You are on page 1of 8

Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 177 (2017) 281 – 288

XXI International Polish-Slovak Conference “Machine Modeling and Simulations 2016”

Assessment of uncertainty in urban traffic noise measurements


Andrzej Bąkowski*, Leszek Radziszewski, Zbigniew Skrobacki
Kielce University of Technology, Al. Tysiąclecia Państwa Polskiego 7, 25314 Kielce Poland

Abstract

This paper presents measurement results recorded by noise and traffic volume monitoring station located in Krakowska Street,
Kielce. Poland. The measurements were performed in the period from January to December 2013, twenty-four hours a day and
with a sampling time of 1 minute. The results were subjected to statistical analysis. A type A uncertainty was evaluated. A new
parameter ‫ ݍ݁ܣܮܸܱܥ‬was proposed for variability assessment of the equivalent sound pressure level.
©©2017
2017Published
The Authors. Published
by Elsevier by Elsevier
Ltd. This Ltd.
is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of MMS 2016.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of MMS 2016
Keywords: noise mesurements; urban traffic;

1. Introduction

The decision [1] that devolves the duty of providing and updating noise maps to local governments has again
drawn public attention to environmental noise as a factor contributing to environmental pollution. Short-term and
long-term noise indicators are used. Their values have to be determined through the measurement of sound levels
and evaluation of uncertainty in the measurements. Determining long-term indicators requires special attention, as
they have to be representative of the period of a calendar year [2]. To do this, cities are building stationary
measurement stations recording sound levels and road, rail and air traffic volumes over a year. Kielce has several
such stations, in the centre and on the outskirts. This paper presents an analysis of the measurement results
equivalent sound pressure levels recorded by one of these stations.

* Corresponding author. Tel.: +48-41-34-24-444; fax: +48-41-34-42-997.


E-mail address: abakowski@tu.kielce.pl pl

1877-7058 © 2017 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of MMS 2016
doi:10.1016/j.proeng.2017.02.226
282 Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288

2. Monitoring station: noise measurement

The study uses the results of sound pressure level measurements conducted at the stationary monitoring station
for the sound pressure level and traffic volume, located in Krakowska Street in Kielce. Kielce is a town in the south
part of Poland, located within a moderate climate zone on the edge of the Świętokrzyskie Mountains. During a year
air temperatures change from about -5°C in January to about +17°C in July. Average monthly precipitation is from
34 mm in October to 96 mm in July. Average wind speed is about 10 km/h, with winds predominantly from the west
and south. Snow remains unmelted for about 70 days a year. Krakowska Street, composed of four traffic lanes, is the
main part of the outward route from the centre of Kielce to Kraków, and carries both urban and transit traffic. The
measurements were made in 2013. The database of sound pressure levels was established in the form of computed
values of the equivalent sound pressure. Owing to various technical problems, the database was incomplete and
included data collected over day, evening and night periods from 8 January to 6 December. The database comprised
905 records. The monitoring station includes a sound level meter, a road radar box and a weather station. Sound
level measurements were made with a SVAN 958A, a Class 1 four-channel sound and vibration analyzer with a
frequency bands from 0.5 Hz to 20 kHz, depending on the microphone used. In the case of a 1/2” prepolarized
capacitive microphone by MIKROTECH GEFELL MK 250, with a sensitivity of 50 mV/Pa, containing a SV 12L
preamplifier, this frequency range is from 3.5 Hz to 20 kHz. The temperature operating range is from -100 C to 500
C. The resolution of the RMS detector is 0.1 dB. The measurements were carried out for 24 hours a day. The sound
level was recorded every 1 s (buffer) and the results were reported every 1 minute. The records were used to
calculate the equivalent sound level for three time intervals: from 6:00 to 18:00, from 18:00 to 22:00 and from 22:00
to 6:00. The traffic volume was measured by WAVETRONIX digital radar with an operating frequency of 245
MHz. The weather data were recorded by means of a VAISALA WTX 510 automatic weather station.
The literature [3] indicates that traffic noise data are rarely normally distributed. This finding was confirmed in
[4], in which the authors proposed non-classical statistical methods for assessing noise measurement uncertainty.
The equivalent sound level LAeq,T is defined as:

ͳ ܶ ‫ ܣ ݌‬ሺ‫ݐ‬ሻ ʹ
‫ ݍ݁ܣܮ‬ǡܶ ൌ ͳͲ ή ݈‫ ݃݋‬൤ ‫ Ͳ׬‬ቀ ቁ ݀‫ݐ‬൨ ݀‫ܤ‬ (1)
ܶ ‫Ͳ݌‬

where:
 - overall measurement time,
’ ሺ–ሻ – A-weighted instantaneous sound pressure level, Pa
’Ͳ - reference sound level of 20 μPa. According to ISO 9612 [5], this parameter can be determined in the following
way [6]

ͳ
‫  ݍ݁ܮ‬ൌ ͳͲ ݈‫ ݃݋‬ቀ σ݅ൌ݊
݅ൌͳ ͳͲ
ͲǤͳ‫ ܣܮ‬ǡ݅
ቁ (2)
݊

where
‫ܣܮ‬ǡ݅ - is the A-weighted sound level measured in the time interval ݅. According to this standard, an average sound
level can be determined as an expected value

ഥ ൌ  ͳ σ݅ൌ݊
‫ܮ‬ ݅ൌͳ ‫ܣܮ‬ǡ݅ (3)
݊

The standard deviation of the measurement can then be calculated from

ߪ ൌට
ͳ
σ݅ൌ݊ ത ʹ
݅ൌͳ ൫‫ܣܮ‬ǡ݅ െ  ‫ ܮ‬൯  (4)
݊െͳ

The equation defining the equivalent sound level can be written as an approximation

ഥ ൅ ͲǤͳͳͷߪ ʹ
‫  ݍ݁ܮ‬ൎ  ‫ܮ‬ (5)
Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288 283

The expanded uncertainty can be determined from

ߪʹ ͲǤͲʹ͸ ߪ Ͷ
ܷ ൌ  േට ൅ ‫ ߙݐ‬Ǣ݊െͳ (6)
݊ ݊െͳ

where coefficient ‫ߙݐ‬Ǣ݊െͳ is the quantile of a t-Student distribution at the confidence level α. The most commonly
accepted value of α is 0.05. According to the authors of [6], relationships (3) to (6) can potentially be used if the null
hypothesis about normal distribution of unimodal measured sound levels ‫ܣܮ‬ǡ݅ and the assumption about independent
variables, adequately large number of data, uncorrelated data, and small standard deviation can be made. In the case
when the skewness and kurtosis are near zero and three, it is possible, based on (5), to determine the equivalent
sound level at an acceptable approximation. The standard deviation value should also be low.
In the studies reported so far, long-term indicators covering the whole calendar years were most commonly
analyzed. In this paper, the authors discuss the results from forty-four week measurements conducted in 2013, split
into seven days of the week and three time intervals during the day-evening-night period. The paper presents the
results of statistical analysis of the equivalent sound level measurements performed with the use of commonly
available R software. The aim of this analysis is to determine days of the week and times of day for traffic in a
medium-sized town (Kielce) when the requirements of ISO [7] are satisfied and when, instead of the equivalent
sound level determined exactly to equation 2, its approximate value computed as the arithmetic mean can be used.

3. Measurement results

Figure 1 compares the results of the equivalent sound pressure measurement (LAeq) determined for each day from 8
January to 9 December divided into three time intervals during a 24-hour period, i.e., from 6:00 to 18:00 – denoted
day, from 18:00 to 22:00 – denoted evening and from 22:00 to 6:00 – denoted night.

Fig.1 Equivalent sound pressure levels for each day from 8 January to 6 December 2013 divided into three time intervals during a 24-hour period

Table 1 details the minimum values, medians and maximum values of LAeq for the overall period of 24 hours and
for three time intervals within this period. The results of Lilliefors and Shapiro – Wilk tests allow rejecting the null
hypothesis about the normal distribution of the measurement data because the calculated limit significance levels, p-
values, are lower than 0.05 – the value adopted as the required level of significance. All investigations of the null
hypothesis discussed further in this paper were performed for this level of significance. In doubtful cases, to check
whether the distribution is normal, the Jarque-Bera test was performed and skewness and kurtosis were calculated.
In the case when the measurement data distribution is approaching a normal distribution, skewness and kurtosis
should be near zero and three, respectively. Since the values of LAeq do not have a normal distribution, measurement
uncertainty determined with the use of the classical ISO method can be seriously vitiated. For this reason, the
authors decided to split the LAeq results for the entire year into seven groups according to the day of the week on
which the measurements were made.
284 Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288

Table 1 Equivalent sound pressure levels determined for each day from 8 January to 6
December 2013 and divided into three time intervals during a 24-hour period
Time of day Minimum Median Maximum Lilliefors test Shapiro Wilk test
dB dB dB
24h 59.12 68.56 72.72 0 0
night 59.12 64.78 67.69 0 0
day 66.64 70.42 72.72 0 0
evening 62.19 68.79 71.69 0.03 0

Each of those groups was divided into three time intervals, i.e., day, evening and night, as shown in Fig. 2. Figure 2
shows the values of LAeq collected on Mondays. Analysis of the box plots in Fig. 2a shows that some data of those
recorded at night can be regarded as atypical. Note that atypical data are those that have values higher than the sum
 െ
of the third quartile and three quartile deviations ( ͵ ൅ ͵ ͵ ͳ ) and lower than the difference between the first
ʹ
 െ
quartile and three quartile deviations (ͳ െ ͵ ͵ ͳ ). The two lines above and below the box represent the limit
ʹ
values for the data regarded as typical (in the case of a sample with data which may be regarded as atypical) or the
locations of the maximum and minimum. The authors of this paper did not remove the atypical data from the
samples analyzed

Fig. 2 The equivalent sound pressure level on Mondays a) box plots, b) split into night, day and evening periods

The minimum values, medians and maximum values of LAeq for the three time intervals are shown in Table 2. For
the data recorded on Mondays, the results of the Lilliefors and Shapiro- Wilk normality tests allow rejecting the null
hypothesis about normal distribution of the measurement data LAeq for the time intervals denoted as night and day.
The mean value of LAeq for the evening period was 68.72 dB, which considerably exceeds the accepted value. The
calculated uncertainty was 0.142 dB. To evaluate the variability of LAeq, the coefficient of variation was used,
determined from
ߪ ‫ݍ݁ܣܮ‬
‫ ݍ݁ܣܮܸܱܥ‬ൌ  ‫ͲͲͳ כ‬Ψ (7)
‫ݍ݁ܣܮ‬

The coefficient of variation for the time interval under analysis for Monday was 1.34%.
For the data recorded on Tuesdays (Table 2), the results of the Lilliefors and Shapiro-Wilk normality tests allow
rejecting the null hypothesis about normal distribution of the measurement data LAeq for the time intervals denoted
night and day. No evidence was found to reject the null hypothesis for the interval denoted as evening. The mean
value of LAeq for the evening period was 68.74 dB, which considerably exceeds the accepted value. The calculated
uncertainty was 0.139 dB. The ‫ “‡ܸܱܥ‬for the time interval analyzed for Tuesdays was 1.32%.
For the data recorded on Wednesdays (Table 2), the results of the Lilliefors and Shapiro-Wilk normality tests do
not allow rejecting the null hypothesis about normal distribution of the measurement data L Aeq for the time intervals
denoted as evening. The mean value of LAeq for the evening period was 68.91 dB, which considerably exceeds the
Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288 285

accepted value. The calculated uncertainty was 0.140 dB. The ‫ “‡ܸܱܥ‬for this time interval was 1.34%.

Table 2 Summarized values of the equivalent sound pressure levels determined for each day of the week between 8 January and 6 December
2013, divided into night, day and evening periods (L – Lilliefors, SW - Shapiro – Wilk, J-B - Jarque-Bera)
Time Min Median Mean Max SD COV L- test SW- Ua J-B- test sk kur
interval dB dB dB dB dB % test dB
Monday
24h 61.42 68.41 72.49 0.00 0.00
night 61.42 64.73 66.21 0.04 0.01
day 67.91 70.68 70.61 72.49 0.94 1.33 0.31 0.08 0.144 .13 .53 4.09
evening 66.91 68.52 68.72 71.05 0.92 1.34 0.19 0.64 0.142 .49. .42 2.69
Tuesday
24h 62.86 68.60 72.61 0.00 0.00
night 62.86 64.52 64.73 67.69 0.95 1.47 0.27 0.22 0.147 .07 .74 3.94
day 68.18 70.77 72.61 0.02 0.008
evening 66.43 68.66 68.74 70.81 0.91 1.32 0.85 0.98 0.139 .93 -.13 3.00
Wednesday
24h 60.18 68.66 71.77 0.00 0.00
night 60.18 64.72 66.67 1.04 1.61 0.08 0.00 0.159
day 68.16 70.62 70.59 71.77 0.71 1.01 0.65 0.05 0.107 .0 -.92 4.68
evening 66.86 68.70 68.91 71.04 0.93 1.34 0.23 0.26 0.140 .51 .42 2.82
Thursday
24h 63.29 68.94 72.72 0.00 0.00
night 63.29 64.93 64.87 66.28 0.70 1.08 0.97 0.96 0.106 .76 -.09 2.49
day 67.30 70.74 72.72 0.02 0.00
evening 67.61 69.16 69.18 71.09 0.83 1.20 0.83 0.54 0.125 .61 .29 2.56
Friday
24h 63.44 69.28 72.24 0.00 0.00
night 63.44 65.11 65.05 66.94 0.70 1.08 0.74 0.97 0.105 .86 .17 3.24
day 68.00 70.51 70.65 72.24 0.79 1.12 0.53 0.15 0.121 .04 -.48 4.60
evening 66.24 69.30 69.43 71.69 0.92 1.32 0.10 0.01 0.140 .0 -.29 5.56
Saturday
24h 62.08 68.49 71.27 0.00 0.00
night 62.08 65.17 67.08 0.002 0.00
day 68.62 69.65 69.67 71.27 0.60 0.86 0.94 0.65 0.091 .63 .35 2.88
evening 62.19 68.49 70.25 0.001 0.00
Sunday
24h 59.12 68.07 70.13 0 0
night 59.12 64.10 63.76 65.75 1.26 1.97 0.07 0.00 0.192 .0 -1.51 6.31
day 66.64 68.61 68.49 70.13 0.76 1.11 0.68 0.64 0.115 .53 -.42 3.02
evening 66.99 68.47 68.55 69.97 0.72 1.05 0.86 0.68 0.109 .83 -.11 2.60

For the data recorded on Thursdays (Table 2), the results of the Lilliefors and Shapiro-Wilk normality tests allow
rejecting the null hypothesis about normal distribution of the measurement data LAeq for the time intervals denoted as
day. Unlike for other days, there is no evidence to reject the null hypothesis for the night and evening periods. The
mean value of LAeq for the night period was 64.87 dB at the measurement uncertainty of 0.106 dB, and 69.18 dB for
the evening period at uncertainty of 0.125 dB, which considerably exceeds the accepted value. The ‫ ݍ݁ܣܮܸܱܥ‬was
1.20% for a day and 1.08% for a night.
Analysis of the data recorded on Fridays (Table 2) allows rejecting the null hypothesis about normal distribution
of the measurement data LAeq for the time intervals denoted as day and evening. The mean value of LAeq for the night
period was 65.05 dB and exceeded the acceptable values considerably. The calculated uncertainty was 0.105 dB.
The ‫ “‡ܸܱܥ‬for the time interval denoted as night for Fridays was 1.08%.
The values of LAeq obtained on Saturdays are shown in Fig. 3.
286 Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288

Fig.3 Sound pressure levels on Saturdays a) box plots, b) split into the night, day and evening periods

Analysis of the data recorded on Saturdays (Table 2, Fig. 3.) allows rejecting the null hypothesis about normal
distribution of the measurement data LAeq for the time intervals denoted as night and evening. There is no evidence to
reject the null hypothesis for the time interval denoted as day. The mean of LAeq for the day period was 69.67 dB,
which exceeds the normalized value considerably. The calculated uncertainty was 0.091 dB. The ‫ “‡ܸܱܥ‬for the
day period on Saturday was 0.86%.

Fig.4 Sound pressure levels on Sundays a) box plots, b) split into the night, day and evening periods

Analysis of the data recorded on Sundays (Table 2, Fig. 4) does not allow rejecting the null hypothesis about
normal distribution of the measurement data LAeq for the time intervals denoted as day. Mean LAeq for the time
interval denoted as day was 68.49 dB, which exceeds the normalized value considerably. The calculated uncertainty
of measurements was 0.115 dB. The ‫ “‡ܸܱܥ‬for this time interval on Sunday was 1.11%. Figure 5 shows the box
plots of the LAeq parameter for particular days of the week divided into night, day and evening periods and represents
a graphical illustration of the results for the equivalent sound level LAeq detailed in Table 2.
Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288 287

Fig. 5 Summarized box plots for the equivalent sound pressure versus day of the week, split into a) night, b) day, c) evening

The LAeq values for the period between 22.00 and 6.00 increase slightly on Thursday, Friday and Saturday and then
decrease on Sunday. The LAeq values for the period between 6.00 and 18.00 fluctuate slightly from Monday to Friday
and then decrease on Saturday and Sunday. The LAeq values for the period between 18.00 and 22.00 increase slightly
on Thursday and Friday to decrease on Saturday and Sunday.

4. Conclusions

The studies of changes in LAeq during a year confirm that the results do not have normal distributions in the majority
of the cases analyzed. This may contribute to imprecise determination of the uncertainty of measurements if ISO
recommendations are followed. For this reason, normality tests should be performed prior to calculation of the
uncertainty. If there is sufficient evidence to reject the null hypothesis, non-classical statistical methods should be
applied. The studies also revealed that in the shorter time intervals chosen in compliance with standards in force
there might be no ground for rejecting the null hypothesis about a normal distribution of LAeq. In the case of data
recorded in Krakowska Street, this problem occurred most frequently for LAeq determined during the day or in the
evening. The calculated uncertainty of measurements ranges from 0.091 dB to 0.142 dB. The ‫ ݍ݁ܣܮܸܱܥ‬values range
from 0.86% to 1.34%.

References

[1] Directive 2002/49/Ec of the European Parliament and of the Council of 25 June 2002 relating to the assessment and management of
environmental Noise, Official Journal of the European Communities – 18.07.2002]
[2] A. Jagniatinskis, B. Fiks, O.Zaporozhets, Annual Assessment of Transport Noise Using Representative Time Measurements, Procedia
288 Andrzej Bąkowski et al. / Procedia Engineering 177 (2017) 281 – 288

Engineering 134 (2016) 301 – 308


[3] T.Wszołek,M. Kłaczyński, Effect of traffic noise statistical distribution on Laeq,T measurement uncertainty ARCH. ACOUST. 31, 4
Supplement (2006), 311–318
[4] W. Batko,B. Stępień, Type A Standard Uncertainty of Long-Term Noise Indicators, ARCH. ACOUST. ,Vol. 39, No. 1, (2014), 25–36 , DOI:
10.2478/aoa-2014-000
[5] ISO 9612 (2009), Acoustics – Determination of occupational noise exposure – Engineering method
[6] B. Przysucha, W. Batko, A. Szeląg, Analysis of the Accuracy of Uncertainty Noise Measurement, ARCH. ACOUST. Vol. 40, No. 2 (2015),
183–189, DOI: 10.1515/aoa-2015-0020
[7] ISO 1996-2:2007 Acoustics – Description, measurement and assessment of environmental noise – Part 2: Determination of environ-mental
noise levels

You might also like