Design of Interchange Loop Ramps and Pavement/Shoulder Cross-Slope Breaks (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Part 1: Design Guidance for Interchange Loop Ramps

1 - ii ACKNOWLEDGMENT The research reported herein was performed under NCHRP Project 3-105, “Design Guidance for Interchange Loop Ramps” as Phases I and II. This report was prepared by Dr. Darren J. Torbic, Ms. Lindsay M. Lucas, and Mr. Douglas W. Harwood of MRIGlobal, and Mr. Marcus A. Brewer, Dr. Eun Sug Park, Dr. Raul Avelar, and Mr. Michael P. Pratt of the Texas A&M Transportation Institute. Mr. Chris A. Fees and Mr. John J. Ronchetto of MRIGlobal, Mr. Dan Walker of the Texas A&M Transportation Institute, Mr. Alan Glenn and Ms. Erin McPherson of Quincy Engineering, and Ms. Heidi Ouren of HQE Incorporated played key roles in this research. The authors wish to thank the State Departments of Transportation of California, Kansas, Missouri, and Texas for their assistance in this research.

1 - iii Abstract The objective of this research was to develop improved design guidance for interchange loop ramps. An observational field study was conducted to investigate the relationship between speed and lane position of vehicles and key design elements of the ramp proper, and the difference in performance on the ramp proper between single-lane and multi-lane loop ramps. Also, comparisons of predicted and observed crash frequencies were performed to validate the Highway Safety Manual (HSM) crash prediction method for loop ramps. Based on the research findings, the primary recommendations regarding the design of the loop ramp proper, applicable to loop ramps at service interchanges in both urban and rural areas, are as follows: (a) For a given radius and design speed, recommended lane and shoulder widths for entrance and exit loop ramps are provided that are expected to induce speeds at or below the ramp design speed. For a given radius and ramp design speed, the recommended lane and shoulder widths are also expected to result in similar levels of safety, and vehicles are expected to stay within their intended travel lane. (b) Outside lane widths of 12-ft for multi-lane entrance loop ramps and 14-ft for multi-lane exit loop ramps are sufficient to accommodate traffic comprised primarily of passenger vehicles. If the outside lane is expected to accommodate a moderate to high volume of trucks, the outside lane width should be increased. (c) When implementing the HSM ramp crash prediction methodology, separate calibration factors should be calculated for diamond ramps and loop ramps.

1 - iv Part 1 Contents ACKNOWLEDGEMENT .............................................................................................................. ii  Abstract ................................................................................................................................. iii  Figures ................................................................................................................................. vi  Tables ................................................................................................................................ vii  Summary ................................................................................................................................. ix  Section 1. Introduction ...................................................................................................................1  1.1 Background ...........................................................................................................1  1.2 Research Objective and Scope ..............................................................................4  1.3 Outline of Report ..................................................................................................5  Section 2. Summary of Literature Review .....................................................................................6  2.1 Current AASHTO Policy on Loop Ramps ...........................................................6  2.2 ITE Freeway and Interchange Geometric Design Handbook ............................14  2.3 Highway Safety Manual (HSM) ..........................................................................15  2.4 Operational Issues for Loop Ramps ....................................................................17  2.5 State Design Manuals on Loop Ramps ...............................................................18  2.6 Geometric Design and Safety .............................................................................18  2.7 Summary .............................................................................................................20  Section 3. Observational Field Study of Loop Ramp Proper.......................................................22  3.1 Study Locations ..................................................................................................22  3.2 Field Data Collection Procedures .......................................................................30  3.3 Data Processing ...................................................................................................32  3.4 Analysis of Observational Data ..........................................................................37  Section 4. Application of the HSM Ramp Crash Prediction Method to Loop Ramps ................67  4.1 HSM Ramp Crash Prediction Method ................................................................67  4.2 Application of the HSM Crash Prediction Method to Specific Ramp Configurations.....................................................................................................68  4.3 Key Considerations in Assessing the Application of the HSM Ramp Crash Prediction Models to Loop Ramps .....................................................................69  4.4 Selection of Study Ramps ...................................................................................70  4.5 Data Collection ...................................................................................................71  4.6 Descriptive Statistics for Ramps Included in the Study ......................................72  4.7 Application of HSM Ramp Crash Prediction Method ........................................80  4.8 Summary Comparison of Predicted and Observed Crash Frequencies ..............81  4.9 Statistical Comparison of Predicted and Observed Crash Frequencies ..............84  4.10 Discussion of Results ........................................................................................87  Section 5. Design Guidance .........................................................................................................89  5.1 Design Vehicles for Loop Ramp Proper .............................................................89  5.2 Recommended Lane and Shoulder Widths for Loop Ramps ..............................90  5.3 Multi-Lane Ramps ..............................................................................................98  5.4 Safety Prediction of Design Alternatives ............................................................98  Section 6. Conclusions and Recommendations for Future Research ........................................100 

1 - v 6.1 Conclusions .......................................................................................................100  6.2 Recommendations for Future Research ............................................................103  Section 7. References .................................................................................................................105  Appendices Appendix A—Sites Included in Observational Field Study of Loop Ramps Appendix B—Recommended Changes for Consideration in the Next Edition of the Green Book

1 - vi Figures Figure 1. General Types of Ramps ..................................................................................................2 Figure 2. Expanded Interchange Configurations with Loop Ramps ................................................3 Figure 3. Recommended Minimum Ramp Terminal Spacing .......................................................12 Figure 4. Typical Design of a Two-Lane Loop Ramp ...................................................................15 Figure 5. I-435/Shawnee Mission Parkway ...................................................................................22  Figure 6. Crossroad Ramp Terminal Types/Classifications ..........................................................27  Figure 7. Sample Field Setup for Observational Study of Ramp Proper .......................................31  Figure 8. Sample Images from Video to Document Vehicle Lane Position on a Single-Lane Ramp .......................................................................................................34  Figure 9. Sample Images from Video to Document Unusual or Critical Maneuvers on Exit Ramps ..............................................................................................................36  Figure 10. Correlations among Variables in the Speed Database for All Sites .............................43  Figure 11. Correlations among variables in the lane position database for all sites ......................60 

1 - vii Tables Table 1. Guide Values for Ramp Design Speed as Related to Highway Design Speed ................11  Table 2. Minimum Acceleration Lengths for Entrance Terminals with Flat Grades of Two Percent or Less .............................................................................................................13  Table 3. Minimum Deceleration Lengths for Exit Terminals with Flat Grades of Two Percent or Less .............................................................................................................13  Table 4. Speed-Change Lane Adjustment Factor as a Function of Grade .....................................13  Table 5. Loop Ramp Characteristics ..............................................................................................14  Table 6. Summary of Ramp Characteristics in NCHRP 15-45 Database Used to Develop Predictive Models for Ramps.......................................................................................15  Table 7. Summary of State Policies Related to Design Speed, Curve Radius, and Widths for Loop Ramps ...........................................................................................................19  Table 8. General Characteristics of Study Locations (Entrance Ramps) .......................................25  Table 9. General Characteristics of Study Locations (Exit Ramps) ..............................................26  Table 10. Characteristics of Ramp Proper (Entrance Ramps) .......................................................28  Table 11. Characteristics of Ramp Proper (Exit Ramps) ...............................................................29  Table 12. Descriptive Statistics of Vehicle Speeds on Ramp Proper (Entrance Ramps) ..............40  Table 13. Descriptive Statistics of Vehicle Speeds on Ramp Proper (Exit Ramps) ......................41  Table 14. Summary Statistics for Speed Data ...............................................................................42  Table 15. Speed Prediction Model to Estimate Speed at the Midpoint of the Controlling Curve on the Ramp Proper of an Entrance Loop Ramp ..............................................44  Table 16. Speed Prediction Model to Estimate Speed at the End of the Controlling Curve on the Ramp Proper of an Entrance Loop Ramp .........................................................45  Table 17. Speed Prediction Model to Estimate Speed at the Midpoint of the Controlling Curve on the Ramp Proper of an Exit Loop Ramp ......................................................46  Table 18. Speed Prediction Model to Estimate Speed at the Beginning of the Controlling Curve on the Ramp Proper of an Exit Loop Ramp ......................................................47  Table 19. Input Data for Ramp Curve Speed Prediction Procedures in ISATe .............................47  Table 20. Comparison of Speeds Estimated by ISATe, Predicted by Model, and Measured in Field (Entrance Ramps) ...........................................................................................52  Table 21. Comparison of Speeds Estimated by ISATe, Predicted by Model, and Measured in Field (Exit Ramps) ...................................................................................................53  Table 22. Descriptive Lane Position Statistics on Ramp Proper (Entrance Ramps) .....................57  Table 23. Descriptive Lane Position Statistics on Ramp Proper (Exit Ramps) .............................58  Table 24. Summary Statistics for Lane Position Data ...................................................................59  Table 25. Lane Position Model for Entrance Ramps .....................................................................61  Table 26. Lane Position Model for Exit Ramps ............................................................................62  Table 27. Summary Statistics of Encroachment and Critical Maneuvers Observed on Exit Ramps ..............................................................................................................64  Table 28. Summary Statistics of Encroachment and Critical Maneuvers Observed on Exit Ramps by Encroachment, Maneuver Type, and Vehicle Type .......................64  Table 29. Summary Statistics of Encroachment and Critical Maneuvers and Key Site Characteristics of the Exit Ramps ................................................................................65  Table 30. Descriptive Statistics for California Rural Exit Ramps .................................................73  Table 31. Descriptive Statistics for California Rural Entrance Ramps .........................................74 

1 - viii Table 32. Descriptive Statistics for California Urban Exit Ramps ................................................75  Table 33. Descriptive Statistics for California Urban Entrance Ramps ........................................76  Table 34. Descriptive Statistics for Washington Rural Exit Ramps ..............................................77  Table 35. Descriptive Statistics for Washington Rural Entrance Ramps ......................................78  Table 36. Descriptive Statistics for Washington Urban Exit Ramps .............................................79  Table 37. Descriptive Statistics for Washington Urban Entrance Ramps .....................................80  Table 38. Total Crash Statistics by State and Ramp Classification ...............................................82  Table 39. Fatal-and-Injury Crash Statistics by State and Ramp Classification .............................83  Table 40. Statistical Significance Levels Associated with First Set of Negative Binomial Regression Models .......................................................................................................85  Table 41. Final Comparisons of Predicted to Observed Total Crash Rates ...................................86  Table 42. Final Comparisons of Predicted to Observed Fatal-and-Injury Crash Rates .................86  Table 43. Recommended Lane and Shoulder Width Combinations for Controlling Curve on Entrance Loop Ramps .............................................................................................93  Table 44. Recommended Lane and Shoulder Widths for Exit Loop Ramps – Simple Curves ..............................................................................................................96  Table 45. Recommended Lane and Shoulder Widths for Exit Loop Ramps – Compound Curves ......................................................................................................97 

1 - ix Summary Interchange projects are among the most complex and expensive projects constructed by highway agencies. Interchanges are comprised of individual ramps. Various ramp configurations are used and can be broadly classified as diagonal, loop, directional, semidirect, outer connection, and one quadrant. Loop ramps are prominent in full cloverleafs, partial cloverleafs, trumpets, and other interchange configurations. Despite their long use, there is little research on the design, safety, and operational characteristics of loop ramps. The objective of this research was to develop improved design guidance for interchange loop ramps. The research focused on developing improved design guidance for the ramp proper portion of the ramp, taking into consideration the connections on either side of the ramp proper. An observational field study of driver behavior and vehicle operations on the ramp proper of single-lane and multi-lane loop ramps was conducted to address two high-priority gaps in knowledge related to the design of loop ramps, namely: (a) the relationship between speed and lane position of vehicles and key design elements of the ramp proper, and (b) the difference in performance on the ramp proper between single-lane and multi-lane loop ramps. Data were collected at 28 loop ramps, including 15 entrance ramps and 13 exit ramps. All of the ramps were located at service interchanges. Speed data were collected at two points on the controlling curve of each ramp; positions of vehicles within the roadway were collected at the midpoint of the controlling curve on the ramp; and unusual or critical behavior near the beginning of the controlling curve in the direction of travel on exit loop ramps were collected. The Highway Safety Manual (HSM) crash prediction method for ramp segments does not separate procedures for specific ramp configurations, such as diamond ramps and loop ramps. Since there is no separate model for loop ramps in the HSM ramp crash prediction method, a key consideration is whether the existing HSM method is effective in distinguishing between the safety performance of loop ramps and other ramp configurations, such as diamond ramps. An investigation was conducted to validate the HSM ramp crash prediction method for loop ramps. Comparisons of predicted and observed crash frequencies for specific combinations of area type, ramp type, and ramp configuration were performed to validate the HSM crash prediction method for ramps. The general conclusions reached based upon the results of this research are as follows: Entrance Ramps  On entrance loop ramps, speeds of vehicles at the end of the controlling curve were found to be slightly higher than speeds at the midpoint of the controlling curve, suggesting that vehicles accelerate as they traverse the length of the controlling curve on the ramp proper.  Key roadway and cross-sectional design elements that significantly influence vehicle speeds at the midpoint of the controlling curve include curve radius, lane width, inside (right) shoulder width, and outside (left) shoulder width. When comparing the impact of

1 - x lane width and shoulder widths on speeds, for a given incremental increase in width (e.g., 1 ft), shoulder widths have a greater influence on speeds than travel lane widths. This may be because as lane widths become sufficiently wide (e.g., 16, 18, 20 ft or more), the relative effect of lane width on speed becomes less in comparison to shoulder width.  Key roadway and cross-sectional design elements that significantly influence vehicle speeds at the end of the controlling curve include curve radius and outside shoulder width. It is logical that the outside shoulder width influences vehicle speeds as drivers approach the end of the curve and begin looking for gaps in freeway traffic. A wider outside shoulder provides drivers with more “forgiveness” in lateral placement as the driving task shifts away from negotiating the curve and shifts toward looking for gaps.  Key roadway and cross-sectional design elements that significantly influence lane position at the midpoint of the controlling curve include lane width, outside shoulder width, superelevation, and grade. As lane width increases, vehicles tend to move farther away from the inside lane line (which is expected). However, they tend to move closer to the inside lane line as either outside shoulder width or superelevation increases. Also, vehicles are positioned closer to the inside lane line on upgrades than on downgrades.  As vehicles traverse an entrance loop ramp, the right tires of passenger vehicles and trucks are positioned approximately an equal distance from the inside lane line; so aside from offtracking issues associated with larger trucks, there are no major concerns associated with differences between the lane positions of trucks and passenger vehicles. Exit Ramps  On exit loop ramps, vehicle speeds were found to be slightly higher at the beginning of the controlling curve than at the midpoint of the controlling curve, suggesting that vehicles decelerate as they transition from the freeway mainline ramp terminal along the ramp proper.  The radius of the controlling curve is the only key roadway and/or cross-sectional design element that significantly influences vehicle speeds at the beginning (i.e., PC) of the controlling curve. This is not surprising since drivers are just transitioning to the ramp proper and have had little time to process the site characteristics and context of the ramp. As vehicles proceed along the ramp proper and reach the midpoint of the controlling curve, other key roadway and cross-sectional design elements (in addition to curve radius) significantly influence vehicle speeds; these include outside shoulder width, type of curvature (simple or compound), and type of mainline freeway ramp terminal. It seems reasonable that vehicle speeds would increase as outside shoulder width increases because drivers may feel more comfortable driving at higher speeds where more recovery area is available on the outside of the ramp. Also, vehicle speeds are higher where the ramp proper is designed with a simple curve rather than a compound curve.  Trucks were found to travel slower than passenger vehicles on exit ramps. Thus, if the ramp proper is designed to accommodate the maximum predicted speeds of passenger vehicles, then the ramp proper will be designed with a greater margin of safety against truck rollovers and skidding.

1 - xi  Key roadway and cross-sectional design elements that significantly influence lane position at the midpoint of the controlling curve include lane width and type of freeway mainline ramp terminal. As lane width increases, vehicles tend to move farther away from the inside lane line (which is expected). Vehicles are positioned farther away from the inside lane line on ramps where the freeway lane is dropped as it transitions to the ramp proper compared to loop ramps following a taper- or parallel type of speed-change lane or weave area. This may be due to fewer lane-changing maneuvers in the immediate vicinity of a freeway ramp that follows a lane drop.  Trucks are typically positioned farther away from the inside lane line than passenger vehicles, and most passenger vehicles are positioned within the travel lane and do not encroach on the inside shoulder. Thus, the lane positioning of trucks does not raise concerns about encroachment onto the inside shoulder of an exit ramp. Multi-Lane Ramps  Vehicles in the outside lane (or lanes) of a multi-lane loop ramp travel at speeds approximately 1 to 2 mph faster than vehicles traveling in the inside lane.  Vehicles traveling in the outside lane of a multi-lane loop ramp are positioned slightly farther from the inside lane line than vehicles traveling in the inside lane. HSM Ramp Crash Prediction Method  The HSM ramp crash prediction methodology is better at predicting diamond ramp crashes than predicting loop ramp crashes. Separate calibration factors for diamond ramps and loop ramps are needed for more accurate comparisons between the safety performances of these different ramp types. Design Guidance The design guidance is applicable to loop ramps at service interchanges in both urban and rural areas.  Consistent with current AASHTO policy in the 2011 Green Book for determining minimum lengths of freeway mainline ramp terminals, it is recommended that passenger vehicles remain the design vehicle for the design of loop ramps.  The alignment and cross section of a loop ramp should be designed so that vehicles traverse the loop ramp at or below the ramp design speed. For curve radii ranging from 100 to 300 ft for design speeds from 20 to 35 mph, recommended lane and shoulder widths for entrance and exit loop ramps are provided that are expected to induce speeds at or below the ramp design speed. For a given radius and ramp design speed, the recommended lane and shoulder widths are also expected to result in similar levels of safety, and vehicles are expected to stay within their intended travel lane. Alternatively, designers may use the speed and lane position prediction models, developed as part of this research, and ISATe to evaluate and design alignments and cross sections of loop ramps.

1 - xii  Based upon speeds and lane positions of vehicles operating in the outside lane of multi- lane loop ramps, no special design considerations are necessary for the design of multi- lane loop ramps to accommodate large differentials in speeds of vehicles traveling in the outside lane compared to the inside lane or to accommodate vehicles in the outside lane that significantly gravitate to the inside and encroach on the inside travel lane. Outside lane widths of 12-ft for multi-lane entrance loop ramps and 14-ft for multi-lane exit loop ramps are sufficient to accommodate traffic comprised primarily of passenger vehicles. If the outside lane is expected to accommodate a moderate to high volume of trucks, the outside lane width should be increased.  As part of implementing the HSM ramp crash prediction methodology to estimate predicted and/or expected crash frequencies for individual ramps, separate calibration factors should be calculated for diamond ramps and loop ramps for more accurate comparisons between the safety performances of these different ramp types.

1 - 1 Section 1. Introduction 1.1 Background Interchange projects are typically among the most complex and expensive projects constructed by highway agencies. A variety of basic interchange configurations have been constructed over the years, including:  Trumpet  Three-leg directional  One quadrant  Diamond  Single-point urban interchange  Full cloverleaf  Partial cloverleaf  All directional four-leg Design decisions on large investments such as interchanges must be made wisely, considering factors such as safety, operations, right-of-way and environmental constraints, and nearby traffic generators. Interchanges are comprised of individual ramps. The primary components of a ramp include the freeway mainline ramp terminal (i.e., an acceleration or deceleration lane or also referred to as a speed-change lane), the ramp proper (i.e., the turning roadway), and the crossroad ramp terminal. Various ramp configurations are used and can be broadly classified as:  Diagonal  Loop  Directional  Semidirect  Outer connection  One quadrant Figure 1 illustrates the general layout of these ramps. Combinations of these types of ramps make up various interchange configurations. Figure 2 illustrates and further classifies the typical range of interchange types in which loop ramps are used:  Full cloverleaf  Parclo-A  Parclo-B  Parclo-AB  Parclo-A/4 quad  Parclo-B/4 quad  Parclo-AB/4 quad  Full diamond with one loop  Directional with two loops Despite their long use, there is little research on the design, safety, and operational characteristics of loop ramps. Research by Bauer and Harwood (1998) indicates that, in some situations, loop ramps have higher crash rates than diagonal ramps which is consistent with the safety performance functions (SPFs) developed recently for the Highway Safety Manual (HSM) (AASHTO, 2014). Chapter 10 of the American Association of State Highway and Transportation Officials (AASHTO) A Policy on Geometric Design of Highways and Streets (2011) (referred to as the Green Book) provides little guidance on the unique challenges of designing multi-lane loop ramps. This report focuses on providing improved design guidance for interchange loop ramps and their connections.

Figure 1. General Typ 1 - 2 es of Ramps (AASHTO, 2011)

1 - 3 Full Cloverleaf Parclo-A Parclo-B Parclo-AB Parclo-A/4 Quad Parclo-B/4 Quad Figure 2. Expanded Interchange Configurations with Loop Ramps

1 - 4 Parclo-AB/4 Quad Full Diamond With One Loop Directional With Two Loops Figure 2. Expanded Interchange Configurations with Loop Ramps (Continued) 1.2 Research Objective and Scope The objective of this research is to develop improved design guidance for interchange loop ramps. The research focused on developing improved design guidance for the ramp proper portion of the ramp, taking into consideration the connections on either side of the ramp proper, namely the freeway mainline ramp terminal and the crossroad ramp terminal. The guidance provided considers the context of the interchange and the safety, operational, and constructability impacts of the design. The research was conducted in two phases. The scope of Phase I was to develop a list of key unresolved issues associated with loop ramp design. Based on information obtained through a literature review and informal survey of private consultants and state highway personnel, the research team identified and prioritized a number of key issues or gaps in knowledge to be investigated as part of this research. The key issues investigated during Phase II of the research included:

1 - 5  The relationship between speed and lane position of vehicles and key design elements of the ramp proper  The impact of key design elements on safety of loop ramps  The difference in performance on the ramp proper of single-lane and multi-lane loop ramps While the findings and recommendations of this research are generally applicable to loop ramps on both service interchanges (i.e., interchanges of a freeway with a surface street [arterial or collector]) and system interchanges (i.e., interchanges between two freeways), the ramps included in the observational and safety studies conducted within this research were located at service interchanges. This research focused on developing improved guidelines for loop ramps in both urban and rural areas. 1.3 Outline of Report This report presents an overview of research conducted to develop improved design guidance for interchange loop ramps and their connections. The remainder of this report is organized as follows: Section 2—Summary of Literature Review Section 3—Observational Field Study of Loop Ramp Proper Section 4—Crash-Based Safety Evaluation of Loop Ramps Section 5—Design Guidance Section 6—Conclusions and Recommendations for Future Research Section 7—References

1 - 6 Section 2. Summary of Literature Review This section summarizes the state of the art and state of the practice in loop ramp design based on a review of the literature including research studies, guidance documents, and design manuals. The section focuses on issues unique to loop ramps – in comparison to other types of ramps such as diagonal (i.e., straight) or direct ramps – which in most cases relate to their sharp horizontal alignment and associated vehicle speeds. This section is organized by the following topics:  Current AASHTO policy on loop ramps  ITE’s Freeway and Interchange Geometric Design Handbook  Highway Safety Manual  Operational issues  State DOT design manuals  Geometric design and safety 2.1 Current AASHTO Policy on Loop Ramps Design guidance for single-lane and multi-lane loop ramps in the Green Book (AASHTO, 2011) is limited to a few sections with brief remarks. Chapter 10 (Grade Separations and Interchanges) provides the most guidance on loop ramp design compared to other chapters of the Green Book, but other chapters provide guidance related to horizontal curve design which is directly applicable to loop ramps. The following discussion addresses geometric criteria/design considerations associated with loop ramp design. 2.1.1 Minimum Turning Paths of Design Vehicles (Green Book Section 2.1.2) Key controls in highway geometric design are the physical and operational characteristics and proportion of vehicles of various sizes using the highway. Chapter 2 of the Green Book presents characteristics for 20 design vehicles within four general classes of design vehicles: passenger cars, buses, trucks, and recreational vehicles. The dimensions and characteristics of the 20 design vehicles presented in the Green Book take into account recent trends in motor vehicle sizes manufactured in the United States and represent a composite of vehicles currently in operation. A key issue associated with design vehicles and loop ramp design relates to minimum turning paths. The boundaries of the turning paths of each design vehicle for its sharpest turns are established by the outer trace of the front overhang and the path of the inner rear wheel (known as offtracking). Green Book Table 2-2 presents the minimum turning radii and Green Book Figures 2-1 through 2-9 and 2-13 through 2-23 illustrate the minimum turning paths for 20 typical design vehicles.

1 - 7 2.1.2 Horizontal Alignment (Green Book Section 3.3) The objective in designing a horizontal curve is to balance the forces acting on a vehicle for safe and comfortable operation at speeds appropriate for the general conditions of the roadway. This is accomplished by considering design speed and selecting appropriate values for superelevation and side friction. The design radius of curvature is based on the design speed for the roadway and practical upper limits for superelevation and side friction. Some basic design elements of a horizontal curve include:  Superelevation  Side friction  Radius of curvature  Distribution of superelevation and side friction Superelevation- Superelevation is the provision of increased cross-slope (i.e., banking) on curves, to limit the portion of a vehicle’s lateral acceleration that must be resisted by tire- pavement friction. The practical limits of superelevation on horizontal curves are controlled by four factors: climate conditions, terrain conditions, adjacent land use, and frequency of slow- moving vehicles. For example, where snow and ice are a factor, the rate of superelevation should not exceed the rate at which vehicles standing or traveling slowly would slide toward the center of the curve when the pavement is icy. When traveling slowly around a curve with high superelevation, negative lateral forces develop and the vehicle is held in the proper path only when the driver steers up the slope or against the direction of the horizontal curve. Because joint factors should be considered when selecting a maximum superelevation rate (emax), there is no universally applicable rate for every region. Highway agencies typically establish their own policies concerning the maximum superelevation rate that will be used on horizontal curves within their jurisdiction. Maximum superelevation rates for horizontal curves typically range from 4 to 12 percent, based on corresponding side friction factors and design speeds. Side Friction Factor- The side friction factor represents a vehicle’s need for side friction and represents the lateral acceleration that acts on a vehicle. The upper limit of the side friction factor is the point at which the tire would begin to skid; this is known as the point of impending skid. The coefficient of friction at impending skid depends upon a number of factors, the most important ones being the speed of the vehicle, the type and condition of the roadway surface, and the type and condition of the vehicle tires. Because highway curves are designed so vehicles can avoid skidding with a margin of safety, the side friction factor used in design should be substantially less than the coefficient of friction at impending skid. The maximum side friction factor (fmax) values used in the design of horizontal curves are based upon the level of centripetal acceleration or lateral acceleration sufficient to cause drivers to experience a feeling of discomfort. In theory, when drivers begin to feel discomfort, they instinctively avoid higher speeds. At low speeds drivers are more tolerant of discomfort, permitting employment of an increased amount of side friction for use in design of horizontal curves. Thus, at low design speeds, higher design side friction factors are used, and as the design

1 - 8 speed increases, lower design side friction factors are specified. Maximum side friction factors range from 0.38 for a design speed of 10 mph to 0.08 for a design speed of 80 mph. Minimum Radius- The minimum radius is a limiting value of curvature for a given speed and is determined from the rate of superelevation and the maximum side friction factor selected for design. The minimum radius of curvature is based on a threshold of driver comfort sufficient to provide a margin of safety against skidding and vehicle rollover. Green Book Table 3-7 presents minimum curve radii for maximum superelevation rates ranging from 4 to 12 percent, based on corresponding side friction factors and design speeds. For design speeds of 25 to 35 mph, a typical range of design speeds for loop ramps, minimum curve radii range from 119 ft (for e = 12 percent) to 154 ft (for e = 4 percent) for a design speed of 25 mph and from 272 ft (for e = 12 percent) to 371 ft (for e = 4 percent) for a design speed of 35 mph. Distribution of Superelevation and Side Friction for Design- For a given design speed, the Green Book presents five methods for sustaining lateral acceleration on curves by use of superelevation or side friction, or both. The methods include:  Method 1: Superelevation and side friction are directly proportional to the inverse of the radius.  Method 2: Side friction is such that a vehicle traveling at the design speed has all lateral acceleration sustained by side friction on curves up to those designed for fmax. For sharper curves, the side friction remains equal to fmax and then the superelevation is used to sustain lateral acceleration until the superelevation reaches emax.  Method 3: Superelevation is such that a vehicle traveling at the design speed has all lateral acceleration sustained by superelevation on curves up to those designed for emax. For sharper curves, the superelevation remains equal to emax and then the side friction is used to sustain lateral acceleration until the side friction reaches fmax.  Method 4: This method is the same as Method 3 but is based on average running speed rather than design speed.  Method 5: Superelevation and side friction are in a curvilinear relation with the inverse of the radius of the curve, with values between those of Methods 1 and 3. Green Book Figure 3-7 illustrates the resulting relationships between the different methods of distributing superelevation and side friction. 2.1.3 Turning Roadways (Green Book Section 3.3.7) Turning roadways include interchange ramps (and intersection curves) for right-turning vehicles. The minimum radius for design should preferably be measured from the inner edge of the traveled way rather than the middle of the vehicle path or centerline of the traveled way. As much superelevation as practical, up to a maximum value, should be developed on ramps to counter skidding and overturning. Superelevation distribution Method 5 is recommended for determining the superelevation for turning roadways with radii greater than the minimum radius for the design speed and selected maximum superelevation rate.

1 - 9 Compound curves can create desirable turning roadway shapes for interchange ramps. When the design speed of the turning roadway is 45 mph or less, compound curvature can be used to form the entire alignment of the turning roadway. When the design speed exceeds 45 mph, the exclusive use of compound curves is often impractical. Compound curves that mislead motorists’ expectations of the sharpness of the curve are undesirable. To avoid such situations, it is preferred that the ratio of the flatter radius to the sharper radius not exceed 2:1. It is estimated that this radii ratio of 2:1 results in a reduction of approximately 6 mph in average running speeds for adjoining curves. Where practical, smaller differences in radii should be used. These guidelines assume travel is in the direction of sharper curvature. Where vehicles are accelerating, the maximum design ratios are not as critical and may be exceeded. 2.1.4 Transition Design Controls (Green Book Section 3.3.8) There are two types of transition sections: 1) a superelevation transition which transitions the roadway cross sections; and 2) the alignment transition which introduces a transition curve (spirals or compound curves) into the horizontal alignment. The desired superelevation rate on a loop ramp may be difficult to achieve due to the length available to transition the pavement cross-slope. 2.1.5 Offtracking (Green Book Section 3.3.9) Offtracking is a characteristic, common to all vehicles, although much more pronounced with large vehicles, in which the rear wheels do not precisely follow the same path as the front wheels when the vehicle traverses a horizontal curve. For loop ramps, typical radii will require additional lane width to accommodate offtracking. 2.1.6 Widths for Turning Roadways at Intersections (Green Book Section 3.3.11) The widths of turning roadways at intersections are governed by the types of vehicles to be accommodated, the radius of curvature, and the expected speed. Turning roadways are classified for operational purposes as follows:  Case I: One-lane, one-way operation with no provision for passing a stalled vehicle is usually appropriate for minor turning movements and moderate turning volumes where the roadway is relatively short.  Case II: One-lane, one-way operation with provision for passing a stalled vehicle is used to allow operation at low speed with sufficient clearance for vehicles to pass a stalled vehicle.  Case III: Two-lane operation, either one- or two-way, is applicable where operation is two way or where operation is one way, but two lanes are needed to accommodate the traffic volume.

1 - 10 Recommended roadway widths are further defined based upon traffic conditions as follows:  Traffic Condition A consists predominantly of passenger vehicles, but some consideration is also given to single-unit trucks (e.g., SU-30).  Traffic Condition B includes sufficient single-unit trucks (e.g., SU-30) to govern design, with some consideration given to tractor-semitrailer combination trucks.  Traffic Condition C includes sufficient tractor-semitrailer combination trucks to govern design. Green Book Table 3-29 provides recommended design widths for turning roadways. 2.1.7 Sight Distance on Horizontal Curves (Green Book Section 3.3.12) For the design of a horizontal curve, the sight line is that of the chord of the curve, and the sight distance is measured from the center of the lane to the center of the lane. Sight distance across the inside of the curve should not be restricted due to walls, cut slopes, or other obstructions. 2.1.8 Cloverleafs (Green Book Section 10.9.3) Cloverleafs are four-leg interchanges that employ loop ramps to accommodate left-turning movements. A loop ramp rarely operates with more than a single line of vehicles, regardless of the roadway width, and therefore, the cloverleaf capacity is limited by the loops. Loops may be made to operate with two lanes abreast, but only by careful attention to design of terminals and the design for weaving, which would need widening by at least two additional lanes at the separation structure. To accomplish this type of design, the terminals should be separated by such great distances and the loop radii be made so large that cloverleafs with two-lane loops generally are not economical from the standpoint of right-of-way, construction, costs, and the amount of out-of-direction travel. Loops that operate with two lanes of traffic are considered exceptional cases. 2.1.9 Ramps (Green Book Section 10.9.6) A loop ramp may have single turning movements (left or right) or double turning movements (left and right) at either or both ends. Ramp design speeds above 30 mph for loop ramps involve large land areas that are rarely available in urban areas. The long loop ramps needed for higher design speeds are costly and require turning vehicles to travel a considerable extra distance. Green Book Table 10-1 (Table 1) provides guidance for ramp design speeds as related to the highway design speed. Ramp design speeds are specified for upper, middle, and lower range values given various conditions and ramp types. Upper-range values of design speed generally are not attainable on loop ramps. Minimum values usually control, but for highway design speeds above 50 mph, the loop design speed preferably should be no less than 25 mph. If less restrictive conditions exist, the loop design speed and the radius may be increased. The lower

1 - 11 values in Green Book Table 10-1 for freeway design speeds of 70 mph require a loop ramp design speed of 35 mph. Values in Green Book Table 10-1 apply to the sharpest, or controlling, ramp curve, usually on the ramp proper. Table 1. Guide Values for Ramp Design Speed as Related to Highway Design Speed (AASHTO, 2011) U.S. Customary Highway design speed (mph) 30 35 40 45 50 55 60 65 70 75 Ramp design speed (mph) Upper range (85%) 25 30 35 40 45 48 50 55 60 65 Middle range (70%) 20 25 30 33 35 40 45 45 50 55 Lower range (50%) 15 18 20 23 25 28 30 30 35 40 Corresponding minimum radius (ft) See Green Book Table 3-7 Experience shows that the practical size of loop ramps resolves into approximate radii of 100 to 170 ft for minor movements on highways with design speeds of 50 mph or less and 170 to 250 ft for more important movements on highways with higher design speeds. In general, increasing the design speed by 5 mph increases the travel distance by 50 percent and right-of-way needs by 130 percent. Although considered exceptional cases, the need for two-lane loop ramps has increased. A two- lane loop ramp should not be preceded or followed by another loop ramp. The radius of the inner edge of traveled way of the loop ramp normally should not be less than 180 to 200 ft. The Green Book refers to ITE’s Freeway and Interchange Geometric Design Handbook for additional details on the design of two-lane loop ramps. Design guidelines for turning roadways at interchanges are discussed in Section 3.3.7 of the Green Book and are directly applicable to the design of ramp curves. Compound or spiral curves are desirable to: (1) obtain the desired alignment of ramps, (2) provide for a comfortable transition between design speeds of through and turning roadways, and (3) fit the natural paths of vehicles. Compound curves should be designed to prevent unexpected and abrupt speed changes. Ramp traveled way widths are governed by the type of operation, curvature, volume, and type of traffic. The roadway width for turning roadways includes the traveled way width plus the shoulder width. Where paved shoulders are provided on ramps, they should be of uniform width for the full length of the ramp. The left- and right-shoulder widths may be reversed if needed to provide sight distance on the inside of the curve. Ramps should have a lateral offset on the right side of the traveled way of at least 6 ft and on the left side of the traveled way of at least 4 ft. Section 3.3.11 on “Widths for Turning Roadways at Intersections” may be referenced for additional guidance on the treatments at the edge of the traveled way. Loop ramps located beyond a structure usually need a parallel deceleration lane prior to the ramp proper curve due to potential sight line restrictions. The speed-change lane should be developed on the near side of the structure and carried across the structure if sight distance is limited. Green Book Figure 10-68 (Figure 3) presents recommended minimum ramp terminal spacing for the various ramp-pair combinations as they are applicable to interchange classifications. The

recomme and shou When an by weavi suggested interchan radii and ramp is d Fi A speed- change in lanes are at a flat a 10-3 (Tab terminals Table 3) grades ar Book Tab ndations are ld be checke entrance ram ng consider in Green B ges, the dist roadway an esirable. gure 3. Rec change lane speed betw : (1) the tape ngle, while le 2) shows , and Green shows minim e present on le 10-4 (Ta based on o d in accorda p is follow ations, and t ook Figure ance betwee d median w ommended should have een the high r type and ( the parallel minimum l Book Table um deceler ramps, spee ble 4). perational ex nce with pr ed by an ex he spacing p 10-68 is not n entrance- idths. Also, Minimum sufficient l way and tu 2) the parall type has an engths of ac 10-5 ( ations lengt d-change le 1 - 12 perience an ocedures in it ramp, the olicy for en applicable t exit ramp no a recovery l Ramp Term ength to ena rning roadw el type. The added lane f celeration d hs for freew ngths shoul d the need f the Highwa absolute mi trance-exit r o cloverleaf ses is prima ane beyond inal Spaci ble a driver ay. Two gen taper type p or changing istances for ay mainline d be adjuste or flexibilit y Capacity M nimum dista amp combin loop ramps rily depend the nose of ng (AASHT to make the eral forms o rovides dire speed. Gre freeway ma exit termin d in accorda y and signin anual (HC nce is gover ations . For these ent on loop the loop exit O, 2011) appropriate f speed-cha ct entry or en Book Tab inline entran als. Where nce with Gr g M). ned ramp nge exit le ce een

1 - 13 Table 2. Minimum Acceleration Lengths for Entrance Terminals with Flat Grades of Two Percent or Less (AASHTO, 2011) Design speed (mph) Speed reached (mph) Acceleration length, L (ft) for entrance curve design speed (mph) Stop 15 20 25 30 35 40 45 50 Initial speed (mph) 0 14 18 22 26 30 36 40 44 30 35 40 45 50 55 60 65 70 75 23 27 31 35 39 43 47 50 53 55 180 280 360 560 720 960 1,200 1,410 1,620 1,790 140 220 300 490 660 900 1,140 1,350 1,560 1,730 – 160 270 440 610 810 1,100 1,310 1,520 1,630 – – 210 380 550 780 1,020 1,220 1,420 1,580 – – 120 280 450 670 910 1,120 1,350 1,510 – – – 160 350 550 800 1000 1230 1,420 – – – – 130 320 550 770 1,000 1,160 – – – – – 150 420 600 820 1,040 – – – – – – 180 370 580 780 Table 3. Minimum Deceleration Lengths for Exit Terminals with Flat Grades of Two Percent or Less (AASHTO, 2011) Highway design speed (mph) Speed reached (mph) Deceleration length, L (ft) for design speed of exit curve (mph) Stop 15 20 25 30 35 40 45 50 For average running speed on exit curve (mph) 0 14 18 22 26 30 36 40 44 30 35 40 45 50 55 60 65 70 75 28 32 36 40 44 48 52 55 58 61 235 280 320 385 435 480 530 570 615 660 200 250 295 350 405 455 500 540 590 635 170 210 265 325 385 440 480 520 570 620 140 185 235 295 355 410 460 500 550 600 – 150 185 250 315 380 430 470 520 575 – – 155 220 285 350 405 440 490 535 – – – – 225 285 350 390 440 490 – – – – 175 235 300 340 390 440 – – – – – – 240 280 340 390 Table 4. Speed-Change Lane Adjustment Factor as a Function of Grade (AASHTO, 2011) Deceleration Lanes Design speed of highway (mph) Ratio of length on grade to length on level for design speed of turning curve (mph)* All speeds 3 to 4% Upgrade0.9 3 to 4% Downgrade 1.2 All speeds 5 to 6% Upgrade0.8 5 to 6% Downgrade 1.35 Acceleration Lanes Design speed of highway (mph) Ratio of length on grade to length on level for design speed of turning curve (mph)* 20 30 40 50 All speeds 3 to 4% Upgrade 3 to 4% Downgrade 40 45 50 55 60 65 70 1.3 1.3 1.3 1.35 1.4 1.45 1.5 1.3 1.35 1.4 1.45 1.5 1.55 1.6 – – 1.4 1.45 1.5 1.6 1.7 – – – – 1.6 1.7 1.8 0.7 0.675 0.65 0.625 0.6 0.6 0.6 5 to 6% Upgrade 5 to 6% Downgrade 40 45 50 55 60 65 70 1.5 1.5 1.5 1.6 1.7 1.85 2.0 1.5 1.6 1.7 1.8 1.9 2.05 2.2 – – 1.9 2.05 2.2 2.4 2.6 – – – – 2.5 2.75 3.0 0.6 0.575 0.55 0.525 0.5 0.5 0.5 * Ratio in this table multiplied by length of acceleration/deceleration distances gives length of acceleration/deceleration distance on grade.

1 - 14 2.2 ITE Freeway and Interchange Geometric Design Handbook The following discussion details the design guidance most applicable to loop ramps in the ITE Freeway and Interchange Geometric Design Handbook (Leisch, 2006). The ITE Handbook can serve as supplemental guidance to the Green Book. 2.2.1 Design Speed Loops serve as the primary forms of left-turn ramps for both service and system interchanges. Loop ramp configurations, particularly regarding their spatial influence upon other ramps within the interchange, limit curve radii and ramp design speeds. Operational experience, including the aspect of extra travel distance, indicates that the practical radii for loop ramps generally range between 150 and 250 ft; approximately equivalent to design speeds of 25 to 30 mph. Table 5 presents differences in several operational measures for radii ranging from 100 to 250 ft. Table 5. Loop Ramp Characteristics (Leisch, 2006) Radius (ft) 100 150 250 Design speed (mph) 20 25 30 Travel distance (ft) 630 950 1,600 Travel time (sec) 22 26 37 On loop ramps limited to 25 mph, the traveled way is usually 18 ft wide. 2.2.2 Two-Lane Loop Ramps Loop ramps are typically single-lane facilities but can be adapted to accommodate multiple lanes (Leisch, 2006). To achieve effective two-lane operation on loop ramps, several features are essential:  A two-lane loop ramp should not be immediately preceded or followed by a loop ramp. Appropriate arrangements are loop ramps associated with directional interchanges with one loop or two loops in opposite alternative quadrants (e.g., Parclo-B and Parclo-A interchanges).  Continuation of the exit terminal should be direct and gradual, forming a stem road speed transition area 600 ft or more in length to align the vehicles in two lanes and then smoothly lead the traffic into the controlling curve of the loop.  The radius of the inner edge of the traveled way should be not less than 200 ft, except under restricted conditions.  The traveled way width should generally be no less than 30 ft (i.e., two 15-ft. lanes). For radii less than 200 ft, a width of 32 ft is appropriate. A typical design for a two-lane loop ramp is illustrated in Figure 4.

2.3 Hi In Nation (2012) de methodo volume e Method f models fo Maine. T database For both total segm developm Table Ramp typ C-D Road Exit Entrance FI crashes Figure ghway S al Cooperat veloped cra logies were dition (i.e., or Ramps) ( r freeways able 6 summ used to deve exit and ent ents and le ent were as 6. Summary Pred e Ramp configurat Segment Connector Diagonal Button hook Loop Connector Diagonal Button hook Loop All = Fatal and all i 4. Typical D afety M ive Highwa sh predictio recently inco Chapter 18 – AASHTO, 2 and intercha arizes the a lop the pred rance ramps ngth) and 19 sociated wit of Ramp C ictive Mode ion Tota numbe segme 20 15 34 5 11 17 32 3 11 1,53 njury crashes esign of a anual (H y Research P n methodolo rporated in Predictive 014). Bonn nges using mount of da ictive mode combined, percent of h loop ramp haracteris ls for Ram l r of nts Tota lengt (mi) 2 18 7 31 7 52 7 6 8 20 7 28 6 50 6 2 7 17 7 228 1 - 15 Two-Lane L SM) rogram (NC gies for fre to the HSM Method for eson et al. ( a sampling o ta by ramp ls for ramps approximate the crashes s. tics in NCH ps (based on l h Segmen range Min .1 0.02 .7 0.02 .9 0.02 .6 0.05 .0 0.02 .5 0.02 .1 0.02 .9 0.04 .9 0.01 .7 oop Ramp HRP) Proj eways and in as a supplem Freeways, a 2012) devel f data from configuratio and collect ly 15 to 16 included in RP 15-45 D Bonneson t length (mi) Max 0.21 2 0.76 0.38 0.41 0.42 0.65 0.42 0.14 0.38 (Leisch, 20 ect 17-45 Bo terchanges ent to the o nd Chapter oped the cra California, n included i or-distributo percent of th the database atabase Us et al., 2012 Volume range (veh/day) Min Ma ,200 29, 440 33, 140 13, 450 1, 160 12, 750 23, 140 13, 740 8, 530 15, 06) nneson et a . The riginal thre 19 – Predict sh predictio Washington n the final r (C-D) roa e sites (i.e., for model ed to Deve ) FI cras 5 yeax 500 15 000 27 200 13 500 6 400 13 800 21 300 10 800 1 300 8 1,17 l. e ive n , and ds. lop hes/ rs 9 1 3 9 7 0 2 1 6 8

1 - 16 The crash prediction models use a SPF in the following form to account for the effect of ramp traffic volume on safety: N = Lr × exp [a + b× ln(c AADTr) + d (c ×AADTr)] (1) where: N = crash frequency per year on the ramp Lr = ramp length (mi) AADTr = average annual daily traffic volume on the ramp (veh/day) a, b, c, d = regression coefficients The SPF uses different regression coefficients for one-lane and two-lane ramps, for fatal-and- injury (FI) and property-damage-only (PDO) crashes, and for multiple-and single-vehicle crashes. The crash modification factors (CMFs) developed for use with the SPFs account for the safety effects of the following factors on ramp segments:  Horizontal curvature  Lane width  Right shoulder width  Left shoulder width  Right side barrier  Left side barrier  Lane addition or drop  Ramp speed-change lane For horizontal curvature, the base condition is a tangent ramp proper, and the CMF value is a function of the radius of curvature, the average entry speed for the curve, and the proportion of the ramp proper with a curvilinear alignment. The CMF value increases (i.e., predicts an increase in crashes) as the radius of curvature decreases, the average entry speed increases, and the proportion of the ramp proper with a curvilinear alignment increases. The curve speed prediction model used with the horizontal curvature CMF is based on data collected at five interchange loop ramp curves and 20 rural two-lane highway curves collected by Bonneson (2000). The primary inputs into the ramp curve speed prediction model include:  Curve radius  Curve length (or segment length)  Average traffic speed on freeway during off-peak periods of the typical day (the default estimate is equal to the speed limit)  Average speed at point where ramp connects to crossroad (the default estimate is 15 mph for ramps with stop-, yield-, or signal-controlled crossroad ramp terminals and 30 mph for all other ramps at service interchanges) Section 3.4.1.2 provides more details related to the calculations of the ramp curve speed prediction procedure used in the crash prediction models for ramps.

1 - 17 Unlike previous studies, Bonneson et al. did not model loop ramps differently from other ramp configurations but developed a detailed methodology accounting for the effect of ramp curvature on crash frequency. ISATe is a spreadsheet based tool that implements the crash prediction models for ramps and the associated ramp curve speed prediction procedure. The tool can be used to evaluate the relationship between various design alternatives of ramps and the expected average crash frequency. The tool also outputs entry and exit speeds of curves estimated as part of the interim calculations for predicting crash frequencies. 2.4 Operational Issues for Loop Ramps Operationally, the ramp proper (i.e., turning roadway) and the ramp terminals do not operate independently from the roadways they connect. Thus, operating conditions on the intersecting roadways can affect operations on the ramp proper and at the ramp terminals, and vice versa. This subsection addresses operational analysis methodologies in the 2010 HCM related to ramps. The operational analysis methodologies in the HCM are not sensitive to ramp type (e.g., loop, diagonal); therefore, this summary focuses on operational issues for ramps in general. The HCM (TRB, 2010) does not specifically address analysis procedures for the ramp proper (i.e., turning roadway). For ramp capacity, a general rule of thumb is that a single-lane ramp (i.e., either an entrance or exit) can handle up to about 1,500 to 2,000 vph, and a two-lane ramp can handle up to 3,600 vph (Leisch, 2006). When considering the operational analysis of the mainline freeway ramp terminal for loop ramps, analysis procedures for weaving areas and independent ramps must be considered. Traffic in weaving areas experiences more lane-changing turbulence than is normally present on basic freeway segments. Chapter 12 of the HCM indicates that three geometric characteristics affect a weaving area’s operating characteristics (TRB, 2010):  Length—the distance between the merge and diverge that form the weaving area  Width—the number of lanes within the weaving area  Configuration—the way the entry and exit lanes are aligned The primary measures of performance for evaluating the operating conditions in a weaving area are density and speed (of both weaving and nonweaving vehicles). At freeway mainline ramp terminals outside of weaving areas (i.e., independent ramps), Chapter 13 of the HCM indicates that the primary geometric characteristics that affect merging and diverging operations include (TRB, 2010):  The length and type (parallel vs. taper) of acceleration or deceleration lane(s)  The free-flow speed of both the ramp and freeway in the vicinity of the ramp  The proximity to other ramps

1 - 18 Level of service criterion for freeway mainline ramp terminals (both entrances and exits) is defined in terms of density. 2.5 State Design Manuals on Loop Ramps Table 7 highlights state policies related to the design of loop ramps. It is based on a review of state design manuals. The table focuses on recommended design guidance associated with design speeds, curve radii, and pavement or traveled way widths for loop ramps. 2.6 Geometric Design and Safety Several studies have investigated the geometric design elements and the safety performance of freeway loop ramps. Relevant findings from the studies are summarized below. In addition, findings from a recent study related to superelevation criteria for sharp horizontal curves on steeps grades are summarized, as they are relevant to the topic of loop ramp design.  The speed differential between merging vehicles and mainline freeway vehicles is nearly the same for both diagonal and loop ramps (Torbic et al., 2012).  There is no substantive difference in operational performance between low-speed loop and high-speed diagonal entrance ramps under free-merge conditions (Torbic et al., 2012).  Deceleration rates of vehicles are greater at loop ramps than diagonal ramps (Torbic et al., 2012).  Drivers exiting loop ramps tend to reduce their speed in the freeway lane more, and decelerate along the speed-change lane at a greater rate, than drivers exiting on diagonal ramps (Torbic et al., 2012).  Comparing diagonal ramps, non-free-flow loop ramps, free-flow loop ramps, and outer connection ramps, results found more crashes occur on exit ramps than entrance ramps by a ratio of three to two (Lord and Bonneson, 2005).  Exit ramps generally have higher crash rates than entrance ramps, and diagonal ramps have lower crash rates than loop ramps. There appears to be an exception for urban, free- flow loop ramps, generally found in cloverleaf interchanges, which had lower crash rates than diagonal ramps for both exit and entrance ramps (Bauer and Harwood, 1998).  Exit ramps have higher crash rates than entrance ramps, and curved ramps have slightly higher crash rates than straight ramps (Lundy, 1965).  Curved ramps, such as loop ramps, have higher crash rates than straight ramps (Yates, 1970).  Replacing an outer connection exit ramp with a free-flow loop ramp and a parclo loop ramp increases crash counts by 68 and 38 percent, respectively (Liu et al., 2009).

1 - 19 Table 7. Summary of State Policies Related to Design Speed, Curve Radius, and Widths for Loop Ramps State Design speed (DS) Radius Width Arizona (ADOT, 2012) DSMin = 30 mph RMin = 230 ft 22-ft wide ramp for one-lane, one- way operation and normal level of truck traffic; increase width for ramps with large truck volumes California (Caltrans, 2012) Radii range from 150 to 200 ft Lane width = 18 ft for ramp radii less than 150 ft (measured along the outside edge of the traveled way); Lane width = 12 ft for ramp radii greater than 300 ft (measured along the outside edge of the traveled way) Connecticut (CT DOT, 2012) For mainline design speeds greater than 50 mph, loop design speed should not be less than 25 mph Radii range from 145 to 275 ft for design speeds of 25 to 30 mph Illinois (IDOT, 2012)  On collector-distributor roadways or in restricted urban conditions, DSMin = 25 mph  Where truck ADT is greater than 15 percent, use minimum design speed of 30 mph for initial curve after exit curve.  For rural loop ramps, 30 mph design speed is preferred.  Use design speed of 40 mph for cloverleaf interchange loop ramps between freeways Indiana (INDOT, 2012) For mainline design speeds greater than 50 mph, loop design speed should not be less than 20 mph. Normally, a loop ramp should not be design for a design speed greater than 35 mph Radii range from 180 to 250 ft For curve radii less than or equal to 300 ft, ramp width should be 28 ft Iowa (Iowa DOT, 2010) Radii range from 150 to 250 ft For radii less than 250 ft, pavement width should be 18 ft; otherwise, pavement width should be 16 ft. 4ft shoulders on left and 6 ft shoulders on right Maine (MaineDOT, 2004) Ramps design speeds range from 25 to 35 mph Massachusetts (MassDOT, 2006) RMin = 1,000 ft for two-lane ramp Michigan (MDOT, 2012) RMin = 260 ft Minnesota (MnDOT 2001) DSMin = 25 mph RMin = 190 ft; Minimum desirable radius is 230 ft (i.e., design speed of 27.5 mph) Pavement width = 28 ft in rural areas (includes 6 ft right shoulder and 4 ft left shoulder) In urban areas, lane width = 24 ft (measured face-of-curb to face-of- curb) for ramp radii between 125 to 160 ft; and lane width = 16 ft for ramp radius greater than 500 ft New Jersey (NJDOT, 2011) DSMin = 25 mph Oregon (ODOT, 2012) Ramps design speeds range from 25 to 35 mph South Dakota (SDDOT, 2012) Design speed of 30 mph is preferred. Texas (TxDOT, 2010) DSMin = 25 mph RMin = 185 ft Notes: DSMin = Minimum design speed RMin = Minimum curve radius

1 - 20  In general, there is no difference between crash rates for trucks and crash rates for all vehicles at either entrance or exit ramps. The exit ramp configurations where truck crash rates exceed the overall crash rates by the greatest amount include parclo loop, free-flow loop, and “other” ramp configurations (Torbic et al., 2012).  Loop ramps on freeway interchanges and direct freeway-to-freeway connections are sometimes subject to truck rollover problems (Seyfried and Pline, 2009).  Rural loop ramps with smaller radii have higher crash rates than rural loop ramps with larger radii, whereas the reverse is true for urban loop ramps (Twomey et al., 1993).  A two-lane highway curve has more crash risk than a ramp curve for any radius smaller than 2,000 ft (Bonneson et al., 2012).  The margins of safety against skidding and rollover are lowest for horizontal curves with low design speeds (40 mph or lower). This suggests that the difference between friction demanded by vehicles in relation to design friction factors (fmax) is lower at low design speeds than higher design speeds (Torbic et al., 2014). 2.7 Summary AASHTO policy and guidance in most state design manuals recommends a design speed in the range of 25 to 30 mph for a loop ramp in urban areas, and in less restrictive conditions (e.g., rural areas), the loop design speed may be increased. The design speed of the loop ramp directly affects the minimum acceleration lane length of an entrance ramp and the minimum deceleration lane length of an exit ramp. Table 2 and Table 3 show the corresponding minimum acceleration/deceleration lane lengths for entrance and exit ramps, respectively. Although the tables show the design speed of controlling curves, minimum acceleration/deceleration lane lengths are one of the few geometric design elements based upon operating speeds (e.g., average running speed on the exit curve) rather than design speeds. Thus, it is important to have accurate data and a better understanding of speeds of vehicles as they transition from the controlling curve of the ramp proper to the acceleration lane of an entrance ramp and as they transition from the deceleration lane to the controlling curve on the ramp proper of an exit ramp to properly design the freeway mainline ramp terminal in conjunction with the ramp proper of a loop ramp. It is also desirable to have a better understanding of the relationship between design speed and operating speeds on the ramp proper of a loop ramp. Much of the design guidance that is provided for loop ramps focuses on the lane widths, shoulder widths, and curve radii to accommodate vehicles as they traverse a loop ramp; however, there is not a good understanding of how key geometric features of the ramp proper such as curve radius, curve length, lane width, shoulder width, maximum superelevation, number of lanes, and the presence of compound curves and/or spirals impact the speeds of vehicles traversing the curves along the ramp proper. In the crash prediction models developed by Bonneson et al. (2012) for ramps, ramp curve speed prediction models are incorporated in the methodology, but the models are based upon data from 25 sites, only five of which were loop ramps. The other 20 sites were curves on rural two-lane

1 - 21 highways. In addition, the speed prediction models only incorporate a few key geometric features including curve length and radius and type of crossroad terminal, so a better understanding of how key geometric design features along the ramp proper of a loop ramp influence vehicle speeds is desirable. In NCHRP Project 17-45, Bonneson et al. (2012) developed a comprehensive methodology to predict crashes for the three primary components of ramps (i.e., freeway mainline ramp terminals, ramp proper, and crossroad ramp terminals). To predict crashes on the ramp proper, Bonneson et al. did not model loop ramps differently from other ramp configurations but developed a detailed methodology to account for the effect of ramp curvature on crash frequency. The methodological approach appears reasonable and valid; however, only approximately 15 percent of the 1,537 segments and 230 miles of ramps and only 19 percent of the 1,178 crashes in the database were from loop ramps. Ideally, the database assembled for modeling ramps would have included a higher proportion of loop ramps. Therefore, validation of the NCHRP Project 17-45 models to verify their appropriateness and accuracy as they relate specifically to loop ramps is important.

Secti Obse This sect the ramp designed ramps, na elements single-lan 3.1 St Observat 13 exit ra in urban perspecti included Figur on 3. rvation ion describe proper of si to address t mely: (a) th of the ramp e and multi udy Loc ional field s mps, in Cal and suburba ve of three r in the obser e 5. I-435/S al Field s an observa ngle-lane an wo of the hi e relationsh proper, and -lane loop r ations tudies were ifornia, Kan n areas, and amps (two e vational fiel hawnee Mi Study tional field d multi-lane gh-priority g ip between (b) the diffe amps. conducted a sas, Missou three of the ntrances an d study wer ssion Parkw S 1 - 22 of Lo study of dri loop ramps aps in know speed and la rence in pe t 28 loop ram ri, and Texa sites were i d one exit) i e located at ay. (Image ervice) op Ram ver behavior . The obser ledge relate ne position rformance o ps, includi s. Twenty-fi n rural areas ncluded in t service inter Credit: G p Prop and vehicle vational fiel d to the des of vehicles a n the ramp p ng 15 entran ve of the sit . Figure 5 s he study. Al changes. oogle Earth er operations d study was ign of loop nd key desi roper betw ce ramps an es were loca hows an aer l of the ram TM Mappin on gn een d ted ial ps g

1 - 23 Site characteristic data for the ramps were collected through a combination of reviewing plans, profile sheets, and aerial images, while other data were collected in the field or estimated based on available information. The following site characteristics were collected for the ramp proper:  Area type (urban, rural)  Type of ramp (entrance, exit)  Design speed, posted speed, and advisory speed of ramp  Curve radius  Curve length  Maximum superelevation  Tangent length  Cross-slope of outside shoulder  Number of lanes  Lane and shoulder widths  Presence of barriers  Presence of compound curves  Grade (upgrade / downgrade)  Presence of ramp meters For ramps that were not simple curves, the respective data elements were collected separately for each curve and/or tangent section along the ramp. All of the ramps connected directly to the freeway. None of them were connected to a collector-distributor (C-D) roadway. In addition, the following site characteristics were collected for the adjoining freeway and ramp connections: Freeway mainline ramp terminals: Crossroad ramp terminals:  Number of freeway lanes (in direction of interest)  Design speed and posted speed limit of freeway  Merge/diverge type (taper, parallel, weave, or lane add/drop)  Length of speed-change lane  Number of speed-change lanes  Type of terminal (4-leg Type A4, 4-leg Type B4, 3-leg Type A2, 3-leg Type B2)  Type of traffic control  Posted speed limit on crossroad  Presence of right-turn channelization Table 8 (entrance ramps) and Table 9 (exit ramps) present general characteristics of the study ramp locations and connecting roadways. Both design speeds and posted speed limits on the freeways ranged from 50 to 70 mph. The ramp design speed is associated with the sharpest curve (i.e., controlling curve) on the ramp proper. Ramp design speeds were estimated based upon an assumed maximum superelevation of 8 percent for Kansas, Missouri, and Texas ramps, 12 percent for California ramps, and Method 5 in the Green Book for distributing superelevation and side friction for sustaining lateral acceleration on curves. With Method 5, superelevation and side friction are distributed in a curvilinear relation with the inverse of the radius of the curve. Method 5 is recommended for the distribution of superelevation and side friction for all curves greater than the minimum radius on rural freeways, urban freeways, and high-speed urban streets, and it is also recommended for the cross-slope design of ramps. Based on Method 5, the ramp design speeds ranged from 15 to 25 mph. Curve warning signs with advisory speeds were posted on several of the ramps. The posted speed limits on the crossroads ranged from 30 to 65 mph. The ramps covered a range of freeway mainline ramp terminal types including parallel, taper, weave, and drop or add lane depending on whether the ramp was an entrance or exit ramp. The crossroad ramp terminals were classified based on terminology used in the HSM (Figure 6).

1 - 24 Recall that the Green Book states that for highway design speeds above 50 mph, the loop ramp design speed should preferably be no less than 25 mph. Of the 28 ramps included in the observational study, one loop ramp (Ramp ID 27) had a ramp design speed of 25 mph, while all the others had design speeds of 15 or 20 mph. In addition, four of the study locations had highway design speeds of 50 mph, while all the others had highway design speeds of 60 mph or higher. Thus, most of the ramps included in the observational speed study did not meet the preferred guidance for loop ramp design speeds as recommended in the Green Book. However, it should also be acknowledged that the loop ramp design speeds and freeway design speeds were not available on most of the plans that were reviewed as part of this research for the study locations, so the loop ramp design speeds and freeway design speeds in Table 8 and Table 9 represent the best estimated values for the respective measure given the information available. Table 10 (entrance ramps) and Table 11 (exit ramps) present the site characteristics of the study locations. The study ramps included a mixture of ramps with simple and compound curves and upgrades and downgrades. Most of the ramps had a single lane, but several of the entrance ramps were two-lane ramps, and one of the exit ramps was a three-lane ramp. The radii of the controlling curves (measured to the inside of the traveled way) ranged from 110 to 312 ft; the lengths of the controlling curves ranged from 220 to 1,434 ft; and the superelevations measured in the field ranged from 0.02 to 0.07. The average lane widths of the controlling curves ranged from 13.1 to 21.0 ft, while the average inside shoulder widths ranged from 1.0 to 12.8 ft, and the average outside shoulders with ranged from 1.0 to 7.5 ft. Overall, ramp length ranged from 461 to 1,477 ft. Appendix A includes aerial images of all the study ramps.

1 - 25 Table 8. General Characteristics of Study Locations (Entrance Ramps) Ramp ID State Nearest city Area type Freeway Crossroad Direction Freeway design speed (mph) Posted speed limit on freeway (mph) Ramp design speed (mph) Advisory speed for ramp (mph) Type of freeway mainline ramp terminal Posted speed limit on crossroad (mph) Type of crossroad terminal 1 MO Kansas City Urban I-435 US-24 NB 70 65 15 None Parallel 35 3-leg; A2 2 KS Kansas City Urban I-435 Shawnee Mission Parkway SB 70 70 20 25 Parallel 55 4-leg; A4 3 KS Kansas City Urban I-435 Shawnee Mission Parkway NB 70 70 20 25 Weave 55 4-leg; A4 4 MO Marshall Junction Rural I-70 MO 65 WB 70 70 15 25 Weave 65 4-leg; A4 5 TX Dallas Urban I-30 Hampton Rd WB 70 60 20 20 Weave 35 4-leg; A4 6 TX Dallas Urban SH 360 E. Mid Cities Blvd SB 70 60 15 None Add lane 40 4-leg; A4 7 TX Dallas Urban US-67 Loop 12 NB 60 60 15 20 Weave 45 4-leg; A4 8 TX Dallas Urban US-67 Loop 12 SB 60 60 15 20 Weave 45 4-leg; A4 9 TX Hillsboro Rural I-35 US-77/SH 579 NB 75 75 15 None Parallel 60 3-leg; B2 10 TX Dallas Urban I-35E Commerce St NB 50 60 15 None Weave 30 4-leg; A4 11 TX College Station Urban FM 2818 Wellborn Rd WB 50 50 20 25 Parallel 45 3-leg; A2 12 CA Sacramento Urban I-5 Arena Blvd SB 65 65 20 None Add lane 40 4-leg; A4 13 CA Sacramento Urban US-50 E. Bidwell EB 65 65 20 None Add lane 45 4-leg; A4 14 CA Sacramento Urban I-80 Elkhorn Rd WB 65 65 20 None Taper 45 4-leg; A4 15 CA Sacramento Urban I-80 Elkhorn Rd EB 65 65 20 None Taper 45 4-leg; A4

1 - 26 Table 9. General Characteristics of Study Locations (Exit Ramps) Ramp ID State Nearest city Area type Freeway Crossroad Direction Freeway design speed (mph) Posted speed limit on freeway (mph) Ramp design speed (mph) Advisory speed for ramp (mph) Type of freeway mainline ramp terminal Posted speed limit on crossroad (mph) Type of crossroad terminal 16 MO Kansas City Urban I-435 US 24 SB 70 65 20 25 Parallel 35 3-leg; B2 17 KS Kansas City Urban I-435 Shawnee Mission Parkway NB 70 70 20 25 Weave 55 4-leg; A4 18 MO Kansas City Urban I-70 US 40 WB 65 55 15 25 Parallel 35 3-leg; B2 19 MO Kansas City Urban I-70 US 40 EB 65 55 15 25 Parallel 35 3-leg; B2 20 KS Kansas City Urban I-435 95th St NB 70 70 20 None Drop lane 40 3-leg; B2 21 MO Marshall Junction Rural I-70 MO 65 WB 70 70 15 25 Weave 65 4-leg; A4 22 TX Dallas Urban I-35E Commerce St SB 50 60 20 20 Taper 30 3-leg; B2 23 TX Dallas Urban I-635 Freeport Pkwy WB 70 60 15 20 Taper 35 3-leg; B2 24 TX Dallas Urban I-635 Freeport Pkwy EB 70 60 15 20 Taper 35 3-leg; B2 25 TX Dallas Urban I-30 Hampton Rd EB 70 60 20 20 Weave 35 4-leg; A4 26 TX Dallas Urban SH 360 E. Mid Cities Blvd NB 70 60 15 30 Drop lane 40 4-leg; A4 27 TX College Station Urban FM 2818 Wellborn Rd EB 50 50 25 30 Parallel 45 3-leg; B2 28 CA Sacramento Urban SR 99 Sheldon Rd SB 65 65 20 30 Parallel 45 3-leg; B2

F igure 6. Crossroad Ramp Termin 1 - 27 al Types/C lassifications (AASHT O, 2014)

1 - 28 Table 10. Characteristics of Ramp Proper (Entrance Ramps) Ramp ID State Freeway Crossroad Type of curve Number of lanes Vertical profile Controlling Curve Overall length of curves on ramp (ft) Overall length of ramp (ft) Radius (ft) Length (ft) Super (%) Avg inside shoulder width (ft) Avg lane width (ft) Avg outside shoulder width (ft) 1 MO I-435 US 24 Comp 1 Up 150 416 0.05 2.5 20.5 4.0 577 1225 2 KS I-435 Shawnee Mission Parkway Comp 1 Down 250 1030 0.05 9.5 14.6 6.8 1318 1418 3 KS I-435 Shawnee Mission Parkway Comp 1 Down 250 1018 0.05 10.1 14.4 7.5 1306 1406 4 MO I-70 MO 65 Comp 1 Up 177 344 0.05 6.3 17.3 5.3 1432 1747 5 TX I-30 Hampton Rd Comp 1 Down 312 1434 0.04 6.0 14.5 4.0 1635 2007 6 TX SH 360 E. Mid Cities Blvd Simple 1 Down 230 940 0.04 6.5 21.0 4.0 940 1445 7 TX US-67 Loop 12 Comp 1 Down 150 333 0.04 1.0 16.0 1.0 910 910 8 TX US-67 Loop 12 Comp 1 Down 125 254 0.04 1.0 16.0 1.0 888 951 9 TX I-35 US-77/SH 579 Comp 1 Up 200 458 0.04 8.0 14.0 2.0 1206 1253 10 TX I-35 Commerce St Comp 1 Up 110 220 0.04 1.0 20.5 3.0 564 564 11 TX FM 2818 Wellborn Rd Simple 1 Up 200 748 0.06 6.0 14.5 2.5 748 911 12 CA I-5 Arena Blvd Comp 2 Down 186 377 0.07 8.0 13.1 3.75 537 787 13 CA US 50 E. Bidwell Simple 2 Down 154 698 0.07 9.5 13.3 3.5 698 1448 14 CA I-80 Elkhorn Rd Simple 2 Down 123 571 0.07 8.0 13.5 4.0 571 667 15 CA I-80 Elkhorn Rd Simple 2 Down 122 566 0.07 6.5 14.0 4.0 566 766

1 - 29 Table 11. Characteristics of Ramp Proper (Exit Ramps) Ramp ID State Freeway Crossroad Type of curve Number of lanes Vertical profile Controlling Curve Overall length of curves on ramp (ft) Overall length of ramp proper (ft) Radius (ft) Length (ft) Super (%) Avg inside shoulder width (ft) Avg lane width (ft) Avg outside shoulder width (ft) 16 MO I-435 US 24 Comp 1 Down 181 424 0.05 3.5 18.7 7.1 706 1384 17 KS I-435 Shawnee Mission Parkway Comp 1 Up 250 1048 0.05 10.3 15.0 6.8 1336 1436 18 MO I-70 US 40 Comp 1 Down 160 411 0.03 4.4 20.0 5.4 890 960 19 MO I-70 US 40 Comp 1 Down 160 447 0.05 4.2 20.9 4.3 822 892 20 KS I-435 95th St Comp 1 Up 300 861 0.05 12.8 15.2 6.6 1107 1307 21 MO I-70 MO 65 Comp 1 Down 218 809 0.04 6.0 17.2 6.25 1106 1263 22 TX I-35E Commerce St Comp 1 Down 104 260 0.05 1.0 20.0 4.0 458 575 23 TX I-635 Freeport Pkwy Simple 1 Up 200 461 0.02 7.0 17.0 2.0 461 546 24 TX I-635 Freeport Pkwy Simple 1 Up 200 461 0.02 7.0 17.0 3.0 461 551 25 TX I-30 Hampton Rd Comp 1 Up 246 970 0.05 6.0 14.0 2.0 1477 1962 26 TX SH 360 E. Mid Cities Blvd Simple 1 Up 230 1050 0.04 7.0 16.0 4.5 1050 1415 27 TX FM 2818 Wellborn Rd Simple 1 Down 250 601 0.06 6.0 14.5 2.0 601 763 28 CA SR 99 Sheldon Rd Simple 3 Up 171 557 0.07 8.5 14.2 3.5 557 1719

1 - 30 3.2 Field Data Collection Procedures The general field data collection procedures conducted along the ramp proper of freeway entrance and exit loop ramps are described below. For entrance loop ramps, the primary measures of interest collected in the field included:  Speeds of vehicles at the midpoint of the controlling curve on the ramp  Speeds of vehicles at the end [i.e., point of tangency (PT)] of the controlling curve in the direction of travel on the ramp  Positions of vehicles within the roadway at the midpoint of the controlling curve on the ramp For exit loop ramps, the primary measures of interest collected in the field included:  Speeds of vehicles at the beginning [i.e., point of curvature (PC)] of the controlling curve in the direction of travel on the ramp  Speeds of vehicles at the midpoint of the controlling curve on the ramp  Positions of vehicles within the roadway at the midpoint of the controlling curve on the ramp  Unusual or critical behavior (e.g., braking, swerving, or use of shoulder) near the beginning (PC) of the controlling curve in the direction of travel on the ramp Data collection activities were similar at both entrance and exit ramps. Field data were collected at study sites on weekdays during non-peak hours. Figure 7 illustrates the general data collection setup on the ramp proper for both entrance and exit loop ramps. Data were collected using laser guns and video recorders. Laser guns were used to collect speeds of subject vehicles on the ramps. Laser guns were hand-held and operated by a data collector on the roadside, located in a position chosen based on several criteria, including:  Safety of data collectors and equipment  Minimal impact of presence of data collectors and equipment on driver behavior or desired speeds  View of as much of the ramp as possible  Minimum angle between the laser gun and the vehicles being tracked Speeds of free-flow vehicles were collected. Two speeds for the same subject vehicle were recorded as the vehicle traversed the study ramps. For entrance loop ramps, the speeds of a subject vehicle were recorded using one or two laser guns at the midpoint of the controlling curve on the ramp and at the end (PT) of the controlling curve in the direction of travel on the ramp. For exit loop ramps, the speeds of a subject vehicle were recorded using one or two laser guns at the beginning (PC) of the controlling curve in the direction of travel on the ramp and at the midpoint of the controlling curve on the ramp. In situations where two laser guns were used,

the opera radio to t vehicle b when it c vehicle a for an ex passenge unit and t At the m inches ap A video r reference of the veh A video r decelerat investiga shoulder tor of the fir he operator eing tracked ame into vie t the midpoi it ramp of th r vehicles (i ractor-semi Figure 7. idpoint of th art in the tra ecorder was markers. Fr icle within ecorder was ion lane to t te and asses beyond the st laser gun of the secon . The opera w along the nt, point of e controllin ncluding lig trailer truck Sample Fie e controlling vel lane(s) positioned om the vide the lane or s also used t he ramp pro s unusual or deceleration would track d laser gun tor of the se ramp. The tangency (P g curve wer ht trucks an s). ld Setup fo curve at th and inside sh on the insid o display, th houlder cou o record man per of an ex critical driv lane) along 1 - 31 a vehicle a the descripti cond laser g operators wo T) for an ent e successful d sport-utilit r Observati e study ram oulder to d e of the curv e type of ve ld be determ euvers of v it loop ramp ing behavio the first sev long the ram on (make, m un would th uld then co rance ramp ly recorded. y vehicles) onal Study ps, referenc etermine the e and zoom hicle traver ined relativ ehicles as th . The prima r (e.g., brak eral hundre p and then odel, and c en track the nfirm that th , and point o Vehicles w or trucks (in of Ramp P e markings w lane positio ed to show sing the cur e to the insi ey transitio ry purpose w ing, swervin d feet of the communicat olor) of the subject veh e speeds of f curvature ere classifie cluding sing roper ere painted ns of vehic a close-up o ve and posit de lane line ned from the as to g, or use of ramp prope e by icle the (PC) d as le- 6 les. f the ion . the r as

1 - 32 this was assessed to be the most likely portion of the ramp where critical driving behavior would occur. Videos from the recorders were viewed in the office for post-processing to record lane position data and unusual or critical driving behavior. 3.3 Data Processing In addition to the site characteristics of the study locations, three types of observational data from the field studies were assembled in databases for analysis:  Speed data,  Lane position data, and  Exiting maneuver data. 3.3.1 Speed Data The database included the following information:  Ramp characteristics,  Vehicle type (passenger vehicle or truck),  Speed of subject vehicle at the midpoint and end (PT) of the controlling curve in the direction of travel (for entrance ramps),  Speed of subject vehicle at the beginning (PC) and midpoint of the controlling curve in the direction of travel (for exit ramps),  Travel lane [for a single-lane ramp, the lane was designated Lane 1; for a two-lane ramp, the inside (right) lane was designated Lane 1 and the outside (left) lane was designated Lane 2; for a three-lane ramp, the inside (right) lane was designated Lane 1, the middle lane was designated Lane 2, and the outside (left) lane was designated Lane 3)]. 3.3.2 Lane Position Data Video recordings were viewed in the office to document the positions of vehicles in the travel lane(s) or on the shoulder at the midpoint of the controlling curve on the ramp proper. Reference markings were painted across the travel lane(s) and shoulder. Figure 8 provides several screen captures of video used to record lane position data for one of the study ramps. Each painted reference marking was 6 inches apart, measured from the inside lane line separating the travel lane from the inside (right) shoulder. The reference markings extended to the middle of the travel lane and across the entire inside shoulder. Based on how many of the reference markings could be seen when a vehicle passed through, the lane position of the right

1 - 33 (passenger side) tire that was closest to the inside lane line (when the vehicle was completely positioned within the travel lane) or the position of the right tire that was farther from the inside lane line when one or more tires encroached onto the inside shoulder was recorded in increments of 6 inches. For example, in the situation where none of the right tires of the vehicle encroached onto the shoulder (i.e., the vehicle was completely within the travel lane of the ramp), then,  if none of the reference markings was visible when the tire closest to the lane line passed the position of the references markings, the distance of the closest tire to the inside lane line was (+) 0 to 6 in;  if one reference marking was visible when the tire closest to the lane line passed the position of the references markings, the distance of the closest tire to the inside lane line was (+) 7 to 12 in;  if two reference markings were visible when the tire closest to the lane line passed the position of the references markings, the distance of the closest tire to the inside lane line was (+) 13 to 18 in; and so on. When at least one tire encroached onto the shoulder, the corresponding distance of the tire farther from the inside lane line was given a negative (-) value. For example,  if the front right tire was in the travel lane (i.e., did not encroach onto the shoulder) and the rear right tire encroached onto the shoulder, but all reference markings on the shoulder were visible, the farther distance of the tire to the inside lane line that encroached onto the shoulder was recorded as (-) 0 to 6 in;  if the first reference marker on the shoulder closest to inside lane line was not visible due to the right vehicle tires but the second reference marker on the shoulder was visible, the farther distance of the tire to the inside lane line that encroached onto the shoulder was recorded as (-) 7 to 12 in; and so on. At multi-lane ramps, the positions of vehicles in the inside lane (Lane 1) were all measured relative to the inside lane line separating the travel lane and the inside shoulder. The positions of vehicles in the outside lane (Lane 2) of a two-lane ramp were all measured relative to the lane line separating the outside lane (Lane 2) and the inside lane (Lane 1). The positions of vehicles in the outside lane (Lane 3) of a three-lane ramp were all measured relative to the lane line separating the outside lane (Lane 3) and the middle lane (Lane 2). Care was taken to only record the lane positions of vehicles that maintained the same lane through the multi-lane ramp. In other words, lane positions of vehicles that changed lanes near the midpoint of the controlling curve were not recorded.

F igure 8. Sample Images from Vide Single 1 - 34 o to Docum -Lane Ram ent Vehicl p e Lane Position on a

1 - 35 3.3.3 Exiting Maneuver Data Video recordings were viewed in the office to document unusual or critical behaviors (e.g., swerving, severe braking) of exiting vehicles that occurred along the first several hundred feet of the ramp proper of the exit loop ramp. For each vehicle documented, the following information was collected from the video: Vehicle type: The vehicle type was recorded as either a passenger vehicle or a truck. Vehicle encroachment: The position of the vehicle was documented as either having encroached onto the inside shoulder or outside shoulder or stayed within the travel lane. Unusual or critical maneuver: Any type of unusual or critical maneuver such as swerving or severe braking was documented. If no unusual or critical maneuver was observed, this was also documented. If a vehicle encroached onto the shoulder but did not display any unusual or critical behavior, this was documented as having not seen any type of unusual or critical maneuver. Time: Time of exiting maneuver. Figure 9 provides several screen captures of video used to record unusual or critical maneuvers at one of the exit study ramps. In the figure, the subject driver makes a last-second decision to exit and nearly misses the ramp. The driver of the minivan applies the brakes in the freeway lane and turns sharply into the painted gore area to be able to complete the maneuver onto the exit ramp. For this sample maneuver, the vehicle was recorded as having swerved and encroached onto the outside shoulder. For each exit ramp during the time for which video data were post-processed, the total number of vehicles and the total number of trucks exiting the ramp was documented and the approximate time during which the counts were made. As such, the percentage of exiting vehicles that encroached onto the shoulders and/or performed unusual/critical maneuvers along the beginning of the ramp proper could be determined.

Figure 9. Sample Images from Video on E 1 - 36 to Documen xit Ramps t Unusual or Critical Maneuvers

1 - 37 3.4 Analysis of Observational Data This section describes the analysis approach and presents descriptive statistics and the analysis results. Three types of observational data from the field studies were analyzed:  Vehicle speed  Lane position  Exiting maneuver 3.4.1 Vehicle Speed The objective of this analysis was to develop speed prediction models to assess the effects of key design elements on vehicle speeds on loop ramps. The data collection procedure resulted in two speed measurements for each subject vehicle along a ramp. The speeds were denoted as follows:  Entrance ramps - Speed 1: speed of subject vehicle at midpoint of controlling curve - Speed 2: speed of subject vehicle at PT of controlling curve in the direction of travel  Exit ramps - Speed 1: speed of subject vehicle at PC of controlling curve in the direction of travel - Speed 2: speed of subject vehicle at midpoint of controlling curve Initially, analyses were conducted to estimate the effects of key design elements on vehicle speeds for single-lane entrance and exit ramps separately. In addition, analyses were conducted to determine the impact of key design elements on Speed 1 and Speed 2, separately, and the difference between Speed 2 and Speed 1 (Speed 2 – Speed 1). Then analyses were conducted on the full dataset which included both single-lane and multi-lane ramps, but separate analyses were still conducted for entrance and exit ramps. The key design elements considered in the analyses included:  Ramp length  Design speed of controlling curve  Radius of controlling curve  Superelevation of controlling curve  Length of controlling curve  Inside shoulder width of controlling curve  Lane width of controlling curve  Outside shoulder width of controlling curve  Freeway speed limit  Crossroad speed limit  Vertical profile (grade, up or down)  Type of speed-change lane  Type of horizontal curvature (simple or compound curve)  Type of crossroad traffic control  Vehicle type Regression models were developed using the restricted maximum likelihood (REML) method to estimate the effect of the key design elements on vehicle speeds. All analyses were performed using the JMP statistical software package from SAS. Given the large number of elements to consider, it was not possible to include all of the elements in the models simultaneously because

1 - 38 of issues of collinearity and confounding. Inclusion of nonrelevant or highly correlated independent variables may render other variables in the model statistically non-significant or lead to counterintuitive signs for some of the estimated model coefficients. Only the most relevant models and findings are presented later in Section 3.4.1.2 (Analysis Results). The speeds measured in the field were also compared to estimated speeds output from the ISATe model. The site characteristics of the study sites were input into ISATe in such a manner that the ISATe spreadsheet tool provided estimated speeds at the midpoint and end (i.e., PT) of the controlling curve for each entrance ramp and the beginning (i.e., PC) and midpoint of the controlling curve for each exit ramp. A comparison of the average speeds measured in the field to the estimated speeds output from the ISATe spreadsheet tool provides an assessment of the accuracy and reliability of the speed prediction procedure incorporated into ISATe. In addition, the site characteristics of the study sites were also input back into the speed prediction models developed to assess the effects of key design elements on loop ramp speeds. The resulting predicted speeds provide another measure to compare the accuracy and reliability of the speed prediction models developed as part of this research to the accuracy and reliability of the speed prediction procedures incorporated in ISATe. 3.4.1.1 Descriptive Statistics Table 12 and Table 13 provide a summary of the speeds recorded on entrance ramps and exit ramps, respectively. In Table 12, the speed data are provided separately for passenger vehicles and trucks, by location along the controlling curve (i.e., midpoint and PT of the controlling curve). For each curve and vehicle type, the number of observations is provided along with the minimum, 50th percentile, 85th percentile, 95th percentile, and maximum speeds. The number of observations for Speeds 1 and 2 were identical (1,535 passenger vehicles and 252 trucks). On two of the four multi-lane ramps (Ramp IDs 12 and 15), speed data were recorded for only one passenger vehicle (and no trucks) in Lane 2. This can be attributed to the fact that few vehicles utilized the outside lane during non-peak hours, and no speed data were collected while ramp meters were operating. With only a few exceptions, vehicles typically traveled at higher speeds at the end (PT) of the controlling curve, as vehicles transitioned from the controlling curve to the acceleration lane, than at the midpoint of the controlling curve. Table 13 provides the corresponding summary of the speed data for exit ramps where data were recorded for 1,433 passenger vehicles and 146 trucks. Speed data for two trucks were recorded on the multi-lane ramp. In most cases vehicle speeds were higher at the beginning of the controlling curve, as vehicles transitioned from the deceleration lane to the ramp proper, than at the midpoint of the controlling curve. The last two columns of Table 12 and Table 13 provide two estimated design speeds of the controlling curves for the given ramps. The first estimate is based on the distribution of superelevation and side friction using Method 5 of the Green Book, while the second estimate is based on Method 2 from the Green Book. In accordance with Method 5, superelevation and side friction are distributed in a curvilinear relation with the inverse of the radius of the curve. Figure 3-8 through Figure 3-13 of the Green Book provide design superelevation rates and curve radii

1 - 39 for maximum superelevation rates ranging from 4 to 12 percent based on Method 5. Corresponding Green Book Tables 3-8 to 3-12 show minimum values of curve radii for various combinations of superelevation and design speeds for maximum superelevation rates ranging from 4 to 12 percent for a full range of common design conditions. Method 5 is recommended for determining the superelevation for turning roadways with radii greater than the minimum radius for the design speed and selected maximum superelevation rate. In accordance with Method 2 curves are designed such that all lateral acceleration is first sustained by side friction up to those designed for fmax, and for sharper curves the side friction remains equal to fmax and then the superelevation is used to sustain lateral acceleration until the superelevation reaches emax. Two points are worth noting about the estimated design speeds:  None of the study locations were designed using the minimum radius for the controlling curve assuming a maximum design superelevation (emax) of either 8 or 12 percent. As a result, the two methods for distributing superelevation and side friction provide different values for the estimated design speeds of the controlling curves. In all cases, the estimated design speeds using Method 5 are 10 to 15 mph lower than the estimated design speeds using Method 2.  At all of the study locations, the observed speeds are higher than the estimated design speeds using Method 5 but are much closer to the estimated design speeds using Method 2. Table 14 lists basic summary statistics for the dependent variables--Speed 1, Speed 2, and their difference; site characteristic variables (continuous and categorical); and vehicle-related variables considered in the analyses. Unless specified otherwise, the summary statistics for entrance ramps are based upon data for 1,787 vehicles, and those for exit ramps are based upon data for 1,579 vehicles.

1 - 40 Table 12. Descriptive Statistics of Vehicle Speeds on Ramp Proper (Entrance Ramps) Ramp ID Lane Speed 1 (midpoint of controlling curve) (mph) Speed 2 (PT of controlling curve) (mph) Estimated design speed (mph) Passenger vehicles Trucks Passenger vehicles Trucks Number of Obs Min Percentiles Max Number of Obs Min Percentiles Max Min Percentiles Max Min Percentiles Max Method 1 50th 85th 95th 50th 85th 95th 50th 85th 95th 50th 85th 95th 5 2 1 Lane 1 132 19 26 28 30 32 25 19 22 25 27 30 22 29 31 33 36 21 26 29 32.2 35 15 25 2 Lane 1 101 30 36 38 40 46 2 29 29 29 29 29 32 39 42 43 47 31 32 32 32 32 20 30 3 Lane 1 99 27 36 38 41 45 5 17 32 33 34 34 22 38 41 42 46 22 35 36 37 38 20 30 4 Lane 1 100 23 30 33 35 38 14 21 25 27 28 30 20 31 34 37 41 22 26 28 30 33 15 30 5 Lane 1 103 25 35 38 40 41 20 25 32 36 37 38 24 36 39 42 43 25 34 38 39 42 20 35 6 Lane 1 96 20 31 33 35 37 28 23 26 29 30 33 24 33 38 40 47 23 29 32 34 37 15 30 7 Lane 1 98 12 20 22 23 28 27 10 16 18 19 23 17 24 27 28 37 14 18 21 23 26 15 25 8 Lane 1 90 14 19 22 22 25 10 14 17 21 22 22 18 26 29 31 35 20 24 28 30 30 15 25 9 Lane 1 88 15 29 31 32 35 24 16 23 26 29 31 13 32 36 38 41 19 27 31 34 34 15 30 10 Lane 1 99 18 25 27 29 34 20 14 20 24 25 25 12 24 27 30 36 11 19 25 28 30 15 25 11 Lane 1 101 18 26 29 31 33 19 14 21 25 26 26 21 30 34 35 41 15 27 31 34 35 20 30 12 Lane 1 106 24 30 33 35 42 18 18 24 29 32 34 23 30 34 36 43 18 27 30 31 33 20 30 Lane 2 1 24 24 24 24 24 0 N/A N/A N/A N/A N/A 27 27 27 27 27 N/A N/A N/A N/A N/A 20 30 Lane 1&2 combined 107 24 30 33 35 42 18 18 24 29 32 34 23 30 34 36 43 18 27 30 31 33 20 30 13 Lane 1 62 26 30 32 33 42 11 20 24 27 30 33 23 29 32 32 40 21 24 27 30 32 20 30 Lane 2 52 26 31.5 34 36 40 2 17 22 25 26 26 26 31 34 36 40 23 24 25 25 25 20 30 Lane 1&2 combined 114 26 31 33 35 42 13 17 24 26 29 33 23 30 33 34 40 21 24 26 30 32 20 30 14 Lane 1 59 21 28 30 32 35 10 20 23 25 25 25 20 28 31 32 35 15 23 25 27 28 20 30 Lane 2 51 22 29 32 34 36 3 17 23 28 29 30 22 29 32 34 38 20 23 28 29 30 20 30 Lane 1&2 combined 110 21 28 31 33 36 13 17 23 25 27 30 20 28 31 34 38 15 23 26 29 30 20 30 15 Lane 1 97 21 25 28 30 34 14 17 22 23 26 26 17 27 30 32 35 18 25 27 29 29 20 30 Lane 2 0 N/A N/A N/A N/A N/A 0 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 30 Lane 1&2 combined 97 21 25 28 30 34 14 17 22 23 26 26 17 27 30 32 35 18 25 27 29 29 20 30 Refer to Table 8 for Corresponding Ramp ID Location Information 1 Estimated design speeds using Method 5 and Method 2 of the Green Book to distribute superelevation and side friction.

1 - 41 Table 13. Descriptive Statistics of Vehicle Speeds on Ramp Proper (Exit Ramps) Ramp ID Lane Speed 1 (PC of controlling curve) (mph) Speed 2 (midpoint of controlling curve) (mph) Estimated design speed (mph) Passenger vehicles Trucks Passenger vehicles Trucks Number of Obs Min Percentiles Max Number of Obs Min Percentiles Max Min Percentiles Max Min Percentiles Max Method 1 50th 85th 95th 50th 85th 95th 50th 85th 95th 50th 85th 95th 5 2 16 Lane 1 100 25 33 36 37 38 26 21 25 28 30 32 23 30 33 35 36 18 24 27 28 31 20 30 17 Lane 1 100 30 40 43 46 69 2 32 33 33 33 33 24 34 37 39 41 27 28 28 28 28 20 35 18 Lane 1 100 22 29 31 33 36 5 20 22 26 27 27 20 27 29 30 32 19 20 22 22 22 15 25 19 Lane 1 100 22 28 30 31 34 8 15 23 25 26 26 18 28 31 33 33 19 23 25 26 27 15 30 20 Lane 1 100 30 41 45 48 50 1 41 41 41 41 41 27 35 39 40 47 36 36 36 36 36 20 35 21 Lane 1 99 22 34 37 38 42 18 21 29 33 34 35 25 33 35 36 38 22 27 30 30 30 15 30 22 Lane 1 95 19 25 28 31 32 7 19 23 27 27 27 12 18 20 21 23 14 15 16 17 18 15 25 23 Lane 1 106 26 42 47 50 53 19 24 31 36 42 46 12 26 28 30.8 33 16 20 23 25 26 15 30 24 Lane 1 105 26 40 43 45 52 20 23 35 38 40 41 21 28 31 33 35 17 22 28 30 33 15 30 25 Lane 1 102 32 44 51 54 60 20 21 40 47 49 49 20 29 33 34 43 16 25 30 31 31 20 30 26 Lane 1 108 30 38 43 46 49 17 28 36 43 44 47 24 32 35 37 40 24 28 35 35 37 15 30 27 Lane 1 49 25 37 42 44 45 1 30 30 30 30 30 20 31 34 37 41 26 26 26 26 26 25 35 28 Lane 1 95 25 37 41 42 46 1 33 33 33 33 33 25 31 33 35 38 26 26 26 26 26 20 30 Lane 2 70 30 38 42 46 49 1 26 26 26 26 26 24 32 35 37 40 25 25 25 25 25 20 30 Lane 3 104 31 38 43 44 47 0 N/A N/A N/A N/A N/A 26 32 35 38 39 N/A N/A N/A N/A N/A 20 30 Lane 1&2&3 Combined 269 25 38 42 44 49 2 26 30 32 33 33 24 32 35 37 40 25 26 26 26 26 20 30 Refer to Table 9 for Corresponding Ramp ID Location Information 1 Estimated design speeds using Method 5 and Method 2 of the Green Book to distribute superelevation and side friction.

1 - 42 Table 14. Summary Statistics for Speed Data  Entrance ramps Exit ramps Variable Min Max Avg Std Dev Min Max Avg Std Dev Speed 1 10 46 27.60 5.85 15 69 35.78 6.98 Speed 2 11 47 29.69 5.84 12 47 29.06 5.28 Speed 2- Speed 1 -12 15 2.09 3.49 -34 11 -6.82 5.92 Site characteristics (continuous variables) FwySpLimit freeway speed limit, mph 50 75 65 6 50 70 62 5 RampLength length of ramp, ft 564 2007 1167 402 546 1962 1212 459 CrSpLim crossroad speed limit, mph 30 65 45 9 30 65 41 9 CCDesSpd design speed of controlling curve, mph 15 20 16 2 15 25 18 3 CCRadius radius of controlling curve, ft 110 312 182 54 104 300 200 46 CCLength length of controlling curve, ft 220 1433 622 334 260 1050 637 251 CC_ISW inside shoulder width of controlling curve, ft 1.00 10.08 5.90 3.04 1.00 12.75 6.62 2.74 CC_LW lane width of controlling curve, ft (Lane 1)1 14 21 16.45 2.51 14 20.92 17.02 2.19 CC_LW lane width of controlling curve, ft (Lane 2)2 11.75 12.00 12.00 0.02 14 14 14.00 0.00 CC_LW lane width of controlling curve, ft (Lane 3)3 14 14 14.00 0.00 CC_OSW outside shoulder width of controlling curve, ft 1.00 7.50 3.72 1.67 2.00 7.08 4.37 1.72 CCSuper superelevation of controlling curve, ft/ft 0.04 0.07 0.05 0.01 0.02 0.07 0.05 0.01 Site characteristics (categorical variables) Categories (proportion) in the data Grade Up (0.35), Down (0.65) Up (0.61), Down (0.39) SCL Type type of speed-change lane Weave (0.38), Parallel (0.28), Add lane (0.21), Taper (0.13) Parallel (0.42), Taper (0.22), Weave (0.22), Drop lane (0.14) RampRadiusType simple or compound curve Simple (0.34), Compound (0.66) Simple (0.44), Compound (0.56) CrTrafCon crossroad traffic control Signal (0.16), Free-flow (0.84) Stop control (0.13), Signal (0.51), Free- flow (0.36) Vehicle-related (categorical variables) VehType vehicle type Car (0.86), Truck (0.14) Car (0.91), Truck (0.09) 1 Summary statistics for lane width of controlling curve (Lane 1) are based on 1,678 vehicles on entrance ramps and 1,404 vehicles on exit ramps. 2 Summary statistics for lane width of controlling curve (Lane 2) are based on 109 vehicles on entrance ramps and 71 vehicles on exit ramps. 3 Summary statistics for lane width of controlling curve (Lane 3) are based on 104 vehicles on exit ramps.

3.4.1.2 This sect elements entrance further ex speeds ba the study estimated Effect of Before de speeds, it which va correlatio F Entrance After rev goodness the ramp data for s normal” Analysis R ion first pre of loop ram and exit ram plained. Fin sed on regr locations to and predict Key Design veloping re is importan riables shou n among va igure 10. C Ramps iewing a nu of fit for es proper of an ingle-lane a in this mode esults sents the ana ps on vehicl ps. Next, th ally, compa essions mod gain a bette ed speeds. Elements gression mo t to first ass ld be includ riables whe orrelations mber of can timating the entrance lo nd multi-lan l. The only lysis results e speeds. Th e speed pred risons are m els develope r understan on Vehicle dels to estim ess the corre ed in the mo re the size o among Va didate mode speed of ve op ramp is e ramps com mild instabi 1 - 43 intended to e analysis r iction proce ade of the s d as part of ding of the a Speeds ate the effe lation amon dels. Figure f the circles riables in th ls, the mode hicles at the shown in Ta bined. The lity in the m identify the esults are pr dures incor peeds outpu this researc ccuracy and ct of key de g variables 10 is a grap is proportio e Speed Da l yielding th midpoint o ble 15. This within-grou odel’s predi effect of ke esented sep porated into t by ISATe h, and meas reliability sign elemen in the datas hical repres nal to the co tabase for e most info f the control model is de p residuals ctors is the m y design arately for ISATe are , the predict ured speeds of the differ ts on vehicl et to identify entation of rrelation. All Sites rmation and ling curve o veloped usi are “textboo oderate ed at ent e the best n ng k

1 - 44 correlations of the coefficient for lane width (CC_LW) and two other predictors. However, these are not substantial, and there are no other major concerns when running basic diagnostics. Table 15. Speed Prediction Model to Estimate Speed at the Midpoint of the Controlling Curve on the Ramp Proper of an Entrance Loop Ramp This speed prediction model can be interpreted as follows. Vehicle speeds at the midpoint of the controlling curve on an entrance loop ramp are expected to:  Increase by approximately 2 mph by incrementally increasing the radius (measured to the inside of the traveled way) of the controlling curve by 50 ft, everything else held constant.  Increase by approximately 0.3 mph for every 1-ft incremental increase of the lane width of the controlling curve, everything else held constant.  Increase by approximately 0.9 mph for every 1-ft incremental increase of the outside shoulder width of the controlling curve, everything else held constant.  Increase by approximately 0.7 mph for every 1-ft incremental increase of the inside shoulder width of the controlling curve, everything else held constant. In addition, vehicle speeds at the midpoint of the controlling curve on an entrance loop ramp in Lane 2 (outside lane) on a multi-lane loop ramp are estimated to be approximately 2 mph faster than speeds in Lane 1 (inside lane) of the same ramp, and truck speeds at the midpoint of the controlling curve on an entrance loop ramp are expected to be approximately 4.3 mph slower than speeds of passenger vehicles. Starting initially with the same variables that were significant in predicting the speeds at the midpoint of the controlling curve, a separate model was estimated to predict the speeds at the end (PT) of the controlling curve. In other words, Speed 1 was the dependent variable for the model to predict speeds at the midpoint of the controlling curve, and Speed 2 was the dependent variable for the model to predict speeds at the end of the controlling curve. Table 16 shows the Linear mixed‐effects model fit by maximum likelihood            AIC      BIC    logLik    9025.673 9075.068 ‐4503.837    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:     1.21075 2.970354    Fixed effects: Speed1 ~ LaneFactor + CCRadius + CC_LW + CC_OSW + CC_ISW + VehType                           Value Std.Error   DF    t‐value p‐value  Significance  (Intercept)        8.359490  3.654970 1769   2.287157  0.0223  Significant at 95% CL  LaneFactor(Lane2)  1.978092  0.633945 1769   3.120292  0.0018  Significant at 99% CL  CCRadius           0.039645  0.007180   11   5.521428  0.0002  Significant at 99% CL  CC_LW              0.313094  0.186978 1769   1.674496  0.0942  Significant at 90% CL  CC_OSW             0.912132  0.291748   11   3.126443  0.0096  Significant at 99% CL  CC_ISW             0.681708  0.190667   11   3.575383  0.0044  Significant at 99% CL  VehType(truck)    ‐4.332671  0.205156 1769 ‐21.118901  0.0000  Significant at 99% CL 

1 - 45 resulting model using data for single-lane and multi-lane ramps combined. Interestingly, only the radius (CCRadius) and the outside shoulder width of the controlling curve (CC_OSW), and vehicle type [VehType(Truck)] remained statistically significant with similar coefficients as in the previous model. This suggests that their influence is roughly unchanged between the two points where Speed 1 and Speed 2 were measured. The lane width (CC_LW) and inside shoulder width of the controlling curve (CC_ISW) are not significant predictors of speed at the end of the controlling curve as in the previous model to predict speeds at the midpoint of the controlling curve (Table 15). Table 16. Speed Prediction Model to Estimate Speed at the End of the Controlling Curve on the Ramp Proper of an Entrance Loop Ramp Exit Ramps The same process was repeated for exit ramps, beginning with developing a speed prediction model to estimate the speed of vehicles at the midpoint of the controlling curve on the ramp proper of an exit loop ramp, resulting in the model shown in Table 17 using data for single-lane and multi-lane ramps combined. The results from Table 17 show that radius, outside shoulder width, vehicle type, type of curvature (simple or compound), and type of speed-change lane are all significant factors that impact vehicle speeds. This speed prediction model can be interpreted as follows. Vehicle speeds at the midpoint of the controlling curve on an exit loop ramp are expected to:  Increase by approximately 2.7 mph by incrementally increasing the radius (measured to the inside of the traveled way) of the controlling curve by 50 ft, everything else held constant.  Increase by approximately 1 mph for every 1-ft incremental increase of the outside shoulder width of the controlling curve, everything else held constant. In addition, vehicles speeds at the midpoint of the controlling curve on an exit loop ramp in Lane 2 or 3 (outside lanes) on a multi-lane loop ramp are estimated to be approximately 1.2 mph faster than speeds in Lane 1 ( inside lane) of the same ramp. Truck speeds at the midpoint of the controlling curve on an exit loop ramp are expected to be approximately 4.9 mph slower than Linear mixed‐effects model fit by maximum likelihood           AIC      BIC    logLik    9717.268 9755.686 ‐4851.634    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:    1.177096 3.614952    Fixed effects: Speed2 ~ LaneFactor + CCRadius + CC_OSW + VehType                            Value Std.Error   DF    t‐value p‐value  Significance  (Intercept)        16.275717 1.1234930 1770  14.486710  0.0000  Significant at 99% CL  LaneFactor(Lane2)   1.443538 0.4498573 1770   3.208879  0.0014  Significant at 99% CL  CCRadius            0.054387 0.0066668   12   8.157932  0.0000  Significant at 99% CL  CC_OSW              1.078759 0.2089150   12   5.163628  0.0002  Significant at 99% CL  VehType(Truck)     ‐4.051046 0.2494547 1770 ‐16.239606  0.0000  Significant at 99% CL   

1 - 46 speeds of passenger vehicles. Vehicles speeds at the midpoint of the controlling curve on an exit loop ramp designed with a simple curve are estimated to be approximately 3.6 mph faster than on loop ramps designed with a compound curve, and the type of speed-change lane significantly impacts vehicle speeds at the midpoint of the controlling curve on an exit loop ramp. Compared to vehicle speeds at the midpoint of the controlling curve on an exit loop ramp with a tapered speed-change lane, vehicle speeds are expected to be approximately 2.9 mph faster following a lane drop, 4 mph faster following a parallel speed-change lane, and 4.3 mph faster following a weave area. Table 17. Speed Prediction Model to Estimate Speed at the Midpoint of the Controlling Curve on the Ramp Proper of an Exit Loop Ramp Initially using the same variables that were significant in predicting the speeds at the midpoint of the controlling curve on an exit loop ramp, a separate model was estimated to predict the speeds at the beginning (PC) of the controlling curve. Table 18 shows the resulting model using data for single-lane and multi-lane ramps combined. In this model, only the radius of the controlling curve (CCRadius) and the vehicle type [VehType(Truck)] remain significant predictors of speed. The outside shoulder width of the controlling curve (CC_OSW), the type of horizontal curvature (RampRadiusType) (simple or compound), and type of speed-change lane (SCLType) are not significant predictors of vehicle speeds at the beginning of the controlling curve of an exit loop ramp. Linear mixed‐effects model fit by maximum likelihood            AIC      BIC    logLik    8057.741 8116.751 ‐4017.871    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:   0.8642518 3.052811    Fixed effects: Speed2 ~ LaneFactor + CCRadius + CC_OSW + VehType + RampRadiusType + SCLType                               Value Std.Error   DF    t‐value p‐value  Significance  (Intercept)             9.511987 1.7298242 1564   5.498817  0.0000  Significant at 99% CL  LaneFactor(Lane2&3)     1.240567 0.3808842 1564   3.257072  0.0011   Significant at 99% CL  CCRadius                0.053491 0.0084193    6   6.353419  0.0007   Significant at 99% CL  CC_OSW                  1.008357 0.2055809    6   4.904916  0.0027   Significant at 99% CL  VehType(Truck)         ‐4.873262 0.2734472 1564 ‐17.821583  0.0000   Significant at 99% CL  RampRadiusType(Simple)  3.550549 0.7998250    6   4.439158  0.0044   Significant at 99% CL  SCLType(Drop lane)      2.911437 1.3000126    6   2.239545  0.0664   Significant at 90% CL  SCLType(Parallel)       3.975007 0.7467030    6   5.323411  0.0018   Significant at 99% CL  SCLType(Weave)          4.334356 1.1546617    6   3.753788  0.0095   Significant at 99% CL 

1 - 47 Table 18. Speed Prediction Model to Estimate Speed at the Beginning of the Controlling Curve on the Ramp Proper of an Exit Loop Ramp Speed Prediction Procedures Incorporated in ISATe ISATe (Bonneson et al., 2010) utilizes the procedure described here for predicting vehicle speed upon entry/exit to a ramp (or C-D road) curve. The procedure is described separately for entrance and exit ramps. The procedure was developed for use with the safety predictive models for ramps and C-D roads and was not necessarily intended for use in other applications or to predict vehicle speed at points other than the start of a curve. Table 19 presents input data for the procedure along with default values for the average traffic speed on the freeway during off-peak periods on a typical day and the average speed of traffic at the point where the ramp connects to the crossroad. The procedure is applicable for entrance and exit ramps at service interchanges. Table 19. Input Data for Ramp Curve Speed Prediction Procedures in ISATe (Bonneson et al., 2010) Variable Description Default value Applicable procedure Xi Milepost of the point of change from tangent to curve (PC) for curve i 1, mi none All Ri Radius of curve i 2, ft none All LC,i Length of horizontal curve i, mi none All Vfrwy Average traffic speed on freeway during off-peak periods of the typical day, mph Estimate is equal to the speed limit All Vxroad Average speed at point where ramp connects to crossroad, mph 15 – ramps with stop-, yield-, or signal-controlled crossroad ramp terminals 30 – all other ramps at service interchanges Entrance ramp, exit ramp, connector ramp at service interchange Notes: 1 If the curve is preceded by a spiral transition, then Xi is the average of the TS and SC mileposts, where TS is the milepost of the point of change from tangent to spiral and SC is the milepost of the point of change from spiral to curve. 2 If the curve has spiral transitions, then Ri is equal to the radius of the central circular portion of the curve. Entrance Ramp Procedure The procedure for entrance ramps involves seven steps. Step 1 - Gather Input Data: The input data for this procedure are identified in Table 19. Linear mixed‐effects model fit by maximum likelihood            AIC      BIC    logLik    9032.565 9059.388 ‐4511.282    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:    3.444244 4.137368    Fixed effects: Speed1 ~  CCRadius + VehType                       Value Std.Error   DF    t‐value p‐value   Significance  (Intercept)      17.515341  4.125814 1565   4.245305   0e+00   Significant at 99% CL  CCRadius          0.090093  0.019539   11   4.610841   8e‐04   Significant at 99% CL  VehType(Truck)   ‐5.967394  0.370594 1565 ‐16.102245   0e+00   Significant at 99% CL

1 - 48 Step 2 - Compute Limiting Curve Speed: The limiting curve speed is computed for each curve on the ramp using Equation 2. vmax,i = 3.24 (32.2 Ri)0.30 (2) where, vmax,i equals the limiting speed for curve i, ft/s. The analysis proceeds in the direction of travel with the first curve encountered on the ramp designated as curve 1 (i=1). The value of vmax is computed for all curves prior to, and including, the curve of interest. The value obtained from Equation 2 is an upper limit on the curve speed. Vehicles may reach this speed if the distance between curves is long enough or the crossroad speed is high. Step 3 - Calculate Curve 1 Entry Speed: The average entry speed at curve 1 is computed using Equation 3. vent,1 = ([1.47 Vxroad]3 + 495 × 5280 X1)1/3 ≤ 1.47 Vfrwy (3) where, vent,1 equals the average entry speed for curve 1, ft/s. The boundary condition of Equation 3 indicates that the value computed (vent,1) cannot exceed the average freeway speed (Vfrwy). Step 4 - Calculate Curve 1 Exit Speed: The average exit speed at curve 1 is computed using Equation 4. vext,1 = (V3ent,1 + 495 × 5280 Lc,1)1/3 ≤ vmax,1 and ≤ 1.47 Vfrwy (4) where, vext,1 equals the average exit speed for curve 1, ft/s. The boundary conditions of Equation 4 indicate that the value computed (vext,1) should not exceed the limiting curve speed (vmax,i) or the average freeway speed (Vfrwy). Step 5 - Calculate Curve i Entry Speed: The average entry speed at curve 2 (and all subsequent curves) is computed using Equation 5. vent,1 = ( V3ext,i-1 + 495 × 5280 [Xi – Xi-1 – Lc,i-1] )1/3 ≤ 1.47 Vfrwy (5) where, vent,i equals the average entry speed for curve i (i = 2, 3, ...), ft/s; and vext,i equals the average exit speed for curve i, ft/s. Step 6 - Calculate Curve i Exit Speed: The average exit speed at curve 2 (and all subsequent curves) is computed using Equation 6. vext,i = (V3ent,i + 495 × 5280 Lc,i)1/3 ≤ vmax,i and ≤ 1.47 Vfrwy (6)

1 - 49 Step 7 - Calculate Speed on Successive Curves: The entry and exit speeds for curve 3 and successive curves are computed by applying Steps 5 and 6 for each curve. Exit Ramp Procedure The procedure for exit ramps involves seven steps. Step 1 - Gather Input Data: The input data for this procedure are identified in Table 19. Step 2 - Compute Limiting Curve Speed: This step is the same as Step 2 for the entrance ramp procedure. A lower curve speed than that obtained from Equation 2 is possible as deceleration may occur along the ramp as the driver transitions from the freeway speed to the crossroad speed. Step 3 - Calculate Curve 1 Entry Speed: The average entry speed at curve 1 is computed using Equation 7. vent,1 = 1.47 Vfrwy – 0.034 × 5280 X1 ≥ 1.47 Vxroad (7) The boundary condition of Equation 7 indicates that the value computed (vent,1) cannot be less than the average speed at the point where the ramp connects to the crossroad (Vxroad). Step 4 - Calculate Curve 1 Exit Speed: The average exit speed at curve 1 is computed using Equation 8. vext,1 = vent,1 – 0.034 × 5280 Lc,1 ≤ vmax,1 and ≥ 1.47 Vxroad (8) The boundary conditions of Equation 8 indicate that the value computed (vent,1) should not exceed the limiting curve speed (vmax,i) and should not be less than the average speed at the point where the ramp connects to the crossroad (Vxroad). Step 5 - Calculate Curve i Entry Speed: The average entry speed at curve 2 (and all subsequent curves) is computed using Equation 9. vent,i = vext,i-1 – 0.034 × 5280 (X1 – Xi-1 – Lc,i-1) ≥ 1.47 Vxroad (9) Step 6 - Calculate Curve i Exit Speed: The average exit speed at curve 2 (and all subsequent curves) is computed using Equation 10. vext,i = vent,i – 0.034 × 5280 Lc,i ≤ vmax,i and ≥ 1.47 Vxroad (10) Step 7 - Calculate Speed on Successive Curves: This step is the same as Step 7 for the entrance ramp procedure.

1 - 50 Comparison of Estimated Speeds Output by ISATe, Predicted Speeds from Regression Models, and Measured Speeds In this section, the estimated speeds output by ISATe, the predicted speeds based on regression models developed as part of this research, and the measured speeds at the study locations are compared to gain a better understanding of the accuracy and reliability of the speed prediction procedures incorporated in ISATe and that of the regression models developed as part of this research (results shown in Table 15 through Table 18). The site characteristics of the study locations were input into ISATe such that ISATe provided estimated speeds at the midpoint and end (PT) of the controlling curve for each entrance ramp and the beginning (PC) and midpoint of the controlling curve for each exit ramp. To obtain speeds at the midpoint of the controlling curve, the controlling curve was divided into two curves of equal length and input as two adjacent curves. If the site characteristics did not meet the input requirements of ISATe, the closest default or minimum/maximum values were used. For example, the minimum shoulder width value that may be input into ISATe is 2 ft. For several ramps the actual inside (right) and/or outside (left) shoulder width was 1 ft. Therefore, the shoulder width was input into ISATe as 2 ft. Similarly, for some ramps, the lane width exceeded the maximum lane width value that may be input into ISATe (i.e., 20 ft). Therefore, the lane width was input into ISATe as 20 ft. Similarly, the site characteristics of the study locations were input into the regression models developed as part of this research to predict speeds. When inputting the site characteristics of the study locations into the regression models, values were specified such that the predicted speed reflected the average speed of passenger vehicles in Lane 1 of the ramp. Table 20 provides a comparison of the estimated speeds output by ISATe, the predicted speeds based on regressions models developed as part of this research, and the average speeds of passenger vehicles as measured in Lane 1 at all of the entrance ramp locations. Columns 2 through 6 of Table 20 present the information for the midpoint of the controlling curve. Column 2 presents the estimated speeds output from ISATe. Column 3 presents the predicted speeds of passenger vehicles in Lane 1 at the entrance ramps based on the regression model in Table 15. Column 4 presents the average measured speeds of passenger vehicles in Lane 1 at the entrance ramps. Column 5 presents the difference between the ISATe estimated speeds and the average measured speeds from the field. Column 6 presents the difference between the predicted speeds based on the regression model in Table 15 and the average measured speeds from the field. For columns 5 and 6, the last row of the table shows the average of the absolute differences in speeds. Columns 7 through 11 of Table 20 present the corresponding information for the end (PT) of the controlling curve. Column 8 presents the predicted speeds of passenger vehicles in Lane 1 at the entrance ramps based on the regression model in Table 16. For columns 10 and 11, the last row of the table shows the average of the absolute differences in speeds. Table 21 provides the corresponding information for exit ramps. Column 3 presents the predicted speeds of passenger vehicles in Lane 1 at the exit ramps based on the regression model in Table 18, and Column 8 presents the predicted speeds of passenger vehicles in Lane 1 at the exit ramps based on the regression model in Table 17. For columns 5, 6, 10, and 11 the last row of the table shows the average of the absolute differences in speeds. From the results shown in Table 20 and Table 21, the following observations can be made:

1 - 51  The ISATe speed prediction procedure for entrance ramps resulted in the same estimated speeds for the midpoint and PT of the controlling curve.  The ISATe speed prediction procedure for entrance ramps provides reasonably accurate estimates of speeds at the midpoint and PT of the controlling curves. The average difference in speeds output from ISATe and those measured in the field is less at the PT (1.8 mph) of the controlling curve than at the midpoint (2.6 mph).  The ISATe speed prediction procedure for exit ramps is not very accurate in estimating vehicle speeds at the beginning (PC) of the controlling curve on an exit loop ramp, at least for ramps where the distance from the gore point to the controlling curve is relatively short (ISATe overpredicts speeds on all study ramps with average difference of 10.6 mph).  In contrast, the predicted speeds at the PC of the controlling curve of an exit loop ramp based on regression models developed as part of this research (average absolute difference of 3 mph) are considerably more accurate than the estimated speeds output from ISATe.  The ISATe speed prediction procedures for entrance and exit ramps provide similar levels of accuracy when estimating speeds at the midpoint of the controlling curve (average absolute difference of 2.6 and 2.1, respectively).  In general, the differences between the predicted speeds and the measured speeds are, on average, slightly lower than those between speeds output from ISATe and the measured speeds. This suggests that the predicted speeds based on regression models developed as part of this research are slightly more accurate than the estimated speeds output from ISATe. This slightly improved accuracy can be attributed to the fact that the regression models were developed using data from the respective sites.

1 - 52 Table 20. Comparison of Speeds Estimated by ISATe, Predicted by Model, and Measured in Field (Entrance Ramps) Ramp ID Speed at midpoint of controlling curve (mph) Speed at PT of controlling curve (mph) Estimated with ISATe Predicted with model Measured ISATe - measured Predicted - measured Estimated with ISATe Predicted with model Measured ISATe - measured Predicted - measured 1 28.16 26.08 25.57 2.59 0.51 28.16 28.75 28.60 -0.44 0.15 2 32.8 35.47 35.86 -3.06 -0.39 32.8 37.15 38.89 -6.09 -1.74 3 32.8 36.50 35.17 -2.37 1.33 32.8 37.96 37.82 -5.02 0.14 4 29.59 29.85 29.69 -0.1 0.16 29.59 31.57 30.69 -1.10 0.88 5 35.05 33.01 34.60 0.45 -1.59 35.05 37.56 35.63 -0.58 1.93 6 31.98 32.13 31.01 0.97 1.12 31.98 33.10 33.73 -1.75 -0.63 7 28.16 20.91 19.69 8.47 1.22 28.16 25.51 23.91 4.25 1.60 8 26.66 19.92 19.09 7.57 0.83 26.66 24.15 26.06 0.60 -1.91 9 30.68 27.95 28.00 2.68 -0.05 30.68 29.31 30.90 -0.22 -1.59 10 25.64 22.56 24.92 0.72 -2.36 25.64 25.49 23.80 1.84 1.69 11 30.68 27.20 25.70 4.98 1.50 30.68 29.85 29.85 0.83 0.00 12 29.25 28.55 30.22 -0.97 -1.67 29.25 29.62 30.38 -1.13 -0.76 13 28.36 28.67 30.53 -2.17 -1.86 28.36 28.43 29.27 -0.91 -0.84 14 26.52 27.03 27.66 -1.14 -0.63 26.52 27.28 27.92 -1.40 -0.64 15 26.45 26.32 25.56 0.89 0.76 26.45 27.28 27.14 -0.69 0.14 Avg 2.61 1.07 Avg 1.79 0.98

1 - 53 Table 21. Comparison of Speeds Estimated by ISATe, Predicted by Model, and Measured in Field (Exit Ramps) Ramp ID Speed at PC of controlling curve (mph) Speed at midpoint of controlling curve (mph) Estimated with ISATe Predicted with model Measured ISATe - measured Predicted - measured Estimated with ISATe Predicted with model Measured ISATe - measured Predicted - measured 16 34.98 33.82 32.60 2.38 1.22 29.80 30.31 30.15 -0.35 0.16 17 40.36 40.04 39.72 0.64 0.32 32.80 34.03 33.65 -0.85 0.38 18 34.98 31.93 28.70 6.28 3.23 28.70 27.51 26.47 2.23 1.04 19 34.98 31.93 28.28 6.70 3.65 28.70 26.42 28.13 0.57 -1.71 20 42.68 44.54 41.38 1.30 3.16 34.64 35.11 35.72 -1.08 -0.61 21 36.14 37.16 33.34 2.8 3.82 31.50 31.81 32.22 -0.72 -0.41 22 57.41 26.89 25.25 32.16 1.64 25.23 19.11 18.11 7.12 1.00 23 60.14 35.53 41.63 18.51 -6.10 30.68 25.78 25.54 5.14 0.24 24 60.14 35.53 39.08 21.06 -3.55 30.68 26.79 27.96 2.72 -1.17 25 50.25 39.68 44.95 5.30 -5.27 32.66 29.02 29.11 3.55 -0.09 26 54.61 38.24 38.37 16.24 -0.13 31.98 32.81 31.82 0.16 0.99 27 50.11 40.04 36.92 13.19 3.12 32.80 32.43 30.90 1.90 1.53 28 48.61 32.92 37.23 11.38 -4.31 29.25 29.71 30.89 -1.64 -1.18 Avg 10.61 3.04 Avg 2.16 0.81

1 - 54 3.4.1.3 Key Findings In summary the key findings from analyses of the speed data for entrance and exit loop ramps are as follows:  General findings - The design speeds of the loop ramps were calculated using Method 5 and Method 2 of the Green Book for distributing superelevation and side friction. At all study locations, the estimated design speeds using Method 5 are 10 to 15 mph lower than the estimated design speeds using Method 2, and the observed speeds are higher than the estimated design speeds using Method 5 and are much closer to the estimated design speeds using Method 2.  Entrance ramps - Speeds at the PT of the controlling curve were slightly higher than speeds at the midpoint of the controlling curve. - Key roadway and cross-sectional design elements that significantly affect vehicle speeds at the midpoint of the controlling curve include radius, lane width, inside shoulder width, and outside shoulder width. - Key roadway and cross-sectional design elements that significantly affect vehicle speeds at the PT of the controlling curve include radius and outside shoulder width. - Truck speeds are approximately 4 mph slower than passenger vehicle speeds. - Vehicles travel approximately 1.5 to 2.0 mph faster in the outside lane of a multi- lane loop ramp than in the inside lane of a multi-lane loop ramp. - The ISATe speed prediction procedure for entrance ramps provides reasonably accurate estimates of speeds at the midpoint and PT of the controlling curves, with slightly better estimates at the PT than at the midpoint of the controlling curve.  Exit ramps - Speeds at the PC of the controlling curve were slightly higher than speeds at the midpoint of the controlling curve. - Key roadway and cross-sectional design elements that significantly affect vehicle speeds at the midpoint of the controlling curve include radius, outside shoulder width, type of curvature (simple or compound), and type of mainline freeway ramp terminal. - Radius is the only key roadway and cross-sectional design element that significantly affects vehicle speeds at the PC of the controlling curve. - Truck speeds are approximately 5 to 6 mph slower than passenger vehicle speeds. - At the midpoint of the controlling curve vehicles travel approximately 1.2 mph faster in the outside lanes of a multi-lane loop ramp than in the inside lane of a multi-lane ramp.

1 - 55 - The ISATe speed prediction procedure for exit ramps is not very accurate in estimating the speeds of vehicles at the beginning of the controlling curve, at least for ramps where the distance from the gore point to the controlling curve is relatively short. - The predicted speeds at the PC of the controlling curve of an exit loop ramp based on regressions models developed as part of this research are considerably more accurate than the estimated speeds output from ISATe. In summarizing these findings, it is important to recognize the limitations of the dataset used in the analysis, namely:  Of the four multi-lane entrance loop ramps, speed data were collected for only 104 passenger vehicles and five trucks in the outside lane (Lane 2).  Only one of the 13 study exit ramps was a multi-lane ramp (3 lane ramp). Speed data were collected for 174 passenger vehicles and one truck in the outside lanes (either Lane 2 or Lane 3) of the ramp. Thus, the amount of speed data for vehicles in the outside lane(s) of multi-lane loop ramps is limited, especially for trucks. 3.4.2 Lane Position The objective of this analysis is to assess the effects of key design elements on the position of vehicles in the traveled way on loop ramps. The positions of vehicles at the midpoint of the controlling curves along the ramp proper of loop ramps were obtained through video field data collection and reduction. The position of the vehicle reflects the position of the right (passenger side) tire that was closest to the inside lane line (either the inside lane line that separates the inside travel lane and the inside shoulder, or the lane line that separates the travels lanes on a multi-lane ramp). A positive value indicates the right tire is within the intended travel lane, while a negative value indicates the right tire encroached onto the shoulder or the inside travel lane. Similar to the analyses for the speed data, analyses were initially conducted to estimate the effects of key design elements on lane position, separately for single-lane entrance and exit ramps. Then analyses were conducted on the full dataset which included data for both single-lane and multi-lane ramps, separately for entrance and exit ramps. The same key design elements considered in the analyses for speed were also considered in the analyses for lane position, namely:

1 - 56  Ramp length  Design speed of controlling curve  Radius of controlling curve  Superelevation of controlling curve  Length of controlling curve  Inside shoulder width of controlling curve  Lane width of controlling curve  Outside shoulder width of controlling curve  Freeway speed limit  Crossroad speed limit  Vertical profile (grade, up or down)  Type of speed-change lane  Type of horizontal curvature (simple or compound curve)  Type of crossroad traffic control  Vehicle type Regression models were developed using the restricted maximum likelihood (REML) method implemented in the JMP statistical package for SAS to estimate the effect of the different key design elements on lane position. Prior to model development, several descriptive statistics were generated and correlations of the key design elements were investigated. Only the most relevant models and findings are presented later in Section 3.4.2.2 (Analysis Results). 3.4.2.1 Descriptive Statistics Table 22 (entrance ramps) and Table 23 (exit ramps) provide basic summary statistics of lane position. For each ramp, lane, and vehicle type, these include number of observations, average lane position, standard deviation of lane position, and minimum and maximum lane positions. In Table 22, lane position statistics are provided by lane and combined across all lanes of a given ramp for multi-lane ramps. Lane position statistics are provided separately for passenger vehicles and trucks and for all vehicles combined. For entrance ramps, lane position was recorded for 1,526 passenger vehicles and 112 trucks. Of the 1,638 lane positions recorded for entrance ramps, 116 observations were for vehicles in the outside lane (Lane 2) of a multi-lane entrance ramp. Table 23 provides the similar basic summary statistics of lane position for exit ramps. Lane position was recorded for 1,465 passenger vehicles and 118 trucks. Of the 1,583 lane positions recorded, 107 observations were for vehicles in the middle and/or outside lane of a multi-lane exit ramp. Table 24 lists site characteristics and vehicle-related variables considered in the analyses to investigate the effect of key design elements on lane position, as well as their basic summary statistics. Unless specified otherwise, the summary statistics for entrance ramps are based on data from 1,638 vehicles, and those for exit ramps are based on data from 1,583 vehicles.

1 - 57 Table 22. Descriptive Lane Position Statistics on Ramp Proper (Entrance Ramps) Ramp ID Lane No. Lane position (inches) Vehicle type Number of Obs Mean Std Dev Min Max 1 1 Passenger vehicle 166 30.8 24.1 -12 102 Truck 19 41.1 24.2 12 96 All 185 31.8 24.2 -12 102 2 1 Passenger vehicle 98 15.1 17.7 -48 66 Truck 6 15.0 24.2 -6 60 All 104 15.1 18.0 -48 66 3 1 Passenger vehicle 99 22.1 26.2 -96 84 Truck 7 13.7 22.4 -12 54 All 106 21.5 26.0 -96 84 4 1 Passenger vehicle 21 30.9 35.5 -36 102 Truck 4 13.5 25.6 -6 48 All 25 28.1 34.2 -36 102 5 1 Passenger vehicle 105 23.8 21.3 -30 90 Truck 3 28.0 9.2 18 36 All 108 23.9 21.0 -30 90 6 1 Passenger vehicle 16 43.9 23.4 12 96 Truck 10 51.0 24.5 24 96 All 26 46.6 23.6 12 96 7 1 Passenger vehicle 90 56.9 15.0 12 84 Truck 16 48.6 19.0 18 84 All 106 55.7 15.8 12 84 8 1 Passenger vehicle 94 45.4 16.0 0 78 Truck 9 41.2 17.1 24 72 All 103 45.1 16.1 0 78 9 1 Passenger vehicle 45 31.7 24.6 -6 78 Truck 12 26.5 24.6 0 78 All 57 30.6 24.5 -6 78 10 1 Passenger vehicle 145 40.7 17.5 0 102 Truck 6 80.0 19.2 54 102 All 151 42.3 19.1 0 102 12 1 Passenger vehicle 168 24.5 23.2 -30 72 Truck 7 18.9 26.5 -12 48 All 175 24.2 23.3 -30 72 2 Passenger vehicle 1 12.0 . 12 12 All 1 12.0 . 12 12 13 1 Passenger vehicle 120 16.6 25.0 -48 78 Truck 1 24.0 . 24 24 All 121 16.7 24.9 -48 78 2 Passenger vehicle 61 19.0 13.2 -12 54 Truck 1 24.0 . 24 24 All 62 19.1 13.1 -12 54 14 1 Passenger vehicle 105 23.5 23.6 -30 78 Truck 7 5.1 20.3 -24 24 All 112 22.3 23.7 -30 78 2 Passenger vehicle 49 30.0 14.1 0 60 Truck 2 30.0 33.9 6 54 All 51 30.0 14.6 0 60 15 1 Passenger vehicle 141 29.3 21.6 -24 78 Truck 2 -6.0 25.5 -24 12 All 143 28.8 22.0 -24 78 2 Passenger vehicle 2 18.0 25.5 0 36 All 2 18.0 25.5 0 36

1 - 58 Table 23. Descriptive Lane Position Statistics on Ramp Proper (Exit Ramps) Ramp ID Lane No. Lane position (inches) Vehicle type Number of Obs Mean Std Dev Min Max 16 1 Passenger vehicle 142 42.4 23.5 0 120 Truck 16 50.3 25.2 18 96 All 158 43.2 23.7 0 120 17 1 Passenger vehicle 135 14.4 22.9 -42 90 Truck 1 36.0 . 36 36 All 136 14.6 22.9 -42 90 18 1 Passenger vehicle 56 52.1 26.1 -6 120 Truck 4 45.0 46.6 12 114 All 60 51.6 27.4 -6 120 19 1 Passenger vehicle 72 35.1 23.9 -12 96 Truck 15 59.6 28.2 18 120 All 87 39.3 26.2 -12 120 20 1 Passenger vehicle 156 38.9 21.1 -6 90 Truck 6 41.0 25.9 0 72 All 162 39.0 21.2 -6 90 21 1 Passenger vehicle 91 25.8 28.8 -30 126 Truck 18 43.3 31.7 -12 96 All 109 28.7 29.9 -30 126 22 1 Passenger vehicle 196 46.4 19.4 12 102 Truck 7 57.4 19.6 36 90 All 203 46.8 19.4 12 102 23 1 Passenger vehicle 121 36.4 20.4 0 84 Truck 12 48.0 24.8 0 78 All 133 37.5 21.0 0 84 24 1 Passenger vehicle 128 27.4 18.2 -30 96 Truck 13 21.2 25.6 -24 72 All 141 26.8 19.0 -30 96 25 1 Passenger vehicle 120 18.8 20.8 -30 78 Truck 17 22.9 28.0 -18 66 All 137 19.3 21.7 -30 78 26 1 Passenger vehicle 49 31.5 22.7 -6 96 Truck 4 10.5 12.4 0 24 All 53 29.9 22.7 -6 96 27 1 Passenger vehicle 19 3.8 21.5 -42 48 All 19 3.8 21.5 -42 48 28 1 Passenger vehicle 74 8.6 22.5 -30 72 Truck 4 -16.5 17.9 -42 0 All 78 7.3 22.9 -42 72 2 Passenger vehicle 42 28.7 14.5 0 66 Truck 1 18.0 . 18 18 All 43 28.5 14.5 0 66 3 Passenger vehicle 64 25.2 14.1 -12 60 All 64 25.2 14.1 -12 60

1 - 59 Table 24. Summary Statistics for Lane Position Data Entrance ramps Exit ramps Variable Min Max Avg Std Dev Min Max Avg Std Dev Lane position (in) -96 102 29.65 24.27 -42 126 32.05 25.32 Site characteristics (continuous variables) FwySpLimit freeway speed limit, mph 60 75 64.56 3.80 50 70 63.07 5.02 RampLength length of ramp, ft 564 2007 1096 400 546 1962 1164 474 CrSpLim crossroad speed limit, mph 30 65 43.32 8.10 30 65 40.11 9.47 CCDesSpd design speed of controlling curve, mph 15 20 16.18 2.13 15 25 17.58 2.62 CCRadius radius of controlling curve, ft 110 312 169 57 104 300 200 54 CCLength length of controlling curve, ft 220 1433 592 329 260 1050 621 264 CC_ISW inside shoulder width of controlling curve, ft 1.00 10.08 5.92 3.33 1.00 12.75 6.50 3.32 CC_LW lane width of controlling curve, ft (lane 1)1 14.00 21.00 16.35 2.44 14.00 20.92 17.09 2.20 CC_LW lane width of controlling curve, ft (lane 2)2 11.75 12.00 12.00 0.02 14.00 14.00 14.00 0.00 CC_LW lane width of controlling curve, ft (lane 3)3 14.00 14.00 14.00 0.00 CC_OSW outside shoulder width of controlling curve, ft 1.00 7.50 3.79 1.63 2.00 7.08 4.54 1.81 CCSuper superelevation of controlling curve, ft/ft 0.04 0.07 0.05 0.01 0.02 0.07 0.04 0.01 Site characteristics (categorical variables) Categories (proportion) in the data Grade Up (0.26), Down (0.74) Up (0.60), Down (0.40) SCL Type type of speed-change lane Weave (0.37), Parallel (0.21), Add lane (0.24), Taper (0.19) Parallel (0.32), Taper (0.30), Weave (0.24), Drop lane (0.14) RampRadiusType simple or compound curve Simple (0.32), Compound (0.68) Simple (0.34), Compound (0.66) CrTrafCon crossroad traffic control Signal (0.11), Free-flow (0.89) Stop control (0.10), Signal (0.50), Free- flow (0.40) Vehicle-related (categorical variables) VehType vehicle type Car (0.93), Truck (0.07) Car (0.93), Truck (0.07) 1 Summary statistics for lane width of controlling curve (Lane 1) are based on 1,522 vehicles for entrance ramps and 1,476 vehicles for exit ramps. 2 Summary statistics for lane width of controlling curve (Lane 2) are based on 116 vehicles for entrance ramps and 43 vehicles for exit ramps. 3 Summary statistics for lane width of controlling curve (Lane 3) are based on 64 vehicles for exit ramps.

3.4.2.2 This sect loop ram effect of assessed represent the corre Figu Entrance After rev goodness of the con some cor freedom, coefficien percent, h Analysis R ion presents ps on lane p key design e to identify w ation of the lation. re 11. Corr Ramps iewing a nu of fit is sho trolling cur relations bet but it provi t of superel ence the na esults the analysis osition of ve lements on hich variab correlation elations am mber of can wn in Table ve on the ra ween estim des reasonab evation mor me of the va results inte hicles. Befo lane positio les should b among varia ong Variab didate mode 25 for estim mp proper o ates, particu le values fo e meaningfu riable as CC 1 - 60 nded to iden re developi n, the correl e included i bles where t les in the L ls, the mode ating the la f an entranc larly the fou r the model l, the units Super100. tify the effe ng regressio ation among n the model he size of th ane Positio l yielding th ne position e loop ramp r estimates coefficients of superelev ct of key de n models to variables in s. Figure 11 e circles is n Database e most info of vehicles . This mode fitted on nin . Also, to m ation were sign elemen estimate the the dataset is a graphic proportiona for All Site rmation and at the midpo l still shows e degrees of ake the changed to ts of was al l to s best int

1 - 61 Table 25. Lane Position Model for Entrance Ramps This lane position prediction model can be interpreted as follows. At the midpoint of the controlling curve on an entrance loop ramp, vehicles are expected to position themselves approximately:  2.3 inches farther away from the inside lane line for each 1-ft increase in lane width, everything else held constant.  2 inches closer to the inside lane line for each 1-ft increase in outside shoulder width, everything else held constant.  5.3 inches closer to the inside lane line for each 1-percent increase in superelevation, everything else held constant.  1.7 inches closer to the inside lane line for each 100-ft increase in controlling curve length, everything else held constant.  15 inches closer to the inside lane line if the loop ramp is on an upgrade, everything else held constant.  10 inches farther away from the inside lane line when traveling in the outside lane of a multi-lane ramp compared to traveling in the inside lane, everything else held constant. The model also shows that lane positioning is not significantly different between passenger vehicles and trucks. Exit Ramps The same process was repeated for exit ramps, beginning with developing a prediction model to estimate the lane position of vehicles at the midpoint of the controlling curve on the ramp proper of an exit loop ramp, resulting in the model shown in Table 26. It is worth noting that CCSuper100 was kept in the model because it significantly affects the goodness of fit, yet its parameter estimate is not statistically significant at the 90-percent significant level, but barely so Linear mixed‐effects model fit by REML            AIC      BIC    logLik    14758.13 14812.09 ‐7369.066    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:     4.07133 21.72377    Fixed effects: LanePos ~ LaneFactor + CC_LW + CC_OSW + CCSuper100 + CCLength + Grade + VehType                         Value Std.Error   DF   t‐value p‐value   Significance  (Intercept)        42.18501 14.053540 1621  3.001735  0.0027  Significant at 99% CL  LaneFactor(Lane2)  10.09629  3.036311 1621  3.325182  0.0009  Significant at 99% CL  CC_LW               2.28886  0.654232 1621  3.498541  0.0005  Significant at 99% CL  CC_OSW             ‐2.03742  1.009476    9 ‐2.018290  0.0743  Significant at 90% CL  CCSuper100         ‐5.26412  1.262833    9 ‐4.168499  0.0024  Significant at 99% CL  CCLength           ‐0.01661  0.006159    9 ‐2.697257  0.0245  Significant at 95% CL  Grade(Up)         ‐14.59897  4.199929    9 ‐3.476004  0.0070  Significant at 99% CL  VehType(Truck)     ‐0.55856  2.180207 1621 ‐0.256196  0.7978   Not significant at 90% CL   

1 - 62 (p-value of 0.1193). It is interesting, however, that the direction of the effect and its standard error are comparable to that in the model for entrance ramps. Table 26. Lane Position Model for Exit Ramps This lane position prediction model can be interpreted as follows. At the midpoint of the controlling curve on an exit loop ramp, vehicles are expected to position themselves:  Approximately 5 inches farther away from the inside lane line for each 1-ft increase in lane width, everything else held constant.  Approximately 13.7 inches farther away from the inside lane line if the speed-change lane preceeding the loop ramp is a drop lane, everything else held constant.  In the outside lane(s) of a multi-lane loop ramp approximately 20 inches farther away from the inside lane line than vehicles traveling in the inside lane, everything else held constant. Also, trucks position themselves approximately 6.6 inches farther from the inside lane line than passenger vehicles. 3.4.2.3 Key Findings In summary, the key findings from analyses of lane position on entrance and exit loop ramps are as follows:  Entrance ramps - Key roadway and cross-sectional design elements that significantly affect lane position at the midpoint of the controlling curve include lane width, outside shoulder width, superelevation, and grade. Linear mixed‐effects model fit by REML           AIC      BIC    logLik    14327.94 14370.84 ‐7155.968    Random effects:   Formula: ~1 | SiteID          (Intercept) Residual  StdDev:    6.570811 22.20448    Fixed effects: LanePos ~ LaneFactor + CC_LW + CCSuper100 + SCLType + VehType                                      Value Std.Error   DF   t‐value p‐value  (Intercept)                    ‐42.16374 20.419250 1567 ‐2.064902  0.0391  LaneFactor(Lane2)               20.34655  3.257214 1567  6.246611  0.0000  CC_LW                            4.88004  0.954767 1567  5.111233  0.0000  CCSuper100                      ‐2.94853  1.730668   10 ‐1.703692  0.1193  SCLType(Drop lane)              13.74523  5.599425   10  2.454757  0.0340  VehType(Truck)                   6.59117  2.162167 1567  3.048412  0.0023 

1 - 63 - Vehicles in the outside lane of a multi-lane loop ramp are positioned approximately 10 inches farther away from the inside lane line than vehicles traveling in the inside lane. - There is no significant difference in the lane positions between passenger vehicles and trucks.  Exit ramps - Key roadway and cross-sectional design elements that significantly affect lane position at the midpoint of the controlling curve are lane width and type of freeway mainline ramp terminal. - Vehicles in the outside lane(s) of a multi-lane loop ramp are positioned approximately 20 inches farther away from the inside lane line than vehicles traveling in the inside lane. - Trucks are positioned approximately 6.6 inches farther away from the inside lane line than passenger vehicles. In summarizing these findings, it is important to recognize the limitations of the dataset used in the analysis, namely:  Only a limited dataset of lane position data was available for vehicles traveling in the outside lane(s) of a multi-lane ramp.  Only a limited dataset of lane position data was available for trucks. 3.4.3 Exiting Maneuvers A qualitative investigation was conducted to determine whether patterns of unusual or critical maneuvers occurred along the first several hundred feet of the ramp proper at nine exit loop ramps included in the observational study, and if so, to determine whether features of the sites could possibly be contributing to that behavior. 3.4.3.1 Descriptive Statistics Table 27 and Table 28 provide summary statistics for exiting vehicles and exit maneuvers observed as part of this research. Approximately 7 hours of video were reviewed to identify erratic or critical maneuvers of exiting vehicles; maneuvers for approximately 1,400 vehicles were recorded in the database, including 105 trucks. In total, 139 encroachments on the shoulders and/or critical maneuvers were identified, accounting for about 10 percent of all exiting maneuvers observed. None of the maneuvers resulted in a crash.

1 - 64 Table 27. Summary Statistics of Encroachment and Critical Maneuvers Observed on Exit Ramps Ramp ID Number of Obs Truck count Percent trucks Time reviewed Number of encroachments and/or critical maneuvers observed Percent encroachments and/or critical maneuvers 16 574 42 7.3% 1:31:01 45 7.8% 17 136 1 0.7% 1:01:00 32 23.5% 18 60 4 6.7% 0:53:04 1 1.7% 20 162 6 3.7% 0:30:00 23 14.2% 21 109 18 16.5% 0:52:40 7 6.4% 24 141 13 9.2% 0:30:00 7 5.0% 25 137 17 12.4% 0:31:45 4 2.9% 26 53 4 7.5% 0:30:20 2 3.8% 27 19 0 0.0% 0:52:28 18 94.7% All 1391 105 7.5% 7:12:18 139 10.0% As shown in Table 28, most of the maneuvers recorded were encroachments onto the inside or outside shoulders. Only three critical maneuvers were observed, two of which involved vehicles swerving and one involving severe braking. Two of the critical maneuvers involved trucks. Of the encroachments observed, 40 percent involved vehicles encroaching onto the inside shoulder, and 60 percent involved vehicles encroaching onto the outside shoulder. Of the 139 encroachments on the shoulders and/or critical maneuvers observed, 90 percent involved passenger vehicles, and 10 percent involved trucks. This indicates that trucks are slightly overrepresented in these shoulder encroachments/critical maneuvers, considering that 7.5 percent of the vehicles observed were trucks. Table 28. Summary Statistics of Encroachment and Critical Maneuvers Observed on Exit Ramps by Encroachment, Maneuver Type, and Vehicle Type Ramp ID Encroachment Critical maneuver Passenger vehicle Truck Total 16 Inside shoulder None 3 2 5 Outside shoulder None 38 1 39 None Severe braking 0 1 1 All All 41 4 45 17 Outside shoulder None 31 1 32 All All 31 1 32 18 Inside shoulder None 1 0 1 All All 1 0 1 20 Inside shoulder None 20 1 21 Outside shoulder None 1 1 2 All All 21 2 23 21 Inside shoulder None 5 0 5 Outside shoulder None 0 2 2 All All 5 2 7 24 Inside shoulder None 3 1 4 Outside shoulder None 0 2 2 Outside shoulder Swerving 1 0 1 All All 4 3 7 25 Inside shoulder None 1 0 1 Outside shoulder None 1 1 2 Outside shoulder Swerving 0 1 1 All All 2 2 4 26 Outside shoulder None 2 0 2 All All 2 0 2 27 Inside shoulder None 18 0 18 All All 18 0 18 All Ramps All All 125 14 139

1 - 65 3.4.3.2 Analysis Results Of the 139 encroachments and critical maneuvers observed, only three involved swerving or severe braking; the remaining 136 maneuvers involved encroachments on the shoulder (51 on the inside shoulder and 85 on the outside shoulder). This indicates that the majority of drivers were able to enter the loop ramps in a controlled manner. Most of the sites had fewer than 10 percent of the observed vehicles encroach on the inside or outside shoulder or perform a critical maneuver. The most notable exception was Ramp 27, on which nearly all of the vehicles encroached on the inside shoulder to exit the freeway. On Ramp 27 the freeway has a posted speed limit of 50 mph, and the exit is on the far side of a vertical curve and at the end of a parallel speed-change lane. Ramp 27 has one of the narrowest lane width (14.5 ft) of all the exit ramps where exiting maneuvers were studied. Ramp 17 had nearly a quarter of the observed vehicles encroach on the outside shoulder. The posted speed limit of the freeway is 70 mph. When comparing all exit ramps where exiting maneuvers were studied, Ramp 17 had the third highest average measured speed at the PC of the controlling curve so drivers were entering the controlling curve at relatively high speeds. Ramp 17 also has a relatively narrow travel lane (15.0 ft) and is located at the end of a weaving area. About 15 percent of vehicles on Ramp 20 encroached on the inside shoulder. The posted speed limit of the freeway is 70 mph, and Ramp 20 had the second highest average measured speed at the PC of the controlling curve when comparing all exit ramps where exiting maneuvers were studied. Ramp 20 also has a relatively narrow travel lane (15.2 ft). Table 29 summarizes the encroachment and critical maneuver statistics in a slightly different manner along with several key site characteristics that may provide insight into the observed behaviors. In Table 29 the ramps are sorted by increasing percentage of encroachments and/or critical maneuvers observed. The third column shows the average speed of vehicles (both passenger vehicles and trucks combined) measured in the field at the beginning of the controlling curve. The other columns present key site characteristics that may influence driver behavior as vehicles transition from the speed-change lane (deceleration lane) to the ramp proper of an exit loop ramp. Table 29. Summary Statistics of Encroachment and Critical Maneuvers and Key Site Characteristics of the Exit Ramps Ramp ID Percentage of encroachments and/or critical maneuvers Avg speed1 (mph) Type of freeway mainline ramp terminal Freeway speed limit (mph) Controlling curve Grade Radius (ft) Inside shoulder width (ft) Lane width (ft) Outside shoulder width (ft) Super (%) 27 94.7% 36.78 Parallel 50 250 6.0 14.5 2.0 5.8 Down 17 23.5% 39.58 Weave 70 250 10.3 15.0 6.8 5.0 Up 20 14.2% 41.38 Drop lane 70 300 12.8 15.2 6.6 5.0 Up 16 7.8% 31.07 Parallel 65 181 3.5 18.7 7.1 5.0 Down 21 6.4% 32.59 Weave 70 218 6.0 17.2 6.3 4.0 Down 24 5.0% 38.19 Taper 60 200 7.0 17.0 3.0 2.0 Up 26 3.8% 38.14 Drop lane 60 230 7.0 16.0 4.5 4.0 Up 25 2.9% 43.80 Weave 60 246 6.0 14.0 2.0 5.0 Up 18 1.7% 28.44 Parallel 55 160 4.4 20.0 5.4 3.0 Down 1 Average speed of vehicles (both passenger vehicles and trucks combined) measured at the beginning of the controlling curve

1 - 66 Several observations related to the patterns of encroachment and/or critical maneuvers that occurred along the first several hundred feet of the ramp proper at the nine exit loop ramps studied can be made as follows:  It is likely that encroachments and/or critical maneuvers are directly related to the speeds of exiting vehicles, as some of the ramps with the highest percent encroachments and/or critical maneuvers observed also have the highest measured speeds of vehicles at the beginning of the controlling curve.  It is likely that encroachments and/or critical maneuvers are directly related to the lane width of ramp proper. When comparing all exit ramps where exiting maneuvers were studied, the three ramps with the highest percent observed encroachments and/or critical maneuvers also had three of the four narrowest lane widths, ranging from 14.5 to 15.2 ft.  The interaction between narrower lane widths and higher speeds likely resulted in higher percent observed encroachments and/or critical maneuvers, the exception being Ramp 25, which had the highest measured speeds of vehicles at the beginning of the controlling curve and the narrowest lane width (14.0 ft). 3.4.3.3 Key Findings In summary, the key findings from the qualitative analysis of exiting maneuvers on exit loop ramps are the following:  A majority of drivers were able to enter the loop ramps in a controlled manner.  Of the 139 encroachments and critical maneuvers observed, only three involved swerving or severe braking and the remaining 136 maneuvers involved encroachments on the shoulder. None of the maneuvers resulted in a crash.  Given that 7.5 percent of the observed vehicles were trucks and that trucks accounted for approximately 10 percent of the total number of encroachments and/or critical maneuvers observed, trucks may be slightly overrepresented in these shoulder encroachments and/or critical maneuvers, but for practical purposes, the percent difference is small enough that trucks should not, in general, be considered as overrepresented in the number of encroachments and/or critical maneuvers that were observed.  Combinations of narrower lane width and higher approach speed at the beginning of the controlling curve showed a higher proportion of observed encroachments and critical maneuvers. In summarizing the findings, it is important to recognize the limitations of the dataset used in the analysis, namely that it included data for approximately 1,400 exiting maneuvers only, with 10 percent of them involving trucks.

1 - 67 Section 4. Application of the HSM Ramp Crash Prediction Method to Loop Ramps This section of the report presents the results of applying the HSM ramp crash prediction method to loop ramps, as well as to a contrasting ramp configuration, diamond ramps. The results provide insight into how the HSM procedures should be applied to loop ramps. 4.1 HSM Ramp Crash Prediction Method A recent supplement to the AASHTO Highway Safety Manual (HSM) (AASHTO, 2014) includes a crash prediction method for ramps at freeway interchanges. This crash prediction method, presented in HSM Chapter 19, was developed in NCHRP Project 17-45. ISATe, a spreadsheet based tool, implements the HSM crash prediction models for ramps. The following discussion addresses the crash prediction methods for roadway segments located on a specific ramp, as opposed to ramp terminal or collector-distributor (C-D) road locations. As in other HSM crash prediction methods, the ramp crash prediction method consists of safety performance functions (SPFs) and crash modification factors (CMFs), plus a local calibration factor. SPFs are provided separately for multiple- and single-vehicle crashes and for fatal-and- injury (FI) and property-damage-only (PDO) crashes at six specific types of ramp segments:  Rural one-lane entrance ramps  Rural one-lane exit ramps  Urban one-lane entrance ramps  Urban one-lane exit ramps  Urban two-lane entrance ramps  Urban two-lane exit ramps The SPFs for multiple-vehicle crashes use the following functional form: Nspf, mv = Lr × exp[a + b×ln(c×AADTr) + d(c×AADTr)] (11) where: Nspf, mv = predicted average multiple-vehicle crash frequency for a ramp segment Lr = length of ramp segment (mi) AADTr = AADT volume of ramp segment (veh/day) a, b, c, d = regression coefficients The SPFs incorporate the effect of average daily traffic volume (AADT) on crash frequency. The values for coefficients a, b, c, and d for use in Equation 11 are presented in HSM Table 19-7.

1 - 68 The SPFs for single-vehicle crashes use the following functional form: Nspf, sv = Lr × exp[a + b×ln(c×AADTr)] (12) where: Nspf, sv = predicted average single-vehicle crash frequency for a ramp segment The values for coefficients a and b for use in Equation 12 are presented in HSM Table 19-8. The CMFs for ramp segments away from ramp terminals and C-D roads address the following geometric design elements:  Horizontal curves  Lane width  Right shoulder width  Left shoulder width  Right side barrier  Left side barrier  Lane add or drop Each of these CMFs is determined with an equation that may have separate tabulated values for multiple- and single-vehicle crashes and for FI and PDO crashes. 4.2 Application of the HSM Crash Prediction Method to Specific Ramp Configurations The HSM crash prediction method for ramp segments does not separate procedures for specific ramp configurations, such as diamond ramps and loop ramps. The same SPFs and CMFs are applied to both diamond and loop ramps. Therefore, any differences in crash predictions for diamond and loop ramps must arise from the geometric features of those ramps. Since the same SPFs are applied, the effect of AADT on crash frequency is assumed by the HSM method to be the same for both diamond and loop ramps and the same SPF coefficient values from HSM Tables 19-7 and 19-8 are used for both diamond and loop ramps. Particular diamond and loop ramps may have the same cross section features – lane width, shoulder width, and roadside barriers, so the fundamental geometric feature that differs between diamond and loop ramps is horizontal curvature. Loop ramps clearly have longer and smaller-radius horizontal curves than other ramp configurations. Therefore, the horizontal curvature CMF appears to be the key variable in the HSM crash prediction method that would distinguish the safety performance of loop ramps from the safety performance of other ramp types.

1 - 69 The CMF for horizontal curves on ramps is computed as: ܥܯܨଵ ൌ 1.0 ൅ a ൈ ଵ,଴଴଴ଷଶ.ଶ ሾ∑ ሺ ௏௘௡௧,௜ ோ೔ ሻ ଶ ൈ ௖ܲ,௜௠௜ୀଵ ሿ (13) where: ܥܯܨଵ = crash modification factor for horizontal curvature on a ramp segment m = number of horizontal curves in the ramp segment ௘ܸ௡௧,௜ = average entry speed for curve i (ft/s) ܴ௜ = radius of curve i (ft) ௖ܲ,௜ = proportion of segment length with curve i The values for coefficient a in Equation 13 are presented in HSM Table 19-24. These coefficient values vary for multiple- and single-vehicle crashes and for FI and PDO crashes, but they do not vary based on ramp or horizontal curve characteristics. The speed component of the horizontal curve CMF in Equation 13 was developed from data for 20 horizontal curves on rural two-lane highways and five interchange loop ramp curves (Bonneson et al., 2012). Thus, while this CMF is applied to ramps in the HSM procedure, its development was not based primarily on ramp data, and certainly not on loop ramp data. Therefore, an evaluation of the appropriateness for loop ramps of this CMF, and the crash prediction method that contains it, appears to be a critical need. 4.3 Key Considerations in Assessing the Application of the HSM Ramp Crash Prediction Models to Loop Ramps Since there is no separate model for loop ramps in the HSM ramp crash prediction method, a key consideration is whether the existing HSM method is effective in distinguishing between the safety performance of loop ramps and other ramp configurations, such as diamond ramps, that are less likely to include long, short-radius horizontal curves. While the investigation of this issue should address the full HSM ramp crash prediction method, clearly the key element of the method being evaluated is the horizontal curve CMF presented in Equation 13. This investigation can be considered a validation of the HSM ramp crash prediction method for loop ramps. The steps in the validation approach include: 1. Select a sample of loop ramps for investigation, including both rural and urban ramps and both entrance and exit ramps 2. Select a similar sample of diamond ramps 3. Review aerial photographs and highway agency records and obtain, for each ramp, all of the data needed to apply the HSM crash prediction method for the “ramp proper” area (i.e., not including ramp terminals) 4. Apply the HSM ramp crash prediction method to each diamond and loop ramp to obtain the predicted number of ramp crashes per year, by severity level

1 - 70 5. From highway agency records, obtain the actual observed crash frequency, by severity level, for the “ramp proper” area, for a five-year period 6. Compare the predicted and observed crash frequencies, by ramp type and ramp configuration, to obtain calibration factors and compare the appropriateness of the HSM predictions for diamond and loop ramps Diamond ramps were selected for comparison to loop ramps because many diamond ramps exist and thus can be readily found and because they typically have fewer, shorter length, and longer radius horizontal curves than loop ramps. This contrast in horizontal curve characteristics provides a good opportunity to assess the applicability of the horizontal curve CMF in Equation 13 to these differing ramp configurations. The following sections review the conduct of this investigation and its results. 4.4 Selection of Study Ramps The database developed in this investigation includes a total of 235 loop ramps and 243 diamond ramps. Ramps were selected in two states – California and Washington – for which data were available from the FHWA Highway Safety Information System (HSIS) to identify the location and configuration of specific ramps. All of the ramps considered were located at freeway-arterial interchanges (i.e., service interchanges). In both states, the sample of loop ramps included both parclo loop ramps (which typically have a stop-controlled ramp terminal at the crossroad) and free-flow loop ramps (which typically have a free-flow speed-change lane at the crossroad). The HSIS data were used to create lists of candidate sites which were reviewed in aerial photographs in Google Earth® to determine their suitability for the study. Candidate sites included the following 12 ramp classifications (defined by a combination of area type, ramp configuration, and ramp type):  Rural free-flow loop entrance ramps  Rural parclo loop entrance ramps  Rural diamond entrance ramps  Rural free-flow loop exit ramps  Rural parclo loop exit ramps  Rural diamond exit ramps  Urban free-flow loop entrance ramps  Urban parclo loop entrance ramps  Urban diamond entrance ramps  Urban free-flow loop exit ramps  Urban parclo loop exit ramps  Urban diamond exit ramps

1 - 71 Criteria for the selection of a ramp as a study site included:  Complete data on area type, ramp configuration, ramp type, and AADT available in HSIS data files  Meets the definition of one of the 12 ramp classifications listed above  No C-D roads or ramp-to-ramp junctions present  No lane additions or lane drops (except at ramp terminals)  No ramp metering present  No unusual or atypical features present The site selection used a quota-sampling method in which ramps were reviewed and selected until a maximum of 30 ramps of any particular type in each state were selected. For many ramp types, fewer than 30 ramps that met all of the required characteristics were found and selected. 4.5 Data Collection Data needed to classify and analyze each ramp were obtained, including all data needed to apply the HSM ramp crash prediction method. Data were obtained from the HSIS data files and from observation and measurement of the ramp in aerial photographs in Google Earth®. For measurement in aerial photographs, electronic pins were placed along each ramp at 50-ft intervals and ramp characteristics were recorded within each 50-ft interval. The data collected included:  Crash history data from 2007 to 2011, including crash severity level and crash type  Area type (urban/rural)  Ramp type (entrance or exit ramp)  Ramp configuration (parclo loop or free-flow loop)  Ramp average daily traffic volume (veh/day)  Ramp terminal characteristics  Ramp length (mi)  Number of lanes  Lane width (ft)  Ramp curve lengths (mi) and radii (ft)  Shoulder widths (right and left) (ft)  Freeway speed limit (mph)  Speed limit at crossroad ramp terminal (mph)  Ramp curve entry speeds (based on the HSM Chapter 19 speed prediction model)  Barrier presence and length (mi) (right and left)  Presence/absence of ramp metering The horizontal curve data were collected and the ramp segments were defined in such a manner that the number of curves on a ramp segment, m in Equation 13, was always equal to either 0 or

1 - 72 1, and the proportion of any curve on a ramp segment, ௖ܲ,௜ in Equation 13, was always either 0 or 1. Crash history data were obtained from the HSIS data files for a five-year period from 2007 to 2011, inclusive. The crash data obtained included all crashes that occurred on the “ramp proper,” and excluded crashes that occurred at the freeway or crossroad ramp terminals. 4.6 Descriptive Statistics for Ramps Included in the Study Basic descriptive statistics (number of ramps, minimum, maximum, mean, and median) for ramp length, AADT, smallest and largest curve radii, freeway speed limit, and minimum and maximum number of lanes of the validation ramps are presented, separately for each ramp configuration, in Table 30 through Table 37. The tables are organized by state, area type, and ramp type as follows:  Table 30. California, rural exit ramps  Table 31. California, rural entrance ramps  Table 32. California, urban exit ramps  Table 33. California, urban entrance ramps  Table 34. Washington, rural exit ramps  Table 35. Washington, rural entrance ramps  Table 36. Washington, urban exit ramps  Table 37. Washington, urban entrance ramps

1 - 73 Table 30. Descriptive Statistics for California Rural Exit Ramps Parameter N Min Max Mean Median Diamond ramps (N = 30) Ramp length (mi) 30 0.17 0.41 0.25 0.23 AADT (veh/day) 30 200 6,400 1,647 1,200 Smallest curve radius (ft) 6 366 594 457 436 Largest curve radius (ft) 6 366 594 457 436 Freeway speed limit (mph) 30 70 70 70.0 70.0 Minimum number of lanes 30 1 1 1.0 1.0 Maximum number of lanes 30 1 2 1.4 1.0 Parclo loop ramps (N = 13) Ramp length (mi) 13 0.08 0.30 0.21 0.23 AADT (veh/day) 13 250 9,600 1,576 990 Smallest curve radius (ft) 13 85 242 135 144 Largest curve radius (ft) 13 85 421 174 145 Freeway speed limit (mph) 13 70 70 70.0 70.0 Minimum number of lanes 13 1 1 1.0 1.0 Maximum number of lanes 13 1 2 1.2 1.0 Free-flow loop ramps (N = 6) Ramp length (mi) 6 0.18 0.23 0.20 0.19 AADT (veh/day) 6 75 3,240 1,244 1,020 Smallest curve radius (ft) 6 137 190 166 168 Largest curve radius (ft) 6 137 197 167 170 Freeway speed limit (mph) 6 70 70 70.0 70.0 Minimum number of lanes 6 1 1 1.0 1.0 Maximum number of lanes 6 1 1 1.0 1.0

1 - 74 Table 31. Descriptive Statistics for California Rural Entrance Ramps Parameter N Min Max Mean Median Diamond ramps (N = 30) Ramp length (mi) 30 0.16 0.48 0.26 0.25 AADT (veh/day) 30 180 7,400 1,546 1,135 Smallest curve radius (ft) 2 654 665 660 660 Largest curve radius (ft) 2 654 665 660 660 Freeway speed limit (mph) 30 70 70 70.0 70.0 Minimum number of lanes 30 1 1 1.0 1.0 Maximum number of lanes 30 1 2 1.4 1.0 Parclo loop ramps (N = 30) Ramp length (mi) 30 0.11 0.31 0.20 0.19 AADT (veh/day) 30 100 4,170 1,321 840 Smallest curve radius (ft) 30 117 228 165 165 Largest curve radius (ft) 30 138 228 173 176 Freeway speed limit (mph) 30 65 70 69.8 70.0 Minimum number of lanes 30 1 1 1.0 1.0 Maximum number of lanes 30 1 2 1.3 1.0 Free-flow Loop ramps (N = 30) Ramp length (mi) 30 0.16 0.28 0.20 0.19 AADT (veh/day) 30 50 9,120 1,577 590 Smallest curve radius (ft) 30 78 191 148 153 Largest curve radius (ft) 30 139 322 192 170 Freeway speed limit(mph) 30 60 70 68.7 70.0 Minimum number of lanes 30 1 1 1.0 1.0 Maximum number of lanes 30 1 1 1.0 1.0

1 - 75 Table 32. Descriptive Statistics for California Urban Exit Ramps Parameter N Min Max Mean Median Diamond ramps (N = 30) Ramp length (mi) 30 0.16 0.33 0.22 0.22 AADT (veh/day) 30 640 17,300 4,845 4,050 Smallest curve radius (ft) 0 na na na na Largest curve radius (ft) 0 na na na na Freeway speed limit (mph) 30 60 60 60.0 60.0 Minimum number of lanes 30 1 2 1.1 1.0 Maximum number of lanes 30 1 2 1.8 2.0 Parclo loop ramps (N = 30) Ramp length (mi) 30 0.13 0.31 0.23 0.23 AADT (veh/day) 30 1,400 30,500 7,208 5,945 Smallest curve radius (ft) 30 73 326 146 144 Largest curve radius (ft) 30 73 366 159 145 Freeway speed limit (mph) 30 60 60 60.0 60.0 Minimum number of lanes 30 1 2 1.2 1.0 Maximum number of lanes 30 1 2 1.8 2.0 Free-flow loop ramps (N = 23) Ramp length (mi) 23 0.06 0.31 0.19 0.19 AADT (veh/day) 23 1,450 17,800 6,020 6,000 Smallest curve radius (ft) 23 73 191 130 130 Largest curve radius (ft) 23 122 305 161 144 Freeway speed limit (mph) 23 60 60 60.0 60.0 Minimum number of lanes 23 1 1 1.0 1.0 Maximum number of lanes 23 1 1 1.0 1.0 na – not applicable

1 - 76 Table 33. Descriptive Statistics for California Urban Entrance Ramps Parameter N Min Max Mean Median Diamond ramps (N = 24) Ramp length (mi) 24 0.14 0.31 0.23 0.24 AADT (veh/day) 24 550 15,680 4,571 3,600 Smallest curve radius (ft) 0 na na na na Largest curve radius (ft) 0 na na na na Freeway speed limit (mph) 24 60 60 60.0 60.0 Minimum number of lanes 24 1 1 1.0 1.0 Maximum number of lanes 24 1 2 1.5 2.0 Parclo loop ramps (N = 22) Ramp length (mi) 22 0.10 0.26 0.18 0.18 AADT (veh/day) 22 900 13,500 4,698 3,800 Smallest curve radius (ft) 22 89 252 144 140 Largest curve radius (ft) 22 89 389 160 150 Freeway speed limit (mph) 22 60 60 60.0 60.0 Minimum number of lanes 22 1 1 1.0 1.0 Maximum number of lanes 22 1 2 1.4 1.0 Free-flow loop ramps (N = 26) Ramp length (mi) 26 0.14 0.45 0.21 0.19 AADT (veh/day) 26 890 11,700 4,648 3,525 Smallest curve radius (ft) 26 109 190 143 142 Largest curve radius (ft) 26 127 292 155 144 Freeway speed limit (mph) 26 60 60 60.0 60.0 Minimum number of lanes 26 1 1 1.0 1.0 Maximum number of lanes 26 1 2 1.0 1.0 na – not applicable

1 - 77 Table 34. Descriptive Statistics for Washington Rural Exit Ramps Parameter N Min Max Mean Median Diamond ramps (N = 26) Ramp length (mi) 26 0.17 0.40 0.24 0.23 AADT (veh/day) 26 250 7,675 2,259 1,309 Smallest curve radius (ft) 13 272 963 610 565 Largest curve radius (ft) 13 440 1,106 753 762 Freeway speed limit (mph) 26 70 70 70.0 70.0 Minimum number of lanes 26 1 1 1.0 1.0 Maximum number of lanes 26 1 2 1.1 1.0 Parclo loop ramps (N = 4) Ramp length (mi) 4 0.23 0.41 0.32 0.32 AADT (veh/day) 4 294 3,049 1,563 1,454 Smallest curve radius (ft) 4 161 310 224 213 Largest curve radius (ft) 4 161 310 224 213 Freeway speed limit (mph) 4 70 70 70.0 70.0 Minimum number of lanes 4 1 1 1.0 1.0 Maximum number of lanes 4 1 2 1.3 1.0 Free-flow loop ramps (N = 0)

1 - 78 Table 35. Descriptive Statistics for Washington Rural Entrance Ramps Parameter N Min Max Mean Median Diamond ramps (N = 26) Ramp length (mi) 26 0.16 0.49 0.30 0.29 AADT (veh/day) 26 243 7,667 2,076 1,242 Smallest curve radius (ft) 12 143 972 463 444 Largest curve radius (ft) 12 143 972 551 542 Freeway speed limit (mph) 26 70 70 70.0 70.0 Minimum number of lanes 26 1 1 1.0 1.0 Maximum number of lanes 26 1 2 1.3 1.0 Parclo loop ramps (N = 6) Ramp length (mi) 6 0.16 0.38 0.26 0.28 AADT (veh/day) 6 301 2,716 1,634 1,523 Smallest curve radius (ft) 6 153 196 175 175 Largest curve radius (ft) 6 156 198 183 193 Freeway speed limit (mph) 6 70 70 70.0 70.0 Minimum number of lanes 6 1 1 1.0 1.0 Maximum number of lanes 6 1 2 1.2 1.0 Free-flow loop ramps (N = 2) Ramp length (mi) 2 0.24 0.26 0.25 0.25 AADT (veh/day) 2 209 2,350 1,280 1,280 Smallest curve radius (ft) 2 193 199 196 196 Largest curve radius (ft) 2 296 304 300 300 Freeway speed limit (mph) 2 70 70 70.0 70.0 Minimum number of lanes 2 1 1 1.0 1.0 Maximum number of lanes 2 1 1 1.0 1.0

1 - 79 Table 36. Descriptive Statistics for Washington Urban Exit Ramps Parameter N Min Max Mean Median Diamond ramps (N = 22) Ramp length (mi) 22 0.14 0.37 0.25 0.24 AADT (veh/day) 22 211 10,692 5,152 4,898 Smallest curve radius (ft) 6 634 1,148 896 886 Largest curve radius (ft) 6 719 1,148 931 931 Freeway speed limit (mph) 22 60 70 61.8 60.0 Minimum number of lanes 22 1 1 1.0 1.0 Maximum number of lanes 22 1 2 1.8 2.0 Parclo loop ramps (N = 11) Ramp length (mi) 11 0.15 0.30 0.23 0.24 AADT (veh/day) 11 806 16,306 5,489 3,481 Smallest curve radius (ft) 11 84 223 155 156 Largest curve radius (ft) 11 104 480 194 158 Freeway speed limit (mph) 11 60 70 60.9 60.0 Minimum number of lanes 11 1 1 1.0 1.0 Maximum number of lanes 11 1 2 1.4 1.0 Free-flow loop ramps (N = 6) Ramp length (mi) 6 0.16 0.41 0.26 0.23 AADT (veh/day) 6 933 8,133 4,253 4,602 Smallest curve radius (ft) 6 170 237 198 201 Largest curve radius (ft) 6 170 283 208 208 Freeway speed limit (mph) 6 60 60 60.0 60.0 Minimum number of lanes 6 1 1 1.0 1.0 Maximum number of lanes 6 1 1 1.0 1.0

1 - 80 Table 37. Descriptive Statistics for Washington Urban Entrance Ramps Parameter N Min Max Mean Median Diamond ramps (N = 22) Ramp length (mi) 22 0.14 0.40 0.29 0.28 AADT (veh/day) 22 208 11,690 5,166 4,829 Smallest curve radius (ft) 5 351 1,003 807 908 Largest curve radius (ft) 5 619 1,003 861 908 Freeway speed limit (mph) 22 60 70 61.8 60.0 Minimum number of lanes 22 1 1 1.0 1.0 Maximum number of lanes 22 1 2 1.4 1.0 Parclo loop ramps (N = 17) Ramp length (mi) 17 0.12 0.33 0.22 0.20 AADT (veh/day) 17 492 12,136 4,241 2,931 Smallest curve radius (ft) 17 101 233 153 155 Largest curve radius (ft) 17 101 250 161 161 Freeway speed limit (mph) 17 60 70 61.8 60.0 Minimum number of lanes 17 1 1 1.0 1.0 Maximum number of lanes 17 1 2 1.2 1.0 Free-flow loop ramps (N = 5) Ramp length (mi) 5 0.19 0.30 0.24 0.25 AADT (veh/day) 5 2,608 20,887 11,723 10,836 Smallest curve radius (ft) 5 155 259 207 226 Largest curve radius (ft) 5 170 288 234 227 Freeway speed limit (mph) 5 60 60 60.0 60.0 Minimum number of lanes 5 1 1 1.0 1.0 Maximum number of lanes 5 1 1 1.0 1.0 Crash counts (for crashes of all types) over the five-year period from 2007 to 2011 were tallied by severity level for each ramp in the validation database. Statistics for total and fatal-and-injury (FI) crash counts over the entire five-year period, by state, area type, ramp type, and ramp configuration, are presented in the following section. These crashes are referred to as observed crash frequencies in subsequent discussions. 4.7 Application of HSM Ramp Crash Prediction Method The HSM ramp crash prediction method was applied to determine the predicted crash frequencies for individual ramps and for specific ramp classifications as a whole. Crash predictions were made separately for FI and PDO crashes, and total predicted crashes were computed as the sum of these individual predictions. These crash predictions were calculated using a program written in SAS®, equivalent to ISATe, developed for this research to apply the HSM Chapter 19 procedure. As in the case of the observed crash data, the predicted crash

1 - 81 frequencies addressed only the “ramp proper” portion of each ramp, and not crashes attributed to either the freeway or crossroad ramps terminals. 4.8 Summary Comparison of Predicted and Observed Crash Frequencies Table 38 (total crashes) and Table 39 (FI crashes) compare the predicted and observed crash frequencies for specific combinations of area type, ramp type, and ramp configuration. The comparisons are made both in terms of predicted and observed crash frequencies per year for a group of ramps as a whole, but also as crash rates normalized for the effects of ramp length and AADT [equivalent to the crash rate per million vehicle miles of travel (MVMT)]. The crash statistics provide a first look at the relative safety performance of specific ramp types. Crash rates per MVMT are appropriate to consider here because the statistical analysis presented in the next section of the report normalizes crash frequencies by MVMT. The tables include, for each state and ramp classification:  Number of ramps for the specific state and ramp classification combination  MVMT per year  Predicted crash frequency per year for all ramps combined  Predicted crash rate per MVMT  Observed crash frequency per year for all ramps combined  Observed crash rate per MVMT  Ratio of predicted crash rate to observed crash rate  Ratio of predicted crash rate for loop ramps to predicted crash rate for diamond ramps  Ratio of observed crash rate for loop ramps to observed crash rate for diamond ramps

1 - 82 Table 38. Total Crash Statistics by State and Ramp Classification State Ramp classification Number of ramps MVMT per year Predicted Observed Crash rate ratios Area type Ramp type Ramp configuration Number of total crashes (per year) Total crash rate (per MVMT) Number of total crashes (per year for 2007-2011) Total crash rate (per MVMT) Ratio of predicted to observed Ratio of loop to diamond (predicted) Ratio of loop to diamond (observed) C al ifo rn ia Rural Exit Diamond 30 4.73 4.54 0.96 4.0 0.85 1.14 na na Parclo 13 1.87 15.79 8.43 3.6 1.92 4.39 8.78 2.27 Free flow 6 0.56 12.11 21.43 1.6 2.83 7.57 22.32 3.35 All loops 19 2.44 27.90 11.44 5.2 2.13 5.36 11.92 2.52 Entrance Diamond 30 4.98 3.54 0.71 0.6 0.12 5.90 na na Parclo 30 2.88 7.67 2.66 3.2 1.11 2.40 3.74 9.21 Free flow 30 3.59 13.92 3.88 2.8 0.78 4.97 5.46 6.48 All loops 60 6.47 21.59 3.34 6.0 0.93 3.60 4.69 7.69 Urban Exit Diamond 30 28.64 25.62 0.89 5.2 0.18 4.93 na na Parclo 30 18.04 110.18 6.11 29.8 1.65 3.70 6.83 9.10 Free flow 23 10.21 95.90 9.39 20.0 1.96 4.79 10.50 10.79 All loops 53 28.25 206.08 7.30 49.8 1.76 4.14 8.16 9.71 Entrance Diamond 24 8.67 6.71 0.77 5.0 0.58 1.34 na na Parclo 20 6.05 20.57 3.40 5.6 0.93 3.67 4.39 1.61 Free flow 26 9.64 38.68 4.01 6.2 0.64 6.24 5.18 1.12 All loops 46 15.69 59.25 3.78 11.8 0.75 5.02 4.88 1.30 W as hi ng to n Rural Exit Diamond 26 4.82 3.77 0.78 5.8 1.20 0.65 na na Parclo 4 0.63 3.65 5.82 1.8 2.87 2.03 7.43 2.38 All loops 4 0.63 3.65 5.82 1.8 2.87 2.03 7.43 2.38 Entrance Diamond 26 5.44 4.83 0.89 2.6 0.48 1.86 na na Parclo 6 0.90 2.32 2.59 0.6 0.67 3.87 2.92 1.40 Free flow 2 0.23 0.53 2.36 0.2 0.89 2.66 2.65 1.85 All loops 8 1.12 2.86 2.54 0.8 0.71 3.57 2.87 1.49 Urban Exit Diamond 22 10.28 7.61 0.74 6.0 0.58 1.27 na na Parclo 11 4.32 29.10 6.73 9.2 2.13 3.16 9.08 3.64 Free flow 6 2.09 17.95 8.58 6.8 3.25 2.64 11.58 5.57 All loops 17 6.42 47.05 7.33 16.0 2.49 2.94 9.90 4.27 Entrance Diamond 22 11.62 7.65 0.66 4.8 0.41 1.59 na na Parclo 17 5.26 19.81 3.77 4.8 0.91 4.13 5.72 2.21 Free flow 5 5.06 12.97 2.56 5.2 1.03 2.49 3.90 2.49 All loops 22 10.32 32.78 3.18 10.0 0.97 3.28 4.83 2.35 na – not applicable

1 - 83 Table 39. Fatal-and-Injury Crash Statistics by State and Ramp Classification State Ramp classification Number of ramps MVMT per year Predicted Observed Crash rate ratios Area type Ramp type Ramp configuration Number of total crashes (per year) Total crash rate (per MVMT) Number of total crashes (per year for 2007-2011) Total crash rate (per MVMT) Ratio of predicted to observed Ratio of loop to diamond (predicted) Ratio of loop to diamond (observed) C al ifo rn ia Rural Exit Diamond 30 4.73 2.16 0.46 1.8 0.38 1.20 na na Parclo 13 1.87 6.89 3.68 1.4 0.75 4.92 8.07 1.96 Free flow 6 0.56 5.22 9.24 0.2 0.35 26.10 20.26 0.93 All loops 19 2.44 12.11 4.97 1.6 0.66 7.57 10.89 1.73 Entrance Diamond 30 4.98 1.51 0.30 0.0 0.00 nc na na Parclo 30 2.88 3.11 1.08 0.6 0.21 5.18 3.56 nc Free flow 30 3.59 5.56 1.55 1.0 0.28 5.56 5.12 nc All loops 60 6.47 8.67 1.34 1.6 0.25 5.42 4.43 nc Urban Exit Diamond 30 28.64 11.69 0.41 1.8 0.06 6.49 na na Parclo 30 18.04 42.36 2.35 11.4 0.63 3.72 5.75 10.06 Free flow 23 10.21 40.76 3.99 6.0 0.59 6.79 9.78 9.35 All loops 53 28.25 83.12 2.94 17.4 0.62 4.78 7.21 9.80 Entrance Diamond 24 8.67 2.93 0.34 1.6 0.18 1.83 na na Parclo 20 6.05 8.15 1.35 2.0 0.33 4.08 3.99 1.79 Free flow 26 9.64 15.25 1.58 1.4 0.15 10.89 4.68 0.79 All loops 46 15.69 23.40 1.49 3.4 0.22 6.88 4.42 1.17 W as hi ng to n Rural Exit Diamond 26 4.82 1.69 0.35 2.2 0.46 0.77 na na Parclo 4 0.63 1.53 2.44 0.8 1.28 1.92 6.98 2.79 All loops 4 0.63 1.53 2.44 0.8 1.28 1.92 6.98 2.79 Entrance Diamond 26 5.44 1.90 0.35 1.0 0.18 1.90 na na Parclo 6 0.90 0.92 1.03 0.2 0.22 4.62 2.95 1.21 Free flow 2 0.23 0.21 0.91 0.0 0.00 nc 2.61 0.00 All loops 8 1.12 1.13 1.01 0.2 0.18 5.66 2.88 0.97 Urban Exit Diamond 22 10.28 3.29 0.32 1.6 0.16 2.05 na na Parclo 11 4.32 11.63 2.69 2.4 0.56 4.85 8.41 3.56 Free flow 6 2.09 7.66 3.66 2.0 0.96 3.83 11.45 6.14 All loops 17 6.42 19.29 3.01 4.4 0.69 4.38 9.40 4.40 Entrance Diamond 22 11.62 3.17 0.27 1.0 0.09 3.17 na na Parclo 17 5.26 7.97 1.52 1.0 0.19 7.97 5.56 2.21 Free flow 5 5.06 5.20 1.03 0.8 0.16 6.50 3.78 1.84 All loops 22 10.32 13.18 1.28 1.8 0.17 7.32 4.69 2.03 na – not applicable; nc – not calculated (division by zero)

1 - 84 4.9 Statistical Comparison of Predicted and Observed Crash Frequencies This section describes the statistical analysis conducted to compare the crash frequencies predicted using the HSM ramp crash prediction method to the crashes observed at the 439 ramps in the study database. Key issues addressed in this analysis include:  How accurately does the HSM ramp crash prediction method predict crashes for diamond ramps and loop ramps?  Are the crash predictions more accurate for some types of ramps than for others?  Can any inaccuracies in the ramp crash predictions be accounted for by calibration of the HSM ramp crash prediction method? Predicted and observed crash rates for each ramp configuration as described in the previous section formed the basis for the statistical approach to validate the HSM ramp model. Crash rates, expressed in number of crashes per year per MVMT, rather than crash counts, were used to account for varying traffic volumes and ramp segment length across ramps. Next, to compare two rates, one typically calculates the ratio of the two rates, not their difference, and compares that ratio to one. Thus, if the HSM model predicts crashes accurately, then the ratio of predicted over observed crash rates should be close to one for a given ramp configuration. In their raw form, these are the ratios shown in the last column of Table 38 (total crashes) and Table 39 (FI crashes). The statistical model developed to test this hypothesis is described next. A negative binomial (NB) regression model to predict crash rates as a function of state, ramp type, ramp configuration, and an indicator variable to indicate whether the crashes were predicted or observed was developed using a generalized linear mixed model with a negative binomial distribution and a logit link. This model allows for estimating the ratio of predicted over observed crash rates for each ramp configuration and test whether that ratio is significantly different from one. The GLIMMIX procedure of SAS (2011) was used for all analyses. All analyses were done separately for total and FI crashes. Because of the small number of free-flow loop ramps in Washington, the research team decided for purposes of this analysis to combine free-flow loop and parclo loop ramps into a single category, resulting in only two ramp configurations for validation: loop ramps and diamond ramps. For consistency, this was also done for California ramps. In a first step, the regression models were built separately for each state. Each model included the following factors and interactions:  Area type (rural/urban)  Ramp type (entrance/exit)  Ramp configuration (diamond/loop)  Indicator variable for predicted vs. observed crashes  All relevant two-way and three-way interactions

1 - 85 Backward elimination was used to identify which factors and interactions were statistically significant at the 5-percent level. Since the objective is to estimate the ratios of predicted over observed crash rates separately for each ramp configuration, ramp configuration, the predicted/observed indicator variable, and their interaction were kept in the models, whether they were statistically significant at the 5-percent level or not. Table 40 shows the p-values associated with the main effects and any significant interactions remaining in the four regression models. Table 40. Statistical Significance Levels Associated with First Set of Negative Binomial Regression Models Severity State Factor or interaction P-value Significant? Total California Area type (rural/urban) 0.16 No Ramp type (On/Off) 0.016 At 5% level Ramp type × ramp configuration 0.024 At 5% level Washington Area type (rural/urban) 0.27 No Ramp type (on/off) 0.055 At 10% level Fatal and injury California Area type (rural/urban) 0.26 No Ramp type (rural/urban) 0.0005 At 5% level Washington Area type (rural/urban) 0.31 No Ramp type (rural/urban) 0.0041 At 5% level As shown in Table 40, area type was not statistically significant at the 5-percent level in any of the four models, while ramp type was always significant or nearly significant at the 5-percent level. Based on these results, another set of models were developed as explained next. In a second step, NB regression models were built separately for each state, ramp type, and severity, resulting in eight individual NB regression models. Rural and urban ramps were combined in these models based on the results shown in Table 40. The regression models are of the following general form: NTot or FI = MVMTr exp[a + b×IRC + c×IPO + d×(IRC × IPO)] (14) where: NTot or FI = number of total or FI crashes per year on the ramp IRC = indicator variable for ramp configuration; 0 if diamond ramp; 1 if loop ramp IPO = indicator variable for predicted vs. observed crashes; 0 if observed; 1 if predicted MVMTr = exposure in million vehicles miles traveled on the ramp per year = Lr×AADTr×365/106 Lr = ramp length (mi) AADTr = average annual daily traffic volume on the ramp segment (veh/day) a, b, c, d = regression coefficients to be estimated from the model Table 41 (total crashes) and Table 42 (FI crashes) show the final results from the eight NB regression models. The tables present sample size, predicted crash rate, observed crash rate, and the estimate and 95-percent confidence limits of the ratio of the predicted to observed crash rates for diamond and loop, off and on ramps. The last column in each table indicates whether the ratio of predicted over observed crash rates is statistically significant at the 5-percent level. This conclusion is based on whether the 95-percent

1 - 86 confidence interval of the ratio estimates includes 1. In other words, if the interval includes 1, then there is not enough evidence to conclude that the predicted crash rate is statistically significantly different from the observed crash rate at the 95-percent confidence level. For example, for California diamond off-ramps, the ratio of predicted to observed crash rates for total crashes is 2.39 with a 95-percent confidence interval of [0.99, 5.75]. Since the interval includes 1, there is not enough evidence to conclude that the predicted total crash rate is significantly different from the observed crash rate. In contrast, for California exit loop ramps, the total crash rate ratio is estimated at 4.78 with a 95-percent confidence interval of [3.39, 6.76]. Since this interval does not include 1, this result indicates that the predicted total crash rate is statistically significantly higher than the observed total crash rate. Table 41. Final Comparisons of Predicted to Observed Total Crash Rates State Ramp type Ramp configuration Number of ramps Total crash rate per MVMT Ratio of predicted to observed total crash rate 95% Confidence limits of ratio Ratio significantly different from 1 at 5% significance level? Predicted Observed Lower Upper CA Off Diamond 60 0.88 0.37 2.39 0.99 5.75 No Loop 72 8.60 1.80 4.78 3.39 6.76 Yes On Diamond 54 1.18 0.86 1.37 0.64 2.91 No Loop 106 3.69 1.49 2.48 1.65 3.72 Yes WA Off Diamond 48 1.40 1.36 1.03 0.56 1.90 No Loop 21 6.57 2.45 2.68 1.53 4.68 Yes On Diamond 48 0.73 0.43 1.69 0.67 4.21 No Loop 30 3.12 0.94 3.30 1.66 6.55 Yes Table 42. Final Comparisons of Predicted to Observed Fatal-and-Injury Crash Rates State Ramp type Ramp configuration Number of ramps FI crash rate per MVMT Ratio of predicted to observed FI crash rate 95% Confidence limits of ratio Ratio significantly different from 1 at 5% significance level? Predicted Observed Lower Upper CA Off Diamond 60 0.42 0.13 3.29 1.11 9.77 Yes Loop 72 3.10 0.62 4.97 3.04 8.15 Yes On Diamond 54 0.42 0.17 2.41 0.53 10.98 No Loop 106 1.39 0.36 3.87 1.77 8.46 Yes WA Off Diamond 48 0.33 0.25 1.34 0.35 5.16 No Loop 21 2.94 0.75 3.94 1.50 10.36 Yes On Diamond 48 0.37 0.17 2.16 0.53 8.76 No Loop 30 1.15 0.33 3.50 1.10 11.11 Yes

1 - 87 4.10 Discussion of Results The analysis was conducted with models that predict crash frequencies (or crash counts per year). Because the total exposure (MVMT) was used as an offset in the analysis, the results can be stated in terms of crash rates per MVMT. The predicted crash rates from HSM procedures for both diamond and loop ramps are higher than the observed crash rates. For total crashes (FI plus PDO), this overprediction of crash rates is statistically significant for loop ramps in all cases, but is not statistically significant for diamond ramps (see Table 41). For FI crashes, the overprediction of crash rates is statistically significant for loop ramps in all cases and is also statistically significant for diamond ramps in one of four cases analyzed (diamond exit ramps in California, see Table 42). Table 41 and Table 42 make clear that, for both California and Washington ramps, the HSM method predicts more crashes than were observed. This is not, in itself, surprising. There are often differences between the number of crashes predicted by HSM methods and observed crashes, and these are addressed through the calibration process. Indeed, the ratio of predicted to observed crashes shown in Table 41 and Table 42 is simply the inverse of the traditional HSM calibration factor. Analysis of contrasts in the data indicates that crash counts for diamond ramps in California are overpredicted by a factor of 2.4 (statistically significant), and diamond ramps in Washington are overpredicted by a factor of 1.4 (not statistically significant). Except for one specific case (discussed above), the overprediction for crashes for diamond ramps is not statistically significant (i.e., the overprediction could be simply the result of chance variations). A similar analysis found that crash rates for loop ramps in California are overpredicted by a factor of 4.5 (statistically significant) and loop ramps in Washington are overpredicted by a factor of 2.9 (statistically significant). This overprediction of loop ramp crashes is statistically significant in both states (i.e., it is so large that it is unlikely to have resulted from chance alone.) The comparison of the overprediction levels in the preceding paragraph suggest that loop ramp crashes in California are overpredicted in comparison to diamond ramps, by a factor of 1.9, and loop ramp crashes in Washington are overpredicted in comparison to diamond ramps, by a factor of 2.1. These results clearly imply that the HSM prediction models for ramp crashes do a better job of predicting diamond ramp crashes than predicting loop ramp crashes. Since the major difference in geometric design between these two types of ramps is longer and sharper curves on loop ramps than on diamond ramps, it is likely that the HSM prediction models for ramp crashes are overrepresenting the true effects of horizontal curvature on ramp crashes. As noted above, the overprediction of the effect of curvature on ramps can be compensated for by calibration of the models. Indeed, the HSM calibration procedure was developed for precisely this purpose. However, the research results indicate that it is vitally important to calibrate the HSM ramp crash prediction models separately for diamond and loop ramps. Otherwise, the loop ramps will appear to have unrealistically high crash frequencies relative to the diamond ramps; this could have the effect of discouraging designers from using loop ramps in situations where they would, in fact, perform well.

1 - 88 In summary, the analysis results indicate that the HSM ramp crash prediction models can be applied to both diamond and loop ramps, but that separate calibration for diamond and loop ramps is necessary to make accurate comparisons between the safety performances of these different ramp types.  

1 - 89 Section 5. Design Guidance The primary components of a ramp include the freeway mainline ramp terminal (i.e., a speed- change lane), the ramp proper (i.e., the turning roadway), and the crossroad ramp terminal. This research focused on developing improved design guidance for the ramp proper component of the loop ramp, taking into consideration the terminals on either side of the ramp proper. This section of the report provides design guidance related to the ramp proper of loop ramps based on the findings from this research and existing design policies. The design guidance is applicable to loop ramps at service interchanges in both urban and rural areas. Topics addressed in this section include:  Design vehicles for loop ramp proper  Recommended lane and shoulder widths for loop ramp proper  Multi-lane ramps  Safety prediction of design alternatives Appendix B provides recommended changes to the text of the 2011 Green Book for inclusion in the next edition, based on the findings and conclusions of this research. 5.1 Design Vehicles for Loop Ramp Proper When selecting the design vehicle or vehicles for the loop ramp proper, two key issues should be considered. First, the horizontal alignment of the loop ramp should be designed to provide reasonable margins of safety against skidding and rollover for all vehicle types. Second, the roadway should be designed to accommodate the turning capabilities of vehicles. This section (Section 5.1) address the first issue in more detail, while Section 5.2 on recommended lane widths and shoulder widths for loop ramps addresses the second issue in more detail. Design values for freeway mainline ramp terminals (minimum acceleration and deceleration lane lengths) provided in the 2011 Green Book are based upon the performance capabilities of passenger vehicles (Torbic et al., 2012). In a recent study of freeway mainline ramp terminals, Torbic et al. (2012) concluded that passenger vehicles should remain the principal design vehicle for freeway mainline ramp terminals, consistent with current AASHTO policy. If acceleration lane lengths were designed to accommodate the acceleration capabilities of trucks, minimum acceleration lane lengths would need to be about 1.8 times greater than the minimum acceleration lane lengths given in the 2011 Green Book (Harwood et al., 2003; Gattis et al., 2008). Some (Ervin et al., 1986; Firestine et al., 1989) have recommended that trucks require deceleration lane lengths 15 to 50 percent longer than the minimum deceleration lane lengths in the Green Book. However, minimum deceleration lane lengths are based on deceleration rates of which trucks are capable (Harwood et al., 2003; Torbic et al., 2012). In recent research on sharp horizontal curves with steep grades, it was found that current horizontal curve design policy appears to provide reasonable lateral margins of safety against skidding for both passenger vehicles and trucks for design speeds greater than 20 mph; however, for design speeds of 20 mph

1 - 90 and less the current approach to horizontal curve design generally overestimates the margins of safety against skidding and rollover for all vehicle types. Considering that the three primary components of a ramp (the freeway mainline ramp terminal, the ramp proper, and the crossroad ramp terminal) should be designed to function in a coordinated fashion as a single entity, it seems most reasonable to recommend that the ramp proper of a loop ramp be designed based on speed, acceleration, and deceleration capabilities of passenger vehicles operating along the ramp which is consistent with current AASHTO policy on designing the freeway mainline ramp terminal component of the ramp. If there is a concern with this recommendation, it would most likely be associated with exit loop ramps, and trucks transitioning from the deceleration lane to the ramp proper at speeds too high for the design and potentially rolling over or skidding off the roadway. If the ramp proper of an exit loop ramp is designed to accommodate the maximum speeds of passenger vehicles, the margins of safety against skidding and rollover may be low for trucks, especially for design speeds of 20 mph or less. However, the ramp proper will be designed to accommodate the vehicle dynamic capabilities of trucks as speed data from this research show that truck speeds are slower on exit loop ramps than passenger vehicle speeds. In summary, consistent with current AASHTO policy in the 2011 Green Book for determining minimum lengths of freeway mainline ramp terminals and conclusions set forth in NCHRP Report 730 (Torbic et al., 2012), it is recommended that passenger vehicles remain the design vehicle for the design of loop ramps. 5.2 Recommended Lane and Shoulder Widths for Loop Ramps The speed prediction and lane positioning models developed as part of this research, along with the safety prediction procedures from ISATe, were used to develop recommended lane and shoulder widths for the ramp proper of entrance and exit loop ramps. The general steps for developing the design recommendations for both entrance and exit loop ramps were: 1. Input a range of design parameters into the speed prediction models to identify the lane and shoulder width combinations that resulted in predicted speeds approximately equal to the design speed. 2. Input the lane and shoulder width combinations identified in Step 1 into ISATe to identify those lane and shoulder width combinations that are predicted to cause the least amount of crashes. 3. Input lane and shoulder width combinations remaining after Step 2 into the lane position models to identify combinations for which vehicles were not likely to encroach on the shoulder or adjacent lane when traversing the ramp. Combinations of lane and shoulder widths remaining after Step 3 are presented as recommended design conditions for a given design speed and curve radius. Each of these steps is discussed in more detail below, first for entrance ramps and then for exit ramps. The recommended lane and shoulder widths provided in this section are applicable to the

1 - 91 controlling curve of the loop ramp. It is recommended that the same, or similar, lane and shoulder widths will be used for all curves along the ramp proper. 5.2.1 Entrance Loop Ramps A sensitivity analysis was conducted to determine how changes in lane and inside and outside shoulder widths impacted vehicle speed, crash frequency, and lane keeping. The first step in the sensitivity analysis was to input a range of values for each design element into the speed prediction model to estimate speeds at the midpoint of the controlling curve (see Table 15). The key design elements included in this model are provided as follows, along with the range of values considered in the analysis:  Radius of controlling curve (100 to 300 ft, in 50-ft increments)  Lane width (14 to 20 ft, in 2-ft increments)  Inside (right) shoulder width (2 to 10 ft, in 1-ft increments)  Outside (left) shoulder width (2 to 8 ft, in 1-ft increments)  Vehicle type (passenger vehicles only)  Lane (for multi-lane ramps, inside lane only) All combinations of lane and shoulder width considered in the sensitivity analysis met these additional criteria:  Total roadway width (sum of lane width, outside shoulder width, and inside shoulder width) did not exceed 28 ft.  Inside shoulder width was greater than or equal to outside shoulder width. The range of values for each design element considered in this sensitivity analysis is consistent with the range of values used in developing the speed prediction model and the guidance specified in the Green Book (AASHTO, 2011), the ITE Freeway and Interchange Geometric Design Handbook (Leisch, 2006), and several state design policies. In particular, the guidance provided in this report was developed for curve radii in the range from 100 ft to 300 ft for design speeds ranging from 20 to 35 mph, which is the typical range of curve radii and design speeds in current design practices. The results of the speed prediction model were evaluated to determine the lane and shoulder width combinations that resulted in speeds at or near the design speed being considered. Specifically, combinations that resulted in expected speeds at the design speed or within 5 mph below the design speed were retained for Step 2. The second step in the analysis was to determine the impact of each remaining lane and shoulder width combination on safety. The ISATe crash prediction models were used to determine relative (rather than absolute) expected crash frequencies between lane and shoulder width combinations. This step was completed so that any combinations that resulted in substantially higher expected crash frequencies could be removed from consideration. However, it was found that nearly

1 - 92 every combination that came out of Step 1 for a given design speed and curve radius produced similar safety estimates. The third step in the sensitivity analysis involved inputting the design conditions into the lane position model developed as part of this research (see Table 25) to assess the expected impact of the design conditions on lane position. Lane and shoulder width combinations expected to result in vehicles encroaching onto the adjacent shoulders and/or travel lanes were removed from consideration. The design elements in the lane position prediction model for an entrance loop ramp include:  Lane width  Outside shoulder width  Superelevation  Grade (upgrade or downgrade)  Vehicle type (passenger vehicle or truck)  Lane [whether the vehicle was located in inside lane (Lane 1) or outside lane (Lane 2) of a multi-lane ramp] In this assessment, both upgrades and downgrades were considered along with superelevation rates of 4 and 8 percent. Superelevation rates of 12 percent were also tested, but 12 percent was outside of the superelevation range from which the lane position prediction model was developed. Results of the analysis indicated that none of the lane and shoulder width combinations considered are expected to result in either passenger vehicles or trucks encroaching onto the adjacent shoulders and/or travel lanes for superelevation rates of 4 and 8 percent; however, for some of the recommended design conditions, analyses associated with a superelevation of 12 percent indicate that vehicles may encroach onto the inside shoulder or inside lane (of a multi-lane ramp). The results of the lane position assessments with 12-percent superelevation were not considered directly in the design recommendations, since 12-percent superelevation was outside of the range for which the lane position prediction model was developed. Table 43 presents all the recommended lane and shoulder width combinations for curve radii ranging from 100 ft to 300 ft, for design speeds from 20 to 35 mph. With only a few exceptions, these recommended lane and shoulder width combinations for the controlling curve are expected to produce average operating speeds along the entrance loop ramp at or within 5 mph below the ramp design speed for the given curve radius. The use of lane and shoulder width combinations larger than the recommended design values provided in Table 43 for entrance loop ramps may result in travel speeds that exceed the ramp design speed for a given curve radius. However, because initial speeds of vehicles at crossroad ramp terminals are relatively low and drivers accelerate along entrance loop ramps, skidding and rollover events are less likely than on exit loop ramps where drivers enter at higher speeds. The use of lane and shoulder width combinations more restrictive than the recommended design values provided in Table 43 for a given ramp design speed and curve radius may result in travel speeds more than 5 mph below the ramp design speed; may negatively impact safety; and/or may result in vehicles encroaching onto the inside shoulder or adjacent inside travel lane.

1 - 93 Table 43. Recommended Lane and Shoulder Width Combinations for Controlling Curve on Entrance Loop Ramps Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 14 - 16 2 2 25 150 14 2 2 - 6 3 3 - 5 16 2 2 - 5 3 3 - 4 18 2 2 - 4 3 3 20 2 2 - 3 25 200 14 2 2 - 3 16 - 20 2 2 30 200 14 2 4 - 10 3 3 - 9 4 4 - 8 5 5 - 6 16 2 3 -10 3 3 - 8 4 4 - 7 5 5 - 6 18 2 2 - 8 3 3 - 7 4 4 - 6 5 5 20 2 2 - 6 3 3 - 5 4 4 30 250 14 2 2 - 8 3 3 - 6 4 4 - 5 16 2 2 - 7 3 3 - 5 4 4 18 2 2 - 6 3 3 - 4 20 2 2 - 5 3 3 - 4 30 300 14 2 2 - 5 3 3 16 2 2 - 4 3 3 18 2 2 - 3 20 2 2 35 300 14 3 4 - 10 4 4 - 9 5 5 - 8 16 2 5 - 10 3 3 - 9 4 4 - 8 5 5 - 7 6 6 18 2 4 - 8 3 3 - 7 4 4 - 6 5 5 20 2 3 - 6 3 3 - 5 4 4

1 - 94 The sum of lane and shoulder widths in Table 43 differ from those presented in the Green Book Table 3-29 for two main reasons. First, the values in Green Book Table 3-29 are specified for turning roadways at intersections, such as channelized-right turn lanes, and are referenced for use in loop ramp design, rather than being presented specifically for loop ramp design. Because of their intended use on turning roadways at intersections, design speed is not considered in the guidance. Vehicles are generally traveling very slowly, often with the anticipation of stopping, at these locations, rather than trying to decelerate from or accelerate to highway speeds within the turning roadway, as they would be on a loop ramp. On turning roadways at intersections, minimum roadway width is provided to accommodate offtracking trucks rather than to allow vehicles to reach a specific design speed. For this reason, lane width requirements increase with decreasing curve radius on turning roadways at intersections to accommodate offtracking, while lane width requirements increase with increasing curve radius on loop ramps to accommodate design speed. Second, the Green Book Table 3-29 provides separate lane width guidance for a scenario in which space for passing a stalled vehicle is provided and a scenario in which that space is not provided. No such distinction is presented in this guidance, and all recommendations instead are based on accommodating design speed, keeping vehicles in their lane, and minimizing negative safety impacts. Despite these differences in the basis of the pavement width guidance, the recommended total roadway widths presented in Table 43 are similar to design pavement widths for turning roadways from the Green Book and the design policies for loop ramps in several states. 5.2.2 Exit Loop Ramps A sensitivity analysis similar to the one described above for entrance loop ramps was conducted for exit ramps. The first step in the sensitivity analysis was to input a range of values for each design element into the speed prediction model to estimate speeds at the midpoint of the controlling curve (see Table 17). The key design elements in the model for exit ramps differs from those in the entrance ramp model. The design elements included in this model and the range of values considered for each are:  Radius of controlling curve (100 to 300 ft, in 50-ft increments)  Outside (left) shoulder width (2 to 8 ft, in 1-ft increments)  Type of curvature (simple or compound)  Type of mainline freeway ramp terminal (taper, parallel, drop lane, and weave) All combinations of lane and shoulder width considered in the sensitivity analysis met these additional criteria:  Total roadway width (sum of lane width, outside shoulder width, and inside shoulder width) did not exceed 28 ft.  Inside shoulder width was greater than or equal to outside shoulder width.

1 - 95 Combinations were evaluated for passenger vehicles on a single-lane ramp or the inside of a multi-lane ramp. The range of values used for each design element is consistent with the range used in developing the model and with guidance specified in the Green Book (AASHTO, 2011), the ITE Freeway and Interchange Geometric Design Handbook (Leisch, 2006), and several state design policies. The results of the speed prediction model were evaluated to determine the lane and shoulder width combinations that resulted in speeds at or near the design speed being considered. The goal was to identify combinations that resulted in expected speeds at the design speed or within 5 mph below the design speed for Step 2. The second step in the analysis was to determine the impact of each remaining lane and shoulder width combination on safety, and was conducted in the same manner as it was for entrance ramps. The ISATe crash prediction models were used to determine relative (rather than absolute) expected crash frequencies between lane and shoulder width combinations for a given design speed and curve radius. This step was completed so that any combinations that resulted in substantially higher expected crash frequencies could be removed from consideration. The variance in expected crash rates among recommended lane and shoulder width alternatives presented below is greater for the smaller curve radii than for the larger curve radii. However, it was found that nearly every combination that came out of Step 1 for a given design speed and curve radius produced similar safety estimates and most combinations were retained for Step 3. The third step in the sensitivity analysis involved inputting the design conditions into the lane position model developed as part of this research (see Table 26) to assess the expected impact of the design conditions on lane position. Lane and shoulder width combinations expected to result in vehicles encroaching onto the adjacent shoulders and/or travel lanes were removed from consideration. The key design elements in the lane position prediction model include:  Lane width,  Vehicle type (passenger vehicle or truck)  Type of mainline freeway ramp terminal (taper, parallel, drop lane, and weave)  Lane [whether the vehicle was located in inside lane (Lane 1) or outside lane (Lane 2) of a multi-lane ramp]  Superelevation In this assessment, superelevation rates of 4, 8, and 12 percent were tested. None of the recommended design conditions were found to result in either passenger vehicles or trucks encroaching onto the adjacent shoulders and/or travel lanes for any of the superelevation rates tested. For exit ramps, recommended lane and shoulder widths vary by type of horizontal curvature (i.e., simple or compound curves), as well as by design speed and curve radius. Table 44 presents recommended lane and shoulder widths for exit loop ramps with simple curves, and Table 45 presents similar information for exit loop ramps with compound curves, for curve radii ranging

1 - 96 from 100 ft to 300 ft and for design speeds from 20 to 35 mph. For exit loop ramps, the minimum recommended lane width is 16 ft. This is primarily based on the qualitative analysis results in which the exit ramps with the three highest percentages of encroachments and/or critical maneuvers had lane widths of 15.2 ft or less. Table 44. Recommended Lane and Shoulder Widths for Exit Loop Ramps – Simple Curves Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 16 2 8 - 10 18 6 - 8 20 4 - 6 25 150 16 2 8 - 10 18 6 - 8 20 4 - 6 25 200 16 2 8 - 10 18 6 - 8 20 4 - 6 30 200 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 30 250 16 2 8 - 10 18 6 - 8 20 4 - 6 35 300 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 5.2.3 Summary of Recommended Lane and Shoulder Widths Table 43 presents recommended lane and shoulder widths for entrance loop ramps; and Table 44 (simple curves) and Table 45 (compound curves) present the recommended lane and shoulder widths for exit loop ramps for curve radii ranging from 100 to 300 ft and for design speeds from 20 to 35 mph. These recommended lane and shoulder widths are expected to produce speeds at or below the ramp design speed for a given curve radius. The recommended lane and shoulder widths are applicable to the controlling curve of the loop ramp, and it is assumed that the same, or similar, lane and shoulder widths will be used for all curves along the ramp proper. The recommended lane and shoulder widths were developed in three general steps: 1. Apply the speed prediction models for estimating speeds at the midpoint of the controlling curve 2. Use ISATe to compare predicted crash frequencies for different alternatives

1 - 97 3. Apply the lane position model to remove alternatives that result in vehicles encroaching on the shoulder or adjacent lane Table 43 through Table 45 are presented for ease of application; however, designers can make use of the speed and lane position prediction models developed as part of this research and ISATe in a similar fashion to evaluate and assess different design alternatives in more detail. Table 45. Recommended Lane and Shoulder Widths for Exit Loop Ramps – Compound Curves Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 16 2 8 - 10 18 6 - 8 20 4 - 6 25 150 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 25 200 16 2 8 - 10 18 6 - 8 20 4 - 6 30 200 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 16 5 5 - 7 18 5 30 250 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 35 300 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 16 5 5 - 7 18 5

1 - 98 5.3 Multi-Lane Ramps It was found that vehicles travel approximately 1 to 2 mph faster in the outside lane of a multi- lane ramp compared to the inside lane. This is logical as vehicles traveling in the outside lane must travel faster to cover slightly longer travel distances and keep pace with vehicles on the inside lane of the ramp. It was also determined that vehicles in the outside lane of a multi-lane loop ramp are positioned slightly farther from the inside lane line (i.e., the lane line separating the outside lane from the inside lane) than vehicles on the inside lane are from the inside shoulder. In general vehicles in the outside lane are positioned 10 to 20 inches farther away from the inside lane line than vehicles traveling in the inside lane. Thus, concerns about vehicles traveling in the outside lane encroaching on the inside lane either due to following a tighter path around the curve or due to offtracking are not founded or supported by the data. At all of the study locations with multi-lane ramps, the inside lane widths ranged from 14.5 ft to 16.0 ft. The outside lane widths ranged from 11.75 ft to 12 ft on entrance loop ramps and 14 ft on exit loop ramps. Only a few trucks were observed on any of the multi-lane ramps; and in particular, very few trucks were observed traveling in the outside lane. Therefore, the available data was not sufficient to make any specific design recommendations regarding trucks on multi- lane entrance loop ramps. Based upon speeds and lane positions of passenger vehicles operating in the outside lane of multi-lane entrance loop ramps, it does not appear that any special design considerations are necessary for the design of multi-lane loop ramps to accommodate large differentials in speeds of vehicles traveling in the outside lane compared to the inside lane of a multi-lane loop ramp or to accommodate vehicles in the outside lane that significantly gravitate to the inside and encroach on the inside travel lane. Thus, 12-ft lane widths for the outside lanes of a multi-lane entrance loop ramp and 14-ft lane widths for the outside lanes of a multi-lane exit loop ramp are sufficient to accommodate traffic comprised primarily of passenger vehicles. If the outside lane is expected to accommodate a moderate to heavy volume of trucks, the width of the outside lane should be increased to reduce the potential for encroachment on the outside shoulder. In general, the outside lane width should be increased as necessary until it was equivalent to the width of the inside travel lane. Table 43 through Table 45 provide guidance on recommended lane and shoulder widths for entrance and exit loop ramps, respectively. 5.4 Safety Prediction of Design Alternatives As part of implementing the HSM ramp crash prediction methodology to estimate predicted and/or expected crash frequencies for individual ramps, it is recommended that separate calibration factors be calculated for diamond ramps and loop ramps. The HSM ramp crash prediction models were developed using a database that consisted of approximately 15 to 20 percent of information on loop ramps, with the rest of the database comprised of information for other types of ramps (e.g., diagonal ramps). The HSM crash prediction methodology for ramps does not include separate procedures for specific ramp configurations, such as diamond ramps and loop ramps. In a comparison of crash frequencies predicted using the HSM ramp crash prediction method to observed crashes, it was found that the HSM ramp crash prediction

1 - 99 methodology does a better job of predicting diamond ramp crashes than predicting loop ramp crashes. Therefore, separate calibration factors should be calculated for diamond ramps and loop ramps. This will allow for more accurate comparisons between the safety performance of these different ramp types.

1 - 100 Section 6. Conclusions and Recommendations for Future Research The objective of this research was to develop improved design guidance for interchange loop ramps. The research focused on developing improved design guidance for the ramp proper portion of the ramp, taking into consideration the connections on either side of the ramp proper. An observational field study of driver behavior and vehicle operations on the ramp proper of single-lane and multi-lane loop ramps was conducted to investigate (1) the relationship between key design elements of the ramp proper and vehicle speed and lane position and (2) the difference in vehicle performance on the ramp proper between single-lane and multi-lane loop ramps. In addition to the field study, a validation effort of the AASHTO Highway Safety Manual (HSM) (AASHTO, 2014) crash prediction method for ramps at freeway interchanges was conducted. This methodology was included in a recent supplement to the HSM. Because the methodology does not provide separate models for each ramp type, an investigation was performed to evaluate the applicability of the HSM ramp methodology to loop ramps in particular. This section of the report presents the general conclusions reached based upon the results of these studies, including a summary of the design guidance, and recommendations for future research. 6.1 Conclusions General conclusions from the research are presented for the following topics: entrance ramps, exit ramps, multi-lane ramps, HSM crash prediction method, and design guidance. Entrance Ramps  On entrance loop ramps, speeds of vehicles at the end of the controlling curve were found to be slightly higher than speeds at the midpoint of the controlling curve, suggesting that vehicles accelerate as they traverse the length of the controlling curve on the ramp proper.  Key roadway and cross-sectional design elements that significantly influence vehicle speeds at the midpoint of the controlling curve include curve radius, lane width, inside (right) shoulder width, and outside (left) shoulder width. When comparing the impact of lane width and shoulder widths on speeds, for a given incremental increase in width (e.g., 1 ft), shoulder widths have a greater influence on speeds than travel lane widths. This may be because as lane widths become sufficiently wide (e.g., 16, 18, 20 ft or more), the relative effect of lane width on speed becomes less in comparison to shoulder width.  Key roadway and cross-sectional design elements that significantly influence vehicle speeds at the end of the controlling curve include curve radius and outside shoulder

1 - 101 width. It is logical that the outside shoulder width influences vehicle speeds as drivers approach the end of the curve and begin looking for gaps in freeway traffic. A wider outside shoulder provides drivers with more “forgiveness” in lateral placement as the driving task shifts away from negotiating the curve and shifts toward looking for gaps.  Key roadway and cross-sectional design elements that significantly influence lane position at the midpoint of the controlling curve include lane width, outside shoulder width, superelevation, and grade. As lane width increases, vehicles tend to move farther away from the inside lane line (which is expected). However, they tend to move closer to the inside lane line as either outside shoulder width or superelevation increases. Also, vehicles are positioned closer to the inside lane line on upgrades than on downgrades.  As vehicles traverse an entrance loop ramp, the right tires of passenger vehicles and trucks are positioned approximately an equal distance from the inside lane line; so aside from offtracking issues associated with larger trucks, there are no major concerns associated with differences between the lane positions of trucks and passenger vehicles. Exit Ramps  On exit loop ramps, vehicle speeds were found to be slightly higher at the beginning of the controlling curve than at the midpoint of the controlling curve, suggesting that vehicles decelerate as they transition from the freeway mainline ramp terminal along the ramp proper.  The radius of the controlling curve is the only key roadway and/or cross-sectional design element that significantly influences vehicle speeds at the PC of the controlling curve. This is not surprising since drivers are just transitioning to the ramp proper and have had little time to process the site characteristics and context of the ramp. As vehicles proceed along the ramp proper and reach the midpoint of the controlling curve, other key roadway and cross-sectional design elements (in addition to curve radius) significantly influence vehicle speeds; these include outside shoulder width, type of curvature (simple or compound), and type of mainline freeway ramp terminal. It seems reasonable that vehicle speeds would increase as outside shoulder width increases because drivers may feel more comfortable driving at higher speeds where more recovery area is available on the outside of the ramp. Also, vehicle speeds are higher where the ramp proper is designed with a simple curve rather than a compound curver.  Trucks were found to travel slower than passenger vehicles on exit ramps. Thus, if the ramp proper is designed to accommodate the maximum predicted speeds of passenger vehicles, then the ramp proper will be designed with a greater margin of safety against truck rollovers and skidding.  Key roadway and cross-sectional design elements that significantly influence lane position at the midpoint of the controlling curve include lane width and type of freeway mainline ramp terminal. As lane width increases, vehicles tend to move farther away from the inside lane line (which is expected). Vehicles are positioned farther away from the inside lane line on ramps where the freeway lane is dropped as it transitions to the ramp proper compared to loop ramps following a taper- or parallel type of speed-change

1 - 102 lane or weave area. This may be due to fewer lane-changing maneuvers in the immediate vicinity of a freeway ramp that follows a lane drop.  Trucks are typically positioned farther away from the inside lane line than passenger vehicles, and most passenger vehicles are positioned within the travel lane and do not encroach on the inside shoulder. Thus, the lane positioning of trucks does not raise concerns about encroachment onto the inside shoulder of an exit ramp. Multi-Lane Ramps  Vehicles in the outside lane (or lanes) of a multi-lane loop ramp travel at speeds approximately 1 to 2 mph faster than vehicles traveling in the inside lane.  Vehicles traveling in the outside lane of a multi-lane loop ramp are positioned slightly farther from the inside lane line than vehicles traveling in the inside lane. HSM Ramp Crash Prediction Method  The HSM ramp crash prediction methodology is better at predicting diamond ramp crashes than predicting loop ramp crashes. Separate calibration factors for diamond ramps and loop ramps are needed for more accurate comparisons between the safety performance of these different ramp types. Design Guidance The design guidance is applicable to loop ramps at service interchanges in both urban and rural areas.  Consistent with current AASHTO policy in the 2011 Green Book for determining minimum lengths of freeway mainline ramp terminals, it is recommended that passenger vehicles remain the design vehicle for the design of loop ramps.  The alignment and cross section of a loop ramp should be designed so that vehicles traverse the loop ramp at or below the ramp design speed. For curve radii ranging from 100 to 300 ft for design speeds from 20 to 35 mph, recommended lane and shoulder widths for entrance and exit loop ramps are provided that are expected to induce speeds at or below the ramp design speed. For a given radius and ramp design speed, the recommended lane and shoulder widths are also expected to result in similar levels of safety, and vehicles are expected to stay within their intended travel lane. Alternatively, designers may use the speed and lane position prediction models, developed as part of this research, and ISATe to evaluate and design alignments and cross sections of loop ramps.  Based upon speeds and lane positions of vehicles operating in the outside lane of multi- lane loop ramps, no special design considerations are necessary for the design of multi- lane loop ramps to accommodate large differentials in speeds of vehicles traveling in the outside lane compared to the inside lane or to accommodate vehicles in the outside lane that significantly gravitate to the inside and encroach on the inside travel lane. Outside lane widths of 12-ft for multi-lane entrance loop ramp and 14-ft for multi-lane exit loop

1 - 103 ramp are sufficient to accommodate traffic comprised primarily of passenger vehicles. If the outside lane is expected to accommodate a moderate to high volume of trucks, the outside lane width should be increased.  As part of implementing the HSM ramp crash prediction methodology to estimate predicted and/or expected crash frequencies for individual ramps, separate calibration factors should be calculated for diamond ramps and loop ramps for more accurate comparisons between the safety performance of these different ramp types. 6.2 Recommendations for Future Research Recommendations for future research related to loop ramps are presented below.  Performance of trucks on multi-lane loop ramps: Few trucks were observed in the outside lane of the multi-lane loop ramps included in this study; thus, it was not possible to make specific design recommendations regarding trucks on multi-lane entrance loop ramps based on the limited dataset. It would be desirable to expand upon this research and further evaluate the difference in performance of single-lane and multi-lane ramps, focusing on the performance of trucks operating in the outside lanes. This gap in knowledge is viewed as high priority.  Capacity of loop ramps: The Highway Capacity Manual (HCM) does not provide a methodology to estimate the capacity of a loop ramp. A general rule of thumb is that a single-lane loop ramp (either an entrance or exit ramp) can handle up to about 1,500 to 2,000 veh/hr, and a two-lane loop ramp can handle up to 3,600 veh/hr (Leisch, 2006). Although there is no formal analysis procedure available to estimate the capacity of a loop ramp, this is generally acceptable to practitioners given that the overall capacity of a ramp is likely limited by the capacity of the ramp terminal connections rather than the capacity of the ramp proper. However, there is more that could be learned about the capacity of a loop ramp, for example, how the capacity of a loop ramp varies with the speed of the ramp. Future research should be conducted to further evaluate the capacity of loop ramps. This gap in knowledge is viewed as medium priority.  Practical size of loop ramps: The Green Book states in its discussion on cloverleaf interchanges (p. 10-48) that, “considering all factors, experience shows that the practical size of loops resolves into approximate radii of 30 to 50 m [100 to 170 ft] for minor movements on highways with design speeds of 80 km/h [50 mph] or less and 50 to 75 m [150 to 250 ft] for more important movements on highways with higher design speeds” (AASHTO, 2011). Yet the loop ramps in this study had radii between 150 and 312 ft. Research is needed to determine what is a “practical size” of a loop ramp, considering the tradeoffs between desired speed and required distance traveled, as well as consideration of construction and right-of-way costs; such research would document the experience of designers. This gap in knowledge is viewed as medium priority.  Speed profile models of loop ramps: The design guidance for recommended lane and shoulder width combinations for a given design speed and curve radius, presented herein, is based partially on speed prediction models that estimate the speed of vehicles

1 - 104 at the midpoint of the controlling curve for a loop ramp. Development of a more detailed speed profile model that predicts the speed of vehicles along the entire length of a loop ramp may be helpful to better inform the design of loop ramps. This gap in knowledge is viewed as low priority.  Difference in performance between single and multi-lane freeway mainline ramp terminals (acceleration or deceleration lanes): In the early phases of this research, the research team attempted to identify multi-lane loop ramps that carried the same number of lanes from the ramp proper all the way through the freeway mainline ramp terminals (acceleration or deceleration lanes), as NCHRP Report 730 (Torbic et al., 2012) notes that differences in the performance of freeway mainline ramp terminals with one lane and more than one lane are unknown. However, very few multi-lane loop ramp freeway mainline ramp terminals with two or more lanes were identified for evaluation. For nearly every multi-lane loop ramp included in the observational field study, the freeway mainline ramp terminal narrowed to a single lane prior to connecting with the freeway. Based on our initial findings, it appears that there are very few existing multi-lane loop ramps with multi-lane freeway mainline ramp terminals in the United States. Differences in the performance between single-lane and multi-lane freeway mainline ramp terminals are unknown and should be evaluated. This gap in knowledge is viewed as low priority.  Safety issues related to crossroad ramp terminals: For parclo-A/4 quad and parclo-B/4 quad crossroad terminals, the crash prediction methodology developed for NCHRP Project 17-45 and incorporated into the supplement of the HSM does not account for crashes associated with the free-flow merge/diverge movement of the loop ramp on the crossroad. The methodology only predicts crashes for those legs of the terminal that are yield, stop, or signal-controlled. Thus, the crash prediction methodology for crossroad ramp terminals within the HSM likely under-predicts the total number of crashes at parclo-A/4 quad and parclo-B/4 quad crossroad terminals. Therefore, future research should be conducted to address this issue. This gap in knowledge is viewed as low priority.  Safety issues related to freeway mainline ramp terminals: The crash prediction methodology developed for NCHRP Project 17-45 and incorporated in the supplement of the HSM for freeway mainline ramp terminals predicts crashes for both the freeway mainline ramp terminal and the adjacent freeway segment combined. Thus, crashes cannot be specifically assigned to the speed-change lane adjacent to the mainline freeway. This limitation is associated with the crash databases for ramps and speed- change lanes. Future research should be conducted to address this gap in knowledge. Because this limitation is associated with the format of available crash data, this gap in knowledge is viewed as low priority.

1 - 105   Section 7. References American Association of State Highway and Transportation Officials (AASHTO), A Policy on Geometric Design of Highways and Streets, Washington, DC, 2011. American Association of State Highway and Transportation Officials (AASHTO), Highway Safety Manual, Volumes 1 – 3, Washington, DC, 2010. American Association of State Highway and Transportation Officials (AASHTO), Highway Safety Manual, Supplement, Washington, DC, 2014. Arizona Department of Transportation, Roadway Design Manual, 2012. Bauer, K. M. and D. W. Harwood, Statistical Models of Accidents on Interchange Ramps and Speed-change lanes, Report FHWA-RD-97-106, Federal Highway Administration, Washington, DC, 1998. Bonneson, J. A. NCHRP Report 439: Superelevation Distribution Methods and Transition Designs. Transportation Research Board, Washington, DC, 2000. Bonneson, J. A., S. Geedipally, M. P. Pratt, and D. Lord, Safety Prediction Methodology and Analysis Tool for Freeways and Interchanges, Final Report for NCHRP Project 17-45, Texas Transportation Institute, 2012. California Department of Transportation (Caltrans), Highway Design Manual, 2012. Connecticut Department of Transportation (CTDOT), Roadway Design Guide, 2012. Ervin , R.D., R.L. Nisonger, C.C. MacAdam, and P.S. Fancher, Influence of Size and Weight Variables on the Stability and Control Properties of Heavy Trucks, Report No. FHWA/RD- 83/029, Federal Highway Administration, 1986. Firestine, M., H. McGee, and P. Toeg, Improving Truck Safety at Interchanges, Report No. FHWA-IP-89-024, Federal Highway Administration, September 1989. Gattis, J.L. M. Bryant, and L.K. Duncan, Acceleration Lane Design for Higher Truck Volumes, Report No. MBTC 2094/3003, Mack-Blackwell Transportation Center, University of Arkansas, U.S. Department of Transportation, 2008. Google Earth™ Mapping Service. Harwood, D. W., D. J. Torbic, K. R. Richard, W. D. Glauz, and L. Elefteriadou, Review of Truck Characteristics as Factors in Roadway Design, NCHRP Report 505, Transportation Research Board, 2003. Illinois Department of Transportation (IDOT), Bureau of Design & Environment Manual, 2012. Indiana Department of Transportation (INDOT), Design Manual, 2012.

1 - 106 Iowa Department of Transportation (Iowa DOT), Design Manual, 2010. Leisch, J.P., Freeway and Interchange Geometric Design Handbook, Institute of Transportation Engineers, Washington, D.C., 2006. Lord, D., and J. A. Bonneson, Estimating the Safety of Four Ramp Design Configurations, 3rd International Symposium on Highway Geometric Design, Chicago, Illinois, Transportation Research Board, 2005. Lundy, R. A. Effect of Ramp Type and Geometry on Accidents, California Department of Public Works, 1965. Maine Department of Transportation (MaineDOT), Highway Design Guide, 2004. Massachusetts Department of Transportation (MassDOT), Project Development & Design Guide, 2006. Michigan Department of Transportation (MDOT), Road Design Manual, 2012. Minnesota Department of Transportation (MnDOT), Road Design Manual, 2001. New Jersey Department of Transportation (NJDOT), Roadway Design Manual, 2011. Oregon Department of Transportation (ODOT), Highway Design Manual, 2012. SAS Institute Inc. SAS/STAT® 9.3 User's Guide. Cary, NC:SAS Institute Inc., 2011. Seyfried, R. K., and J. L. Pline, Guidelines for the Determination of Advisory Speeds, 2009. South Dakota Department of Transportation, Roadway Design Manual, 2012. Texas Department of Transportation (TxDOT), Roadway Design Manual, 2010. Transportation Research Board, Highway Capacity Manual 2010, National Research Council, Washington, DC, 2010. Torbic, D. J., J. M. Hutton, C. D. Bokenkroger, D. W. Harwood, D. K. Gilmore, M. M. Knoshaug, J. J. Ronchetto, M. A. Brewer, K. Fitzpatrick, S. T. Chrysler, and J. Stanley, Design Guidance for Freeway Mainline Ramp Terminals, NCHRP Report 730, Transportation Research Board, National Research Council, Washington, DC, 2012. Torbic, D. J., M. K. O’Laughlin, D. W. Harwood, K. M. Bauer, C. D. Bokenkroger, L. M. Lucas, J. R. Ronchetto, S. Brennan, E. Donnell, A. Brown, and T. Varunjikar, Superelevation Criteria for Sharp Horizontal Curves on Steep Grades, NCHRP Report 774, Transportation Research Board, National Research Council, Washington, DC, 2014. Twomey, J. M., M. L. Heckman, J. C. Hayward and R. J. Zuk, Accidents and Safety Associated with Interchanges, Transportation Research Record 1385, Transportation Research Board, National Research Council, Washington, DC, 1993.

1 - 107 Yates, J. G. Relationship Between Curvature and Accident Experience on Loop and Outer Connection Ramps, Highway Research Record 312, Highway Research Board, 1970

1A - 1 Appendix A Sites Included in Observational Field Study of Loop Ramps

1A - 2 This appendix provides aerial views of each loop ramp included in the observational study described in Section 3 of this report. More than one study site may be illustrated in a single figure.

1A - 3

1A - 4

1A - 5

1A - 6

1A - 7

1A - 8

1A - 9

1A - 10

1B-1 Appendix B Recommended Changes for Consideration in the Next Edition of the Green Book

1B-2 This appendix provides recommended changes to the 2011 edition of the Green Book for consideration in the next edition of the Green Book. The recommendations are based on the findings and conclusions of this research. Recommended text is specified for relevant sections of the Green Book as follows. Recommended Text for the next edition of the Green Book (Page 10-48) Cloverleafs … The advantages of increased speed should be weighed against the disadvantages of increased travel time, distance, and right-of-way. It should also be noted that large trucks may not be able to operate as efficiently on curves with smaller radii curves. Considering all factors, experience shows that the practical size of loops resolves into approximate radii of 30 to 50 m [100 to 170 ft] for minor movements on highways with design speeds of 80 km/h [50 mph] or less and 50 to 75 m [150 to 250 ft] for more important movements on highways with higher design speeds. A continuous additional lane is needed for deceleration, acceleration, and weaving between the on- and off-loop ramps. Additional structure width or length is usually needed for this lane. The cloverleaf involves weaving maneuvers as discussed in “Weaving Sections” of Section 10.9.5. The presence of weaving maneuvers is not objectionable when the left-turning movements are relatively light, but when the sum of traffic on two adjoining loops approaches about 1,000 vph, interference mounts rapidly, which results in a reduction in speed of through traffic. The weaving lengths presented in “Minimum Lengths Measuresd between Successive Ramp Terminals” of Figure 10-68 should be provided on low-volume cloverleaf interchanges. When the weaving volume in a particular weaving section exceeds 1,000 vph, the quality of service on the main facility deteriorates rapidly, thus generating a need to transfer the weaving section from the through lanes to a collector-distributor road. A loop rarely operates with more than a single line of vehicles, regardless of the roadway width, and thus has a design capacity limit of 800 to 1,200 vph, the higher figure being applicable only where there are no trucks and where the design speed for the ramp is 50 km/h [30 mph] or higher. Loop ramp capacity is, therefore, a major control in cloverleaf designs. Loops may be made to operate with two lanes abreast, but only by careful attention to design of the terminals and design for weaving, which would need widening by at least two additional lanes through the separation structure. To accomplish this type of design, the terminals should be separated by such great distances and the loop radii should be made so large that cloverleafs with two-lane loops generally are not economical from the standpoint of right-of-way, construction, cost, and amount of out-of- direction travel. Loops that operate with two lanes of traffic, therefore, are considered exceptional cases. …

1B-3 (Page 10-50) Partial Cloverleaf Ramp Arrangements … Figure 10-29 illustrates the manner in which the turning movements are made for various two- and three-quadrant cloverleaf arrangements. When ramps in two quadrants are adjacent and on the same side of the minor road, as shown in Figures 10-29A and 10-29BC, or diagonally opposite each other, as shown in Figures 10-29E and 10-29F, all turning movements to and from the major road are accomplished by right turns. Any decision between the arrangement in Figure 10-29A and its alternate arrangement (ramps in the other two quadrants) will depend on the predominant turning movements or the availability of right-of-way, or both. When the ramps in two quadrants are adjacent but on the same side of the major road (Figures 10-29B and 10-29D), four direct left turns fall on the major road. This arrangement and its alternate are the least desirable of the six possible arrangements, and their use should be avoided. … (Page 10-89) Loop Ramps (No changes recommended.) (Page 10-90) Two-Lane Loop Ramps (No changes recommended.) (Page 10-102) Ramp Traveled Way Widths Width and cross section—Ramp traveled way widths are governed by the type of operation, curvature, and volume and type of traffic. It should be noted that the roadway width for a turning roadway includes the traveled way width plus the shoulder width or equivalent offset outside the edges of the traveled way. Section 3.3.11 on “Widths for Turning Roadways at Intersections” may be referenced for additional discussion on the treatments at the edge of the traveled way. Recommended widths of travel lanes, left shoulders, and right shoulders of loop ramps are provided in Table 10-X1 for entrance loop ramps, Table 10-X2 for exit loop ramps with simple curves, and Table 10-X3 for exit loop ramps with compound curves. The values in these tables are based on research that observed and documented operating speeds on loop ramps of various radii and design speeds across the United States. The values for lane and shoulder width in Tables 10-X1 through 10-X3 were developed to produce operating speeds within -8.0 to 0 km/h [-5 to 0 mph] of the corresponding design speed.

1B-4 For a given radius (measured to the inside of the traveled way) and ramp design speed, the recommended lane and shoulder widths are expected to result in similar levels of safety, and vehicles are expected to stay within their intended travel lane. The total roadway width (sum of lane width and width of both shoulders) is limited to 8.5 m [28 ft] in these tables, and the right shoulder width is always equal to or greater than the left shoulder width. The recommended lane and shoulder widths are applicable to the controlling curve of the loop ramp, and it is recommended that the same, or similar, lane and shoulder widths be used for all curves along the ramp proper. Table 10-X1a. Recommended Lane and Shoulder Widths for Controlling Curves on Entrance Loop Ramps Metric Design speed (km/h) Radius (m) Lane width (m) Outside shoulder width (m) Inside shoulder width (m) 30 30 4.3 – 4.8 0.6 0.6 40 45 4.3 0.6 0.6 – 1.80.9 0.9 – 1.5 4.8 0.6 0.6 – 1.50.9 0.9 – 1.2 5.4 0.6 0.6 – 1.20.9 0.9 6.0 0.6 0.6 – 0.9 40 60 4.3 0.6 0.6 – 0.94.8 – 6.0 0.6 0.6 50 60 4.3 0.6 1.2 – 3.0 0.9 0.9 – 2.7 1.2 1.2 – 2.4 1.5 1.5 – 1.8 4.8 0.6 0.9 – 3.0 0.9 0.9 – 2.4 1.2 1.2 – 2.1 1.5 1.5 – 1.8 5.4 0.6 0.6 – 2.4 0.9 0.9 – 2.1 1.2 1.2 – 1.8 1.5 1.5 6.0 0.6 0.6 – 1.8 0.9 0.9 – 1.5 1.2 1.2 50 75 4.3 0.6 0.6 – 2.4 0.9 0.9 – 1.8 1.2 1.2 – 1.5 4.8 0.6 0.6 – 2.1 0.9 0.9 – 1.5 1.2 1.2 5.4 0.6 0.6 – 1.80.9 0.9 – 1.2 6.0 0.6 0.6 – 1.50.9 0.9 – 1.2 50 90 4.3 0.6 0.6 – 1.50.9 0.9 4.8 0.6 0.6 – 1.20.9 0.9 5.4 0.6 0.6 – 0.9 6.0 0.6 0.6 60 90 4.3 0.9 1.2 – 3.0 1.2 1.2 – 2.7 1.5 1.5 – 2.4

1B-5 4.8 0.6 1.5 – 3.0 0.9 0.9 – 2.7 1.2 1.2 – 2.4 1.5 1.5 – 2.1 1.8 1.8 5.4 0.6 1.2 – 2.4 0.9 0.9 – 2.1 1.2 1.2 – 1.8 1.5 1.5 6.0 0.6 0.9 – 1.8 0.9 0.9 – 1.5 1.2 1.2 Table 10-X1b. Recommended Lane and Shoulder Widths for Controlling Curve on Entrance Loop Ramps U.S. Customary Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 14 - 16 2 2 25 150 14 2 2 - 6 3 3 - 5 16 2 2 - 5 3 3 - 4 18 2 2 - 4 3 3 20 2 2 - 3 25 200 14 2 2 - 3 16 - 20 2 2 30 200 14 2 4 - 10 3 3 - 9 4 4 - 8 5 5 - 6 16 2 3 -10 3 3 - 8 4 4 - 7 5 5 - 6 18 2 2 - 8 3 3 - 7 4 4 - 6 5 5 20 2 2 - 6 3 3 - 5 4 4 30 250 14 2 2 - 8 3 3 - 6 4 4 - 5 16 2 2 - 7 3 3 - 5 4 4 18 2 2 - 6 3 3 - 4 20 2 2 - 5 3 3 - 4 30 300 14 2 2 - 5 3 3 16 2 2 - 4 3 3 18 2 2 - 3 20 2 2 35 300 14 3 4 - 10

1B-6 4 4 - 9 5 5 - 8 16 2 5 - 10 3 3 - 9 4 4 - 8 5 5 - 7 6 6 18 2 4 - 8 3 3 - 7 4 4 - 6 5 5 20 2 3 - 6 3 3 - 5 4 4 Table 10-X2a. Recommended Lane and Shoulder Widths for Controlling Curve on Exit Loop Ramps – Simple Curves Metric Design speed (km/h) Radius (m) Lane width (m) Outside shoulder width (m) Inside shoulder width (m) 30 30 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 40 45 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 40 60 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 50 60 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 4.8 0.9 2.1 – 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 50 75 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 60 90 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 4.8 0.9 2.1 – 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 Table 10-X2b. Recommended Lane and Shoulder Widths for Controlling Curve on Exit Loop Ramps – Simple Curves U.S. Customary Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 16 2 8 - 10 18 6 - 8 20 4 - 6 25 150 16 2 8 - 10

1B-7 18 6 - 8 20 4 - 6 25 200 16 2 8 - 10 18 6 - 8 20 4 - 6 30 200 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 30 250 16 2 8 - 10 18 6 - 8 20 4 - 6 35 300 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 Table 10-X3a. Recommended Lane and Shoulder Widths for Controlling Curves on Exit Loop Ramps – Compound Curves Metric Design speed (km/h) Radius (m) Lane width (m) Outside shoulder width (m) Inside shoulder width (m) 30 30 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 40 45 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 4.8 0.9 2.1 - 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 4.8 1.2 1.8 – 2.4 5.4 1.2 - 1.8 6.0 1.2 40 60 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 50 60 4.8 0.9 2.1 - 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 4.8 1.2 1.8 – 2.4 5.4 1.2 - 1.8 6.0 1.2 4.8 1.5 1.5 - 2.1 5.4 1.5 50 75 4.8 0.6 2.4 - 3.0 5.4 1.8 - 2.4 6.0 1.2 - 1.8 4.8 0.9 2.1 - 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 4.8 1.2 1.8 – 2.4 5.4 1.2 - 1.8 6.0 1.2

1B-8 60 90 4.8 0.9 2.1 - 2.7 5.4 1.5 - 2.1 6.0 0.9 - 1.5 4.8 1.2 1.8 – 2.4 5.4 1.2 – 1.8 6.0 1.2 4.8 1.5 1.5 - 2.1 5.4 1.5 Table 10-X3b. Recommended Lane and Shoulder Widths for Controlling Curves on Exit Loop Ramps – Compound Curves U.S. Customary Design speed (mph) Radius (ft) Lane width (ft) Outside shoulder width (ft) Inside shoulder width (ft) 20 100 16 2 8 - 10 18 6 - 8 20 4 - 6 25 150 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 25 200 16 2 8 - 10 18 6 - 8 20 4 - 6 30 200 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 16 5 5 - 7 18 5 30 250 16 2 8 - 10 18 6 - 8 20 4 - 6 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 35 300 16 3 7 - 9 18 5 - 7 20 3 - 5 16 4 6 – 8 18 4 – 6 20 4 16 5 5 - 7 18 5

1B-9 Design widths of ramp traveled ways for turning roadways under various conditions are also given in Table 3-29, which may be used to determine widths for ramps with design characteristics not shown in Tables 10-X1, 10-X2, and 10-X3. Values in Table 3-29 are shown for three general design traffic conditions, as follows:  Traffic Condition A—predominantly P vehicles, but some consideration for SU trucks  Traffic Condition B—sufficient SU vehicles to govern design, but some consideration for semitrailer vehicles  Traffic Condition C—sufficient buses and combination trucks to govern design Traffic conditions A, B, and C are described in broad terms because design traffic volume data for each type of vehicle are not available to define these traffic conditions with precision in relation to traveled way width. In general, traffic condition A has a small volume of trucks or only an occasional large truck, traffic condition B has a moderate volume of trucks (in the range of 5 to 10 percent of the total traffic), and traffic condition C has more and larger trucks. Shoulders and lateral offset—Design values for shoulders and lateral offsets on the ramps are as follows:  When paved shoulders are provided on ramps, they should have a uniform width for the full length of ramp. For one-way operation, the sum of the right and left shoulder widths are typically ranges from between3.0 to 4.3 m [10 to 14 ft]. A paved shoulder width of 0.6 to 1.2 m [2 to 4 ft] is desirable on the left with the remaining width of 2.4 to 3.0 m [8 to 10 ft] used for the paved right shoulder.  The left and right shoulder widths may be reversed if needed to provide additional sight distance.  The ramp traveled way widths from Table 3-29 for Case II and Case III should be modified when paved shoulders are provided on the ramp. The ramp traveled way width for Case II should be reduced by the total width of both right and left shoulders. However, in no case should the ramp traveled way width be less than needed for Case I. For example, with condition C and a 125-m [400-ft] radius, the Case II ramp traveled way width without shoulders is 6.4 m [21 ft]. If a 0.6-m [2-ft] left shoulder and a 2.4-m [8-ft] right shoulder are provided, the minimum ramp traveled way width should be 4.8 m [15 ft].  Directional ramps with a design speed over 60 km/h [40 mph] should have a paved right shoulder width of 2.4 to 3.0 m [8 to 10 ft] and a paved left shoulder width of 0.3 to 1.8 m [1 to 6 ft].  For freeway ramp terminals where the ramp shoulder is narrower than the freeway shoulder, the paved shoulder width of the through lane should be carried into the exit terminal. It should also begin within the entrance terminal, with the transition to the narrower ramp shoulder accomplished gradually on the ramp end of the terminal. Abrupt changes should be avoided.

1B-10  Ramps should have a lateral offset on the right outside of the edge of the traveled way of at least 1.8 m [6 ft], and preferably 2.4 to 3.0 m [8 to 10 ft], and a lateral offset on the left of at least 1.2 m [4 ft] beyond the edge of traveled way.  Where ramps pass under structures, the total roadway width should be carried through the structure. Desirably, structural supports should be located beyond the clear zone. As a minimum, structural supports should be at least 1.2 m [4 ft] beyond the edge of paved shoulder. The AASHTO Roadside Design Guide (4) provides guidance on clear zone and the use of roadside barriers.  Ramps on overpasses should have the full approach roadway width carried over the structure.  Edge lines or some type of color or texture differentiation between the traveled way and shoulder is desirable.