An update to the WHO 2018 Environmental Noise Guidelines exposure response relationships for annoyance from road and railway noise Benjamin Fenech 1 , Sierra Clark, Georgia Rodgers Noise and Public Health team, Radiation Chemical and Environmental Hazards, Science Group, UK Health Security Agency, United Kingdom

ABSTRACT The systematic review on environmental noise and annoyance commissioned by the World Health Organization (WHO) to inform its 2018 Environmental Noise Guidelines proposed new aggregate exposure response relationships (ERRs) for transportation noise, based on studies published between 2000 and 2014. A subsequent scoping review commissioned by the UK’s Interdepartmental Group on Costs and Benefits Noise Subject Group identified 12 new studies for road and nine new studies for rail, published between 2014 and 2019, and recommended that an update to the WHO ERRs may be warranted. This paper proposes updated aggregated ERRs for the percentage of people highly annoyed (%HA) from road and railway traffic noise incorporating data from a subset of recent studies, published between 2014-2022, that met certain eligibility criteria. In light of the ongoing debate on the temporal stability of community annoyance (observed or modelled), we also discuss some of the important considerations that need to be taken into account when deciding if and how often a reference exposure-response relationship is updated for policy and practice. 1. INTRODUCTION

The World Health Organization (WHO) 2018 Environmental Noise Guidelines (ENG) [1] were informed by systematic reviews of scientific literature on the health effects of noise published between 2000 and 2014/2015. The systematic review for noise annoyance identified 26 studies on road-traffic noise (total of 34,211 respondents), and 10 studies for railway noise (total of 10,970 respondents) published between 2000 and 2014 [2]. The exposure response relationships (ERR) between road and rail traffic noise and annoyance were derived predominantly from studies conducted in Europe and Asia [2]. For the road-traffic noise studies, the authors of the systematic review identified a subgroup of studies from Asia and the Alpine valleys in Austria with several study characteristics that could influence the results: the range of noise exposure levels investigated, housing characteristics linked to ventilation and air conditioning, geographical terrain, and annoyance scale cut-offs used to determine the highly annoyed category. Therefore, two ERRs were presented: one using the full dataset of 26 studies and one excluding five Alpine studies and the ten Asian studies. For the railway studies, a study of highspeed rail (Shinkansen) was excluded from the estimation of an average rail ERR because it differed considerably from the other studies of conventional rail.

Since the WHO-commissioned systematic reviews, numerous new studies with a diverse geographical spread have been published. The purpose of this present review and meta-analysis was

1 Corresponding author benjamin.fenech@phe.gov.uk

to derive new ERRs between road and railway-traffic noise and annoyance including new studies published between 2014-2022.

2. MATERIALS AND METHODS

2.1. Search for additional studies published between 2014 – 2022

The search for new studies, screening, and extraction of the data occurred in two separate phases. First, a scoping review commissioned by the UK Department for Environment, Food & Rural Affairs (Defra) on behalf of the Interdepartmental Group on Costs and Benefits Noise Subject group (IGCB(N)) and undertaken by van Kamp et al. [3] identified several new studies published between January 2014 and June 2019, which we refer to as the 1 st update. Next, a follow-up search was conducted by the UK Health Security Agency (UKHSA) in February 2022 to find additional studies published between 2019 and early 2022, referred to as the 2 nd update.

2.2. Database search and paper screening

A detailed description of the protocol for searching through bibliographic databases and title and abstract screening of retrieved papers for the 1 st update can be found in van Kamp et al. [3]. In brief, Scopus, MEDLINE (PubMed), EMBASE, and PsycInfo bibliographic databases were searched in the Spring/Summer of 2019. Search terms followed the protocol of the WHO evidence reviews as much as possible (see Table 1). All retrieved Dutch, English, French and German papers were screened independently by two researchers for relevance to the topic (based on the search protocol), and papers that did not match the inclusion criteria were excluded.

Table 1. S ear ch profile (adapted from [3]). 1 Published or Accepted Papers in Peer-Review Journals 2 Published papers in conference proceedings 3 Individual studies (no reviews, meta-analyses or “commentaries”) 4 Language: Dutch, German, English and French 5 Population: General population 6 Setting: Environmental exposure at home (No exposure to noise in occupational setting nor in health care setting (e.g., hospital)) 7 Study design: Study design: observational studies only (No experimental studies following the WHO protocol) 8 Relevant outcome: Annoyance (s pe c if ically highly annoyed)

The UKHSA conducted an additional search in early 2022 to identify English language studies published between 2019 and 2022. The search protocol followed the same procedures as the 1 st update (See Appendix 1 for search terms), with the exception of the language requirement. The search and study selection process was documented for the 2 nd update by following the PRISMA reporting method (Appendix 2) [4].

2.3. Full text screen and eligibility criteria

For consistency with the WHO-commissioned systematic review [2] and guidelines [1], studies were screened in the 1 st and 2 nd update for inclusion based on the following criteria:

(1) Study type: cross-sectional or longitudinal surveys, using an explicit protocol for selecting

respondents. (2) Participants: Studies including members of the general population. (3) Exposure type: Long-term outside noise levels which are either expressed in L Aeq,24h , L dn ,

L den or its components, or can be easily converted from similar acoustic variables AND:

a. The level is based on a reliable calculation procedure, using the actual traffic volume,

composition, and speed per 24 h per road/railway/airport as input; OR

b. The level is based on measurements, or a combination of measurements and statistical

modelling, which were judged to be collected in a representative way across different time periods and capture sufficient variability across the study area. While the WHO commissioned systematic review [2] specified that measurement based studies would be excluded if the duration was less than seven days, we took a more nuanced approach [1] if the measurements were deemed likely to give a reliable estimation of the sound environment across space and time. (4) Outcome measure : The base of the outcome measure is the individual annoyance response

made during a standardized survey. The annoyance question and the response format either follow the recommendations given by ICBEN and/or ISO TS 15666 directly or are very close to them. (5) The paper gives at least one original table, formula, or graph which can be used to derive an

ERR based on grouped data (i.e., %HA).

2.4. Data extraction

After reaching consensus on the studies included in the updated analysis, the following data were extracted, coded, and imported into tables (see Appendix 3): Author(s) and publication year; country/region; sample population; sample size and response rate; exposure source and type (and range); outcome type; equation, table, or figure representing data that can be used to construct an aggregated ERR of %HA. 2.5. Data processing and modelling

The WHO systematic review for annoyance conducted a comprehensive set of analyses to inform the grading of the evidence, including meta-analyses of study ERRs, correlation between noise levels and annoyance raw scores, Odds Ratios for a 10dB level increase, funnel plots to assess publication bias and investigations of sources of heterogeneity across studies. In this study we only look at aggregating study ERRs, as these constitute the most relevant input for conducting health impact analyses [1, 5].

Equations, tables of values, and/or graphical representations of the %HA as a function of L den in 5dB steps were used to estimate an aggregated ERR across studies. Where a noise metric other than L den was used, a conversion was applied based on Brink et al, 2018 [6]. Additionally, if a study reported the %HA within a given range, such as 55-60 dBA, the midpoint value was used (e.g., 57.5 dBA) as the assigned noise level. The data for the studies identified in the WHO systematic review were kindly provided by the original authors. The aggregated ERRs derived in the WHO systematic review were based on a quadratic regression between L den and the individual study ERRs, which were weighted according to the square root of the respective study sample size. We carried out separate quadratic regression analysis on both the weighted and unweighted datasets (see Section 4 for further details). The regression analysis was carried out in Microsoft 365 Excel using the LINEST function. 3. RESULTS

The number of publications retrieved and abstract/title screened in the 1 st updated search are described in detail in van Kamp et al., 2020 [3]. For the 2 nd updated search, 220 publications were initially retrieved from the bibliographic databases, and after 92 duplicates were removed, 128 remained for title and abstract screening. At this stage, 107 publications were removed as they were not relevant to the analysis, a further two were removed as they were not published in English, and one was removed as it was already captured in the 1 st update.

A total of 33 publications from the 1 st and 2 nd updates were screened on full text against the inclusion criteria. After the screening process, seven [7-13] and eight [14-21] publications on road and/or rail traffic and annoyance were included from the 1 st (2014-2019) and 2 nd (2019-2022) updated

searches, respectively. Furthermore, one paper [22] which was published in 2016 but not captured in the 1 st update was additionally included in the review, resulting in a total of 16 publications (Appendix 3). Six of the publications reported associations for both road and railway noise. We excluded the rail estimate from one paper from Canada [11] as the exposure assessment method was unlikely to capture intermittent (short-duration) rail noise events. One publication from Japan included associations for 29 individual studies of road and railway (conventional and highspeed) noise, with data collected between 1994-2017: we included the results from these individual studies if they were published after the year 2000.

Half of the publications reported studies carried out in Europe (n=8), while the remaining publications covered Asia (n=5), Africa (n=1), South America (n=1) and North America (n=1). Six publications used a 72% cut-off for highly annoyed based on an 11-point numeric scale, six used a 60% cut off based on a 5-point verbal scale, and three publications used a 5-point verbal scale but weighted the responses to assign 72% of the participants into the highly annoyed category. All these are valid methods of defining %HA according to ISO/TS 15666:2021 [23]. An additional study used a 75% cut off based on a 4-point verbal scale. Appendix 3 lists our chosen source (equation, figure or table) for the annoyance ERR for each study.

We derived aggregated ERRs for road traffic and conventional railway noise. For road traffic noise we restricted the analysis to the exposure range 40-80 dB L den , as only the Hyena studies [24] had associated %HA values at 85dB L den . For railway noise we restricted the analysis to 40-85 dB L den . We excluded the estimates for high-speed rail [18, 19] from the updated aggregated rail ERR analysis because the ERRs for conventional rail and high speed rail have been shown to be dissimilar [19]. Potential causes of this dissimilarity could be from an infrastructure change effect, elevated tracks, proximity of residential properties to railway tracks, and perceptible vibrations [2, 19]. The identified evidence base for high speed rail ERRs originated from only one country (Japan Shinkansen), and an aggregated high-speed rail ERR based on these studies may have limited generalisability to other countries.

3.1 Aggregated exposure-response relationships

Figure 1 shows the scatterplot and quadratic regressions of the relation between L den and %HA for road traffic noise. For the majority of studies, the datapoints represent predicted values estimated from the regression analysis for the respective study. For exposures below 55dB L den , most of the datapoints for studies published post-2014 fall below the WHO aggregated curve. At higher exposure levels, the datapoints pre- and post- 2014 exhibit a similar spread of %HA. Therefore, our aggregate regression curve predicts a lower %HA below 55dB L den , and slightly higher %HA above 65dB L den , compared to the WHO ERR. This trend was also observed by the authors of the WHO systematic review when they excluded Alpine and Asian studies from their analysis [2]. The weighted and unweighted regression analysis gave very similar results, with some slight deviations at the highest exposures. The variance explained by the unweighted quadratic regression is R 2 = 0.518 (squared fit), which is slightly lower than that reported in the WHO systematic review (0.546). Figure 2 shows the scatterplot and quadratic regressions of the relation between L den and %HA for railway traffic noise. The majority of datapoints from studies post 2014 for noise levels above 60dB L den fall above the WHO aggregate curve, and our aggregate curve predicts a higher %HA for all exposures above 45dB L den . The weighted and unweighted regression analysis gave very similar results, with some slight deviations at the highest exposures. The variance explained by the unweighted quadratic regression is R 2 = 0.789 (squared fit), which is similar to that reported in the WHO systematic review (0.79).

Figure 1: Scatterplot and quadratic regression of the relation between L den and the calculated %HA for road traffic noise. Blue dots refer to studies identified by the WHO systematic review. Orange squares and red triangles refer to studies identified from the (2014- 19) and (2019-22) publication periods, respectively. The dashed lines are from the regression on all studies.

Ae te a 4 ee pavioulig, ye hy o oat val] th Leal ene pakou a, hy

Figure 2: Scatterplot and quadratic regression of the relation between L den and the calculated %HA for conventional railway traffic noise. Blue dots refer to studies identified by the WHO systematic review. Orange squares and red triangles refer to studies identified from the (2014-19) and (2019-22) publication periods, respectively. The dashed lines are from the regression on all studies.

The equation for the derived aggregate ERRs can be expressed as:

2 (1)

Estimated %HA = 𝑎 0 + 𝑎 1 𝐿 den + 𝑎 2 𝐿 den

where the polynomial coefficients for road and railway traffic noise are given in Table 2.

2S ASE ERISZISS007 9

Table 2. Aggregate ERR regression equation polynomial coefficients. Noise source quadratic regression coefficients for eq. (1)

regression

R 2 a 0 a 1 a 2 road traffic 57.256 -2.5731 0.0312 0.518 railway traffic 39.216 -2.1835 0.0311 0.789 4. DISCUSSION

unweighted analysis

We carried out regression analysis on exposure response relationships (ERRs) derived by studies on annoyance from road and railway traffic noise published between 2000 and 2022, using similar methods to the WHO ENG2018 systematic review on annoyance [2]. The aggregated studies considered in this paper represent a near doubling in number of survey respondents for road traffic noise (N ≈ 68,000) and a near tripling for railway traffic noise (N ≈ 29,000). Furthermore, the updated road traffic dataset represents a more diverse range of geographies, including studies from North America, South America, and Africa. Our updated search was limited to studies published in English. Conducting regression analysis on modelled ERR curves (as opposed to the original individual observed data) is inherently sensitive to the curve fitting methodology of the individual studies. For example, some studies force the ERR to converge to zero at the lower exposure levels. This masks any increases or levelling off in observed high annoyance at the lowest exposures, a phenomenon that can be observed in at least three studies included in this review [7, 13, 20]. Therefore, we hypothesise that our aggregate ERRs likely underestimate the true observed %HA at the lowest noise exposures. The two literature searches post-WHO identified a higher proportion of studies which characterized noise exposure based on measurements (sometimes combined with statistical land use regression models) and predominantly conducted outside of Europe. Setting and reviewing inclusion criteria for such studies is more complicated, as the risk of exposure misclassification depends on a number of factors, including the temporal and spatial variation of the noise source, measurement duration, topography, building density, and the location of the respondents in relation to measurement locations. However if inclusion criteria penalise studies based on measured exposure, aggregated ERRs would likely be biased towards European results (countries in the EU are required to carry out strategic noise mapping [25]), and potentially limit the global generalizability of the curve. We did not assign a quality rating to each of the included studies, although we did observe variations across studies in the quality and/or lack of reporting of specific aspects, such as the exposure modelling/measurement, how the participants sample was drawn, and response rates. The WHO systematic review used the square root of the sample size as a proxy for study quality to generate a weighted regression curve. However, some of the larger studies, whilst considered “good quality”, had specific issues that are relevant to the generalisability of the aggregated ERR. For example, the study by Yli-Tuomi et al. [20] ( N =6,754) assessed noise annoyance specifically when indoors with the windows closed, as this aligned better with the noise policy requirements for Finland. Given that weighting based on the square root of study size resulted in a very small effect on the derived ERRs, we suggest the use of the aggregate ERRs derived from unweighted studies. The publication of the WHO systematic review on annoyance sparked an academic debate [26, 27] on the specific methodology for conducting the ERR meta-analysis. One of the debated topics was the study inclusion criteria, including annoyance questionnaire wording, and whether respondents had to be representative of the general population or members of the general population. In our systematic review we noticed that very tight inclusion criteria would have resulted in only a handful of studies being screened in. We argue that every socio-acoustic survey will have its unique characteristics, and these differences contribute to the generalisability of an aggregate ERR (as long as multiple studies do not have the same systematic deviations from the “ideal” criteria). The merits of deriving an aggregated ERR when heterogeneity is very high is a matter of debate [26, 27]. Guski et al. found that a large part of the total variance in the regression analysis conducted for the WHO is due to “true” variance between studies [2]. The WHO ENG 2018

acknowledge that this is an important issue when considering the transferability and generalizability of ERRs, and recommend that data and ERRs derived in a local context should be used whenever possible [1]. Our literature review identified three large-scale studies carried out in Denmark, Japan and Switzerland [7, 19, 22] for this purpose. However relatively few countries have access to such datasets. For example, in the UK there have been no large-scale socio-acoustic studies for road and railway traffic noise since 1984 [28]. The WHO ENG 2018 note that generalized ERRs can be used where “local” data is not available. One may question whether generalized ERRs should include all international studies that meet certain quality criteria, or whether it would be better to “pick and choose” studies to match a specific situation and context. However, defining a robust set of criteria a priori for this purpose is likely to be very challenging. For example the type/age of vehicles/rolling stock, ventilation and sound insulation characteristics of housing stock, and cultural acceptance of a specific noise source, can vary considerably between countries within the same continent [1, 29-31] (and sometimes even between different regions within the certain country). We consider an aggregated ERR based on all eligible international studies as a strength of a generalised curve that reflects a wide range of situations and contexts. Given that our aggregated ERRs for road and railway noise deviate from those derived in the WHO systematic review, the question of how often ERRs should be updated inevitably arises. On the one hand, policy and decisions on new transport infrastructure should be based on sound, up-to-date evidence. On the other hand, infrastructure development needs a degree of policy stability, especially if any operating restrictions are linked to predicted health impacts. Changes to the ERR can also impact project cost-benefit analyses [32], with repercussions on the business case for a specific intervention. The answer to the question about update frequency depends somewhat on whether there is a time-invariant “true” ERR, that we can get closer to approximating as the number and quality of studies included in a meta-analysis increases. We argue that this is not likely to be the case. Noise annoyance is a complex multi-faceted human response [2], associated, modified, and moderated by both noise exposure and non-acoustic factors [33]. There are many reasons why annoyance reactions may vary over time [28, 34], and in our view regression analysis should be limited to recent studies. We followed the WHO systematic review’s approach of not including studies pre-2000 in the regression analysis. For future updates the chosen time period needs to balance the need for only including recent evidence whilst ensuring a sufficient number of studies to make the evidence representative and generalisable. Although we did not conduct a formal grading of the quality of evidence, the evidence published since 2014 still exhibits a large amount of heterogeneity between studies, and there are still inconsistencies in the criterion for defining highly annoyed. Therefore, the GRADE rating is likely to remain “low” (road) or “moderate” (rail), as per the WHO systematic review grading. In line with the WHO systematic review approach [2], we did not estimate confidence intervals around the aggregated ERRs, though the variability between study ERRs is clearly shown by the scatter plots in Figures 1 and 2. Confidence intervals would have been useful in determining whether our aggregated ERRs are statistically significantly different from those derived by the WHO systematic review. This highlights the need for a centralised repository of individual annoyance responses from socio-acoustic surveys carried out across the world, an initiative currently being explored by ICBEN [35]. 5. CONCLUSIONS

We carried out regression analyses on exposure response relationships between road and railway traffic noise (expressed in L den ) and the percentage of people highly annoyed, derived from studies published between 2000 and 2022. Our aggregated ERRs deviate from the ERRs derived in the systematic review that informed the WHO Environmental Noise Guidelines 2018, especially at low exposures for road noise and at high exposures for railway noise. We argue that it is important that generalised ERRs for annoyance from transport noise are updated regularly as newer evidence is published.

6. ACKNOWLEDGEMENTS

The authors would like to thank Rainer Guski for kindly providing the dataset used in the WHO systematic review for annoyance. 7. REFERENCES

1. WHO, Environmental Noise Guidelines for the European Region . 2018. 2. Guski, R., D. Schreckenberg, and R. Schuemer, WHO Environmental Noise Guidelines for the European Region: A Systematic Review on Environmental Noise and Annoyance. International Journal of Environmental Research and Public Health, 2017. 14 (12). 3. van Kamp, I., et al., Evidence Relating to Environmental Noise Exposure and Annoyance, Sleep Disturbance, Cardio-Vascular and Metabolic Health Outcomes in the Context of IGCB (N): A Scoping Review of New Evidence. International Journal of Environmental Research and Public Health, 2020. 17 (9). 4. Moher, D., et al., Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ, 2009. 339 : p. b2535. 5. van Kamp, I., et al., Study on methodology to perform environmental noise and health assessment . 2018, National Institute for Public Health and the Environment (RIVM). p. 112. 6. Brink, M., et al., Conversion between noise exposure indicators Leq24h, LDay, LEvening, LNight, Ldn and Lden: Principles and practical guidance. International Journal of Hygiene and Environmental Health, 2018. 221 (1): p. 54-63. 7. Brink, M., et al., A survey on exposure-response relationships for road, rail, and aircraft noise annoyance: Differences between continuous and intermittent noise. Environment International, 2019: p. 277-290. 8. Bunnakrid, K., T. Sihabut, and W. Patthanaissaranukool, The relationship between road traffic noise and annoyance level in Phuket province, Thailand. Asia-Pacific Journal of Science and Technology, 2017. 22 . 9. Licitra, G., et al., Annoyance evaluation due to overall railway noise and vibration in Pisa urban areas. Science of the Total Environment, 2016. 568 : p. 1315-1325. 10. Pennig, S. and A. Schady, Railway noise annoyance: Exposure-response relationships and testing a theoretical model by structural equation analysis. Noise & Health, 2014. 16 (73): p. 388-399. 11. Ragettli, M.S., et al., Annoyance from Road Traffic, Trains, Airplanes and from Total Environmental Noise Levels. International Journal of Environmental Research and Public Health, 2016. 13 (1). 12. Sung, J.H., et al., Relationship of Transportation Noise and Annoyance for Two Metropolitan Cities in Korea: Population Based Study. Plos One, 2016. 11 (12). 13. Vianna, K.M.D., M.R.A. Cardoso, and R.M.C. Rodrigues, Noise pollution and annoyance: An urban soundscapes study. Noise & Health, 2015. 17 (76): p. 125-133. 14. Bouzid, I., A. Derbel, and B. Elleuch, Factors responsible for road traffic noise annoyance in the city of Sfax, Tunisia. Applied Acoustics, 2020. 168 . 15. Gilani, T.A. and M.S. Mir, A study on the assessment of traffic noise induced annoyance and awareness levels about the potential health effects among residents living around a noise-sensitive area. Environmental science and pollution research international, 2021. 28 (44): p. 63045-63064. 16. Lechner, C., C. Kirisits, and S. Bose-O'Reill, Combined annoyance response from railroad and road traffic noise in an alpine valley. Noise & health, 2020. 22 (104): p. 10-18. 17. Lechner, C., D. Schnaiter, and S. Bose-O'reilly, Combined effects of aircraft, rail, and road traffic noise on total noise annoyance-a cross-sectional study in Innsbruck. International Journal of Environmental Research and Public Health, 2019. 16 (18): p. 3504. 18. Morihara, T., S. Yokoshima, and Y. Matsumoto, Effects of noise and vibration due to the hokuriku shinkansen railway on the living environment: A socio-acoustic survey one year after the opening. International Journal of Environmental Research and Public Health, 2021. 18 (15): p. 7794. 19. Yokoshima, S., et al., Representative exposure-Annoyance relationships due to transportation noises in Japan. International Journal of Environmental Research and Public Health, 2021. 18 (20): p. 10935. 20. Yli-Tuomi, T., et al., Exposure-Response Functions for the Effects of Traffic Noise on Self-Reported Annoyance and Sleep Disturbance in Finland: Effect of Exposure Estimation Method. International Journal of Environmental Research and Public Health, 2022. 19 (3): p. 1314. 21. Puyana-Romero, V., et al., Community Response to Noise from Hot-Spots at a Major Road in Quito (Ecuador) and Its Application for Identification and Ranking These Areas. International Journal of Environmental Research and Public Health, 2022. 19 (3): p. 1115. 22. Fryd, J. and T.H. Pedersen, Noise annoyance from urban roads and motorways , in Inter.noise . 2016: Hamburg. 23. ISO, ISO/TS 15666:2021 , in Acoustics — Assessment of noise annoyance by means of social and socio-acoustic surveys . 2021. p. 18.

24. Babisch, W., et al., Annoyance due to aircraft noise has increased over the years-Results of the HYENA study. Environment International, 2009. 35 (8): p. 1169-1176. 25. European Comission. Environmental Noise Directive (as amended). [cited 2022 03/05]; Available from: https://ec.europa.eu/environment/noise/directive_en.htm . 26. Guski, R., et al., Comment on Gjestland, T. A Systematic Review of the Basis for WHO's New Recommendation for Limiting Aircraft Noise Annoyance. Int. J. Env. Res. Pub. Health 2018, 15, 2717. International Journal of Environmental Research and Public Health, 2019. 16 (7). 27. Gjestland, T., A Systematic Review of the Basis for WHO's New Recommendation for Limiting Aircraft Noise Annoyance. International Journal of Environmental Research and Public Health, 2018. 15 (12). 28. Fenech, B. and G. Rodgers, Valuing impacts of noise on health-exposure response relationships in current UK guidance and the WHO Environmental Noise Guidelines 2018 , in ICA . 2019: Aachen, Germany. 29. COST Action TU0901 – Building acoustics throughout Europe. Volume 1: Towards a common framework in building acoustics throughout Europe , ed. B. Rasmussen and M. Machimbarrena. 2014. 30. ACEA. Average age of the EU vehicle fleet, by country . 2020; Available from: https://www.acea.auto/figure/average-age-of-eu-vehicle-fleet-by-country/ . 31. Kang, J., Urban Soundscape , in Urban Sound Environment . 2007. 32. HM Treasury, The Green Book: appraisal and evaluation in central government . 2013, UK Government. p. 148. 33. Fenech, B., et al., Development of a new ISO Technical Specification on non-acoustic factors to improve the interpretation of socio-acoustic surveys , in ICBEN . 2021: Stockholm, Sweden. 34. Guski, R., R. Schuemer, and D. Schreckenberg, Aircraft noise annoyance - Present exposure-response relations , in Euronoise . 2018. 35. ICBEN. International Comission on Biological Effects of Noise . [cited 2022 03/05]; Available from: http://www.icben.org/index.html .

8. APPENDICES

8.1. Appendix 1: Bibliographic database search strings

Table A1. Scopus searched Feb 4 th 2022. TITLE-ABS-KEY (noise W/5 (rail* OR aircraft OR airport* OR road* OR traffic* OR automobile* OR vehicle* OR motorcycle*)) AND (TITLE(annoyance) OR KEY(noise-annoyance)) AND PUBYE AR > 20 19

Table A2. EMBASE (via OVID). Searched on Feb 14 th 2022. Query Results #11 Limit 10 to yr=“2019-current” #10 #6 AND #9 #9 #8 OR #7 #8 annoyance.ti #7 exp annoyance/ #6 (#1 OR #2 OR #3 OR #5) AND (#1 OR #4 OR #5) #5 exp traffic noise/ #4 *noise/ OR *sound/ #3 exp railway/ OR exp motor vehicle/ #2 exp *“traffic and transport”/ #1 noise Adj5 (rail* OR road* OR traffic* OR automobile* OR vehicle* OR motorcycle* OR transport*)

Table A3. PubMed (MEDLINE). Searched on Feb 4 th 2022. # Search 1 noise[tiab] AND (rail*[tiab] or road*[tiab] or traffic*[tiab] or automobile*[tiab] or vehicle*[tiab] or ʺ motor cycle* ʺ [tiab] or motorcycle*[tiab] or transport*[tiab]) 2 ʺ Transportation ʺ [majr] 3 ʺ Railroads ʺ [mh:noexp] OR ʺ Motor vehicles ʺ [mh] 4 ʺ Noise ʺ [mj:noexp] 5 ʺ Noise,transportation ʺ [mh] 6 (#1 OR #2 OR #3 OR #5) AND (#1 OR #4 OR #5) 7 annoyance[ti] 8 #6 AND #7 9 #8 AND 2019:2022[dp]

Table A4. PyscINFO. Searched on Feb 4 th 2022. # Searches

1 (noise adj5 (rail* or road* or traffic* or automobile* or vehicle* or motorcycle*)).mp. [mp=title, abstract, heading word, table of contents, key concepts, original title, tests & measures] 2 traffic.mp. or railroad trains/ or transportation/ or motor vehicles/ 3 exp Noise Effects/ 4 exp Auditory Stimulation/ 5 (noise*).ti,ab. 6 (1 or 2) and (1 or 3 or 4 or 5) 7 (annoyance).mp. [mp=title, abstract, heading word, table of contents, key concepts, original title, tests & measures] 8 6 and 7 9 limit 8 to yr= ʺ 2019 ‐ current ʺ

8.2. Appendix 2: PRISMA flow chart for the 2 nd updated search.

Figure A1. PRISMA flow chart for the 2 nd update (papers published between 2019-2022).

Records identified through database search (n=220) Duplicates removed (n=128) Records screened Records excluded ‘on abstract and (n=110 title (n=128) ( ) Full text articles Full text articles assessed for excluded with eligability (n=18) reasons (n=10) Included 8 papers

8.3. Appendix 3: Characteristics of included studies Table A5. Cha r acteristics of included studies on road and/or railway traffic noise and annoyance.

Pub. Country (Study population)

Sample size (response rate %*); Sampling strategy

Exposure type and assessment

HA question (cut off %)

Source of ERRs 1 st update (2014-2019) [3] Brink et al, 2019 Switzerland (Adults aged 19-75)

5592 (31%) Stratified random sampling

Figure 5

Modelled road, rail, air traffic noise (L den and Intermittency Ratio) within range of 30 to 70 dBA

11-point numeric scale (72%) ICBEN/ISO TS 15666 wording/scale

Sung et al, 2016 South Korea (Adults)

967 (53%) – Ulsan 869 (47%) – Seoul 1836 - Total Stratified random sampling

Table 2

Modelled road and aircraft noise (L dn ) categorized into <55, 55- 65, and >65 dBA

11-point numeric scale (72%) ICBEN/ISO TS 15666 wording/scale

180 (65-75%) Unknown sampling procedure

Figure 8

de Paiva Vianna, K.M et al, 2015

Portugal (Adults aged 20 +)

Modelled road-traffic noise (L den ) within the range of 48.5 to 72.8 dBA

4-point verbal scale (75%)

Pennig et al, 2014 Germany (Adults aged 18-97)

320 (22%) Sampling based on proximity to noise source

Figure 4

Modelled rail traffic noise (L den ) within the range of 44 to 90 dBA

5-point verbal scale (72%) ICBEN/ISO TS 15666 wording/scale

Licitra et al, 2016 Italy (Adults aged 35 to 70)

119 (*) Random sampling of subjects who lived near noise measurement sites

Figure 7

Measured and modelled rail traffic noise (L den, L night ) within the range of <55, 55-60, 60-65, 65-70, and >70 dB

11-point numeric (72%) and 5-point verbal scale (60%) ICBEN/ISO TS 15666 wording/scale

Bunnakrid et al, 2017 Thailand (Unknown)

253 (*) Sampling based on proximity to measurement sites

Figure 5

Measured road-traffic noise 3 times for 24 hours at 9 locations and respondents lived around measurement locations for more than 1 year. Measurements (L dn ) within the range of 74 – 80 dBA

5-point verbal scale (60%) ICBEN/ISO TS 15666 wording/scale

Ragettli et al, 2015 Excluded the association for railway exposure from the analysis due to the 2- minute integrating time-window for sound level measurements.

4336 (46.8%) Stratified random sampling

Figure 4

Canada (Adults aged 18 yrs +)

Measured and modelled (via land use regression) environmental noise (road) (L Aeq24hr and L den ). Measurement data for models was based on at least one week of sampling at 204 locations. ERRs given within the range of 45 to 75 dBA

5-point verbal scale (60%)

2 nd update (2 0 19-2022) Bouzid et al, 2020 Tunisia (Youth and adults aged 13 – 79)

1272 (58%) Stratified random sampling

Figure 9

Measured road traffic noise (3, 15-minute long measurements at 221 sites) (L den, L day , L evening , L night ) within the range of 40 to 75 dBA

5-point verbal scale (60%) ICBEN/ISO TS 15666 wording/scale

Gilani et al, 2021 India (Adults aged 20 to 60+)

565 (*) Random sampling

Modelled road traffic noise (L dn ) within the range of 42.5 to 75 dBA

5-point verbal scale (60%) ICBEN/ISO TS 15666 wording/scale

Equation 2

Lechner et al, 2020 Austria (Adults aged 18 to 60+) Lower Inn Valley Route

1003 (54%) Stratified random sampling

Modelled road and rail traffic noise (L den , L night ) within the range of 35 to 65 dBA

11-point numeric scale (72%) ICBEN/ISO TS 15666 wording/scale

Figure 1 / Figure 4

Lechner et al, 2019 Austria (Adults aged 18 – 94) Innsbruck

1031 (48%) Stratified random sampling

Modelled road and rail traffic noise (L den ) within the range of 35 to 65 dBA

11-point numeric scale (72%) ICBEN/ISO TS 15666 wording/scale

Figure 7 / Figure 2 from Lechner 2020 Puyana-Romero et al, 2022

225 (60.2%) Sampling based on proximity to noise source

Table 4

Ecuador (Adults aged 18-80)

Modelled road traffic noise (L den ) within the range of 53 to 80 dBA, but ERRs valid within the range of 68 to 78 dBA

5-point verbal scale (60%) ICBEN/ISO TS 15666 wording/scale

Yli-Tuomi et al, 2021 Finland (Adults aged 25 yrs+)

7321 (45.7%) - Total 6754 - Road 3331 - Rail Random sampling

Modelled road traffic and rail noise (L den ) above or equal to 45 dB and up to ~ 73 dB for road and ~ 58 dB for rail

5-point verbal scale (60%) Modified ICBEN/ISO wording to specify being annoyed at the home environment and with windows closed

Table 3 / Figure 4

Yokoshima et al, 2021 Multiple individual studies RT: Road-traffic CR: Conventional rail HR: High speed rail Removed JPN014CR2002 (Takeshita et al, 2003) as suspected to contain the same data/population as the Yano et al, 2005 study which was included in the WHO Guski et al, 2017 review Removed HKR107HR2016 as included as separate publication (Morihara et al, 2021)

1601 (*) JPN011RT2000 Modelled road traffic noise (DENL) within the range of 45 to 75 dBA; Modelled conventional rail traffic noise (DENL) within the range of 35 to 75 dBA; Modelled highspeed rail traffic noise (DENL) within the range of 35 to 70 dB

Table 6 272 (*) JPN016RT2003 1358 (*) JPN021RT2004 371 (*) ISK101RT2007 189 (*) STM104RT2011 1442 (*) JPN012CR2001 1490 (*) JPN017CR2003 1357 (*) JPN021CR2004 601 (*) KMM102CR2009 1028 (*) KMM103CR2011 162 (*) STM104CR2011 693 (*) KMM108CR2016 1101 (*) JPN013HR2001 715 (*) JPN015HR2003 1306 (*) JPN018HR2003 174 (*) JPN022HR2005 1031 (*) KMM103HR2011 293 (*) NGN105HR2013 691 (*) KMM108HR2016

Japan (People aged <40, 40- 60 and >60 yrs)

5-point verbal scale (72%) ICBEN/ISO TS 15666 wording/scale

Morihara et al, 2021

1022 (52%) Sampling based on proximity to noise source

Japan (Youth and adults aged 10 to 70 yrs +)

Modelled highspeed rail traffic noise (L den ) within the range of <46, 46-49, 50-53, and <=53 dB

5-point verbal scale (72%) ICBEN/ISO TS 15666 wording/scale (for all)

Table 6 in Yokoshi ma 2021 Other studies

Fryd et al, 2016 Denmark (Adults)

6761 (48) Even sampling along motorways and urban roads

Figure 3

Modelled road traffic noise (L den ) within the range of 47.5 to 75 dB

11-point numeric scale (72%) ICBEN/ISO TS 15666 wording/scale

*Study did not report a survey response rate. In the case of Yokoshima et al, 2021 survey response rates are likely contained within individual studies if these are published.