Effects of traffic speed reduction interventions on noise-induced annoyance and self-reported sleep disturbances: a longitudinal study in Zurich Mark Brink Federal Office for the Environment, 3003 Bern, Switzerland Simone Mathieu and Stefanie Rüttener City of Zurich, Department of Health and Environment, 8001 Zurich, Switzerland ABSTRACT For the purpose of evaluating acceptance and effects of permanent speed reductions on noise level, noise annoyance and self-reported sleep disturbance, we surveyed about 1300 randomly sampled inhabitants, before and after a speed regime changeover from 50 km/h to 30 km/h along 15 small and mid-sized city streets in Zurich. Concurrently, individual noise exposure calculations based on traffic counts and on-site speed measurements were carried out. The results show a decrease of road traffic noise levels at the loudest façade point by an average of 1.6 dB during the day and 1.7 dB at night, a significant decrease of noise annoyance and of self-reported sleep disturbances as well as a significant but moderate increase of the perception of road safety. Most importantly, the exposure-response relationships for annoyance and sleep disturbance were shifted towards lower effects in the 30 km/h condition by, depending on receiver point, between about 2 and 4 dB during the day and about 4 dB at night, indicating lower effects at the same average level. We conclude that besides the lower average level alone, additional factors related to the lower driving speed must play a role in the reduction of annoyance and sleep disturbance.

1 INTRODUCTION

From a public health perspective, road traffic is one of the most important causes of injuries and non-communicable diseases – including those caused by air pollution and noise [1]. The adverse health effects of road traffic are even greater in urban areas, where both population and traffic densities are high and general mobility demands are increasing, partly as a result of projected population and job growth. As Switzerland's most populous city, Zurich is particularly affected by this problem. The legal exposure limits for road traffic noise are exceeded on a network of about 230 km length. In order to improve the health and quality of life of the city's residents and, above all, as a means of reducing noise, the city of Zurich has for several years pursued the goal of reducing road traffic noise in a first step by reducing the speed limit from 50 km/h to 30 km/h on parts of the street network. This is a cost-effective noise abatement measure with an estimated noise reduction potential of up to 3 dB for the average level (Leq) and up to approx. 5 dB for the maximum level [2, 3]. In recent years, the city has signalized a 30 km/h speed limit on almost 40 km of the street network with

iced ral inter.noise mee DOOD

excessive noise exposure. For the purpose of evaluating this measure, an empirical longitudinal survey and corresponding individual exposure calculations were carried out in the city in the years 2017-2020 among about 1300 randomly contacted residents living on a total of 15 street sections where the change from speed limit 50 km/h to speed limit 30 km/h took place in the corresponding time period. A before-after comparison was realized by interviewing the same persons twice, about one to two years apart. The study goals were to investigate whether and to what extent the speed reduction led to a reduction in average noise exposure, noise annoyance and self-reported sleep disturbance due to noise, as well as an increase in the perceived road safety in the respective street. In addition, it was investigated how exposure-response relationships differ between the 50 km/h and 30 km/h regime. The study focused in particular on the question of whether the (hypothesized) reduction in annoyance and noise-induced sleep disturbance as a result of the speed reduction can be explained solely by the achieved reduction in the average noise level or possibly also by other/additional factors.

2 METHODS

The study was carried out between 2017 and 2020 on 15 street sections in the city of Zurich that in this period were affected by the changeover from 50 km/h to 30 km/h speed limit. On these street sections, one randomly determined resident per household (N=3732) were contacted by post before the changeover (fill-out date of questionnaire on average 95 days before) and asked to complete a questionnaire on their perception of street traffic noise in their street. Respondents were then contacted again after the changeover (fill-out date of questionnaire on average 393 days after) and asked to complete another questionnaire (with largely the same items/questions). The participants were asked about the noise pollution they experienced and about their sleep disturbances in relation to a retrospective period of 6 months (before the questionnaire was completed, and as opposed to the usual 12-month period according to the ICBEN recommendation). For the after-survey, only persons who had returned the questionnaire of the previous survey (N=1311) were again contacted. This procedure allowed for a pairwise before-and-after comparison of responses, in other words, a repeated measures statistical design (Figure 1).

Figure 1: Study design with “before”-survey during the previous 50 km/h regime (orange bar) and “after”-survey with the same persons after the changeover to 30 km/h (blue bar) (in the after-survey, only those persons were contacted who had participated in the before-survey). A four-page paper and pencil questionnaire was used for the survey. The questionnaire was entitled "Survey on the perception of road traffic noise among residents of the city of Zurich" and was sent out together with a cover letter. A before and an after version of the questionnaire was created in each

6 months 6 months i, —— 60) [before 1130) [after | referring time period referring time period a = 1 year or more since changeover

case, which were identical apart from a few exceptions. The questionnaire started with some general questions (age, sex, family and housing circumstances, length of residence, household size, housing type, house/apartment ownership), bedroom location relative to street (facing street, sideways, or away from street) followed by questions about annoyance from different noise sources and about noise-induced sleep disturbance. The degree of annoyance was measured using both the verbal 5-point ICBEN scale with the scale values "not at all", "slightly", "moderately", "very", "extremely" [4, 5]. People with answers on one of the top two scale points ("very" or "extremely") were defined as "highly annoyed" (HA=1). The degree of noise-induced sleep disturbance was also measured using an 11-point scale. "Highly sleep disturbed” (HSD=1) were defined people with responses with a scale value ≥ 8. The exchange of the speed limit signalizations on the 15 street sections relevant for this study were carried out in several phases between 18 Sept 2017 and 17 May 2019 by the Traffic Department of the City of Zurich, according to their planned schedule. The mail-outs of the before and after questionnaires followed this schedule. In order the respondents had ample time to get used to the new lower speed regime at their street, the mail-outs for the after survey were executed only after a period of at least 300 days has passed after the changeover. The changeovers included new traffic signs, new pavement markings, and in few cases, the installation of speed bumps, but no further measures. On all street sections affected by the changeover, we calculated the respective street emission levels based on automated traffic counts. And based on these, we determined the exposure at the faintest (min), loudest (max), and bedroom façade for each individual dwelling unit, both in the before and after condition, for the daytime (Leq between 06 und 22 hrs) and nighttime period (Leq between 22 and 06 hrs). The immissions on the building façades and at the presumed bedroom window (identified for each case individually with the help of Google Streetview) were calculated with the software CadnaA, based on the digital terrain model of the official cadastral survey (DTM-AV) and the 3D city model of the city of Zurich. To have available the exposure data for both the before and after condition made it possible to elucidate to what extent the exposure-effect relationships differ between the regime with 30 km/h and the regime with 50 km/h speed limit signalization (see chapter 3.3). The (anonymized) data from the questionnaires were linked with the noise exposure data. The final dataset contained two records per person, one for the before and one for the after condition, with complete exposure values as well. The below analyses based on this dataset include descriptive statistics, inferential linear and hierarchical logistic models, and derived graphs, both on the exposure and effects side. The statistical analyses were carried out in R version 3.5.1 [6]. For all statistical analyses, the significance level was set at a value of 5% (p<0.05).

3 RESULTS

3.1 Response statistics

A total of 3732 questionnaires were sent out in the before survey (i.e. the first wave). Of these, 1311 were returned, resulting in a response rate of 35%. Since for the after-survey only persons were contacted who had returned the questionnaire of the previous survey, 1311 questionnaires were sent out for the after-survey. Of these, 886 were returned, thereof 880 with a validated address. In the end, we had paired data for 880 individuals, i.e. 1660 data records for statistical analysis of before-after changes.

3.2 Noise exposure before and after the changeover

Table 1 shows mean values and standard deviations of the measured speeds during the before and after survey (averaged over all street sections and associated subsections; total N=48). As can be seen, the new speed limit was well respected on average.

Table 1: Mean values and standard deviations of the average speed (V average ) during the day and night in the 50 km/h and the 30 km/h speed regime, and their differences [in km/h].

50 km/h day 30 km/h day Difference day 50 km/h night 30 km/h night Difference night

Mean value 39.88 30.97 8.91 41.23 31.51 9.72

Standard dev. 1.99 2.69 2.78 1.90 2.70 2.26

The density distributions of the calculated exposure levels during the day and at night are shown in Figure 2. The mean achieved reductions of the exposure amounted to 1.58 dB during day (loudest façade) and 1.33 dB during night (bedroom façade). The average level reduction of slightly less than 2 dB corresponds with the average speed reduction of just under 10 km/h, which is quite in line with expectations.

2 Den: 0.075 0.050 Day Speed regime 50 km/h 30 km/h a 0.04 Density 0.02 Night Speed regime

Figure 2: Density distribution of the average road traffic noise level during the day (left) and at night (right) for the two different speed regimes. The dashed lines indicate the respective mean value. In this plot, the distribution of LDay and LNight, respectively, is not mapped in discrete categories but estimated continuously, with the area under the density distribution normalized to the value 1.

3.3 Scores for annoyance, self-reported sleep disturbance, and self-assessed road safety in the before (50 km/h) and after (30 km/h) condition

The following analyses show before-after comparisons of the main dependent variables, i.e. annoyance, self-reported sleep disturbance, and self-assessed road safety at residence. Annoyance and sleep disturbance. Figure 3 shows mirrored histograms of the frequency of selected response categories for the variables annoyance and noise-induced sleep disturbance in the before (50 km/h) and after (30 km/h) condition. This form of presentation makes it possible to directly compare the change in the frequency of a particular scale value between the before and after surveys. In particular, it becomes apparent that higher annoyance values ("very" and "extremely" on the 5- point scale) occur only about half as often during the 30 km/h speed regime as during the 50 km/h speed regime. A similar effect can be observed with sleep disturbance, measured on an 11-point scale (Figure 3, left).

——w = 7 50 km/h 30 km/h 0.000 0.00 35 40 45 50 55 60 65 70 25 30 35 40 45 50 55 60 65 LDay [dB] max. fagade LNight [dB] at bedroom fagade

300

200

200

100

Number of responses

Number of responses

100

0

not_at_all slightly moderately very extremely

0

100

-100

Speed regime:

200

Speed regime:

50 km/h 30 km/h

50 km/h 30 km/h -200

300

0 1 2 3 4 5 6 7 8 9 10 Sleep disturbance [11-point scale]

0 1 2 3 4 Annoyance [05-point scale]

Figure 3: Mirrored histograms showing the distribution of annoyance responses on the 5-point ICBEN scale (left) and of noise-induced sleep disturbance responses on an 11-point scale (right), during the before (50 km/h, red) and the after survey (30 km/h, green). N=2×880 Figure 4 shows the distributions of annoyance and sleep disturbance ratings in the before and after surveys as violin plots with mean (●). Mean differences between the speed regimes were examined using two-tailed t-tests for dependent samples and were both significant (Annoyance: ∆=0.2 points on the 5-point scale, t=5.7629, df=838, p<.01; Sleep disturbance: ∆=0.55 points on the 11-point scale, t=6.3577, df=854, p<0.01).

Annoyance

Sleep disturbance

10

4

9

Sleep disturbance [11-point scale]

8

Annoyance [05-point scale]

3

7

6

5

2

4

3

1

2

1

0

0

Speed 50 km/h Speed 30 km/h

Speed 50 km/h Speed 30 km/h

Figure 4: Density distribution of raw scores for annoyance (5-point ICBEN scale) and noise induced sleep disturbance (11-point scale) in the before (50 km/h) and after (30 km/h) surveys. Violin plots with mean (●). Road safety. Respondents were asked in the before and after survey about their subjectively perceived road safety in the street affected by the new speed limit. The response scale included 5 levels ("not safe", "not much safe", "moderately safe", "rather safe", "very safe"). In order to determine whether the average subjective perception of traffic safety increased in a statistically significant manner after the change to 30 km/h, the scale was transformed into a 1-5 interval scale (for the sake of simplicity, the five verbal scale values were assumed to be equally spaced) and the difference in means was tested using a t-test. The result shows a slight increase in the perception of road safety at 30 km/h compared to 50 km/h, with the difference in mean proving to be significant (two-tailed t-test for dependent samples: ∆=0.25 points, t=7.714, df=851, p<0.01). A corresponding violin plot is shown in Figure 5.

Percieved road safety

very safe

Estimated Road Safety

rather

moderately

not much

not safe

Speed 50 km/h Speed 30 km/h

Figure 5: Perceived road safety during the before (50 km/h) and after (30 km/h) conditions. Violin plots with mean (•).

3.4 Exposure-effect relationships for %HA and %HSD in the 50 km/h and 30 km/h condition

The analyses presented so far show the observed changes in the mean and the distribution of some key study variables. They indicate a decrease in annoyance and self-reported sleep disturbance, and a slight but significant increase in perceived road safety after switching from the 50 km/h to the 30 km/h speed regime. However, these analyses do not show whether and to what extent the exposure- response relationships for annoyance and self-reported sleep disturbance changed or shifted after the changeover. Thus, the remaining question is whether the observed reduction in annoyance and sleep disturbance is adequately explained by the reduction in the average exposure alone.

Percent highly annoyed (%HA) 30 10 50 55 60 LDay [dB] max. facade 70

Figure 5: Exposure-response curves for the percentage highly annoyed (%HA, left) and highly sleep disturbed (%HSD, right) during the speed regime 50 km/h (before intervention) and speed regime 30 km/h (after intervention). Scales used – Left: 5-point ICBEN scale with cutoff point 60%; Right: 11-point ICBEN- type scale with cutoff point 73%. Statistical modelling: Multilevel logistic regression with adjustment for age and sex; plotted are the centered curves (at the mean of age and sex) and the 95% confidence intervals It is quite possible that, in addition to the reduction of the average exposure level, factors (acoustic and non-acoustic) such as less rapidly rising pass-by levels or lower maximum levels additionally reduce annoyance and/or sleep disturbance. In this case, the entire exposure-response curves would have to shift toward lower effects under the 30 km/h regime.

Percent highly sleep disturbed (%HSD) 50 40 30 10 25 30 35 «40 45 50 55 60 LNight [dB] bedroom facade

To test this, we calculated corresponding statistical models and plotted the respective exposure- response curves for %HA and %HSD for the "before" condition (50 km/h speed limit) and the "after" condition (30 km/h speed limit), respectively, side by side in Figure 5 above. To do so, hierarchical logistic regression models were computed, with the speed regime condition (i.e., whether the questionnaire was completed before or after the change to 30 km/h) defined as a fixed and the respondent as a random intercept effect. Parameter estimates of these models are given in Table 2. Table 2: Parameter estimates of the logistic models for %HA regressed on LDay at the loudest façade and %HSD regressed on LNight at the bedroom façade, and other predictors. Legend: B=unstandardized coefficient; SE=standard error of B; VIF=variance inflation factor. Significant p-values are highlighted in bold.

Effect B SE p-value VIF

Model for highly annoyed (%HA):

Intercept -15.6176 1.9451 <0.01 0.00 Speed regime 30 km/h (vs. 50 km/h) -0.5173 0.1616 <0.01 1.02

LDay in dB (loudest façade point) 0.2373 0.0314 <0.01 1.02

Age 0.0016 0.0065 0.80 1.01 Male sex (vs. female) 0.1583 0.2260 0.48 1.01

Model for highly sleep disturbed (%HSD):

Intercept -8.6125 1.2132 <0.01 0.00 Speed regime 30 km/h (vs. 50 km/h) -0.5607 0.2023 <0.01 1.00 LNight in dB (bedroom façade) 0.1434 0.0225 <0.01 1.02 Age -0.0169 0.0078 0.03 1.01 Male sex (vs. female) -0.2472 0.2646 0.35 1.02 Figure 5 above shows that for both outcomes %HA and %HSD the two curves are shifted against each other. In both models, the variable speed regime turned out to be significant. This means that, with an average exposure level remaining constant, the proportions of both highly annoyed and highly sleep disturbed were higher in the 50 km/h speed regime. The 30 km/h regime therefore reduced these outcomes irrespective of the average exposure levels. Expressed in dB, this shift of the curves can be determined as the ratio of the unstandardized regression coefficients for the effect of the speed regime (see Table 2) to the coefficient for the effect of the average exposure level (see Table 2), which results in a shift of 2.18 dB for %HA on the loudest façade and by 3.91 dB for %HSD at the bedroom façade. These figures clearly show that in the 30 km/h condition, the percentage of highly annoyed as well as the percentage highly sleep disturbed at the same level is lower than during the 50 km/h speed regime. Thus this decrease, obviously, cannot be explained simply by the lower average noise exposure (LDay or LNight respectively) in the 30 km/h situation, because the effect of (lower) noise exposure is already statistically accounted for in these models. Effect modification. In the course of modeling exposure-response relationships, the potential influence of a range of effect modifiers on %HA was investigated, which yielded the following results: length of residence, soundproof windows, and time spent on a private outdoor patch (e.g. balcony) showed no significant modifying effect on %HA. Furthermore, no effects biasing the HA results could be found for the weather (operationalized as individual average outdoor temperature over 90 days before questionnaire fill out) and the Corona crisis (operationalized as number of daily

new cases in a 30 or 90-day period before questionnaire fill out) which already prevailed for about half of the respondents during the after survey. However, we found a strong significant effect for the orientation of the bedroom towards the nearest street. Figure 6 below shows that the new speed regime mainly benefitted people whose bedrooms were oriented towards the street or to the side of the street. For persons whose dwelling (or at least whose bedroom) points away from the street (e.g. faces an inner courtyard), the introduction of the new speed limit did not cause a shift in the exposure-relationship relationship. Table 3 lists the parameter estimates for the curves underlying Figure 6.

Figure 6: Exposure-response curves for %HA at the loudest façade point during the day period for the two speed regimes and bedrooms (dwellings) oriented towards/facing the street (left), pointing sideways to street (middle), or away from the street (right), determined using the 5-point ICBEN scale. Basis: hierarchical logistic model with adjustment for age and sex. Shown are the centered curves (at the mean of age and sex) and the 95% confidence interval.

Table 3: Parameter estimates of the logistic model for the probability of being highly annoyed (HA, determined using the 5-point ICBEN scale with cutoff point 60%), regressed on LDay at the loudest façade point and other predictors, with bedroom orientation as an effect modifier . Legend: B=unstandardized coefficient; SE=standard error of B; VIF=variance inflation factor. Significant p-values are highlighted in bold.

Effect B SE p-value VIF

Intercept -13.3332 1.9321 <0.01 0.0000 Speed regime 30 km/h (vs. 50 km/h) -0.7784 0.2610 <0.01 1.58

LDay in dB (loudest façade point) 0.2139 0.0314 <0.01 1.03 Age 0.0022 0.0066 0.74 1.00 Male sex (vs. female) 0.1565 0.2290 0.50 1.01 Bedr. orientation sideways (vs. on street) -0.8124 0.3292 0.01 1.19

Bedr. orientation away from street (vs. on street) -2.1235 0.3482 <0.01 "

Interaction term: 30 km/h speed regime x orientation sideways 0.0228 0.3928 0.95 1.40

Interaction term: 30 km/h speed regime x orientation away from street 0.8031 0.3854 0.04 "

WHA 40 50 0 towards/facing street [N=259] ‘sideways to street [N=243] away from street [N=338] ‘Speed regime: - ‘Speed regime: = ‘Speed regime: — 50 kmh F] | — sok S} | — sok — 30kewn @ | | — 30k | | — 30k <8 8 Fe & 4 4 50 55 60 65 70 4 45 50 85 60 6570 4 45 50 55 60 65 70 LDay [48] max. facade LDay [48] max. facade LDay {4B] max. facade

4 DISCUSSION

Empirical studies on the effects of changes in signalized speed limits are very scarce, as the recent WHO systematic review of intervention studies in noise [7] could show. Thus, the present study is probably one of the few of its kind that used a longitudinal repeated measures design to investigate whether a speed reduction to 30 km/h (from a previous 50 km/h) also results in a reduction of adverse effects of traffic noise and, most importantly: whether exposure-response relationships differ between the two speed regimes. Our study showed a post-changeover decrease of road traffic noise levels at the loudest façade point by an average of 1.6 dB during the day and 1.7 dB at night. This level reduction is within the expected range for an average speed reduction of just under 10 km/h. While these values are certainly not very large, noise level reductions of this magnitude can be considered perceptible. Beyond the global effect of reducing the average exposure, the exposure-response relationships in the lower speed regime were shifted by a few dB toward smaller effects, i.e., lower annoyance and less sleep disturbance. We estimate this effect – depending on the receiver point – to be between about 2 and 4 dB during the day and about 4 dB at night. Thus, at the same average level, annoyance and noise induced sleep disturbances are lower at 30 km/h than at 50 km/h. This is an indication that, in addition to the decrease of the average level, other factors related to the lower driving speed additionally reduce noise annoyance and sleep disturbance. Candidates for such factors are the lower maximum levels in the 30 km/h condition and a less steep slope of rise of level during passbys of vehicles. The outcomes of potentially effect-modifying factors were also examined. Here it was found that especially those residents benefited from the introduction of the 30 km/h speed limit whose bedrooms were oriented towards the street, while for persons with apartments or bedrooms facing away from the street towards e.g. a backyard, the introduction of the 30 km/h speed limit did not result in an additional reduction of annoyance. No effects biasing the results could be found for the weather prevailing at the respective survey periods and the Corona pandemic (which already prevailed for some of the respondents during the followup phase after the changeover to the speed limit 30 km/h). Like all similar studies, this one is characterized by strengths and weaknesses. The strengths of the study certainly include the apllication of repeated measurements and the comparatively precise calculation of exposure in both the before and after surveys, especially on the bedroom façade. On the downside, the lack of a control group (i.e. a sub-sample of persons without speed regime change, but that are also interviewed twice) in this study must be regarded as a certain disadvantage. With a control group, it could have been examined whether, at best, an unobserved influencing factor could have been held responsible for the decrease in annoyance in the after survey compared to the before survey (and not the introduction of the new speed limit per se). Possible candidates for such (uncontrollable) influences are political events, media coverage of the topic of reducing speed limits, weather effects, and other unmeasured and therefore unknown influences. However, there were some underlying conditions during the initiation and planning of the study design that spoke against the implementation of a control group, which cannot be discussed in detail here. However, our analyses of possible biasing effects – some of them mentioned in Chapter 3.4 – did not indicate any serious overinterpretation of the decrease in annoyance and sleep disturbance due to the lack of a control group.

5 CONCLUSIONS

The present study clearly showed that city dwellers are relieved by a reduction of the permissible speed to 30 km/h (from a previous 50 km/h) on their residential street by a subsequent reduction of the Leq of 1.6 dB during the day and 1.7 dB at night. As a result, noise annoyance and noise-induced sleep disturbances are reduced in a statistically significant manner, and the perception of road safety also increases significantly, although not very markedly. Beyond the global effect of reducing the

average level (LDay and LNight), the exposure-response relationships between average level and effect shift by a few dB toward smaller effects, i.e., lower annoyance and less sleep disturbance. We (conservatively) estimate this effect to be at about 2 dB during the day and about 4 dB at night. Thus, at the same average level, annoyance and the sleep disturbances are lower at 30 km/h than at 50 km/h. Further information: A more detailed report (in German) of this study can be downloaded here .

6 REFERENCES

1. Nieuwenhuijsen, M.; Khreis, H., Advances in Transportation and Health . Elsevier: 2020. 2. EKLB Tempo 30 als Lärmschutzmassnahme Grundlagenpapier zu Recht – Akustik – Wirkung ; 2015. 3. Heutschi, K. Grundlagenpapier zu Tempo 30 auf Strassen: Teil B: Akustikgrundlagen. Untersuchungsbericht Nr. 5214.00.7157 [Dübendorf: Empa] ; Empa: Dübendorf, 2015. 4. Brink, M.; Giorgis-Allemand, L.; Schreckenberg, D.; Evrard, A.-S., Pooling and Comparing Noise Annoyance Scores and “High Annoyance” (HA) Responses on the 5-Point and 11-Point Scales: Principles and Practical Advice. International Journal of Environmental Research and Public Health 2021, 18, (14). 5. Fields, J. M.; De Jong, R. G.; Gjestland, T.; Flindell, I. H.; Job, R. F. S.; Kurra, S.; Lercher, P.; Vallet, M.; Yano, T.; Guski, R.; Felscher-Suhr, U.; Schumer, R., Standardized general-purpose noise reaction questions for community noise surveys: Research and a recommendation. Journal of Sound and Vibration 2001, 242, (4), 641-679. 6. R Core Team, R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. In 2018. 7. Brown, A.; van Kamp, I., WHO Environmental Noise Guidelines for the European Region: A Systematic Review of Transport Noise Interventions and Their Impacts on Health. International Journal of Environmental Research and Public Health 2017, 14, (8), 873.