H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No.

1 (2011) 21–33

A model to predict capacity

of multi-lane roundabouts under
high demand flows in Bahrain
H.M.N. AL-MADANI
Department of Civil Engineering & Architecture, College of Engineering, University of Bahrain, Bahrain.

Abstract
An exponential model is developed for capacity estimate of roundabouts with triple circulating lanes based on
given circulating flows. The necessary data for the development of the model and for the comparison purposes
between the model and other international models are gathered from 13 roundabouts in Bahrain. The geometric
data were gathered from the actual drawings, scaled aerial photographs and from the field. The necessary traffic
data were gathered during morning and evening peak periods. The developed model falls well in between the
tested international models and matches the actual data reasonably well. Substantial differences in estimating
capacities were observed between the various available international methods. Such variations make the judgment
of accepting or rejecting the estimated capacities difficult. They will also make the gating strategies and traffic
assignments unreliable. The methods with complicated input parameters, extensive equations and tedious calcula-
tions, such as aaSIDRA, UK RODEL, French and Indian methods, did not prove to be better than much simpler
ones, such as HCM or FHWA methods, in estimating roundabout capacities when compared with actual data.
The influence of most of the geometric parameters of roundabouts on capacity during rush hours is limited. The
findings are quite essential for traffic planners in making judicious decisions regarding roundabouts’ performance.
There is a real need for a more consistent model for the capacity estimation of roundabouts.
Keywords: circulating or conflicting flow, roundabout capacity, triple circulating lanes, entry capacity.

1 Background
Current roundabouts came as a replacement of traffic circles (rotaries) to overcome some of the
practical deficiencies associated with circles such as locking up of the traffic and development of
long queues in the circles. This is because of the right of way given to the vehicles entering the cir-
cles. In order to avoid such blocking queues, the entering traffic should yield, or give way, to
circulating ones. This is the main principle of the current roundabouts. Roundabouts are quite popu-
lar in erstwhile British colonial countries. All the traffic circles in Bahrain were converted into
conventional roundabouts, soon after Britain, in the 1960s. Roundabouts are widely spreading in
many other countries because of their advantages over other types of intersection control. The prin-
cipal reason was the profit of safety over other cross-roads [1]. Reduction in incapacitating crashes
was observed in the USA, Denmark, France and Germany after adopting roundabouts in replace-
ment to other types of intersection control [2]. This is due to reduction in approaching speeds, fewer
conflict points and no direct left turning. Some of the other advantages include minimum mainte-
nance cost and a nice landscape. However, they also carry several serious disadvantages, many of
which are usually overlooked. Some of them are as follows: drivers frustration due to unpredictable
long queues during rush hours, when compared with traffic signals [3], limitation of bicycle and
pedestrian facilities, high construction cost, large land requirements, high tail end accidents, fre-
quent rutting failures at approach entries of flexible pavements, limitation of the technological
support or software compared with traffic signals and limitation of entry and circulating lanes.

1.1 History of roundabouts

Traffic circles or rotaries were part of the transportation system in the United States and in some of the
European countries long back. High crash experience and congestion in the circles led to the rotaries
© 2011 WIT Press, [Link]
ISSN: 1743-7601 (paper format), ISSN: 1743-761X (online), [Link]
DOI: 10.2495/SDP-V6-N1-21-33
22 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

falling out of favor in America since the mid-1950s. Internationally, the experience with traffic circles
was equally negative, with many countries experiencing circles that locked up as traffic volumes
increased [4]. The roundabouts were developed in the United Kingdom. It adopted a mandatory ‘Give-
Way’ rule at all circular intersections, which required entering traffic to give way, or yield, to circulating
traffic. Wardrop [5], Britain TRL researcher, developed some models related to roundabout capacity in
1957. The ‘Give-Way’ rule prevented circular intersections from locking up by not allowing vehicles
to enter the intersection until there were sufficient gaps in the circulating traffic. The roundabout rep-
resents a substantial improvement, in terms of operations and safety [2], when compared with older
rotaries. Therefore, many countries have adopted them as a common intersection form and some have
developed extensive design guidelines and methods to evaluate the operational performance of modern
roundabouts. Al-Madani [3] found that roundabouts perform better than traffic signals when the traffic
demand is low. However, as the demand increases at a roundabout so does the queue length and the
delays. Beside drivers’ frustration due to the unpredictable delays at congested roundabouts, when
compared with traffic signals, long queues are inevitable. The latter leads to police intervention in order
to control the queues and direct the traffic flow. Such phenomena led governments of Bahrain and
Qatar to convert most of the major roundabouts on the main roads into signalized intersections. Some
were constructed during the 1960s. Akçelik [6] also observed roundabouts to perform well at low to
medium flow conditions. Clear capacity reduction was observed at high demand levels.

1.2 Roundabout capacity

Roundabout capacity is the main determinant parameter for the performance measure of many other
parameters such as delay and queue length. It is the maximum sustainable entry flow rate that an approach
can accommodate during a specific period under prevailing traffic, geometric and control conditions.
Capacity is service rate and is different than maximum volume that an intersection can handle [6], which
is the practical capacity under high demand volume, not under prevailing conditions.
Most of the capacity models are either analytical ones based on gap acceptance, with no actual
observations, or empirical regression ones based on observed geometric and flow parameters (Fig. 1).
Both techniques are considered to be macroscopic ones. Kimber [7] stated that capacity estimates
based on gap acceptance models are not suitable for application in England. This was due to the
problems related to human behavior. Russell and Rys [8] also questioned the validity of gap accept-
ance models at near capacity conditions. On the other hand, Fisk [9] finds regression models to be
difficult for frequent application due to large number of data requirements. Microscopic models
typically simulate traffic system on a vehicle-by-vehicle basis by updating position, speed, accelera-
tion, lane position and other variables on small time steps such as a one-second interval or less [10].
Stanek and Milan [11] recommended use of macroscopic methods, such as FHWA, RODEL and
aaSIDRA, for the capacity use for unsaturated conditions. For oversaturated conditions microscopic
methods such as Paramics and VISSIM are preferred. However, Stanek and Milan [11] placed their
recommendations based on very limited tested roundabouts. Furthermore, they utilized two famous
macroscopic models, RODEL and aaSIDRA, for the comparison purposes.
The maximum flow rate that can be accommodated at a roundabout entry depends mainly on the
following factors: the circulating flow on the roundabout that conflicts with the entry flow, exiting
flow and the geometric elements of the roundabout. When the circulating flow is low, drivers at the
entry are able to enter the roundabout without any significant delay. The larger the gaps, i.e. the
headways, in the circulating flow, the more useful they are for the drivers entering the roundabout.
In fact, more than one vehicle may enter in each gap. As the circulating flow increases, the size of
the gaps in the circulating flow decreases, and the rate at which vehicles enter the roundabout
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 23

Figure 1: Geometric parameters needed for the roundabout capacity models.

decreases. The geometric elements of the roundabout also affect the rate of entry flow. The most
important geometric elements are the width of the entry, the width of the circulatory roadway or the
number of lanes around the central island. Wider circulatory roadways allow vehicles to travel along-
side, or follow, each other in tighter bunches and so provide longer gaps between bunches of vehicles.
The flare length also affects the capacity. The inscribed circle diameter and the entry angle have
minor effects on capacity. In fact, TRL invented mini roundabouts that performed better than some
larger ones in terms of capacities [12].
There are several analytical and empirical models for the capacity estimations. Some are very well
known and others are less popular. Many countries utilize models developed by their own research-
ers to meet their needs. However, capacities estimated through these models widely differ from one
model to another. Some are very simple and require minimum data entry as the Swiss, the US High-
way Capacity Manual (HCM) and Federal Highway Administration (FHWA) models, others are far
more complicated requiring extensive data gathering and tedious calculations such as UK RODEL,
Australia aaSIDRA, French GIRABASE and Indian models. The UK RODEL and the Australian
aaSIDRA models are the most famous ones. The question of how good the capacity estimate of each
model is requires further investigations. Pratelli [13] found clear differences in capacity estimates
when he used French and Swiss models as compared with actual data in Italy. Overestimation of
25–79% was observed in the capacity estimates. The need for capacity evaluation developed through
the various models and software programs, available worldwide, was stressed by Jacquemart [14].
24 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

1.3 International roundabout capacity models

Most of the capacity models are developed in the West European countries and in Australia. There
are several models currently used by both the researchers and the operators. Some are widely used,
others are not. Stanek and Milan [11] stated that most of the known capacity models can be cali-
brated by modifying the intercept values, as in RODEL model, and follow up headway, as in
aaSIDRA model.

1.3.1 UK model (RODEL)

It is based on the work carried out by Kimber and Hollis [15] and Kimber [12] for TRL. This method
has been incorporated into a software packages widely known as RODEL [16]. The model involves
extensive geometric requirements. The basic model is as follows:

qe,max = k*(F – fc*qc) for qe > 0 else qe = 0

F = 303x2; fc = 0.21TD*(1+0.2x2); S = (e – v)/l′
x2 = v+((e – v)/(1 + 2*S)); TD = 0.5/(1 + exp((Di – 60)/10))
k = 1 – 0.00347(ϕ– 30) – 0.978 (1/r – 0.05)

1.3.2 Australian model (aaSIDRA)

Detailed capacity expressions published in Australia are available in studies carried out by Akçelik
[17, 18], and have been incorporated into a widely known software called aaSIDRA. Many parameters
are employed in aaSIDRA [6] for sensitivity analysis to count for driver behavior.

qe,max = (max fod  qg, qm)

qg = (3600/b)*(1– Δc (qc/3600))+0.5b*ϕc(qc/3600)*exp(–l*(α – Δc)/3600)
q = min(qe, 60nm), fod = 1 – fqc (PqdPcd)
l = (ϕcqc/3600)/(1 – Δcqc/3600) for (qc/3600) ≤ (0.98/Δc)
= (49ϕc/∆c) Else
ϕc = exp(–3.0*qc/3600) for nc = 2 or 3, for 0.01 < ϕc < 1.0
qc = qcr + qco= flow on inner and outer circulating flow, rds = qd/qs
b = bd = b′o – 3.94*10–4*qc for 1.2 < b < 4.0
b′o = 3.3 – 0.0208Di + 0.889*10–4Di2 –0.395ne + 0.388nc subject to 20 < Di < 80
a = (3.6135 –3.137*10–4*qc –0.339*e′ –0.277*nc)*b for qc ≤ 1200
= (3.2371 – 0.339*e′ – 0.2775*nc)*b Else
subject to 3.0 ≥ a/b ≥ 1 and 2.2 ≤ α ≤ 8.0
fqc = 0.04+0.00015*qc for qc < 600
= 0.55 for qc > 1800
= 0.00035*qc – 0.08 for 600 ≤ qc ≤ 1800

1.3.3 Australian model (NAASRA)

The NAASRA model was developed earlier than SIDRA model. The model is summarized as
follows [19–21]:

qe,max = nc*qc*exp(–qc *T/3600)/(1 – exp(–qc*T0/3600)),

where T = 6 and T0 = 3.
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 25

1.3.4 German model

The German Highway Capacity Manual has officially introduced in 2001 [22] the Tanner-Wu capac-
ity equation [23]. The formula for entry flow [24, 25] is as follows:
qe,max = (3600/tf)*(nenf/(nf+1)S)*(1 – (Δc*qc/3600)/nc)nc*exp(–(qc/3600)*(a – (tf/2) – Δc))
Utilized default values: α = 3.3s, tf = 3.1s, Δc = 1.8s, nf =1.4 vehicles.

1.3.5 French model

The model is based on the work carried out by Louah [26] which was later incorporated into a model
known as GIRABASE [27].
qe,max = A*exp(–CB*qg)
qg = qaka(1 – (qa/(qc+qa))) + qcikt,i +qcekt,e, A = 3600/tf(e/3.5)
qci = conflicting flow on inner lane (default 0.4*qc)
qce = conflicting flow on outer lane (default 0.6*qc)
CB = 3.525 for urban area,
Li,max = 4.55 √(R+(w/2))
Ka = (R/(R+ wl)) – (Li/Li,max) for Li < Li,max, Ka = 0 Else
kt,i = Min((160/(w*(R + w)) Else = 1
kt,e = Min((1 – ((w – 8)/w)*(R/(R + w)2))) Else = 1

1.3.6 US HCM model

There are two methods currently found in the US literature. The first is found in the Highway Capacity
Manual – HCM [28]. The second is a simplified British linear regression method cited in the Federal
Highway Administration (FHWA) Roundabout Guidebook [29]. The former is as follows:
qe,max = 1230*nc*exp(–0.0009*qc)    for multi-lanes

1.3.7 US FHWA model

US FHWA model is as follows [29]:
qe,max = 2424 – 0.71*qc    for ne =2;    Di >50 m

1.3.8 Indian model

The basic concept of the model was based on Wardrop literatures of the late 1950s. The modified
equations are as follows [30]:
qe,max = ((280*w)*(1 + (e/w))*(1 – (p/3))/(1 + (w/l))
Average e = (e + e2)/2, p = qc/(qc + qe)
Subject to 0.1 < e/w < 0.4; 0.1 < w/l < 0.4; 0.4 < p < 1.0; and 18 < l < 90

1.3.9 Simulation models

Simulation techniques cover modeling of complicated traffic operation [10]. Simulation has not yet
been used much in studying roundabout performance. However, not many simulation software are
flexible enough to allow the user to model roundabouts [31]. Multi-lane simulations are very limited
because procedure for setting out priority rules for roundabouts entry are quite complicated [31]. Cube
Dynasim and VISSIM have been used to simulate many urban network including some roundabouts.
26 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

However, no clear simulation program has yet been utilized for roundabouts during congested condi-
tions [10]. Simulation models require more input data and are more time consuming when compared
with macroscopic ones.
Further to the earlier mentioned models Al-Madani and Saad [32] have developed a capacity
model for roundabouts in Bahrain with triple entry and circulating lanes. The model is as follows:
qe = 2952.9e–0.0007Qc
The model fall within the earlier mentioned models, but is much simpler than many of them, and
matched the actual data reasonably well.

2 Objectives
The main objective is to develop a capacity model for roundabouts with triple circulating lanes under
the saturated traffic demands in Bahrain. The developed model is compared with eight international
capacity models.

3 Methodology
The capacity model is developed using data from 13 existing major roundabouts, out of 15 selected ones,
in Bahrain. These cover triple circulating lanes by either dual or triple entry lanes. The roundabouts in
Bahrain are designed to high international standards. The model is developed based on the maximum entry
flow, i.e. actual capacity, and the corresponding circulating flow. The former will be used as the dependent
variable while the latter as the independent one. The data are collected during the peak hours to ensure
higher saturation flows necessary for the development of capacity models. Furthermore, they provide bet-
ter basis for comparison [33] with other models. Least square regression models are used to develop the
model. The data will be regressed based on linear, logarithmic, polynomial, power and exponential
­methods. The model with best fit in terms of highest R2 values will be selected to represent the data.
All the necessary traffic and geometric data are carefully gathered and properly considered to
avoid misinterpretations of the parameters, which are used in the various models, as they may vary
from one method to another. These parameters are unified and given the same abbreviated letters.
Afterward, the roundabouts’ maximum entry flows are determined for each approach of the selected
roundabouts for nine international methods. These are determined using the advanced EXCEL fea-
tures. The estimated maximum entry flows are based on the equations stated earlier. None were
produced through the developed software corresponding to the earlier mentioned methods, in case
there are any. This is to avoid unrealistic assumptions and default values utilized in the software for
simplification purposes. The maximum entry capacity is analyzed per approach entry flow, not per
individual lane, since most of the considered models utilize the traffic per approach entry. This is just
not to add further complications to the models that are already complex. Furthermore, researchers
prefer them over individual lane determination [33]. The calculated capacities are compared with
that determined for the developed model.
The circulating and exiting flows are quite difficult to measure. Therefore, they are measured
indirectly. The circulating flow for the South approach in Fig. 1, for example, consist of through
traffic flowing from the West approach, left turning traffic flowing from the North approach and the
U-turning traffic from the East approach. The latter is usually very limited during rush hours. Exiting
traffic is considered in the same way, as well. Data from both a.m. and p.m. traffic are considered for
the development of the capacity model.

4 Data gathering
A total of 15 roundabouts were first selected across Bahrain for the investigation. The selected
roundabouts carried saturated traffic flows during peak periods, relatively large inscribed diameters,
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 27

i.e. over 60 m, and similar approaching lane widths. These were short-listed to 13 roundabouts based
on the following requirements: multi-entry lanes, multi-circulating lanes, either four or five approach-
ing legs, being on the main roads, and minimal approaching grades. These are meant to lead to better
consistency in the results and fair comparison between the considered models. The geometric param-
eters require careful attention and high caliber to measure them as to fulfill the needs for the various
international methods. Many are very tedious to acquire. The process of gathering the geometric
parameters was not an easy one. The geometric parameters of the roundabouts were gathered from
actual drawings of the roundabouts, GIS maps with scale of (1:2000), scaled aerial photos and actual
field measurements. These were necessary to cross-check the data extracted from one source with
another, to measure the missing geometric parameters from the original drawings and to compare the
proposed drawings with the actual ones. The traffic flow counts during both morning and evening
peaks were gathered for each approach.
The range of measured geometric parameters for the considered roundabouts, along with the aver-
ages, are presented in Table 1. Majority of the parameters fall within the recommended values by the
UK Ministry of Transport, as has been discussed by Salter [34].
It is quite essential to mention that many methods require measurement of many tedious param-
eters such as flare length, entry angles splitter width and weaving length. Accuracy in measuring
them may vary between one reading and another for over one meter, no matter how accurate your
measurement is even when digital tools are utilized. This is because of difficulties in locating the
beginning and the end of the measured parameter. However, such differences did not affect the esti-
mated entry capacity much.

5 Model development
Various types of models are regressed for the estimated maximum entry flows as being the depend-
ent parameter versus the circulating flows as being the independent one. The models best fitting the
gathered maximum entry flow versus circulating flow data along with their corresponding R2 values
are presented in Table 2. These covered the following five main types: linear, logarithmic, exponen-
tial, quadratic and power regression models. The exponential model fits the tested data best when

Table 1: Geometric parameters of the roundabouts used in the study.

Minimum and maximum Averages of measured

Parameter measured values values
Number of circulating lanes 3 3
Number of entry lanes 2 and 3 2 and 3
Inscribed diameter (m) 63–150 105
Entry angle 8–32.5 19
Entry radius from edge (m) 11.5–120 45
Flare effective length (m) 10–55 21
Approach half width of lanes (m) 6–10 7.6
Entry width all lanes (m) 6.5–16 9.3
Width of non-weaving (m) 9–14 6.1
Width weaving section or circulating width (m) 7.5–20 11
Length weaving section (m) 24–99 61
Radius of central island (m) 24–64.5 40
Width of splitter island (m) 18–82 31
28 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

Table 2: Developed models for roundabout capacity.

Model type Developed models R2

Exponential 2768e–0.0007Qc 0.559
Quadratic 2534.9 – 1.1216Qc + 0.0001Qc2 0.536
Logarithmic 6323.1 – 714.45ln Qc 0.501
Linear 2142 – 0.5785Qc 0.480
Power 146,510Qc–0.7076 0.415

compared with the other models because it holds the highest R2 value (0.56). However, the quadratic
relationship does not differ much from the exponential model in terms of R2(0.54). The R2 values for
the models though being insufficiently high, they look fine for such dispersive nature of data and
relatively large number of data points. It is quite interesting to mention that as the extreme data
points, i.e. those beyond 1.8 of the standard deviation around the model, are removed the R2 value
improves to around 0.68 with little change in the model’s constant values.

The maximum entry flow rates predicted from the developed model is compared with eight other
international models. As can be seen from Figs 2 and 3 the following observation can be made
regarding the developed model:

1. The model falls well in between the tested international methods.

2. The model fairly matches the Australian NAASRA model at low circulating flow rates but
clearly falls above it at medium and high circulating flow rates. In fact, the NAASRA model
falls below all the other models.
3. The Australian aaSIDRA model falls far above the developed models, the actual data and most
of the other models.
4. The UK RODEL model clearly showed higher maximum entry flows when compared with the
developed model.
5. The German model matched the developed model quite well. It is quite interesting to mention
that German roundabouts showed actual capacities less than that found in Britain [35]. This
match well with the results found here.
6. The French model falls above the developed model, especially for circulating flows below
2800 veh/h per approach, and above most of the actual data. The two match well beyond
2800 veh/h. The French model falls below the aaSIDRA and the RODEL.
7. The US HCM model matches the developed model quite well. However, it showed slightly
lower predictions at medium to high circulating flows.
8. The US FHWA method also showed quite close match with the model developed here. How-
ever, one may observe higher estimated entry flow values using FHWA at medium circulating
flow rates.
9. The Indian model showed far high estimations of entry flows compared with all the other ­models
and with the actual data.

6 Comparison between the international models

The results for the various roundabout capacity methods versus circulating flows, for the different
roundabouts, as shown in Figs 2 and 3, clearly show the Indian method to be highly overestimating
the entry capacity for given circulating flow and corresponding geometric characteristics when
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 29

7000
UK RODEL

6000 NAASRA

aaSIDRA
5000 actual entry flow
Entry flow (veh/h)

Expon. (2768e-0.0007x)
4000

3000

2000

1000

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Circulating flow (veh/h)

Figure 2: Developed capacity model for multi-lane roundabouts along with UK, NAASRA and
aaSIDRA models.

7000
Indian
US HCM
6000 US FHWA
German

5000 French
actual entry flow
Entry flow (veh/h)

Expon. (2768e-0.0007x)
4000

3000

2000

1000

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Circulating flow (veh/h)

Figure 3: Developed capacity model for multi-lane roundabouts along with HCM, FHWA, German
and French models.

c­ ompared with the actual demand and with those estimated from other tested methods. The results
also indicated wide variations in capacity estimates of roundabouts among the various international
methods. Some showed to be more reasonable than others when compared with the actual demand
during rush hours. Such differences confuse the practicing engineers and consultants in properly
assigning traffic to the various routes. Gating policies will also suffer just decisions.
The Australian NAASRA method showed the lowest capacity estimates compared with other
international methods and compared with the actual data. The Australian aaSIDRA method, on the
other hand, showed the highest capacity estimates, excluding the Indian method, for circulating
flows up to 2300 veh/h. In other words, the maximum entry flows estimated by the two methods
30 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

showed two extreme sides of predictions. One on the far high side and the other on the low side.
It is also interesting to mention that very high estimates of entry capacities are observed for low
circulating flow values when either the Australian aaSIDRA or the French methods are considered.
They may require careful calibration and threshold limits for low circulating flows. Similar to the
two Australian methods, i.e. NAASRA and aaSIDRA, the two US methods FHWA and HCM also
showed a high and a low estimate of maximum entry flows, respectively, for the various given
parameters when compared with the actual data. In general, the estimated entry flows using both
methods fall between the envelopes of the two Australian methods, especially when the circulating
flows falls are less than 2700 veh/h. The estimated roundabout entry capacity through the US HCM
method showed to be closely matching that determined through the German method. Both tend to
show low capacity estimates for given circulating flows when compared with the UK RODEL, the
US FHWA, the French and the aaSIDRA methods. The maximum entry flows estimated through the
French method falls below that estimated through the UK RODEL method.
The complicated methods, i.e. those requiring many parameters and involving several tedious
equations, such as the aaSIDRA, the UK RODEL, the French GIRABASE and the Indian methods,
did not show better capacity estimates than much simpler ones, such as the US FHWA or the HCM
methods, when compared with the actual demands during peak periods. The latter match the demand
flow reasonably well. The probable explanation is that motorists will be approaching and crossing
the roundabouts at relatively low speeds during the forced flow of the rush hours. This makes the
influence of many of the geometric parameters, such as the entry angle, flare length, entry radius and
lane width, on capacity very limited, simply because motorists have limited space for maneuvering.
The influence of geometric parameters might be more crucial to the motorists during nonforced flow
conditions, through the vehicles being forced to slow down. The motorists have better approaching
and crossing choices compared with congested conditions during rush hours.

7 Conclusions and recommendations

An exponential model for the capacity estimates for triple circulating lanes’ roundabouts is devel-
oped here. The model predicts the maximum entry capacity of an approach knowing the circulating
flow. The model matches the actual demand during the peak periods reasonably well and falls well
in between the international models. Substantial differences in the capacity estimations were
observed among the various available international methods. Such differences make the judgment of
accepting or rejecting the estimated capacities very difficult. They also cause confusion to practicing
engineers. While the capacities estimated through the Australian NAASRA method showed to be the
lowest among those tested, that determined through the Australian aaSIDRA method showed to be
among the highest. The US FHWA and HCM also showed a high and a low estimated entry flows,
respectively, compared with the actual data and with the other tested methods. The aaSIDRA and the
French methods showed nonrealistic high capacity estimates at low corresponding circulating flows.
The estimated capacities through the HCM method matched that determined through the German
method closely. However, they tend to show low estimated values when compared with UK RODEL,
aaSIDRA and USFHWA methods. The French method falls below the UK RODEL method.
The complicated methods, those involving measuring many parameters and extensive calcula-
tions, such as the aaSIDRA, the RODEL, the French, and the Indian methods, did not necessarily
show better capacity estimates than those much simpler ones as the HCM or the FHWA methods.
This is probably because of the low speed of vehicles during rush hours, which make the influence
of the geometric parameters minimal.
The findings are quite essential for the traffic planners in making judicious decisions regarding
roundabouts’ performance. There is a real need for a more consistent model for the capacity estimation
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 31

of roundabouts. Such a model should utilize parameters that can easily be measured and compre-
hended by the users.

Nomenclature
qe,max maximum entry flow for an entry lane (veh/h)
qg minimum entry flow (veh/h)
qc, Qc conflicting flow (veh/h)
qci conflicting flow on inner lane (default 0.4qc) (veh/h)
qce conflicting flow on outer lane (default 0.6qc) (veh/h)
qe entry arrival flow (veh/h)
qa exiting flow (veh/h)
nm minimum entry flow (veh/min)
ne number of entry lanes
nc number of lanes in circulating (conflicting) flow
rds ratio of dominant and subdominant flow in the entry
fod origin-destination adjustment factor
Pcd  Pqd proportion of total circulating flow 0.5–0.8 (0.6 used)
Δc minimum headway in circulating traffic (s)
a critical headway (s)
tf follow-up time (2.05 s in French model, 3.1 s in German model, 2.6–3.1 in HCM model)
l arrival headway distribution factor (veh/s)
ϕc proportion of unbunched conflicting vehicles in circulatory stream
b follow-up headway (s)
Di inscribed diameter (outer diameter of the roundabout) (m)]
R radius of central island (m)
e width at entry all lanes (m)
e′ average entry width per lane (m)
e2 width of non-weaving section (m)
w width of circulating lanes (m)
l length of weaving section between the ends of the channelized islands (m)
r narrowest radius of the right edge at the entry (m)
l′ effective length of the funnel-shaped flare (m)
v width of the lane on the approaching street (m)
Li width of the splitter island (m)
nf short lane length (veh)
S measure of the degree of the flaring (°)
ϕ angle between ring and entry (°)
p proportion of conflicting (weaving) traffic

Acknowledgement
The author wishes to thank the Deanship of Scientific Research at the University of Bahrain for
supporting part of this study. The author also thanks Dr. Y. Najjar, Prof. S. Alvi, Mrs. Z. Saleh,
Mr. A. Jaafar and H. Qadhi for their assistance during the study.

REFERENCES
[1] Guichet, B., Evolution of roundabouts in France. National Roundabout Conference, RA 056A,
TRB: Colorado, USA, 2005.
32 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011)

[2] Pratelli, A. & Souleyrette, R.B., Visibility, perception and roundabout safety. WIT Transactions
on the Built Environment, vol. 107. Proceedings Urban Transport XV Urban Transport and the
Environment, ed. C.A. Brebbia, WIT Press: UK, pp. 577–588, 2009.
[3] Al-Madani, H.M.N., Dynamic vehicular delay comparison between a police-controlled
roundabout and a traffic signal. Transportation Research, 37A, pp. 681–688, 2003.
[4] Taekroutok, T., Modern roundabouts for Oregon. Department of Transportation, Research
Unit, OR98-SRS22, Salem, USA, 1998.
[5] Wardrop, J.G., The traffic capacity of weaving sections of roundabouts. Proceedings of the
First International Conference on Operational Research, English Universities Press: London,
UK, 1957.
[6] Akçelik and Associates, aaSIDRA In Put Guide: Roundabout Data. Revised Version, Akçelik
and Associates Pty Ltd., Greythorn, Melbourne, Australia, 2005.
[7] Kimber, R.M., Gap acceptance and empiricism in capacity prediction. Transportation Science,
23(2), pp. 100–111, 1989. doi:10.1287/trsc.23.2.100
[8] Russell, E. & Rys, M., Modeling Traffic Flows and Capacities at Roundabout, MBTC 1099,
Kansas State University: Manhattan, USA, 2000.
[9] Fisk, C.S., Traffic performance analysis at roundabouts. Transportation Research, 25B,
pp. 89–102, 1991. doi:10.1016/0191-2615(91)90016-C
[10] Boxill, S.A., An Evaluation of 3-D Traffic Simulation Modeling Capabilities. Report 167621-1,
Center for Transportation Training and Research, Texas Southern University: Houston, USA.
[11] Stanek, D. & Milan, R.T., High-capacity roundabout intersection analysis: going around in
circles. National Roundabout Conference, RA 057B, TRB: Colorado, USA, 2005.
[12] Kimber, R.M., The Traffic Capacity of Roundabouts. TRRL, LR942, Crowthorne, UK, 1980.
[13] Pratelli, A., Design of modern roundabout in urban traffic system. WIT Transactions on
the Built Environment, vol. 89. Proceedings Urban Transport XII Urban Transport and the
Environment, ed. C.A. Brebbia, WIT Press: UK, 2006.
[14] Jacquemart, G., Modern Roundabout Practice in the United States. Synthesis of Highway
Practice 264, Transportation Research Board: Washington, DC, USA, 1998.
[15] Kimber, R.M. & Hollis, M., Traffic Queues and Delays at Road Junctions, Transport and Road
Research Laboratory, LR909: Crowthorne, Berkshire, UK, 1979.
[16] Crown, R.B., RODEL1: Interactive Roundabout Design (Software Manual), RODEL Software
Ltd.: Stoke-on-Trent, UK, 1987.
[17] Akçelik, R., Chung, E. & Baseley, M., Performance of roundabouts under heavy demand
conditions. Road and Transport Research, 5(2), pp. 36–50, 1996.
[18] Akçelik, R., Chung, E. & Baseley, M., Roundabouts: Capacity and Performance Analysis.
Research Report APR 321, ARRB, Transport Research Ltd.: Vermont South, Australia, 1998.
[19] National Association of Australian State Road Authorities (NAASRA), Guide to Traffic
Engineering Practice- Roundabouts. Part 6, AP-G11.6, National Association of Australian
State Road Authorities: Sydney, Australia, 1998.
[20] National Association of Australian State Road Authorities (NAASRA), Guide to Traffic
Engineering Practice- Intersections at Grade. Part 5, National Association of Australian State
Road Authorities: Sydney, Australia, 1998.
[21] Adams, M., Barker, D. & Rye, T., Roundabouts. Unit 9, BE71008: Highway Planning and
Design, 3rd edn, Napier University, School of the Built Environment: Scotland, UK, 2004.
[22] German Highway Capacity Manual (Forschungsgesellschaft fur StraBen- und Verkehrswesen,
Handbuch fur die Bemessung von StraBenverkehrsanlagen, HBS), FGSV, no. 299, Verlag
GmbH, Köln, Germany, 2001.
 H.M.N. Al-Madani, Int. J. Sus. Dev. Plann. Vol. 6, No. 1 (2011) 33

[23] Brilon, W., Wu, N. & Bondzio, L., Unsignalized intersections in Germany- A state of the art
1997. In: Proceedings of the Third International Symposium on Intersections Without Traffic
Signals, ed. M. Kyte, University of Idaho: Portland, Oregon, USA, 1997.
[24] Wu, N., Capacity of shared- short lanes at unsignalized intersections. In: Proceedings of the
Third International Symposium on Intersections Without Traffic Signals, ed. M. Kyte, ­University
of Idaho: Portland, Oregon, USA, 1997.
[25] Wu, N.A., Universal procedure for capacity determination at unsignalized (Priority-Controlled)
intersections. Transportation Research B, 35, pp. 3, 2001.
[26] Louah, G., Panoroma critique des models Francais de capacite des carrefours giratoires.
Proceedings of the Seminar Roundabouts 92, SETRA, BP 100, F 92223 Bagneux, Nantes,
France, 1992.
[27] GIRABASE, Calculation of roundabout capacity, [Link], 2008.
[28] Transportation Research Board, Highway Capacity Manual. National Research Council, TRB,
SR 209: Washington DC, 2000.
[29] Federal Highway Administration (FHWA), Roundabouts: An Information Guide. FHWA-
RD-00-067, US Department of Transport, Virginia, USA, 2000.
[30] Kadiyalli, L.R., Traffic Engineering and Transport Planning, Khanna Publisher: New Delhi,
India, 1999.
[31] Bared, J.G. & Edra, P.I., Simulated capacity of roundabouts and impact of roundabout within a
progressed signalized Road. National Roundabout Conference, TRB: Colorado, USA, 2005.
[32] Al-Madani, H.M.N. & Saad, M., Analysis of roundabout capacity under high demand flows.
WIT Transactions on the Built Environment, vol. 107. Proceedings Urban Transport XV
Urban Transport and the Environment, ed. C.A. Brebbia, WIT Press: UK, pp. 223–234, 2009.
[33] Seiberlich, E.L., A Formulation to Evaluate Capacity and Delay of Multilane Roundabouts
in the United States for Implementation into a Trend Forecasting Model. [Link]. Thesis, The
University of Wisconsin: Milwaukee, USA, 2001.
[34] Salter, R.J., Highway Traffic Analysis and Design, 2nd edn, Macmillan: London, UK, 1989.
[35] Brilon, W. & Stawe, B., Capacity and Design of Traffic Circles in Germany, Transportation
Research Record 1398, TRB, NRC: Washington DC, USA, 1993.