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.
123 The approach to design criteria and technical content in the Green Book needs to be updated. The following is a review of the key findings: ⢠Context mattersâand it varies, particularly with respect to the transportation service for vulnerable road users; ⢠AASHTO dimensional criteria should be based on proven, known measurable performance effects; ⢠Speed is an essential input to the determination of design values and dimensions; ⢠Some AASHTO criteria are not sensitive to key context attributes that are proven influencers of performance and cost effectiveness, specifically traffic volume and road type; ⢠Some AASHTO criteria are overly simplistic in their formulation, or are based on rational models lacking a proven basis in science; ⢠AASHTO design criteria produce inconsistent outcomes with respect to performance; ⢠AASHTO criteria should reflect known interactive safety and operational effects of geometry; and ⢠Dimensional guidance should be replaced with direct performance guidance (i.e., dimensions derived from performance metrics) where possible within the AASHTO policy. The suggested geometric design process still requires published dimensional criteria or meth- ods for deriving such criteria for roads on new alignment, and reconstructed roads on existing alignment, which substantially change the road type. This section of the report addresses the core geometric design criteria as currently published in the AASHTO Green Book, which would presumably apply to new road design. The most significant finding, which is at the core of the concept of cost effectiveness in design policy, is the lack of sensitivity to traffic volume and road type in the formulation and application of AASHTO design criteria. Table 3 summarized a review of current AASHTO policy short comings in this area. Design criteria lacking traffic volume sensitivity and formulated in a manner that treats all contexts the same are strong candidates for significant revision. Design criteria that ignore potentially important interactions with other geometric variables are also candidates for study and potential revision. In reference to Table 1, the following recommendations are made. 6.1 Overview of Contents of 2011 AASHTO Policy on Geometric Design The matrix (Appendix B) summarizes the results of a complete review of the 2011 AASHTO Green Book. The research team listed each direct reference to quantitative guidance on design of geometric features. The matrix summarizes the location of each reference in the policy, and then classifies the background behind each reference. C h a p t e r 6 Updating the Technical Guidance on Geometric Design in the AASHTO Policies
124 a performance-Based highway Geometric Design process Geometric design guidance is broadly classified as being based on either empirical research or rational engineering models. In the case of the former, the research basis for a design dimension or value can be further classified as relating to any of the following: ⢠Spatial or operational characteristics of motor vehicles, ⢠Human factors (drivers, pedestrians, or cyclists), ⢠Substantive or quantitative safety research, ⢠Substantive traffic flow or operations research, and ⢠Roadway infrastructure conditions. The matrix also summarizes the overall quality or robustness of the research basis, as defined by its context sensitivity, and currency of the research. There are many geometric elements whose derivation relies on rational engineering models. Many of these models were developed and formulated years ago, and remain as the fundamental basis for design. Such models may include factors or assumptions with a research basis, but the fundamentals of the models were developed independently of actual operations and as such represent hypothetical measures. In some cases the models reflect general observations of an anecdotal nature, but lack any formal testing or validation through published research. (For example, selection of a maximum superelevation rate for curves is based on avoiding a vehicle sliding down a curve at low speeds or stopped when the pavement friction is very low, as would be the case under icy conditions. Such behavior is observable and explainable in engineering physics terms, but its relative frequency and severity has never been documented.) Those geometric elements based on rational models have as their basic assumptions one or more of the following attributes: ⢠Hypothesized safety performance, ⢠Hypothesized traffic operational performance, and ⢠Aesthetics. In some cases, most notably SSD, the basis for design values is a combination of a rational engineering model and empirical research to develop parameters of the model. The matrix confirms that there is much technical dimensional guidance in the AASHTO policy that is not based on empirical substantive safety or operational research, but rather on rational engineering models. Note that the relative importance of the models and geometric criteria vary widely. The last column of the table provides a qualitative judgment of how important each criterion is with respect to the cost or impact on design and construction. For those criteria rated low, a simple hypothetical model may be sufficient and not in need of formal validation. (A separate issue to consider is whether the geometric guidance is really needed.) The following summarizes the review of the AASHTO Green Book: 1. Design dimensions for many elements are based on simple and direct operational measures of vehicle dimensions and performance (e.g., offtracking). AASHTO continually adds design vehicles. For such design elements and dimensions current policy guidance is sufficiently robust. 2. Certain key design elements that greatly affect the cost of the roadway are based on hypothe- sized simple models and not empirical research. This includes SSD, horizontal curvature, and width dimensions for roadways other than two-lane rural highways. These are highlighted in the matrix. 3. Some criteria are based only on aesthetics. Although these have negligible influence on the cost of design, they are potential candidates for elimination as they may unnecessarily
Updating the technical Guidance on Geometric Design in the aaShtO policies 125 restrict the ability of designers to implement geometric solutions. As a minimum, should AASHTO retain these, it should be made clear that the guidance has no meaningful safety or operational basis. 4. The operational and safety needs of pedestrians and bicyclists are not expressly addressed in the formulation of geometric design criteria. Lane and roadway widths for certain urban road types and contexts for both separate bike lanes and shared lanes are lacking. Current NCHRP research is addressing these gaps. Also, maximum grades for urban roadways in certain contexts should reference gradability of the road for cyclists. 5. Recommended LOS policy for freeways in urban contexts is outdated. Urban freeway recon- struction carries the highest costs per mile, and as design LOS directly influences decisions on the sizing of such facilities, revising current published guidance is a critical need. Design for LOS E has been commonplace for over 20 years across the U.S. Indeed, LOS E is consistent with well-established design solutions such as HOV/HOT lanes and ramp metering. Design of such roadways should be included in the Green Book. 6. The current design speed model is vehicle centric and focused on directing designers to select a high enough speed such that speed-sensitive design dimensions will be properly designed. Criteria for selection of lower speeds as being consistent with a multimodal and pedestrian- centric context are needed. 7. Criteria based on hypothesized models that have been disproven or found not critical are can- didates for removal or substantial change. Foremost among these is the guidance for selection of ramp design speeds. Cross-slope rollover is another design criteria that should be changed based on actual research. 8. The majority of design elements are context insensitive in their formulation and applicabil- ity. The prevailing variable in many of these is design speed, which presumably reflects the context. However, as noted above the design speed process itself does not satisfy all urban contexts. Context insensitivity also relates to the assumptions regarding the design vehicle and driver. For many design criteria the passenger car is assumed for all conditions and contexts. 9. AASHTO has incorporated recent research to correct flawed hypothetical models and pro- duce more reasonable design criteria. Examples include changes in the ISD models (based on gap acceptance studies) and the updating of the object height for SSD to represent a more realistic condition based on actual events. Research completed since the last edition should be referenced to update design of exit and entrance ramp terminals, and spacing between successive ramps on freeways. Following is a discussion of the individual geometric design elements of the policy. 6.2 Lane and Traveled Way Widths Lane-width criteria for two-lane rural highways were developed based on cost-effective analy- ses of safety and operational performance and initial construction costs (Zegeer and Neuman 1993). They reflect some consideration of the interactive effects of shoulder width. The research was based on older statistical methods for crash analysis and may warrant updating. Finally, the criteria as published in the Green Book are among the very few that differentiate roads to be reconstructed versus new roads. Lane-width criteria for multilane rural roads and urban/suburban arterials do not reflect traf- fic volume sensitivities and are not based on explicit analysis of safety performance. The research basis for the HSM methods suggested no sensitivity of lane width to serious crash frequency or severity (although the lack of much mileage of multilane roads in rural areas with lane widths less than 12 feet makes firm conclusions difficult). Gaps in current knowledge are primarily
126 a performance-Based highway Geometric Design process associated with lane-width requirements and interactive effects of lane widths given incorpora- tion of bicycle and transit lanes with general purpose lanes. This specific topic is being addressed in ongoing NCHRP research. Lane-width criteria for freeways per both AASHTO and FHWA remain fixed at 12 feet under all circumstances and traffic volumes. Interestingly, there are sufficient freeway segments in the United States with less than 12 feet of width (these would all have been constructed as design exceptions) such that statistically meaningful safety performance measures of varying freeway lane widths were able to be confirmed in the research that has become part of the HSM (Bonneson et al. 2012). Figure 41 shows the crash modification factor for lane widths, with 12 feet of width being the base condition. Essentially, the examples of less than 12-foot lane widths are reconstruction projects, many of which included addition of general purpose lanes, managed lanes, or both within restricted right-of-way. That projects incorporating narrower lane widths continue to be proposed by agencies, approved by FHWA, constructed, and operated in an acceptable manner demonstrates the reasonableness of dimensions less than 12 feet. The relatively modest difference in safety performance between 11- and 12-foot lanes confirms that they can be reasonable designs. The operational effects of varying lane width and shoulder width on freeway throughput are known and codified in the HCM, and these are similarly small in terms of the difference in performance. Moreover, the HSM research confirmed other research (Lord et al. 2004, Harwood et al. 2013) that documented a relationship between traffic density and crash frequency on freeway seg- ments. Where narrower lane widths are used to enable the addition of a lane at little additional cost, the net safety effects may actually be positive, or at least negligible (i.e., the adverse effects of narrower lanes may be offset by the decrease in density that results in lower crash frequencies). See Tables 19 and 20. Finally, it is noteworthy that Australian design policy allows for the metric equivalent of 11-foot lanes for lower design speeds on freeways (Austroads nd). Source: HSM Figure 41. CMF for lane widths on freeways. Source: HCM Table 15-7. Table 19. CMF for lane and shoulder widths.
Updating the technical Guidance on Geometric Design in the aaShtO policies 127 To summarize, the application of less than 12-foot lane widths has become commonplace for high-volume urban freeways, despite the stated criteria in the AASHTO Policy. There is an adequate knowledge base on both safety and operations that would enable the develop- ment of a performance-based process and criteria with more flexibility to address limitations in right-of-way and high marginal costs of widening that are typical on urban and suburban freeways. 6.3 Shoulder Widths Shoulder width criteria for two-lane rural roads were developed with the lane-width crite- ria (Zegeer and Neuman 1993). For rural highways, the shoulder is the initial portion of the roadside, and criteria for shoulder width reflect research on the safety performance of roadside encroachments. Some agencies employ shoulders on multilane suburban roads with higher volumes and in some cases with curbs. There is limited knowledge on the safety effects of shoulders in such cases, but this design practice is relatively uncommon in true suburban contexts. Design criteria for shoulders and shoulder widths on urban freeways have been unchanged for many years. Current AASHTO criteria call for full shoulders (generally 10 feet or more) on the right, and 4 feet on the left or inside shoulder for four-lane freeways, and full shoulders for both sides on freeways with six or more lanes. As with lane widths, the practice of accepting design exceptions for shoulder widths is com- monplace and has increased in recent years. Agencies have treated shoulders as spatial oppor- tunities to increase throughput with little or no expense. Some of the trade-offs associated with such design decisions can be characterized objectively. The recent research on freeway crash prediction that is in the HSM provides SPFs for varying freeway cross sections and CMF for varying shoulder widths (AASHTO 2010). The operational effects of shoulders with respect to throughput also are understood and codified in the HCM. Table 21 summarizes the O&M functions associated with shoulders, and the extent to which the shoulder width is critical to enabling these functions. Some functions require a full width (generally 10 feet or more), while others may be provided with lesser widths. For the latter, the value derived may be related to the width provided (e.g., a 6-foot shoulder provides more offset for roadside safety than a 3-foot shoulder), but any dimension provides some measur- able value. Some functions are important for all contexts (e.g., emergency access, law enforcement) and others only in some contexts (e.g., snow storage and removal). Also, the relative importance of all functions clearly varies by road type, context, and traffic volume. The ability to access crashes Shoulder Width (ft) Shoulder Type 0 1 2 3 4 6 8 Paved 1.00 1.00 1.00 1.00 1.00 1.00 1.00 Gravel 1.00 1.00 1.01 1.01 1.01 1.02 1.02 Composite 1.00 1.01 1.02 1.02 1.03 1.04 1.06 Turf 1.00 1.01 1.03 1.04 1.05 1.08 1.11 Note: The values for composite shoulders in this table represent a shoulder for which 50 percent of the shoulder width is paved and 50 percent of the shoulder width is turf. Source: HSM Table 10-10 Table 20. CMF for shoulder types and shoulder widths on roadway segments.
128 a performance-Based highway Geometric Design process and incidents is greatly facilitated by full shoulders. Road user benefits include reductions in non-recurring congestion, and survivability of crash victims by access to and egress from the site by fire, police, and ambulances. Agencies have ministerial duties to maintain roads in a reasonable state of good repair. Many maintenance activities involve the roadside, such as mowing, culvert cleaning and ditch mainte- nance, sign replacement and repair, utility maintenance, and guardrail and barrier replacement. The presence and available width of shoulders can greatly affect the ability and cost associated with routine maintenance activities. The maintenance and operational costs of either having full shoulders to work with, or having to close lanes (and often perform maintenance in off hours, and weekends) also will vary by road type and context. Some functions requiring full widths (enforcement, emergency stops, and incident manage- ment) may be suitably accommodated by intermittent pull outs. In an urban design context such pull outs may be strategically placed wherever space is available within the right-of-way. This concept may provide some measure of value at minimal additional cost. Finally, the ability to substitute ITS solutions for shoulder functions should be investigated. Managed lane facilities without shoulders are virtually impossible to enforce using conventional techniques, but automated enforcement using overhead cameras may enable an agency to no longer require or rely on shoulders as much as has been the case. There is a great need for systematic research on the cost effectiveness of shoulders and shoulder width dimensions on high-volume urban freeways. Such research would include full investigation of the operational and safety trade-offs associated with narrowing lane widths. It also should include the relative effectiveness of intermittent turnouts as a substi- tute for full shoulders, including the frequency of such turnouts. This knowledge could serve to provide design guidance for projects in which alternatives to full shoulder dimensions are needed. Shoulder Function Effectiveness for Given Width Condition Partial (< 10 ft.) Full (10 feet or more) Law Enforcement None* Yes Emergency Stopping None Yes Snow Storage Yes Yes Ponding and Water Storage During Intense Rainfall Yes Yes Roadway Capacity Yes Yes Routine Roadside Maintenance None Yes Roadside Safety (Clear Zone) Yes Yes Pavement Support Yes Yes Maintenance of Traffic During Construction Some (Depends on Total Width Available) Yes Provide Sight Lines to Vertical Obstructions Yes Yes Incident Management None Yes * Some agencies use turnouts for enforcement Table 21. Effectiveness of shoulder widths for shoulder functions.
Updating the technical Guidance on Geometric Design in the aaShtO policies 129 6.3.1 Framework for Design Criteria Developmentâ Lane Width and Shoulders Design values for lane width increased over the years as vehicle dimensions increased and speeds increased. For many agencies a 12-foot lane width is considered a standard minimum. As research has demonstrated, though, the contribution of incremental width to safety perfor- mance is minimal for certain road types (two-lane roads, freeways) and nonexistent for others (collectors, local roads, urban arterials). The incremental benefits of 12-foot versus 11-foot (or 11-foot versus 10-foot) lanes relate primarily to operational efficiency and capacity. The basic principles of cost effectiveness and knowledge of safety performance should pro- duce lane-width policy that properly recognizes the relative importance in transportation per- formance of lane width to other roadway features. Except for the highest class roadways in more rural, high-speed contexts, design policy should favor or promote the use of narrower lanes for the following reasons: ⢠The design width of lanes affects the entire length of the project, including right-of-way foot- print and construction cost. ⢠Better uses for the additional 1 or 2 feet of lane width may be in shoulder width, provision of bike lanes, or greater offsets to roadside objects. ⢠To the extent that lane width influences driver speed behavior, design policy should promote the use of lane widths consistent with the desired speeds, which in many cases are lower. ⢠Every project ends by tying to an existing alignment and cross section. The agency may have no plans to reconstruct the roadway for many years. A cross section with greater width in the middle of a roadway with lesser widths on either end may be more costly and produce counter- productive operations. Figure 42 shows a straw-man road type and context framework for lane-width design criteria. Roadway Type Rural Natural Zone Rural Zone Suburban Zone GeneralUrban Zone Urban Center Zone Urban Core Zone Local Collector Arterial Freeway 10- to 11- widths; greater dimension where bicycles, on-street parking, bus and loading zones occur 10-12 ; addiÂonal width where bicycles are to be considered 12- lane widths; full right shoulders 11- to 12- lanes; consider total width of shoulders and develop opÂmal soluÂon given right-of-way, maintenance and performance analysis 10- to 11- widths; greater dimension where bicycles, on-street parking, bus and loading zones occur 10-12 ; addiÂonal width where bicycles are to be considered 10- to 11- widths; greater dimension where bicycles, on-street parking, bus and loading zones occur Total road width based on operating characterisÂcs of vehicle; 9 minimum lanes may suffice Total road width based on providing minimum LOS and reflecÂng expected crash risk; 10- lanes should suffice for most volume ranges Range of 10 to 12 may apply based on volume, context (terrain, trucks, environmental); shoulder dimensions of 2 or more based on crash risk and maintenance costs 12 lane widths for most cases; in extreme context constraints 11- to 11.5 may be considered 10- minimum; addiÂonal width where bicycles are to be considered Figure 42. Straw-man framework for design of lane widths for road types by context.
130 a performance-Based highway Geometric Design process 6.3.2 Local Roads in Rural Contexts Local roads in rural contexts are low-volume in nature and lower speed. Lane widths of 9 to 10 feet are sufficient. Operation of special vehicles such as agriculture combines may require greater widths, but these could include total width of roadway and shoulder, and may be limited to widening along curves. 6.3.3 Collector Roads and Arterials in Rural Contexts Design criteria for lane and shoulder width as currently in the AASHTO policy are based on performance-based analysis (Zegeer and Neuman 1993). The safety performance research basis preceded advances in methodologies for crash analysis that are prevalent in the HSM and now considered state-of-the art. Although the framework and approach reflect safety performance, there may be a need to update and potentially revise the findings based on current research best practices. 6.3.4 Urban Nonfreeway Roads The regime encompassing all land use contexts for local, collector, and arterial road types should be viewed as inherently multimodal in nature, with designer choices for many cross- section features, including bike lanes, on-street parking, medians, loading zones, and transit stops. Cross-section design in such cases can only be assessed holistically (i.e., by considering the width needs and interactive effects of the elements being considered). NCHRP Project 03-112 is just beginning. This project should address the knowledge gaps and provide performance-based input for such roadways. The geometric design process framework should (1) acknowledge that lane widths above 12 feet provide no discernible safety performance benefits; (2) recognize that the availability of width for the roadway within the right-of-way will always be limited, requiring choices and trade-offs; and (3) reflect the fact that lesser lane widths of 10 to 11 feet may provide substantial flexibility, enabling inclusion of other desired features. 6.3.5 Urban Freeways Urban freeways carry a vastly disproportionate share of regional travel of all types. In virtually every major city, existing urban freeways operate at traffic volumes exceeding their design capac- ity. Despite published AASHTO guidance, urban freeway reconstruction projects are routinely undertaken in which the number of lanes is limited such that the design is LOS E, often through much of a typical day, for reasons of unavailable space and unaffordable costs. Indeed, the prin- ciple value of managed lane projects is on corridors in which operation at low levels of service is accepted because of context constraints. Current design policy per AASHTO and FHWA places strict limitations on lane width for freeways, with no allowance under policy for lane widths less than 12 feet. Of course, there are sufficient examples of freeway projects in which lesser lane widths have been constructed or reconstructed such that the relative safety performance of lane widths less than 12 feet can be assessed. The design process for high-volume urban freeways should allow for the evaluation of alter- native cross-section designs that include lane widths of less than 12 feet. There are quantifiable benefits associated with narrower lanes that include the ability to provide additional lanes within limited space and the cost to reconstruct what are the most costly facilities in the highway system. The Interchange Safety and Analysis Tool Enhanced (ISATe) HSM and 2010 HCM procedures reveal the relative value of different freeway cross sections using the same total width (lanes and shoulders) but allocating the dimensions such that one additional lane of travel is provided. Case
Updating the technical Guidance on Geometric Design in the aaShtO policies 131 Study 5 in Chapter 7 is an example of a freeway where using narrow lanes and shoulders to allow an additional lane provides better performance. Urban freeway corridor reconstruction projects can have initial capital costs on the order of $50 million to $100 million per mile. For a 10-lane segment with full shoulders (164 feet ±), a reduction in lane width from 12 to 11 feet reduces the footprint by 10 feet, or 6 percent. The cost savings associated with such a design should be expected to exceed 6 percent given lessening of right-of-way effects and improved constructability. But even assuming only a 6 percent reduc- tion, the savings can be $3 to $6 million per mile. Although some new urban freeways are being planned, designed, and constructed, the vast majority of expenditures in the U.S. on urban freeways involve reconstruction. In many cases, agencies seek to add capacity for either general purpose traffic or managed (HOV, tolled) traffic. As a matter of transportation policy, and depending on the problem(s) being addressed in the reconstruction of the urban freeway, either design solution may be considered optimal from a corridor, network, or regional perspective. A revised design process applied specifically for reconstruction of urban freeways should allow for the use of lane widths less than 12 feet, particularly in cases where additional freeway capacity can be provided without increasing the total width of the cross section. Given the ability to assess the operational and safety performance of an array of lane-width values, there is no reason why the design process for urban freeways cannot promote and allow a site-specific analysis that allows for lane widths as low as 10 feet. The context of such a design is for freeway and freeway distributors in the urban core where speeds throughout the day may be 40 mph or less. 6.4 Roadside Design Roadside design policy as described in the AASHTO Roadside Design Guide is based on safety performance research including traffic volume sensitivity. Indeed, it is AASHTO and FHWA policy that the selection and application of roadside design in the rural environment (clear zone and slopes) be determined on a project basis because of the wide range in contexts encountered. AASHTO recently implemented further context-sensitive changes to design policy on urban streets in the most recent updates. For urban conditions in which speeds are lower and right-of- way more limited, the current policy is based on safety performance research on the frequency and severity of impacts on roads with and without curbs. The criterion lateral offset with varying dimensions of a 2-foot minimum (between curb and object) and a 4-foot minimum (between edge of uncurbed traveled way and object) now forms the basis for roadside design on urban and suburban streets. Similarly, the criteria for placement and design of barrier systems are based on well-established performance criteria that reflect the vehicle fleet. There are currently NCHRP projects investigating roadside design and roadside quantitative safety. NCHRP 17-54, âConsideration of Roadside Features in the Highway Safety Manualâ and NCHRP 17-55, âGuidelines for Slope Traversabilityâ are in progress. No recommendations are made for further research and development pending completion of these projects. 6.5 Alignment and Sight Distance The design criteria that influence alignmentâboth horizontal and verticalârepresent the greatest need for improvements to bring about a more cost-effective approach to road design. Two specific criteria represent the greatest opportunitiesâhorizontal curvature and sight distance.
132 a performance-Based highway Geometric Design process 6.5.1 Horizontal Curvature Horizontal alignments consist of intersecting tangents (the point of intersection or PI), with simple circular curvature providing a smooth pathway connecting the two tangents. The geom- etry of the circular curve is defined by the radius and central angle between the tangents. For any given radius, the greater the central angle, the longer the curve. Other geometric features include the superelevation, superelevation transitions, and spirals. The primary design criteria governing horizontal alignment is the minimum curve radius, which is associated with the roadâs design speed and maximum superelevation, which is set by policy of the owning agency. According to AASHTO, maximum superelevation rates between 4 percent and 12 percent are allowable. Common practice among highway agencies is to use maximum rates between 6 percent and 8 percent for open, high-speed alignment. The following summarizes the current approach to design of horizontal curvature according to AASHTO: ⢠It applies consistently across all road types and contexts. ⢠It is volume insensitive (i.e., curves are designed the same regardless of the design traffic volume). ⢠It assumes passenger car operation and assumes the driver tracks the curve exactly as it is designed. ⢠The model assumes constant speed (design speed). ⢠The model incorporates only radius of curve and not length (or central angle). ⢠Design policy considers curvature independent of other geometry (grade, cross section). ⢠Design policy assumes simple tangent-curve geometry (i.e., no spiral transitions). ⢠Design policy for the friction factor, which expresses the lateral acceleration is set by AASHTO, and varies with speed. ⢠Although labeled as friction factor, the lateral acceleration component to the curve design policy is not based on measures of the tire/pavement interface, but rather on the comfort- based reaction of drivers tracking curves at varying speeds. ⢠The research basis for the comfort factor is field studies that are over 70 years old. ⢠Design policy is independently established for a range of maximum superelevation values from e = 0.04 to e = 0.12. ⢠Establishment of maximum superelevation values is assumed necessary to prevent a vehicle from sliding down a curve under icy (low pavement friction) conditions in stopped or low- speed operation. 6.5.1.1 Critique of Current AASHTO Curve Design Model The adoption of a comfort-based approach to curve design was made in the 1940s, predating all the research on safety and operations. It is consistent with the general design philosophy of the day and limitations in knowledge of the day. The vast majority of roads were two-lane rural. The research behind the model reflected the driving population and vehicle characteristics of the day (suspensions, weight, center of gravity, tires, etc.). The rational model that assumes tracking the curve (center of lane) at the design speed was appropriate for the time, given the lack of knowledge on driver/vehicle behavior. However, it is impossible for a driver/vehicle system to track a curve as the model assumes. This would require an instantaneous steering behavior and vehicle response when the vehicle reaches the point of curvature (PC) and a similar response in curve departure at point of tangency (PT). The model also is based on anecdotal evidence of vehicle behavior on superelevation on icy pavement. The risk of vehicles sliding down a curve has never been characterized in quantitative
Updating the technical Guidance on Geometric Design in the aaShtO policies 133 terms. This behavior is inherently low speed. It is associated with the exposure to icing condi- tions that vary greatly by context. Finally, the model ignores the potential interaction of grade with curvature on operations. Research conducted in the 1980s characterized driver/vehicle behavior in the approach and tracking through the curve (Glennon et al. 1985). Drivers on the tangent approach to an unspiralled curve are actually required to steer in the opposite direction of the impending curve to counteract the effect of the pavement rotation as superelevation is being developed on the tangent. Once at the PC, the alignment changes instantaneously from tangent to the full curve radius. As drivers cannot instantaneously steer this radius, but rather take some time to steer from a tangent to curve path, the path of the vehicle itself differs measurably from that described by the geometry of the curve. The actual path that is tracked is a spiral (which is mathematically defined as a curve for which the radius is changing at a constant rate). In the initial 100 to 150 feet of the curve, the path generates radii that are greater than that of the curve itself. Because such a path if not corrected would result in the driver running off the road, driver steering behavior corrects the vehicle path downstream of the first 100 to 150 feet. This corrective response results in the actual minimum tracked radius at some point having to be smaller than the road curve itself. This steering behavior was observed on hundreds of vehicles; it occurred within the lane that varied in width. All drivers operating on unspiralled, horizontal curves produce this behavior that was referred to as overshooting by Glennon et al. (1985). In addition, the research demonstrated that the extent or severity of overshooting behavior (which would be an indicator of the responsive- ness of the driver) is independent of the driverâs selected speed. The net effect of this behavior is that drivers undergo significantly greater lateral acceleration, and generate greater friction demand at the tire/pavement interface than the AASHTO design model assumes and the com- pensating effect of superelevation (assumed to be 1:1) is less than intended by AASHTO. The extent to which driving curved alignment represents a greater challenge and higher risk to drivers means that the amount or length of alignment that is curved versus tangent may be of concern in design for horizontal alignment. The relevant geometric variable influencing the amount of curvature is the central angle between tangents. The relationships among central angle, radius, and length are shown in Figure 43. Figure 43. Curve geometry.
134 a performance-Based highway Geometric Design process Additional research performed in the 1990s characterized horizontal alignment in terms of risk associated with the number of curves and the sum of the central angles of the curves over a given length of roadway (Glennon et al. 1985). This research reinforced the importance of not only the radius but also the central angle (or length) in curve design policy. As currently configured, AASHTO design policy does not explicitly recognize the risk associated with greater central angles, but rather focuses the designerâs attention solely on the appropriate or minimum radius of curve for the design speed. Another concern with respect to curve operations is the behavior of vehicles with high cen- ters of gravity. Research has confirmed that trucks on curves will overturn at speeds lower than the critical skidding speed (Lamm et al. 1986). Moreover, the propensity to overturn occurs at speeds much lower than the critical skid speed for passenger cars as well. This explains the anec- dotal evidence (yet to be quantified through formal research) that suggests that truck overturns on interchange loop ramps (which have very short radii and large central angles) are a recurring issue for many agencies. Finally, there is clear evidence of a dynamic effect associated with the combination of curvature and grade, such effect producing a greater adverse effect than would be expected. 6.5.1.1.1 Summary of Knowledge Base on Curve Operations. According to the HCM there is no direct effect of curvature on the speed or other performance measures such as density or LOS for any road type. An indirect effect associated with the terrain (level, rolling, mountainous) may be inferred, but this appears to be mostly attributable to grade. The presence and length of upgrades and downgrades have measurable effects for which there are adjustment factors in calculations of flow rates and capacity, but no such adjustment factors exist with curvature. There are anecdotal observations of curvature effects on speed and car-following behavior on high-volume urban freeways. The extent to which this can be characterized is a potential research area. Any meaningful effect on density and operations of high-volume freeways should be included in derivation of design criteria for such facilities. Indeed, for very high-volume freeways, it may be that the most important performance effect of curvature is on its effect on capacity, which could be translated to substantial user operating and travel time cost differences. It also is established that vehicles (tires and brakes) undergo more wear on curved versus tangent alignments. The user costs over time will be greater on alignment composed of curves. Where isolated, curves represent limiting speed elements, and differences in travel time may be associated to such curves when they occur on roads with higher free operating speeds. 6.5.1.1.2 Summary of Knowledge Base on Safety Performance of Horizontal Curves. Curves have long been recognized as being overrepresented in crashes. Research now incor- porated into the HSM provides a basis for understanding the safety performance of horizontal curvature (AASHTO 2010). The following is basic background information that should inform a review and updating of horizontal curve design policy: ⢠The relative risk of a crash is a function of both the radius of curve and length of curve (or stated differently, the central angle of curve). ⢠Figure 44 shows the relationship of curve geometry and traffic volume to crash frequency risk for two-lane rural highways. ⢠The interrelationship between curve geometry and roadside design (shoulder width and slope or barrier) to crash risk is significant; curve crashes are run-off-road by type and the outcome of the roadside incursion is affected by the quality of the roadside including the shoulder. ⢠Spiral transitions have a small but significant positive effect on quantitative curve crash risk. ⢠The effect of curve radius on crash risk is evident for two-lane rural highways and freeways, less so for multilane rural highways, and not apparent for urban arterials.
Updating the technical Guidance on Geometric Design in the aaShtO policies 135 The manner in which the AASHTO model is applied to curve design policy produces incon- sistent outcomes. This effect is based on the variance in maximum superelevation rates that are allowed by AASHTO. Curve design policy produces different designs for the same design speed. For example, for a design speed of 60 mph, the following are superelevation rates to be used for a radius of curve of 1,600 feet for the range of emax policies per AASHTO: emax = 4%â4% emax = 6%â5.6% emax = 8%â7.1% emax = 10%â8.1% In other words, a horizontal curve with identical horizontal geometry, that looks the same in all cases and is intended to be driven at the same speed is designed differently and hence feels differently to the driver depending on what state he/she is driving (i.e., on which superelevation policy is employed by that stateâs transportation agency). This inconsistent outcome of how AASHTO prescribes design of horizontal curves is illogical. Indeed, if superelevation is impor- tant, and the amount of superelevation for a given curve is important (which is confirmed in the HSM), it is hard to make the argument that the four design solutions noted above are equivalent. The background presumption of setting maximum superelevation policy is based on the objective of avoiding sliding down an icy curve. This dynamic can be easily calculated for very low pavement friction values. However, this fundamental control in how curves are designed has never been objectively quantified; it deserves objective study. The type of event or collision is by definition low speed and hence low severity. Even in states with colder climates the major- ity of travel occurs on roads that are not icy or subject to this behavior, thus this is in most cases a low frequency occurrence. Given that research confirms a marginal benefit of increased Figure 44. Potential crash effect of the radius, length, and presence of spiral transition curves in a horizontal curve.
136 a performance-Based highway Geometric Design process superelevation, the overall risk in setting of design policy seems to be in providing not enough rather than too much superelevation. 6.5.1.2 Context Insensitivity of AASHTO Curve Design Perhaps the most significant concern regarding the AASHTO curve model is its context insen- sitivity. The model is applied identically across all road types and across all traffic volume ranges. Curves are designed the same, with the same costs incurred, regardless of whether the design year traffic volume is 500 vpd, 5000 vpd, or 50,000 vpd. A comfort-based operational model may be appropriate for two-lane rural highways, but a different model may be better suited to high-volume freeways or lower speed and lower-volume local roads. The marginal operational costs (safety, delay, vehicle operations) of one curve design versus another depend on the traffic volume exposure to the curve. The lack of a traffic volume exposure measure in horizontal curve design policy prevents the development of cost-effective alignment design, particularly in reconstruction settings, without having to resort to design exceptions. 6.5.1.3 Potential Approaches to Horizontal Curve Design Policy Design policy for horizontal curvature should be thoroughly reviewed with new models and approaches proposed and tested. Measures of risk exposure per traffic volume and unique con- text features should be included. Differences in operational performance that may be present on different road types should be researched. The notion that the same design model should apply to all road types and contexts should be challenged. The following represents a straw-man proposed structure that illustrates the breadth and depth of potential policy changes. 6.5.1.3.1 Design of Curves on New Roads. Roads designed on new alignment should have a design policy that reflects the understanding of differences in operational and safety perfor- mance, and in the cost of construction based on the context (primarily location and terrain). Consider, for example, the following possible design policy framework for horizontal alignment, such framework representing a tailored approach to curve design as opposed to the current singular one-size-fits-all approach: ⢠Two-lane Rural HighwaysâDesign policy for curves could be based on safety-effectiveness analyses of crash risk (which would reflect traffic volume and length of curve); operational effects (speed and capacity, vehicle operating costs, and travel time value); and construction costs associated with terrain. The assumption of design speed independent of adjacent align- ment and grade could be replaced by use of iterative speed-profile analyses. Design policy could potentially include incorporation of shoulder width and roadside as an interactive design feature (e.g., allowance for smaller radii with wider shoulders) and use of spirals (e.g., allowance for smaller radii if spirals are used). The basis for designing the combination of curves and superelevation could be revisited to eliminate the inconsistencies in outcomes noted above, with one set of superelevation rates developed for the full range of maximum superelevation up to 12 percent. Finally, the importance or criticality of truck operations on curves may apply in setting design policy to certain two-lane rural curves (e.g., those function- ally classified as arterials with sufficiently high truck volumes). ⢠Multilane Rural Highways and FreewaysâDesign policy for curves could be based on safety-effective analyses of crash risk (which would reflect traffic volume and length of curve); operational effects (speed and capacity, vehicle operating costs, and travel time value); and construction costs associated with terrain. Policy could include specific analysis of trucks. Policy could include incorporation of shoulder width and roadside as a feature (e.g., allow- ance for smaller radii with wider shoulders) and spirals (e.g., allowance for smaller radii if
Updating the technical Guidance on Geometric Design in the aaShtO policies 137 spirals are used). For such highways, the notion of curve design policy based on truck opera- tions may be appropriate. ⢠Urban ArterialsâDesign policy for curves could be based on vehicle tracking requirements and driver behavior of large vehicles (trucks, buses) when operating at the speed limit; opera- tional analyses of capacity effects of curvature; and safety effects of intersections with curves on approaches (human factors and safety studies). ⢠Multimodal Urban ArterialsâDesign policy for curves could be based on vehicle tracking requirements and driver behavior of large vehicles (trucks, buses) when operating at the speed limit; operational analyses of capacity effects of curvature; safety effects of intersections with curves on approaches (human factors and safety studies); and operating costs, requirements, and characteristics of light rail, articulated buses, and streetcar operations. ⢠Urban FreewaysâDesign policy for curves could be based on safety effectiveness of crash risk, operational effects of curvature on uninterrupted flow under high-volume conditions, and construction costs. Regarding operational effects, the influence of both horizontal and vertical geometry on throughput and capacity may be significant. For high-volume freeways, the effect of milder curves and grades may compute to significant operational benefits over a 50 to 75 year project life. Such operational benefits also translate to safety benefits per the established relationship of crash frequency to volume-to-capacity (v/c) ratio. Finally, effects of high-volume flow pavement through curves (rutting, wear) on pavement life (maintenance costs) that may vary by curvature could influence urban freeway design criteria. ⢠Interchange RampsâDesign policy for curves could be based on safety effectiveness of crash risk; including potential interactive effects of grade and potential effects of truck safety per- formance on loop ramps; actual basis for loop ramp design may be truck operations, or may depend on the volume of trucks expected (e.g., through setting a forecast daily truck volume as a design policy threshold). 6.5.1.3.2 Curves on Reconstructed Existing Roads. Design policy for horizontal align- ment for reconstructed roads should follow the fundamental principle of addressing the prob- lem. The problem may or may not be related to the cross section or alignment; and the problem may or may not be evident over the entire project limits. For example, the presence of curve geometry that does not meet design policy for new roads should not require such a curve to be reconstructed to the new road policy nor require a design exception to justify not reconstruct- ing the curve. To re-emphasize, existing curve geometry that does not meet current design criteria for new roads does not in and of itself constitute a problem. The policy for reconstruction should be structured to utilize the site-specific data (speeds, crashes by severity, other geometry, and road corridor context). The policy should be process based rather than dimensionally based. Consider the following possible geometric design policy framework: ⢠Basic Problem Is OperationalâIf the problem being addressed is operational, and the nature of the operational problem is independent of the curve or curves, the policy should promote presumption of the adequacy of the existing geometry. If the curve or curves are part of the problem, the policy should outline (1) the specific performance measures to focus on; (2) solutions to improve performance that may include but are not limited to revisions to curve geometry; and (3) the diagnostic evaluation process to arrive at the optimal (most cost- effective) solution. For example, the combination of alignment and cross section may con- tribute to a high v/c ratio that produces low speeds and higher travel times. A single, sharp curve may create a bottleneck in an otherwise high-quality alignment. Operational models that predict the effects of increasing curve radius, widening lanes and shoulders through the curve, or both could serve as the basis for evaluating geometric design alternatives to the current alignment.
138 a performance-Based highway Geometric Design process ⢠Basic Problem Is SafetyâThe policy should direct a safety performance analysis potentially employing the same models or methods used above in derivation of new alignment criteria. The quantitative safety models would be adjusted to incorporate site-specific data through Empirical Bayes or other methods. The process would include but not be limited to revi- sions to curve geometry. Shoulder and/or roadside improvements would be allowable without them being considered mitigation of a design exception. The cost-effectiveness analysis would include site-specific construction and right-of-way costs (this consideration alone would pro- vide focus on the applicability of curve flattening as a potential solution). The design process for reconstructed curves may result in different outcomes than application of the new con- struction design criteria because of the influence of site-specific benefit estimates and cost calculations. Appendix C provides an example of a process to evaluate the curve design based on safety performance. Central to the geometric design process for existing curves is the base condition, which is the existing curve itself, comprising the radius, length, and central angle. The cost effectiveness of any revised design is measured by the expected benefits (reduced crashes, lower operating cost) compared with the reconstruction costs of the alternative. As both benefits and costs are uniquely defined by the base condition, the geometric design process cannot rely on dimensional guidance, but rather must rely on a cost-effectiveness analysis process. 6.5.1.3.3 Treatment of Curves on 3R Projects. A project for which the primary problem is infrastructure condition may still include consideration of design alternatives that address the curve geometry or cross section. The following process could be applied: ⢠Site analysis demonstrates a PSIâAn agencyâs HSIP or special safety program for curvature may not include a specific curve as its crash history does not reach the agency threshold; yet it may suggest a PSI. For 3R projects the presumption is that existing geometric features are adequate. Based on a curve-safety diagnostic review process, considering a full range of low- cost curve improvements that may include some geometric revisions, the 3R project may offer the opportunity to incorporate such improvements based on their expected benefit/cost analysis. Note that this process may apply on a site-specific basis or through a systemic safety program implementation (x, y). ⢠Site analysis demonstrates no PSIâWhere site-specific data are such that the curveâs PSI is zero, the 3R project should be performed without incorporating any improvements that would add substantially to the projectâs cost. An exception to this may be for agencies employ- ing a systemic curve-safety program that is typical for lower traffic volume roadways. Curves that fall within the guidance of the systemic program may have systemic improvements, which are by definition very low cost, included as part of the 3R project. 6.5.1.4 Framework for Design Criteria DevelopmentâHorizontal Alignment Given the wide range in context and in transportation priority based on road type, there is no reason to believe that a single model or approach for horizontal curve design will produce uniformly cost-effective or optimal solutions. The current horizontal curve design policy, which is based on driver comfort, is context insensitive (i.e., applies to all road types and all terrain); is volume insensitive; based on outdated research; and applies to both new design as well as reconstruction. There are multiple rational approaches to design for horizontal alignment, any of which may produce cost-effective solutions reflective of the context and project type. Figure 45 shows the AASHTO design approach and others that may be considered the basis for more performance-based design of curves.
Updating the technical Guidance on Geometric Design in the aaShtO policies 139 6.5.1.4.1 Alternative Design Vehicles. For some road types or conditions, a larger vehicle may be more appropriate as the basis for design. Vehicles with high centers of gravity overturn before they skid. Loop ramps, or two-lane roads with high proportion of truck traffic, or truck- only facilities may be candidates for an operational model based on avoidance of overturning. Special purpose roads by definition may be designed explicitly for special vehicles such as agricultural or resource recovery. The design vehicle characteristics of interest may vary. In some cases the offtracking character- istics may be the controlling operational condition. In other cases, the speed at which overturn occurs for design vehicles with higher centers of gravity may control. 6.5.1.4.2 Alternative Functional BasisâLoss of Control to Skidding. As an alternative to driver comfort, curvature could be controlled based on providing a margin of safety against loss of control from skidding (many designers are under the impression that this in fact is the basis for AASHTO curve design). This approach would entail modeling of critical driver behavior based on research and studies of available pavement friction or established pavement perfor- mance policies of an agency. For example, Glennon et al. (1985) demonstrated that passenger car drivers do not track curves as designed, but rather overshoot them (i.e., track a radius significantly smaller than the design). Functional Basis for Curve Design Design Vehicle Assumption Speed Input Assumptions Potential Geometric Interactions Comments Research Issues Driver Comfort Passenger Car Requires Design Speed Assumption None Current AASHTO approach, requires updated data and model Replicate studies using current vehicles and drivers, or potentially use SHRP2 naturalistic driver database Vehicle Overturn Potential Single Unit or Semi-trailer Requires Design Speed Assumption Could be combined with grade May be appropriate for special purpose roads, loop ramps, or roads with high proportion of large vehicles Determine relationship of curvature to overturn risk Driver Loss of Control Passenger Car Requires Design Speed Assumption Could be combined with grade Apply models of actual driver behavior through curves; establish margin of safety for range of pavement friction based on studies or agency policy Apply models of vehicle path and speed behavior (validate and update), potentially use SHRP2 naturalistic database; collect pavement performance data Offtracking of Critical Design Vehicle Semi-trailer or other long vehicle Noneâwould by definition apply to low-speed roads with minimal risk of severe crashes Could be combined with roadway or lane width May be appropriate for very low-speed and/or low-volume roadways Develop radius and width for low- speed turns based on AUTOTURN or other computer models Offtracking at Speed of Frequent Design Vehicle Bus, semi-trailer, or single unit truck Noneâwould by definition apply to moderate speed roads irrespective of speed Could be combined with roadway or lane width May be appropriate for collectors and urban arterials up to 40 to 45 mph Confirm and validate insensitivity of horizontal curvature to crashes on urban and suburban arterials, Conduct field studies observing offtracking at moderate speeds Cost Effectiveness Analysis, Quantitative Safety and Operating Cost vs. Construction and Maintenance Cost None Noneâprocess tests incrementally larger radii curves for their quantitative benefits Could be combined with shoulder width and roadside, automatically incorporates radius and length (or central angle) May be appropriate for two- lane highway reconstruction projects Model operating costs (fuel consumption, wear and tear), incremental safety benefits using HSM models for various road types Cost Effectiveness Analysis, Quantitative Safety and Operating Cost vs. Construction and Maintenance Cost, including effects of curvature on capacity and throughput None Noneâprocess tests incrementally larger radii curves for their quantitative benefits Could be combined with shoulder width and roadside; automatically incorporates radius and length (or central angle) May be appropriate for reconstruction of high-volume urban freeways Model operating costs (fuel consumption, wear and tear), incremental safety benefits using HSM models for various road types, study effects of curvature on capacity and include these Figure 45. Potential bases for development of horizontal alignment policy.
140 a performance-Based highway Geometric Design process This overshoot behavior was found to be independent of the speed. Krammes and Otteson (2000) showed that the 85th percentile speed at which drivers track curves at speeds is signifi- cantly greater than the AASHTO assumptions; with such speed behavior particularly variant in the 45 to 65 mph speed range. For a given design speed, criteria for new or reconstructed roads could be developed to speci- fications reflective of the agencyâs operating environment and policies: ⢠Select design friction supply associated with dry pavements in place (by road type if desired). ⢠Reduce the design friction supply if desired by the agency to reflect the significant frequency of wet or icy pavements, using historic weather records; and be consistent with the agencyâs asset management policies and approaches related to pavement condition and skid resistance. ⢠Establish by policy a margin of safety threshold for the maximum design friction demand. ⢠Select or establish the maximum superelevation for design policy. ⢠Establish the design tracking percentile behavior (say, 95th) and calculate the path radius tracked for the range of design radii. ⢠Compute the speed at which the margin of safety threshold is reached for the design track- ing behavior, which becomes the nominal maximum design speed for the curve radius and superelevation. As part of subsequent work in Phase II the above approach could be developed and compared to current design policy. This approach in all likelihood would allow for smaller radius curves, particularly on roads with lower design speeds. It is not clear what this approach may yield for higher design speeds. Clearly, the assumed friction and margin of safety will heavily influence the outcome. 6.5.1.4.3 Alternative Functional BasisâLarge Vehicle Offtracking. For road types and contexts in which there is no evidence of operational or safety criticality related to curvature, the basis for curvature could simply be providing for the offtracking and turning of the most critical legal vehicles that would use the road. Low-volume roads and roads in low-speed environments may be designed with this functional basis. Indeed, even higher class roads in urban environ- ments may be designed using vehicle turning and offtracking characteristics. For higher type and higher-volume roads, studies may be needed to characterize the paths drivers of such vehicles take at moderate speeds, to develop a policy that facilitates operation along curves under traffic. This approach also may be used to develop minimum roadway widths based on, for example, the total needs of design vehicles operating in opposing traffic. 6.5.1.4.4 Alternative Functional BasisâCost-Effectiveness Analysis. Curve design for reconstruction of existing roads may be best determined through a cost-effective analysis that incorporates site-specific data and conditions. The existing alignment is the baseline condi- tion, with traffic volume and composition, speed, and crash history available for the analysis. Incremental flattening of the curve to different radii could be tested by modeling the operating benefits (vehicle operating costs and delay) and safety benefits (using an appropriate SPF for curvature and Empirical Bayesâ methodology). The design approach could include the ability for the designer to test variances in shoulder width and road widening through the curve. The incre- mental construction and maintenance costs could be determined, and an annualized benefit/ cost analysis performed using a suitably long project life (e.g., 50 years). The agency could estab- lish by policy a level of benefit to cost that reflected their priorities and system affordability. The cost-effective approach may appear involved, but it can readily be streamlined using tai- lored software. Moreover, it has the benefit of directly incorporating site-specific conditions that influence the cost. The benefits being directly related to traffic volume, it will produce different
Updating the technical Guidance on Geometric Design in the aaShtO policies 141 outcomes for lower-volume than higher-volume roads, all other factors being equal. The out- come by policy is one that reflects the known attributes of system performance and agency poli- cies. Such a design outcome is exceptional rather than a design exception. 6.5.2 Summary of Revised Approach to Horizontal Curve Design The above approach to design of horizontal curves represents substantial change over the cur- rent comfort-based, one-size-fits-all approach. This approach will produce different solutions in different states and locations within a state, an outcome some may find difficult to accept. What must be understood is that the current process already produces different outcomes (based on different emax policies). The current process is based on an oversimplified operational model that ignores the effects of approach geometry on speeds, the combined effects of vertical alignment and cross section with curvature, and it does not differentiate vehicle types. Finally, the notion of comfort as a basis for investment and environmental impact decisions related to roadway design is inappropriate. Driver comfort when combined with the characteristics of vehicles and pavement conditions is presumably related to safety performance. The relationship, however, is tenuous, varies by context, and is a means to the ultimate end, which is safety and operational performance. The use of direct models, methods, and measures of curve-safety performance, including diagnostic approaches, when applied appropriately, should produce greatly improved performance and significant aggregate cost savings. Figure 46 is offered as a straw-man proposal to illustrate how design policy for curvature could be established within the recommended context proposal. Roadway Type Rural Natural Zone Rural Zone Suburban Zone General Urban Zone Urban Center Zone Urban Core Zone Local Based on oï¬-tracking requirements of larger design vehicles (nominal DS = 20 to 30 mph) or loss of control from skidding (DS = 40 mph) Based on loss of control from skidding Based on oï¬-tracking requirements of typical large vehicles (perhaps vary by road type and context zone) at very low speeds; urban buses, single unit trucks, semi-trailers Collector Based on loss of control from skidding Arterial Based on volume-sensitive, cost- eï¬ective design criteria derived from safety performance, operating cost; and infrastructure life-cycle cost; include interactive eï¬ects of grade as appropriate Based on loss of control from skidding Based on oï¬-tracking requirements of typical large vehicles (perhaps vary by road type and context zone) at moderate speeds Freeway Based on volume-sensitive, cost-eï¬ective design criteria derived from safety performance, operating cost, and throughput/capacity; and infrastructure life-cycle cost; include interactive eï¬ects of grade; include consideration of decision or stopping sight distance limited by horizontal curvature and median barriers Figure 46. Straw-man proposal for design of horizontal curvature for road types by context.
142 a performance-Based highway Geometric Design process 6.6 Sight Distance The concept of sight lines providing sufficient distance to drivers to perceive and react to conditions ahead is a central requirement of geometric design. The following summarizes the current approach to design for sight distance in the AASHTO Green Book: ⢠There are four distinct types of sight distanceâstopping, passing, intersection, and decision sight distance. These apply in varying ways to the roadway system. ⢠The functional models for each form vary, but all models have speed as a central param- eter; and all models require assumptions about location of the driverâs eye, the object being observed, and the human driving response. ⢠SSD is the most fundamental type of sight distance. According to AASHTO, it should apply to all locations and road types (with the narrow exception of the sight distance at the approach to a signalized intersection). ⢠The AASHTO SSD model is a rational operational model. It applies equally to all roadways and contexts. It is volume independent and context independent. ⢠The SSD model assumes passenger car parameters and operation at design speed. ⢠The passing sight distance (PSD) model is a rational operational model. It applies only to two-lane rural highways. Provision for geometry that provides PSD is not a requirement but rather a design choice that influences the operation (capacity) of the road. ⢠The ISD model is a series of rational operational models based on field studies of driver behav- ior and intersections with varying types of traffic control. ⢠The ISD model is context sensitive (i.e., it reflects the type of control, the design parameters of the intersection including number of lanes crossed or turning into, and grade). ⢠The DSD is a series of rational operational models based on human factors studies of driver behavior. ⢠DSD is not a requirement but rather is recommended for certain specific contexts or conditions. 6.6.1 Critique of AASHTO Sight-Distance Design Policy and Models The fundamental concept of sight distance is applied in varying ways in design policy. Some types of sight distance are required (SSD and ISD) and others optional (PSD and DSD). SSD applies in all contexts and road types; the others (ISD, PSD, and DSD) apply only in certain spe- cific cases. Per the above, SSD has the greatest influence on roadway design. The design of both horizontal and vertical alignment is directly influenced by the provision for SSD. 6.6.1.1 SSD The AASHTO SSD model was developed over 70 years ago and has remained essentially unchanged in its form since that time. The basic concept of the roadway providing sufficient sight lines for a driver to perceive and avoid a collision is fundamental. The execution of the con- cept requires either data or assumptions regarding the parameters of the design vehicle (height of eye, braking, or steering capability), the driver (perception and reaction, braking, or steering response), the conditions (assumed speed, pavement condition), and the nature of the event or object representing collision risk. The AASHTO SSD model is simple in its construct. It was developed to fit the limited data and knowledge in existence in the 1940s. The SSD model recognizes speed as a basic parameter. It is context insensitive (i.e., it applies equally to all road types, and conditions, irrespective of traffic volume). Over time as changes in the vehicle fleet were observed, the inputs to the model were revised, but the model itself has remained unchanged. Neuman pointed out the shortcomings of the current model in a paper published more than 25 years ago (Neuman 1989). The author proposed a risk-based framework for SSD design
Updating the technical Guidance on Geometric Design in the aaShtO policies 143 that included road type, traffic volume, and road context features. The model was described in theoretical terms; at the time the data and knowledge base for its development were lacking (The model was actually adopted by the authors of TRB Special Report 214 on 3R design criteria). The AASHTO SSD model adopts a crash avoidance performance metric as its basis (vehicle must be able to come to a full stop to avoid striking the object in the road ahead). The con- sensus that has evolved in transportation policy regarding safety now focuses on avoidance or elimination of crashes that produce serious injuries and fatalities, and not total crashes. As noted previously, mandated advances in vehicle technology to improve safety performance have sig- nificantly changed the relationship between injury risk and speed of collision. The principles associated with a risk-based SSD approach were alluded to in NCHRP Report 400 (Fambro et al. 1997) and referenced in work for AASHTO on design criteria for VLVLR. A maneuver sight-distance (MSD) model was proposed for application to very low- volume situations. The notion was that in low-risk cases, drivers need not necessarily come to a full stop to avoid collision; but rather could slow and maneuver around the object or prob- lem. While this model also was theoretical, it is useful in that it challenged the conventional, simple AASHTO model. The AASHTO SSD model is a simple human-factor based model that considers human response to perception of a conflict in the road ahead, and also relies on a human response mechanism (rate of deceleration or braking). The advent of eventual implementation of self- driving features in cars designed to perceive and adapt to risks is a long-term issue to consider. Provision for SSD per the current design policy can have significant impact on reconstruction of existing rural roads. Designers and those responsible for design decision making acknowledge that the SSD model is overly conservative. Crest vertical curvature, which is based on SSD, is among the most frequent design criteria for which a design exception is approved (Mason and Mahoney 2003). The AASHTO SSD model clearly has little applicability to the urban road context in which intersections are frequent, driveways frequent, and conflicts with crossing pedestrians and vehi- cles is the greatest risk factor. The majority of severe crashes are angle related or pedestrian related. For such facilities, the ISD model is more appropriate, and mid-block SSD as defined by an object in road is irrelevant for all but the most extreme terrain conditions. 6.6.1.2 Decision Sight Distance. Another aspect of the AASHTO sight-distance policy is the extent to which the DSD criteria are generally ignored. Few agencies (Caltrans being a notable exception) adopt DSD as a design control comparable to SSD in its importance. The principle behind DSD is the provision of sufficient notice to drivers to enable them to make deci- sions and take driving actions. The decision making and navigating tasks are more complex, take longer time, and require more sight distance to accomplish. The human response parameters input to DSD design were obtained from research in the 1970s and early 1980s. However, there is no direct evidence of the influence of DSD (or lack of DSD) on safety or operational perfor- mance of roadways. This is not surprising given the lack of direct evidence associated with the much shorter SSD dimensions. Despite the lack of direct evidence, there is indirect evidence of the potential influence of sight distance on the operation of freeways. Lane changing and the intensity of lane changing (frequency over a given length of roadway) are associated with increased crash risk and reduced throughput. This knowledge base is associated with the measurable effects of weaving traffic and weaving sections on freeways. The extent to which the presence or absence of DSD may influence the actual performance of lane changes is theoretical per the DSD model but yet to be quantified. Concentration of lane changing over shorter distances limited by lack of DSD may have a measurable effect compared
144 a performance-Based highway Geometric Design process with such lane changing occurring over greater lengths of roadway. High-volume freeways (those with average daily traffic in excess of 150,000 vpd) experience traffic densities consistent with LOS D to E throughout much of a typical weekday. Differences in the intensity of lane chang- ing influenced by lack of DSD may potentially translate to observable differences in throughput, which in turn may result in quantifiable operational benefits associated with DSD. DSD is either important (i.e., it offers measurable performance benefits) or it is not as a design control. Most likely its importance is restricted to higher-volume facilities operating at higher speeds. Proponents of DSD cite the human factors basis for DSD. Guide signing and advances in GPS technology to aid navigation may mitigate or eliminate the need for DSD to be incorpo- rated as part of geometric design. To determine the extent to which providing DSD marginally increases the cost of a design, more observable and quantifiable benefits should be determined. 6.6.1.3 Passing Sight Distance. AASHTO recently revised the PSD model to reflect research that corrected the assumptions imbedded in the previous model with respect to the critical nature of the passing model. The revised model produces shorter passing zones than the overly conservative previous model. PSD applies only to two-lane highways. PSD is somewhat unique in that it expresses or characterizes an operational attribute of the highway rather than a design requirement. Roads do not have to include passing zones, but the roads that include passing zones provide a higher LOS than those that donât. In this respect, PSD comes closest to a purely operational design criterion than others. Table 22 shows the effect of no-passing zones on the operation of a road, as published in the HCM. Designers may theoretically develop design alternatives that provide varying levels of PSD, measure the difference in LOS and travel time, determine the difference in costs, and make informed, optimization decisions on the objective operational and cost data. 6.6.1.4 Intersection Sight Distance. ISD criteria were recently revised to reflect performance- based research (Fambro et al. 1997). The HSM suggests a relationship between ISD and safety per- formance for unsignalized intersections. Current models for ISD are focused on vehicle-vehicle conflicts, as defined by the duration of accepted gaps. The appropriate critical gaps for design was found in NCHRP Report 383 (Harwood et al. 1996) to vary with the maneuver being made and the vehicle type making the maneuver. Research is currently underway in NCHRP Project 17-59 to determine a relationship between ISD and safety performance. In urban contexts, particularly collector and arterial conditions with speeds in the 30 mph to 40 mph range, the frequency of potential crossing conflicts and demands on driver atten- tion may present greater risks (or stated differently, longer driver reaction times) than is cur- rently used. Providing continuous sight lines for drivers related to crossing conflicts (both at Opposing Demand Flow Rate, Percent No-Passing Zones v0 (pc/h) ⤠20 40 60 80 100 FFS ⥠65 mi/h ⤠100 1.1 2.2 2.8 3.0 3.1 200 2.2 3.3 3.9 4.0 4.2 400 1.6 2.3 2.7 2.8 2.9 600 1.4 1.5 1.7 1.9 2.0 800 0.7 1.0 1.2 1.4 1.5 1,000 0.6 0.8 1.1 1.1 1.2 1,200 0.6 0.8 0.9 1.0 1.1 1,400 0.6 0.7 0.9 0.9 0.9 ⥠1,600 0.6 0.7 0.7 0.7 0.8 Table 22. Effect of no-passing zones.
Updating the technical Guidance on Geometric Design in the aaShtO policies 145 intersections and mid-block) may be the best performance basis for geometric design of urban roads and streets. Additional research is needed to confirm and quantify this issue. 6.6.1.5 Maneuver Sight Distance. The concept of MSD was proposed as part of NCHRP Report 400 (Fambro et al. 1997) on SSD. The concept was formulated to address the perceived low risk associated with full SSD for low-volume conditions. It expressed a sight distance necessary for a driver to perceive an object ahead, but rather than brake to a full stop, decel- erate and maneuver around the object. The concept assumed that no opposing conflict was present and that the roadway (lanes and shoulders) provided sufficient space to allow for such maneuver. MSD was proposed as a reasonable basis for VLVLRs (Neuman 1998). It is an example of criteria developed to reflect variable levels of risk and preclude unnecessary greater dimensions for SSD. 6.6.2 Potential Approaches to Sight-Distance Design Policy As long as the human driver is a necessary element of operation on the road system, the prin- ciple of sight distance in some fashion is central to geometric design. What is needed is a more objectively variable set of criteria reflecting differences in risk associated with traffic volume, road type, speed, and circumstances along the highway. Indeed, the development of a unified sight-distance approach seems appropriate. Figure 47 is offered as a straw-man proposal for sight distance based on the suggested context framework. 6.6.2.1 New Roads Sight-distance criteria for new roads may be developed using the above framework. In devel- oping new alignments, the design task of balancing the earthwork required for horizontal and vertical alignment is such that providing marginally longer vertical curvature can often be done with little additional construction cost, at least for roads in level and rolling terrain. The direct inclusion of traffic volume and risk relative to circumstances along the roadway could be translated into varying vertical curve lengths from basic (i.e., no unusual circumstances) to complex (potential conflict within the zone of limited sight distance). A skilled designer for new road alignment would have the flexibility to adjust the vertical alignment, or to relocate the Roadway Type Rural Natural Zone Rural Zone Suburban Zone GeneralUrban Zone Urban Center Zone Urban Core Zone Local Collector Arterial Freeway Intersection Sight Distance; Speed Regime assumed by policy Decision Sight Distance or risk-based stopping sight distance based on proven traffic operaÂonal and/or safety benefits of longer sight lines Risk-based Stopping Sight Distance (volume and roadway context sensiÂve); Passing Sight Distance for 2-lane and Intersection Sight Distance at Intersections Intersection Sight Distance; Speed Regime assumed by policy Risk-based Stopping Sight Distance (volume and roadway context sensiÂve) Maneuver Sight Distance (low speed and low volume) Figure 47. Straw-man proposal framework for design of sight distance for road types by context.
146 a performance-Based highway Geometric Design process potential conflict (intersection, driveway, horizontal curve) away from the zone of limited sight distance, allowing the use of a basic crest vertical curve. The tools of crash prediction and speed profiles/capacity would be applied to various designs, and quantities and right-of-way computed for each, enabling development of an optimal solution. To the extent that availability of PSD and DSD can be demonstrated by research, designers would have the information to test the addition of these at select locations and perform the optimization. 6.6.2.2 Reconstructed Roads The marginal costs of lengthening sight distance through alignment alternations are com- pletely different in the reconstruction environment. Any change in horizontal or vertical align- ment will typically be costly, and often include additional right-of-way. A central aspect of developing cost-effective solutions is linking the observed crash history with the presence and location of sight restrictions. Here, the use of SSD profiles as shown in Figure 48 becomes an essential tool. Where sufficient observed crashes coincide with the zone of limited sight dis- tance, a designer has the necessary evidence to justify a design change and its attendant costs. As such evidence would be expected to be observed on higher-volume roads, agencies would find themselves avoiding reconstruction to increase SSD on lower-volume roads, and only on select higher-volume roads for which the evidence of an SSD problem is translated to observed crashes. 6.6.2.3 3R Roads Under the assumption that a project designated as 3R in nature is purely related to infra- structure replacement or repair, the only sight-distance related improvements considered would include very low-cost signing, striping, and lighting. Dynamic warning signs at intersections near sight restrictions may be considered, for example. Source: HSM Figure 48. SSD profiles.
Updating the technical Guidance on Geometric Design in the aaShtO policies 147 6.7 Vertical Alignment Vertical alignment is composed of tangent grades and parabolic vertical curvature. Grades are expressed as a percent, and vertical curves as crest (an upgrade followed by a downgrade) or sag (a downgrade followed by an upgrade). The engineering profession adopted the use of parabolic (as opposed to circular) vertical curvature because it facilitated the calculation of elevations to two decimal places in the field. 6.7.1 Grade AASHTO road design policy considers the need for minimum grades and maximum grades for O&M reasons. Minimum grades (generally at least 0.3 percent) are intended to assure drain- age of the pavement; such criteria apply to all roads in all contexts. Maximum grade criteria are set to enable operation of low-powered trucks carrying full loads. The performance criterion is the ability of a truck with a given wt/hp ratio to climb an upgrade. The AASHTO criteria are sensitive to road type and terrain, but include no component of traffic volume. There is no reference to a design volume of heavy vehicles in the formulation of the criterion, nor is there reference to the length of the upgrade in the setting of the maximum grade criteria. Design policy does acknowledge the combination of grade and length of grade in determining need for auxiliary truck climbing lanes. Current grade criteria are based solely on motor vehicle operations (passenger cars and trucks). However, both steepness and length of grade substantially affect the operation of bicyclists. 6.7.1.1 Operational and Safety Effects of Grade Most roadways are two way in nature, so the operational effects of grade differ by direction of travel. The exception to this is one-way roads and one-way interchange ramps. The AASHTO policy provides qualitative guidance on downgrade versus upgrade ramps, but does not offer definitive values, and does not differentiate between entrance and exit ramps with respect to desirable grades. Grades have a quantifiable influence on the throughput of roadways as documented in the HCM. The determination of capacity uses the percent of heavy vehicles (although these are not specifi- cally defined with respect to wt/hp ratio). As heavy vehicles slow on steep grades, the throughput of vehicles is reduced. This is expressed in the passenger car equivalents (PCE). The effect of grade on capacity varies by road type (two-lane rural highways, multilane roads, freeways); two-lane highways are shown in Table 23. Grades also have a quantifiable influence on the safety performance of certain road types, including two-lane rural highways (Table 24). The effect of grade on urban and suburban arterials under most typical conditions has not been established. The effect of grade on safety Source: HCM, Exhibit 11-13 Table 23. PCEs for trucks and buses on specific grades.
148 a performance-Based highway Geometric Design process performance of freeways and interchanges is suspected but not known, as this geometric vari- able was excluded in the research that developed the safety prediction models for freeways and interchanges for the HSM. The grade and length of grade has a substantial effect on the ability of bicyclists to use the road. Table 25, shows design criteria for grade for reasonable bicycle operations. For certain roads in certain contexts, the gradability of bicycle operations may be a suitable or even controlling basis for vertical alignment design. 6.7.1.2 Construction Costs and Grade The construction cost of a new road is heavily influenced by the earthwork needed to con- struct the road. A fundamental design principle is to minimize the amount of earthwork, and balance the amount of cut and fill to eliminate the need to import or haul away earth. Assuming a given cross section, the primary geometric design means of cost minimization is through the use of grades to fit the terrain. It is common design practice for a preliminary alignment to be set, earthwork balances computed, and realignment performed as an iterative exercise to reduce the earthwork. This design exercise typically uses cross-section templates that themselves are devel- oped to minimize earthwork. Steeper cross-slopes for fill heights above certain dimensions are used, with flatter cross-slopes applied in more moderate terrain. The above typical design practice commonly focuses only on iterating grades and vertical curvature to minimize the earthwork. The use of varying cross-section templates also translates to different roadside designs, and specifically application of guardrail, barrier, and attenuation devices. Because such features in design are a relatively small proportion of the initial construc- tion cost, they are not typically included in the earthwork cost and design iteration. The cost of hauling earthwork can vary widely based on the context. Factors include the avail- ability of suitable sites near the project, the quality of available material, and the overall amount of material needed. Construction practices and costs have changed significantly over the years with the development of larger and more efficient earth-moving equipment. For rural roads on new alignment, the overall cost of earthwork is typically 15 to 30 percent of the total construction cost. Source: HSM Table 10-11 Table 24. CMF for grade of roadway segments. Grade Length (feet) 5 to 6% 800 7% 400 8% 300 9% 200 10% 100 ê11% 50 Source: AASHTO (1999) Table 25. Bicycle gradability.
Updating the technical Guidance on Geometric Design in the aaShtO policies 149 Reconstruction projects have less flexibility in design of grades than roads on new alignment. In most cases, the reconstruction design, which may include widening to add lanes, will closely follow the existing grades. Changes in vertical curvature and grade changes associated with realignment of the horizontal geometry may be made, but the reconstructed road will mostly follow the vertical alignment of the existing road. Significant grade changes over longer segments create problems in maintaining traffic flow and driveway or local road access to adjacent proper- ties during construction, making such changes unusual. 6.7.2 Crest Vertical Curvature The design basis for crest vertical curvature is provision for SSD. The AASHTO model for vertical curve design has remained unchanged since its inception. Vertical curves are designed to provide full SSD. The design parameters required to compute the necessary length include the design speed, the algebraic difference in the two grades, the assumed height of the driverâs eye, and the assumed height of the object for which crash avoidance is the design objective. The original model formulation in the 1940s was actually based on the concept of cost effec- tiveness in design criteria. Figure 49 summarizes the SSD model applied to crest vertical curve length. The driver eye height was established through surveying the vehicle fleet in use at the time. The object height parameter (4 inches) was derived from studies of the marginal costs of earthwork as a function of varying object heights, with the lower boundary being zero. Changes in object height (from 4 inches to 6 inches) were made in conjunction with observed changes in the vehicle fleet over time that required a lowering of the eye height. The rationale by AASHTO was to maintain calculated vertical curve lengths, as there was no evidence of safety problems with the curve length criteria. Contrary to the understanding of many engineers, the setting of object height was never intended to model any particular type of object until it was revised to 2 feet to account for taillight height. It was derived as a mathematical solution to a cost-optimization problem. More recent research to update the AASHTO model confirmed its lack of a relationship between crash risk and the amount of SSD provided by the roadway geometry (Fambro et al. 1997). Increasing the object height from its previous arbitrary 0.5 foot dimension to a 2 foot dimension addressed a portion of this issue, but the simplicity of the model and its lack of a traffic volume component to SSD design values call for a fresh look at SSD design. There was substantial resistance to the changes in AASHTO policy to 2 feet from 0.5 feet by many who did not understand the original basis for vertical curve design. SSD is volume, road type, and context insensitive. Vertical curve design is similarly insensi- tive to these considerations. Although designers may use longer than minimum SSD (just as Source: AASHTO Green Book Figure 49. AASHTO SSD model for vertical curvature showing eye and object height.
150 a performance-Based highway Geometric Design process horizontal curves milder than minimum controlling curves are used), the practice is unusual. A longer vertical curve may be used to lower a control elevation for an overcrossing structure, or perhaps to provide DSD, or to tie to another alignment such as a crossing facility at grade. Typi- cally designers default to the minimum vertical curve length per criteria. NCHRP Report 400 (Fambro et al. 1997) recommendations were adopted by AASHTO. As a result, the minimum vertical curve lengths for design purposes are now shorter than previously. While this has reduced the number of vertical curves that may be considered substandard, there remain many miles of alignment for which vertical curve lengths do not provide minimum AASHTO dimensions. 6.7.2.1 Operational and Safety Effects of Crest Vertical Curvature The geometry of vertical alignment produces greatly varying design outcomes for SSD. SSD profiles (Figure 49) are a useful, but not routinely used tool that show the amount of SSD provided by the vertical alignment at each location along the roadway. SSD profiles provide insights to the effects of a design. As drivers proceed on the upgrade, the effect of the grade gradually reduces the amount of SSD provided. Just prior to reaching the crest of the verti- cal curve minimum, limiting amount of SSD per design policy is provided. Once the vehicle proceeds beyond the crest, the sight lines open up quickly and the amount of SSD increases rapidly. The SSD profile is directional. The nature of the profile is defined by the relative difference in the two grades connected by the vertical curve. When a horizontal curve and roadside features coincide with the vertical curve, the sight line may be further affected. In such cases a three- dimensional sight line can be produced. Although SSD profiles have been presented and discussed by AASHTO for many years, they are not routinely plotted nor studied as part of the preliminary design phase, despite the fact that the road design software used by most agencies allows for their automatic production. Even a cursory review of SSD profiles such as are illustrated in Figure 49 provides a number of important insights: ⢠The length of the road by direction along which actual minimum sight distance provided is relatively short; for much of the alignment more than minimum SSD is provided to the driver. ⢠The location of the minimum sight distance by direction varies based on the grades and is unique to the context. ⢠The nature of the roadway itself within the limits of minimum SSD should be of greatest concern to the designer. Regarding the last point, the presence of geometric elements that may pose a risk to driver behavior or operations (e.g., the PC of a horizontal curve, intersection or major driveway, bus stop, pedestrian crosswalk, narrow bridge approach) coinciding with the location of minimum SSD should be flagged by the designer as a condition to be redesigned. Neuman (1989) hypothesized that not all crest curves presented equal risk profiles in his con- struction of an alternative SSD design model. The risk associated with any crest vertical curve was asserted to be a function of core exposure measures including traffic volume, the amount of the sight restriction, the length of highway over which the sight restriction occurred; and the presence (or absence) of risk-increasing features. Research on FHWAâs 13 controlling criteria for design, presented in NCHRP Report 783 (Harwood et al. 2014), confirmed that the relationship of SSD to safety is highly situational. The research examined the safety performance of Type I crest vertical curves (hillcrests) on
Updating the technical Guidance on Geometric Design in the aaShtO policies 151 rural, two-lane highways with SSD above and below the AASHTO SSD criteria and found that the presence of SSD below AASHTO criteria was not sufficient to produce elevated crash fre- quencies. However, when SSD limited the approaching driverâs view of a critical feature, such as a horizontal curve, intersection, or driveway, elevated crash frequencies were observed. This illustrates that SSD is much more important to substantive safety at some locations than others. This situational variation in the importance of SSD should be considered in the geometric design process, particularly for projects on existing roads with proven safety histories. Fambro et al. (1997) demonstrated the lack of quantitative safety relationship in crest curve geometry. Others have shown some evidence of the added risk based on unique features within the crest curve as postulated by Neuman. Given the above insights and importance of SSD and vertical alignment, further research to characterize the relative risk of varying lengths of crest vertical curvature is considered a high priority. Another consideration for crest vertical curves on low-speed access roads, alleys, and drive- ways where the control is to maintain enough clearance so the underside of a vehicle does not drag on the surface. The guidance such as that presented in NCHRP Report 659 (Gattis 2010) could be included in the revised process. 6.7.2.2 Sag Vertical Curvature The design basis for length of parabolic sag vertical curvature has consistently been simple operational models. The primary basis is to provide a minimum length of curve such that a headlight beam can show the roadway ahead and provide SSD. For roadways that are lit, a sec- ondary design control is to provide for a minimum length of sag curve based on studies of driver comfort. This secondary control produces minimum sag curve lengths of about half the head- light beam dimension. As with crest vertical curvature, the design basis for sag vertical curvature is independent of traffic volume, road type, and context. The geometry of sag vertical curves is rarely critical in alignment design. Designers can use vertical curves longer than the minimum, and many do so in situations where this improves the construction cost minimization or facilitates a vertical tie with another alignment. The most typical critical aspect of design for sag curvature is in managing drainage. The rationale for the use of headlight beam distance as a key basis for sag vertical curve design is unclear. Headlight beam distance is irrelevant during daytime operations. At night, most drivers outdrive their headlight beam distance on level tangents (i.e., choose a travel speed at which SSD exceeds headlight beam distance), and presumably do so on sag vertical curves as well. Fambro et al.âs (1997) work in NCHRP Report 400, leading to the change to a 2-foot object height for SSD representing the taillight height of a vehicle, suggested that the object that most needs to be seen by drivers on sag vertical curves will be illuminated (a special situation applies where an overpass structure is present on a sag vertical curve, as the structure will limit SSD). There is no known nor hypothesized relationship between sag vertical curve length and crash frequency. There is a small but measurable effect of sag vertical curvature on the operating speeds of vehicles accelerating onto a relatively steep (greater than 4 percent) grade. Longer sag vertical curves provide better transition to the grade and aid in maintaining speeds as the vehi- cle begins natural deceleration on the upgrade. The combination of grade and length of grade represents operationally limiting geometry that may be a meaningful, controlling feature of a design. A design criterion for sag curvature that involves longer than current design values may marginally but significantly influence throughput on roadways such as higher-volume freeways operating near their capacity. This potential operational-performance criteria for such facilities bears further study.
152 a performance-Based highway Geometric Design process 6.8 Vertical Clearance Vertical clearance is the dimension between the roadway surface and overcrossing structure. Sufficient clearance is necessary to enable legal-height vehicles to pass beneath the structure. Vertical clearance dimensions influence the vertical alignment or profiles of each roadway. The difference in elevation between the two roadways translates to earthwork and construction costs. The operational and cost effects of providing minimum vertical clearance dimensions can be multiplicative at system interchanges on freeways. The geometry and cost to construct four or five level interchanges is highly influenced by the vertical clearance dimensions employed. The consequences of low vertical clearance are collision of the vehicle with the overhead structure. In most cases such hits involve overheight vehicles or those with improper loads, but while rare, such collisions can be catastrophic. In addition to creating a severe crash, such colli- sions can damage and in some cases destroy the bridge or require its complete reconstruction. Legal-height limitations vary by state and are in the range of 13.5 to 14 feet. AASHTO specifies a 16-foot minimum clearance dimension for new structures. Many states apply criteria that add 0.5 feet to this minimum dimension to provide for future pavement overlays. AASHTO allows for retention of existing structures for which the vertical clearance is less than 16 feet, with the lowest dimension allowable being 14 feet. The 16-foot dimension originated with the 1956 Interstate System of Defense Highways Act. The design criteria were set to enable transport of military vehicles and equipment. This provi- sion remains in force. There were many bridges built to lesser dimensions that became part of the Interstate system. Even when reconstructed, lesser dimensions were retained because of the difficulty of revising crossing roadway vertical alignment and/or designing a bridge to a depth to allow the full dimension. States must either provide this dimension for all Interstate highway bridges or provide an alternate route through or around a city that fully provides this dimension. The full 16-foot dimension is no longer needed for the originally intended purpose. Vertical clearances of 15 feet adequately address the requirements of the legal vehicle fleet. The marginal costs of constructing bridges to 16 or 16.5 feet versus 15 feet are negligible; embankment costs and design complexity in setting the vertical alignment are the greatest potential savings. Also, where multiple levels are constructed the aggregate differences in cost may be somewhat greater. The consequences of low vertical clearances would appear in databases showing bridge hit crashes and bridge repair records. The design process for reconstructed bridges or reconstructed roads with lesser vertical clearances should involve interrogation of such databases and, as neces- sary, interviews with agency staff to determine the actual risk of retaining lower clearances. The costs and difficulties of increasing clearance even by a foot or less can be substantial depending on the context. 6.9 Medians and Access Control Current AASHTO design policy treats the presence and type of median as a necessary feature for fully controlled access facilities (freeways), but as a design choice for all other types. The types (raised, mountable, flush, and types of barriers used) and dimensions of medians may vary widely, with the width based on provision for other geometric elements such as shoulders and horizontal sight distance; and inclusion of other roadway functions such as drainage and locat- ing overhead signs, lighting, and other similar structures. Medians serve to separate high-speed traffic and enforce limitations or better facilitate access, in particular, for left-turning access into and out of adjacent land uses.
Updating the technical Guidance on Geometric Design in the aaShtO policies 153 There is substantial research on the safety performance of medians by their type and relation- ship to driveway and intersection frequency and level of traffic (AASHTO 2010). The AASHTO HSM crash prediction methods include median presence and type as significant variables for all road types. The trade-offs involved with median design decisions will typically include the relative safety performance vs. both inconvenience created by out of direction travel, and economic loss attrib- utable to changes in accessibility to certain types of commercial land uses. Such trade-offs are generally associated with reconstruction projects, including conversion of one road type to another. Operational inconvenience costs to users are theoretically possible to model and quan- tify at the project level. Although research has been done on the effects of access management to business, the potential economic losses are less quantifiable on a case-by-case basis. For roads on new alignment the marginal costs and benefits of including (or not) medians with given dimensions and type should be quantified and incorporated into design criteria for new roads.
