HAL Id: hal-03233736
https://hal.archives-ouvertes.fr/hal-03233736
Submitted on 13 Jun 2021
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Discussion of Measurement Uncertainties in Soundscape
Studies
Andre Fiebig
To cite this version:
Andre Fiebig. Discussion of Measurement Uncertainties in Soundscape Studies. Forum Acusticum,
Dec 2020, Lyon, France. pp.1453-1456, �10.48465/fa.2020.0529�. �hal-03233736�
DISCUSSION OF MEASUREMENT UNCERTAINTIES IN
SOUNDSCAPE STUDIES
André Fiebig
11 Department of Engineering Acoustics, Technische Universität Berlin, Germany
andre.fiebig@tu-berlin.de ABSTRACT
An increasing interest in soundscape studies can be noted indicated by the steady rise of soundscape publications and new releases of soundscape standards and technical specifications. The basic tenet of the soundscape approach is the consideration of the context, where perception of acoustics environments is inevitably embedded in. This calls for context-sensitive surveys, where the collection of physical and perceptual data has to be performed in situ. With this special emphasis on field related investigations, the consideration of measurement uncertainties and reliability aspects is of particular importance. How to measure reliably perception and sound in rather uncontrolled conditions? How to decide whether a measurement is meaningful, representative or even valid or not?
1. INTRODUCTION
The soundscape approach and its application gain in significance indicated by the increasing number of publications [1]. The availability of the ISO 12913-1 Standard [2] and the Technical Specifications ISO/TS 12913-2 [3] and ISO/TS 12913-3 [4] defining the way to collect, analyze, and report soundscape data seems to be an additional catalyst for the increased use of the soundscape approach and the performance of soundscape investigations.
In contrast to the classical noise control and environmental noise assessment approach, the soundscape approach demands to primarily focus on perception. The definition of the term soundscape provided by the ISO 12913-1 highlights this paradigm shift: Soundscape means an
”acoustic environment as perceived or experienced and/or understood by a person or people, in context” [2].
However, this measurement paradigm shift – moving from the measurement of sound pressure levels (and interpreting them as indicator of noise annoyance) to the measurement of perception – implicates serious challenges.
Besides numerous aspects, such as the discussion on the adequacy of methods, the operationality of perception, the derivation of descriptors and indicators, the triangulation of qualitative and quantitative data, the estimation of measurement uncertainties becomes more important. The aspect of measurement uncertainties is of particular interest for the further development of national and
international standards or technical specifications, for the general acceptance of those documents and finally the implementation of the soundscape method. The discussion of measurement uncertainties and how to quantify them in detail are increasingly vital as it is expected that the standardization developments will improve the level of compatibility of soundscape studies and allow for performing meta-analysis. This makes a reliable estimation of measurement uncertainties even more important. Altogether, the aspect of measurement uncertainties, despite all the progress in soundscape research, is not conclusively discussed so far and needs further attention.
2. CONCEPTS OF MEASUREMENT UNCERTAINTIES AND TRUE VALUE Measurement uncertainty means “a parameter, associated
with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand” [5].
In general, the result of a measurement is determined on the basis of series of observations obtained under repeatability conditions and that variations in repeated observations are assumed to arise because influence quantities are not held completely constant affecting the measurement result - the source of uncertainties. [5] The uncertainty in the result of a measurement can be generally evaluated by statistical methods or by other means. This means that for all types of soundscape studies information about measurement uncertainties specifying the magnitude of uncertainty is relevant. However, the determination of uncertainty does not seem as straightforward as expected and a detailed consideration of terms is necessary. According to the ISO 3534-2 it is important to differentiate between uncertainty and estimation, because “uncertainty is distinguished from an
estimate attached to a test or measurement result that characterizes the range of values within which the expectation is asserted to lie. The latter estimate is a measure of precision rather than of accuracy and should be used only when the true value is not defined.” [6]
Thus, measurement uncertainties provides information about the precision of a measurement, i.e. the level of agreement between independent measurement results obtained under stipulated conditions. This information does not imply information about the accuracy or trueness (i.e. validity) of the measurement result, because the term
trueness indicates the correspondence between the (expectation of) a measurement result and the true value. Unfortunately, the true value is usually not known and we have to assess measurement uncertainties on the basis of estimated values.
This means in a strict sense, even if we can quantify measurement uncertainties, we cannot conclude from it the validity of the measurement result. This means that, on the one hand, we have to question the reliability and validity of a soundscape measurement result, if the measurement uncertainties are large. On the other hand, a reasonable measurement uncertainty is necessary, but not sufficient for validity.
3. MEASUREMENT UNCERTAINTIES OF SOUND AND PERCEPTION IN SOUNDSCAPE
STUDIES
In the context of soundscape investigations, uncertainties have to be discussed for acoustical and perceptual measurements. However, it is obvious that the con-sideration of measurement uncertainties for acoustic measurements is more straightforward in comparison to perceptual measurements (judgmental data). In the following, the uncertainties regarding physical measure-ments are discussed only with respect to the investigated locations and sites. Of course, the measurement equipment itself with its inherent data processing possesses uncertainties as well. This uncertainty aspect is not subject of the following considerations.
Figure 1 shows the coefficient of variation cv (the standard
deviation divided by the sample mean of a data set) for the measurement of the psychoacoustic loudness using measurements at multiple locations with different measurement time intervals. The coefficient of variation indicates the variation of measurement results inde-pendently of the unit. It can be seen that the uncertainty for loudness measurement results according to the DIN 45631/A1 [7] decreases in average with longer measure-ment time intervals per location. This means that the similarity of loudness results is 20 % closer if the measurement time interval is increased from 1 to 3 minutes AS expected this observations indicates the critical issue of too short measurement time intervals. Accordingly, the Technical Specification ISO/TS 12913-2 [3] demands as a minimum requirement that the measurement time interval shall not be shorter than 3 minutes.
However, figure 2 illustrates the variation of measurement results considering multiple acoustical parameters at the same time for different sites. Again the coefficient of variation is used to indicate the degree of variation. It can be seen that the magnitude of dispersion of (physical) measurement results depend on the respective site with its specific character. Some locations show only small variations regarding the similarity of repeated acoustical measurement results, whereas other locations show much larger variations. This means that the amount of uncertainties simply depend on the type of the investigated
soundscape. Thus, the ISO/TS 12913-2 proposes to choose the measurement time interval “[…] sufficiently long to
encompass all sound situations that are needed to obtain a representative picture of the soundscape”. [3]
Figure 1. Average of coefficient of variations of loudness measurement results of 13 repeated measurements at eight sites in dependence of measurement duration [8] As it is mostly unknown which time period must be chosen to obtain a “representative picture” of an investigated soundscape, the performance of preliminary studies and using the knowledge of local experts (persons who are familiar with the area under scrutiny [3]) seem to be mandatory. Preparatory, preliminary investigations allow for estimating the degree of variability of a soundscape (in terms of acoustical properties) and the acoustical measurement conditions can be adapted to sufficiently address measurement uncertainties.
Figure 2. Average of coefficient of variations considering 35 (psycho)-acoustical parameters of 13 repeated measurements over locations loudness measurement results of 13 repeated measurements over location [8] This means that the measurement uncertainty regarding the measurement of psychoacoustic loudness of a specific site has can be comprehensively quantified.
0,00 0,05 0,10 0,15 0,20 0,25 0,30
1 Min 2 Min 3 Min
coef fi cient of v ariatio n Duration of measurement Coefficient of variation regarding
loudness measurements 0,00 0,05 0,10 0,15 0,20 0,25 0,30 1 2 3 4 5 6 7 8 co ef ficien t o f va riatio n Location
Coefficient of variation of acoustical measurement results over locations
Of course, if perception and judgment respectively (e.g. of loudness) are considered, then measurement uncertainty in terms of a dispersion of ratings can be discussed as well. Figure 3 compares the variation of loudness judgments over repeated measurements (during soundwalks) with the respective variation of the measured psychoacoustic loudness values using the ISO 532-1 [9] for calculating the time-variant loudness based on the time signals collected during the soundwalks. In total 8 soundwalks with 57 participants were performed and used for the analysis of coefficient of variation. The participants were requested to listen in silence three minutes and then to provide their ratings. The same time period was used for the performance of acoustical measurements.
Figure 3. Top: Coefficient of variation over location regarding psychoacoustic loudness measurements according to ISO 532-1 [9] based on 8 repeated measurements (soundwalks). Bottom: Coefficient of variation over location regarding in-situ loudness judgments based on 8 repeated measurements
It can be seen that the coefficients of variation over the locations vary in a similar way; although the coefficient of variation concerning the loudness judgments is much greater than the variations of measured psychoacoustic loudness values. The coefficients of variation between the
acoustical and the judgmental data correlates with r=0.84. This means that the locations, which vary strongly in their psychoacoustic loudness, provoke at the same time loudness ratings with greater variations.
The correlation of all single loudness judgments with the corresponding psychoacoustic loudness measurements
according to ISO 532-1 (using N5 as the representative
loudness value for overall loudness) exhibits a high relationship of r=0.74**. Altogether it seems that at least for sensory related judgments quantifying the magnitude of a certain auditory sensation, the measurement uncertainties of physical measurements correspond to uncertainties of perceptual measurements. However, as the perception of acoustic environments is not limited to auditory sensations (like loudness, sharpness or tonality), the question remains whether this presupposition is also true for perceived affective qualities of acoustic environments. For example, the ISO/TS 12913-2 proposes a questionnaire, among others, where attributes related to the pleasantness (valence dimension) and eventfulness (arousal dimension) are considered, which represent well established dimensions in environmental psychology [10]. These two independent dimensions are considered to be important as they reflect evolutionary needs, promoting survival by preferring certain environments and avoiding others [11].
Figure 4. Coefficient of variation regarding in-situ unpleasant judgments based on 8 repeated measurements (soundwalks) over location
Figure 4 illustrates the variation of unpleasantness ratings in terms of coefficient of variation over the 8 repeatedly visited measurement locations. As expected, the valence related judgments vary stronger from person to person and over the repeated measurements than the sensory related judgments, which are obviously less affected by personal aspects. At all locations, the unpleasantness ratings vary stronger than the loudness ratings. Moreover, the correlation of all single unpleasantness judgments with the corresponding psychoacoustic loudness measurements according to ISO 532-1 drops remarkably down to
0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1 2 3 4 5 6 7 8 co ef ficien t o f va riatio n Location
Coefficient of variation of loudness measurements according to ISO 532-1
0,00 0,10 0,20 0,30 0,40 0,50 0,60 1 2 3 4 5 6 7 8 co ef fi cien t o f variatio n Location
Coefficient of variation regarding loudness judgments 0,00 0,10 0,20 0,30 0,40 0,50 0,60 1 2 3 4 5 6 7 8 coefficien t of variation Location
Coefficient of variation regarding unpleasantness judgments
r=0.64**. According to Guski, it is assumed that only
30 % of noise annoyance data can be explained by sound pressure level [11]. Interestingly, the amount of explained variance of unpleasantness by the measured
psycho-acoustic loudness (N5)or sound pressure level (LAeq)falls
into the expected range of 41 % and 36 % explained variance respectively. Because during a soundwalk only short-term measurement time intervals are considered the amount of explained variance by acoustic indicators describing the perception of loudness is slightly higher compared to noise annoyance surveys.
The relationship between the coefficient of variation of psychoacoustic loudness measurements with the variations observed with respect to unpleasantness over the 8 locations is with r=0.65 less strong as for the sensory related ratings (e.g. loudness).
4. CONCLUSIONS
The consideration of measurement uncertainties is imperative for assessing the quality and reliability of measurements in general. Therefore, as the soundscape approach gains in significance the discussion of measurement uncertainties becomes more important. The long-term objective of soundscape research must be to develop measurement quality indicators, which help to assess the magnitude of measurement uncertainties in soundscape studies with respect to acoustical as well as perceptual measurands.
In the context of measurement uncertainties, it appears important to distinguish systematic effects and random
effects of uncertainty, which contribute to the dispersion of
data. The observations made above indicate that some “uncertainties” (variations) in the measurement of perception follow the “uncertainties” (variations) of the acoustical properties of a location. However, the degree of variation depends also on the aspect and category of perception investigated. The “random” component of uncertainty refers to the dispersion in the (loudness) ratings influenced by “uncontrolled” (unknown) variables. Systematic effects vary in a predictable manner, whereas random errors vary in an unpredictable manner. Further research in the field of soundscape needs to separate more in detail the systematic effects from the random effects. This will lead to a better understanding and management of measurement uncertainties in soundscape studies. Finally, it is important to note that the measurement uncertainties in soundscape studies should not be understood as a kind of error, because if we deal with perception, which is generally of a probabilistic nature, the observation of dispersion is always a valuable information.
5. REFERENCES
[1] To, W.M., Chung, A., Vong, I., Ip, A.(2018). Opportunities for soundscape appraisal in Asia, Euronoise 2018, Proc., Crete, Greece
[2] ISO 12913-1 (2014). Acoustics-Soundscape-Part 1: Definition and conceptual framework. International Organization for Standardization, Geneva, Switzer-land
[3] ISO/TS 12913-2 (2018). Acoustics-Soundscape-Part 2: Data collection and reporting requirements. International Organization for Standardization Geneva, Switzerland
[4] ISO/TS 12913-3 (2019). Acoustics-Soundscape-Part 3: Data analysis. International Organization for Standardization Geneva, Switzerland
[5] JCGM 100 (2008). Evaluation of measurement data - Guide to the expression of uncertainty in measurement, Joint Committee for Guides in Metrology
[6] ISO 3534-2 (2006). Statistics. Vocabulary and symbols. Part 2: Applied statistics. International Organization for Standardization, Geneva, Switzer-land
[7] DIN 45631/A1 (2010). Calculation of loudness level and loudness from the sound spectrum - Zwicker method - Amendment 1: Calculation of the loudness of time-variant sound, Beuth, Berlin, Germany [8] Fiebig, A. (2016). Reliability of in-situ measurements
of acoustic environments, DAGA 2016, Proc., Aachen, Germany
[9] ISO 532-1 (2017). Acoustics - Methods for calculating loudness. Part 1: Zwicker method, Inter-national Organization for Standardization, Geneva, Switzerland
[10] Mehrabian,, A., Russell J.A. (1974). An approach to environmental psychology, MIT Press, Cambridge, Massachusetts and London
[11] van den Bosch, K.A.M., Welch, D., Andringa, T.C. (2018). The evolution of soundscape appraisal through enactive cognition. Frontiers in Psychology, July 2018, Vol. 9, 1129
[12] Guski, R. (1999). Personal and social variables as co-determinants of noise annoyance, Noise and Health 1999, 3, 45-56
