Transportation Research Part D 49 (2016) 313–322

Contents lists available at ScienceDirect

Transportation Research Part D

journal homepage: www.elsevier.com/locate/trd

A new probability statistical model for traffic noise prediction

on free flow roads and control flow roads
Feng Li a, Shaoyi Stephen Liao b, Ming Cai c,⇑
a
School of Automotive Engineering, GuangDong Polytechnic Normal University, Guangzhou, China
b
Department of Information Systems, City University of Hong Kong, Hong Kong, China
c
School of Engineering, Sun Yat-sen University, Guangzhou, China; Guangdong Provincial Key Laboratory of Intelligent Transportation System, Guangzhou,
China; and Guangdong Provincial Engineering Research Center for Traffic Environmental Monitoring and Control, Guangzhou, China

a r t i c l e i n f o a b s t r a c t

Article history: A new traffic noise prediction approach based on a probability distribution model of vehi-
Available online 10 November 2016 cle noise emissions and achieved by Monte Carlo simulation is proposed in this paper. The
probability distributions of the noise emissions of three types of vehicles are obtained
Keywords: using an experimental method. On this basis, a new probability statistical model for traffic
Traffic noise noise prediction on free flow roads and control flow roads is established. The accuracy of
Prediction model the probability statistical model is verified by means of a comparison with the measured
Probability distribution
data, which has shown that the calculated results of Leq, L10, L50, L90, and the probability dis-
tribution of noise level occurrence agree well with the measurements. The results demon-
strate that the new method can avoid the complicated process of traffic flow simulation but
still maintain high accuracy for the traffic noise prediction.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The current development of road traffic has brought about serious noise pollution in cities. Accurate traffic noise predic-
tion is a precondition of traffic noise control and an important task of environmental management. There are currently two
types of popular traffic noise prediction models: steady-state calculation models and dynamic simulation models. A steady-
state calculation model can be applied to predict the average traffic noise over a period of time on the basis of the average
traffic volume, average traffic speed, etc. Researchers have performed a large body of work in traffic noise prediction and
established various influential steady-state calculation models, such as the FHWA highway traffic noise prediction model
in the United States (Rochat and Fleming, 2002), the CRTN model in the United Kingdom (Department of Transport and
Welsh Office UK, 1988), the RLS90 model in Germany (Road Construction Section of the Federal Ministry for Transport,
1990), the ASJ RTN-model in Japan (Yamamoto, 2010) and the CNOSSOS model in Europe (Kephalopoulos et al., 2012). In
addition, some specific models for intersections (Makarewicz et al., 1999; Makarewicz and Kokowski, 2007; Abo-Qudais
and Alhiary, 2007) and roundabouts (To and Chan, 2000; Covaciu et al., 2015) also belong to the steady-state calculation
model category. Although all of the above models can predict the equivalent noise level Leq or the time average noise level
LAT over a given period of time, they cannot calculate the dynamic change of noise.
A dynamic simulation model can be used to predict not only the Leq over a period of time but also the second-by-second
dynamic changes of the noise level (Jacobs et al., 1980; Chevallier et al., 2009; Guarnaccia, 2013). This approach generally

⇑ Corresponding author.
E-mail address: caiming@mail.sysu.edu.cn (M. Cai).

http://dx.doi.org/10.1016/j.trd.2016.10.019
1361-9209/Ó 2016 Elsevier Ltd. All rights reserved.
314 F. Li et al. / Transportation Research Part D 49 (2016) 313–322

works by combining a dynamic traffic simulation model, a vehicle noise emission model, and a sound propagation model
(Can et al., 2010; Luo et al., 2012). Dynamic simulation models are more and more widely used in predicting traffic noise
with the development of computer technology. For instance, Suzuki et al. proposed a dynamic traffic noise simulation model
for a large intersection (Suzuki et al., 2003). Chevallier et al. developed a new traffic simulation tool for roundabouts and
combined it with noise emission laws and a sound propagation model to study the dynamic noise near a roundabout
(Chevallier et al., 2009). Li et al. established a dynamic model for signalized road intersections among buildings and studied
the effect of traffic lights on the traffic noise (Li et al., 2011). Cai et al. used a dynamic simulation model to simulate the
dynamic change and study the characteristics of traffic noise near a signal-controlled pedestrian crossing junction (Cai
et al., 2011). However, almost all of the dynamic simulation models rely on a complex traffic simulation model. Furthermore,
almost all of the vehicle noise emission models within the dynamic simulation models are fixed algebraic expressions
related to the speeds and types of vehicles, which makes it difficult to represent the varied noise emission in the actual
situation.
The Monte Carlo method is a computational algorithm that relies on repeated random sampling to obtain numerical
results. It can be used to solve any problem having a probabilistic interpretation (Kroese et al., 2014). In transportation engi-
neering, it has been applied for traffic speed forecasting (Jeon and Hong, 2016), traffic flow simulation (Waldeer, 2003), and
traffic noise simulation based on speed distribution of vehicles (Alberto and Efraín, 2013; Iannone et al., 2013). To simplify
the process of microscopic traffic flow simulation and to preserve the diversity and accuracy of the prediction parameters,
this paper proposes a new traffic noise prediction approach that is based on a probability distribution model of vehicle noise
emissions and achieved by Monte Carlo simulation. First, we used an experimental method to obtain the probability distri-
butions of the noise emissions of three types of vehicles; then, on this basis, we established a probability model for traffic
noise prediction on free flow roads and control flow roads; finally, we verified the accuracy of the model with measured data.

2. Model development

2.1. Model for free flow roads

Whenever a car is travelling on a long straight road, as shown in Fig. 1, the sound pressure level at the receiving point can
be expressed as:
Lj ¼ Lj;0 þ Lj;D þ Lj;Other ð1Þ
where Lj,0 denotes the sound pressure level of a single vehicle measured at a standard distance, which can be seen as the
noise emission of a single vehicle; Lj,D denotes distance attenuation; and Lj,Other denotes the noise attenuation caused by other
factors, such as air absorption and shielding by barriers.
According to the Chinese Standard ‘‘GB 1495-2002” (Anon., 2002), Lj,0 can be expressed as the sound pressure level of a
single vehicle measured at a standard distance d0 = 7.5 m. In past studies, Lj,0 is related to the vehicle type and speed and is
commonly presented as a fixed expression that is fitted to a large amount of experimental data (Steele, 2001). However, the
noise emissions of different vehicles are not fixed but instead fluctuant within a certain distribution, even if the vehicles tra-
vel at the same speed. Therefore, we assume that the noise emission of each type of vehicle in each certain speed range obeys
a certain distribution, whose probability density function can be expressed as:
y ¼ f ðLj;0 Þ ð2Þ
By classifying the vehicles into several types according to their weight and further classifying each type of vehicle into
several speed grades, the probability density function f(Lj,0) can be obtained from a large number of experimental data
analysis.
Distance attenuation can be expressed as:

Fig. 1. Traffic noise propagation for the case of a long straight road.
 F. Li et al. / Transportation Research Part D 49 (2016) 313–322 315

2 2
d0 d0
Lj;D ¼ 10 lg ¼ 10 lg ð3Þ
d
2 x2 þ r2

where d0 denotes the reference distance when measuring the noise emission of a single vehicle, taken as 7.5 m in the Chinese
Standard ‘‘GB 1495-2002” (Anon., 2002); and d denotes the distance from the vehicle to the receiving point, dependent on
the coordinates of the vehicle on the road (denoted by x) and the vertical distance from the receiving point to the road
(denoted by r). Assuming that the length of the road affecting the receiving point is l (in meters), x obeys the uniform dis-
tribution within the range of l, whose probability density function is:

1
f ðxÞ ¼ ð4Þ
l
To exclude the impact of vehicles outside the studied road segment, the viewing angle of the receiving point to the road
segment is suggested to be larger than 150°. In this model, the value of l is 400 m.
Traffic flow on the road is divided into two directions, and then the noise generated by direction 1 at any second can be
calculated by summing all the sound powers of the vehicles within direction 1 via:
X
m
Li;1 ¼ 10 lg 100:1Lj ð5Þ
j¼1

where m denotes the number of vehicles on the road. In the case of free flow on a straight road, m obeys a Poisson distri-
bution (Roess et al., 2010), whose probability density function can be written as:

kk ek
Pfm ¼ kg ¼ ; k ¼ 0; 1; 2; . . . ð6Þ
k!
where k is the average number of vehicles on the road and is related to the hourly traffic volume Q and speed V (in m/s):

Q l
k¼ ð7Þ
3600  V
The noise generated by direction 2 at any second (Li,2) can be calculated using the same method. Then, the instantaneous
sound level can be obtained by summing the sound powers of the two directions:

Li ¼ 10 lgð100:1Li;1 þ 100:1Li;2 Þ ð8Þ

2.2. Model for controlled flow roads

For controlled flow roads, we divide the traffic lane into two segments, l1 and l2 (in meters), by the stop line. Similarly, to
ensure that the viewing angle of the receiving point to the road segment is greater than 150°, l1 and l2 are suggested to be
200 m. The running process of a vehicle under the control of traffic lights is shown in Fig. 2. Assuming that the signal period

Fig. 2. Sketch of a controlled flow traffic lane and time-space diagram.

316 F. Li et al. / Transportation Research Part D 49 (2016) 313–322

of the traffic light is T, the red time and green time are TR and TG, and t denotes the tth second of the signal period from when
the red light is on. Thus, t obeys the uniform distribution within the range (0, T). At the tth second of any period, we have:
When the traffic light is red (0 < t 6 T R ), vehicles within l1 are queuing and waiting to be released. In the case that the
traffic is not congested, the number of queuing vehicles at the moment m1,R is related to the hourly traffic volume Q and
t and obeys the Poisson distribution with k ¼ Qt=3600. The distance from the vehicle to the stop line, lx1, obeys the uniform
distribution within the range ðm1;R  ls ; 0Þ, where ls denotes the average space headway. At the moment, the head-vehicle of
the queue released during the last green time has traveled for TG + t seconds, and the tail-vehicle has traveled for t seconds.
Therefore, the number of vehicles m2,R within l2 becomes:

ðl2 =V  tÞ=ht ; l2 > Vt
m2;R ¼ ð9Þ
0; l2 6 Vt
where ht (in s/veh) denotes the average time headway of the traffic flow, namely the time interval between two consecutive
vehicles in the queue; V (in m/s) denotes the average speed of the leaving traffic flow; and l2/V denotes the required time for
the vehicle to leave l2. Similarly, lx2 obeys the uniform distribution within the range ðVt; l2 Þ.
When the traffic light is green (T R < t 6 T), the queuing is releasing, and the number of vehicles m1,G can be calculated by:
t  TR
m1;G ¼ m1;t  ð10Þ
ht
where m1,t obeys the Poisson distribution with k ¼ Qt=3600 and lx1 obeys the uniform distribution within the range
ðm1;G  ls ; 0Þ. The number of vehicles m2,G within l2 at the moment is considered in two cases: when t  TR < l2/V , all the
vehicles released during the present green time are in l2; when t  T R P l2 =V, some of the vehicles released during the pre-
sent green time have left l2. Therefore, we have:

ðt  T R Þ=ht ; t  T R < l2 =V
m2;G ¼ ð11Þ
l2 =Vht ; t  T R P l2 =V
Similarly, lx2 obeys the uniform distribution within the range ð0; Vðt  T R ÞÞ.
Eqs. (9)–(11) give the calculation approach for the number of vehicles on the road at any given moment. On this basis, the
instantaneous sound level generated by a controlled flow road can be calculated by referring to Eqs. (1), (3), (5) and (8).

2.3. Determination of the noise emission generated by a single vehicle

Experimental methods were used to obtain the noise emission of a single vehicle, Lj,0. The experimental scenario was set
according to the Chinese standards GB 1495-2002 (Limits of external noise of an accelerating vehicle and relevant measure-
ment methods) (Anon., 2002) and GB 1496-79 (Methodology of Surveying Traffic Noise of Individual Vehicles) (Anon., 1979).
A straight road with low traffic and around 100 m without obvious reflectors is chosen. The sound level meters were set
1.2 m above the ground and 7.5 m from the traffic lane on each side of the road. The vehicles were divided into three types:
light vehicles, with a length less than 6 m and a weight less than 2 tons; heavy vehicles, with a length more than 10 m or a
weight more than 12 tons; and medium vehicles, i.e., anything other than a light vehicle or heavy vehicle. The noise emis-
sions of 3300 vehicles in total (including 2429 light vehicles, 354 medium vehicles and 517 heavy ones) were measured
when they passed the sound level meter. The speed of each vehicle was recorded by a radar speed indicator.
The speeds of the vehicles were divided into three grades: low speed (0–20 km/h), medium speed (20–50 km/h), and high
speed (more than 50 km/h). A normality test for the noise emission of each type of vehicle in each speed grade was carried
out using SPSS software by the Kolmogorov-Smirnov Method and Shapiro-Wilk Method, as shown in Table 1. The hypothesis
is that each group of noise emission data obeys a normal distribution. The results of the normality test show that the signif-
icance level of each group of noise emission data is greater than 0.05, which means that it cannot refute the hypothesis.
Although the significance levels of a few groups of data are relatively low (such as heavy vehicles with low speed and med-

Table 1
Normality test of the vehicle noise emissions data.

Vehicle Type Speed Sample size Kolmogorov-Smirnov Shapiro-Wilk

Statistic Sig. Statistic Sig.
Heavy Low 121 0.091 0.060 0.976 0.092
Heavy Medium 290 0.047 0.092 0.992 0.117
Heavy High 106 0.047 >0.200 0.985 0.269
Medium Low 82 0.148 0.073 0.952 0.169
Medium Medium 133 0.049 >0.200 0.996 0.918
Medium High 139 0.067 >0.200 0.983 0.087
Light Low 492 0.079 >0.200 0.979 0.142
Light Medium 776 0.037 >0.200 0.995 0.778
Light High 1161 0.031 0.053 0.998 0.437
 F. Li et al. / Transportation Research Part D 49 (2016) 313–322 317

ium vehicles with high speed), we still believe that the hypothesis is true and that the data approximately obey normal dis-
tributions. Based on statistical analysis, the average values l and the standard deviations r of the noise levels of the three
types of vehicles in the three speed grades are obtained (Fig. 3). Thus, formula (2) (the probability density function of Lj,0) can
be expressed as:
ðL lÞ
2
1  j;0
f ðLj;0 Þ ¼ pffiffiffiffiffiffiffi e 2r2 ð12Þ
2pr

2.4. Calculation steps of the model

2.4.1. For free flow roads

Traffic noise at any given moment can be simulated using the Monte Carlo method by the following steps:

Step 1: generate a random number m that denotes the number of vehicles and obeys a Poisson distribution according to
Eq. (6).
Step 2: randomly generate m groups of vehicle parameters (representing m vehicles, such as vehicle types, coordinates,
and speeds) according to the probability distribution of each parameter.
Step 3: repeat Step 1 and Step 2 to generate the parameters for the vehicles running in the other direction.
Step 4: randomly generate the noise emission of each vehicle according to the speeds generated in Step 2 and the noise
emission model presented in Section 2.3. Then, calculate the noise level at the receiving point produced by each vehicle,
which usually requires calculating the distance attenuation using Eq. (3) and the noise attenuation caused by other fac-
tors. Finally, sum the noise powers of all vehicles to obtain the noise level at the given moment.
Step 5: repeat the above calculation N times to obtain N instantaneous noise levels.
Step 6: calculate the equivalent continuous sound level Leq, the statistical sound levels L10, L50, and L90, and the probability
distribution of traffic noise over a period of time based on these instantaneous noise levels.

Fig. 3. Statistical result of the vehicle noise emissions.

318 F. Li et al. / Transportation Research Part D 49 (2016) 313–322

2.4.2. For control flow roads

The calculation steps for control flow roads are similar to the ones for free flow roads:

Step 1: generate a random number t that denotes the tth second of the signal period following the distribution t  U[0,T].
Step 2: generate the random numbers m1,R, m2,R or m1,G, m2,G, which denote the number of vehicles according to Eqs. (9)–
(11).

The remaining steps are the same as Steps 2 through 6 for free flow roads, which are given above.

3. Model validation

Field measurements for model validation were taken at two typical free flow roads (South Jinzhou Road and East Xiaoz-
hou Road) and two typical controlled flow roads (Binjiang Road and East Xingang Road) in Guangzhou city. Noise levels were
recorded by sound level meters, which were mounted 5 m from the road side and 1.2 m above the ground. For the typical
controlled flow roads, the sound level meters were simultaneously set 10 m, 30 m, and 50 m downstream of the stop line.
The simultaneous traffic flows were recorded by video cameras, and the two-way volumes of light, medium, and heavy vehi-
cles were count. The vehicle speeds were recorded by radar speed indicators to obtain the speed distributions of the roads.
The geometric parameters, such as the widths of the roads and the number of lanes of the roads, were measured. Meanwhile,
the red time, green time, and period of the traffic light were recorded.
The speeds of about 1/4 number of the vehicles were measured and performed a statistical analysis, as shown in Fig. 4. On
the free flow roads (South Jinzhou Road and East Xiaozhou Road), the vehicle speeds are mainly concentrated in 40–70 km/h.
On the controlled flow roads (Binjiang Road and East Xingang Road), the vehicle speeds are divided into two cases, deceler-
ated flow before the stop line and accelerated flow after the stop line. The distribution of the speed in the decelerated flow is
wider than the ones in the accelerated flow. The reason is that the traffic is less, so that some of the vehicles need to slow
down in front of the stop line, and some of the vehicles do not need to slow down but can through directly. The traffic volume
and speed level of each investigated road is shown in Table 2. It should be noted that, in regard to the controlled flow roads,
Table 2 just gives the distributions of the measured speeds when vehicles were entering or leaving the junction, and the
speeds when vehicles were about to stop or had just started to run were classified as ‘‘low speed” (which is not reflected
in Table 2 but considered in the calculation).

Fig. 4. Speed distribution obtained from the field measurement. The data of decelerated flow were measured at 100 m before the stop line, and the
accelerated ones were measured at 100 m after the stop line.
 F. Li et al. / Transportation Research Part D 49 (2016) 313–322 319

Table 2
Statistical table of the measured traffic volume (veh), speed distribution (%).

Roads Time Direction Traffic Volume (veh) Speed Distribution (%)

Heavy Medium Light Low Medium High
South Jinzhou Road 14:00–15:30 S to N 27 36 310 2 48 50
N to S 38 31 342 0 41 57
East Xiaozhou Road 9:00–12:00 E to W 224 138 1323 0 39 61
W to E 195 122 1160 4 45 51
Binjiang Road 16:00–17:50 E to W 45 39 426 10 72 18
W to E 59 42 532 8 69 23
East Xingang Road 9:00–12:00 E to W 135 108 1373 8 70 22
W to E 112 85 985 5 62 33

Each 30 min, as an experimental period, the traffic parameters obtained from the field measurements, such as traffic vol-
ume, probability distribution of the vehicle speeds, probability distribution of the vehicle types, and geometric parameters of
the road, were substituted into the model. Then, the random instantaneous noise levels for each second (1800 noise level
values for each experimental period) were calculated. In addition, these instantaneous noise level values were processed
to calculate the equivalent continuous sound levels Leq and the statistical sound levels L10, L50, and L90 of each experimental
period, and the results were compared with the measured ones.
Whether for the free flow roads or for the controlled flow roads, the calculated values of Leq, L10, L50, and L90 are in good
agreement with the measured ones, and the comparison results are shown in Figs. 5 and 6. We can see that the mean abso-
lute errors of all the evaluation parameters are less than 2 dB. The prediction accuracy of L50 is higher than that of Leq, L10, and
L90. On the whole, the prediction accuracy of the evaluation parameters for the free flow roads is higher than that for the
controlled flow roads. The main reason for this could be the complexity of the traffic on the controlled flow roads. The simple
probability model cannot simulate the traffic flow on the controlled flow roads accurately. Even so, the accuracy of the cal-
culated traffic noise on the controlled flow roads is acceptable. These results indicate that the traffic noise can be predicted
accurately by the probability model proposed in this paper, and the more complex evaluation parameters, such as the traffic
noise index (TNI) and noise pollution level (NPL), can be predicted by processing the calculated instantaneous noise levels
further.

Fig. 5. Comparison of the calculated and measured Leq, L10, L50 and L90 near a free flow straight road with 30 min as a period.
320 F. Li et al. / Transportation Research Part D 49 (2016) 313–322

Fig. 6. Comparison of the calculated and measured Leq, L10, L50 and L90 near a controlled flow straight road with 30 min as a period.

The probability distribution statistics regarding the noise level values are carried out, and the results are compared with
both the results given by the dynamic traffic noise simulation method presented in References Li et al. (2011) and Cai et al.
(2011) and the measured results. The main difference between the two approaches is that the method presented in Refer-
ences Li et al. (2011) and Cai et al. (2011) adopts the microscopic traffic flow simulation model in the ‘‘Paramics” software to
simulate the dynamic motion of the vehicles, while we adopt the Monte Carlo simulation method to do that in this paper.
Figs. 7 and 8 show the probability distribution of noise level values collected in 1 h (3600 data points in total) for the free
flow roads and the controlled flow roads, respectively. It is obvious that the probability distribution curve calculated by the
presented model is in good agreement with those obtained by the method in References Li et al. (2011) and Cai et al. (2011)
and by the measurement. This means that the probability model we proposed can be applied to predict the probability dis-
tribution of traffic noise levels. From the figures, we can see that the main difference in the results is that the sound level

Fig. 7. Comparison of the results given by the presented model, a dynamic traffic simulation method and measurement of the probability distribution of
noise level occurrence near a free flow straight road.
 F. Li et al. / Transportation Research Part D 49 (2016) 313–322 321

Fig. 8. Comparison of the results given by the presented model, a dynamic traffic simulation method and measurement of the probability distribution of
noise level occurrence near a controlled flow straight road.

distribution given by the dynamic simulation method is more concentrated in some areas and with a higher peak value,
while the result given by the presented model is more random and closer to the physical reality. This is because the noise
emission law in the dynamic simulation method is more unitary and with a weak dispersion when the traffic flow is stable.
One attraction is that the noise levels near a free flow road show a ‘‘single-peak” distribution, while the noise levels near a
controlled flow road have a ‘‘dual-peak” distribution. The results given by these two methods and the measurement simul-
taneously confirm the phenomenon. The reason is that the traffic flow is not continuous under the control of traffic lights,
and the traffic noise levels during the green light are significantly higher than those during the red light.

4. Discussion and conclusion

There are some errors between the simulated and measured data, especially in the probability distribution of traffic noise
levels. The main reasons for the error may include the following: first, that the speeds of the vehicles are simply divided into
three grades, and the noise emission in each grade does not strictly obey a normal distribution; second, that the parameters
of the actual traffic flow, such as position, speed, and number of vehicles, cannot be accurately simulated; and third, that
some influences due to the ground, plants, etc. are not considered in the calculation.
It should be noted that the model is suitable for roads with less traffic because it is established based on the assumption
that the traffic volume is low. The applicable condition of the model is that, for free flow roads, the traffic volume must be
low enough that the time headway of vehicles approximately obeys a Poisson distribution; similarly, for control flow roads,
all the queued vehicles must be released in a green period, i.e., the traffic is not congested. The error of the model may
increase when the traffic volume is high, which will be tested in a follow-up study.
On the whole, a new theoretical method based on probability models has been proposed to predict the traffic noise on
free flow roads and controlled flow roads. A probability model of the noise emission of three types of vehicles has been estab-
lished through an experimental method. Compared to the dynamic traffic noise simulation methods mentioned in Refer-
ences Li et al. (2011) and Cai et al. (2011), the new method presented in this paper avoids the complicated process of
traffic flow simulation but still maintains high accuracy in the traffic noise prediction. The calculated results of Leq, L10,
L50, L90, and the probability distribution of noise level occurrence agree well with the measurements, which implies that this
method could be applicable to traffic noise management in cities.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 11574407) and the Science and Tech-
nology Planning Project of Guangdong Province, China (No. 2015B010110005). Comments and suggestions from the review-
ers and editor are highly appreciated.

References

Abo-Qudais, S., Alhiary, A., 2007. Statistical models for traffic noise at signalized intersections. Build. Environ. 42 (8), 2939–2948.
Alberto, R., Efraín, D., 2013. Modeling urban traffic noise with stochastic and deterministic traffic models. Appl. Acoust. 74, 614–621.
Anon., 1979. Methodology of Surveying Traffic Noise of Individual Vehicles. China Environmental Science Press, Beijing.
Anon., 2002. Limits and Measurement Methods for Noise Emitted by Accelerating Motor Vehicles. China Environmental Science Press, Beijing.
322 F. Li et al. / Transportation Research Part D 49 (2016) 313–322

Cai, M., Li, F., Liu, J.K., 2011. Dynamic simulation and characteristics analysis of traffic noise at signal-controlled pedestrian crossing junction. Noise Control
Eng. J. 59 (5), 549–555.
Can, A., Leclercq, L., Lelong, J., Botteldooren, D., 2010. Traffic noise spectrum analysis: dynamic modeling vs. experimental observations. Appl. Acoust. 71 (8),
764–770.
Chevallier, E., Can, A., Nadji, M., Leclercq, L., 2009. Improving noise assessment at intersections by modeling traffic dynamics. Transp. Res. D-Transp. Environ.
14 (2), 100–110.
Chevallier, E., Ledercq, L., Lelong, J., Chatagnon, R., 2009. Dynamic noise modeling at roundabouts. Appl. Acoust. 70, 761–770.
Covaciu, D., Florea, D., Timar, J., 2015. Estimation of the noise level produced by road traffic in roundabouts. Appl. Acoust. 98, 34–51.
Department of Transport and Welsh Office UK, 1988. Calculation of Road Traffic Noise. Her Majesty’s Stationery Office, London.
Guarnaccia, C., 2013. Advanced tools for traffic noise modelling and prediction. WSEAS Trans. Syst. 12 (2), 121–130.
Iannone, G., Guarnaccia, C., Quartieri, J., 2013. Speed distribution influence in road traffic noise prediction. Environ. Eng. Manage. J. 12 (3), 493–501.
Jacobs, L.J.M., Nijs, L., Willigenburg, J.J.V., 1980. A computer model to predict traffic noise in urban situations under free flow and traffic light conditions. J.
Sound Vib. 72, 523–537.
Jeon, S., Hong, B., 2016. Monte Carlo simulation-based traffic speed forecasting using historical big data. Future Gener Comp Sy 65, 182–195.
Kephalopoulos, S., Paviotti, M., Anfosso-Ledee, F., 2012. Common Noise Assessment Methods in Europe (CNOSSOS-EU) EUR 25379 EN. Publications Office of
the European Union, Luxembourg.
Kroese, D.P., Brereton, T., Taimre, T., Botev, Z.I., 2014. Why the Monte Carlo method is so important today. WIREs Comput. Stat. 6, 386–392.
Li, F., Cai, M., Liu, J.K., Yu, Z., 2011. Dynamic traffic noise simulation at a signalized intersection among buildings. Noise Control Eng. J. 59 (2), 202–210.
Luo, W.L., Cai, M., Li, F., Liu, J.K., 2012. Dynamic modeling of road traffic noise around building in an urban area. Noise Control Eng. J. 60 (4), 353–362.
Makarewicz, R., Kokowski, P., 2007. Prediction of noise changes due to traffic speed control. J. Acoust. Soc. Am. 122 (4), 2074–2081.
Makarewicz, R., Fujimoto, M., Kokowski, P., 1999. A model of interrupted road traffic noise. Appl. Acoust. 57 (2), 129–137.
Road Construction Section of the Federal Ministry for Transport (RCSFMT), 1990. Directives for Anti-Noise Protections along Roads. Ministry for Transport,
Berlin.
Rochat, J.L., Fleming, G.G., 2002. Validation of FHWA’s Traffic Noise ModelÒ (TNM): phase 1. U.S. Department of Transportation, Washington DC.
Roess, R.P., Prassas, E.S., McShane, W.R., 2010. Traffic Engineering. Prentice Hall, p. 133. ISBN: 0136135730.
Steele, C., 2001. A critical review of some traffic noise prediction models. Appl. Acoust. 62 (3), 271–287.
Suzuki, T., Tsukui, K., Oshino, Y., Tachibana, H., 2003. Road traffic noise prediction model around signalized intersections. In: Proceedings of Inter-Noise
2003.
To, W.M., Chan, T.M., 2000. The noise emitted from vehicles at roundabouts. J. Acoust. Soc. Am. 107 (5), 2760–2763.
Waldeer, K.T., 2003. The direct simulation Monte Carlo method applied to a Boltzmann-like vehicular traffic flow model. Comput. Phys. Commun. 156 (1),
1–12.
Yamamoto, K., 2010. Road traffic noise prediction model ‘ASJ RTN-Model 2008’: report of the research committee on road traffic noise. Acoust. Sci. Tech. 31,
1–55.