figure 9. Examples of rural four-lane 4-ft flush median roadway and identified locations of rural four-lane 4-ft flush median roadways segments. ...........................................................................19 Figure 10. Examples of rural four-lane TWLTL roadway and identified locations of rural fourlane TWLTL roadways segments. .................................................................................................20 Figure 11. Examples of rural four-lane non-traversable roadway and identified locations of rural four-lane non-traversable roadways segments...............................................................................21 Figure 12. An example of counting the access points along a sampled roadway segment. ..........24 Figure 13. Examples of overlaid homogeneous TWLTL roadway segment with 2018 crash data (Left); and overlaid homogeneous 4-ft Flush Median roadway segment with 2018 crash data (Right) before the linkage. (Grey dots represent the crashes) .......................................................25 Figure 14. Example of overlaid homogeneous TWLTL roadway segment with all crash data (Left); and overlaid homogeneous 4-ft Flush Median roadway segment with all crash data (Right) after the linkage. (Grey dots represent the crashes) ..........................................................26
x

Figure 15. Spatial distribution of AADT stations (black dots) that are linked to the sampled roadway segments..........................................................................................................................27 Figure 16. Scatter plot of vehicle AADT and truck percentage. ...................................................29 Figure 17. Number of crashes per mile along the rural four-lane undivided segments.................36 Figure 18. Number of crashes per mile along the rural four-lane 4-ft flush median segments. ....37 Figure 19. Number of crashes per mile along the rural four-lane TWLTL segments. ..................38 Figure 20. Number of crashes per mile along the rural four-lane non-traversable segments........39 Figure 21. Number of access points per mile along the sampled rural four-lane segments (a) undivided; (b) 4-ft flush median; (c) TWLTL; (d) non-traversable. .............................................41 Figure 22. Distribution of the crash rates: (a) under 1000 crashes per 100-million VMT. ...........46 Figure 23. Seven Key Steps in Microsimulation Analysis (Wunderlich et al., 2019)...................72
xi

LIST OF TABLES Table 1. A summary of highly relevant studies for rural as well as urban/suburban 4-lane roadways. .......................................................................................................................................11 Table 2. Data procured. .................................................................................................................14 Table 3. Frequency of homogeneous segments for safety analysis by segment length. ...............23 Table 4. Crashes statistics for undivided, 4-ft flush median, TWLTL, and non-traversable roadway segments..........................................................................................................................33 Table 5. Crashes statistics for different median facility types in 2018..........................................35 Table 6. Number of access points and density by median type.....................................................40 Table 7. AADT and truck percentage for 4L roads with different median types from 2013 to 2018. ..............................................................................................................................................43 Table 8. Average traffic volume (AADT) for different median types. .........................................44 Table 9. snapshot of segments frequency and percentage by average traffic volume (AADT) and truck percentage for different median types. .................................................................................44 Table 10. Crash rates for undivided, 4-ft flush median, TWLTL, and non-traversable segments by 100-million VMT......................................................................................................................46 Table 11. Crash rates by KABCO. ................................................................................................47 Table 12. Crash rates by crash type. ..............................................................................................49 Table 13. Crash rates for single- and multi-vehicle crashes. .........................................................50 Table 14. SPFs by crash severity...................................................................................................52 Table 15. CMFs based on AADT with UR segments as base (Access point density - 10 per mile). .......................................................................................................................................................55
xii

Table 16. CMFs based on AADT with UR segments as base (Access point density - 20 per mile). .......................................................................................................................................................56 Table 17. CMFs based on AADT with UR segments as base (Access point density - 30 per mile). .......................................................................................................................................................57 Table 18. CMFs based on AADT with UR segments as base (Access point density - 40 per mile). .......................................................................................................................................................58 Table 19. CMFs based on AADT with UR segments as base (Access point density - 50 per mile). .......................................................................................................................................................59 Table 20. Average CMFs for given road/traffic conditions (KABCO crashes). ...........................61 Table 21. Average crash reduction per year for given road/traffic conditions. .............................62 Table 22. Safety benefits for different crash severities (GDOT, 2019a).......................................62 Table 23. Crash severity distributions on four cross-section types. ..............................................63 Table 24. Safety Benefits (in $million) for 4FMs, TWLTLs, and NTMs compared with URs. ...64 Table 25. The per-mile cost differences for roadways with 4FMs, TWLTLs, and NTMs are compared with URs. ......................................................................................................................65 Table 26. BCRs for 4FMs, TWLTLs, and NTMs (the base cross-section is UR).........................66 Table 27. Preliminary criteria for recommending cross-section on four-lane rural roadways. .....69
xiii

LIST OF ABBREVIATIONS

AADT

Annual Average Daily Traffic

AASHTO American Association of State Highway and Transportation Officials

ADT

average daily traffic

AIC

Akaike Information Criterion

APD

access point density

BCR

Benefit-cost ratio

BLS SIO Bureau of Labor Statistics Southeast Information Office

CMF

Crash Modification Factors

DOT

Department of Transportation

DON

Department of Numbers

FDOT

Florida Department of Transportation

FHWA

Federal Highway Administration

FM

Flush Median

GDOT

Georgia Department of Transportation

GRIP

Governors' Road Improvement Program

GIS

Geographic Information System

HPMS

Highway Performance Monitoring System

HSM

Highway Safety Manual

ITD

Idaho Transportation Department

MPO

metropolitan planning organization

NB

negative binomial

NPV

Net Present Values

xiv

NTM SCDOT SPF TWLTL TxDOT UR TP 4FM 4L

non-traversable medians South Carolina Department of Transportation Safety Performance Function two-way left-turn lane Texas Department of Transportation undivided roadways truck percentage four-feet flush median four-lane

xv

Chapter 1: INTRODUCTION
Background The Governors' Road Improvement Program (GRIP) in Georgia was initiated in 1989 to improve the transportation infrastructure across the state. The current extent of the GRIP system consists of 19 corridors with more than 3,300 centerline miles of roadway. When completed, the GRIP system will place 98 percent of Georgia's population within 20 miles of a four-lane highway and connect 95 percent of cities with a population of more than 2,500 to the Interstate System (GDOT, 2018a). Studies indicate that the GRIP fosters economic growth in several ways, including expanding access to markets, reducing shipping costs, increasing per capita income, decreasing unemployment rates, and adding buying power (Bachtel et al., 1998; Humphreys, 2003).
Per the GRIP, the Georgia Department of Transportation (GDOT) is committed to convert existing primary routes and truck access routes into four-lane (4L) roadways. When a 4L road is divided with non-traversable medians (NTMs), it provides more effective and efficient transportation, and safer travel compared to a two-lane highway (Council et al., 1999). However, due to the cost of right-of-way acquisition, construction costs, and/or terrain restrictions in certain areas, 4L Undivided Roadways (UR) or 4L divided roadways with 4-feet wide Flush Medians (4FM) are sometimes proposed and constructed. The environmental impact-related costs should also be considered as part of the project cost; such costs may include the traffic delay costs, and extra fuel consumption and increased emissions due to a project construction (Clavenger and Kociolek, 2006). Studies have found that 4L-UR can experience a degradation of service and/or safety as traffic volumes increase (Knapp and Giese, 2001). The crash rate on 4L-UR can be higher than on two-lane roads because 4L roads carry greater traffic volumes, have a higher frequency of
1

intersections and other access points, and have greater development of adjacent land (AASHTO, 2011). Within 4L roads, divided roads with NTMs have measurably lower collisions compared to URs (Mohamedshah et al., 1994; TxDOT, 2020). 4L-UR should be proposed only under reasonable circumstances that consider the posted speed limits, traffic volume, truck percentage, access points and other factors that could affect traffic safety (Shea et al. 2000).
A 4L roadway with a two-way left-turn lane (TWLTL) may be considered in place of a 4L-UR section as it has an added benefit of providing a separate lane for left-turning traffic that provides the safety benefit of separating left-turning traffic from higher speed and higher volume of through traffic. As the number of access points increases, both safety and operational benefits are recognized with the use of a TWLTL on 4L roads (Ballard and McCoy, 1983; Hovey and Chowdhury, 2005). When average daily traffic volumes exceed 20,000 ~ 24,000 and the demand for mid-block turns is high, a raised or non-traversable median rather than a TWLTL would be recommended (Kentucky DOT, 2019; TxDOT, 2020).
It is essential to understand under what roadway and traffic conditions a given median type yields maximum safety benefits considering its cost effectiveness especially on 4L rural roads. This could help design engineers choose an optimal median type and be proactive in ensuring the safety of drivers.
2

Project Objectives The objectives of this project are: (1) to examine the safety performance of existing rural 4L roadways with NTM, TWLTL, 4FM, and UR in Georgia, and (2) to develop criteria to determine under what conditions these four median types yield maximum safety benefits while considering the cost of construction. Specifically, this project:
1. Investigated characteristics of crashes on rural 4L roads with NTM, TWLTL, 4FM, and UR in Georgia;
2. Developed safety performance functions (SPFs) with key factors that significantly influence crash frequencies on 4L roads;
3. Identified key characteristics such as traffic volume, truck percentage, posted speed limit, lane width and access point density which are related to safety sharp deterioration to an undesirable level on rural 4L roads with NTM, TWLTL, 4FM, and UR;
4. Developed a set of guidance criteria to determine when it would be appropriate to consider an NTM, TWLTL, 4FM, and UR on 4L roads in rural areas.
To achieve the objectives of this study, a data-driven approach was adopted following the recommendations listed out in the Highway Safety Manual (HSM).
The rest of the report is organized as follows: literature review, data and methodology, descriptive analysis of the data, safety performance functions, crash modification factors, criteria development, and summary.
3

Chapter 2: LITERATURE REVIEW A detailed review of the literature was conducted to examine studies that were relevant to this study's objectives and are presented in this section. The authors split the review results into DOT and scholarly guidelines.
State DOTs Iowa DOT issues the Design Manual for highways in Iowa (Iowa DOT, 2019). Per Chapter 6, Geometric Design, TWLTLs are recommended only for suburban and urban areas as they could be confused with passing lanes in rural areas. The average daily traffic (ADT) over 10,000 to 12,000 vehicles per day (vpd) would warrant consideration of a TWLTL on a 4L road. When the ADT on a street exceeds about 17,000 vpd, 4L with raised medians or five-lane roadways with TWLTL is more appropriate designs. The limit for TWLTL facilities is approximately 24,000 ADT. TWLTL should generally not be used in when ADT is greater than 24,000 vpd.
Florida DOT sponsored a study to investigate the safety issues related to TWLTLs and the study analyzed more than 1600 road segments with TWLTLs and identified critical traffic volumes by access point density for 4L roads (Pernia et al., 2004). Figure 1 shows the critical AADT (Annual Average Daily Traffic) values for safety improvements at road segments with TWLTLs.
4

Figure 1. Critical AADT values for 4l roads for the higher speed category (Pernia et al., 2004).
Alabama DOT studied approximately 300 miles of roadways in Alabama and proposed 4 different alternatives to improve both operational and safety deficiencies. They relied on previously conducted studies to make alternative determinations (Barnett and Wallace, 2010).
Kentucky DOT's Highway Access Management provided the following guidelines for TWLTL on multi-lane roads in urban/suburban areas: 1) projected ADT of less than 24,000 vpd 2) access point density between 10 and 85 per mile, 3) left turn volume less than 100 vph. Further, they also recommend flush medians to be used when access point density is less than 10 per mile. However, the flush medians are to provide a consistent cross-session rather than for safety or traffic
5

operations (KDOT, 2019). Note, TWLTLs are also considered as flush medians. In this reference, the width of flush medians is not specified.
Texas DOT's Roadway Design Manual suggests the use of TWLTL on 4L suburban roads when the future ADT is around 6,000 vpd and with over 10 entrances per mile (TxDOT, 2020). Another study conducted by Texas Transportation Institute in 1993 states that there are no statistical differences in accident rates of highways with TWLTLs and highways with flush medians when driveway densities are low (e.g., less than 9 driveways per mile) (Balke and Fitzpatrick, 1993). There were also no differences in the way that flush medians and TWLTLs function on rural 4L roads. They recommended the use of flush medians only on highways where the frequency and spacing of driveways permit individual median openings at each driveway. In cases where this is not possible, they recommended the use of TWLTLs on 4L rural highways. Note, TWLTLs are also considered as flush medians. In this reference, the width of flush medians is not specified.
Oklahoma DOT recommends the consideration of TWLTLs in urban and suburban areas, when there is a large number of driveways per mile (> 45 driveways total per mile on both sides), there is a commercial area with high left-turn volumes, and the ADT is between 10,000 and 25,000 (ODOT, 2019).
Georgia DOT has recommendations for TWLTLs in urban and suburban areas (GDOT, 2019c; GDOT, 2019d). The TWLTLs would be recommended when design speeds are 45 mph or lower, and current ADT is less than 18,000 or future ADT is less than 24,000 vpd. Research data showed
6

that after the construction of TWLTLs, the roadway capacity increased by 30%, delay decreased by 30%, and traffic crashes decreased by 35%.
South Carolina DOT recommends TWLTL for 45 mph facilities with four or less lanes, when there are 10 to 35 driveways per mile on both sides of the road, in high-density commercial areas with substantial mid-block left turns. The TWLTLs could be recommended for rural highways, but they are typically near suburban areas or roads passing through small towns. Median width less than 12-feet was not recommended where posted speeds are greater than 35 mph and the percentage of trucks, buses and recreational vehicles is greater than 5 percent of the AADT (SCDOT, 2019).
Idaho DOT considered TWLTL when AADT is equal to or less than 28,000 vehicles per day in urban or suburban areas. The TWLTLs should only be considered in places where commercial driveways are the majority of driveways along a road section and the percentage of vehicles turning left at peak hours is at least 20%. The TWLTLs should not be used in areas that are expected to remain rural in the foreseeable future, or on roadways with posted speeds in excess of 45 mph. In urban and suburban applications, the reduction in left-turn and rear-end crash rates may be as much as 35%. The crash reduction in rural applications is not as dramatic, but if properly used at higher crash locations, the TWLTLs may result in significant safety benefits (ITD, 2019).
Researchers Fitzpatrick and Blake (1995) evaluated operational and safety differences between flush medians and TWLTL medians on 4L rural roads. They found no differences in these two median types at
7

lower driveway densities. However, their study sites are very limited (Fitzpatrick and Balke, 1995). Note, TWLTLs are also considered as flush medians. In this reference, the width of flush medians is not specified. Gattis et al. (2005) examined the effects of median treatment and access density on safety outcomes on rural and suburban 4L highways in Arkansas. With an increase in median width (up to 60 to 80 ft), the crash rate decreased while with an increase in access density, the crash rates increased. Their study recommended depressed medians if the access point density is less than 20 and TWLTL if the access point density is greater than 40. Between 20 and 40, a narrow median is recommended (Gattis et al., 2005). Phillips (2004) compared four-lane median divided highways with TWLTLs. They applied several filters to the data to study suburban settings (Phillips, 2004). Figure 2 summarizes the majority of their results.
Figure 2. Collision difference for varying ADT and approach density (Phillips, 2004). 8

The NCHRP report 794 "Median Cross-Section Design for Rural Divided Highways" provided additional criteria (see Figure 3 and Figure 4) based on operation cost savings and accident cost savings (Graham et al., 2014):
Figure 3. Cost-effectiveness of TWLTL based on operation cost savings (Graham et al., 2014).
9

Figure 4. Cost-effectiveness of TWLTL based on accident cost savings (Graham et al., 2014).
After a thorough literature review, it was clear that traffic volume, access point density, truck percentage, turning volume, and spacing of access points are key factors that influence safety on 4L roads besides median type. In summary, there is more information available in the literature for the TWLTLs, NTM, and UR than for the 4FM. Besides, according to our literature search, limited information is available for rural areas compared to urban or suburban areas. We provided Table 1 as a summary of highly relevant studies for rural 4-lane roadways compared to urban or suburban roadways.
10

Table 1. A summary of highly relevant studies for rural as well as urban/suburban 4-lane roadways.

Title
Highway safety manual 2010/2011 Edition

Study location (State or City)
--

Two-way left-turn lane guidelines for urban four-lane roadways (McCoy et al., 1988).

Nebraska

Continuous raised medians Empirical collision model for 4-lane median divided and 5-lane with TWLTL segments Safety issues related to TWLTL (TWLTL FDOT) A policy on geometric design of highways and streets 2011

Iowa --
Florida --

TWLTL - four-lane roadway study_8-9-10 undivided four-lane roadway safety review

Alabama

Continuous two-way left-turn lanes (TWLTLs)

--

Accident comparison of raised median and two-way left-turn lane median treatments (Squires and Parsonson, 1989)

Georgia

Rural / urban / suburban
Suburba n

Median type 4-ft flush TWLT media L n
Yes Yes

Other median types
--

Urban --

Yes

--

Urban --

Yes

--

Guidance criteria

Traffic volume Vehicle

(ADT - vehicles per hour

per day)

(vph)

AADT of 15,000 or more

--

Cost-effective:

>6,200 ~ 6,600

Cost savings: 10,500 ~16,200 --

Accident cost savings alone: >7100 10,000~ 28,000 --

Left turn volume or percentag e

Access points per mile

--

--

Intersection density
--

--

--

--

--

--

--

Urban --

Yes

--

--

--

--

25 ~ 90 --

--

--

Yes

--

--

--

--

--

--

Urban --

Urban & Rural

--

Urban & suburban

--

Yes

--

Yes

--

Yes

--

Urban --

Yes

--

--

--

--

--

--

Peak hour

volume

--

> 1750 & High

--

--

< 2400

vph

In excess of 10,000 ~ 12,000

--

--

--

--

Unsignalized

intersections:

maximum of

--

--

--

>75

five or six

depending on

signals per

mile

11

Evaluation of flush medians and two-way, left-turn lanes on four-lane rural highways Evaluation of flush medians and two-way, left-turn lanes on four-lane rural highways

Lufkin, Texas

Rural

Nacogdoches , Texas

Rural

Yes Yes Yes --

Raised (or depressed -) medians,

--

--

--

--

--

--

Kentucky median type guidance: white paper

Urban &

Kentucky

Suburba Yes Yes

--

n

<24000 for TWLTL

--

<100

Effects of rural highway median treatments and access

Arkansas

Roadway design manual section 3: suburban roadways

--

Benefits of installing medians Georgia

Addition to road design

manual

four-lane vs. five-lane

--

Comparison/Recommendatio

n

Rural & Suburba n

Yes <8ft

Yes

Suburba n

--

Yes

Urban & suburban

--

Yes

Rural --

Yes

Depressed median; Raised median; Barrier; Wide flush median; Narrow flush median
Raised median
--
4-lane undivided
4-lane divided
5-lanes

--
6,000 for an existing fourlane suburban roadway Current traffic volume of < 18,000 & Future traffic volume projected at < 24,000 Curbed: DHV<1,000vph ADT<30,000vp d
10' Shoulders: DHV<1,000vph

--

--

--

--

--

--

Curbed: 45 mph or higher
-10' Shoulders :

12

--

--

--

--

10 ~ 85 for TWLTL;
->10 for flush median

<20 for

depresse

d

median;

20 - 40 for

--

narrow

median;

>40 for

TWLTLs

6 or more

--

--

--

>4

>45

intersections/m

i

Roadway Design Manual

Chapter 7: Cross section

--

element

--

--

Yes

Iowa Department of

Transportation's Design

--

Manual

--

--

Yes

An Evaluation of Flush Medians And Two-Way, Left-Turn Lanes -on Four-Lane Rural Highways

--

Yes Yes

Design Guidelines for

Provision Of Median Access on Principal

--

Urban Yes Yes

Arterials

Road Diet information guide (Knapp et al., 2014)

--

--

--

Yes

Median Cross-Section Design for Rural Divided Highways

--

Rural Yes --

Practical Solution for Highway Design

--

Rural & Urban

--

Yes

Safety Evaluation of

Arizona,

Installing Center Two-Way California,

Left-Turn Lanes on Two-

Illinois,

Rural --

Yes

Lane Roads (Persaud et al. North

2008)

Carolina

Notes: "--" = not available.

with

ADT<20,000vp 55/65

TWLTL d

mph

--

--

--

--

--

17,000-24,000

Raised or

depressed

median; flush

--

median;

TWLTL

Raised median

24,000

--

--

--

--

> 45 mph --

--

<20,000

--

--

--

Base < 18,000 Design = 24,000

--

--

--

< 28,000

--

< 20%

--

--

--

--

10 to 35

on both sides of

--

the street

--

--

--

--

--

--

--

--

--

--

--

--

--

--

13

Chapter 3: DATA & METHODS This chapter details the specifics of all the data that was used for the study and also statistical methods adopted. Table 2 summarizes the data the team has obtained, reviewed and processed.

Table 2. Data procured.

Data 2013-2018 Traffic crash data

Source GDOT

2013-2018 Traffic count data

GDOT

TWLTL layer data (shapefile)

GDOT

Mileposts of 4FMs and TWLTLs GDOT

Highway Performance Monitoring FHWA

System (HPMS) data

website

Use of the data To obtain crash frequencies by type and severity at sites of interest. To provide exposure information in safety analysis and SPF development. To identify roadway segments with TWLTLs. To show locations for 4-ft flush medians and TWLTLs. To identify additional locations for 4-ft flush medians and TWLTLs.

Traffic Crash Data The crash data provide details about crashes that occurred on Georgia roads from 2013 to 2018. The dataset contains over 130 variables related to persons/vehicle occupants, vehicles, and crashes. All crashes are geo-referenced with longitudes and latitudes, except a smaller portion of crashes missing the geocodes. The geo-references are critical for this project, to link the crashes to roadway segments of focus. Therefore, crashes that are not linked to the segments of interest are out of the project scope and are not be included in the safety performance analysis.

Traffic Count Data The team obtained traffic count data from GDOT for all stations from 2013-2018. The data were downloaded from the website: https://gdottrafficdata.drakewell.com/publicmultinodemap.asp. Figure 5 shows an example of the data. Variables that are available in the dataset include: station
14

ID, functional class, latitude and longitude, AADT for each year from 2013 to 2018, Truck percentage for each year from 2013 to 2018. The traffic count data can be linked to the roadway segments with Station ID or the geo-references (i.e., longitude and latitude). Figure 6 shows an example of the 2018 traffic counts on Georgia roadways. The data look reasonable, as high-volume roads appear to be in urban areas, especially in the metropolitan area of Atlanta.

Station Functional

AADT_ Truck% AADT_ Truck% AADT_ Truck% AADT_ Truck% AADT_ Truck% AADT_ Truck%

ID

Class

Lat/Long Lat

Long 2018 _2018 2017 _2017 2016 _2016 2015 _2015 2014 _2014 2013 _2013

001-0183 4R : Rural Mi3n1o.8r3A1r5te60ri,a-l3812..80381654670 -82.0865 970 18.7 990 18.2 940 18.2 880 18.1 850 19.2 870 16.4

001-0185 5R : Rural Ma3j1o.r60C2o8ll5e0c,t-o38r12..63002881560 -82.3082 410 11.7 450 16 480 17.7 490 17.8 480 17.7 520 15

003-0132 3R : Rural Pri3n1c.i3p0a7l1A10rt,e-3r8i12a..l39-00O7115t1h40er-82.9015 4760 22.9 4660 23.5 4600 23.5 4550 23.7 4360 24.5 4360 23.6

003-0138 3R : Rural Pri3n1c.i2p9a5l8A00rt,e-r8i32a1.l8.-239O958t8h00er-82.8398 4520 23.1 4470 23.6 4390 23.7 4310 21.7 4170 24.4 4110 24.7

005-0125 3R : Rural Pri3n1c.i6p0a8l6A10rt,e-3r8i12a..l64-06O8168t1h60er-82.4619 5380 16.4 5300 17 5140 17.3 4840 19 4620 19.7 4480 19.6

009-0156 4R : Rural Mi3n3o.0r8A6r9te90ri,a-l3833..01876291970 -83.1722 8030 7.4 8410 6.7 9190 8.7 9070 9.2 8710 8.6 8610 8.1

011-0103 3R : Rural Pri3n4c.i2p7a9l3A30rt,e-3r8i43a..l24-76O9533t3h00er-83.4653 13900 7.9 13700 7.4 12700 7.4 12000 7.4 11300 7.4 10900 7.1

013-0036 4U : Urban M3i3n.o97r 7A4r9te0,ri-a38l33..96757343900 -83.6533 8190 5.9 8180 6.1 8300 5.8 8110 5.5 7860 5.2 8090 4.2

015-0118 3U : Urban P3ri4n.c2i0p4a3l6A0r,t-e38r44i..a28l01-40O3562th0e-r84.8105

45200

44900 4.9 43600 4.9 42200 4.9 41500 5.7

015-0178 3U : Urban P3ri4n.c2i1p4a8l5A0r,t-e38r44i..a27l19-44O8155th0e-r84.7942 10300 8.5 10600 9.4 10800 9.1 9870 9.1 9480

9 9370 8.3

015-0276 1U : Urban P3ri4n.c2i1p8a7l1A0r,t-e38r44i..a27l15-82I72n14t0er-s8t4a.t7e522 78700 25.6 78600 22.5 75200 21.4 73600 24.5 70700 24.5 66600 17.9

017-0143 3U : Urban P3ri1n.c7i2p8a9l3A0r,t-e38r13i..a72l24-87O9136th0e-r83.2472 3480 4.7 3410 6.7 3470 6.6 3420 6.3 3410 6.4 3510 5.7

021-0116 4R : Rural Mi3n2o.8r0A6r3te20ri,a-l3823..88046039230 -83.8409 4080 7.1 4110 7.1 4310 6.8 4200 6.5 4240 6.7 4360 6.6

021-0132 3U : Urban P3ri2n.c8i1p5a9l7A0r,t-e38r23i..a87l10-52O9778th0e-r83.7028 19900 4.1 20300 4.3 20800 4.1 20900 4.1 20800 3.8 21300

3

021-0158 3U : Urban P3ri2n.c7i3p6a1l3A0r,t-e38r23i..a76l35-65O1636th0e-r83.6557 28100

6 28000 7.5 27600

7 26500

7 25400 7.2 25600

7

Figure 5. Traffic count data example.

Roadway Data GDOT provided two datasets that were used to extract the roadway-related information: 1) Layer data of TWLTL routes in Georgia; and 2) Mileposts of existing 4FM and TWLTLs (reported by GDOT Districts). In addition to GDOT-provided data, the team utilized the publicly available roadway data from the Highway Performance Monitoring System (HPMS) to identify more segments that fit into the project scope. The data can be downloaded from the Federal Highway Administration (FHWA) at https://www.fhwa.dot.gov/policyinformation/hpms/shapefiles.cfm. The HPMS data are in shapefile format and contain a comprehensive list of roadway attributes for

15

all public roads in the country, including the median type and median width information, traffic volume data (AADT and truck volume) and number of lanes.
Figure 6. Traffic count data example. Figure 7 shows the HPMS data visualized in a Geographic Information System (GIS) platform. In addition to GDOT-provided data, the team looked into HPMS data to identify additional segments. Due to some extent of the data inaccuracy and missing information for some segments in the HPMS data, the project team carefully examined the HPMS data. First, according to the HPMS variables, segments that are clearly not within the project scope (rural four-lane segments with speed limit 50 mph or higher) are not included in the analysis. Only segments that are in rural
16

areas, with four-lanes and high-speed limits (i.e., 50 mph or higher) are kept. We used the same characteristics (rural or urban, number of lanes, posted speed limit) to remove segments out of the project scope and keep the ideal segments for this project. Then, the project team used Google Maps to manually check these locations, and only verified segments are merged into the shapefiles for four median types detailed in the following sections.
Figure 7. HPMS data. Undivided roadway segments Given the potentially similar safety performance of rural 4L roads with flush median width less than 3 ft, these roads are considered as UR for this study. Figure 8 exhibits examples of rural fourlane undivided roadways and the statewide identified undivided roadway segments.
17

Figure 8. Examples of rural four-lane undivided roadway and identified locations of rural four-lane undivided roadways segment.
4-ft Flush Medians If the flush median width is roughly between 3 and 5 feet, we considered it a 4-ft flush median (divided roadway). The team manually verified GDOT provided 4L roads with 4FM using Google
18

Maps. A shapefile was created to document the verified roadways with 4FM. In addition, the team looked into the HPMS data to identify any additional segments. Figure 9 exhibits examples of rural four-lane 4-ft flush median roadways and the statewide identified 4-ft flush median roadway segments.
Figure 9. Examples of rural four-lane 4-ft flush median roadway and identified locations of rural four-lane 4-ft flush median roadways segments. 19

TWLTLs The same segment identification method was used to locate the rural four-lane TWLTL segments in Georgia. Figure 10 exhibits examples of rural four-lane TWLTL roadway and the statewide identified TWLTL roadway segments.
Figure 10. Examples of rural four-lane TWLTL roadway and identified locations of rural four-lane TWLTL roadways segments. 20

Non-traversable Using the same segment identification method, the project team was able to locate the rural fourlane non-traversable segments in Georgia. Figure 11 exhibits examples of rural four-lane nontraversable roadway and the statewide identified non-traversable roadway segments.
Figure 11. Examples of rural four-lane non-traversable roadway and identified locations of rural four-lane non-traversable roadways segments. 21

After data cleanup, there were four different shapefiles for the four interested median types on 4L roads including undivided, 4-ft flush median, TWLTL, and non-traversable. The shapefiles consisted of road segments that are as short as 0.001miles and as long as 6.137 miles. Hence, before developing SPFs, it is necessary to refine this data by creating homogenous roadway segments that have similar road characteristics. The research team manually looked up the characteristics of roadways within the shapefile using Google Maps and QGIS tools. Shorter segments were merged with adjacent segments if they had similar features. Similarly, a homogenous longer road segment was divided into shorter segments if major intersection (i.e., signalized four-leg intersections) were identified, particularly for non-traversable segments. Major intersections (i.e., signalized four-leg intersections) are used as slicers to dissect longer road segments into shorter ones.
Subsequent to all the merges and splits, a shapefile was created with homogeneous road segments to help with the rest of the data analysis. According to the HSM 2010, the desirable minimum sample size for the calibration database for one predictive model is 30 to 50 sites. For segments, each site should be between 0.1 and 1.0 mile in length. Lengths in this range should be long enough to have statistical validity and short enough to be realistically homogeneous (AASHTO, 2010). Note that some segments are relatively short because they are isolated in the rural areas and cannot be linked to other segments, though HSM recommends the minimum length is 0.1 miles. Similarly, some rural segments are longer than 1 mile, because they are homogeneous in the sense that they cannot be divided into shorter segments by major intersections or other traffic controls. All segments are high-speed rural 4-lane segments, greater than 45 mph. Table 3 shows descriptive data statistics of the homogeneous road segments created in terms of segment length.
22

Table 3. Frequency of homogeneous segments for safety analysis by segment length.

Length (miles) 0.09 - 0.19 0.20 - 0.49 0.50 - 0.99 1.00 - 1.99 2.00 - 6.14 Average Segment length

UR 9 39 23 7 1 0.529 (miles)

4FM 14 80 51 18 2 0.555 (miles)

TWLTL 109 263 150 33 3 0.444 (miles)

Non-traversable 5 158 431 458 124 1.151 (miles)

Access Points The number of access points is an important parameter for this study. However, there is no existing database that provides such information. Hence, the research team extracted the number of access points for each identified road segment by using Google Maps. Major intersections were used to separate long segments, and only minor intersections and driveways are counted towards the total access points along a segment. Figure 12 shows a screenshot of how access points were counted. The descriptive data statistics of the access point density variable are detailed in Chapter 4.

Data Linkages to Segments Finally, the team linked the crash records (from 2013 to 2018) and traffic volume to sampled segments based on the geo-references (i.e., longitudes and latitudes). The latitude and longitude in the crash data were used to locate crashes; GIS shapefiles were created for the crashes from 2013 to 2018. Not all latitude and longitude information were valid in the crash datasets; therefore, crashes with invalid latitude and longitude information (e.g., null latitude and longitude, (-1, -1), (44.956954, -69.937925)) are removed from the datasets (less than 0.5% of the total sample).

23

Figure 12. An example of counting the access points along a sampled roadway segment.
There are a few crashes that did not fall exactly on the roadway (within the right-of-way), but were close to the sampled segments. For such crashes, the research team verified the road name, road class, and other information that can help locate the crash, to make sure crashes are correctly linked to the sampled segments. A small number of crash records do not contain the geo-reference information. It remains unknown how many of these crashes are related to the sampled segments in this study. The statistics presented in this report might be underrepresented given this common issue, which is not unique to this study. The homogeneous roadway segment shapefile and crash
24

data shapefile were overlaid; crashes were geographically linked to roadway segments, as shown in Figure 13. Figure 14 shows examples of data linkage results - crashes that occurred on sampled road segments.
Figure 13. Examples of overlaid homogeneous TWLTL roadway segment with 2018 crash data (Left); and overlaid homogeneous 4-ft Flush Median roadway segment with 2018 crash data (Right) before the linkage. (Grey dots represent the crashes)
25

Figure 14. Example of overlaid homogeneous TWLTL roadway segment with all crash data (Left); and overlaid homogeneous 4-ft Flush Median roadway segment with all crash
data (Right) after the linkage. (Grey dots represent the crashes) Similarly, the AADT stations were also linked to sampled segments using the same near analysis method to obtain the AADT information for sampled roadway segments. Figure 15 presents the AADT stations that are located in the sampled roadway segments. We do acknowledge that some of the segments do not have an associated AADT. We used AADT along the nearby segments to interpret an AADT value in these cases.
For the final safety analysis, four different shapefiles for four different median types (UR, 4FM, TWLTL, NTM) were created that have traffic volume, truck percentage, access points, segment length, and number of crashes by severity information.
26

Figure 15. Spatial distribution of AADT stations (black dots) that are linked to the sampled roadway segments.

Modeling Background

According to the HSM, the standard or most common form of an SPF for roadway segments is

described as below (AASHTO, 2010):

= e + ln () or = e + ln () + ln ()

(1)

where is the predicted number of crashes on a segment; is the length of the segment;

is annual average daily traffic volume; and b are regression coefficients to be estimated using

historical crash data. In Equation (1), the segment length L is included as a multiplier, which 27

assumes that the crash frequency on a segment is simply proportional to the segment length.

However, this assumption may be inappropriate in some cases. Traveling on a road segment, a

driver experiences homogeneous road conditions (including the number of lanes, the shoulders,

etc.). Driving on a relatively longer road segment with unchanging circumstances may be different

from driving on a relatively shorter road segment with frequent variations of circumstances.

Therefore, another common form of SPFs is also suggested by transportation professionals:

= e + ln () + ln ()

(2)

where c is a parameter indicating the relationship between crash frequency and segment length. If

the estimate of c is close to 1, then Equation (2) is identical to equation (1). If c is significantly

different from 1, then it shows that the road segment length is not simply proportional to crash

frequencies.

Multiple regression models are estimated, providing parameters (a, b, and c) of the SPFs. It is

assumed, in the HSM, that crash frequencies follow a negative binomial (NB) distribution

(AASHTO, 2010). The negative binomial distribution is an extension (capturing over-dispersion)

of the Poisson model:

~ ( , )

(3)

where is the observed crash frequency on a segment; is the expected crash frequency,

and is the NB over-dispersion parameter. A larger value of implies greater over-dispersion in

data. If = 0, then the data follows a Poisson distribution (where mean = variance). In such a

situation, the Poisson and Negative Binomial model provide identical estimates of parameters (a,

b, and c). If is significantly greater than 0, an NB model is preferred. Formally, can be

viewed as a log link function of a set of independent variables (Liu et al., 2017):

28

( ) = + ln () + ln ()

(4)

In addition to the variables of AADT and segment length in the standard SPF, other factors can

also be included in the form. The truck percentage has a complex relationship with the crash

frequency due to its intercorrelated relationship with the traffic volume AADT. Figure 16 shows

the scatter plot of traffic volumes (AADT) and the truck percentages. Higher truck percentages are

more likely to exist in lower traffic volumes. By adding the interaction term in modeling, we can

capture the main effect of truck percentages on crash frequencies and also the intercorrelation

between the AADT and truck percentage.

Figure 16. Scatter plot of vehicle AADT and truck percentage.
An expanded SPF form, including the truck percentage and access point density can be expressed as:
(5
( ) = + ln () + ln () + + + ln ()
) where is the predicted number of crashes on a segment ; is the length of the segment ; is annual average daily traffic volume of the segment ; and b are regression coefficients
29

to be estimated using historical crash data; is the truck percentage and is the access point density along a segment ; parameters and are the coefficients for truck percentage and access point density; is a coefficient for the interaction term between truck percentage and AADT.

In recognizing the improved prediction performance of SPFs, the research team applied an advanced random-intercept modeling technique to estimate SPFs. This technique can account for the unobserved heterogeneity that may exist in the model data. Both traditional NB models and random-intercept NB models are estimated. The models can be compared by calculating the Akaike Information Criterion (AIC), which is often reported in statistical applications.

Methodology to Estimate CMFs

When a site or segment is different from the base condition, or a new treatment is added to a

segment, the Crash Modification Factors (CMFs) are used to adjust the SPF-predicted crash

frequency. A CMF is a measure of the safety effectiveness of a particular treatment or design

element. This project developed CMFs for 4FM, TWLTLs, and NTM by comparing the safety

performance with the baseline rural 4L-UR. The CMFs estimated based on the following

equation:



=

(6)

The target segments are the rural 4L roadways with 4FMs or TWLTLs or NTMs; the base segments

are the rural 4L-URs. As defined in SPF, the crash frequency is a function of traffic volume

(AADT), truck percentage, access point density and segment length. Therefore, CMFs are likely

to vary across different values of AADT, segment length, truck percent and access point density,

indicating that the safety effectiveness of a treatment on roadways is likely to vary across different

30

situations. The CMFs need to be estimated by controlling all other factors constant (i.e., the same values across different segment types). This study predicted the crash frequencies for one mile of segments. For the other three factors (AADT, truck percent and access point density), their values change simultaneously across all segment types in the CMF estimations.
31

Chapter 4: DESCRIPTIVE ANALYSIS Based on the collected and prepared data, the team conducted descriptive statistics analysis on the sampled segments and the results were presented in this Chapter. Furthermore, this chapter presents the dimensions of the data, which help better understand the roadway, traffic, and safety conditions of the sampled segments.
Traffic Crashes From 2013 to 2018, a total of 1,611,889 crashes were reported on Georgia roadways. On sampled rural 4L roadway segments, there are 24,564 crash records on TWLTL segments, 1,562 on 4FM segments and, 4,973 on undivided segments. Table 4 provides the crash frequency distributions by severity, collision type, road type, number of vehicles involved, and year, respectively. Overall, there are more crashes on the non-traversable roadway segments due to the larger sample of nontraversable roadway segments.
In terms of KABCO injury severity (K Fatal Injury, A Suspected Serious injury, B Suspected Minor or Visible Injury, C Possible Injury or Complaint, O No Apparent Injury), the majority of the crashes were no apparent injury (O) crashes (at least 59%). Fatal injury and suspected serious injury crashes (KA) had higher proportions on rural four-lane undivided segments (2.44% and 17.4%, respectively) compared to those of other median type segments.
In terms of the manner of collision for crashes on sampled segments, rural four-lane TWLTL segments had a higher proportion (23.91%) of angle crashes compared to other median type segments. Rural four-lane 4-ft flush median segments had a higher proportion (42.23%) of rear-
32

end crashes; while rural four-lane non-traversable segments had a higher proportion (51.69%) of crashes that were not a collision with other motor vehicles. Note that some manner of collision information was missing for some of the crashes.

In terms of the number of vehicles involved in the crashes on sampled segments, rural four-lane non-traversable segments had a higher proportion of (53.26%) of single-vehicle crashes; while rural four-lane 4-ft flush median and TWLTL segment had higher proportions of multi-vehicle crashes (68.12% and 64.69%, respectively).

The crashes also presented temporal trends along the study period from 2013 to 2018. Results in Table 4 indicate that number of crashes occurred on the sampled segments exhibited an overall increasing trend from 2013 to 2018. Particularly, the number of crashes increased by about 51.7% from 2013 to 2018 for rural four-lane undivided segments compared to that of about 40% for the other three median types of roadway segments. Table 4 also indicates that NTM roadways have the second-highest percentage of fatalities and the highest number of fatalities in the study period. This is most likely the posted speed limit is higher on NTM roadways, which may be a contributing factor to high fatality frequency along NTM roadways (see Appendix A for more details).

Table 4. Crashes statistics for undivided, 4-ft flush median, TWLTL, and non-traversable

roadway segments.

Median Type Crash Features Total Crashes
Fatal Injury

UR (Freq & %)

4FM (Freq & %)

615

1,518

KABCO

15

18

TWLTL (Freq & %)
4,868
65

NTM (Freq & %)
13,694
201

33

(K) Suspected Serious Injury
(A) Suspected Minor or
Visible Injury (B)
Possible Injury or Complaint (C)
No Apparent Injury (O)
Angle
Head-On
Read-End Sideswipe Same
Direction Sideswipe Opposite
Direction Not a Collision with a
Motor Vehicle
Single-Vehicle
Multi-Vehicle
2013
2014
2015
2016
2017
2018

(2.44%) 107
(17.4%)
66 (10.73%)

(1.19%) 210
(13.83%)
188 (12.38%)

49 (7.97%)

117 (7.71%)

367

971

(59.67%)

(63.97%)

Manner of Collison

114

233

(18.54%)

(15.35%)

16

40

(2.60%)

(2.64%)

149

641

(24.23%)

(42.23%)

43

122

(6.99%)

(8.04%)

14

11

(2.28%)

(0.72%)

267

451

(43.41%)

(29.71%)

Single / Multiple Vehicles

281

484

(45.69%)

(31.88%)

334

1,034

(54.31%)

(68.12%)

Year

87

235

(14.15%)

(15.48%)

97

233

(15.77%)

(15.35%)

87

254

(14.15%)

(16.73%)

104

253

(16.91%)

(16.67%)

105

209

(17.07%)

(13.77%)

132

332

(21.46%)

(21.87%)

(1.34%) 505
(10.37%)
506 (10.39%)
461 (9.47%)
3,257 (66.91%)
1,164 (23.91%)
106 (2.18%) 1,285 (26.4%)
496 (10.19%)
68 (1.40%) 1,648 (33.85%)
1,719 (35.31%)
3,419 (64.69%)
738 (15.16%)
731 (15.02%)
765 (15.71%)
771 (15.84%)
840 (17.26%)
1,023 (21.01%)

(1.47%) 1,968 (14.37%)
1,584 (11.57%)
868 (6.34%)
8,942 (65.3%)
1,854 (13.54%)
217 (1.58%) 3,152 (23.02%) 1,117 (8.16%)
89 (0.65%) 7,078 (51.69%)
7,293 (53.26%)
6.401 (46.74%)
2,063 (15.06%)
1,942 (14.18%)
2,053 (14.99%)
2,429 (17.74%)
2,339 (17.08%)
2,868 (20.94%)

34

Notes: Adding the subcategory number of crashes might not be equal to the total number of crashes due to that some of the crashes may have unknown values in those categories (e.g., some category values are coded as 99 or 98).

The per-mile crash frequency is also calculated and aggregated at the segment level for undivided, TWLTL, 4-feet flush median, and non-traversable segments from 2013 to 2018 (see Table 5).

Table 5. Crashes statistics for different median facility types in 2018.

Facility Type Per mile Crashes Frequency
0-9 10-19 20-29 >30 Average Per mile Maximum Per mile

UR
31 (39%) 29 (37%) 8 (10%) 11 (14%)
15.45 75.19

Frequency of Segments

4FM

TWLTL

193 (56%) 35(21%) 11 (7%) 26 (16%)
20.97 450.78

238 (43%) 127 (23%) 77 (14%) 116 (21%)
22.55 472.22

NTM
771 (66%) 241 (20%) 88 (7%)
76 (6) 11.02 144.15

For sampled TWLTL roadways, the per-mile crash frequency can be as high as 472 crashes per mile. In addition to the descriptive data statistics of the crashes, the research team provided additional spatial distribution of those crashes. Per mile crash frequency (number of crashes per mile) are presented in Figures 17, 18, 19, and 20 for the four types of facilities, including rural four-lane undivided, 4-ft flush median, TWLTL, and non-traversable segments. The visualized results show that sampled rural four-lane segments had diversified per mile crash frequency across the state of Georgia. For example, for sampled 4-ft flush median roadway segments, the per-mile crash frequency was higher in northwest rural areas. Figure 20 shows that sampled non-traversable segments in the segments surrounding big cities (i.e., Atlanta) seem to have a greater per-mile crash frequency.

35

Figure 17. Number of crashes per mile along the rural four-lane undivided segments. 36

Figure 18. Number of crashes per mile along the rural four-lane 4-ft flush median segments.
37

Figure 19. Number of crashes per mile along the rural four-lane TWLTL segments. 38

Figure 20. Number of crashes per mile along the rural four-lane non-traversable segments.
Access Points Access point density (access points per mile) is calculated by dividing the segment length from the number of access points, Frequency distribution of access points per segment, access point density per mile, and frequency distribution of access point density are presented in Table 6. The rural four-lane TWLTL segments had the highest average access point density. In addition, this
39

study also visualized the access point density for the sampled rural four-lane segments. No clear spatial patterns were identified for the access point density along those sampled segments.

Table 6. Number of access points and density by median type.

Number of Access Points

0 - 5

6 - 10

11 - 30

31 - 100

Average number of Access Points per segment

Min

APD - Access Point Density

Median

(Number of Points per Mile)

Mean

Max

0 - 1.99

2 - 4.99

5 - 9.99

10 - 19.99

20 - 49.99

50 - 90

UR 41 16 21 1 7.76 0 15.15 16.93 53.96 5 5 13 35 19 2

4FM 111 39 15
0 4.84
0 8.82 10.38 37.04 13 29 52 48 23
0

TWLTL 228 162 159 9 8.74 0 18.68 21.44 87.38 14 25 74 188 233 24

NTM 661 299 207
9 6.50
0 4.61 6.52 42.65 163 471 321 176 45
0

40

(a)

(b)

(c)

(d)

Figure 21. Number of access points per mile along the sampled rural four-lane segments (a) undivided; (b) 4-ft flush median; (c) TWLTL; (d) non-traversable.

41

Traffic Volume Table 7 presents the AADT summary statistics for sampled segments from 2013 to 2018. During this period, the average traffic volumes increased about 10.9% for rural four-lane undivided segments, 16.1% for rural four-lane 4-ft flush median segments, 13.3% for rural four-lane TWLTL segments, and 19.7% for rural four-lane non-traversable segments.
Table 8 further presents the descriptive data statistics for AADT and truck percentage. The results show that the average AADT on the rural four-lane TWLTL segment was the highest (9,775) among all four median facilities. While the average truck percentage on the rural four-lane nontraversable segments was the highest (17.8) among all four median facilities.
Table 9 presents a snapshot of segment frequency and percentage by AADT and truck percentage for four facility types from 2013 to 2018. In general, the majority of the sampled rural four-lane segments had AADT less than 12,000 vehicles per day (i.e., 94%, 85%, 71%, and 79% for undivided, 4-ft flush median, TWLTL, and non-traversable segments, respectively). In other words, more TWLTL, and non-traversable segments had AADT that were 12,000 vehicles per day or greater compared to undivided and 4-ft flush median segments. Similarly, this project study found that the majority of the sampled segments had a truck percentage in the range 10 - 19.9% (i.e., 58%, 58%, 48% and 48% for undivided, 4-ft flush median, TWLTL, and non-traversable segments, respectively).
42

Table 7. AADT and truck percentage for 4L roads with different median types from 2013

to 2018.

UR Year
2018 AADT (Truck %) 2017 AADT (Truck %) 2016 AADT (Truck %) 2015 AADT (Truck %) 2014 AADT (Truck %) 2013 AADT (Truck %)
4FM Year
2018 AADT (Truck %) 2017 AADT (Truck %) 2016 AADT (Truck %) 2015 AADT (Truck %) 2014 AADT (Truck %) 2013 AADT (Truck %)
TWLTL Year
2018 AADT (Truck %) 2017 AADT (Truck %) 2016 AADT (Truck %) 2015 AADT (Truck %) 2014 AADT (Truck %) 2013 AADT (Truck %)
NTM Year
2018 AADT (Truck %) 2017 AADT (Truck %) 2016 AADT (Truck %) 2015 AADT (Truck %) 2014 AADT (Truck %) 2013 AADT (Truck %)

Min
1,610 (5.5) 870 (5.5) 860 (5.6) 830 (5.6) 810 (5.6) 1480 (5.6)
Min
610 (3.8) 460 (3.8) 450 (2) 430 (2) 410 (3.3) 520 (0)
Min
1,600 (2.6) 1,370 (2.8) 1,330 (2.9) 1,280 (2.4) 1,220 (2.4) 1,230 (2.4)
Min
1,340 (2.5) 1,330 (5) 1,130 (2) 1,090 (2) 1,040 (4) 450 (3.4)

Median
6,630 (9.9) 6,500 (9.4) 5,670 (13.6) 5,580 (13.6) 5,290 (13.7) 5,410 (13.2)
Median
7,130 (9.2) 6,650 (9.7) 6,890 (11.9) 6,030 (15.6) 5,740 (15.7) 6,060 (15.6)
Median
8,935 (11.4) 8,760 (11.9) 8,460 (11.8) 8,330 (11.8) 7,800 (12.1) 8.045 (11.8)
Median
6,490 (16.6) 6,270 (16.4) 5,920 (11.1) 5,965 (16.8) 5,570 (16.9) 5,450 (16.7)

Mean
6,981 (10.85) 6,707 (10.7) 6,567 (12.79) 6,421 (13.6) 6,187 (14.2) 6,297 (12.4)
Mean
8,997 (11.2) 8,754 (9.9) 8,714 (12.33) 8,231 (15.8) 7,775 (16.8) 7,748 (13.7)
Mean
10,387 (12.4) 10,235 (12.4) 10,113 (12.5) 9,638 (12.9) 9,106 (13.3) 9,171 (12.6)
Mean
8,652 (17.7) 8,416 (17) 8,186 (18.1) 7,802 (17.8) 7,480 (17.9) 7,227 (11.42)

Max
15,900 (21.4) 15,800 (17.1) 14,500 (24.2) 13,900 (33.7) 13,200 (35.4) 13,500 (20.9)
Max
32,300 (30.3) 32,600 (20) 31,400 (25.8) 31,500 (37.9) 30,500 (39.9) 30,500 (25)
Max
29,100 (34) 28,500 (32.6) 32,500 (37.1) 31,600 (43.9) 28,300 (46.1) 28,300 (39)
Max
36,400 (39.5) 33,800 (38.1) 35,300 (40.3) 34,100 (43.9) 32,100 (46.1) 29,100 (39)

43

Table 8. Average traffic volume (AADT) for different median types.

Median Type
UR 4FM TWLTL NTM

Observations
79 165 558 1,176

Median Type
UR 4FM TWLTL NTM

Observations
79 165 558 1,176

Annual Average Daily Traffic (AADT)

Mean
6,526 8,370 9,775 7,972

Standard Deviation
2,983 5,347 4,665 5,801

Minimum
1,077 480 1,338 1,198

Truck Percentage (%)

Mean
12.9 13.5 12.5 17.8

Standard Deviation
5.2 5.4 6.0 7.7

Minimum
5.6 2.7 2.6 4

Maximum 13,650 31,633 29,550 32,967
Maximum 34.4 25.4 38.8 40.0

Table 9. snapshot of segments frequency and percentage by average traffic volume (AADT) and truck percentage for different median types.

Median Type
UR 4FM TWLTL NTM
Median Type
UR 4FM TWLTL NTM

Segments Frequency and Percentage by Traffic Volume (AADT)

0-5,999
49 (62%) 71 (43%) 122 (22%) 592 (50%)

6,000-11,999
25 (32%) 70 (42%) 273 (49%) 335 (28%)

12,00017,999 5 (6%) 11 (7%) 126 (23%) 170 (14%)

18,00023,999 0 (0%) 9 (5%) 31 (6%) 48 (4%)

24,000+
0 (0%) 4 (2%) 6 (1%) 31 (3%)

Segments Frequency and Percentage by Truck Percentage

0-4.9%
0 (0%) 8 (5%) 29 (5%) 7 (1%)

5-9.9%
25 (32%) 43 (26%) 189 (34%) 197 (17%)

10-19.9%
46 (58%) 96 (58%) 266 (48%) 562 (48%)

20-29.9%
7 (9%) 18 (11%) 71 (13%) 344 (29%)

30%
1 (1%) 0 (0%) 3 (1%) 66 (6%)

44

Chapter 5: CRASH RATES

The crash rate is commonly used to evaluate the current safety ratings of the roadway facilities in

practice. We adopted the measures of exposures to estimate the crash rate by vehicle miles traveled

(VMT) and its equation is formulated as follows:

100,000,000

= 365

(7)

Where:

= Crash rate per 100-million VMT;

= Total number of crashes in the 6-year study period;

= Average Traffic volume using Annual Average Daily Volumes (AADT) in the 6-year

study period;

= Total number of years in the study period; and

= Length (in miles) of the roadway segment.

All Crashes Table 10 shows that, on average, the rural four-lane non-traversable segments had a lower crash rate than segments with the other three median types. Besides that, the rural four-lane 4-ft flush median segments had the second-lowest crash rate among all median type fatalities. The average crash rate of the rural four-lane undivided roadway segment is the highest (110.85) among all four median types. Figures 22 further shows the distributions of the crash rates by 100-million VMT. The majority of sampled segments had a crash rate of 200 crashes per 100-million VMT or lower (i.e., 93%, 86%, 89%, and 97% for the rural four-lane undivided, 4-ft flush median, and nontraversable segments, respectively).

45

Table 10. Crash rates for undivided, 4-ft flush median, TWLTL, and non-traversable

segments by 100-million VMT.

Median Types UR 4FM TWLTL NTM



Observations
79 165 558 1,176

Mean
110.85 92.69 100.46 65.62

Standard Deviation 77.62 130.76 133.53 84.71

Minimum
0 0 0 0

Maximum
332.64 1,330.48 1,333.75 1,900.19

Figure 22. Distribution of the crash rates: (a) under 1000 crashes per 100-million VMT.
Crash Severity We also estimated the crash rates for crashes of different severity levels in the KABCO scale.
K Fatal Injury A Suspected Serious injury B Suspected Minor or Visible Injury C Possible Injury or Complaint O No Apparent Injury
46

Table 11 shows that, speaking of the average fatal injury (K) and suspected serious injury (A) crash rates, rural four-lane undivided segments were associated with a significantly higher crash rate compared to the other three median types. In terms of suspected minor or visible injury (B) crash rate, the rate of rural four-lane 4FMs segments was found to have a slightly higher rate; but the rural four-lane 4FMs segments had a significantly higher rate of possible injury or complaint (C) injury crashes. In addition, the posted speed limits are further evaluated to explore the potential correlations with injury crash rates (more details can be found in Appendix A).

Table 11. Crash rates by KABCO.

Median Types

Observations

K- Fatal Injury

UR

79

4FM

165

TWLTL

558

UTM

1,176

A Suspected Serious injury

UR

79

4FM

165

TWLTL

558

UTM

1,176

B Suspected Minor or Visible Injury

UR

79

4FM

165

TWLTL

558

UTM

1,176

C Possible Injury or Complaint

UR

79

4FM

165

TWLTL

558

UTM

1,176

O No Apparent Injury

UR

79

4FM

165

TWLTL

558

UTM

1,176

Mean
3.58 0.79 1.65 1.03
19.12 15.48 10.91 9.92
10.32 11.28 11.22 7.72
8.42 12.16 9.19 3.75
67.17 52.17 65.52 42.59

Standard Deviation

Minimum

Maximum

12.28

0

2.83

0

6.73

0

3.43

0

91.85 20.19 66.23 39.78

26.16

0

20.77

0

18.91

0

15.80

0

126.84 120.85 128.84 234.59

17.00

0

18.52

0

23.36

0

16.02

0

72.71 97.23 285.94 375.35

15.63

0

78.07

0

18.53

0

9.81

0

82.04 997.86 193.27 211.13

56.90

0

68.71

0

97.78

0

55.83

0

277.20 471.19 1,176.83 1,079.12

47

Crash Type In addition, we estimated the crash rates for a few typical crash types, using the same measure. The following crash types are used in the crash data:
Angle Head-on Rear-end Sideswipe same direction Sideswipe opposite direction Other (including non-collision) The crash rates of six crash types for four median facility types are listed in Table 12. The results show that, on average, rural four-lane TWLTL segments had a higher rate in angle, sideswipe same direction crashes; while rural four-lane undivided segments had a higher rate in head-on, sideswipe opposite direction, and others (including non-collision) crashes; and rural four-lane 4-ft flush median segments had a higher rate of rear-end crashes.
48

Table 12. Crash rates by crash type.

Median Types

Observations

Angle

UR

79

4FM

165

TWLTL

558

UTM

1,176

Head-on

UR

79

4FM

165

TWLTL

558

UTM

1,176

Rear-end

UR

79

4FM

165

TWLTL

558

UTM

1,176

Sideswipe same direction

UR

79

4FM

165

TWLTL

558

UTM

1,176

Sideswipe opposite direction

UR

79

4FM

165

TWLTL

558

UTM

1,176

Others (including non-collision)

UR

79

4FM

165

TWLTL

558

UTM

1,176

Mean
23.73 2.11 25.41 9.36
3.54 2.12 2.28 1.08
22.99 38.15 25.13 12.77
5.28 7.07 9.72 4.99
1.84 0.67 1.35 0.48
49.73 29.63 34.37 36.07

Standard Deviation

Minimum Maximum

35.60

0

2.19

0

51.28

0

17.03

149.77 48.19 601.49 234.59

12.26

0

6.19

0

8.63

0

3.74

0

91.85 48.19 103.77 70.38

32.31

0

113.59 0

51.42

0

37.04

0

158.54 1,330.48 601.49 844.53

9.62

0

12.66

0

20.69

0

11.22

0

30.06 68.79 214.62 211.13

6.05

0

3.54

0

4.79

0

2.63

0

30.36 29.92 41.61 46.63

50.23

0

31.44

0

44.03

0

38.79

0

246.18 181.72 450.96 469.18

Single- and Multi-vehicle Crashes We also estimated the crash rates for single- and multi-vehicle crash rates. As shown in Table 13 across the first three median types of segments (i.e., UR, 4FM, TWLTL), the average crash rates for multi-vehicle crashes were greater than rates for single-vehicle crashes. However, the single-

49

vehicle crash rate on the non-traversable segments was higher than the rate for multi-vehicle crashes.

Table 13. Crash rates for single- and multi-vehicle crashes.

Median Types Single-vehicle crash UR 4FM TWLTL UTM Multi-vehicle crash UR 4FM TWLTL UTM

Observations Mean

79 165 558 1,176
79 165 558 1,176

54.46 32.46 35.69 37.61
56.39 60.23 64.77 28.00

Standard Deviation

Minimum Maximum

53.25

0

33.31

0

43.98

0

40.24

0

246.18 201.92 450.96 539.56

52.77

0

123.00

0

112.04

0

58.73

0

221.96 1,330.48 1,225.29 1,360.63

50

Chapter 6: SAFETY PERFORMANCE FUNCTIONS This section summarizes the SPFs that were developed for the following four segment types on rural 4L roads: UR, TWLTL, 4FM, NTM. The SPFs were developed by the team using the manipulated data. Both traditional Negative Binomial (NB) models and the random-parameter (RP) NB models were estimated. The RP-NB models are able to account for the unobserved heterogeneity due to the limited number of variables in the modeling. The RP models appear to outperform the traditional NB models, according to Akaike's Information Criterion, as shown in Table 14. Therefore, the SPFs estimated from RP-NB models are used for crash frequency prediction, finally to calculate CMFs. In general, these parameters look reasonable. The traffic volumes (AADT) and segment length, as the exposure variables, are positively related to crash frequencies. The factor of access point density is positively related to the crash numbers, because of more access points, more potential conflicts between vehicles on a segment.
51

Table 14. SPFs by crash severity.

SPF Paramete rs

Constant

Ln (AADT)

Seg. length

Truck Pct.

APD

Interaction

N

AIC

All Crashes (KABCO)

UR

-10.689*** 1.286***

0.886*** 0.243

0.009

-0.027

79 202

4FM

-16.338*** 1.887***

0.720*** 0.105

0.015

-0.013

165 456

TWLTL -8.527*** 1.049***

0.883*** 0.064

0.017*** -0.012

558 1,515

NTM

-5.229*** 0.706***

0.946*** 0.015

0.012*** -0.006*** 1,176 3,668

Injury Crashes (KAB)

UR

-12.479*

1.377*

1.288*** 0.255

0.006

-0.029

79 112

4FM

-14.692** 1.557**

0.633*** 0024

0.010

-0.002

165 234

TWLTL -10.820*** 1.113***

0.914*** 0.211

0.021*** -0.026

558 658

NTM

-5.107*** 0.539***

1.004*** -0.04

0.013* 0.001

1,176 1,815

No Injury Crashes (O)

UR

-11.658

1.256

0.934*** 0.150

0.002

-0.014

79 111

4FM

-21.560*** 2.397***

0.869*** 0.099

0.028* -0.012

165 317

TWLTL -8.794*** 1.050***

0.916*** -0.005

0.014*** -0.005

558 1,245

NTM

-6.268*** 0.784***

0.938*** 0.033

0.011** -0.009

1,176 3,027

Possible Injury or No Injury Crashes (CO)

UR

-11.227** 1.318***

0.866*** 0.294

0.010

-0.034

79 170

4FM

-17.622*** 1.994***

0.812*** -0.148

0.024* 0.014

165 343

TWLTL -8.856*** 1.076***

0.922*** -0.031

0.015*** -0.002

558 1332

NTM

-6.227*** 0.791***

0.935*** 0.022

0.009* -0.007

1,176 3074

Fatal Injury Crashes (K)

UR

-7.390

0.400

1.029** -0.136

0.031

0.021

79 33

4FM

-28.176** 2.829**

1.313*** 1.186

-0.042 -0.136

165 38

TWLTL -4.658

0.111

0.792*** -0.318

0.022** 0.034

558 115

NTM

-3.138

-0.039

0.970*** -0.357** 0.003

0.040**

1,176 301

Notes: "*", "**", and "***" indicate significance at 10%, 5%, and 1% level, respectively, N = number of

observations; AIC = Akaike's Information Criterion. "-" indicates not available. SPFs for fatal crashes are

estimated, however, it is not recommended to apply these SPFs due to unreliable model estimates with

insufficient data.

52

Chapter 7: CRASH MODIFICATION FACTORS The CMFs are calculated to show the effectiveness of a cross-section compared to the base crosssection which is a 4L-UR. CMFs are likely to vary across different values of AADT, segment length, truck percent, and access point density, indicating that the safety effectiveness of a treatment on roadways is likely to vary across different situations. This study predicted the crash frequencies for one mile of segments. CMFs are estimated by keeping one of the factors (AADT, truck percent, access point density) constant and varying the other two factors. CMFs were calculated for typical values of AADT, truck percentage, and access point density. Tables 15 to 19 show the results for different access point densities. According to these tables, the overall key findings are as follows:
The segments with NTMs outperform the other three segment types across all AADTs, truck percentages and access point densities, except the very low AADT around 3,000 where the 4FMs appear to be associated with an improved condition. When traffic volumes are low, the NTMs may create a "too perfect" free-flow condition for the traffic, and speeding-related crashes are frequent on low-volume rural 4L roadways.
AADT Is a principal factor to show the effectiveness of a cross-section compared to the base. In general, the 4FMs appear to have an improved performance than UR segments when the AADT is low (around or lower than 9,000); when the AADT increases (around or higher than 9,000), the TWLTLs have an improved performance than UR section and 4FMs; when the AADT reaches 20,000, the NTMs may be considered.
Truck percentage Is positively related to the safety effectiveness of 4FM (when AADT < 12,000), TWLTLs and NTMs compared to the base segment type. In other words,
53

converting a 4L-UR segment into one of the other three cross-section types can result in an even further improved condition for traffic with a higher truck percentage. Access point density Is negatively related to the safety effectiveness of 4FMs, TWLTLs and NTMs compared to the base segment type. In other words, the safety effectiveness of these three types of medians decreases with the increase in the number of access points along a segment. Besides the CMF tables, Appendix B presented more details regarding the trends of CMFs along with the AADT, for different truck percentages and access point densities. The CMF plots are developed for crash severity groups including KABCO, KAB and CO crashes.
54

Table 15. CMFs based on AADT with UR segments as base (Access point density - 10 per

mile).

Truck % AADT
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000

5%
1 1 1 1 1 1 1 1
0.407 0.648 0.850 1.032 1.198 1.354 1.502 1.643
1.051 0.940 0.881 0.841 0.811 0.788 0.769 0.752
1.713 1.232 1.015 0.885 0.796 0.730 0.678 0.637

10%

15%

UR 1 1 1 1 1 1 1 1
4FM 0.360 0.602 0.813 1.007 1.188 1.361 1.526 1.685
TWLTL 0.792 0.747 0.722 0.705 0.692 0.682 0.673 0.665
NTM 1.266 0.979 0.842 0.756 0.696 0.651 0.614 0.585

1 1 1 1 1 1 1 1
0.318 0.559 0.778 0.983 1.179 1.367 1.550 1.727
0.597 0.594 0.592 0.591 0.590 0.590 0.589 0.588
0.936 0.778 0.698 0.646 0.609 0.580 0.557 0.537

20%
1 1 1 1 1 1 1 1
0.282 0.520 0.744 0.960 1.169 1.373 1.574 1.771
0.450 0.472 0.486 0.496 0.503 0.510 0.515 0.520
0.692 0.618 0.579 0.552 0.533 0.517 0.504 0.493

25%
1 1 1 1 1 1 1 1
0.249 0.483 0.712 0.937 1.159 1.380 1.599 1.816
0.339 0.375 0.398 0.416 0.429 0.441 0.451 0.460
0.511 0.491 0.480 0.472 0.466 0.461 0.457 0.453

55

Table 16. CMFs based on AADT with UR segments as base (Access point density - 20 per

mile).

Truck % AADT
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000

5%
1 1 1 1 1 1 1 1
0.433 0.689 0.905 1.098 1.275 1.441 1.599 1.749
1.132 1.012 0.948 0.906 0.874 0.848 0.828 0.810
1.764 1.269 1.046 0.912 0.820 0.752 0.699 0.656

10%

15%

Undivided

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

4FM

0.383

0.339

0.640

0.595

0.865

0.828

1.072

1.046

1.265

1.254

1.448

1.455

1.624

1.649

1.793

1.838

TWLTL

0.853

0.643

0.805

0.640

0.778

0.638

0.759

0.637

0.745

0.636

0.734

0.635

0.724

0.634

0.716

0.634

NTM

1.304

0.964

1.008

0.801

0.867

0.719

0.779

0.666

0.717

0.627

0.670

0.597

0.633

0.573

0.602

0.553

20%
1 1 1 1 1 1 1 1
0.300 0.553 0.792 1.021 1.244 1.461 1.675 1.885
0.484 0.508 0.523 0.534 0.542 0.549 0.555 0.560
0.712 0.637 0.596 0.569 0.549 0.533 0.519 0.508

25%
1 1 1 1 1 1 1 1
0.265 0.514 0.757 0.997 1.234 1.468 1.701 1.933
0.365 0.404 0.429 0.447 0.462 0.475 0.486 0.495
0.526 0.506 0.494 0.486 0.480 0.475 0.471 0.467

56

Table 17. CMFs based on AADT with UR segments as base (Access point density - 30 per

mile).

Truck % AADT
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000

5%
1 1 1 1 1 1 1 1
0.460 0.733 0.963 1.168 1.357 1.534 1.701 1.861
1.219 1.090 1.021 0.975 0.941 0.914 0.891 0.872
1.817 1.307 1.077 0.939 0.845 0.774 0.720 0.675

10%

15%

UR 1 1 1 1 1 1 1 1
4FM 0.407 0.682 0.921 1.140 1.346 1.541 1.728 1.908
TWLTL 0.918 0.866 0.838 0.818 0.802 0.790 0.780 0.771
NTM 1.343 1.038 0.893 0.803 0.739 0.690 0.652 0.620

1 1 1 1 1 1 1 1
0.360 0.633 0.881 1.113 1.335 1.548 1.755 1.956
0.692 0.689 0.687 0.685 0.684 0.684 0.683 0.682
0.993 0.825 0.740 0.686 0.646 0.615 0.591 0.570

20%
1 1 1 1 1 1 1 1
0.319 0.589 0.842 1.087 1.324 1.555 1.782 2.006
0.522 0.547 0.563 0.575 0.584 0.591 0.598 0.603
0.734 0.656 0.614 0.586 0.565 0.549 0.535 0.524

25%
1 1 1 1 1 1 1 1
0.282 0.547 0.806 1.061 1.313 1.562 1.810 2.056
0.393 0.435 0.462 0.482 0.498 0.511 0.523 0.534
0.542 0.521 0.509 0.501 0.494 0.489 0.485 0.481

57

Table 18. CMFs based on AADT with UR segments as base (Access point density - 40 per

mile).

Truck % AADT
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000

5%
1 1 1 1 1 1 1 1
0.490 0.780 1.025 1.243 1.444 1.632 1.810 1.980
1.312 1.174 1.100 1.050 1.013 0.984 0.960 0.939
1.872 1.346 1.110 0.968 0.870 0.798 0.741 0.696

10%

15%

UR 1 1 1 1 1 1 1 1
4FM 0.433 0.725 0.980 1.213 1.432 1.640 1.839 2.030
TWLTL 0.989 0.933 0.902 0.880 0.864 0.851 0.840 0.831
NTM 1.383 1.069 0.920 0.827 0.761 0.711 0.672 0.639

1 1 1 1 1 1 1 1
0.384 0.674 0.937 1.184 1.420 1.647 1.867 2.082
0.745 0.742 0.740 0.738 0.737 0.736 0.735 0.735
1.022 0.850 0.763 0.706 0.665 0.634 0.608 0.587

20%
1 1 1 1 1 1 1 1
0.339 0.626 0.897 1.156 1.408 1.655 1.897 2.134
0.562 0.590 0.606 0.619 0.629 0.637 0.644 0.650
0.756 0.675 0.632 0.603 0.582 0.565 0.551 0.539

25%
1 1 1 1 1 1 1 1
0.300 0.582 0.857 1.129 1.397 1.663 1.926 2.188
0.423 0.469 0.497 0.519 0.536 0.551 0.563 0.574
0.558 0.537 0.524 0.516 0.509 0.504 0.499 0.495

58

Table 19. CMFs based on AADT with UR segments as base (Access point density - 50 per

mile).

Truck % AADT
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000
3000 6000 9000 12000 15000 18000 21000 24000

5%
1 1 1 1 1 1 1 1
0.521 0.830 1.090 1.323 1.537 1.737 1.926 2.107
1.413 1.264 1.184 1.131 1.091 1.059 1.033 1.011
1.928 1.386 1.143 0.997 0.896 0.822 0.764 0.716

10%

15%

UR 1 1 1 1 1 1 1 1
4FM 0.461 0.772 1.043 1.291 1.524 1.745 1.956 2.160
TWLTL 1.065 1.005 0.971 0.948 0.930 0.916 0.904 0.894
NTM 1.425 1.102 0.948 0.851 0.784 0.732 0.692 0.658

1 1 1 1 1 1 1 1
0.408 0.717 0.997 1.260 1.511 1.753 1.987 2.215
0.802 0.799 0.796 0.795 0.794 0.793 0.792 0.791
1.053 0.875 0.786 0.728 0.685 0.653 0.627 0.605

20%
1 1 1 1 1 1 1 1
0.361 0.667 0.954 1.230 1.499 1.761 2.018 2.271
0.605 0.635 0.653 0.666 0.677 0.686 0.693 0.699
0.778 0.696 0.651 0.622 0.599 0.582 0.568 0.555

25%
1 1 1 1 1 1 1 1
0.319 0.619 0.912 1.201 1.486 1.769 2.050 2.329
0.456 0.505 0.536 0.559 0.577 0.593 0.607 0.619
0.575 0.553 0.540 0.531 0.524 0.519 0.514 0.510

59

Chapter 8: CRITERIA DEVELOPMENT When selecting a median type for GRIP roadways, recognition of average annual daily traffic, truck needs, and property access demands are to be considered. On certain projects, more than one cross-section may be necessary or desirable, and consideration of these needs should be made along the corridor to confirm the suitable cross-section choice. The length of the project will influence the selection of the cross-section. Short highway sections should limit the number of cross-section changes; whereas, longer highway sections may consist of many cross-section changes. Changes in cross-sections should be limited to logical transition points, such as natural boundaries or political boundaries.
The criteria are intended to be given for ranges of road or traffic conditions, i.e., AADT, truck percentage, and access point density, as shown in Table 20. However, within each cell, there can be thousands of combinations of these three factors. Tables 15 and 19 only showed a few CMFs, for example conditions. 1,000 possible conditions were randomly generated within each cell in Table 20 and estimated the CMFs to show the effectiveness of three cross-sections compared to the base cross-section UR under these 1,000 random conditions. Table 20 shows the average CMFs within each cell, showing the average effectiveness of improving safety performance by converting an undivided road to a target cross-section. The CMFs greater than 1.00 are not shown in the Table. For AADTs under 5,000 (regardless of the access point density and truck percentage), the average CMFs range from 0.2 to 0.42. It means that converting an UR to a cross-section with 4FMs could reduce the crash numbers by 58% to 80%. For AADTs between 5000 to 10,000, the crash numbers could be reduced by 11% to 43%. From Table 20, in general, the 4FMs seem to be effective in improving safety performance when AADTs are under 10,000, and they are more
60

effective for traffic with higher truck percentages. When AADTs are above 10,000, TWLTLs and NTMs have comparable CMFs.

Table 20. Average CMFs for given road/traffic conditions (KABCO crashes).

Vehicle AADT

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Truck Percentage

Access Point Density, AP/mile

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

<=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

0.36 0.39 0.40 0.42 0.75 0.79 0.84 0.89 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% 0.31 0.33 0.35 0.37 0.70 0.74 0.78 0.84 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

4-ft Flush Median >10% to <=15% 0.26 0.30 0.31 0.33 0.65 0.69 0.74 0.79 0.98 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% 0.24 0.25 0.27 0.29 0.62 0.65 0.69 0.74 0.95 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

0.20 0.22 0.24 0.25 0.57 0.61 0.65 0.69 0.92 0.98 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5% >5% to <=10% >10% to <=15% >15% to <=20%
>20%

N/A N/A N/A N/A 0.98 N/A N/A N/A 0.88 0.95 N/A N/A 0.83 0.90 0.97 N/A 0.78 0.85 0.92 1.00 0.76 0.82 0.88 0.96 0.94 N/A N/A N/A 0.78 0.85 0.92 0.99 0.73 0.79 0.86 0.93 0.70 0.76 0.82 0.89 0.68 0.74 0.80 0.86 0.66 0.72 0.78 0.84 0.67 0.72 0.78 0.85 0.62 0.67 0.73 0.79 0.61 0.66 0.71 0.77 0.60 0.65 0.70 0.76 0.59 0.64 0.69 0.75 0.58 0.63 0.68 0.74 0.48 0.52 0.56 0.61 0.50 0.54 0.58 0.63 0.50 0.54 0.59 0.64 0.51 0.55 0.59 0.64 0.51 0.55 0.60 0.65 0.51 0.56 0.60 0.65 0.35 0.37 0.41 0.44 0.40 0.43 0.46 0.50 0.42 0.45 0.49 0.53 0.43 0.47 0.51 0.55 0.44 0.48 0.52 0.56 0.45 0.49 0.53 0.57

<=5%

N/A N/A N/A N/A N/A N/A N/A N/A 0.94 0.96 0.99 N/A 0.78 0.81 0.83 0.86 0.68 0.71 0.73 0.75 0.62 0.63 0.65 0.67

>5% to <=10% N/A N/A N/A N/A N/A N/A N/A N/A 0.81 0.83 0.85 0.88 0.70 0.72 0.74 0.76 0.63 0.65 0.66 0.69 0.58 0.59 0.61 0.63 Non-traversable >10% to <=15% N/A N/A N/A N/A 0.82 0.85 0.87 0.89 0.69 0.72 0.74 0.76 0.62 0.64 0.66 0.68 0.57 0.59 0.61 0.63 0.54 0.55 0.57 0.59

>15% to <=20% 0.92 0.95 0.98 N/A 0.67 0.69 0.71 0.73 0.59 0.62 0.63 0.65 0.56 0.57 0.59 0.61 0.53 0.54 0.56 0.57 0.50 0.52 0.53 0.55

>20%

0.64 0.66 0.68 0.70 0.54 0.56 0.58 0.59 0.51 0.53 0.54 0.56 0.49 0.51 0.53 0.54 0.48 0.50 0.51 0.53 0.47 0.49 0.50 0.52

Notes: The CMF calculation base cross-section is undivided roadways. Cells with "N/A" are estimated with

average CMFs greater than 1.00. The results are applicable to roadways with posted speed limits of 50 mph

and higher. These CMFs shown in this table are for all crashes (KABCO). Appendix C provides estimated

CMFs for KAB, and CO crashes.

The CMFs are calculated by comparing predicted crash numbers between two types of crosssections. The base crash numbers matter for the magnitudes of CMFs. The CMFs can be used to indicate the percent reduction in crash numbers. CMFs are useful when applying different countermeasures under the same road or traffic conditions. When it comes to recommending countermeasures for different conditions, it may be more reasonable to compare the concrete crash reduction numbers. Table 21 shows the average reduced annual crash frequencies if converting a UR to a cross-section with 4FMs, TWLTLs, and NTMs, respectively.

61

Table 21. Average crash reduction per year for given road/traffic conditions.

Vehicle AADT

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Truck Percentage

Access Point Density, AP/mile

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

<=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

-0.35 -0.36 -0.38 -0.39 -0.54 -0.48 -0.38 -0.25 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% -0.44 -0.45 -0.48 -0.5 -0.66 -0.6 -0.52 -0.43 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

4-ft Flush Median >10% to <=15% -0.54 -0.56 -0.61 -0.64 -0.77 -0.73 -0.69 -0.59 -0.05 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% -0.66 -0.71 -0.75 -0.79 -0.87 -0.86 -0.83 -0.77 -0.16 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

-0.82 -0.84 -0.93 -0.98 -0.98 -0.98 -0.96 -0.91 -0.25 -0.05 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5% >5% to <=10% >10% to <=15% >15% to <=20%
>20%

N/A N/A N/A N/A -0.06 N/A N/A N/A -0.52 -0.22 N/A N/A -1.13 -0.74 -0.25 N/A -1.89 -1.41 -0.82 -0.02 -2.77 -2.27 -1.54 -0.59 -0.10 -0.04 N/A N/A -0.52 -0.41 -0.25 -0.04 -1.1 -0.92 -0.69 -0.37 -1.74 -1.53 -1.23 -0.82 -2.46 -2.21 -1.87 -1.36 -3.20 -2.94 -2.55 -1.94 -0.29 -0.26 -0.23 -0.19 -0.89 -0.84 -0.75 -0.65 -1.5 -1.43 -1.31 -1.14 -2.12 -2.03 -1.89 -1.67 -2.74 -2.65 -2.47 -2.21 -3.36 -3.25 -3.04 -2.73 -0.49 -0.50 -0.49 -0.47 -1.20 -1.20 -1.19 -1.14 -1.79 -1.8 -1.77 -1.7 -2.33 -2.34 -2.3 -2.21 -2.85 -2.85 -2.79 -2.68 -3.33 -3.33 -3.27 -3.11 -0.71 -0.70 -0.76 -0.77 -1.44 -1.50 -1.19 -1.57 -1.98 -2.04 -2.08 -2.09 -2.43 -2.49 -2.52 -2.53 -2.83 -2.88 -2.92 -2.91 -3.19 -3.25 -3.27 -3.24

<=5%

N/A N/A N/A N/A N/A N/A N/A N/A -0.32 -0.20 -0.07 N/A -1.42 -1.39 -1.35 -1.25 -2.78 -2.83 -2.90 -2.90 -4.32 -4.54 -4.69 -4.80

>5% to <=10% N/A N/A N/A N/A -0.02 N/A N/A N/A -0.82 -0.76 -0.72 -0.65 -1.77 -1.80 -1.83 -1.83 -2.88 -2.98 -3.10 -3.13 -4.03 -4.26 -4.45 -4.61 Non-traversable >10% to <=15% N/A N/A N/A N/A -0.44 -0.43 -0.38 -0.35 -1.18 -1.21 -1.21 -1.21 -2.00 -2.07 -2.15 -2.22 -2.85 -2.99 -3.13 -3.27 -3.73 -3.94 -4.14 -4.36

>15% to <=20% -0.17 -0.17 -0.15 -0.12 -0.81 -0.82 -0.85 -0.85 -1.45 -1.53 -1.59 -1.65 -2.11 -2.23 -2.33 -2.44 -2.76 -2.92 -3.07 -3.25 -3.40 -3.61 -3.82 -4.02

>20%

-0.44 -0.43 -0.47 -0.48 -1.10 -1.16 -1.23 -1.29 -1.65 -1.76 -1.85 -1.95 -2.15 -2.29 -2.42 -2.57 -2.63 -2.79 -2.97 -3.15 -3.07 -3.27 -3.48 -3.70

Notes: The base cross-section is undivided roadways. Cells with "N/A" are estimated with no crash

reduction from the base cross-section. The results are applicable to roadways with posted speed limits of

50 mph and higher. The results are for all crashes (KABCO).

Further, project decisions are often made based on the benefits and costs of alternative designs or countermeasures. In this project, the crash reductions could be translated as the safety benefits of a highway project or a countermeasure on a roadway. Crashes have different severities, which associate with uneven safety benefits. Table 22 shows the benefits of reductions for crashes of different severities (GDOT, 2019a).

Table 22. Safety benefits for different crash severities (GDOT, 2019a).

Crash Severity

$ Benefit

K

$

10,450,271.99

A

$

2,285,054.32

B

$

500,966.66

C

$

109,889.46

O

$

23,701.65

62

To consider the crash severity in the safety benefit estimation, we estimated the crash reductions for different severities using the crash severity distributions and the total crash reductions. As shown in Table 23, URs are associated with the greatest percentages of fatal and suspected severe crashes (KA) among all crashes on undivided roadways. Within the other three types of crosssections, segments with non-traversable medians are linked with greater shares of fatal and severe crashes.

Table 23. Crash severity distributions on four cross-section types.

Cross-section Type Undivided
4-ft Flush Median TWLTL
Non-traversable

K 2.48% 1.20% 1.36% 1.48%

A 17.72% 13.96% 10.53% 14.51%

B 10.93% 12.50% 10.55% 11.68%

C 8.11% 7.78% 9.62% 6.40%

O 60.76% 64.56% 67.94% 65.93%

Though the speed limits across four types of cross-sections are 50 mph or higher (i.e., 55 mph, 65 mph) in this project, the travel speeds of vehicles on roadways with NTMs may be higher than those on the other three types of roadways. Higher speeds are associated with severer crashes. The information on vehicle speeds prior to crashes is currently not available to the project team, but it would be worthwhile to investigate the relationships between the vehicle travel speeds and the median types under the same speed limits.

With the predicted crash reductions for different severities and the severity-based safety benefits per crash, the authors estimated the safety benefits for 4FMs, TWLTLs, and NTMs compared with URs as the base. The safety benefits are estimated for the 20 years and all safety benefits are converted to Net Present Values (NPVs).

63



=

20 (1 + ) (1 + )

= 1

(1 + )

(8)

where the Base Benefits are the estimated safety benefits from the crash reductions in the current

year; is the number of years. The inflation rate used in this project is 2% (U.S. BLS SIO, 2019),

the income discount rate is 2.5% (DON, 2018), and the discount rate is 4% (GDOT, 2019b).

Table 24 shows the estimated safety benefits in net present values for 4FMs, TWLTLs, and NTMs compared with URs as the base. The numbers shown in Table 24 are the reduced crash costs in 20 years. Note that, positive values, i.e., negative safety benefits, are not shown in the table. Besides, the safety benefits estimated in this project do not consider the changes in traffic volumes and road environments (i.e., access points) in the future.

Table 24. Safety Benefits (in $million) for 4FMs, TWLTLs, and NTMs compared with URs.

4-ft Flush Median

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 to >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 30

<=5%

$7 $7 $8 $6 $16 $17 $17 $16 $17 $15 $13 $18 $8 $2 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% $8 $9 $9 $8 $18 $18 $18 $17 $17 $16 $14 $18 $8 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>10% to <=15% $10 $10 $11 $9 $19 $20 $20 $18 $18 $16 $15 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% $12 $12 $13 $11 $20 $21 $22 $20 $18 $17 $15 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

$14 $15 $16 $13 $22 $23 $24 $21 $18 $17 $16 $18 $7 $3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5%

$3 $2 $2 $3 $14 $14 $13 $15 $30 $30 $29 $30 $49 $49 $48 $49 $71 $71 $70 $69 $93 $95

>5% to <=10% $5 $5 $5 $5 $19 $19 $19 $18 $35 $35 $35 $34 $51 $53 $53 $50 $69 $71 $72 $67 $87 $90

>10% to <=15% $8 $8 $8 $7 $23 $23 $24 $22 $37 $39 $40 $36 $52 $54 $56 $49 $66 $69 $71 $63 $80 $84

>15% to <=20% $10 $11 $11 $10 $26 $27 $28 $24 $39 $41 $43 $37 $51 $53 $56 $48 $62 $65 $68 $59 $73 $77

>20%

$13 $14 $15 $12 $28 $30 $32 $26 $39 $42 $44 $37 $49 $52 $55 $46 $57 $61 $64 $54 $65 $69

$95 $91 $92 $84 $87 $77 $80 $69 $73 $61

<=5% >5% to <=10% Nontraversable >10% to <=15% >15% to <=20%
>20%

N/A N/A N/A N/A $3 $2 $1 $3 $19 $20 $20 $19 $41 $43 $44 $39 $66 $70 $73 $62 $93 $99 $106 $87 N/A N/A N/A N/A $9 $9 $9 $9 $25 $26 $26 $23 $43 $45 $48 $40 $62 $66 $71 $59 $83 $89 $95 $78 $2 $2 $2 $2 $14 $15 $15 $13 $28 $30 $31 $27 $43 $46 $49 $40 $59 $63 $67 $55 $74 $80 $85 $69 $6 $6 $6 $5 $18 $19 $20 $17 $31 $33 $35 $29 $43 $46 $49 $40 $54 $58 $62 $51 $66 $71 $76 $61 $9 $10 $10 $9 $22 $24 $25 $20 $32 $35 $37 $30 $42 $45 $48 $39 $50 $54 $58 $47 $58 $62 $67 $54

Notes: The base cross-section type is undivided roadways. The safety benefits cover reductions in all crash

types. The safety benefits of 20 years are estimated. Cells with "N/A" are estimated with negative safety

benefits. The results are applicable to roadways with posted speed limits of 50 mph and higher.

64

In addition to the benefits that a proposed project or countermeasure can bring, the cost of a project is also a major factor that decision-makers consider. The cost often includes construction cost and all ancillary costs such as right-of-way, utilities, etc. A benefit-cost ratio (BCR) is often used as an indicator to show the relationship between the relative costs and benefits of a proposed project, expressed in monetary or qualitative terms. If a project has a BCR greater than 1.0, the project is expected to deliver a positive net present value to the investment. An alternative project or countermeasure with a greater BCR is often preferred and selected for construction or implementation.

With the assistance of GDOT, the research team obtained the cost data for several recent projects related to the median types of interest in this research project. The data include the project costs for expanding two-lane roadways to 4L roadways of different cross-sections (e.g., 4FMs, TWLTLs, and NTMs) in rural areas. In addition, from other states, we obtained the project cost data for expanding two-lane roadways to 4L undivided roadways (Capitol Fax.com, 2010; Hillsborough MPO, 2014; FDOT, 2019a; FDOT, 2019b). All these cost data allowed to roughly compare the cost differences between UR and 4FMs, TWLTLs, and NTMs.

Table 25. The per-mile cost differences for roadways with 4FMs, TWLTLs, and NTMs are

compared with URs.

Cross-section Type 4-ft Flush Median TWLTL Non-traversable

Additional Cost (million $) from undivided roadways, per mile

Lower Bound $0.48M $2.73M $1.21M

Upper Bound $2.65M $3.98M $6.85M

65

Using the upper bound cost differences (to be conservative) and assuming there is a proposed project to convert four-lane undivided roadways to 4FM, TWLTLs, or NTMs, the authors calculated the BCRs for three types of cross-sections, as shown in Table 26. The results show that, though the roadways with non-traversable medians are seen a great number of crash reductions (Table 21), the BCRs for such cross-sections are not equally promising due to the cost of constructing them (especially compared with the TWLTLs).

Table 26. BCRs for 4FMs, TWLTLs, and NTMs (the base cross-section is UR).

4-ft Flush Median

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

2.5 2.7 2.9 2.4 6.2 6.2 6.2 6.0 6.4 5.7 4.8 6.9 3.0 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% 3.1 3.3 3.4 2.8 6.7 6.8 6.9 6.5 6.6 6.0 5.2 6.9 2.9 1.0 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>10% to <=15% 3.7 3.9 4.2 3.4 7.2 7.4 7.6 6.9 6.7 6.2 5.5 6.9 2.8 1.1 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% 4.4 4.7 5.1 4.1 7.7 8.0 8.2 7.4 6.7 6.3 5.8 6.9 2.6 1.2 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

5.3 5.6 6.1 4.9 8.2 8.5 8.9 7.8 6.7 6.5 6.0 6.9 2.5 1.3 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5%

N/A N/A N/A N/A 3.6 3.4 3.2 3.7 7.6 7.5 7.2 7.6 12.4 12.4 12.1 12.2 17.7 17.8 17.7 17.3 23.5 23.8 23.8 22.9

>5% to <=10% 1.3 1.3 1.2 1.3 4.7 4.8 4.7 4.6 8.7 8.8 8.9 8.4 12.9 13.2 13.4 12.5 17.4 17.8 18.1 16.8 22.0 22.6 23.1 21.2

>10% to <=15% 2.0 2.0 2.0 1.8 5.7 5.9 6.0 5.4 9.3 9.7 10.0 8.9 13.0 13.5 13.9 12.4 16.6 17.3 17.9 15.9 20.2 21.1 21.8 19.3

>15% to <=20% 2.6 2.7 2.9 2.5 6.4 6.8 7.1 6.1 9.7 10.2 10.7 9.2 12.7 13.4 14.0 12.1 15.6 16.4 17.1 14.8 18.3 19.2 20.1 17.3

>20%

3.3 3.5 3.8 3.1 7.1 7.6 8.0 6.6 9.9 10.5 11.1 9.3 12.3 13.0 13.7 11.5 14.4 15.3 16.1 13.6 16.4 17.4 18.3 15.4

<=5%

N/A N/A N/A N/A N/A N/A N/A N/A 2.8 2.9 2.9 2.8 6.0 6.3 6.5 5.7 9.6 10.1 10.7 9.0 13.6 14.5 15.4 12.8

>5% to <=10% N/A N/A N/A N/A 1.3 1.3 1.2 1.3 3.6 3.7 3.9 3.4 6.2 6.6 6.9 5.9 9.1 9.7 10.3 8.6 12.2 13.0 13.9 11.4 Nontraversable >10% to <=15% N/A N/A N/A N/A 2.0 2.1 2.2 2.0 4.1 4.3 4.6 3.9 6.3 6.7 7.1 5.9 8.5 9.1 9.7 8.0 10.8 11.6 12.4 10.1
>15% to <=20% N/A N/A N/A N/A 2.7 2.8 3.0 2.5 4.5 4.8 5.1 4.2 6.2 6.7 7.1 5.8 7.9 8.5 9.1 7.4 9.6 10.3 11.1 8.9

>20%

1.3 1.4 1.5 1.3 3.2 3.4 3.7 3.0 4.7 5.0 5.4 4.4 6.1 6.5 7.0 5.6 7.3 7.8 8.4 6.8 8.5 9.1 9.8 7.9

Notes: The base cross-section type is undivided roadways. Cells with "N/A" are estimated with BCRs under

1.0 (i.e., benefits < costs) which are not shown in the table. The results are applicable to roadways with

posted speed limits of 50 mph and higher.

In highway project practices, some states use BCR 1 as the decision-making point, and some other states user higher BCRs ( 5) for safety projects. To be conservative, the team chose the BCRs 5 for recommendations. Based on Table 26, we developed preliminary criteria to recommend which cross-section (median type) is suitable for a traffic and road condition. The
66

research team designed the following rules to recommend three cross-sections (over the undivided roadways) in each cell (representing a group of road/traffic conditions).
Undivided: Recommended If a BCR for all three alternative cross-sections is under 3, the base cross-section (undivided roadway) is recommended. Microsimulation suggested If a BCR for any one of alternative cross-sections is between 3 and 5. Not recommended If a BCR for any one of alternative cross-sections is above 5.
4-ft Flush Median: Recommended If a BCR for this cross-section type is greater than 5 and BCRs for the other two cross-sections are under 5. Microsimulation suggested If a BCR for this cross-section type is between 3 and 5 (simulations for UR and segments with 4FMs), or if a BCR for any other cross-sections is also greater than 5 (simulations for segments with 4FMs and TWLTLs and/or NTMs). Not recommended If a BCR for this cross-section type is under 5.
Two-Way Left-Turn Lane: Recommended If a BCR for this cross-section type is greater than 5 and BCRs for the other two cross-sections are under 5.
67

Microsimulation suggested If a BCR for this cross-section type is greater than 5 and BCRs for the other two cross-sections are also greater than 5 (simulations for segments with TWLTLs and 4FMs and/or NTMs).
Not recommended If a BCR for this cross-section type is under 5. Non-traversable Median:
Recommended If a BCR for this cross-section type is greater than 5 and BCRs for the other two cross-sections are under 5, or if the average daily traffic volume is greater than 20,000 vpd (considering the mobility needs of rural 4L roadways with NTMs).
Microsimulation suggested If the average daily traffic volumes are under 20,000 vpd, a BCR for this cross-section type is greater than 5 and BCRs for the other two crosssections are also greater than 5 (simulations for segments with NTMs, 4FMs and/or TWLTLs).
Not recommended If a BCR for this cross-section type is under 5.
The following chart (Table 27) provides preliminary criteria for selecting an appropriate crosssection for rural 4L roadways. Note that, to ensure the consistency of recommendation, if a recommendation in one cell changes from one cross-section type to another, the recommendation would not reverse back to the previous cross-section recommendation. The chart provides recommendations for four typical cross-section types on rural four-lane roadways, including:
A. Four-Lane Undivided B. Four-Lane Divided w/4-ft Flush Median C. Four-Lane Divided w/TWLTL D. Four-Lane Divided w/Non-Traversable Median or Barrier
68

When multiple letters (e.g., A/B, or B/C) are given in a cell, simulations are suggested to further examine and compare the performance of alternative cross-section designs. The performance may be examined from safety, mobility, and environmental aspects.

Table 27. Preliminary criteria for recommending cross-section on four-lane rural

roadways.

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

A A A A B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>5% to <=10% A/B A/B A/B A/B B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>10% to <=15% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>15% to <=20% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

>20%

B B B B B/C B/C B/C B/C B/C B/C/D B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

Notes: A: Four-Lane Undivided

B: Four-Lane Divided w/4-ft Flush Median

C: Four-Lane Divided w/TWLTL

D: Four-Lane Divided w/Non-Traversable Median or Barrier

It is important to note that, the chart in Table 27 provides approximate cross-section recommendations merely based on the estimated crash reductions (by comparing the safety performance of a target cross-section with the base cross-section four-lane undivided roadways), empirical crash costs, and project costs from the previous projects. The crash reduction estimation is based on the achieved crash data from 2013 to 2018 and sampled roadway segments. The results may change significantly when different years of data or roadway segments are selected in the analysis. The crash costs are also affected by numerous factors including economic inflation, which varies from time to time. The project cost is highly project-specific, and it can vary
69

substantially from location to location in a state due to various factors including the prior conditions of a project site, local labor cost, and material cost, etc. It is highly recommended that decision-makers or practitioners use the safety benefits from Table 24 to compare with the cost of a specific proposed project and make a more realistic recommendation of the cross-section type. Also, Table 24 includes only the safety benefits, other benefits such as mobility and environmental benefits also need to be considered in the decisionmaking process. Application of access management principles such as consolidation of driveways to a single point where auxiliary lanes may be provided to reduce access point density. Roadway intersections and higher volume driveways should be provided with auxiliary lanes per adopted warranting criteria, see Section 7.4 of the GDOT Design Policy Manual (GDOT, 2019c). Where provided, depressed median widths should follow Section 6.12 of the GDOT Design Policy Manual (GDOT, 2019c).
70

Chapter 9: GUIDELINES for MICROSIMULATION Microsimulation helps to evaluate the impact of different access management practices on the operation and safety of the network. FHWA has created "Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software" to assist practitioners in assessing traffic performance (Wunderlich et al., 2019). The team extracted high-level information from this document and presented it in a simple and concise manner. For additional details, the above-mentioned document should be referenced.
The overall process for developing and applying a microsimulation model to a specific transportation analysis problem consists of the following seven major steps, as shown in Figure 23: Step 1 Microsimulation Analysis Planning
In planning phase, the research questions that need to be answered through simulation should be phrased. In this context, an example question will be: o Which cross-section provides improved traffic performance for the given traffic and road conditions?
An effective plan should include study "scope/objectives, hypotheses, performance measures, scope, technical approach, and an estimate of resources required for the study".
Performance measures such as delay, queue length, travel time, average speed, etc. could be used. It would be better to consider measures that could be easily captured in the microsimulation models.
Once the study objectives and performance measures have been identified, the next step is to identify the scope - both in terms of geographic and temporal limits.
71

Figure 23. Seven Key Steps in Microsimulation Analysis (Wunderlich et al., 2019).
Microsimulation models are data-intensive. They take more effort than macroscopic simulation models (e.g., transportation planning models). The model results are sensitive to different vehicle performance characteristics and differing driver behavior characteristics. 72

The resource requirements for the development, calibration, and application of microsimulation models will vary according to the complexity of the project, its geographic scope, temporal scope, number of alternatives, and the availability and quality of the data. Adequate time should also be allotted to conduct a successful analysis. Data collection, coding, error checking, and calibration are critical tasks for completing a calibrated model.
Step 2 Data Collection & Analysis This step is to identify data sources and gather data needed to develop a microsimulation model for a specific project analysis. In this project, typical data elements include: o Road geometry (number of lanes, lane widths, segment length, shoulder type, shoulder width, slope, curvature, etc.) o Access management (median type, median width, access points, turn lanes, pavement markings, etc.) o Traffic volume or travel demand (AADT, truck percentages, turning volumes) o Traffic controls (speed limit sign, stop sign, yield sign, signals, pavement marking, etc.) o Vehicle characteristics (vehicle mix, vehicle dimensions, and vehicle performance characteristics e.g., maximum acceleration). o Driver characteristics (driver aggressiveness, reaction time, desired speeds, and acceptable critical gaps for lane changing, merging and crossing). If the alternatives are to be evaluated in target future year (especially for major construction projects, e.g., road widening), forecasted data needs to be procured, such as the travel demand. Local metropolitan planning organization (MPO) travel demand models can provide travel demand data for future years.
73

Additional data may be needed for model calibration. Calibration is the process of systematically adjusting model parameters so that the model emulates observed traffic conditions in the study area.
It is important to note that, data quality needs to be checked before using it. The data error checking can be done by field inspection and surveys.
Step 3 Base Model Development When resources are constrained, it may not be cost-effective to model every condition. A base model is developed to represent the most typical conditions. The base model is used to compare the travel conditions where alternatives (e.g., two cross-sections) are likely to have significant impacts. While developing the base model, Default values can be assumed for driver behavior, gap acceptance, etc. as it is hard to obtain this information. Any assumptions, default values, or deviations from default values should be discussed and documented. The link-node diagram is the blueprint for constructing the microsimulation model. The diagram identifies which streets and highways will be included in the model and how they will be represented. A route needs to be created (as it is in the field) or imported into the microsimulation software. Road geometry (number of lanes, lane widths, etc.), traffic volume, and signage information are assigned to the designed route.
Step 4 Error Checking The error correction step is essential in developing a working model so that the calibration process does not result in parameters that are distorted to compensate for overlooked coding errors.
74

Error checking involves various reviews of the coded network, coded demand, and default parameters.
The analyst should review the software and user group web sites to ensure that he or she is aware of the latest known "bugs" and user workarounds for the software. The analyst should ensure that he or she is using the latest version and "patch" of the software, if any.
There might be errors in the input data, such as the network connectivity, link attributes (e.g., free-flow speed, facility type, lane numbers, etc.), traffic controls, vehicle mix proportions, percentages of truck volumes and turn movements in traffic, Driver behavior and vehicle characteristics.
Animation output enables the analyst to see the vehicle behavior that is being modeled and assess the reasonableness of the microsimulation model itself. A two-stage process can be followed in reviewing the animation output. Run the animation at an extremely low demand level (so low that there is no congestion). Once the extremely low demand level tests have been completed, then run the simulation at 50 percent of the existing demand level.
Step 5 Model Calibration It is important to check if the developed model can represent the field. If not, the model parameters need to be changed (within a defined range) until the model is comparable. The model calibration is also to ensure that the microsimulation model will function as an accurate predictor of transportation system performance in alternatives analysis. Every microsimulation software program comes with a set of user-adjustable parameters for the purpose of calibrating the model to local conditions. Therefore, the objective of
75

calibration is to find the set of parameter values for the model that best reproduces observed measures of system performance. An effective calibration requires at least two key performance measures. Travel time or speed could be used as calibration parameters for 4L rural roadways. Step 6 Alternatives Analysis The objective of this project is to evaluate different median alternatives on 4L roadways. Multiple microsimulations run by varying random number seeds must be performed for each design alternative and the outcomes need to be captured. Some analyses require the explicit consideration of system performance in future years. In these cases, the analyst should make a forecast of future year travel demand. Forecasts of future travel demand significantly different from current conditions are best obtained from a travel demand model developed and maintained by the local metropolitan planning organization. Statistical methods (e.g., Welch's t-test) could be used to test a significant difference in the performance of the design alternatives. A sensitivity analysis is a targeted assessment of the reliability of the microsimulation results, given the uncertainty in the input or assumptions. The analyst identifies certain input or assumptions about which there is some uncertainty and varies them to see what their impact might be on the microsimulation results. Step 7 Final Report Making a clear and concise presentation of analytical findings is a critical element of a successful microsimulation analysis. The results from the microsimulation could be summarized and reported as needed.
76

The report should begin with the analytical objective and the context of the study. While preparing the final report, it is important to remember your audience and clarify
what was analyzed and not analyzed. Microsimulation can provide the analyst with valuable information on the performance of the existing transportation system and potential improvements. However, microsimulation can also be a time-consuming and resource-intensive activity. It is recommended that the operational performance will be evaluated only in conditions when the safety outcomes of the two proposed alternatives are relatively unclear. This project provides criteria for the median uses on the rural 4L segments and the criteria are based on the traffic volumes (AADT), truck percentage, and access point density. According to the criteria chart in Table 27, there are cases that the microsimulation may be conducted to assist the decision-making of cross-section selection. The simulation results can offer insights into other performance measures (besides the crash outcomes), such as delay, queue length, travel time, etc.
77

Chapter 10: SUMMARY & CONCLUSIONS NTMs mostly yield improved safety performance compared to other median types such as undivided, 4FM, TWLTL. However, constructing NTMs is costly because of the extra right-ofway that needs to be procured and additional construction costs involved. The goal of this study is to identify cost-effective median types, that maximize safety benefits, for four-lane rural roads in Georgia for a given set of roadway and traffic conditions. The authors thus came up with a criterion to determine under what conditions these four median types yield maximum safety benefits while considering the cost of construction.
Extensive data cleaning and manipulation were adopted to refine the data. Data were pooled from various sources such as GDOT, FHWA, and Google maps. Traffic volume data, crash data, roadway data, etc. were merged using several tools. Manual data extractions were also performed because of the lack of data such as access points. Also, manual verification of median types was carried out to ensure the authors include the right section in the desired median type.
Safety performance functions and Crash modification factors were used to create the criteria. SPFs were developed for all the four median types that were considered in the study. AADT, truck percentage, and access point density were considered as independent variables in the SPFs. SPFs were developed for different crash-severity combinations; eventually, KAB severities were used in the final criteria development as they are believed to influence the benefits (in terms of crash costs). The CMFs are calculated to show the effectiveness of a cross-section compared to the base cross-section which is a 4L-UR. CMFs are likely to vary across different values of AADT, segment length, truck percent, and access point density, indicating that the safety effectiveness of a
78

treatment on roadways is likely to vary across different situations. This study predicted the crash frequencies for one mile of segments. CMFs are estimated by keeping one of the factors (AADT, truck percent, access point density) constant and varying the other two factors. CMFs were calculated for typical values of AADT, truck percentage, and access point density. The key findings based on CMFs are as follows:
The segments with NTMs outperform the other three segment types across all AADTs, truck percentages and access point densities, except at very low AADT around or lower than 5,000 where the 4FMs appear to be associated with an improved condition.
AADT Is a principal factor to show the effectiveness of a cross-section compared to the base. In general, the 4FMs appear to have an improved performance than UR segments when the AADT is low (around or lower than 10,000); when the AADT increases (around or higher than 10,000), the TWLTLs have an improved performance than UR section and 4FMs; when the AADT reaches 20,000, the NTMs may be considered.
Truck percentage Is positively related to the safety effectiveness of 4FM, TWLTLs and NTMs compared to the base segment type. In other words, converting a 4L-UR segment into one of other three cross-section types can result in an even further improved condition for traffic with a higher truck percentage.
Access point density Is negatively related to the safety effectiveness of 4FMs, TWLTLs and NTMs compared to the base segment type. In other words, the safety effectiveness of these three types of medians decreases with the increase in the number of access points along a segment.
79

Note that, the safety performance on roadways may be associated with the posted speed limits. The estimation of SPFs/CMFs in this study did not consider the impact of speed limits. According to a preliminary study in Appendix A, a higher speed limit is likely associated with a greater rate for fatal and serious injury crashes. Besides estimating CMFs, to facilitate criteria development and make recommendations on median type, the authors estimated average annual crash reductions (compared with URs). These crash reductions were converted into monetary values using the average crash costs by severities. The safety benefits (shown below) and the project costs could be used to estimate benefit-cost ratios. Safety Benefits (in $million) for 4FMs, TWLTLs, and NTMs compared with URs as the base.
Notes: The base cross-section type is undivided roadways. The safety benefits cover reductions in all crash types. The safety benefits of 20 years are estimated. Cells with "N/A" are estimated with negative safety benefits. The results are applicable to roadways with posted speed limits of 50 mph and higher.
80

The research team gathered project costs from GDOT and several openly available sources and estimated BCRs considering the upper bound of the project costs and came up with the following criteria.

Preliminary criteria for recommending cross-section on four-lane rural roadways.

Vehicle AADT

Truck Percentage

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

A A A A B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>5% to <=10% A/B A/B A/B A/B B B B B B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>10% to <=15% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C B/C C/D C/D C/D C/D D D D D D D D D

>15% to <=20% A/B A/B A/B A/B B/C B/C B/C B/C B/C B/C B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

>20%

B B B B B/C B/C B/C B/C B/C B/C/D B/C/D B/C/D C/D C/D C/D C/D D D D D D D D D

Notes: A: Four-Lane Undivided

B: Four-Lane Divided w/4-ft Flush Median C: Four-Lane Divided w/TWLTL

D: Four-Lane Divided w/Non-Traversable Median or Barrier

When multiple letters (e.g., A/B, or B/C) are given in a cell, simulations are suggested to further examine and compare the performance of alternative cross-section designs. The performance may be examined from safety, mobility, and environmental aspects. It is important to note that, the criteria table provides rough cross-section recommendations merely based on the estimated crash reductions, empirical crash costs, and project costs from the previous projects. The project cost is project-specific, and it can vary substantially from location to location in a state due to various factors including the prior conditions of a project site, local labor cost, and material cost, etc. It is recommended that decision-makers or practitioners use the safety benefits table to compare with the cost of a specific proposed project and make a more realistic recommendation of the crosssection type.

81

REFERENCES American Association of State Highway and Transportation Officials (AASHTO), 2010. The
Highway Safety Manual, American Association of State Highway Transportation Professionals, Washington, D.C., http://www.highwaysafetymanual.org. American Association of State Highway and Transportation Officials (AASHTO), 2011. A Policy on Geometric Design of Highways and Streets. The American Association of State Highway and Transportation Officials, AASHTO Green Book, Washington D.C. Bachtel, D.C., Ragsdale, M. and Dowd, K.E., 1998. AN ANALYSIS OF THE GOVERNOR'S ROAD IMPROVEMENT PROGRAM (GRIP) FOR THE GEORGIA DEPARTMENT OF TRANSPORTATION. Online at http://www.dot.ga.gov/InvestSmart/GRIP/Resources/grip.pdf. Balke, K. N. and Fitzpatrick, K., 1993. An evaluation of flush medians and two-way, left-turn lanes on four-lane rural highways. Interim Report (No. FHWA/TX-94/1293-1). Online at https://static.tti.tamu.edu/tti.tamu.edu/documents/1293-1.pdf. Ballard, J. L. and McCoy, P. T., 1983. Operational effects of two-way left-turn lanes on twoway, four-lane streets. Transportation Research Record, 923, pp. 54-57. Barnett, T. E. and Wallace, W. L., 2010. Undivided Four- Lane Roadway Safety Review. Report (No. MM-2010-01-SAF). Capitol Fax.com, 2010. Generic Cost Per Mile Models. Online at https://capitolfax.com/summary.pdf. Center for Transportation Research and Education (CTRE). Continuous Raised Medians. Online at https://www.gtcmpo.org/sites/default/files/pdf/2007/ContinuousRaisedMedians.pdf. Accessed on 10.30 2019.
82

Clavenger, A.P. and Kociolek, A.V., 2006. Highway median impacts on wildlife movement and mortality: state of the practice survey and gap analysis (No. F/CA/MI-2006/09). California. Dept. of Transportation. Online at http://www.elkhornsloughctp.org/uploads/files/1182459531highway%20median%20imp acts.pdf.
Council, F.M. and Richard Stewart, J., 1999. Safety effects of the conversion of rural two-lane to four-lane roadways based on cross-sectional models. Transportation Research Record, 1665(1), pp.35-43.
Department of Numbers (DON), 2018. Georgia Household Income. Online at https://www.deptofnumbers.com/income/georgia/.
Florida Department of Transportation (FDOT), 2019a. Widen Existing 2 Lane Arterial to 4 Lanes Undivided; Add 1 Lane to Each Side; 7' Paved Shoulders (Project: WURA24-R-21BBO). Online at https://fdotwww.blob.core.windows.net/sitefinity/docs/defaultsource/programmanagement/estimates/lre/costpermilemodels/r21.pdf?sfvrsn=17d805c8_ 6.
Florida Department of Transportation (FDOT), 2019b. New Construction Divided Rural 4 Lane Interstate with Paved Shoulders 10' Outside and 4' Inside (Project: NDRI4L-R-05-BB). Online at https://fdotwww.blob.core.windows.net/sitefinity/docs/defaultsource/programmanagement/estimates/lre/costpermilemodels/r05.pdf?sfvrsn=2b3610f4_ 6.
Fitzpatrick, K. and Balke, K., 1995. Evaluation of flush medians and two-way, left-turn lanes on four-lane rural highways. Transportation research record, pp. 146-152.
83

Gattis, J., Balakumar, R. and Duncan, L. K., 2005. Effects of rural highway median treatments and access. Transportation research record, 1931(1), pp. 99-107.
Georgia Department of Transportation (GDOT), 2018a. The Governor's Road Improvement Program (GRIP): Fact Sheet. Online at http://www.dot.ga.gov/InvestSmart/GRIP/Resources/GRIPSystemSummaryFactSheet.pdf .
Georgia Department of Transportation, 2018b. GDOT Design Traffic Forecasting Manual. Online at http://www.dot.ga.gov/PartnerSmart/DesignManuals/Planning/GDOT%20Design%20Tra ffic%20Forecasting%20ManGual.pdf.
Georgia Department of Transportation (GDOT), 2019a. Highway Safety Plan FY 2020 Georgia. Online at https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/documents/ga_fy20_hsp.pdf
Georgia Department of Transportation (GDOT), 2019b. Pavement Type Selection Manual. Online at http://www.dot.ga.gov/PartnerSmart/DesignManuals/Pavement/Pavement%20Type%20S election%20Manual.pdf.
Georgia Department of Transportation, 2019c. GDOT Design Policy Manual. Online at: http://www.dot.ga.gov/PartnerSmart/DesignManuals/DesignPolicy/GDOT-DPM.pdf.
Georgia Department of Transportation (GDOT), 2019d. Benefits of Installing Medians Improving Safety and Reducing Crashes. http://www.dot.ga.gov/DriveSmart/SafetyOperation/Pages/Medians.aspx#tab-2. Accessed on 10.30 2019.
84

Graham, J.L., Harwood, D.W., Richard, K.R., O'Laughlin, M.K., Donnell, E.T. and Brennan, S.N., 2014. Median Cross-Section Design for Rural Divided Highways (No. Project 2221A).
Hillsborough MPO Metropolitan Planning for Transportation, 2014. LRTP Needs Assessment: Major Project Cost Estimates. Online at http://www.planhillsborough.org/wpcontent/uploads/2014/10/Capacity-Cost-Estimates-Tech-Memo.pdf.
Hovey, P.W. and Chowdhury, M.A., 2005. Development of crash reduction factors. Online at https://www.researchgate.net/profile/Mashrur_Chowdhury/publication/292148719_Devel opment_of_Crash_Reduction_Factors/links/56aada1b08ae8f3865666418/Developmentof-Crash-Reduction-Factors.pdf.
Humphreys, J.M., 2003. The economic benefits of the governor's road improvement program (GRIP). Online at https://athenaeum.libs.uga.edu/bitstream/handle/10724/36394/grip_study_october_2003.p df?sequence=1.
Iowa Department of Transportation Office of Design. Design Manual Chapter 5 Roadway Design: Access Management. Online at https://intrans.iastate.edu/app/uploads/sites/15/2018/09/5L-3.pdf. Accessed on 10.30 2019.
Iowa Department of Transportation Office of Design. Design Manual Chapter 6 Geometric Design: Continuous Two-Way LeftTurn Lanes (TWLTLs). Online at https://iowadot.gov/design/dmanual/06c-06.pdf. Accessed on 10.30 2019.
85

Idaho Transportation Department (ITD). Practical Solutions for Highway Design. Online at https://apps.itd.idaho.gov/apps/manuals/DesignSolutions/files/DesignSolutionsPrintable.p df. Accessed on 10.30 2019.
Kentucky Transportation Cabinet (KDOT). Rationale for Median Type Recommendations. Online at https://transportation.ky.gov/CongestionToolbox/Documents/Kentucky%20Median%20Type%20Guidelines.pdf. Accessed on 10.30 2019.
Knapp, K.K. and Giese, K., 2001. Guidelines for the conversion of urban four-lane undivided roadways to three-lane two-way left-turn lane facilities. Online at http://publications.iowa.gov/2888/1/4to3lane.pdf.
Knapp, K., Chandler, B., Atkinson, J., Welch, T., Rigdon, H., Retting, R., Meekins, S., Widstrand, E. and Porter, R.J., 2014. Road diet informational guide (No. FHWA-SA-14028). United States. Federal Highway Administration. Office of Safety.
Liu, J., Khattak, A. J., and Wali, B., 2017. Do safety performance functions used for predicting crash frequency vary across space? Applying geographically weighted regressions to account for spatial heterogeneity. Accident Analysis & Prevention, 109, pp. 132-142
McCoy, P. T., Ballard, J. L., Eitel, D. S. and Witt, W. E., 1988. Two-way left-turn lane guidelines for urban four-lane roadways. Transportation Research Record, 1195, pp. 1119.
Mohamedshah, Y. M., and Kohls A. R., 1994. Accident rates using HSIS. Public Roads 58 (1). Online at https://www.fhwa.dot.gov/publications/publicroads/94summer/p94su44.cfm.
86

Oklahoma Department of Transportaion (ODOT). Roadway Design Mannual: 4-lane vs. 5-lane Comparison/Recommendation. Online at http://www.okladot.state.ok.us/p-rdiv/us81corridor/appendh.pdf. Accessed on 10.30 2019.
Pernia, J., Lu, J.J. and Peng, H., 2004. Safety Issues Related to Two-way Left-turn Lanes (No. Final Technical Report).
Persaud, B., Lyon, C., Eccles, K., Lefler, N., Carter, D. and Amjadi, R., 2008. Safety evaluation of installing center two-way left-turn lanes on two-lane roads (No. FHWA-HRT-08-042). United States. Federal Highway Administration.
Phillips, S. L., 2004. Empirical Collision Model for Four-Lane Median Divided and Five-Lane with TWLTL Segments. Online at https://repository.lib.ncsu.edu/bitstream/handle/1840.16/2021/etd.pdf?sequence=1.
Shea, J.K., Machemehl, R.B. and Rioux, T.W., 2000. Design Guidelines for Provision of Median Access on Principal Arterials (No. FHWA/TX-0-1846-1). University of Texas at Austin. Center for Transportation Research.
South Carolina Department of Transportation (SCDOT). Roadway Design Manual: Connecting People and Places. Online at https://www.scdot.org/business/pdf/roadway/2017_SCDOT_Roadway_Design_Manual.p df. Accessed on 10.30 2019.
Squires, C. A. and Parsonson, P. S., 1989. Accident comparison of raised median and two-way left-turn lane median treatments. Transportation Research Record, 1239, pp. 30-40.
Texas Department of Transportation (TxDOT), 2020. Roadway Design Manual. Online at http://onlinemanuals.txdot.gov/txdotmanuals/rdw/rdw.pdf. Accessed on 10.30 2020.
87

U.S. Bureau of Labor Statistics Southeast Information Office (U.S. BLS SIO), 2019. Changing Compensation Costs in the Atlanta Metropolitan Area December 2019. Online at https://www.bls.gov/regions/southeast/newsrelease/2020/employmentcostindex_atlanta_20200204.htm.
Wunderlich, K.E., Vasudevan, M. and Wang, P., 2019. TAT Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software 2019 Update to the 2004 Version (No. FHWA-HOP-18-036). United States. Federal Highway Administration.
88

APPENDIX A. Speed and Safety Performance on Rural Four-lane Roadways. This appendix provides some quick results on the relationship between the posted speed limits and safety performance on rural four-lane roadways in Georgia. The safety performance on roadways may be associated with the posted speed limits. We linked the roadway segments sampled in this project with the speed limit information in the Highway Performance Monitoring Database. Four types of cross-sections were investigated:
1. Four-lane Undivided 2. Four-lane Divided w/4-ft. flush paved median 3. Four-lane Divided w/Two-Way Left-Turn Lane 4. Four-lane Divided w/Non-traversable median or median barrier
Table A-1 shows the numbers of observations for four cross-section types under three possible speed limits. Note, the scope of this project is limited to high-speed rural roadways, > 45 mph. In the sampled segments, there is no observation for Undivided, 4-ft Flush Median and TWLTL with a speed limit of 65 mph. The analysis is limited to the observation available in sampled roadway segments.
89

Table A-1. Number of observations.

Undivided 4-ft Flush Median TWLTL Non-Traversable

Speed Limit
50 55 50 55 50 55 50 55 65

Number of observations (segments) Total 54 25 19 146 103 455 22 783 369

The crash rate for each segment is estimated using the equation below:

100,000,000

= 365

(1)

where,

= Crash rate per 100-million VMT;

= Total number of crashes in the 6-year study period;

= Average Traffic volume using Annual Average Daily Volumes (AADT) in the 6-

year study period;

= Total number of years in the study period; and

= Length (in miles) of the roadway segment.

Note that, the crash rates can be sensitive to the exposure that includes the traffic volume and the segment length. For segments with small exposures (low AADT and short segment), the crash rates can be extreme especially with a limited number of observations. To show an unbiased comparison of safety performance across observations of different exposures, the crash rates are weighted by the exposure (AADT Segment Length). In other words, observations with greater
90

exposure are weighted more in the average weighted crash rates for a segment group. Table 2 shows the weighted crash rates for four types of cross-sections under different speed limits. In addition to the rates for total crashes, Table A-2 also shows the rates for different crash severities including:
K Fatal Injury A Suspected Serious injury B Suspected Minor or Visible Injury C Possible Injury or Complaint O No Apparent Injury

Table A-2. Weighted crash rates (per 100 million vehicle miles).

Speed Limit

Weighted* crash rate (per 100 million vehicle miles)

Total K

A

B

C

O

Undivided

50

93.005 2.608 16.225 10.430 10.430 57.367

55

107.730 2.284 19.414 11.420 11.420 64.334

4-ft Flush Median

50 55

60.604 97.653

0.522 9.404 4.180 4.180 33.959 1.196 13.508 12.664 12.664 63.742

TWLTL

50

87.440 1.154 9.575 7.960 7.960 58.485

55

93.683 1.277 9.795 10.144 10.144 63.833

50

93.079 1.103 5.294 7.940 7.940 66.611

Non-Traversable

55

61.719 0.807 9.014 7.127 7.127 40.798

65

55.682 1.012 8.332 6.796 6.796 36.342

*Weighted by AADT Segment Length

The results show that the total crash rates are greater for higher speed limits for the first three cross-sections (Undivided, 4-ft Flush Median, and TWLTL), and the total crash rates drop for higher speed limits for the last cross-section type with non-traversable medians. Across all four sections, it seems that a higher speed limit is associated with a greater rate for crashes of high severities (K and A).

91

Note that, the results are limited to the available observations in the sampled roadway segments. Further, the highway safety performance can be affected by many factors, and the results presented in this document are not intended to attribute the crash rate differences to speed limit only. For example, for a given 50 mph non-traversable roadway we may experience a higher rate than the 65 mph roadways if we consider certain conditions that may exist on those segments, such as a greater number of driveways/intersections, horizontal/vertical alignment changes that are more pronounced on the lower speeds, and other characteristics (lighting, shoulder widths, land use, etc.). Therefore, a comprehensive investigation would be needed to uncover the complete picture of contributing factors of safety performance on roadways.
92

APPENDIX B. CMF Plots along AADT for different Truck Percentages and Access Point Densities.
93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

APPENDIX C. Estimated average CMFs for KAB and CO crashes.

Estimated average CMFs for KAB crashes

Vehicle AADT

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Truck Percentage

Access Point Density, AP/mile

<=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30 <=10 >10 >20 >30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

0.41 0.43 0.45 0.40 0.57 0.59 0.62 0.55 0.65 0.67 0.70 0.62 0.70 0.73 0.76 0.68 0.75 0.78 0.81 0.72 0.79 0.82 0.85 0.76

>5% to <=10% 0.37 0.38 0.40 0.36 0.60 0.62 0.65 0.57 0.73 0.76 0.79 0.70 0.83 0.86 0.90 0.80 0.91 0.95 0.99 0.88 0.99 N/A N/A N/A

4-ft Flush Median >10% to <=15% 0.34 0.35 0.37 0.32 0.63 0.65 0.68 0.60 0.82 0.85 0.89 0.79 0.98 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% 0.31 0.32 0.34 0.30 0.66 0.69 0.71 0.63 0.92 0.96 1.00 0.89 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

0.28 0.29 0.31 0.27 0.69 0.72 0.75 0.67 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5% >5% to <=10% >10% to <=15% >15% to <=20%
>20%

0.77 0.90 N/A 0.66 0.52 0.60 0.70 0.45 0.45 0.53 0.61 0.39 0.41 0.48 0.56 0.36 0.39 0.45 0.53 0.33 0.37 0.43 0.50 0.32 0.68 0.81 0.92 0.59 0.48 0.55 0.64 0.41 0.42 0.49 0.56 0.36 0.39 0.45 0.52 0.33 0.36 0.42 0.49 0.31 0.35 0.40 0.47 0.30 0.62 0.72 0.83 0.53 0.44 0.51 0.59 0.38 0.39 0.45 0.52 0.33 0.36 0.42 0.48 0.31 0.34 0.39 0.46 0.29 0.32 0.38 0.44 0.28 0.55 0.63 0.74 0.47 0.40 0.46 0.54 0.34 0.36 0.42 0.48 0.31 0.33 0.39 0.45 0.29 0.32 0.37 0.43 0.27 0.30 0.35 0.41 0.26 0.49 0.57 0.66 0.42 0.37 0.43 0.49 0.32 0.33 0.38 0.45 0.28 0.31 0.36 0.42 0.27 0.29 0.34 0.40 0.25 0.28 0.33 0.38 0.24

<=5%

N/A N/A N/A N/A 0.86 0.92 0.99 0.80 0.57 0.61 0.65 0.53 0.44 0.47 0.50 0.41 0.36 0.39 0.41 0.34 0.31 0.33 0.35 0.29

>5% to <=10% N/A N/A N/A N/A 0.68 0.73 0.78 0.63 0.49 0.52 0.56 0.45 0.39 0.42 0.45 0.37 0.33 0.36 0.38 0.31 0.29 0.32 0.34 0.27 Non-traversable >10% to <=15% N/A N/A N/A N/A 0.54 0.58 0.62 0.50 0.42 0.45 0.48 0.39 0.35 0.38 0.41 0.33 0.31 0.33 0.36 0.29 0.28 0.30 0.32 0.26

>15% to <=20% 0.77 0.80 0.87 0.70 0.43 0.46 0.49 0.40 0.36 0.38 0.41 0.33 0.32 0.34 0.36 0.29 0.29 0.31 0.33 0.27 0.27 0.29 0.31 0.25

>20%

0.47 0.50 0.54 0.44 0.34 0.37 0.39 0.32 0.30 0.33 0.35 0.28 0.28 0.30 0.33 0.26 0.27 0.29 0.31 0.25 0.26 0.28 0.30 0.24

Notes: The CMF calculation base cross-section is undivided roadways. Cells with "N/A" are estimated with

average CMFs greater than 1.00. The results are applicable to roadways with posted speed limits of 50 mph

and higher.

Estimated average CMFs for CO crashes

Vehicle AADT

<=5,000

>5,000 to 10,000

>10,000 to 15,000

>15,000 to 20,000

>20,000 to 25,000

>25,000

Truck Percentage

Access Point Density, AP/mile

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

>10 >20

<=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30 <=10

>30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

to 20 to 30

<=5%

0.29 0.33 0.39 0.25 0.72 0.83 0.95 0.63 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>5% to <=10% 0.22 0.25 0.29 0.19 0.67 0.78 0.89 0.59 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

4-ft Flush Median >10% to <=15% 0.16 0.19 0.22 0.14 0.63 0.73 0.84 0.55 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>15% to <=20% 0.13 0.15 0.17 0.11 0.60 0.69 0.79 0.52 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

>20%

0.10 0.11 0.13 0.08 0.56 0.64 0.74 0.49 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

TWLTL

<=5% >5% to <=10% >10% to <=15% >15% to <=20%
>20%

N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A 0.95 0.96 N/A N/A 0.91 0.93 N/A N/A 0.88 0.95 N/A N/A 0.91 0.94 N/A N/A 0.90 0.94 N/A N/A 0.90 0.94 N/A N/A 0.90 0.94 N/A N/A 0.89 0.94 N/A N/A 0.89 0.63 0.66 0.70 0.60 0.77 0.81 0.85 0.74 0.84 0.88 0.93 0.80 0.88 0.93 0.98 0.84 0.92 0.97 N/A 0.88 0.95 N/A N/A 0.90 0.43 0.45 0.47 0.41 0.63 0.67 0.70 0.60 0.75 0.78 0.83 0.71 0.83 0.87 0.92 0.79 0.90 0.95 N/A 0.86 0.96 N/A N/A 0.91 0.29 0.31 0.32 0.28 0.52 0.55 0.57 0.49 0.66 0.70 0.74 0.63 0.78 0.82 0.86 0.74 0.88 0.93 0.97 0.84 0.97 N/A N/A 0.92

<=5%

N/A N/A N/A N/A N/A N/A N/A N/A 0.98 0.98 0.97 0.99 0.84 0.83 0.83 0.85 0.75 0.74 0.73 0.76 0.68 0.68 0.67 0.69

>5% to <=10% N/A N/A N/A N/A N/A N/A N/A N/A 0.90 0.89 0.88 0.91 0.81 0.80 0.79 0.82 0.74 0.74 0.73 0.75 0.70 0.69 0.68 0.70 Non-traversable >10% to <=15% N/A N/A N/A N/A 0.91 0.90 0.89 0.92 0.83 0.82 0.81 0.83 0.77 0.77 0.76 0.78 0.74 0.73 0.72 0.75 0.71 0.70 0.70 0.72

>15% to <=20% 0.85 0.84 0.83 0.85 0.78 0.77 0.76 0.79 0.76 0.75 0.74 0.76 0.74 0.74 0.73 0.75 0.73 0.73 0.72 0.74 0.72 0.72 0.71 0.73

>20%

0.60 0.59 0.59 0.61 0.67 0.66 0.65 0.67 0.69 0.69 0.68 0.70 0.71 0.71 0.70 0.72 0.73 0.72 0.71 0.73 0.74 0.73 0.72 0.75

Notes: The CMF calculation base cross-section is undivided roadways. Cells with "N/A" are estimated with

average CMFs greater than 1.00. The results are applicable to roadways with posted speed limits of 50 mph

and higher.

124