Office of Operations
21st Century Operations Using 21st Century Technologies

Recurring Traffic Bottlenecks: A Primer
Focus on Low-Cost Operational Improvements

Understanding Bottlenecks

What Exactly is a “Traffic Bottleneck”?

Webster’s Dictionary defines a “bottleneck” as: i) a narrow or obstructed portion of a highway or pipeline; or ii) a hindrance to production or progress. Certainly the elemental characteristics of traffic bottlenecks exist in these descriptions. However, a road does not necessarily have to “narrow” for a bottleneck to exist (e.g., witness bottlenecks caused by a weave condition, sun glare, or a vertical climb). A bottleneck is distinguished from congestion in that it occurs at a specific location, and not pervasively along the entire corridor.

Traffic bottlenecks (hereafter, bottlenecks) have a myriad of causes. The most egregious ones tend to be freeway-to-freeway interchanges, but we all know that smaller, lesser chokepoints are frustrating too. Bottlenecks can be areas where traffic is merging, diverging, or weaving, or where other physical restrictions exist like narrow lanes, lack of shoulders, steep grades, and sharp curves. The fact that many recurring locations are “facility determinate,” i.e., the design condition contributes to the resulting congestion. Facility design is a tangible feature that can always be improved; however the cost or the necessary right-of-way may be prohibitive. Alternately, demand can be reduced so that the bottleneck performs better. The LBR program is focused on the infrastructure side.

Photograph - This image shows freeway congestion due to a bottleneck.

“Good News” and “Bad News” About Fixing Bottlenecks

The FHWA estimates that 40 percent of all congestion nationwide can be attributed to recurring bottlenecks (i.e., inadequate physical capacity) and another 5 percent is attributable to inefficient traffic signalization. The good news is that all these things are potentially correctable with mitigation strategies and roadway improvements. The bad news is that there are many, many candidate locations, and agencies are fiscally constrained on how much they can do. A tabulation of the top 25 bottlenecks, compiled by INRIX in the National Traffic Scorecard 2010 Annual Report, is shown in Exhibit 2. Their analysis uses raw data which comes from their historical traffic data warehouse along with discrete “GPS-enabled probe vehicle” reports from vehicles traveling the nation’s roads – including taxis, airport shuttles, service delivery vans, long-haul trucks, and consumer vehicles.

Exhibit 2. The Worst Physical Bottlenecks in the United States (2010)
2010 Rank Area Road/Direction Segment/Interchange State Length (Miles) Hours Congested Average Speed When Congested
1 New York Cross Bronx Expressway WB/I-95 SB Bronx River Parkway/Exit 4B NY 0.35 116 11.3
2 New York I-95 NB U.S. 9/U.S. 1/U.S. 46/Exit 72 NJ 0.42 109 9.2
3 Chicago Dan Ryan Expressway/I-90/I-94 WB Canalport Avenue/Cermak Road/Exit 53 IL 0.40 105 11.3
4 New York Cross Bronx Expressway WB/I-95 SB I-895/Sheridan Expressway/Exit 4A NY 0.51 133 13.0
5 New York Cross Bronx Expressway WB/I-95 SB White Plains Road/Exit 5 NY 0.28 105 12.1
6 New York Harlem River Drive SB 3rd Avenue NY 0.16 98 10.6
7 Chicago Dan Ryan Expressway/I-90/I-94 WB Ruble Street/Exit 52B IL 0.12 115 14.5
8 Chicago Dan Ryan Expressway/I-90/I-94 WB 18th Street/Exit 52C IL 0.41 107 13.4
9 New York Cross Bronx Expressway WB/I-95 SB Westchester Avenue/Exit 5 NY 1.15 91 11.7
10 Los Angeles Hollywood Freeway/U.S. 101 SB Vermont Avenue CA 0.62 117 16.7
11 Los Angeles San Diego Freeway/I-405 NB I-10/Santa Monica Freeway CA 1.23 91 14.1
12 New York Harlem River Drive SB 2nd Avenue/125th Street/Exit 19 NY 0.22 110 13.0
13 Chicago Kennedy Expressway/I-90/I-94 EB Ohio Street/Exit 50B IL 0.38 100 14.2
14 Chicago Dan Ryan Expressway/I-90/I-94 WB Roosevelt Road IL 0.22 111 16.4
15 New York Van Wyck Expressway/I-678 NB Hillside Avenue/Exit 6 NY 0.12 103 15.2
16 New York Van Wyck Expressway/I-678 NB Liberty Avenue/Exit 4 NY 0.52 86 12.8
17 Chicago Kennedy Expressway/I-90/I-94 EB Lake Street/Exit 51A IL 0.43 107 15.3
18 Los Angeles Hollywood Freeway/U.S. 101 NB Alameda Street CA 0.27 102 14.0
19 Los Angeles Hollywood Freeway/U.S. 101 NB Spring Street CA 0.14 110 16.4
20 San Francisco CA 24 WB Gateway Boulevard/Exit 7A CA 1.12 66 11.8
21 New York Harlem River Drive NB Lower Level Washington Bridge NY 0.11 108 14.1
22 New York I-95 NB NJ 4 NY 0.81 81 12.1
23 New York Major Deegan Expressway/I-87 NB 153rd Street/River Avenue/Exit 6 NY 0.29 79 11.6
24 Los Angeles Hollywood Freeway/U.S. 101 SB Melrose Avenue CA 0.35 97 17.3
25 New York Gowanus Expressway/I-278 EB NY 27/Prospect Expressway/Exit 24 NY 1.32 107 16.5

Source: INRIX National Traffic Scorecard 2010 Annual Report.

Understanding Merging at Recurring Bottlenecks

This guidance document focuses on “localized” recurring bottlenecks (i.e., point-specific or short corridors of congestion due to decision points such as on- and off-ramps, merge areas, weave areas, lane drops, tollbooth areas, and traffic areas); or design constraints such as curves, climbs, underpasses, and narrow or nonexistent shoulders.

So You Think You Can Merge?

Are you a “profiteering” lane merger, who seeks only your own personal gain, or are you an “altruistic” driver who yields to others for the benefit of all? Are you an “early merger” (upstream of the point of confluence) or “late merger” (at the last possible moment)? Are you “left-brain” or “right-brain; ” Republican or Democrat; plastic or paper? In the end, there is no right or wrong, legally speaking. When and how one merges is more a study in human behavior, and less a study in efficiency.

The Difference in Merging for Recurring and Nonrecurring Conditions

Merging maneuvers at recurring bottlenecks are essentially “cat herding” with implicit rules (often local in culture or habit) at best. Typically, not much guidance is given – everyone is “on their own.” Drivers “suddenly” encounter taillights ahead and slow, then “swarm” to get past it, whereas, in a nonrecurring event, there is more apt to be advance warning and instruction in the form of orange cones, signs, flagmen, or police. There is often direction to motorists how (“Take Turns”), where (“Merge Here”), and even what (“All Lanes Thru”) to do/expect, and there can even be enforcement (of lane jumpers) or simply order (traffic cops) from chaos. One might argue “What’s the difference? I’m in bumper-to-bumper traffic regardless!” The great difference is the greater potential (in nonrecurring) for herding those cats.

Controlling the chaos of lane merging is fundamental to advanced traffic operations strategies. Ramp metering has long been used to limit the number of merges at a recurring bottleneck in order to prevent breakdown of traffic flow. In nonrecurring situations the “dynamic lane merge” or “lane control” is increasingly used where an incident or work zone has “stolen” a lane. This strategy proactively directs motorists to both slow down and to get into the appropriate travel lane well in advance of the problem. Active Traffic Demand Management (ATDM) strategies take and advantage of lane control as well as other types of actions to balance demand and available capacity. Several U.S. examples of this strategy already exist and more are planned: I-35W in Minneapolis and U.S. 2 in Seattle have functioning systems. I-66 in Virginia has an older system in place that will be upgraded over the next few years

Which is Best? “Early” or “Late” Merging?

Can a better recurring merge be developed? Merging takes place at-speed or “at-crawl.” The former is most often associated with free flow on-ramp maneuvers, while the latter is most often associated with bumper-to-bumper congestion. In either condition the motorist has the additional choice to merge “early” (upstream) or “late” (at point of confluence). This creates a matrix of four possible merge conditions; 1) at-speed “early;” 2) at-speed “late;” 3) at-crawl “early;” and 4) at-crawl “late.” To further complicate things, guidance concerning where, when, and how best to merge can vary from modest to no forewarnings in recurring conditions to fully deployed Traffic Control Plans (TCP) in nonrecurring conditions. Given that this Primer is focused on the recurring bottleneck genre, the purpose of this section was to research if early or late merging was best for these noncontrolled situations; i.e., when no active TCP exists.

What Instruction is Given to Motorists?

On the whole, drivers are typically left to their own strategies as to how to merge together at recurring chokepoints. Personal preference, impatience, frustration, speed differentials, and other human and vehicular traits conspire to influence safety and reduce efficiency. Altruistic drivers are unselfish and yield – in varying degrees – to proactive drivers, who seek only their own benefit to cut in line. The only real conclusion that can be drawn is suggested by the similarity in methodologies used in the work zone studies. Specifically, in setting up “Dynamic Work Zones” these are essentially systems that are “on” when traffic volumes are high and “off” when traffic volumes are low. The mere fact that all of these trials presumed to set up – and study – a “late merge” scenario speaks to the engineering community’s penchant towards this method over the “early merge” option for stop-and-go conditions. One theme, however, remained constant. Regardless of the amount of forewarning and direction given to motorists (e.g., “light” guidance in recurring situations and “heavy” guidance in nonrecurring situations) personal preference seemed to win the day. Absent absolute enforcement, motorists were observed to – or opined to – merge when and how they preferred, with less regard for any instruction.

Recurring Congestion –

“Close to half of all congestion happens day after day at the same time and location.”

Source: Focus on Congestion Relief

Early Attempts to Direct Motorists How to Merge

When the Interstates were built in the 1960s and 70s there was often “instruction” by local engineers and the media of how to engage Interstate ramps, acceleration and deceleration lanes, etc. Of course, at that time, traffic was less congested on the whole, and the merging and diverging were essentially lessons in how to enter and exit Interstates. Academia has touched on queue theory, gap analysis, and related safety-oriented aspects, but none of these studies have focused much on educating motorists how to merge efficiently, unless one considers a “queue” or a “traffic stream” as an entity that can deduce instruction. Nevertheless, the academic community has essentially confirmed, via queuing theory and microsimulation that the discharge rate after the merge governs congestion on the segment. In layman’s terms, there is a finite capacity of the single lane downstream of the constriction. Very little of what happens upstream can refute the laws of physics; that only one vehicle can occupy the discharge space at a time; and in a jammed situation, the lead vehicle does so from essentially a crawl speed.

Excepting for some basic, generic instruction in states’ drivers manuals (“wait for a safe gap in traffic” – typ.) little has been done at the national level to educate drivers how to merge safely and efficiently, as compared to other national education efforts promoting seat belt compliance, school zone safety, traveler information, or pedestrian rights and practices. The perceived reason for this may simply be the expectation that there will always be drivers who feel they know best how and when to merge in a queue, irrespective of any instruction to the contrary. The altruistic view is to leave gaps, yield to your neighbor, take your turn but don’t force your turn, and generally don’t deny him or her entry into your lane. The more proactive view is to take first opportunity to cut in line, perhaps “line jump” to chase whichever line seems to be moving, and scuttle the principles of any orderly manner. Anecdotal evidence from many local traffic blogs and an Internet search finds strong sentiment from both camps as to why they think their method is best.

Merge Principles

How can we increase the efficiency of merging prior to the discharge point? In two words – be orderly. Not surprisingly, safety improves too. It is repeatedly shown that traffic is inherently safer when all vehicles are traveling at or near the same speed. Think of an orderly progression on a crowded escalator. Everyone is safely cocooned because they are going the same speed. Now imagine the bumping and chaos that would occur if impatient folks pushed past others.

Photograph - This image shows a densely packed freeway traffic.

Principle #1: “Go Slow to Go Fast”

“Go slow to go fast” is an increasingly trendy expression in traffic circles. It speaks to the seemingly paradoxical idea that if we slow down the rate of our “mixing” we can get past a constriction faster. A well known example (actually the winning entry in a 2006 contest to demonstrate the meaning of “throughput maximization”) is the “rice” experiment. In the first case, dry rice is poured all at once into a funnel. In the second case, the same amount is poured slowly. Repeated trials generally conclude about a one-third time savings to empty the funnel via the second method. And, it should be noted, there is a tipping point reached as one graduates from a v-e-r-y slow pour, to a medium pace, and so on. What lesson does the rice experiment teach us about traffic? The densely packed rice (or traffic) in the first trial creates friction in the literal sense and the practical sense, respectively. The denser the traffic, the smaller the safety cushion around each driver, and the more cautious (i.e., slower) he becomes. A classic “bell curve” diagram also serves to explain how traffic throughput reaches an apex up to the point where traffic friction and conflict conspire to begin a decline in the rate of throughput and speed. There exist some examples of validation of this principle at intersections (e.g., traffic signalization, roundabouts, vehicle detection) that demonstrates that slowing or stopping some traffic benefits the aggregate flow, and is far better than the free-for-all converse. In the bottleneck and corridor genres, we have ramp metering and speed harmonization, respectively, providing examples on freeways.

Principle #2: Keep Sufficient Gaps

Keeping sufficient (or ideally, the largest possible) gaps leads to uniform and free(er) traffic flow. Gaps allow for small adjustments in braking, accelerating, and drifting. The larger the gap, the lesser the “ripple” affecting adjacent and following vehicles, which otherwise would react by slowing. Gap maintenance (and thus, lane reliability) is achieved on-purpose in high-occupancy vehicle (HOV) lanes or high-occupancy toll (HOT) lanes; by selective admittance in the former, and by dynamically shifting the price every few minutes in the latter. The target benefit is to allow qualifying vehicles the guarantee of a free flow trip, versus the hit-or-miss prospect in the adjacent general purpose (GP) lanes. Both cases have the added (and intended) benefit of removing vehicles and or person-trips from the GP lanes too; so all traffic streams win when these practices are employed. Absent out-and-out violators who can muck up the system, agencies can tweak the lane mandates to keep the systems running at optimum levels. How does this apply to localized bottlenecks? Theoretically, the same “gapping” principles would hold true in backups; to wit, leaving progressively larger gaps would allow for progressively better progression. (Taken to the extreme, no “bottleneck” would even exist!) The point is that in congested situations the constant brake-tapping in bumper-to-bumper traffic works to self-perpetuate the problem. No one can get much momentum before he or she has to react to the vehicle directly ahead or adjacent. The ripple effects are short, abrupt, and inefficient. The obvious problem with this is that human nature simply won’t allow for the patience and orderliness to make this work. The second that I create a sufficient gap between me and the car ahead, some “profiteering” lane jumper will fill it. Which is a nice segue into the next principle; zippering.

Photograph - This image shows an example of “zippering”.

Principle #3: Zippering

Unlike principle #2, which is noted to be fairly impractical to expect, this one could easily be melded into our regular practice; namely, to take turns, or “zipper” merge at the front of the line. The fairness – and simple visualization – of this principle speaks for itself. And there is already precedence that we have been schooled in this; witness the “Yield” condition and many recurring locations where this is the unwritten rule; newcomers quickly adapt! Advocates of zipper merging are proponents of “late” merges; i.e., staying in your lane until the last possible moment and taking turns to get through the chokepoint nozzle. One enterprising fellow in California has gone so far as to adorn his car with a zipper graphic and messages promoting this method.

Is Murphy Right? Does the Other Lane “Always Move Faster”?

How many times have you observed (or seemed to observe) that “the other lane is moving faster” only to get into that lane and then watch the first lane move past you? Actually, you are at the whim of “observation selection bias” which essentially opines that one will selectively conclude a result only on the basis of a distortion of data; in this case, your distorted sampling of only the cars that are moving, and less so the ones that aren’t. So, does cutting in line help you?

Imagine two lanes of cars. The left lane (L) is the continuous lane and the right lane (R) is dropping. You are 6th in line in R lane. If everyone stays put and “zippers” then the zipper order is L, R, L, R, etc. Your neighbor to your left is 11th and you will be 12th to merge. If, however, you “early merge” and cut in front of him into the L line, then you will now be 11th to merge, the person behind you (formerly 14th) moves up to 12th, and you neighbor drops to 13th. You win. Your neighbor loses. But the guy behind you benefits most.

Now consider the same scenario except the zipper order is R, L, R, L, etc. In the orderly scenario you would be 11th and your neighbor is 12th. If you cut in front of him, the guy behind you moves up to 11, you are now 12th, and your neighbor is now 14th. You neighbor really loses (drops two slots) and the guy behind you (formerly 13) really wins; he gains two spots – again.

Congratulations! In both scenarios you have definitely improved the slot for the guy behind you! You may or may not have improved your slot. And in either case, you made your neighbor mad! And in the end, all the jockeying you have done may have been canceled by someone ahead of you. So maybe it’s better to leave Murphy’s Law to “anything that can go wrong, will” and let zippering be the fair and simple solution to traffic backups.

Photograph - This image shows freeway traffic that performs lane-changing.

Principles Put Into Practice: Variable Speed Limits and Speed Harmonization

Variable speed limits (mostly tried in work zones; i.e., nonrecurring conditions) and the European concept of “speed harmonization” (nonwork zones) both intend to “harmonize” traffic by regulating speeds. In the latter case, a series of overhead gantries gradually adjust speeds through congested highway segments in order to flatten the sinusoidal effect of traffic speeds bouncing between open sections and interchanges. Speed harmonization is typically effected as the open highway approaches the denser central business district. A great expense is incurred by the cost of the overhead, spanned gantries, the necessary detectors, the interconnectivity, the necessary operational overhead, and the sheer number of gantries required along the multikilometer corridor. “Go slow” (harmonize) can therefore be used as a strategy as a means to move more traffic than otherwise might have gotten by. Several tests of speed harmonization are in the planning stages throughout the United States.

For example, the Minnesota DOT has deployed a variable speed limit system on I-35W in Minneapolis in conjunction with a “priced dynamic shoulder lane” (PDSL). Exhibit 3 shows a schematic of how the system operates. The features of this comprehensive system include:

  • During the off-peak hours the lanes are not tolled and open to general traffic with the exception of northbound from 42nd Street to downtown;
  • Two-plus carpools, vanpools, transit, motorcycles travel toll free;
  • Dynamically priced based on demand;
  • PDSL operates as a priced lane during peak periods to maximize capacity on existing roadways;
  • Electronic signs alert drivers whether the PDSL is open or closed; and
  • Variable speed limits are set in the adjacent nontolled lanes.

Exhibit 3. Typical Section of MN I-35W Northbound Priced Dynamic Shoulder Lane (PDSL)

Planned Overhead Signage Showing Priced Dynamic Shoulder Lane and Variable Speed Limits

The first image in Exhibit 3 is a schematic that shows the operational features of the comprehensive system of the Priced Dynamic Shoulder Lane (PDSL) on Minnesota Interstate 35 West Northbound. It includes the Planned Overhead Signage showing PDSL and Variable Speed Limits (VSL).

Source: MnDOT.

PDSL Opened

The second image in Exhibit 3 is a schematic that shows the operational features of the comprehensive system of the Priced Dynamic Shoulder Lane (PDSL) on Minnesota Interstate 35 West Northbound - a simulated snapshots of the Opened PDSL.

Source: Simulated Photos.

PDSL Closed

The third image in Exhibit 3 is a schematic that shows the operational features of the comprehensive system of the Priced Dynamic Shoulder Lane (PDSL) on Minnesota Interstate 35 West Northbound - a simulated snapshots of the Closed PDSL.

Source: Simulated Photos.

Office of Operations