*2.4. Count of the Number of People Exposed to WT Sound*

Building exposures were cross-referenced using a GIS software (QGIS [28]) with the populations provided by the national database of land use files MAJIC updated in 2016 [29]. Only residential buildings were included in the counting.

Finally, the number of people exposed to WT sound for day- and night-time propagation conditions was determined at the scale of the whole metropolitan France, as well as of the 13 administrative regions in metropolitan France.

#### *2.5. Comparison of WT Sound Exposure with Other Environmental Noise Sources*

A comparison of WT sound exposure with transportation noise exposure was conducted. Transportation noise exposure data were derived from calculations performed in 2017 [30] as part of the European directive related to the assessment and management of environmental noise [31].

#### **3. Results**

#### *3.1. Evaluation of the Wind Turbine Sound Prediction Model*

The distribution of differences between the predicted and measured sound levels was obtained from 77 and 79 measurements where the wind speed was greater than 6 m/s and the distance from the wind turbine ranged from 500 m to 1500 m, for daytime and night-time propagation conditions respectively. The bias was −4.4 dBA and −0.2 dBA, and the standard uncertainty was 3.9 dBA and 4.7 dBA, for daytime and night-time propagation conditions, respectively (Figure 4).

**Figure 4.** Boxplot of the difference between predicted and measured sound levels, for two typical meteorological conditions during the day and night. The wind speed was above 6 m/s and the distance from the wind turbine ranged from 500 m to 1500 m. Number of samples (n), bias (mean), standard uncertainty (sd).

## *3.2. Wind Turbine Sound Exposure in Metropolitan France*

The total number of people in metropolitan France exposed to WT sound levels above 30 dBA was 685,770 people for daytime propagation conditions and 721,559 people for night-time conditions (Table 1). Taking into account the standard uncertainty from the model validation resulted in a range of this estimate of [430,036; 905,967] people for daytime meteorological conditions and [303,976; 1,029,390] for night-time. These results corresponded to 1.0% [0.6%; 1.3%] of people living in France in 2017 for daytime, and 1.0% [0.4%; 1.5%] for night-time. It should be noted that the range of these estimates did not correspond to confidence intervals, but to estimates based on the lower and upper scenarii.


**Table 1.** Number of people in metropolitan France exposed to WT sound levels above 30 dBA for daytime and night-time propagation conditions. The lower/upper scenarii correspond to +/−1 standard deviation on the Harmonoise sound level estimate.

Finally, 48% and 61% of the people exposed to sound levels above 30 dBA were exposed to sound levels below 35 dBA for daytime and night-time propagation conditions respectively, and 82% and 93% of the people exposed to sound levels between 30 dBA and 40 dBA for both daytime and night-time propagation conditions respectively (Figure 5).

**Figure 5.** Number of people in metropolitan France exposed to WT sound as a function of WT sound level for daytime and night-time meteorological conditions, normalized by the total number of people exposed to WT sound level above 30 dBA.

### *3.3. Wind Turbine Sound Exposure by Region*

Most of the population exposed to WT sound levels above 30 dBA was located in the Hauts-de-France region (daytime: 265,227 people, night-time: 275,846 people). Bretagne, Grand-Est and Normandie regions accounted for 62,728 to 86,770 people for daytime, and for 65,285 to 94,742 people for night-time (Figure 6).

**Figure 6.** Number of people exposed to WT sound level above 30 dBA, by region. Error bars account for the uncertainty of +/−1 standard deviation on the Harmonoise sound level estimate.

For daytime, 39% of the French population exposed to WT sound levels above 30 dBA lived in Hauts-de-France (38% for night-time) (Figure 7). Bretagne, Grand-Est and Normandie represented 9% to 13% (daytime) and 9% to 13% (night-time) of the population exposed to WT sound levels above 30 dBA.

**Figure 7.** Number of people exposed to WT sound levels above 30 dBA for daytime and nighttime meteorological conditions, by region, normalized by the total number of exposed people in metropolitan France. Error bars account for the uncertainty of +/−1 standard deviation on the Harmonoise sound level estimate.

Considering the exposed population in relation to the total population of each region, two northern regions had a higher percentage of people exposed to WT sound: Hautsde-France and Normandie accounted for 4% and 6% for daytime, and 5% and 6% for night-time, respectively, (Figure 8) of the total population of each region.

**Figure 8.** Number of people exposed to WT sound level above 30 dBA for daytime and night-time meteorological conditions, by region, normalized by the total population of each region. Error bars account for the uncertainty of +/−1 standard deviation on the Harmonoise sound level estimate.

Figure 9 shows a comparison of the WT sound level distributions of the exposed populations between all regions of France, for both daytime and night-time meteorological conditions. Except for Corse and Provence-Alpes-Côte d'Azur (PACA) regions, for all other regions, the majority of exposed people are exposed to WT sound levels below 35 dBA.

**Figure 9.** Number of people exposed to WT sound as a function of WT sound levels, normalized by the total number of people exposed to WT sound levels above 30 dBA: daytime (**a**) and night-time (**b**) meteorological conditions.

#### *3.4. Comparison of WT Sound Exposure with Transportation Noise Exposure*

Figure 10 shows a comparison of the proportion of people exposed to transportation sound levels above 40 dBA during the night (Lnight) [30] with the population exposed to WT sound levels above 40 dBA for night-time propagation conditions. The French population in 2017 exposed to night-time noise was 10,394,293 for road traffic noise, 5,113,159 for railway noise and 463,611 for aircraft noise, which can be compared to 53,752 people for night-time WT sound exposure; i.e., 15.0%, 7.0%, 0.7% and 0.08% of the 2017 French population, respectively.

**Figure 10.** Proportion (%) of the 2017 French population exposed to sound levels above 40 dBA during the night, for four noise sources: aircraft noise (Air), railway noise (Rail), road traffic noise (Road) and wind turbine noise (WT).

#### **4. Discussion**

The aim of the research presented in this paper was to quantify the number of windfarms' residents in France exposed to wind turbine sound. The first objective was to validate a sound level prediction model for the calculation of WT sound levels. The performance of the Harmonoise model was evaluated through the quantification of its uncertainties on the WT sound levels predictions. For both daytime and night-time propagation conditions, the bias between predicted and measured WT sound levels was regarded as sufficiently small (bias between −4.4 dBA and −0.2 dBA) to allow a correction of predicted sound levels. The value of the standard uncertainty was considered sufficiently low to validate

the use of Harmonoise model for the prediction of WT sound levels in this research. It is consistent with what is encountered for engineering models of outdoor noise prediction, for which the standard uncertainty can typically range from 2 to 4 dBA, depending on the model [12,32–34].

The second objective was to count the number of people in metropolitan France exposed to WT sound levels. The proportion of those exposed to WT sound levels above 30 dBA represented a small proportion of the total population in France (1% for both daytime and night-time propagation conditions). The WT sound exposure of these people was very moderate: the majority of people were exposed to sound levels below 35 dBA, and more than 80% were exposed to sound levels below 40 dBA, for both daytime and night-time propagation conditions. Most of the exposed population was located in the Hauts-de-France region (about 40%). A very large majority of the exposed population was located in the North of France: Bretagne, Grand-Est, Normandie and Hauts-de-France represented more than 75% of the people exposed to WT sound.

Compared to other regions, the two northern regions, Hauts-de-France and Normandie have more people exposed to WT sound compared to the total population of each region. Nevertheless, this difference was rather small (less than 5%) and could be partly explained by the combination of several regional factors: the number of wind farms, the number of people living in rural areas, where wind farms were generally installed, and the spatial distribution of people within each region.

The distributions of sound levels between regions were compared in order to investigate regional specificities in the WT sound exposure. This could happen, for example, if WT would be noisier or closer to residents' dwellings in some regions. Except for Corse and PACA regions, the distributions of sound levels of exposed populations seemed very similar across regions, for both daytime and night-time meteorological conditions. In metropolitan France, people exposed to WT sound were exposed in a similar manner, with no regional specificity. In addition, for almost all regions, the sound exposure of the majority of exposed people was between 30 dBA and 35 dBA. The cases of Corse and PACA were marginal given the small number of exposed people for those regions.

If any significant health effects from WT sound were to be demonstrated in the future, it would be useful to assess this potential public health issue in comparison to other exposures to common noise source for which health effects have already been demonstrated [3]. The comparison of the proportion of the French population in 2017 exposed to sound levels above 40 dBA during the night for transportation noise and for WT sound showed that far few people were exposed to WT sound than to transportation noise (0.08% for WT sound, compared to 15.0%, 7.0%, 0.7% for road traffic noise, aircraft noise and railway noise respectively). This is especially true as the sound levels of the population exposed to transportation noise were underestimated. Indeed, only major transportation infrastructures, and cities with more than 100,000 inhabitants were considered (see [31] for more details). Conversely, WT sound exposure was probably overestimated here (see below). The fact that the proportion of people exposed to WT sound was much lower than that exposed to transportation noise could be explained by the much smaller number of wind farms on the metropolitan territory, compared to transportation infrastructures, by the installation of wind farms in rural areas where the population density is lower, by sound levels at dwellings that are lower for WT, and finally by distances between dwellings and noise sources that are greater for WT due to French legislation about wind farms implantation.

The methodology for assessing the population exposed to WT sound had some limitations. The first limitation, as briefly mentioned above, was that the count of exposed people was likely to be overestimated because the WT sound levels were estimated by considering (i) a constant wind speed corresponding to the nominal operation of the wind turbines, as if WT were constantly in operation throughout the day or night, (ii) the sound exposure based on the maximum sound level among all the sound levels predicted for the different wind directions, (iii) the sound level of the most exposed façade of each dwelling and (iv) the meteorological conditions corresponding to favorable conditions for sound propagation. The initial objective of the paper was to investigate the worst-case scenario that maximized noise exposure levels. Therefore, only one wind speed value was considered (7 m/s at 10 m height). This corresponded to the maximum sound emission of the wind turbine and to the nominal WT operation. It should be noted that the sound emission is constant for wind speeds above 7 m/s. Three scenarii were then considered: the best, the worst and the average scenario, named in the paper lower, upper and average scenario in order to bound the estimates of population counts. The range of estimates could sometimes be wide, but this was not a major problem, the most important being the orders of magnitude of the population counts, and also the values corresponding to the lower scenario, which indicated whether there are enough people exposed to different and relatively contrasting WT sound levels to conduct an epidemiological study and to demonstrate, in a statistically rigorous way, an association between a health condition and the level of wind turbine noise, if such an association exists. In the on-going RIBeolH project [35], information from regional wind statistics will be used to take into account the regional influence of wind direction and speed on exposure. Considering that all inhabitants of a house were assigned to the most exposed façade could also lead to an overestimation of the number of people exposed to the WT sound. This overestimation was not a major concern in this research since one of its original aims was to obtain an upper limit of WT sound exposure. Conversely, if the www.thewindpower.net (accessed on 1 December 2021) database (was the most complete publicly available database of WT in France in 2017, pending the exhaustive database that the French Ministry of the Environment is building and which should be accessible in 2022, its completeness was not perfect and the missing wind farms could be a cause of underestimating the number of residents exposed to WT sound in this study. A comparison of national wind capacity from the database with other data sources [36,37] showed that the underestimation could nevertheless be limited because the database contains 94% of the installed capacity as of mid-2017.

A second limitation was that the predicted sound exposure did not correspond to an actual exposure over a day or night period, as would for example an equivalent noise level, but took into account typical propagation conditions of daytime and night-time periods. However, they could possibly be interpreted as an equivalent sound level exposure if it was assumed that the meteorological conditions and wind speed were the same during all the periods. While this was not the most realistic situation, it nevertheless did provide useful information on the upper limit of the number of people exposed to WT sound.

Another limitation of this research concerned the estimation of uncertainties. Estimating the uncertainties in the population counts (associated with the level of reliability and the percentage of error) was rather difficult because there was no simple relationship between sound exposure and population count: this relationship differed strongly from one wind farm to another due to the propagation influence (e.g., the presence or absence of obstacles), and the spatial distribution of the population around each wind farm. Some procedures based on Monte Carlo or bootstrapping techniques could have been considered, but they were deemed too time-consuming in relation to the time available for this research. They will nevertheless be explored for the update of this research in the framework of the RIBEolH project. Thus, in order to give a range of estimates of the population counts, three noise exposure scenarii (upper scenario, average scenario and upper scenario) were preferred for estimating these uncertainties.

The aim of this study was to deal only with audible noise, which is a first step before going further. It was out of the scope to deal with all phenomena involved in WTN, in particular:


2021). Moreover, the few available research models on amplitude modulation cannot be applied at the scale of a national territory because of their complexity and because they require input information that is not available at this large geographical scale.

The methodology proposed in this paper could be improved: the introduction of annual wind speed and direction statistics for each region in the calculation process could provide a more realistic sound exposure for daytime or night-time periods. An investigation of the low frequency noise exposure could also be done. An update of this work is in progress as part of the preparation of the epidemiological study planned in the RIBEolH project [35], to adapt to the rapid growth in the number of wind farm installations in France and the evolution of the French population, but also to improve the calculation process and to extent the assessment to low frequency noise exposure.

The results presented above, being the first assessment of exposure to wind turbine noise in France, may lead to a relevant epidemiological study in the RIBEolH project. Indeed, in order to conduct an epidemiological study and to demonstrate, in a statistically rigorous way, an association between a health condition and the level of wind turbine noise, if such an association exists, it would first be necessary to recruit a sufficient number of individuals exposed to different and relatively contrasting WT sound levels. It will be possible to reach this first step thanks to the count of the number of people exposed to contrasted levels of WT sound for all wind farms of the whole metropolitan France. Then, the quality of epidemiological studies assessing the risks associated with environmental exposures depends in part on the quality of the estimation or measurement of participants' exposure. However, in a large-scale epidemiological study, noise measurements in a large number of residents will not be feasible because of the cost and it will be necessary to use noise prediction models. The use of Harmonoise prediction model proposed and validated in this paper will thus make it possible to estimate the exposure to wind noise at the home of each participant who will be included in the epidemiological study of the RIBEolH project.

## **5. Conclusions**

The objective of the Cibelius feasibility study was to propose a methodology for calculating WT sound exposure at a national geographical scale and to identify the number of people exposed to WT sound in metropolitan France.

The total number of people exposed to WT sound was approximately 686,000 during the day and 722,000 during the night, thus about 0.1% of the 2017 French population. More than 80% of the population exposed to WT sound levels above 30 dBA was exposed to levels below 40 dBA. It is important to note that, due to some assumptions (wind speed corresponding to WT nominal operation, wind turbines constantly in operation throughout the day), the sound exposure, and therefore the number of people exposed to WT sound, was probably overestimated.

These results constitute the first assessment of WT sound exposure at a national geographical scale, and more specifically for metropolitan France. The results and methodology proposed in this paper were the first step in preparing an epidemiological study in France in the few next years. This study will be conducted in the RIBEolH project, which is investigating the impact of WT sound on human health and annoyance. The results of this study will provide useful information for public authorities to assess whether or not regulations concerning WT should be adapted.

**Author Contributions:** Conceptualization, D.E.; Formal analysis, D.E.; Funding acquisition, D.E. and A.-S.E.; Investigation, D.E.; Methodology, D.E.; Project administration, A.-S.E.; Software, D.E., P.D. and G.G.; Supervision, A.-S.E.; Validation, D.E. and P.D.; Writing—original draft, D.E., L.G.-A. and A.-S.E.; Writing—review & editing, D.E., L.G.-A. and A.-S.E. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Ifsttar (now Gustave Eiffel University), Cerema and by a grant provided by the Environment-Health-Work Program of Anses (French Agency for Food, Environmental and Occupational Health & Safety) with the support of the ministries in charge of ecology (EST-2016/1/007).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data for each wind farm is publicly available at www.thewindpower.net, but the complete database was purchased from that site by the authors for this project under a research agreement. The topography and building databases are freely available online (see ref. [22]).

**Acknowledgments:** The authors thank the French Agency for Food, Environmental and Occupational Health & Safety (Anses) who funded this study.

**Conflicts of Interest:** The authors declare no conflict of interest.
