Evaluation of the renovation of an urban highway viaduct using citi- zen science low-cost noise monitoring Luc Dekoninck 1 Ghent University – Departement of Informationtechnology - Waves Technologiepark Zwijnaarde 126, 9052 Ghent, Belgium

ABSTRACT A main highway in Belgium (E17) transect the urban area of the city of Ghent with a viaduct since 1970. This highway is the main north-south artery for heavy traffic from and to the ports of Antwerp (BE) and Rotterdam (NL).The viaduct has significant maintenance issues and was subjected to mul- tiple noise control actions in the past. Since 2014 a local citizen initiative named ‘Viadukaduk’ - mimicking the sound of the viaduct- is organizing the public engagement to resolve the environmental issues in a structured approach. The condition of the viaduct was rapidly declining in the last years. In March 2020, the Flemish government initiated a major renovation project. Prior to this renovation (2015), low-cost noise measurements were performed at two dwellings at either side of the viaduct. In 2020, at the beginning of the renovation, the noise measurements were restarted at the two dwell- ings. The renovation suffered from several delays and was completed in October 2021. The new sit- uation, post-renovation can be assessed on the data collected from November 2021 onwards. The two datasets, 2015 and 2020/2022 are used to perform a citizen science based noise evaluation of the renovation of the viaduct. At the south side of the viaduct a gap in the noise screens was closed. The impact of the changes in noise barrier type, road surface and bridge joints can be evaluated in the measurement point at the north side of the viaduct. Preliminary results show an effect of 7 dBA at the south side. The impact of the renovation itself is limited to 1 dBA. These types of citizen engagements are common in the air pollution field and that public engagement resulted in significant improvements awareness, the related legislation and policy actions. This citi- zen based initiative for noise exposure shows that similar public engagement and measurement cam- paigns can add value in the noise exposure field as well.

1. INTRODUCTION

Noise monitoring is a standard method to evaluate the state of the environment with many aspects. Within the acoustic community, high quality noise monitoring is the standard approach for all appli- cations related to noise regulations. Yet, since over a decade, the number of applications of low cost noise monitoring has grown significantly. Numerous cities already implemented larger noise moni- toring networks, sometimes a combination of high quality and low quality measurement equipment (Barcelona, Lille, Paris, New York, etc...). Within the Flemish region in Belgium, the potential of low-cost noise monitoring has been included in the newly defined Regional Environmental Noise Indicator Framework (in Dutch) [1]. A summary of this framework is presented at Forum Acusticum 2020 [2].

The low-cost noise monitoring equipment used for this application was developed at the Ghent University more than a decade ago [3, 4]. This equipment was also used in several mobile applications

1 Luc.dekoninck@ugent.be

worm 2022

[5-8]. A large part of these applications included simultaneous measurements of noise and air pollu- tion [5, 6, 9, 10]. A large part of the work fitted in the PhD of the author [11]. A summary of the PhD is available as a separate publication [12].

The purpose of this publication is to present a long-term noise monitoring application to extend and/or facilitate the population exposure monitoring component in the Regional Environmental Noise Indicator Framework. It fits in a wider set of applications to collect data at a larger scale. The ensem- ble of different monitoring cases will support the future implementation of the framework. In prior work, another example of large-scale deployment combined education and citizen science through a STEM project in schools [13].

2. THE MONITORING CONTEXT

2.1. The physical situation

A main highway transects a suburban area (Gentbrugge) of the city of Gent. The viaduct is built in the early seventies as a part of the European highway network. This highway is the main connection between the port of Antwerp (BE) and the port of Rotterdam (NL) towards the south, direction France. Noise screens were implemented in the early 2000’s, with a focus on the areas with the highest pop- ulation density. The design included a significant gap near an area with semi-public gardening func- tion, without permanent inhabitants. The viaduct is deteriorating and several renovation actions took place over time. During a major renovation more than 10 years ago, the road surface and the bridge joints were modified. The results were not well received by the nearby population. In 2014, a local initiative emerged with a name mimicking the sound of the viaduct and referring to a Flemish word referring to ’kaput’: ‘Viadukaduk’. Their focus was both noise and air pollution exposure with a focus on air pollution.

worm 2022

Figure 1: Noise screens until 2020 and the monitoring locations near the viaduct in Gentbrugge (city of Ghent, Belgium). The blue transects show the start and the end of the viaduct (PRE).

In this situation, combined noise and Black Carbon monitoring was performed in the spring of 2015. The results were presented at ICBEN 2017 [9]. The physical situation and the position of the

two monitoring locations is visualized in Figure 1.The noise monitoring during this exercise is used as the prior situation, referred to as ‘PRE’ in the manuscript.

The viaduct caught bad publicity by concrete rot. At several point along the bridge, the concrete ruptured, pieces of concrete fell on cars and near pedestrians. An emergency renovation was neces- sary. The planned works were accelerated at the start of the pandemics. Less traffic due to the COVID lockdown reduced the impact on traffic flow during the renovation. The noise screens would be re- placed and extended, the road surface refurbished and the bridge joints – the origin of the name of the citizen action group- would be improved. At that moment, the citizens’ active in Viadukaduk asked to restart the measurements. The same noise monitoring equipment was mounted at (almost) the same locations (one set was mounted at a different corner of the dwelling, facing to the highway as the first setup). A dataset could be collected during and after the road works. The prior screens had a height of 3 m with a special screen-top ‘Mice’, the new screens have a height of 4 m without mod- ified screen top. The renovation itself took longer than expected due to several technical issues. Fail- ing joints triggered several extensions on the road works. The final state was achieved at October 15, 2021. The new situation with the extended screens is shown in Figure 2.

Within this publication only the pre and post situations will be evaluated. The situation during the traffic works was divers: removal and rebuilding of the noise screens, a significant speed reduction (from 90 to 50 km/h), 2x3 lanes to 2x2 lanes with several changing cross-overs during the road works, diversions of traffic etc. This part of the measurement campaign is not included in the evaluation.

The monitoring equipment at the south of the viaduct (Ooievaarsnest) failed and was replaced beginning of December 2021.

worm 2022

Figure 2: Noise screens after completion of the renovation between March 2020 and October 2021 (POST).

2.2. The measurements and data processing Two measurement locations are choosen at both sides of the viaduct, both 300 m from the infrastruc- ture. The data is collected in third octave bands (20-20 kHz) at a 8 Hz sample rate. This raw data is processed to 15 minute intervals resulting in several statistical levels: L 01 , L 05 , L 10 , L 25 , L 50 , L 75 , L 90 , L 95 and L 99 . The meteorological data is gathered from an official weather station at 10 m height less than 10 km from the measurement locations.

The data for the evaluation of the prior situation ranges from January 2015 till May 2015 (with some missing data). The prior situation includes data from November 1 st till March 31th for the first loca- tion (MP1: Leeuw van Vlaanderenstraat) and from December 6 th till February 15 th in MP2: Ooievaarsnest). The limited set in MP2 is caused by equipment failure.

2.3.Traffic data Traffic data is available in an aggregated version. The data for the prior situation is available for working days only (excluding weekends, school holidays and school holiday periods per hour, week- day and month). Unfortunately this data is not available for the post-period at the time of submission. No traffic comparisons can be presented. The results will be available in the presentation. 3. THE RESULTS

worm 2022

3.1. Noise exposure at the dwellings: diurnal pattern and meteorological variability The long-term data is analyzed by using generalized additive models (GAMs). The noise parameters are evaluated by four covariates: hour of the day, wind direction, wind speed and temperature. The dominant noise source for the measurement locations is the viaduct and the differences between the different noise parameters is not high. In the section we restrict the presented data to L 5 0, a good measure to evaluate a continuous source. In this manuscript we don’t focus on the in-period features of the splines but we focus on the differences between the pre and post periods. It is evident that the limitations in the data collection still can affect specific features in the individual models. In Figure 3, the results are shown for MP1, North of the viaduct. The splines show the typical diurnal pattern, the variation in function of the wind direction (200° is the downwind condition), the increas- ing noise levels with higher wind speeds and the lower noise levels at higher temperatures.

Figure 3: MP1(Leeuw van Vlaanderenstraat), GAM model for L A50,15min for four covariates: prior

renovation (top) and post renovation (bottom) The diurnal pattern shows significant changes. In the POST model, the morning peak is much less expressed compared to the PRE model. This change is with a high probability, related to the changed traffic situation due to the pandemic. A significant part of the employees perform their jobs in a

remote context, potentially affecting the morning rush hour. The traffic data to confirm this is not available yet (see section 2.3). The wind direction spline is also significantly different. In the PRE model, the noise levels show an asymmetric form, while in the POST model, a symmetric form is visible. In the prior situation, the exposure under East wind direction is relatively high. This matches with the short screen in this di- rection, while in the new situation, the extended screen does reduce exposure at MP1. For the next two splines, no important comment can be made, apart from the fact the collected data in the post-situation is largely collected during winter which affects both the wind speed and temper- ature spline. The intercept of the models is shown in the function above the splines. This is a first indication of the absolute difference in the noise exposure. In Figure 4, the same data is presented for MP2, south of the viaduct. IN the location the L A50,15min value is highly affected, with a typical effect of close to 10 dB. The collected data is affected by an equipment failure. The post dataset might e partially affected by Christmas holiday period and the downwind condition for this location is not the most abundant meteorological situation. These results have to be used with the proper caution. They do reflect the physical situation of closing the gap in the screen.

worm 2022

Figure 4: MP2 (Ooievaarsnest), GAM model for L A50,15min for four covariates: prior renovation

(top) and post renovation (bottom) The splines for the the hour of the day and the wind direction haven’t changed decreased a little for the diurnal pattern and increased a little for the shape, but the range increased in the post situation. In the other spline, no significant changes are visible. The range of available temperatures illustrate the lack of data in the location across seasons.

3.2. The impact of the renovation on the noise emission of the bridge joints The noise emitted by the bridge joint is the origin of the citizens’ structured response to coordinate the communication on the government and the road agency. The recent renovation was extended due to continued issues with those bridge joints. Assessing this nuissance is not evident. The relevant measuring location is MP1. The peak exposure is available in the L 01 statistical level but this property is also affected by other factors. The most important factor is local traffic. The local traffic is not that

abundant but a bus-line passes along the dwelling, providing a very constant traffic source during the day. The measurement location was, for practical reasons (power supply) to another corner of the dwelling. The dwelling is on a corner of a crossing with one road virtually without traffic (access to few local dwellings only) and the local road with the bus-line, with little local traffic. The new meas- urement position location is moved to a corner of the house away from the road with the bus-line. If L 01 is defined by the local bus traffic, a reduction of L 01 can be expected. This is exactly what is found in the data (See Figure 5-top row, more details on this graphs in the next section). During the night, L01 is almost identical before and after the intervention. If L 01 is a relevant parameter to evaluate the joints, no effects is detected. It is very well possible that L 01 is not the proper parameter to evaluate the bridge joints. Further investigations are necessary.

3.3. Comparing noise exposure after meteorological correction To exclude the variation in the data due to meteorological features, the prediction function of the gam is used to extract a diurnal pattern for two relevant meteorological situation:, upwind and downwind condition for the viaduct to the measurement locations. The results are presented in Figure 5.

worm 2022

Figure 5: Diurnal patterns for five statistical levels under upwind and downwind conditions for MP1 (top row) and MP2 (bottom row). Full lines represent the post situation, the dashed lines the

prior situation.

The most relevant conditions are the downwind version, top-right and bottom left. The post situation at MP1 shows an improvement of 1 dB for L Aeq (red lines) for most of the day except for the early morning (4-6 AM). The highest difference occurs during the morning peak (2 dB). All other statistical levels follow similar patterns except the L 01 and L A95 .

Downwind Upwind

Pre Post Delta Pre Post Delta

MP1: Leeuw van Vlaanderenstraat

LA10 56.1 55.5 -0.6 49.2 51.1 1.9 LA50 52.3 51.7 -0.5 43.6 46.6 3.0 LA95 49.2 48.6 -0.6 41.4 44.0 2.6 LAeq 54.4 53.5 -1.0 48.6 49.6 1.0

MP2: Ooievaarsnest

LA10 64.4 56.4 -8.0 55.9 45.9 -10.0 LA50 62.0 53.9 -8.1 52.6 43.8 -8.8 LA95 59.0 51.4 -7.6 49.4 42.3 -7.1 LAeq 62.5 55.3 -7.2 53.8 45.1 -8.7

Table 1: summary of th e results for bot h locatio n s for d o wnwin d and u p wind c o nditions. In Table 1 the effect of the intervention are summarized. For downwind conditions, the effect at MP1 is between -0.5 and 1.0 dB for all parameters. The effect on MP2, with the closure of the gap in the screens ranges between -7.2 and -8.1 dB. 4. DISCUSSION

The results also show the sensitivity to small changes in the setup before and after the intervention, affecting the usability of the L01 noise indicator in MP1. Other issues in the application of these data exist. One of the main issues is the use of the absolute noise levels in the communication. Varying façade reflections for site to site might affect the reports and have to be included in the evaluation process, fitting the application of the data. In this case, the evaluation of an intervention, the façade reflection is cancelled in the comparison. The absolute levels have to be adjusted to fit the general practice. The road agency performs validation measurements also but the protocol is restricted measurements at multiple locations for only 15 minutes during the day, mainly on locations much closer to the source than the data presented in this manuscript. A short test was performed at MP1, but The traffic on the highway has changed due to the change mobility after the pandemic. This effect couldn’t be included in this stage of the evaluation. 5. CONCLUSIONS

This zero budget measurement campaign, initiated by citizens, shows the potential of long-term low- cost noise measurements. The extension of the noise screen has a significant effect. The replacement of the screen has only a small effect. Using basic acoustical parameters, no effect was found for the noise emission related to an improvement of the bridge joint. Further analysis on this topic is neces- sary. It is clear that this citizen initiative combined with available semi-professional low-cost moni- toring system can add value to assess interventions and can be used for both citizens and governments to understand and document the effect of large projects on citizens.

worm 2022

6. REFERENCES

1. https://www.vlaanderen.be/publicaties/geluidshinder-evaluatie-van-de-mira-indicatorset 2. Dekoninck, L., & Botteldooren, D. (2020). Regional environmental noise indicators: SWOT anal-

ysis and future developments. In e-Forum Acusticum 2020 (pp. 1663-1665). 3. Van Renterghem, T., Botteldooren, D., Thomas, P., Can, A., Dauwe, S., Dhoedt, B., ... & O'Dono-

ghue, R. (2010). Towards an extensive noise and air quality measurement network. In 39th Inter- national Congress and Exposition on Noise Control Engineering (Inter-Noise 2010) (pp. 1-6). Inter-Noise. 4. Botteldooren, D., Van Renterghem, T., Oldoni, D., Samuel, D., Dekoninck, L., Thomas, P., ... &

Dhoedt, B. (2013, June). The internet of sound observatories. In Proceedings of Meetings on Acoustics ICA2013 (Vol. 19, No. 1, p. 040140). Acoustical Society of America. 5. Dekoninck, L., Botteldooren, D., Panis, L. I., Hankey, S., Jain, G., Karthik, S., & Marshall, J.

(2015). Applicability of a noise-based model to estimate in-traffic exposure to black carbon and particle number concentrations in different cultures. Environment international, 74, 89-98. 6. Dekoninck, L., Yang, Q., Zhao, H., Ross, J., Jack, D., & Chillrud, S. (2019, September). An

international application of the city-wide mobile noise mapping methodology: Retro-active traffic attribution on a bicycle commuters health study in New York City. In INTER-NOISE and NOISE-CON Congress and Conference Proceedings (Vol. 259, No. 6, pp. 3265-3276). Institute of Noise Control Engineering. 7. Ramos, T. D. C., Silva, A. N. R. D., Souza, L. C. L. D., Dekoninck, L., Botteldooren, D., &

Teixeira, I. P. (2020). Cyclists’ noise exposure in a Brazilian medium-sized city. Ciência & Saúde Coletiva, 25, 2891-2902. 8. Teixeira, I. P., da Silva, A. N. R., Schwanen, T., Manzato, G. G., Dörrzapf, L., Zeile, P., ... &

Botteldooren, D. (2020). Does cycling infrastructure reduce stress biomarkers in commuting cy- clists? A comparison of five European cities. Journal of Transport Geography, 88, 102830. 9. Dekoninck, L., & Botteldooren, D. (2017). Uncorrelated components in noise and PM exposure

give opportunities to disentangle health effects of noise and air pollution. In 12th International Congress on Noise as a Public Health Problem (ICBEN 2017). 10. Dekoninck, L., & Hofman, J. (2020). Multi-disciplinary sensing for personal exposure assess-

ments: Quantifying the impact of traffic interventions and meteorological variability. In e-Forum Acusticum 2020 (pp. 3085-3090). 11. Dekoninck, L. (2015). Spatiotemporal modelling of personal exposure to traffic related particu-

late matter using noise as a proxy (Doctoral dissertation, Ghent University). 12. Dekoninck, L., Botteldooren, D., & Int Panis, L. (2017). Extending participatory sensing to per-

sonal exposure using microscopic land use regression models. International journal of environ- mental research and public health, 14(6), 586. 13. Dekoninck, L., Van Beggenhout, W., & Sterken, M. (2021, August). Merging Science Education,

Citizen Science and General Population Noise Exposure data collection in the battle against noise pollution. In INTER-NOISE and NOISE-CON Congress and Conference Proceedings (Vol. 263, No. 3, pp. 2996-3007). Institute of Noise Control Engineering.

worm 2022