Temperature influence on tire/road noise measurements: recently
collected data and discussion of various issues related to standard
Grolimund + Partner AG – environmental engineering, Thunstr. 101a, CH-3006 Bern, Switzerland
Grolimund + Partner AG – environmental engineering, Thunstr. 101a, CH-3006 Bern, Switzerland
Swedish National Road and Transport Research Institute (VTI), SE-58195 Linköping, Sweden
Air, road, and tire temperatures substantially affect tire/road noise emission. For measuring
purposes, one would like to normalize measurements to a reference temperature by means of a
reliable correction procedure. Current studies show that temperature effects remain an important
source of uncertainty in tire/road noise measurements and tire testing, even after applying the
correction terms provided in the various standards. This seems to be the case for the measurement
methods used in OBSI, CPX, SPB, and various regulations or directives based on ECE R117. This
paper examines a new dataset consisting of 7.5 million temperature measurements aimed at
contributing to a better understanding of temperature effects and the ways they relate to air, road,
and tire temperatures. It is assumed that tire temperatures are the most relevant for noise
corrections; therefore, special studies are made for how tire temperatures relate to air and road
(test surface) temperatures. A profound analysis is provided on how these relationships vary over
different day times, seasons, and climatic regions. Based on this analysis, the authors provide
suggestions for improvement of temperature normalization in current tire/road noise and tire
testing standards. Special considerations are devoted to measurements on test tracks having ISO
10844 reference surfaces.
1. INTRODUCTION AND PURPOSE
Air, road, and tire temperatures substantially affect tire/road noise emission. It has appeared that
temperature effects are an important source of uncertainty in tire/road noise measurements, whether
the object is the pavement or the tire. For measuring purposes, one would like to normalize
measurements to a reference temperature by means of a reliable correction procedure. Many
international regulations have for about two decades used a procedure specified in  which was
determined in the 1990:s based on a minimum of previous research. To provide a more precise and
more useful temperature correction, ISO recently has developed two procedures for this purpose;
one specialized for the use of the reference tires for CPX measurements , the other for more
general cases, such as SPB measurements . Nevertheless, even after applying the correction
terms provided in the standards, uncertainty due to the temperature effects are significant and need
to be reduced. To this end, more knowledge must be gained about the temperature-influencing
processes, the relation between various relevant temperatures (air, road, tire) and the influence on
The purpose of this paper is to report research results based on a huge amount of noise- and
temperature-related data in Swiss CPX measurements, in cooperation with the project leader for
development of the ISO Technical Specifications mentioned above (Sandberg). The results shed
light on the relation between the temperatures of ambient air, road surface and tire surface, about
their variation over day-night times and seasons, and also about the influence on tire/road noise
frequency spectra; the latter measured with the CPX method,  and .
2. UNDERLYING ASSUMPTIONS
Tire/road noise is generated in the tire. It is assumed here that noise is not emitted from the road
surface; at least not to a substantial amount. It follows that the noise emission must depend on the
temperatures in or on the tire. Therefore, the most relevant temperature should be the tire tempera-
ture. However, both air and road temperatures affect tire temperature. The noise propagation may
also influence the noise received by a microphone or an observer; hence air temperature may be of
some importance also in this respect. The road temperature cannot be totally excluded from
influence on the pavement, since the pavement stiffness, which may affect tire/road noise genera-
tion, is influenced by temperature. The latter is, however, a negligible effect for asphalt and
concrete pavements, unless the pavement is extremely soft due to its content of substantial
proportions of rubber as in so-called poroelastic pavements. The effect of air temperature on the
propagation of noise is negligible for frequencies below about 10 kHz and if one is within about 10
m from the source, but may be significant at distances exceeding 10-20 m. For the measurements
which we consider in this paper (CPX = Close ProXimity, OBSI = On-Board Sound Intensity, SPB
= Statistical Pass-By, or coast-by according to ECE R 117), we are within 7.5 m from the source
and frequencies above 10 kHz are unimportant and thus we can neglect the air temperature effect on
What remains is thus the tire temperature. Unfortunately, things are much more complicated, since
tire temperatures vary substantially from location to location on the tire surface, also from location
to location within the tire structure. These variations also change with time: they change when the
tire is in operation due to energy losses and they change due to the air and road surface temperatures
and all this with different time histories in various parts of the tire. The total picture is extremely
complicated. In this project we have had to use only the tire surface temperature continuously
measured with an IR thermometer in the middle of the tire tread. Since it is known that tire/road
noise is essentially generated in the tread, including its shoulders, and is emitted mainly from tread
surface, it is logical that this is an important test location, if one has to choose one.
Figure 1 – The various temperatures involved and its influencing factors in a schematic overview.
tire temperature (Ttire)
depends on: Tair, Troad, friction,
air temperature (Tair; )
depends on: climate/season/weather, surface radiation
(~90% of contact area tire)
road temperature (Troad)
depends on: solar radiation surface
albedo, heat storage capacity, wind
(~10% of contact area tire)
conduction friction (adhesion and hysteresis)
most relevant temperature when
correcting for temperature effects
As mentioned above, we know that ambient air, road, and tire temperatures substantially affect
tire/road noise emission. This is illustrated in Figure 1. The tire temperature at rolling without any
load is affected by air temperature by convection (and maybe some ventilation too) and road
temperature by conduction. Convection operates on the entire tire (tread and sidewalls). The
conduction is from the road surface to the tire tread, which in the case of no load is close to zero,
and thus the tire should assume the same temperature as the ambient air.
When the tire is loaded, things change dramatically: the tire is deflected and gets a significant
contact surface with the road surface. Although, generally, conduction is more effective for trans-
ferring heat than convection, the tire/road contact area is just a fraction of the overall tire area which
is subject to convection (less than 10 %). Thus, it is assumed here that ambient air temperature is
more related to tire temperature than the road surface temperature is. Whether this assumption is
reasonable is treated later in the paper.
The load creates a deflection in the tire, a deflection which causes energy losses due to hysteresis
effects in the rubber compound and its interaction with the structure (steel or textile). These losses
are known to be most significant in the tire shoulders. The energy losses are converted to heat,
which heats up the tire, more or less depending on the location in or on the tire, and sooner or later,
depending on the time after load application and the thermal capacities.
The tire, when in operation, will thus be a heat source and will soon be warmer than the ambient air.
It may not necessarily be hotter than the road surface at all times, since black asphalt surfaces may
in some weather conditions reach temperatures of up to 70 oC. When the tire is warmer than the
ambient air, it will lose heat by convection but also by some radiation to the surroundings. The
radiation to the tire (from the sun) is less significant as the tires are usually shaded by a vehicle
body, but radiation from the road surface to the tire may be more significant if the road surface is
really hot. However, a black body, as asphalt, will not radiate so much, so it is assumed here that we
can neglect the radiation effect and that conduction in the contact patch (possibly also convection)
is much more important.
The temperatures and the balances in heat transfers between tire, air and road surface are worth
much more discussion than provided here and should be closely studied in the future.
How is noise affected by the temperature? The tire rubber gets softer with increasing temperatures,
causing most significantly changes in material behavior, change in stiffness, change in contact
patch, and change in air pressure. This influences the generation mechanisms; potentially it may be
different influences for the different mechanisms. For example, it was reported that tire P1 (in
ISO/TS 11819-3) became softer by 2.5 units Shore A per 10 °C in temperature rise .
Nevertheless, one shall not mix this temperature-caused effect with the effect caused by higher
stiffness caused by mechanical and chemical ageing of rubber, which is dealt with in for example
3. TEMPERATURE CORRECTION APPROACHES IN INTERNATIONAL STANDARDS
A way to express the temperature correction (which is the opposite to the considered temperature
influence on noise) is the expression in Equation 1, which uses a linear correction to normalize the
measured noise levels at temperature (T) to a reference temperature (Tref). All noise-related
temperature corrections schemes have adopted a reference temperature of 20 oC. Equation 1 reads
where is the noise level correction for a measurement temperature (T) and for tire (t), is the
temperature coefficient for tire t (in dB/°C). T and can be either air, road surface or tire
temperature depending on the procedure considered. Note that when the coefficient is negative,
the correction () will be positive in Eq. 1, when T is higher than Tref.
The first published procedure was the one on which most international and national regulations are
based, presently expressed in Regulation 117 of the United Nations Economic Commission for
Europe (ECE), which has been used for coast-by tire/road noise measurements since these were first
standardized some 20 years ago . Note that even though it is an ECE regulation, many nations
outside Europe have adopted an essentially similar method. Refer to Table 1 for more information.
It was never publicly reported how this correction was developed but it would have been suggested
by the tire industry. In a later ISO version of the coast-by procedure developed by the tire industry,
ISO 13325, the ECE R117 corrections were copied.
ISO recently has developed two Technical Specifications for temperature correction of noise levels;
one specialized for the use of the two reference tires for CPX measurements , the other for more
general cases, such as SPB measurements . The temperature corrections depend on speed, tire
size and pavement type; see Table 1. The correction procedures provided in these standards were
developed essentially based on a literature review from 2014 , summarizing the empirical
research on temperature effects from 17 different studies. New research on the variation of
temperature effects with speed , as well as a study with controlled measurements on a laboratory
drum, involving a considerably large selection of different tires, provided a valuable basis for the
development of the correction scheme.
These ISO specifications, where the purpose is to measure the noise properties of pavements (but
13471-2 also can be used to measure noise properties of tires), have been issued as first attempts to
reduce the measurement uncertainty due to the temperature influence. It is recognized that they will
have to be improved when more knowledge is gained about the relation between noise and
temperature. Table 1 summarizes how different standards and procedures handle temperature
corrections. Note that all of them are applied equally over the entire frequency spectrum.
When OBSI measurements are used , the temperature coefficient is 0.072 for all pavements
(transferred from oF to oC). However, this is close to an average of what ISO/TS 13471-1 specifies,
and as the latter is assumed to take more details in the conditions into consideration, the ISO
Technical Specification could equally well be used for OBSI as for CPX measurements, but the
AASHTO procedure still specifies its own coefficient.
Table 1 – The temperature coefficients for noise in different standards and other published procedures. Note
that the corrections to noise levels will have the opposite sign to the coefficients.
ECE R117 and
-0.06 at 5-20 + -0.03 at 20-50 oC
-0.02 at 5-50 oC
P1 & H1
-0.10 for C1 & C2, -0.06 for C3
-0.05 for C1 & C2, -0.04 for C3
-0.07 for C1 & C2, -0.06 for C3
AASHTO T 360:
-0.072 at 4-38 oC
Symbols and abbreviations: v = vehicle speed in km/h. P1 and H1 are the reference tires defined in ISO/TS 11819-3,
where P1 is the ASTM 16” reference tire. C1 = car tires, C2 = van tires, C3 = heavy vehicle tires.
In addition, the SAE (Soc. of Automotive Engineers) has a specification SAE J57 in which coast-by
measurements are made and where the temperature correction is applied by measurements at one
temperature below 20 oC and one above 20 oC (the reference temperature) and then the noise level
at the reference 20 oC is determined by interpolation. The temperatures required to interpolate to the
reference temperature may seriously limit the time and season when measurements can be done.
4. MEASUREMENT METHODS APPLIED IN THE TWO STUDIES IN THIS PAPER
4.1. Measurements on temperature effects (measurements of temperatures and noise)
This study provides a further evaluation of data acquired during an extensive measurement
campaign on temperature effects carried out with the CPX method ( and ) in Switzerland in
2011 (see  for more details). Figure 2 shows a picture of the measurement system and the
Figure 2 – CPX measurement trailer from G+P in Switzerland (left) with reference tires
P1 (ASTM SRTT C1 “16) at the left and H1 (Avon AV4) on the right
Noise levels were recorded at the “mandatory” microphone positions in 200 mm distance from the
inner tire sidewalls (height = 100 mm) to the front and the back of the tire-road contact area.
Variation of speed was kept to a minimum using automatic speed control (standard deviation <1.5
km/h). At the time of measurements, the reference tires P1 and H1 had been in use for about six
months and were run-in over a distance of at least 1000 km. For all tires, the manufacturing date
was less than two years prior to the measurement survey. Rubber hardness was within the limits
specified by : 66 Shore A for the P1 tire and 68 Shore A for the H1 tire. The tires were inflated to
a pressure of 200 kPa (measured in a climatic chamber at 20 °C).
A total of 9 measurement series were carried out in Switzerland on clear days in late summer 2010.
At this time of the year, variation of temperature over the course of a day is substantial.
Measurement series were started at sunrise (at around 06:00) and were continuously repeated until
the maximum day temperature was reached (at around 17:00). Air temperature ranged between
10 °C and 30 °C. Measurement runs were performed with sets of reference tires and speeds of 80
km/h and 50 km/h by turn. The tires were warmed-up as follows: (A) At the beginning of a
measurement series, the tires were brought to operating temperature by driving for at least 15
minutes. (B) Prior to each measurement run, the tires were brought to a stable temperature state by
driving a constant distance of at least 2 km. A schematic view of the measurement set-up is given in
Figure 3. In each of the 9 measurement series, several adjacent road surfaces were measured.
4.2. Study of relations between temperatures
In addition to the temperature effect dataset described in section 4.1, an extensive new dataset
comprising 7.5 million temperature measurements is introduced to investigate the relationships
between the different temperatures. The data was collected from all CPX measurement runs
undertaken by G+P within the scope of Swiss projects between 2010 and 2020. Figure 4 illustrates
how the different temperatures were measured.
Figure 3 – Schematic overview of measurement set-up.
Warming-up (at least 2 km)
Section of measurements
- Tires: P1 and H1
- Speeds: 80 km/h, 50 km/h
road surface α
road surface β
road surface γ
Figure 4 – Measurements of ambient air temperature (blue), road surface temperature (green) and tire tread
temperature (orange) on-board of the CPX-measurement system by G+P.
All temperatures – ambient air temperature (Tair), road surface temperature (Troad) and tire tread
temperature (Ttire) – were continuously measured and averaged for every 20 m segment (the
standard evaluation length for CPX measurements in ). These measurements are stored, together
with tire/road noise measurements and pavement data (type and age) in a geo database.
5. DATABASE COLLECTED IN SWISS PROJECTS
5.1. Temperature influence on noise (measured in study 2010)
In a total of 9 measurement series carried out in Switzerland on clear days in the late summer of
2010, 32 road surfaces were assessed, yielding a total of 124 tire/pavement/speed combinations (see
section 4.1 for more information on the measurement system and set-up). An overview of the 2010
dataset is given in Table 2.
Table 2 – Overview of 2010 dataset on temperature effects
late autumn 2010, full exposure to solar radiation (road surface), wind speed < 5 m/s
automatic cruise control, 50 km/h, 80 km/h
- Data set:
Repeated measurements of tire/road noise and temperature for 124
tire/pavement/speed combinations (reduced to 91; see the text)
- Temperature range:
Tair: 7 to 34 °C, Troad: 8 to 48 °C, Ttire: 15 to 49 °C
- Pavement types:
12 types: DAC 10 (n=4), DAC 11 (n=8), DAC 16 (n=1), ACMR 8/11 (n=1), ACMR 4/8
(n=1), HRA 11/16 (n=1), MA 11 (n=2), MA 16 (n=1), SMA 11 (n=4), CC (n=5), PAC 11
(n=8), Thin Layer 2/4 (n=3)
39 different pavements: homogeneous, undamaged, minimum length of 200 m, age
range: 3 months to 15 years, void range: 2 to 20 %
standard deviation of speed < 1.5 km/h
- Acoustic data:
sound pressure levels for third-octave bands between 315 and 5000 Hz
Calculated for each microphone position (front/back of contact patch) and third-octave-
band, for each tire/pavement/speed combination and temperature (Tair, Troad, Ttire)
- Descriptors of data
coefficient of determination, standard error of regression model, standard error of
regression slope, 95 % confidence intervals
GPS antenna for georeferencing
2 microphones each side
IR sensor road surface temperature Troad
(continuous meas. every 125 ms,
average over area with diameter ~10 cm)
IR sensor tire tread temperatureTtire
(continuous meas. every 125 ms,
average over area with diameter ~10 cm)
air temperature sensor Tair
(continuous, every 30 s, time-interpolation,
height above surface: 150 cm) DAQ modul
PC with measurment software
Measurements at constant
speeds of 50 and 80 km/h
(undertaken from 2010 to 2020)
Strong linear relationships between noise levels and the three temperatures were found at all speeds,
for all pavement types and both tires (P1 and H1). On a few road sections, however, weak correla-
tions were determined, especially for low and mid frequencies. This was due to a parasitic
phenomenon caused by oscillation of the measurement trailer due to small unevenness of the road
surface. For a robust determination of temperature effects on tire/road noise, it is of great
importance to exclude such contaminated data from the analysis. Therefore, only noise-temperature
slopes were considered which met the criteria for non-contaminated measurement data (standard
error < 0.2 dB(A) or coefficient of determination R² > 0.9, see also ). The number of
tire/pavement/speed combinations was thus reduced from 124 to 91 non-contaminated
5.2. New dataset comprising 7.5 million temperature measurements
Our study is supplemented by a new dataset with millions of temperature measurements collected
during Swiss projects between 2010 and 2020, covering a total measurement distance of around
50 000 km. To reduce the degree of complexity to a minimum, the focus in this paper is laid on data
collected at the reference speed 50 km/h, covering a total distance of 30 951 km and a share of 62 %
of all measurements. An overview on the dataset is given in Table 3.
Table 3 – Overview of a new extensive dataset with temperature measurements
Years 2010 to 2020, with the bulk of measurements undertaken between March and
November; in all parts of Switzerland (see map), all measurements undertaken by G+P
automatic cruise control, 50 km/h, 80 km/h (but results at 80 not included in this paper)
- No. of data points:
7.5 million temperature measurements (2.5 million measurements per Tair, Troad, Ttire)
- Area coverage:
50 000 km (of which 30 951 km at 50 km/h)
- Temperature range:
air: 0 to 34 °C, road surface: 0 to 54 °C, tire: 6 to 47 °C
- Road surf. types:
Virtually all pavement types used in Switzerland:
- dense asphalts, e.g. DAC, SMA, MA, SD (780 000 data points, ~15 500 km)
- porous /semi-dense asphalts, e.g. SDA, PAC (675 000 data point ~12 500 km)
- cement concrete (9 000 data points ~35 km)
Between temperature Tair, Troad, separate analysis for daytime (with solar radiation) and
nighttime (without solar radiation)
- Descriptors of data
coefficient of determination, standard error of regression model, standard error of
regression slope, 95 % confidence intervals
A map with the geographic distribution of the temperature measurements collected during Swiss
projects in 2010-2020 is provided in Figure 5.
Figure 5 – Geographic distribution of temperature measurements during Swiss projects from 2010-2020
(measurements at a speed of 50 km/h only).
As Figure 5 shows, the temperature measurements cover all regions (Cantons) in Switzerland. The
bulk of measurements, however, originate from densely populated areas in the Swiss plateau
(Mittelland) and Western Switzerland in a 300 km long strip between Geneva and Zurich. A
moderate continental climate is typical for this area. The cumulated number of temperature
measurements collected at the reference speed 50 km/h amounts to a total of 4 642 680 data points.
The temperature data collected at the reference speed 80 km/h (roughly one third of the total data)
will be presented in a forthcoming paper with extended analyses but is expected to differ
significantly from the ones at 50 km/h only with regard to tires being somewhat warmer and air
providing more convection.
6. PROCESSING OF RESULTS AND DISCUSSION
6.1. How well do Tair, Troad and Ttire explain temperature effects?
A dataset from a study in 2010 comprising temperature effect measurements for 124 tire/pave-
ment/speed combinations was further analyzed. This section examines the explanatory power of the
temperatures of ambient air (Tair), road surface (Troad) and tire tread surface (Ttire) with regard to
temperature effects on tire/road noise frequency spectra measured with the CPX method. The left
graph in Figure 6 illustrates the strength of relationship between temperature and tire/road noise
levels, while the graph on the right shows the robustness of the resulting linear regression models
for different third-octave bands: the standard error of the slope (noise vs temperature) gives valu-
able information on the expected error when applying the correction factors provided in section 6.2.
Figure 6 – Average R2 between temperatures and tire/road noise spectra (left) and error of the slope for the
linear regression models (average over 28 tire/pavement combinations at reference speed 50 km/h)
The left graph in Figure 6 suggests that all temperatures (Tair, Troad and Ttire) explain temperature
effects on the tire/road noise equally well over the noise spectra. This is not an unexpected result,
because the measurements for this temperature effect dataset were carried out under homogeneous
conditions (full exposure of pavement to solar radiation; measurements undertaken in same climatic
region and season). To investigate and compare the performance of Tair, Troad and Ttire in predicting
(i.e., correcting for) temperature effects, such an analysis should be undertaken with data including
different weather conditions and seasons.
Figure 6 reveal very strong relationships (R2 > 0.8) for the high-frequency range (≥1250 Hz), where
typically stick-slip, stick-snap, air pumping and pipe resonance related sound generation
mechanisms dominate. Moderate to strong relationships (R2 between 0.5 and 0.7) are observed for
mid frequencies with a drop at 1000 Hz (in this frequency range typically a broad mix of sound
generation mechanisms are present). Moderate relationships (R2 between 0.44 and 0.55) can be
found for the lower frequency range in which vibrational sound, caused by the impact of road
surface texture and tire tread, dominates.
The graph on the right of Figure 6 shows that highest standard errors of the slope are found for Tair,
followed by Ttire. The lowest standard errors are found for Troad. This does not directly correlate with
the uncertainty of temperature corrections, as the slopes also differ substantially; for example, the
slopes for air temperature are generally almost twice as large as those of road temperatures. Given
the magnitude of the temperature effects reported for the different frequencies and temperature
variables (see section 6.2), the linear models can be considered as reasonably robust. To estimate
the uncertainty of the different correction approaches based on these three temperatures, the here
provided standard errors need to be weighted with the common ranges of each temperature variable
(see also section 0).
The average profiles for Tair, Troad and Ttire over the 9 days measurement campaign in August/Sep-
tember 2010 displayed in Figure 7 versus the time of day give us a better understanding on how
these temperatures relate to each other.
Figure 7 – Average temperature profiles during a 9 day measurement campaign (with 95 % confid. intervals)
An important observation from the temperature profiles in Figure 7 is that the profile of Tair and Ttire
follow a similar path over the course of the day, while Troad shows a different behavior. When
comparing the profile of Troad to the one of Ttire – with Ttire considered as the most relevant
temperature for explaining temperature effects – it becomes evident that it will be difficult
explaining temperature effects based on Troad alone, as the influence of the air surrounding the tire
cannot be neglected. The explanatory power of Tair alone seems not to be entirely sufficient either:
The impact of solar radiation on Ttire through heating-up of the pavement becomes clearly visible, as
the offset of Ttire from Tair starts to gradually increase after the sun rises at around 06:40. In the early
morning Ttire is 9 °C warmer than Tair, increasing to 15 °C in the afternoon, when the maximum
offset is reached at 15:00. From the above observations it can be expected, that neither Troad nor Tair
alone can be used for a reliable prediction of Ttire. The effect this may have on the accuracy of
existing temperature correction approaches will be investigated later in section 6.3.
6.2. Temperature effects and suggested correction factors for Tair, Troad and Ttire
From the same dataset of the 2010 study average temperature effects and suggested spectral
correction factors were calculated for each of the three temperatures. The results from this analysis
are presented in Figure 8 with separate averages calculated for the microphone positions to the front
and the rear of the tire/road contact patch, to allow for a better interpretation of the involvement of
different sound generation mechanisms (further discussed in section 7).
Figure 8 shows that the temperature effects vary over the noise spectrum and shows a fairly
consistent pattern for the three temperatures. Smaller disparities can be found between the
microphone positions to the front and the rear of the contact patch. This was expected, as there are
differences regarding the nature and contribution of sound generation mechanisms. The data show
that the high frequency range is most affected by temperature effects (up to -0.27 dB/°C for Tair),
while the low and mid frequency range average at around -0.1 dB/°C for Tair with dips at around
500 and 1000 Hz at the front microphone. At the rear microphone, there is an additional dip at
400 Hz while the dip at 1000 Hz is less pronounced.
Time of day [h]
Figure 8 – Average spectral temperature effects and suggested correction factors with 95 % confidence
intervals derived for Tair, Troad and Ttire (average over 28 tire/pavement combinations at speed 50 km/h)
The temperature effects derived for the three temperatures are summarized in Table 4.
Table 4 – Average temperature correction factors derived for Tair, Troad and Ttire (average over 28
tire/pavement combinations at the reference speed 50 km/h)
Overall average effect
Spectral variation (315 to 5k Hz)
-0.01 to -0.27 dB/°C
-0.02 to -0.11 dB/°C
-0.03 to -0.16 dB/°C
It should be stated that the above factors were derived based on a comprehensive measurement
campaign involving two sets of two tires. The effects that were measured in the right and left wheel
track (not shown here) showed a very consistent behavior, suggesting a high robustness of the
determination of temperature effect.
The magnitude of the temperature correction factors derived for Troad amount to only 44 % and the
ones of Ttire to 63 % in comparison to the average factor for Tair. This is directly linked to the greater
variance these variables show, due to influence of solar radiation (see section 0 for more on this).
6.3. Impact of different correction approaches on measurement uncertainty
An independent validation dataset from a study in 2017 was used to validate the temperature
correction factors derived in section 6.2 and to test the different correction approaches when using
Tair, Troad or Ttire based correction factors. The dataset comprises 32 consecutive CPX measurement
runs undertaken during one single day between 04:00 and 17:00 with two different tires (P1 and
H1) on a road section of 5 km length with 7 different pavements (four semi-dense asphalts SDA
with void contents between 12 and 18%, two DAC 11 and one SMA 8). The temperature ranges
covered during each of those runs are presented in the upper part of Figure 9 with each data stacks
representing a measurement run. To investigate the robustness of the different temperature
correction approaches based on either Tair, Troad or Ttire, the distribution of standard deviations is
assessed after applying each of those correction approaches (see Figure 9 lower part).
As shown in Figure 9, the run-to-run variation for each 20 m segment (percentage of 20 m segments
with a remaining standard deviation >0.5 dB) can be reduced from 64 to 50 % for Tair, if spectral
correction approaches are applied instead of flat corrections that are used equally over the entire
frequency spectrum, but for road and tire temperatures, the spectral correction gives only a marginal
improvement compared to flat correction. Interestingly, the correction approaches based on Troad
and Ttire (green and yellow curves coinciding) perform considerably better than the ones for Tair.
Figure 9: Distribution of standard deviation (run-to-run variation for each 20 m segment) on independent
validation dataset, when different correction approaches are applied. The percentage values AUC (Area
Under the Curve) represent the share of 20 m segments that show a standard deviation >0.5 dB.
A numeric summary of the above graph is given together with the median standard errors for each
of the tested correction approaches in Table 5.
Table 5 – Performance of the tested correction approaches with regard to the median standard error and
percentage (“area”) above standard error > 0.5 dB.
Median standard error in dB
Area of std. dev. > 0.5 dB
Flat correction, air (current standard)
It must be stressed that this apparent advantage of road temperatures is based on highly unfair
conditions, since the air temperatures cannot change from each 20 m long road segment to segment
(i.e., from second to second), while road temperatures can change over such short stretches. Tire
tread surface temperatures would probably change between the 20 m segments in a way which is
somewhere between that of air and road temperatures. Probably, also the air temperatures may
“flutter” from segment to segment due to air turbulence. Note also that test speed was only 50 km/h
here. Nevertheless, the differences in median standard error in Table 5 are small (only 0.07 dB).
6.4. Which temperature is most suitable to use for correction?
A new extensive dataset with millions of temperature measurements was analyzed to investigate
how the different temperatures Tair, Troad and Ttire relate to each other in different weather conditions
and as a function of time, i.e., over day/night and seasons. The dataset was obtained between 2010
and 2020 within the scope of Swiss projects covering a total measurement distance of around
50 000 km. In Figure 10 the relationship between the different temperatures and the relevant Ttire are
investigated, focusing on data collected at the reference speed 50 km/h (accounting for 62 % of all
data points in the dataset).
Figure 10 – Relationships between Ttire and Tair (upper part), and Troad (lower part) during daytime with
varying amount of solar radiation (left) and during nighttime i.e., without solar radiation (right).
To investigate the relationships between Ttire and Tair or Troad, different models were tested. The
goodness of fit (GoF) values for the different models indicate that the linear model represents the
data best. For the linear models presented in Figure 10 it can be observed that the relationship
between Tair and Ttire show different slopes for daytime than for nighttime: the slope for the data at
daytime, i.e. with solar radiation (right), is considerably steeper than the one at nighttime. This
means that for any correction approach based on Tair alone, this would result in a certain error if one
makes measurements over both day and night, as the impact of solar radiation on tire temperature
cannot be fully accounted for. This is an important finding for standardization, because standards
should provide robust temperature correction procedures that work accurately at different times and
weather conditions (i.e., cloudy versus sunny).
The difference in slope is far less pronounced for the relationship between Troad and Ttire (see bottom
graphs). In this second case, however, the regression line is steeper during nighttime, as the
additional heating of the pavement during daytime only partially transfers to the tire temperature. In
the case of correction approaches solely based on Troad, the expected error for weather conditions
with varying amounts of solar radiation seems to be far less important, as this impact is already
reflected in the measurement of Troad. An important result in Figure 10 is that the tire temperatures
seem to be somewhat more closely related to road surface than to air temperatures, as the R2 values
are higher for road than for air temperatures.
To estimate the expected uncertainties for correction approaches based on Tair and Troad alone, the
impact of the varying slopes with Ttire are multiplied with the respective temperature correction
factors for Ttire from section 6.2. The results of this error estimation are presented for common
temperatures in Table 6.
Table 6 – Estimated uncertainties for correction approaches based on Tair, Troad and Ttire at typical temperature
Variation of relationship to Ttire (nighttime within
Expected typical additional uncertainty
(multiplication of variation with corr. factor Ttire)
4.5 °C @ 30 °C Tair
(2.9 °C @ 20 °C Tair)
-1.4 °C @ 30 °C Troad
(-0.55 °C @ 20 °C Troad)
The additional uncertainties (resulting from an insufficiently explained tire temperature by the other
temperature variables) presented in Table 6 can be considered as important, but not necessarily
critical. It should be stated, however, that the above figures are derived from measurements that
were undertaken in a moderate climate (in the plains of Switzerland) with a total annual direct
normal solar irradiation of around 1315 kWh/m2. The listed uncertainties will likely increase for
warmer climatic regions with higher solar radiation amounts (at latitudes <47°).
6.5. Consequences for temperature correction of noise levels
The following considerations and recommendations regarding the use of different temperatures in
correction approaches in tire/road noise standards can be made from the results presented above:
▪ Ttire: Tire temperature constitutes arguably the most relevant temperature for explaining (and
correcting for) temperature effects on noise and performs clearly better than Tair and almost
equally well as Troad when testing it in the correction approaches here (see analyses in section
6.3). The challenge for correction approaches based on Ttire is that tire temperatures vary
substantially from location to location on the tire surface and from location to location within
the tire structure. The matter becomes rather complex, as these variations also change with time
(as the tire warms up and cools off). Therefore, more research is needed to assess the
explanatory power of different tire temperatures, the ways they best can be measured, and how
they should be incorporated in future correction approaches.
▪ Troad: Road surface temperature performs equally well in correction approaches as Ttire and
considerably better than Tair (see analyses in section 6.3). However, this is based on
measurements over 20 m segments and for only 50 km/h. For longer test sections and for higher
speeds, this advantage may well be different and even opposite. The drawback from correction
approaches based on Troad is that it is not entirely possible to predict the relevant Ttire, since the
cooling of the tire by air cannot be accounted for.
▪ Tair: Air temperature is the least performant temperature of the three temperatures investigated
in this study (see analyses in section 6.3), with its limitations as stated in the previous point. A
considerable disadvantage of using correction approaches solely based on Tair is that it does not
sufficiently account for the impact of solar radiation on Ttire via conduction from the road
▪ The above considerations are based on cases where road temperatures may vary substantially
between 20 m long segments due to shading from roadside objects. In open areas where this is
not the case, the road temperature may be less important for corrections.
Based on the same dataset, it was investigated whether a combined model with the easy to measure
temperatures Tair and Troad would lead to an improved prediction of Ttire (and, hence, also for
temperature effects). The results can be summarized as follows:
▪ Combined model with Tair and Troad: As we have learned from the analysis in the previous
sections, tire temperature is highly influenced by both Tair and Troad. The advantage of
incorporating Tair in the correction approach on the one hand, is that it shows a very similar
behavior to the relevant Ttire (see section 6.1). By incorporating Troad in the correction approach
in the other hand, this may allow to account for the additional heating of the tire at daytime
(especially at sunny days in seasons/regions with high solar radiation). The impact of surface
heat is also likely to occur during nighttime, as the pavement stores energy from the day or
cools down as it emits more energy than it receives (at clear nights). For the combined
multivariate linear model of Equation 2, R2 was increased from 0.92 to 0.96 in comparison to
the correlation with air temperature alone. This corresponds to a reduction in non-controlled
variance from 8 % to 4 %; i.e., a significant improvement. Therefore, it is a promising idea to
establish temperature correction procedures for standardization with Tair and Troad combined,
which should be further assessed before implementation.
The resulting combined multivariate model with Tair and Troad to predict Ttire is expressed in
Equation 2 for measurements at 50 km/h.
The weightings in Equation 2 suggest, that the road temperature is slightly more important in
predicting the tire temperature, for our dataset, with its limitations. The interpretation of the
coefficients or even the multivariate modelling must be considered with care, as all the temperatures
are highly correlated among each other, and thus leading to multicollinearity.
6.6. Recent results and recommendations from the STEER project
The authors are involved in an ongoing project named Strengthening the Effect of quieter tires on
European Roads (STEER), sponsored by CEDR, which is an organization of European national
road administrations. The purpose of STEER is to suggest how the EU noise labelling system for
(car) tires can be improved. Part of this is how we can reduce the uncertainty in noise measurements
on ISO pavements due to improved temperature corrections. A significant part of the report deals
with the relation between air and road surface temperatures and how to convert corrections from
using air to road temperatures. It is noteworthy that it is found that tire temperature is better related
to air than to road temperature in the data reported in the STEER report, which is opposite to the
conclusions in the present study. The reason for this should be explored. The report about these
activities is under publication, but briefly, the following are suggested:
• Preferably change from basing the correction on surface temperature to air temperature
• If this is not politically possible, one can retain the old procedure but using a more progressive
correction than today, which is part-linear; i.e., it includes two linear relations under and above a
certain “knee” temperature
• Conduct research to try to find a suitable and more relevant correction based on tire temperature
for the future. In the meantime, it is envisioned that the use of an average of air and surface
temperatures (maybe weighted) would improve the correction, but this must be tested further
• The allowed temperature range is limited to a maximum temperature of 40 oC for both air and
• To facilitate the above, measures are proposed to color the usually “black” asphalt ISO surfaces
in a light greyish color and/or to use a temperature-regulating pipe system built into the
pavement to heat and/or cool the pavement by means of conventional heating or cooling systems.
By implementing these suggestions, it is expected that the uncertainty contribution of temperature
deviating from the reference condition is reduced significantly.
7. TEMPERATURE EFFECT ON NOISE GENERATION MECHANISMS
Tire/road noise generation mechanisms are complicated and there are (arguably) three major
mechanisms, some of which act in different frequency ranges, and some also act in opposite ways
on the noise generation . Therefore, by looking into how temperatures influence the frequency
spectra, the chances of finding physical explanations for the effects are increased. The following is
a brief discussion about probable causes for the temperature effects on noise generation.
In the lower frequency (LF) range (Figure 8) the temperature coefficient is around -0.10, except at
500 and 1000 Hz, the latter of which may be tread impact frequencies and harmonics. The tread
impact seems to be much less influenced by the temperature. It is speculated that this is because it
creates “air pumping” in grooves and other tread cavities. It is difficult to imagine that air pumping
is influenced significantly by temperature. The reduced noise at LF with temperature may poten-
tially be caused by the hysteresis energy losses in the rubber being reduced when temperatures
increase, something which is even more influential in rolling resistance. Tire inflation is known to
increase with hotter air inside the tire and if inflation is not adjusted for increasing temperatures, it
will stiffen the tire and thus potentially cause less deflections, normally resulting in less vibrational
excitation of noise.
In the higher frequency (HF) range, temperature effects are the highest. This may be caused by
stick-slip and stick-snap actions in the contact patch, which will cause tangential vibrations but may
also increasingly excite air pumping . For example, it is reported that a higher pavement,
ambient air, and contained air temperatures resulted in a lower hysteretic friction of tires for a given
pavement surface . Again, hysteresis may be involved, but also friction will decrease which
means that stick-slip vibrations will be released at lower tangential forces.
8. CONSIDERATIONS IN STANDARDIZATION OF TEMPERATURE CORRECTIONS
The tire industry seems to favor the use of road surface temperature for the temperature corrections.
However, although the results of this study support that principle, independent results reported in
, indicate that tire surface temperatures are more influenced by ambient air than by road
temperatures. Therefore, it should be considered to change the ISO Technical Specifications (or
future standards) to be based on both air and road temperatures, as it is quite obvious that each one
has its unique effects. Further studies should be conducted to verify Eq. 2 in our study, for both 50
and 80 km/h. Also, the noise measurements on which tire noise limits and labelling levels are based
should be changed from being based on road surface to a mix of air and road temperatures. This is
equally valid for the part of vehicle noise testing (ISO 362 or similar) which is dominated by
tire/road noise, which are not presently subject to temperature corrections.
This study also gives insight into the temperature influence on tire/road noise frequency spectra. It
is logical if the various noise generation mechanisms acting in tire/road noise generation are
differently influenced by temperature and thus will give inconsistent influences over the frequency
range. In future standardization, not only the A-weighted overall levels but also the frequency
spectra should be normalized to 20 oC in order to further improve the correction efficiency
whenever frequency spectra are reported. This study constitutes a significant contribution to
determination of such a spectral correction, but more data must be collected.
This paper reported research results based on a huge amount of noise- and temperature-related data
in Swiss CPX measurements and incorporates data from the ongoing project STEER. The results
provided a deeper insight into the relationships between the temperatures of ambient air, road
surface and tire surface, about their variation over day-night times and seasons, and also about the
influence on tire/road noise frequency spectra.
As a result, we found reasonably consistent data on the shortcoming of using ambient air and road
surface temperatures as a single variable in temperature correction procedures. A promising solu-
tion may be to combine both variables in temperature correction approaches to ensure that both
important influences, the impact of solar radiation and the impact of ambient air, are accounted for.
An alternative may be, to develop correction approaches directly based on tire temperature (where
this is practicable), although much more research will be needed on this. To further improve the
correction accuracy, in future standardization, not only the A-weighted overall levels but also the
frequency spectra should be normalized to 20 oC. It is recommended for the responsible
standardization bodies, to review their current temperature correction approaches and invest into
more progressive correction procedures to further reduce measurement uncertainty.
Bühlmann and Schlatter would like to acknowledge the financial support from the Swiss Federal
Office for the Environment FOEN (dataset from study 2010) and the Canton of Aargau (validation
dataset from study 2017) at the stage of data collection. Sandberg recognizes that his work is jointly
sponsored from the CEDR project STEER and the International Standardization project #20229
sponsored by Trafikverket in Sweden.
 ECE R117 (2011): Uniform provisions concerning the approval of tyres with regard to rolling sound emissions and
to adhesion on wet surfaces and/or to rolling resistance. ECE Regulation 117, United Nations Economic
Commission for Europe, Geneva, Switzerland.
 ISO/TS 13471-1 (2017): Acoustics - Temperature influence on tyre/road noise measurement - Part 1: Correction for
temperature when testing with the CPX method. Geneva, Switzerland: International Organization for
 ISO/DTS 13471-2 (2021): Acoustics - Temperature influence on tyre/road noise measurement - Part 2: Correction
for temperature when testing with pass-by methods. Geneva, Switzerland: International Organization for
 ISO 11819 - 2 (2017): Acoustics - Measurement of the influence of road surfaces on traffic noise - Part 2: The
Close-proximity method. Geneva, Switzerland: International Organization for Standardization.
 ISO/TS 11819-3 (2017): Acoustics - Measurement of the influence of road surfaces on traffic noise - Part 3:
Reference tyres. Geneva, Switzerland: International Organization for Standardization.
 WEHR, R., FUCHS, A., AICHINGER, C. (2018): A combined approach for correcting tyre hardness and
temperature inﬂuence on tyre/road noise, in: Applied Acoustics 134 (pp. 110-118).
 BÜHLMANN, E. & VAN BLOKLAND, G. (2014): Temperature effects on tyre / road-noise – A review of
empirical research. Proc. of Forum Acoustica 2014.
 BÜHLMANN E., SANDBERG U., MIODUSZEWSKI P. (2015): Speed dependency of temperature effects on road
traffic noise. Proc. of Inter-Noise 2015, San Francisco, USA.
 AASHTO (2016): Standard method of test for measurement of tire-pavement noise using the on-board sound
intensity (OBSI) method. AASHTO T360, Americ. Assoc. of State Highw. Transp. Officials, Washington , USA.
 BÜHLMANN E. & ZIEGLER, T. (2011): Temperature effects on tyre/road noise measurements. Proc. Inter Noise
2011, Osaka, Japan.
 SANDBERG, U.; EJSMONT, J.A. (2002): Tyre/Road Noise Reference Book. Informex (www.informex.info).
 ANUPAM, K.; SRIRANGAM, S.K.; SCARPAS, A.; KASBERGEN C. (2014): Influence of Temperature on Tire-
Pavement Friction: Analyses, in: Transportation Research Record Journal of the TRB 2369(4):114-124.
 SANDBERG, Ulf; VIEIRA, Tiago (2021): Temperature effects on tyre/road noise measurements. Technical
Report R3.1 in WP 3 of Project STEER by CEDR (under publication, will be published in http://vti.diva-portal.org).