The smoothing effect of carpool lanes on freeway bottlenecks

Abstract

Real data show that reserving a lane for carpools on congested freeways induces a smoothing effect that is characterized by significantly higher bottleneck discharge flows (capacities) in adjacent lanes. The effect is reproducible across days and freeway sites: it was observed, without exception, in all cases tested. Predicted by an earlier theory, the effect arises because disruptive vehicle lane changing diminishes in the presence of a carpool lane. We therefore conjecture that smoothing can also be induced by other means that would reduce lane changing. The benefits can be large. Queueing analysis shows that the smoothing effect greatly reduces the times spent by people and vehicles in queues. For example, by ignoring the smoothing effect at one of the sites we analyzed one would predict that its carpool lane increased both the people-hours and the vehicle-hours traveled by well over 300%. In reality, the carpool lane reduced both measures due to smoothing. The effect is so significant that even a severely underused carpool lane can in some instances increase a freeway bottleneck's total discharge flow. This happens for the site we analyzed when carpool demand is as low as 1200Â vph.

Full text

Page 1

The Smoothing Effect of Carpool Lanes on Freeway Bottlenecks
Michael J. Cassidy, Kitae Jang and Carlos F. Daganzo
WORKING PAPER
UCB-ITS-VWP-2008-9
December 2008

Page 2

The Smoothing Effect of Carpool Lanes on Freeway Bottlenecks

Michael J. Cassidy
Kitae Jang
Carlos F. Daganzo




Abstract
Real data show that reserving a lane for carpools on congested freeways induces a smoothing effect that is
characterized by significantly higher bottleneck discharge flows (capacities) in adjacent lanes. The effect
arises because disruptive vehicle lane changing diminishes in the presence of a carpool lane. The effect is
reproducible across days and freeway sites: it was observed, without exception, in all cases tested.
Queueing analysis shows that the effect greatly reduces the times spent by people and vehicles in
queues. By ignoring the smoothing effect at one of the sites we analyzed, for example, one would predict
that its carpool lane increased both the people-hours and the vehicle-hours traveled by well over 300%;
when in reality the carpool lane and its attendant smoothing reduced both measures. The effect is so
significant, in fact, that even a severely underused carpool lane can in some instances increase a freeway
bottleneck’s total discharge flow. This happens for the site we analyzed when carpool demand is as low
as 1200 vph. It follows that strategies designed to induce smoothing by other means also hold promise
for managing congestion, both for freeways that have carpool lanes and those that do not. Possible
strategies of this kind are discussed.

Page 3

1
1. Introduction
Carpool lanes are deployed on urban freeways for the exclusive use of vehicles that carry more than a
predetermined number of occupants. The usefulness of these lanes seems to be a subject of debate. On
one hand, they tend to be underused, and as a consequence a number of studies report that carpool lanes
unduly penalize Low Occupancy Vehicles (LOVs) by creating congestion in non-carpool lanes (e.g.
Schofer and Czepiel, 2000; Chen, et al, 2005; Kwon and Varaiya, 2008). And since an underutilized
carpool lane wastes a freeway’s queue storage space, it extends the queue length in adjacent lanes.
On the other hand, the damage done by this queue extension effect tends to be small for most
freeways (Daganzo and Cassidy, 2008). And we know, of course, that by enabling high occupancy
vehicles to bypass LOV-queues, carpool lanes can reduce the time that people collectively spend
commuting (e.g. Cassidy, et al 2006).
Moreover, there is limited evidence suggesting that these lanes, even when underutilized, can
diminish freeway congestion. Menendez and Daganzo (2007) have predicted based on simulation
experiments that carpool lanes diminish lane-changing maneuvers, and that this, in turn, smoothes (and
increases) bottleneck flows in adjacent lanes. This conjecture is consistent with earlier work showing that
disruptive lane changes cause capacity drops at bottlenecks without carpool lanes (Cassidy and
Rudjanakanoknad, 2005; Laval and Daganzo, 2006). These findings are intriguing: if the smoothing
effect turns out to be real, it would mean that carpool lanes can sometimes benefit all freeway commuters,
and not just carpoolers; and would shed light on new ways to control freeway congestion.
The present paper uses detailed video data to demonstrate the existence of the smoothing effect,
and to unveil the mechanism that causes it (sec. 2). Detector data from all suitable sites in the San
Francisco Bay Area are next used to show that the effect arises consistently, significantly and
reproducibly across days and sites (sec. 3). Queueing theory is then used to quantify its impacts (sec. 4).
Finally, the paper discusses how to exploit the effect on freeways with and without carpool lanes (sec. 5).


2. The Effect and its Causal Mechanism
Traffic data collected from videos are used below to demonstrate (i) the existence of the smoothing effect
at a freeway merge bottleneck; and (ii) the role in this played by a carpool lane that runs through the
bottleneck.

Page 4

2
2.1 The merge bottleneck and evidence of smoothing
This section examines a day in which a queue formed in the early portion of a rush, before the carpool
restriction went into force; and demonstrates that the queue discharge rate increased when the carpool
restriction did take effect. The underlying causal mechanism is unveiled in Sec. 2.2.
Our study site is shown in Fig. 1. The median lane (lane 1) of that freeway is reserved for
carpools on weekdays during the morning rush (5:00 to 9:00), and again in the afternoons (15:00 to
19:00). The remaining lanes are labeled 2 through 4.
Video cameras were erected on the over-crossings for pedestrians and for Tennyson Rd., and
these cameras recorded traffic during part of an afternoon rush (on July 19, 2006). Vehicle arrival times
at locations X1, X2 and X3 were manually extracted from the videos and, as is customary, cumulative
curves of vehicle count for all lanes combined were plotted on an oblique coordinate system (O-curves),
as shown in Fig. 2. Note that the slopes of the O-curves are the excess flows over a background flow,
which is 6800 vph in the present case; and that the curves in Fig. 2 were constructed in such ways that
superimposed curves indicate free-flow traffic (flow = demand), and separated curves indicate delays: the
wider the separation the longer the delays (see Cassidy and Windover, 1995; and Muñoz and Daganzo,
2002).
In Fig. 2, curves 2 and 3 are superimposed, and below curve 1. Thus, traffic was freely flowing
between X2 and X3, but delays arose between X1 and X2; i.e. a bottleneck formed between the latter two
locations. The curves at these two locations diverged for good at about 14:43 hrs when a disruption
temporarily reduced the total flow at X2. Less than 3 minutes later (at approximately 14:45:30) and well
before the carpool restriction was activated, flow dropped further to about 6950 vph. Thus, the carpool
lane did not contribute to the bottleneck formation and capacity drop. Instead, and as is typical of merge
bottlenecks without carpool lanes, the queue first formed in the shoulder lane and then spread to all lanes;
see Cassidy, et al (2006) for more details. The carpool lane did begin to exert influence a short time later,
however; and the influence was favorable.
Figure 3 displays an O-curve for the median (carpool) lane measured at X3. As one might expect,
flow diminished both before and after 15:00 hrs, as LOVs exited the lane. Surprisingly, a comparison of
Figs. 2 and 3 from 14:52 to 15:10 reveals that the total flow across all lanes (including the carpool lane)
remained quite steady at rates approaching 7000 vph, even as the carpool lane was being vacated.
These patterns indicate that the diminished carpool-lane flow was compensated by increased
queue discharge rates (capacity) in adjacent lanes. The effect was sustained from then on, and this is
underscored by extending some of the curves in Figs. 2 and 3 beyond 15:10.

Page 5

3


Figure 1 Study Site: I-880 North, Hayward, California
Shaded segments were subject to video surveillance



0
14:40
20
40
60
80
100
120
140
160
Oblique Vehicle Count (Lanes 1 ! 4 Combined)
14:4514:5014:5515:00 15:0515:1015:15
Time (measured at X 3)
X1
X2 & X3
7,560
vph
6,900
vph
6,980 vph
t =14:57
t =15:05
Flow Disruption
t =14:43
Capacity Drop
t =14:45:30
t =14:52 6,950
vph
6,980
vph
t =15:00
6,700
vph


Figure 2 O-Curves for Lanes 1 – 4 at X1 through X3 (July 19, 2006)

Page 6

4
0
14:40
20
40
60
80
100
120
14:4514:50 14:55
Time
15:00 15:0515:10 15:15
2,150
vph
1,600
vph
1,370
vph
t =14:57
t =15:05
Capacity Drop
t =14:45:30
t =14:52
1,920
vph
1,790
vph
t =15:00
1,640
vph
Oblique Vehicle Count


Figure 3 O-Curve of Median (Carpool) Lane at X3 (July 19, 2006)


2.2 The cause of smoothing
We show here that the increase in discharge flows is due to a decline in lane changing rates caused by the
carpool restriction. Lane-changing rates were extracted from videos over the 0.4-km segment between
the over-crossings in Fig. 1. We consider first the connection between discharge rates and lane-changing
rates, and then show that the decline in the latter can be attributed to the carpool lane.
To understand the connection between lane changing and discharge flow, we examined the lanes,
one at a time, starting with lane 2. The boldfaced O-curve in Fig. 4a shows that lane-changing maneuvers
in and out of that lane began to diminish minutes before 15:00 hrs, the carpool lane activation time (as
highlighted by the downward-bending dashed line in the figure). An abrupt increase in the lane’s
discharge flow followed close on the heels of this event, as revealed by the thin O-curve of vehicle count
(and highlighted by the upward-bending dashed line). Figure 4b reveals that a similar pattern was
observed a few minutes earlier in lane 3: lane changing diminished and discharge flow rose very soon
thereafter, beginning sometime around 14:52 hrs. The phenomenon was not observed in lane 4, however;
see Fig. 4c.
Thus, we see that in each of the two lanes closest to the carpool lane, a reduction in lane-changing
rate was closely followed by an increase in discharge flow. The timing of these events so close to 15:00
hrs strongly suggests that they were caused by the carpool restriction.

Page 7

5
0
14:40
20
40
60
80
100
(All Maneuvers in & out of Lane 2)
120
14:4514:50 14:55
Time
15:00 15:0515:10
0
20
40
60
80
100
Oblique Lane Change Count
Oblique Vehicle Count in Lane 2
t =15:00
1,790 vph
2,050 vph
1,280/hr/km
990/hr/km
Lane Changes
Dicharge Flows
Capacity Drop
t =14:45:30
Reduced lane -
changing rate
Increased
discharge flow
(a)


0
14:40
30
60
90
120
150
180
210
(All Maneuvers in & out of Lane 3)
240
14:4514:5014:55
Time
15:00 15:05 15:10
(b)
0
20
40
60
80
100
120
140
160
180
200
Oblique Lane Change Count
Oblique Vehicle Count in Lane 3
t =14:52
1,330 vph
1,600 vph
2400/hr/km
1850/hr/km
Lane Changes
Dicharge Flows
Reduced lane -
changing rate
Increased
discharge flow
Capacity Drop
t =14:45:30


Figure 4 Oblique Cumulative Curves of Lane-Changing and Discharge Flow (July 19, 2006);
(a) Lane 2; (b) Lane 3

Page 8

6
0
20
40
60
80
100
120
140
(All Maneuvers in & out of Lane 4)
160
14:4014:4514:5014:55
Time
15:0015:05 15:10
(c)
0
20
40
60
80
100
120
140
160
1,810 vph
Oblique Vehicle Count in Lane 4
Oblique Lane Change Count
1,370/hr/km
1,270/hr/km
Lane Changes
Dicharge Flows
Capacity Drop
t =14:45:30


Figure 4 (cont’d) Oblique Cumulative Curves of Lane-Changing and Discharge Flow (July 19, 2006);
(c) Lane 4

Appendix A takes a more detailed look at the data and shows that the observed patterns
(including the 8-min discrepancy between the pattern changes in lanes 2 and 3) were indeed caused by the
migration of vehicles away from the carpool lane, further solidifying the idea that the carpool restriction
is at the root of the reductions in lane changing and improvement in discharge flow.
The next section shows that the smoothing effect arises consistently and reproducibly at different
sites.


3. Repeated Observations
We examined the entire network of carpool facilities in California’s San Francisco Bay Area during
multi-week study periods, and identified all the sites in which a bottleneck was active for at least 30 mins
before and after its carpool restriction switched on or off. This filtering method is logical, since we are
comparing the bottleneck’s center-lane discharge flows with and without the carpool lane, while holding
all else approximately constant. Although we found only two suitable sites (the site in Fig. 1 and one

Page 9

7
additional site), multiple instances passed our filter at each site. The smoothing effect arose in every
instance. We show this for our first site in sec. 3.1; and for the second site in sec. 3.2.

3.1 Reproducibility across days: Site 1
This site was examined every weekday in Aug. and Sept. 2007, and eight suitable instances turned up.
These eight periods are in addition to the one used in sec 2; four came from late portions of the morning
rush, and four from early portions of the afternoon rush. No other instances were found in which the
bottleneck’s active period overlapped both the carpool lane’s active and inactive periods.
Table 1 summarizes the data. For each of the eight periods, it presents 30-min average
bottleneck discharge flows in the two center lanes combined, with and without the carpool restriction.
(Discharge flows from 5-min transition periods on each side of the carpool lane activation and
deactivation instants are excluded.) The table shows that the smoothing effect arose without exception,
and did so significantly and consistently. The resulting rise in center-lane discharge flows ranged from
9.5% to 13%, with an average of 10.5%, in the early afternoons; and from 18% to 21%, with an average
of 19.5%, in the late mornings. The late morning and early afternoon differences are statistically
significant, so something must be causing them. As we shall see momentarily, a similar discrepancy
arises at the second site.

Table 1 Discharge Flows from Lanes 2 and 3 Combined at I-880 North Study Site

(1) (2) (3) (4) (5)
Observation
Date
Flows with
carpool lane
(vph)
Flows without
carpool lane
(vph)
Increase due to
smoothing
(vph)
%
increase
Early
Afternoons
2007-08-02
15:05 ~ 15:35 14:25 ~ 14:55
3670 3350 320 9.5
2007-08-03
2007-08-23
3690
3740
3370
3400
320
340
9.5
10
2007-09-11
Late
Mornings
2007-08-21
3650 3240 410 13
8:25 ~ 8:55 9:05 ~ 9:35
3290 2750 540 20
2007-08-22 3530 2990 540 18
2007-08-23 3370 2830 540 19
2007-09-04 3430 2840 590 21

Page 10

8
3.2 An additional site
The second site is shown in Fig. 5. A bottleneck forms at the entrance to the curved section during the
afternoon rush. The site was canvassed for suitable study instances from May through September 2007.
Four instances were found: all during the afternoon rush. Two straddled the carpool lane’s activation
time (at 15:00 hrs) and two its deactivation time (at 19:00 hrs).
Again, the smoothing effect emerged without exception; see Table 2 which presents the discharge
flows measured in the two center lanes for each of the four cases. The effect is again significant and
consistent: discharge flows increased by 8% and 12% in the beginning of the afternoon rush; and by 18%
and 19% at the end of the rush; and the discrepancies between early and late measurements are
statistically significant.
Since for both sites, we see that the smoothing effect is less significant at the start of the
afternoon rush, we look more deeply at the data in Tables 1 and 2 and see that all the discharge flows are
significantly higher at this time of day than at the end of a rush. The pattern indicates that early-afternoon
drivers are more aggressive, perhaps because they are trying to “beat the rush” for the remainder of their
trips and are less affected by lane changes. We therefore conjecture that, during the beginning of a
morning rush, center-lane discharge flows would increase by 10% due to smoothing, as they do near the
start of the afternoon rush; and that 15% might be a good average to use for planning purposes.


Bottleneck
Location
Amador St .San Pablo Ave .
: Detector
Lane 2
Lane 3
Center Lanes
Median (Carpool ) Lane 1
Shoulder Lane 4


Figure 5 Second Study Site: I-80 East, Richmond, California

Page 11

9
Table 2 Discharge Flows from Lanes 2 and 3 Combined at I-80 East Study Site

(2) (3) (1) (4) (5)
Observation
Date
Flows with
carpool lane
(vph)
Flows without
carpool lane
(vph)
Increase due to
smoothing
(vph)
%
increase
Early
Afternoons
2007-05-30
15:05 ~ 15:35 14:25 ~ 14:55
3880 3590 290 8
2007-07-23
Late
Afternoons
2007-07-10
3870 3450 420 12
18:25 ~ 18:55 19:05 ~ 19:35
3660 3100 560 18
2007-09-11

3350 2820 530 19

4. The Real Impacts of Carpool Lanes on People and Vehicle Delay
This section explores the real impacts of carpool lanes; i.e. by recognizing smoothing. We compare the
PHT and VHT for an afternoon at the site in Fig. 1 under three scenarios: (i) no carpool restriction; (ii) a
carpool lane with realistic consideration of smoothing; and (iii) the hypothetical (and unrealistic) case of a
carpool lane that does not induce smoothing. Predictions were made with the queueing/kinematic wave
model of Newell (1993). Details are provided in Appendix B. Inputs to the analysis were estimated from
the site’s data: discharge flows were set equal to the average rates over multiple afternoons; and input
flows were set equal to those measured at the upstream detector station during an afternoon when the
queue did not grow beyond these detectors. This allowed us to measure upstream demand precisely, but
corresponds to a day with lower than usual congestion. Thus, our results underestimate the differences
that arise between our three scenarios on more typical days. Results are shown in Table 3.
Note from columns 2 through 4 which compare system performance with and without the carpool
lane, the carpool lane reduces PHT by 30% compared to the case of no carpool lane. This is reassuring,
since PHT-reduction is a commonly-cited reason for deploying carpool lanes in the first place (Turnbull
and Capelle, 1998; Bracewell, et al. 1999; Henderson, 2003). But more remarkably, and thanks to the
smoothing effect, the carpool lane reduces VHT by 15%.
Let us now see what a conventional analysis (wrongly ignoring the smoothing effect) as in
Dahlgren (1998, 2002) and Kirshner (2001) would have predicted. The result is shown in column 5. By
ignoring the smoothing effect, one would incorrectly attribute very large delays to the carpool lane. One

Page 12

10
would be predicting increases well in excess of 300% both for PHT and VHT when instead the carpool
lane would reduce both. This example clearly shows that one cannot assess the real impacts of a carpool
lane without accounting for smoothing. A question of interest then is: what fraction of traffic must be
carpools to justify a carpool lane?

Table 3 Predicted PHT’s and VHT’s

(3) (1) (2) (4) (5) (6)

No carpool
restriction
Carpool lane with
(real)
Smoothing effect
%
difference
Carpool lane without
smoothing effect
(hypothetical)
% difference
with (2)
PHT
(person-hrs)
VHT
(veh-hrs)
2450 1900
– 30 10340 + 322
1950 1700
– 15 9080 + 365

Queueing analysis also shows that the carpool lane is beneficial even when demands for that lane
are quite low. The boldfaced curves in Figs. 6(a) and (b) show the PHT and VHT obtained at a 4-lane site
like ours, with and without a carpool lane, as a function of the percentage of freeway demand that is
comprised of carpools, α. Note from Fig. 6(a) that the carpool lane reduces PHT when α is as low as
17%; and from Fig. 6(b) that only a slightly higher α (17.3%) is required to reduce the VHT. In the
present case, α ≈ 17% corresponds to carpool-flows that are less than 1200 vph. So we see that, with
smoothing, even a very underused carpool lane can reduce this freeway’s PHT and VHT, with its
attendant externalities, and therefore be a win-win proposition for society.

0
500
1000
1500
2000
VHT
2500
3000
3500
1617 181920 21
0
500
1000
1500
2000
2500
3000
3500
16 171819 2021
_ (%)
(b)
_ (%)
PHT
(a)
Without Carpool Lane
With Carpool Lane
Without Carpool Lane
With Carpool Lane
_ = 17.5 %
Table 3 Benchmark
_ = 17.5 %
Table 3 Benchmark


Figure 6 Predictions with and without Carpool Lane as functions of α, (a) PHT’s; (b) VHT’s

Page 13

11
5. Conclusions and Recommendations
This paper has shown that carpool lanes passing through bottlenecks significantly increase the discharge
flows in lanes adjacent to the carpool lanes. The effect was consistently reproduced across days and sites.
The effect is so pronounced that even an underutilized carpool lane can increase a bottleneck’s total
discharge rate. Queueing analysis illustrates that carpool lanes with flow as low as 1200 vph can reduce
not only people delay, but even vehicle delay, and that the influence of smoothing on all this is very large.
Thus, one cannot realistically assess the impacts of a carpool lane without accounting for smoothing.
Given that smoothing was not observed furthest from the carpool lane in lane 4, the effect when induced
by a carpool lane may be especially strong on narrow freeways with few lanes.
The findings suggest that freeway congestion could also be reduced by inducing the smoothing
effect through other means. For example, roadside signing and (solid) painted lane striping might be used
near certain bottlenecks to limit the disruptive impacts of lane changing. Disruptive lane changing might
also be reduced in some cases by sorting drivers (and vehicle classes) across lanes according to their
preferred travel speeds; or in other cases by inducing a more even distribution of flows across lanes; and
these outcomes might be achieved by imposing lane-specific speed limits, based perhaps on real-time
measurements of traffic. The above measures could be deployed on any freeway, whether or not it
includes a carpool lane. For freeways with severely underused carpool lanes, one might even try to
induce smoothing by rescinding carpool restrictions near bottlenecks at certain times only, e.g., as
described in Daganzo, et al (2002). Though this latter dynamic strategy may be unconventional,
simulations in Menendez and Daganzo (2006) indicate that it can significantly increase bottleneck
capacity. Field experiments to test some of these ideas are now being planned.


References
Bracewell, D., Sayed, T. and Shalaby, A. (1999) High-Occupancy Vehicle monitoring and evaluation framework,
Transpn Res. Rec. 1682, 38-45
Caltrans (2004) Report on Bay Area HOV lanes. California Dept. of Transportation, District 4, Oakland, Office of
Highway Operations.
Cassidy, M.J., Daganzo, C.F., Jang, K. and Chung, K. (2006) Empirical reassessment of traffic operations: Freeway
bottlenecks and the case for HOV lanes. UCB-ITS-RR2006-6, Institute of Transportation Studies, University of
California, Berkeley, CA.
Cassidy, M.J. and Rudjanakanoknad, J. (2005) Increasing the capacity of a merge by metering its on-ramp. Transpn
Res Part B 39, 896-913.

Page 14

12
Cassidy, M.J. and Windover, J.R. (1995) Methodology for assessing dynamics of freeway traffic flow. Transpn Res.
Rec. 1484, 73-79.
Chen, C., Varaiya, P. and Kwon, J. (2005) An empirical assessment of traffic operations. In: Mahmassani, H.S.
editor, Transpn and Traffic Theory, Procs of 16th International Symp on Transpn and Traffic Theory.
Daganzo, C.F. and Cassidy, M.J. (2008) Effects of high occupancy vehicle lanes on freeway congestion, Transpn
Res Part B 42. 861-872
Daganzo, C.F., Laval, J.A. and Munoz, J.C. (2002) Ten strategies for freeway congestion mitigation with advanced
technologies, Traffic Engineering and Control 43, 397-403.
Dahlgren, J. (1998) High occupancy vehicle lanes: Not always more effective than general purpose lanes, Transpn
Res Part A 32, 99-114
Dahlgren, J. (2002) High-occupancy/toll lanes: where should they be implemented? Transpn Res Part A 36, 239-255
Henderson, D. (2003) State of the practice in High-Occupancy Vehicle system performance monitoring, Transpn
Res. Rec. 1856, 152-160
Kirshner, D. (2001) Efficiency gains from value pricing: case study of Sunol Grade, Transpn Res. Rec. 1747, 21-28
Kwon, J. and Varaiya, P. (2008) Effectiveness of California’s High Occupancy Vehicle (HOV) system, Transpn Res
Part C 16, 98-115.
Laval, J.A. and Daganzo, C.F., (2006) Lane-changing in traffic streams. Transpn Res Part B 40, 251-264.
Lawson, T.W., Lovell, D.J. and Daganzo, C.F. (1997) Using inpit-output diagrams to determine the spatial and
temporal extents of a queue upstream of a bottleneck. Transpn Res. Rec. 1572, 140-147
Menendez, M. and Daganzo, C.F. (2007) Effects of HOV lanes on freeway bottleneck. Transpn Res Part B 41, 809-
822.
Muñoz, J.C. and Daganzo, C.F. (2002) Fingerprinting traffic from static freeway sensors. Cooperative Transpn
Dynamics 1, 1-11.
Newell, G.F. (1993) A simplified theory of kinematic waves in highway traffic, Part II: Queueing at freeway
bottlenecks, Traspn Res Part B 27, 289-303.
Schofer, J.L. and Czepiel, E.J. (2000) Success factors and decision issues for High-Occupancy Vehicle facilities,
Transpn Res. Rec. 1711, 13-22
Turnbull, K.F. and Capelle, D.G. (1998) Development of an HOV systems manual, National Cooperative Highway
Research Program, Transportation Research Board, National Research Council

Page 15

13
Appendix A: The Connection between the Carpool Restriction and Lane Changing
Patterns
The evidence presented in this appendix indicates that favorable lane-changing patterns (i.e. patterns that
ultimately induced the higher discharge flows in lanes 2 and 3) were triggered by the carpool restriction.
This is explained with Fig A1. It uses arrows to illustrate the time-varying lane-changing patterns
measured over the 0.4-km stretch upstream of the I-880 bottleneck (the darker shaded area in Fig. 1).
Thin, solid arrows denote initial lane-changing rates; thick arrows increased rates; and dashed arrows
diminished rates. Note from the line-weights of the arrows how the lane-changing rates diminished when
comparing the period before 14:52 (before the carpool restriction came into play) to the period after 15:05
(when the restriction was in effect), as was mentioned in the text. Let us now examine the sequence of
events in more detail.

14:52
(Crowding)
14:5715:0015:05
Lane 2
Lane 3
Lane 1(Carpool Lane )
Lane 4 (Shoulder Lane )

Figure A1 Lane-Changing Patterns Observed over Time

Consider first the vehicle maneuvers made out of lane 1 (the carpool lane) and into lane 2. This
migration rate increased from 14:52 to 15:00 hrs, as depicted with the two boldfaced arrows that project
from lane 1 to 2. We attribute this temporary increase to the impending carpool restriction since LOVs
are required to vacate lane 1 by 15:00 hrs: early responses are to be expected since LOV-drivers risk fines
for carpool violations.
The impending carpool restriction also discouraged drivers from maneuvering into lane 1 from
lane 2. These movements started to decline at 14:57, and continued to diminish after the carpool lane
activated, as depicted with the dashed arrows from lane 2 to 1.
The imbalance in lane-changing rates between lanes 1 and 2 created crowded conditions in lane 2
promptly after 14:52, and this had two effects. First, for a time the crowding pushed vehicles from lane 2
into lane 3. As shown by the thicker arrows, this push subsided at 15:00 hrs, when the heightened
migration from lane 1 subsided as well. Second, the crowding discouraged maneuvers made into lane 2
from lane 3, as depicted with the dashed arrows from lane 3 to 2. Measurements quantifying the patterns
of Fig. A1 are furnished in Figs. A2 – A4.