Bus station2.

qxd 27/4/06 17:28 Page 182

182 Calculating bus station capacity

Design of bus station: a

case study in Brighton
Bhargav Adhvaryu, University of Cambridge, UK

This paper1 investigates the design of bus data collection is primarily done manually and the swept paths of buses, taxis, and lorries is
stations in a physically constrained area through is supplemented by videography. The analysis used to formulate the proposed physical layouts
a case study of Brighton Bus Station, which is of capacity of the existing bus station is for the bus station. The proposals were
situated in the historic core of the city abutting completed using microsimulation software. In evaluated using a simple evaluation
the railway station complex. Bus and passenger addition, a graphic simulation tool for tracking methodology developed in this study.

here are two key concerns in developing the proposals. Time periods associated with a bus stop

T The first is to generate workable solutions that are com-

patible with the existing physical constraints of the
area and the operational constraints of buses and taxis. The
A bus arrives at a bus stop and stops. The doors open and
passenger exchange takes place (ie, alighting and boarding of
passengers), which may be serial or parallel. Doors then close
second is to make the bus station more accessible to wheel- and the bus departs. The time elapsed during this period is
chair users. The study emphasises that the process of bus sta- known as dwell time (tdwell). Tyler (2002) has distinguished
tion design could be enhanced by use of simple analytical six time-periods associated with a bus stop, which are ex-
tools and hence need not be purely driven by architectural plained briefly as follows:
and urban design considerations.
1 Queuing time (tq) is the time spent by a bus in a queue
prior to entering the bus stop.
INTRODUCTION 2 Clearance time (tclear) is the minimum [theoretical] time
between the departure of one bus and the arrival of a sub-
This paper is based on a study undertaken on design of a bus sequent bus in the bus berth.
station in Brighton - a town about 100 km south of London. 3 Dead time (tdead) is the time that elapses in opening and
The author: The study covered the review of models for bus stop capacity closing of doors plus the time associated with extending
Bhargav Adhvaryu estimation, issues realting to adapting such models for esti- ramps/lifts/other equipment to facilitate alighting and
is Research mation of bus station capacity, use of a microsimulation boarding of disabled persons, if any.
Associate at the model for estimating capacity of existing situation and the 4 Passenger service time (tpst) is the time elapsed while pas-
Martin Centre for proposals, use of a graphic simulation tool for formulating senger exchange takes place.
Architectural and proposed physical layouts for the bus station, developing a 5 Internal delay (tint) is the time spent by a bus waiting to
Urban Studies, methodology to evaluate bus station designs, and taxi rank leave the bus berth after it is ready to leave, but cannot
Department of capacity estimation and its physical layout to the extent that actually leave as it is obstructed by other bus(es) in the
Architecture, it enhances the performance of the bus station. However, in bus stop area.
University of this paper the discussions on development of a method for 6 External delay (text) is the time spent by a bus waiting to
Cambridge since evaluation of bus station designs and capacity estimation for leave the bus berth after it is ready to leave, but cannot
Dec 2004. taxi ranks and its alternative physical layouts are omitted, be- actually leave as it is obstructed by other traffic outside
cause it would require a separate dedicated space. Also, only the bus stop area.
Before joining the
two layouts of bus stations out of the proposed four are in-
Martin Centre, he
cluded in this paper. The discussion on models for estimat- If we reflect back on the definition of dwell time discussed
was in India
ing bus stop capacity have able been considerably shortened above, it would constitute a sum of time periods 3 to 6
where he spent
in this paper. (above). Time periods 5 and 6 often occur concurrently, in
four years as a which case the higher of the two would be used in estimation
lecturer and four of dwell time. Clearance time has to be added to dwell time
years in an urban ESTIMATION OF BUS STOP CAPACITY: A to take into account the time taken to enter and leave the
planning BRIEF REVIEW berth. Thus, the simplest possible theoretical model for esti-
consultancy firm mating bus stop capacity may be given as:
as a project Estimation of bus stop capacity is now a well-researched
manager. He can field. There are various mathematical models available for
be contacted by estimating bus stop capacity. The theoretical basis for esti-
email at ba247 mating bus stop capacity is briefly explained below. How- The degree of saturation, which is a key performance mea-
@cam.ac.uk ever, firstly the time periods associated with a bus stop need sure of a bus stop defined as the ratio of flow to the theoreti-
to be highlighted. cal capacity can be found as:

tec MAY 2006

Bus station2.qxd 27/4/06 17:29 Page 183

183

Saturation = (Q/Qc) x 100 (%), where Q is the observed bus It appears that tackling the issue of not using gross values
flow. could be dealt with by use of microsimulation models. In
that, observed individual values could be used instead of
Some models for estimating bus stop capacity grossed or averaged ones, allowing the user to adapt the
The Convoy Model model to the context it is being used for. IRENE (Gibson et
The Convoy Model which is based on the São Paulo (Brazil) al, 1989 as reviewed in Fernández, 2001 and PASSION [PAral-
experience is given as: lel Stop SimulatION] (Fernández, 2001)) are both microsimu-
lation models. The assumption in IRENE that buses arrive at
random and that almost any bus that arrives satisfies the
waiting passenger seems reasonable when the bus system is
deregulated or informal with many services sharing the same
where, routes and high bus flows and passenger demands. However,
there is no empirical evidence that this is the actual case, and
many not be true if services, frequencies, and demand are
lower (Fernández, 2001). In addition, the assumption that
the mean arrival rate of passengers remains constant during
the simulation period might not be applicable when passen-
ger arrivals are dependent on other services such as a railway.
The Highway Capacity Manual (HCM) 2000 Model In a nutshell, IRENE applies well in situations with high and
The HCM 2000 Model (TRB, 2000) introduces a new concept homogeneous bus and passenger arrivals, but does not repre-
of using failure rate. Failure rate, which is derived from basic sent passenger arrivals in detail.
statistics, is the probability that a queue of buses will not be To overcome this particular deficiency of IRENE, a different
formed behind a bus stop. Za represents the area under one approach in modelling is required. Elaluf (1994) made some
tail of the normal curve beyond the acceptable levels of prob- advances on these lines. Her study recognised that the
ability that a queue will form at a bus stop. Typical values of processes at a bus stop happen in parallel and hence it would
Za are obtained from standard tables. Suggested values of Za be reasonable to use a parallel computing technique to simu-
for CBD stops range from 1.44 to 1.04 resulting in probabili- late bus stop operations. Fernández (2001) furthered this ap-
ties of 7.5% to 15.0%, respectively that queues will develop. proach by developing a microsimulation program PASSION.
For outlying stops, 1.96 is suggested as the value of Za result- PASSION addresses the deficiencies of IRENE, by having a
ing in queues beyond bus stops only 2.5% of the times. dedicated input section for passenger arrival data. Owing to
However, in general, Za values of 1.44 (representing 7.5% these merits PASSION was used for estimating bus station ca-
probability) are acceptable. pacities in this study.
The theoretical capacity of the berth is then given as:
ESTIMATION OF BUS STATION CAPACITY

The brief literature review of capacity estimation models in-

dicates that these models are primarily used for bus stops.
No model specifically refers to estimating capacity of a bus
station. Prima facie, one can argue that bus station capacity
could be estimated by summing up the individual capacities
of the bus stops/berths within the bus station. HCM 2000
(TRB, 2000) mentions the use of discounted values as a multi-
plier instead of the number of berths, ie instead of multiply-
ing the capacity of a bus stop by three for a three-berth bus
station, it suggests using 2.45 and 2.60 respectively for in-line
and off-line bus stops, to get the effective capacity. In other
The value of Neb has to be discounted for more than one words, this multiplier represents the effective number of
berth depending on whether the bus stop is in-line or off- berths. The author feels that there are issues that are differ-
line. It is suggested that values of off-line would be lesser ent in bus stops and bus stations (briefly discussed in the
than in-line. next section ) and hence the worthiness of this ‘natural ex-
tension’ is an area warranting further research. However, in
Microsimulation Models this study PASSION was used for each of the three bus stops
Fernández (2001) tested the Convoy Model for a one-berth and bus station capacity was derived by summation of the ca-
bus stop at Manor House Station, London. The results were pacities of the three bus stops in case of independent bus
2.7 times the observed data, and hence the model did not de- stops and using discounting factors in case of interdependent
liver good results in this case. It was pointed out that the bus stops (see Table 1).
constants assumed in the formula, which are essentially ap-
plicable to high capacity busways, were not applicable to a
bus stop in the UK urban environment, and hence the over-
IS ESTIMATING BUS STOP CAPACITY ANY
estimation in capacity. DIFFERENT FROM ESTIMATING BUS
The HCM 2000 model is simple and straightforward and STATION CAPACITY?
has a good conceptual underpinning. However, its inherent
nature of using gross values makes it too simplistic for appli- Estimation of bus stop capacity uses dwell time and clearance
cation to real-life situations. In addition, since the gross val- time. However, in case of bus stations, two aspects emerge as
ues used in the model are derived mainly from US-based em- being different: [1] the additional time of manoeuvring the
pirical studies, its application to the UK may not render ap- bus from bus station entry to bus berth and from bus berth to
propriate results. bus station exit, which could be termed at the internal run-

tec MAY 2006

Bus station2.qxd 27/4/06 17:30 Page 184

184 Calculating bus station capacity

Figure 1: Existing situation

Existing layout Brighton station has three bus stops (see Figure 1). However,
and bus stops
N in practice, stop A and B act as one bus stop (stop A is merely

B
a flag-poled stop, without shelter) and buses for stop A often
stop at B if B is empty. In the 90-minute survey period only
three buses were observed at stop A. Hence, for the existing
situation stops A and B were combined. The key outputs of
the microsimulation runs, with a traffic signal at exit, are

C A
shown in Table 1. Table 2 shows the outputs with free-exit
for comparison. Please note that microsimulation runs with
obstructed exit were also performed but are not presented in
this paper.
From the microsimulation outputs it is revealed that the
saturation for the existing situation is 33% (16% to 51%),
which is much below the generally acceptable maximum de-
sign value of 60% (Tyler, 2002), although queuing was ob-
served on some occasions. Thus, adding new berths to the
bus station was not necessary and hence the proposed de-
signs consider only three bus berths.

Proposed situation
The congestion and queuing observed at Brighton Bus Sta-
tion can be attributed to the following factors:
Factor 1 Internal delay
(due to interdependency of berths)
ning time2 and [2] a component of the internal delay attrib- (ie, a bus from a stop cannot depart if a bus is parked at
uted to jaywalking passengers (if any), which could be preceding stop and a wrongly parked bus at the following
termed as additional internal delay. The latter might be par- stop will delay a bus wishing to enter the preceding stop)
ticularly high in case of an island type layout but could also Factor 2 Internal delay (due to pedestrian jaywalking)
exist in perimeter or concourse layouts if pedestrians are Factor 3 External delay (due to hindrance from taxis
undisciplined or the design of the station is such that it en- queuing at bus station exit)
courages jaywalking. Contrastingly, these two components Factor 4 External delay (due to taxis and cars using bus
in case of bus stops would be nonexistent or minimal (if station area)
any). This discussion suggests two things: either bus stop ca- Factor 5 External delay (due to traffic signal at exit)
pacity estimation models need appropriate modifications in Factor 6 High passenger service time (probably attrib
order to consider these two components or a separate simula- uted to cultural factors, ie Brighton being slow
tion model for estimating bus station capacity should be de- paced as against London which is a fast-paced city!)
veloped. Factor 7 Irregular bus arrival pattern
Table 1:
Summary of It can be appreciated that factors 1 to 4 can be manipu-
simulation runs CAPACITY OF BRIGHTON BUS STATION lated easily, ie these are generally within the scope of a bus
with signal-at-exit station design exercise. How these have been achieved
The bus station capacities for existing and proposed situa- through the proposed design is discussed later in the paper.
Table 2: tions were estimated using PASSION. In that, separate input Factors 5 to 7 could be changed, but would generally be part
Summary of files were created for each of the bus stops for both existing of a larger transport planning exercise, which would com-
simulation runs and proposed situations. A primary survey was conducted to prise rerouting of buses, area-wide traffic management, and
with free-exit3 record bus arrivals and passenger alighting and boarding (for signal redesign.
each route) for a period of 90 minutes. To address Factor 1 (above), which is of prime importance,

Table 1: Table 2:

tec MAY 2006

Bus station2.qxd 27/4/06 17:30 Page 185

185

the three berths are construed as being independent of each

other in the proposals. The arithmetic sum of individual N N
berths is the capacity of the bus station. The three berths
cater to the same twelve routes that come to the station at "Off-line" "In-line"
present. However, some minor reallocation of routes was Bus Station Area Bus Station Area
done based on their frequency and destinations.
Separate input files were created for simulation of the pro-
posed situation using exactly the same bus arrivals, passenger
arrivals, and route opted by each passenger as used in the
input files for simulation of existing situation (as summarised
in Tables 1 and 2).
As can be seen from Table 1, the capacity of the bus station
for the proposed situation turns out to be 127 buses/h as
against the existing capacity of 82 buses/h. The increase in ca-
pacity can be attributed to making the berths independent (ie,
the presence of a bus at any of the berths does not affect func-
tioning of the other two berths). In addition, the reallocation
of bus routes also helped in bringing the saturation level down
to 23% (14% to 43%) as against the existing value of 33% (16%
to 51%). Based on the analysis using PASSION, it could be said wheelchair users). However, owing to limitations on time Figure 2:
that because of the independence of berths, the overall delay, and resources, this process was carried out only once in the
The concept of
which was occurring in the existing situation, is eliminated in study. Thus further refinements to the proposed layouts are
off-line and in-
the proposed situation. The possible physical layouts for the certainly possible. As a special note it is worth mentioning
line bus stations
proposed situation are discussed in the next section. that currently the MISS software is programmed to simulate
in context of
the swept paths of the vehicle regardless of the external con-
ditions such as the type of kerb. This means that if a Kassel
Brighton
THE DESIGN APPROACH kerb4 is put in place then the steering mechanism and the
manoeuvrability of the bus changes, which currently is not
Conceptual bus station layouts reflected in MISS. Should such modifications be made to this
Conceptually two different layouts could exist with reference software then it could still become a more powerful tool for
to the location of the bus station in the context of the access designing bus station layouts.
road. In the off-line layout, the road(s) that leads to the bus Drawings depicting the physical layout of the bus station
station abounds the station area. In other words, no part of and how it has been achieved by use of swept-path simula-
the road space is used as bus berths. On the other hand, in tion can be referred to in drawings labelled BO1.1, BO1.2,
the in-line layout, part of the road leading to the bus station and BI3.1. Drawing showing swept-paths for achieving the
itself is used as bus berths. Off-line and in-line layouts in layout for BI3.1 are not included in this paper. In the evalua-
context of Brighton Bus Station can be explained by help of tion methodology developed in this study for comparing the
the schematic diagram in Figure 2. proposed design (not discussed in this paper), the proposal
Generally, most major bus stations would be off-line. BO1 ranked the highest. There was not room to display all of
However, in-line bus stations might have to be considered if the bus station layouts and the proposed taxi rank alterna-
the site is in a physically constricted area-a case common tives in this paper.
with inner-city bus stations. Brighton Bus Station is indeed
situated in the historic core of the city and hence it was nec-
essary to explore the possibility of an in-line type layout.
HOW THE PROPOSED LAYOUTS WOULD
REDUCE CONGESTION AND QUEUING?
Vehicle swept-path simulation
After having decided the conceptual layout, a software used This section discusses how the above-discussed factors have
by London Buses, Transport for London (TfL) called MISS been addressed in the proposed layouts.
(Mobil Interactive Swept-path Simulator) was employed. Ve- Factor 1 Internal delay (due to interdependency of
hicle parameters such as the outer dimensions, distance be- berths)
tween front and rear wheel axles, and steering sensitivity are Dealt with by making the three berths independent. How-
inputs to the MISS. On the site area in the drawing, a path is ever, in case of in-line type layouts bus drivers’ error could
traced along which one expects the bus to be driven. Using make it interdependent.
the ‘calculate path’ command, the swept path of the vehicle Factor 2 Internal delay (due to pedestrian jaywalking)
as it moves is simulated (see Drawing BO1.2 ). Dealt with by making more area available to pedestrians
The first step is to mark out tentative areas for the bus plat- and by having designated pedestrian crossing points.
form islands or the areas where bus should not run (based on Factor 3 External delay (due to hindrance from taxis
the design concept discussed above) before running the ‘elec- queuing at bus station exit)
tronic bus’ on the bus station site in the drawing. Then vehi- Dealt with by having a higher ‘turnover’ rate for taxis
cle swept paths are obtained. Bus platforms and channelisers thereby reducing incidence of queuing and by having more
are then formed such that they are clear off the bus swept space for taxi set-down.
paths. The bus platforms so formed are checked against the Factor 4 External delay (due to taxis and cars using bus
minimum standards of space required for wheelchair ma- station area)
noeuvre and for providing a shelter (if necessary). This Dealt with by having a higher ‘turnover’ rate for taxis
process ideally has to be repeated until the optimum layout is thereby reducing incidence of queuing and by having a
obtained (ie, the one that minimises space requirements and physical layout that deters taxis and private cars from en-
more importantly satisfies the horizontal gap between the tering the bus station area and stopping there (which is oc-
bus and the platform of accessibility considerations of the curring frequently in the existing situation).

tec MAY 2006

Bus station2.qxd 27/4/06 17:31 Page 186

186 Calculating bus station capacity

I would like to
acknowledge the
SCOPE FOR FURTHER RESEARCH port departments could adapt the framework developed in this
study to get a reasonably accurate assessment of the existing
academic inputs The topic of bus station design has been little dealt with in situation of their bus stations. Such an assessment could then
from Professor the literature. Most literature on the subject relating to ca- be used to formulate schematic alternatives for the purpose of
Nick Tyler, UCL; pacity estimation is on bus stops and does not explicitly refer budgeting and terms of reference for detailed design.
practical inputs, to a bus station. Thus, development of a microsimulation
base map and software, for bus stations or the likely adaptation of a soft- Footnotes
supporting data ware like PASSION specifically for bus stations, needs to be 1
Is based on an MSc Transport dissertation at Imperial College
from Len further researched. In addition, modifications to the swept- London & UCL, under the supervision of Professor Nick Tyler, UCL.
Holloway, path simulation software that are sensitive to external condi- 2
Coined by the author.
Brighton & Hove tions could make it a more powerful tool for enhancing the 3
The purpose of comparing signal-at-exit and free-exit conditions is
City Council; help physical layouts of bus stations. Lastly, further research on to understand the impact of free-exit conditions on capacity and
with MISS from development of a framework or guidelines for evaluation of saturation. From the simulation results, it can be seen that the
Hugh Bailey and bus station designs would certainly help the cause of bus sta- individual berth capacities and hence the bus station capacity and
Martin Edgar, tion design. the saturation level are markedly improved. Although, it is a
Transport for utopian idea to totally eliminate other traffic from the bus station
area unless it is part of a major town center revitalization project,
London; and CONCLUSIONS investment in bus-actuated signals could markedly improve the
assistance in
capacity.
primary surveys
This study tries to demonstrate that bus station design should 4
A type of kerb that has been used in some areas of London and
from Taku
and can be a more scientific process. Capacity and accessibility some cities in the UK, which enables the bus to be aligned with the
Fujiama, UCL.
considerations should be at the forefront. In other words, the bus platform without any horizontal gap.
layout of the bus paths, bus berths, pedestrian infrastructure,
kerb design, bus platform area, etc need not be purely driven
by architectural and urban design considerations.
REFERENCES
In case of Brighton, the existing constraints, namely lack of
space and not drastically disturbing the activity pattern in the Elaluf, Silva Monica (1994). Transputer-based microscopic
area, were the key considerations in developing the proposals. simulation of bus-stop operations. Unpublished MSc dissertation,
However, if such a project is undertaken the opportunities of University of London.
looking into a wider context should not be foregone. In other Fernández, R E (2001). Modelling bus stop interactions.
words, since Brighton Bus Station is located in the historic core Unpublished PhD dissertation, University of London.
of the city, aspects such as decongesting the area and making it Transportation Research Board (TRB) (2000). Highway capacity
more pedestrian friendly should be dovetailed with such pro- manual 2000. Washington, DC: TRB.
jects, if possible. Tyler, Nick [ed] (2002). Accessibility and the bus system: from
The author believes that local planning councils and trans- concepts to practice. London: Thomas Telford.

tec MAY 2006

Bus station2.qxd 27/4/06 17:32 Page 187

187

These drawings
were produced
with software
used by London
Buses,Transport
for London (TfL),
called MISS,
(Mobile
Interactive
Swept-path
Simulator).

tec MAY 2006