U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-16-064    Date:  November 2016
Publication Number: FHWA-HRT-16-064
Date: November 2016


Traffic Bottlenecks: Identification and Solutions


Definition of Bottlenecks

For at least 15 years, FHWA has stated that 40 percent of all congestion nationwide is attributable to bottlenecks, while another 5 percent is attributable to poor signal timing, which is also a tangible problem and not a random event.(5) These two causes together are said to define recurring congestion and account for 45 percent of congestion. While this number has been oft-quoted in FHWA literature and widely used by others, it is purportedly based primarily on a congestion study undertaken mostly for, or using data mostly from, the I-95 Corridor Coalition. (See references 24, 35, and 36–39.) Whether or not recurring congestion is 45 percent of all congestion is less the issue than having confidence that recurring causes were correctly identified and aggregated because it is also safe to say that said study did not break out, identify, or otherwise stratify the indicators.(5) In short, recurring congestion just exists, and all other congestion is considered nonrecurring (e.g., special events, bad weather, work zones, and traffic incidents) (see figure 1). Regardless, it is fair to say that no confirmation or additional study has been made since. There is little consensus on a bottleneck definition in the literature other than to say that bottlenecks, by whatever cause, result in the formation of queues upstream of the bottleneck and free-flowing traffic downstream. According to a theory proposed by Kerner, traffic is categorized into three phases: free-flow, synchronized flow, and wide-moving jams.(40) Free-flow is an unrestricted condition. Synchronized flow entails localized congestion. Kerner’s synchronized traffic is fixed at a downstream front, which is a defacto bottleneck. Wide-moving jams are more or less systemic congestion, as the downstream traffic velocity moves, even as traffic may propagate upstream through bottlenecks.

The Merriam-Webster Dictionary defines a bottleneck as: (1) a narrow route or (2) a point of traffic congestion.(41) However, a road does not necessarily have to narrow for a bottleneck to exist (e.g., bottlenecks caused by a weaving condition, sun glare, rubbernecking, or a vertical climb). A layman’s understanding of a bottleneck might be too many cars trying to use a highway at the same time. A bottleneck is distinct from congestion because it occurs at a specific location and not pervasively along the entire corridor. The term “traffic bottleneck” infers a localized congestion problem, not a systemic congestion problem. A physical bottleneck cause (e.g., a lane drop or other operationally deficient design on the roadway) only manifests itself when traffic demand overwhelms the available roadway capacity. Otherwise, the design operates safely (otherwise it would have drawn attention on that merit) and unobtrusively, and no one is delayed. The takeaway here is that in the case of a recurring bottleneck location, it takes both a deficiency in the design as well as traffic overburden for the delay to occur, whereas a nonrecurring event is the sole causal event that causes traffic.

Attributions of Bottlenecks

Many papers have referenced the definition used by Daganzo, which attempts to explain a bottleneck’s signature; namely, a bottleneck is considered to be “active” if traffic is detected to be “queued upstream of the location and unqueued downstream.”(pg. 106)(42) Geroliminis et al. provides another useful bottleneck definition, as follows: “A phenomenon where the full performance level (capacity) of an entire system cannot be realized due to an abnormality at a single component of the system. The performance at one location thus brings down the performance of the entire system.”(pg. 5)(43) In supporting Geroliminis’ contention, it is also FHWA’s Office of Operation’s observation that locally subordinate locations are often indeed the defacto “congestion” that is unfairly attributed to the entire facility (e.g., the highway is congested when in fact only the vicinity of eastbound (EB) Exit 12 is the problem.).(7) This is akin to one bad apple unjustly tainting the entire crop.

Severe congestion can be caused at locations where a physical reduction in roadway width (e.g., lane blockage or lane drop) occurs. This cause may be called a “physical bottleneck.” Physical bottlenecks have been the focus of transportation improvements for many years for the simple reason that they are tangible and correctable by redesign in contrast to nonphysical causes of slowdown like rubbernecking, weather, or weaving. Regardless, motorists only care about the result, not the reason behind the bottleneck. Some of these locations are notorious to the extent that they have acquired colorful nicknames by the local motorists, such as the following:

NCHRP Report 3-83, Low-Cost Improvements for Recurring Freeway Bottlenecks, emphasizes that hidden bottlenecks occur when downstream demands are metered by upstream bottlenecks.(44) When hidden downstream bottlenecks are activated by improvements to upstream bottlenecks, it potentially causes negative system-wide impacts. This makes the case for using predictive tools and real-time data to drive bottleneck solutions even more compelling.

The degree of congestion at a bottleneck is related to its physical design. Some bottlenecks exist on roadways constructed many years ago using designs that were appropriate at the time but are now considered antiquated. Others have been built to sufficiently high design specifications but are simply overwhelmed by traffic. In the FHWA report, Recurring Traffic Bottlenecks: A Primer Focus on Low-Cost Operational Improvements, FHWA identified the worst physical bottlenecks in the country and examined the potential benefits that improving them could have for travel times, safety, emissions, and fuel consumption.(2) Ultimately, many definitions and attributes for bottlenecks arose from the literature. However, Margiotta and Spiller were able to summarize the different attributions of bottlenecks (see table 3).

Table 3. Traffic bottleneck attributes.(2)
Bottleneck Characteristic Description
  • Recurring: Predictable in cause, location, time of day, and approximate duration.

  • Nonrecurring: Random (in the colloquial sense) as to location and severity. Even if planned in some cases, like work zones or special events, these occurrences are irregular and are not predictably habitual or recurring in location.

Causes Recurring operational causes: A facility determinate condition wherein a fixed condition (the design or function of the facility at that point) allows surging traffic confluence to periodically overwhelm the roadway’s physical ability (i.e., capacity) to handle the traffic, resulting in predictable periods of delay.
Examples Recurring: Ramps, lane drops, weaves, merges, grades, underpasses, tunnels, narrow lanes, lack of shoulders, bridge lane reduction, curves, and poorly operating signals.
Supplementary terms
(applies to both recurring and nonrecurring bottlenecks)
  • Active bottlenecks: When traffic released past the bottleneck is not affected by a downstream restriction (i.e., queue spillback) from another bottleneck.

  • Hidden bottlenecks: When traffic demand is metered by one or more upstream bottleneck(s) (i.e., either a lesser or nonexistent bottleneck that would increase or appear, respectively) if only unfettered.

Identification methods
(applies to both recurring and nonrecurring bottlenecks)
Motorists typically refer to bottlenecks in terms of added time delay when compared to the same nondelayed trip, but engineers and agencies also measure performance data: average speed (travel time), lane densities, queue lengths, queue discharge rates, vehicle miles traveled, and vehicle hours traveled.
Measurement methods
(applies to both recurring and nonrecurring bottlenecks)
Data are collected using manual techniques (e.g., floating cars, aerial photography, or manual counts from video recordings) or from dynamic surveillance (e.g., detectors, radar, video, etc.) collected in real time. Modeling, especially microsimulation, can be used to study the impacts of bottleneck remediation on upstream and downstream conditions.
  • Recurring—type I: Demand surge, no capacity reduction (typically at freeway on-ramp merges).

  • Recurring—type II: Capacity reduction, no demand surge (typically changes in freeway geometry; lane drop, grade, or curve).

  • Recurring—type III: Combined demand surge and capacity reduction (typically in weaving sections).

  • Nonrecurring: Usually classified by event type (e.g., incident or work zone) and impact severity (e.g., duration of the number of lanes lost, closed, or impassable).

Signature trigger
  • Recurring: Bottleneck is due to overdemand of volume (i.e., peak-hour conditions). The bottleneck clears from the rear of the queue as volume declines.

  • Nonrecurring: Bottleneck is due to loss of capacity due to an incident, or short-term overdemand due to a spot event. The bottleneck clears from the front or rear of the queue, depending on whether the cause is incident-related (former) or volume-related (latter), respectively.

Dissipation criteria
  • Recurring: When volume overdemand drops back to manageable levels for available capacity (i.e., when off-peak conditions return).

  • Nonrecurring: When dynamic event is removed, queue should dissipate thereafter.


Bottleneck Identification Framework

Bottlenecks can have various and interrelated causes. This section focuses on identifying geometric or operational challenges associated with a bottleneck location and tailoring a solution or set of solutions to alleviate congestion. A given bottleneck location can have one or more causes. One cause can have multiple solutions, and one solution may address multiple causes, which is discussed further in the next section. Figure 29 shows a decision tree to help transportation agencies identify potential causes of an active bottleneck. The decision tree is divided into geometric and operational challenges. Further subdivisions within the two categories delve into exactly what aspects of geometry or operations may be causing bottleneck conditions.

The appendix contains a detailed description of each potential bottleneck cause in figure 29 and identifies potential solutions. Note that the numbers in the figure correlate to the numbers provided in the appendix. Each subcategory has a description or definition, a list of theoretical or empirical effects of that potential cause, and identification of existing and proposed solutions, where possible.

  1. Try to classify the bottleneck according to the seven categories listed in figure 29.

  2. Browse the available bottleneck causes listed under the chosen category.

  3. Refer to the appendix to obtain additional information on associated mitigation strategy (or strategies) for the chosen bottleneck cause.

  4. Consider the applicability of several mitigation strategies given in chapters 4 and 5 whose details are embedded within those sections.

For example, suppose a bottleneck is periodically identified on a section of freeway with none of the typical physical characteristics that tend to cause bottlenecks (e.g., lane drops, nearby ramps, horizontal or vertical curvature, etc.). If further study determines that this bottleneck tends to form during periods of poor weather, this could potentially be classified under the “Geometric—Roadway Specific,” “Operational Challenges—Agency Related,” and “Operational Challenges—Driver Related” categories listed in figure 29. Regarding the “Geometric—Roadway Specific” bottleneck cause sun glare, which is a weather-related issue, one suggestion from the appendix is the installation of redundant traffic control devices. Regarding “responses to weather” under “Operational Challenges—Agency Related,” the appendix lists options including education campaigns, reflective lighting, roadside lighting, roadside shelters for pulling over, signage (e.g., lights on when raining), salting or other material to increase traction, and restrictions during weather events. Regarding “unsafe vehicle conditions for weather conditions present” under “Operational Challenges—Driver Related,” the appendix discusses vehicle inspection requirements and electronic enforcement. Regarding chapters 4 and 5, none of these solutions specifically target weather-induced, nonrecurrent bottlenecks. Therefore, the framework and playbook provided in figure 29 and the appendix produce a useful set of considerations and/or options in this case.

This flowchart provides a framework to identify causes of bottlenecks. The top is labeled “Active Bottleneck” that flows to seven boxes labeled (from left to right) (1) Geometric—Roadway Specific, (2) Geometric—Facility Specific, (3) Geometric—Specific to Interchanges, (4) Geometric—Intersections/TCD/ITS, (5) Operational Challenges—Agency Related, (6) Operational Challenges—Driver Related, and (7) Operational Challenges—Nonmotorists Related. Each of the seven boxes flows into an additional box that includes a list of items. Under the first box, (1) Geometric—Roadway Specific, is the following list: 1. Design Speed, 2. Number of Lanes, 3. Lane Width, 4. Presence and type of shoulders, 5. Lane drops, 6. Lane Reduction Transition, 7. Hz clearance, 8. VI Clearance, 9. Sun Glare, 10. Hz Alignment, 11. VI Alignment, 12. SSD, 13. Pavement friction/surface, 14. Cross Slope, 15. Superelevation, 16. Access Points, 17. Mid-block Crossing, 18. Medians, 19. Lighting/Glare, 20. Marking, 21. Bicycle Lanes, and 22. Separation Type of Managed Lanes. Under the second box, (2) Geometric—Facility Specific, the list is as follows: 1. Bridges, 2. Tunnels and underpass, and 3. Collector-distributor network. Under the third box, (3) Geometric—Specific to Interchanges, the list is as follows: 1. Merge/Diverge Sections, 2. Auxiliary Lanes, 3. Weaving Areas, 4. On-ramp/Off-ramp, and 5. Acceleration/Deceleration lanes. Under the fourth box, (4) Geometric—Intersections/TCD/ITS, the list is as follows: 1. Intersection Sight Distance, 2. Left-turn and RT lane overflow, 3. Parking, and 4. TCD (signal, stop sign, etc.). Under the fifth box, (5) Operational Challenges—Agency Related, the list is as follows: 1. Managing Demand, 2. Intersection Spacing, 3. Interchange control, 4. Policy on entry/exit ramp placement, 5. Posted Speed limit (static/dynamic),. 6. Signal Timing, 7. Traffic Composition, 8. Work Zone, 9. Roadway Closure, 10. Incident Management and clearance, 11. Ramp Metering, 12. Heavy vehicles exclusion/prohibition for certain lanes/routes, 13. Managing lanes, 14. Response to weather, 15. Over height management policy, 16. Congestion pricing, 17. Toll booths operation, 18. Service patrols, 19. Law enforcement policy/location, and 20. Forecasting traffic demand. Under the sixth box, (6) Operational Challenges—Driver related, the list is as follows: 1. Bunching vehicle, 2. Roadside Distractions/rubbernecking, 3. Nonroadside distractions, 4. Unsafe vehicle condition for weather condition, 5. Aggressive lane change/weaving, 6. Driving unauthorized roadway section, 7. Driver performance in wz, 8. Driver performance when involved in an accident, 9. Driver performance on a roadway with an incident, 10. Driver performance with regard to emergency vehicle, 11. Driver performance wrt TCD, 12. Driver performance wrt conventional and alternative intersections, 13. Driver performance wrt peds, cyclists, 14. Driver performance wrt animal crossing, and 15. Driver performance wrt commercial and heavy vehicle operation. Under the seventh box, (7) Operational Challenges—Nonmotorists Related, the list is as follows: 1. Sub-optimal pedestrians and bicyclist performance. A key in the bottom left-hand corner indicates that Hz representable horizontal, Vl represents “vertical,” and wrt represents “with respect to.”

Figure 29. Flowchart. Framework to identify causes of bottlenecks.




Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101