Functional Pavement Design – Erkens et al.

(Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02924-8

Comparing in-lab and on-site measurement of pavement

acoustic absorption

Filippo Giammaria Praticò & Domenico Vizzari

University Mediterranea of Reggio Calabria, Reggio Calabria, Italy

ABSTRACT:  As is well known, pavement features affect traffic noise. In more detail, rolling
noise depends on generation factors and acoustic absorption. This latter can be assessed
through in-lab (ISO 10534-2) and/or on-site (ISO 13472-1; ISO 13472-2) tests but several
issues arise when comparing measures carried out according to different measurement meth-
ods. Consequently, the objectives and scope of the study described into this paper were to
compare different methods in order to derive relationships to use in practical applications.
The following devices were used: in-lab Kundt tube; on-site Kundt tube; Adrienne device.
Based on measurements, absorption spectra were derived. Analyses and results focused also
on composite indicators and first-pick frequencies and values. Outcomes can benefit both
practitioners and researchers. Future research will address a number of issues arouse during
experiments and analyses and will address a better understanding of phenomena and reasons
which originated the relationships which emerged from the study.

1  INTRODUCTION

As is well known, noise pollution is one of the most important environmental problems in
Europe (see, 2002/49/CE; SMILE workshop 2003; Freitas et al, 2012; Praticò and Anfosso,
2012; Kihlman et al, 2014). Power unit and tyre-road contact are the main sources, and in the
mid-to-high speed range (above 40 km/h–80 km/h) the main contributor to traffic noise is
rolling noise (Sandberg and Ejsmont, 2002). These phenomena generate acoustical pressures
(Weyl-Van Der Poel’s equation) which generate loudness (Attenborough, 1983; Praticò, 2001;
Frías et al, 2011; Jimenez-Espadafor et al., 2011). Tyre/road noise may vary more than 15 dB,
based on tyre and pavement type, and noise reduction at the source can be more cost-effective
than treatments on the buildings or on the propagation path (e.g., noise barriers). Genera-
tion factors (pavement texture, tyre type, etc.) and acoustic absorption govern rolling noise
(Luong et al, 2014). In more detail, pavement composition, volumetric and surface proprie-
ties affect generation, absorption and propagation (Praticò, 2001; Praticò et al, 2010; Praticò
and Moro, 2011; Praticò and Vaiana, 2013; Praticò and Vaiana, 2012; Vazquez et al, 2016).
Based on the above, road agencies require satisfactory level of acoustic absorption, which
is usually quite appreciable in porous asphalt concretes.
Acoustic absorption can be assessed through in-lab (ISO 10534-2) and/or on-site (ISO
13472-1; ISO 13472-2) tests but several issues arise when comparing measures carried out
according to different measurement methods.

2  OBJECTIVES

Based on the above, the objectives and scope of the study described into this paper were to
compare different methods in order to derive relationships to use in practical applications.

1001
Figure 1.  Main tasks.

3  DESIGN OF EXPERIMENTS

Figure 1 illustrates the flow chart of the experimental investigation.

Task 1 focused on on–site investigations according to ISO 13472-1 and ISO 13472-2.
A motorway, located in Southern Italy was investigated. Measurements were carried out
on the hard shoulder lane and on the inside lane, but not on the outside (overtaking) lane.
The surface layer main characteristics are the following: nominal maximum aggregate size:
16 mm; asphalt binder content, percent by total weight of mixture: 4.7%; air voids content:
20%; clogging not noticed. In task 2, after core extraction (from the sections corresponding
to the above on-site measurements), cores were labelled and transported to the laboratory.
There, each core was preliminary cleaned, swept and controlled (for imperfections in the
lower face). Afterwards it underwent the nondestructive tests (in-lab Kundt, ISO 10534-2,
etc.).
Task 3 dealt with data analysis (absorption).
Finally, in task 4 experimental spectra were compared.

4  RESULTS AND OBSERVATIONS

As abovementioned, the absorption coefficient was measured through three different devices:
AD: Adrienne device (ISO 13472-1, see also Londhe et al, 2009); KS: on site Kundt device
(ISO 13472-2, method for reflective surfaces); KL: in-laboratory Kundt device (ISO 10534-2,
see also Wolkesson, 2013). Note that the hypothesis which underlies this analysis is the homo-
geneity of the surface.
Figures 2–7 and Tables 1–7 illustrate main results and analyses.
By referring to the probability density function (PDF, y-axis) of the absorption coefficient
(x-axis) for the different devices, for each frequency, Figure 2 illustrates the main results.
In particular it shows the probability density function of the acoustic absorption coef-
ficient, for a given frequency (e.g., 250  Hz). Frequencies in the range 250–1600  Hz are
considered.
At 250 Hz, note that the in-lab device (KL) provides the lowest value (about 0.1) while the
on-site Kundt tube provides the highest value (about 0.3). Furthermore note that the high-
est PDF corresponds to the KS, which entails, in this case, that the standard deviation is the
1002
Figure  2.  Probability density function of the absorption coefficient for the three devices, for each
frequency. Symbols. PDF: probability density function. AD: Adrienne; KS: on-site Kundt tube; KL:
in-lab Kundt tube.

lowest (highest precision). In contrast, the lowest PDF corresponds to the AD, which implies
the lowest precision. Based on these data, no evidence can be associated towards the best
accuracy (degree of closeness of measurements to true value).
When different frequencies are considered, the following observations can be pointed out:
i) the precision decreases in the range 250–800 Hz. Afterwards, it slightly increases in 800–
1600 Hz; ii) in-lab measures usually yield lower values of absorption; iii) extended surface
method (AD) usually yields higher values of acoustic absorption coefficient.
1003
Figure 3.  Probability density function of the absorption coefficient, for each frequency domain. Sym-
bols. 400–630: frequency range (Hz); 800–1250: frequency range (Hz); PDF: probability density. func-
tion. AD: Adrienne; KS: on-site Kundt tube; KL: in-lab Kundt tube.

Figure  4.  Curve envelopes of the absorption coefficient (maximum and minimum) for the three
devices. Symbols. α: absorption coefficient (dimensionless, defined in 0, 1). Max: envelope curve of the
highest values for the given frequency (i.e., 630 Hz), for the given device (i.e., AD). AD: Adrienne; KS:
on-site Kundt tube; KL: in-lab Kundt tube.

Figure 5.  Average absorption coefficient and lower specification limits. Symbols. AD: Adrienne; KS:
on-site Kundt tube; KL: in lab Kundt tube; α: absorption coefficient.

As is well known, contract specifications often require to merge different frequencies into
an average value (e.g., ANAS, 2010: 400–630; 800–1250; 1600–2500 Hz). To this end, by refer-
ring to the probability density function of the absorption coefficients for the different devices,
for each frequency range, Figure 3 illustrates the main results (the range 1600–2500 Hz is not
considered because KS and KL don’t provide data for these frequencies).
Even when frequency ranges are considered, note that KL usually provides the lowest
values of absorption.

1004
Figure  6.  Absorption coefficient in 400–630  Hz (averages). Symbols. AD: Adrienne; KS: on-site
Kundt tube; KL: in lab Kundt; dotted line: equality line.

Figure  7.  Absorption coefficient in 800–1250  Hz (averages). Symbols. AD: Adrienne; KS: on-site
Kundt tube; KL: in lab Kundt; dotted line: equality line.

Table 1.  Main results for the linear trendlines applied to Figures 6 and 7.

y = mx + q
Devices Range [Hz] m q R2 y = m1 × m1

AD(KL) 400–630 0.4101 0.381 0.3281 1.5966

AD(KS) 400–630 0.6946 0.2707 0.3364 1.6403
KS(KL) 400–630 0.5304 0.1347 0.5517 0.9121
AD(KL) 800–1250 0.3468 0.4167 0.5731 1.1606
AD(KS) 800–1250 0.4151 0.3314 0.2137 0.9581
KS(KL) 800–1250 0.6154 0.2617 0.5464 1.1747

By referring to the highest and lowest values of the absorption spectra (peaks), for each
frequency, for each device, Figure 4 illustrates the main results.
In particular it shows the envelope curves for the three devices. By referring to the maxi-
mum values (Max), note that all the devices yield a maximum envelope curve which has a
relative minimum (or/and an appreciable variation of the first derivative towards zero) at
about 630 Hz. In contrast the three devices yield a relative maximum at about 500 Hz and an
absolute maximum at about 800–1250 Hz. It may be observed that the lower the frequency is
the higher the discrepancy among the three outputs becomes.
For the envelopes which refer to the minimum values (Min), note that local and global
extrema appear quite uncorrelated. For example the ADMin curve presents a local minimum
at about 1250  Hz which appears absolutely unrealistic and which doesn’t have any corre-
spondence with the remaining measurements.
1005
 Table 2.  Main statistics for Adrienne measurements.

Adrienne 250 Hz 315 Hz 400 Hz 500 Hz 630 Hz 800 Hz 1000 Hz 1250 Hz 1600 Hz

Average 0.24 0.31 0.36 0.43 0.58 0.68 0.60 0.45 0.46
St dev 0.09 0.05 0.05 0.14 0.15 0.17 0.19 0.14 0.11
St dev/average 0.39 0.17 0.14 0.32 0.25 0.25 0.32 0.32 0.25
Numb of meas 19 19 19 19 19 19 19 19 19
Min -0.09 0.13 0.26 0.27 0.35 0.31 0.23 0.06 0.28
Max 0.39 0.37 0.50 0.86 0.88 0.90 0.87 0.76 0.69
P (0.5) 0.24 0.31 0.36 0.40 0.57 0.70 0.56 0.47 0.45
Skewness -2.07 -2.03 0.86 2.03 0.36 -0.54 -0.10 -0.51 0.67
Observation n* n* p* p* q* n* q* n* p*
Kurtosis 7.29 5.27 3.24 4.50 -0.44 -0.41 -1.05 2.16 -0.15
Observation Leptokurtic Leptokurtic Leptokurtic Leptokurtic Mesokurtic Mesokurtic Platykurtic Leptokurtic Mesokurtic

Symbols. n*: negative skewed (that is to say, left-skewed, left-tailed, skewed to the left: The left tail is longer); p*: positive skew (i.e., right-skewed, right-tailed, skewed
to the right: The right tail is longer); q*: quasi-symmetric distribution; P(0.5): 50% percentile

1006
Table 3.  Main statistics for on-site Kundt measurements.

On-site Kundt 250 Hz 315 Hz 400 Hz 500 Hz 630 Hz 800 Hz 1000 Hz 1250 Hz 1600 Hz

Average 0.30 0.26 0.25 0.27 0.33 0.47 0.58 0.54 0.41
St dev 0.03 0.03 0.04 0.09 0.13 0.16 0.17 0.20 0.12
St dev/average 0.09 0.11 0.17 0.32 0.39 0.34 0.29 0.36 0.30
Numb of meas 31 31 31 31 31 31 31 31 31
Min 0.24 0.20 0.17 0.16 0.17 0.23 0.27 0.25 0.20
Max 0.34 0.31 0.39 0.62 0.64 0.75 0.88 0.96 0.61
P (0.5) 0.30 0.27 0.24 0.24 0.29 0.44 0.56 0.59 0.44
Skewness -0.86 -0.39 1.11 2.34 0.84 0.40 -0.14 0.09 -0.20
Observation n* q* p* p* p* q* q* q* q*
Kurtosis 0.44 -0.56 2.58 7.27 -0.32 -1.00 -0.80 -1.12 -1.17
Observation Leptokurtic Leptokurtic Leptokurtic Leptokurtic Mesokurtic Mesokurtic Platykurtic Leptokurtic Mesokurtic

Symbols. n*: negative skewed (that is to say, left-skewed, left-tailed, skewed to the left: The left tail is longer); p*: positive skew (i.e., right-skewed, right-tailed, skewed
to the right: The right tail is longer); q*: quasi-symmetric distribution; P(0.5): 50% percentile
 Table 4.  Main statistics for in-lab Kundt measurements.

On-site Kundt 250 Hz 315 Hz 400 Hz 500 Hz 630 Hz 800 Hz 1000 Hz 1250 Hz 1600 Hz

Average 0.08 0.11 0.18 0.28 0.39 0.52 0.46 0.33 0.31
St dev 0.02 0.04 0.09 0.14 0.13 0.19 0.16 0.10 0.10
St dev/average 0.28 0.36 0.50 0.52 0.34 0.37 0.35 0.30 0.34
Numb of meas 19 19 19 19 19 19 19 19 19
Min 0.05 0.06 0.09 0.12 0.16 0.21 0.15 0.20 0.18
Max 0.12 0.22 0.44 0.64 0.63 0.79 0.77 0.61 0.58
P (0.5) 0.07 0.10 0.14 0.22 0.40 0.61 0.48 0.33 0.31
Skewness 0.39 1.08 1.55 1.34 0.04 -0.19 0.04 1.16 0.91
Observation q* p* p* p* q* q* q* p* p*
Kurtosis -1.14 0.45 2.07 1.30 -0.52 -1.37 -0.31 1.99 0.61
Observation Platykurtic Mesokurtic Leptokurtic Leptokurtic Platykurtic Platykurtic Mesokurtic Leptokurtic Leptokurtic

Symbols. n*: negative skewed (that is to say, left-skewed, left-tailed, skewed to the left: The left tail is longer); p*: positive skew (i.e., right-skewed, right-tailed, skewed
to the right: The right tail is longer); q*: quasi-symmetric distribution; P(0.5): 50% percentile

Table 5.  Main statistics for each range of frequency.

1007
Adrienne On-site Kundt In-lab Kundt
Range 400–630 Hz 800–1250 Hz 1600–2500 Hz 400–630 Hz 800–1250 Hz 400–630 Hz 800–1250 Hz
Average 0.46 0.58 0.54 0.28 0.55 0.29 0.42
St dev 0.09 0.10 0.07 0.08 0.12 0.13 0.14
St dev/average 0.19 0.18 0.14 0.31 0.22 0.45 0.33
Numb of meas 19 19 19 31 31 19 19
Min 0.34 0.34 0.36 0.17 0.32 0.10 0.17
Max 0.64 0.74 0.65 0.59 0.74 0.54 0.67
P(0.5) 0.45 0.61 0.54 0.26 0.57 0.25 0.38
Skewness 0.46 -1.00 -0.75 1.84 -0.37 0.56 0.02
Observation q* n* n* p* q* p* q*
Kurtosis -0.61 0.72 0.42 4.62 -0.82 -0.81 -0.79
Observation Platykurtic Leptokurtic Mesokurtic Leptokurtic Platykurtic Platykurtic Platykurtic
Threshold 0.25 0.50 0.25 0.15 (+) 0.43 (+) 0.15 0.30
Ratio 1.84 1.16 2.16 1.87 1.28 1.93 1.4

Symbols. n*: negative skewed (that is to say, left-skewed, left-tailed, skewed to the left: The left tail is longer); p*: positive skew (i.e., right-skewed, right-tailed, skewed
to the right: The right tail is longer); q*: quasi-symmetric distribution; P(0.5): 50% percentile. (+): working hypothesis.
 Table 6.  Conclusions.

AD KS KL

St dev/average 0.14–0.39 0.09–0.39 0.28–0.52

Precision
Accuracy

Figure 5 shows how measurements carried out by the three abovementioned devices match
with lower specification limits (reference: ANAS, 2010).
The same pavement presents slight issues when analysed through the methods AD and
KL, while it would be considered totally compliant in the case KS.
Figures 6 and 7 refer to the comparison among the three devices in the range 400–630,
800–1250 and report linear regressions.
In more detail, Figure 6 refers to the estimate of the acoustic absorption coefficient in the
range 400–630 Hz and highlights how linear regressions (with zero intercept or not) compare
with equality line (dotted line) and data (averages).
Note that even if non-zero intercept linear regressions yield the best results (highest
R-square values), when Kundt-type measurements are considered (i.e., KS-in situ, and KL-in
the laboratory), zero-intercept curve has a slope quite close to one and curve seems to explain
an appreciable percentage of data variations.
This occurrence is very relevant because of the high speed of KS with respect to KL meas-
urements and call for further research and investigation. Importantly, it refers to a domain
in which Adrienne-type measurements can present accuracy and precision issues and draw-
backs (see Figure 3).
Figure 7 refers to the range 800–1250, where human hearing is generally considered to be
far more sensitive to sounds. Note that KL measurements are matched quite well by both
AD and KS. In more detail, KS and AD curves seem to overestimate lower values even if
they explain 50–60% of data variance. Importantly, zero-intercept curves have a slope quite
close to one.
Tables 2–7 summarise the main statistics of data gathered.
Table 1 shows how y-intercepts (q) are quite high if compared to the range of α (0.1). In
contrast, m1 values (slopes when the intercept is assumed zero), are quite encouraging except
than for AD. Note that in the case of zero-intercept, the coefficient of determination resulted
always very low (<0.2).
Tables 5 refers to ranges of frequency.
As for dispersion indicators (see tables 2–5), note that standard deviation of the three indi-
cators give information about how much spread are the data and yield that the three devices
are comparable in terms of precision. Unfortunately, standard deviation is affected by the
magnitude of the data (average), and higher values yield higher standard deviations. In order
to limit this bias effect the coefficient of variation has been derived (see again tables 2–5).
Based on the results obtained it is possible to observe that the in-lab measurements (KL)
result less precise than the remaining two methods, because of the coefficient of variation
in the range 0.3–0.5 (see tables 4 and 5). In contrast, in terms of symmetry (skewness), the
on-site method provides outputs which are, on average, quite symmetric (see table 3) except
that for 500 Hz. In contrast, both the Adrienne and the in-lab method are right-skewed and
usually present a longer right tail., i.e., a few, very high values (see tables 2 and 4). In terms
of kurtosis, the three methods perform similarly, with a certain tendency towards leptokurtic
distributions. Importantly, in table 5 a working hypothesis is formulated about the threshold
value for the on-site measurements carried out through the KS device. This hypothesis origi-
nates from the assumption of having a ratio (in 400–630 and in 800–1250 Hz) which complies
with the ones obtained through the remaining two devices. Under the hypothesis of statistical
significance, it turns out that having the above thresholds (+) would imply, on average, a com-
mon ranking of pavement acoustic absorption, based on one out of the three devices.

1008
5  CONCLUSIONS

The work presented in this paper was carried out by considering a very high amount of data
and further analyses and checks are necessary. The following preliminary conclusions may
be drawn. The rationale which underlies the on-site Adrienne method is that the quantity of
energy reflected by the pavement surface depends on its acoustic absorption. This device per-
forms measurements in 0–4000 Hz, with a coefficient of variation of 0.1–0.4. This method is
time-consuming (about 30 minutes are necessary to perform one test). Its performance seems
to better match KL for high frequencies more than for low frequencies.
The rationale behind on-site Kundt tube is that pressures into the tube are affected by
the sample absorption. It is supposed to work properly only for reflective surfaces (as per
standard specifications). This notwithstanding its correlation with the in-lab device (which
is usually supposed to represent the best method) is relevant and appreciable. Importantly,
zero-intercept linear regression (KS versus KL) works quite properly and provides a straight-
forward tool for practitioners.
In-lab impedance tube is based on the same principle as on-site Kundt tube. Such as the
Adrienne method, it is time consuming (coring, transport, drying, core fixing in the tube).
Issues in fixing the core inside the tube can occur and can generate false response. Based
on previous studies, it seems to provide an outstanding estimate of the real spectrum of
absorption coefficient, as per theoretical models estimates (Stinson, Bereinger, etc.), even if
other validations of this assumption are required. Importantly, it has the worst coefficient
of variation.
Authors are aware that coring process might have affected results (Howard and Doyle,
2013; Praticò et al, 2014). Even if further studies are needed, from this study (see table below)
an overall scenario emerges in which the best accuracy is associated with the worst precision
(KL), the worst accuracy is associated with a questionable precision (AD), and the least
appreciated device (KS) provides reasonable (even if to correct), high-speed measurements.

REFERENCES

Attenborough, K. 1983. Acoustical characteristics of rigid fibrous absorbents and granular materials,
JASA, Vol.73.
Capitolato speciale d’appalto ANAS S.p.A. 2010 – Norme Tecniche.
European Commission. Directive on environmental noise, 2002/49/CE, 25th June 2002, DOCE L 89,
7/18/02.
Freitas, E., Mendonça, C., Santos, J.A., Murteira, C., Ferreira, J.P. 2012. Traffic noise abatement:
How different pavements, vehicle speeds and traffic densities affect annoyance levels, Transportation
Research Part D 17 321–326.
Frías, M., Jiménez-Mateos, J.M., Pfretzschner, J., Olmeda, J., Rodríguez, R.M., Sánchez de Rojas, M.I.
2011. “Development of blended cement mortars with acoustic properties using petroleum coke”,
Construction and Building Materials 25, pp 1086–1092.
Howard, I.L., Doyle, J.D. 2013. Investigating the consistency of asphalt density measurement methods
over a wide range of air voids. Journal of Testing and Evaluation, 42 (3).
Jimenez-Espadafor et al., 2011. Optimal design of acoustic material from tire fluff, Materials and Design
32, pp 3608–3616.
Kihlman, T., Kropp, W., Lang, W. 2014 Quieter cities of the future. Lessening the severe health effects
of traffic noise in cities by emission reductions. International Council of Academies of Engineering
and Technological Sciences, CAETS.
Luong, J., Bueno, M., Vázquez, V. F., Paje S. E. 2014. Ultrathin porous pavement made with high viscos-
ity asphalt rubber binder: A better acoustic absorption? Applied Acoustics 79 117–123.
Niranjan Londhe, Mohan D. Rao, Jason R. Blough. 2009. Application of the ISO 13472-1 in situ tech-
nique for measuring the acoustic absorption coefficient of grass and artificial turf surfaces, Applied
Acoustics, Volume 70, Issue 1. Pages 129–141.
Praticò, F.G. 2001. Roads and loudness: a more comprehensive approach, Journal of Road Materials
and Pavement Design, Volume2, issue 4/2001.

1009
Praticò, F.G., Anfosso-Lédée F. 1994. Trends and Issues in Mitigating Traffic Noise Through Quiet
Pavements, SIIV – 5th International Congress – Sustainability of Road Infrastructures, Procedia –
Social and Behavioral Sciences, 2012. E201994 – issn: 1887–0428.
Praticò, F.G., Vaiana, R., A study on volumetric versus surface properties of wearing courses, Construc-
tion and Building Materials, Volume 38, January 2013, Pages 766–775.
Praticò, F.G., Moro, A., Ammendola, R. 2010. Potential of fire extinguisher powder as a filler in bi-
tuminous mixes, Journal of Hazardous Materials, 173 (1–3), pp. 605–613.
Praticò, F.G., Moro, A. 2011, In-lab and on-site measurements of hot mix asphalt density: Convergence
and divergence hypotheses, Construction and Building Materials, 25 (2) 2011, pp. 1065–1071.
Praticò, F.G., Vaiana, R. 2013, A study on volumetric versus surface properties of wearing courses,
2013, Construction and Building Materials, 38, pp. 766–775.
Praticò, F.G., Vaiana, R. 2012, Improving infrastructure sustainability in suburban and urban areas: Is
porous asphalt the right answer? And how?, 2012, WIT Transactions on the Built Environment.
Pratico, F.G., Vaiana, R., Moro, A. 2014. Dependence of volumetric parameters of hot-mix asphalts on
testing methods. Journal of Materials in Civil Engineering. Volume 26, Issue 1, Pages 45–53.
Sandberg, U., and Ejsmont, J. A. 2002. Tyre / Road noise reference book, Informex, Kisa, Sweden.
SMILE workshop (2003). Guidelines for road traffic noise abatement. Berlin, Germany. <http://www.
smile-europe.org/>.
Vázquez,V. F., Luong, J., Bueno, M., Terán, F., Paje, S. E. 2016, Assessment of an action against envi-
ronmental noise: Acoustic durability of pavement surface with crumb rubber. Science of the Total
Environment 542 (2016) 223–230.
Wolkesson M. 2013. Evaluation of impedance tube methods. A two microphone in-situ method for
road surface and the three microphone transfer function method for porous materials. Department
of Civil and Environmental Engineering, division of Applied Acoustics, Vibroacoustics Group. Chalmers
University of Technology. Goteborg, Sweden 2013.

1010