Bijlage 18 Detail 11b Windververkenning RIVM.pdf

The nonresponders were more often women (60.6%) than men (39.4%), and less of them had a view of a WT compared to respondents (61.5% vs. 81.6%). There were no differences in attitude towards the WT between the responders and non-responders. WT sound was recorded by the residents in this study and distance was also available as a proxy for exposure. This longitudinal study did not find empirical evidence for an association between annoyance or acceptance of WT and distance to the residence at both measurement points. More residents complained about physical and psychological symptoms due to road sound than WT sound (16%, two years later the same) than from WT sound (10%, two years later 7%). There is no numerically strong relationship between noise Page 29 van 73

RIVM report 2020-0150

annoyance and the distance to the nearest WT or the estimated sound pressure level. Fairness showed to be the best predictor and annoyance was found to decrease over time. These findings are in line with previous evidence. However, WT sound (recorded by some residents) showed to be an important indicator of annoyance and stress responses. One of the key causes for WT noise annoyance might be the amplitude modulation (AM). The authors conclude that the reason why AM is so strongly linked to annoyance is the fact that short-term amplitude changes attract the attention and thus disturb current behaviour. Krogh et al (2019) performed a qualitative study (risk of bias not relevant) among 67 study participants: 28 had vacated/abandoned their home because of the presence of a wind farm within 10 km; 31 were contemplating to do so; 4 pre-emptively vacated their home before the wind farm started operating; and 4 had decided to remain. Preliminary results showed that people with pre-existing medical conditions were concerned that living near a WT would exacerbate their symptoms. Although this study is not focussed on annoyance per se, these concerns affecting moving behaviour might also be of relevance for annoyance reactions. 3.2

Sleep disturbance Evidence regarding the effect of night time WT sound on sleep was inconclusive in 2017. The results at the time did not allow a definite conclusion regarding both subjective and objective sleep indicators. However, studies did find a relation between self-reported sleep disturbance and annoyance from wind turbines.

3.2.1

Reviews on sleep disturbance Based on the recently published WHO evidence review of Basner and McGuire (2018), we know that there is evidence of sufficient strength for self-reported and objective indicators of sleep disturbance due to environmental noise in general. Studies investigating the association between noise and sleep disturbance are usually based on the percentage of highly sleep disturbed (%HSD) as measured with a semistandard question with reference to the noise source. Objective measures include motility data (movements while sleeping) and awakenings (as measured by EEG). As part of their review, Basner and McGuire conclude that the evidence for sleep disturbance from wind turbine sound is only emerging and no EER is available yet. This statement was based on self-reported sleep in six studies published in the period between 2000 and 2015 that had to meet the rigid selection criteria used. Meta-analysis was performed for five out if these six studies and led to the inconclusive results in line with several earlier reviews including our own review in 2017. A distinction was made between questions in which self-reported sleep disturbance referred to noise or sound, and studies that did not refer to WT sound in the question. This forms a potential source of bias according to Basner and McGuire. In four studies a significant association was confirmed.

A metaanalysis was performed on five of the six studies based on the odds ratios for sleep disturbance for a 10 dBA increase in outdoor predicted SPL levels. Results show a non-significant association on the pooled data with an odds ratio of 1.60 (95% CI: 0.86–2.94). Two studies were Page 30 of 73

RIVM report 2020-0150

identified which used an objective method (actigraphy) to evaluate sleep disturbance due to WT sound (Lane, 2016 and Michaud, 2016c). The study by Lane was too small and the large study of Michaud concluded there was no significant association between wind turbine sound levels and sleep measured with actigraphy. In an update of studies that could expand the WHO sleep review (van Kamp et al, 2020a, 2020b) it was concluded that since 2015 a number of studies with large size and of good quality on windturbine sound was published and this justifies a meta-analysis. The search from mid-2015 to mid-2020 identified 14 new articles on sleep disturbance (11 with self-reported measures and 3 using objective measures). A new metaanalysis on subjective and objective sleep measures was suggested to assess the relation between sleep disturbance and WT sound. The review of Micic et al published in 2018 also focused on sleep disturbance. This is a review of potential mechanisms, rather than a review of current evidence for an association between WT sound and sleep disturbance.

According to the authors only a few studies have shown an association, but they consider it as plausible that WT sound leads to sleep disturbance via two mechanisms 1) chronic sleep fragmentation from frequent physiological arousals due to sensory disturbances in sleep; and 2) chronic insomnia that could develop in individuals with higher sensory acuity and/or those prone to annoyance from environmental noise. Between 2000 and mid 2018 Freiberg et al (2019) identified 19 studies on sleep that met their criteria (described in 3.1). Most of the studies included measures of self-reported sleep disturbance and some objective sleep parameters measured with polysomnography. In higher quality studies WT sound was not associated with self-reported or objective sleep disturbance, which contrasts – at least partly – with findings from lower quality studies that more often suggest there is an association. The conclusions are broadly in line with those of Basner and Macquire (2018). Below, the results of original studies are summarised. Some papers fall outside the time frame of 2017-2020, but are included, since they were missed in our previous review and are considered as relevant for the current state of the art. 3.2.2

Original studies on sleep disturbance Lane et al (2016) performed a field experiment with a case-control design with sleep measures and diaries over a period of five nights. 27 individuals participated in the experiment of whom 15 were from a WT exposed area. The response rate was 50%. Exposures were estimated based on the distance to the nearest WT and sound levels were measured during the period of the experiment. Sleep measures included sleep onset latency (SOL), wake after onset, total sleep time, time in bed, number of awakenings and sleep efficiency. Subjective sleep was measures by the standard but adapted Pittsburgh sleep quality index. No statistically significant differences were found between the two groups on any of the objective and subjective sleep measures after adjustment for gender and age. The authors concluded that either there is no effect of WT sound on sleep, or the number of participants was too Page 31 van 73

RIVM report 2020-0150

small to find such an effect, or the effect was masked by unknown factors. It was suggested that annoyance (which was not measured in this experiment) could be an important mediator between sound level and sleep quality, hereby referring to findings of Pedersen et al (2011), Persson Waye et al (2007) and Bakker et al (2012). The Danish cross-sectional study (Poulsen et al, 2019a) made use of a cohort of 583,968 addresses and studied the association between modelled WT sound levels above 24 dB at the façade and low frequency sound level indoor and the use of prescribed sleep medication. Age, gender, income, education, marital status, type of dwelling and distance to a nearby road were included as important confounders. Results showed that a five-year averaged outdoor night time WT sound level of 42 dBA or more was weakly associated with use of sleep medication with an odds ratio (OR) of 1.14 [95% confidence interval (CI]:0.98, 1.33) per 10 dB increase. No association was found that was related to the indoor sound level. Further analysis showed the strongest associations for the older age groups. The risk of bias was estimated to be medium, since this is an ecological study, in which the data are analysed at group/population level, rather than at individual level.

In a Finish cross sectional study (Radun et al, 2019) with a low risk of bias among 429 people (response rate 57%) the association between indoor WT sound levels and self-reported sleep was studied. WT sound level was modelled and categorised (intervals: [25–30], [30–35], [35– 40] and [40–46] dB Lden). Trust in authorities and operators, visibility, economic benefits, age, gender, education, type of dwelling, and distance were included as important confounders. This yielded a significant, but weak association between indoor sound level class and subjective sleep disturbance with an OR of 1.38 (1.16, 1.65) and (Nagelkerke pseudo R2=.50). However, health concerns from participants had a bigger influence on sleep disturbance than WT sound level did. Morsing et al (2018) performed two laboratory experiments with six healthy students during three consecutive nights. Sound exposure consisted of recordings of wind turbine sound with variations in sound pressure level, amplitude modulation strength and frequency, spectral content, turbine rotational frequency and beating behaviour. Sleep was measured by polysomnographic indicators as well as questionnaires. Results showed some indications that WT sound led to objective sleep disruption, reflected by an increased frequency of awakenings, a reduced proportion of deep sleep and reduced continuous sleep stage N2.

This corresponded with increased self-reported sleep disturbance. However, there was a high degree of heterogeneity between the two studies, preventing firm conclusions regarding effects of WT sound on sleep. Furthermore, there was some limited evidence from the second study that wakefulness increased with strong amplitude modulation and lower rotational frequency. The deepest sleep was adversely affected by higher rotational frequency and strong amplitude modulation, and disturbance of light sleep increased with high rotational frequency and acoustic beating. As described below, these findings were used in the development of a larger-scale sleep study (Smith et al, 2020) in a more Page 32 of 73

RIVM report 2020-0150

representative study population and exposing participants to recordings of more naturalistic WT sound. In 2020 a paper was published on a large-scale sleep experiment studying the effects of WT sound (Smith et al, 2020). For this Swedish laboratory experiment study participants were recruited from the general population aged 30-70, with a body mass index (BMI) below 30 kg/m2, habitual sleep times between 23:00 and 07:00 and a mean sleep duration of about eight hours. Other exclusion criteria were use of sleep medication, sleep apnoea and self-reported auditory acuity which was confirmed during the first pilot night of the study. The total experiment lasted 3 nights with 1 habituation night and two nights with an exposure and control condition. WT sound was synthesized but based on many field recordings of WT sound. The AM modulation parameters used were frequency-dependent modulation depth and WT rotor speed. Random variations in time were also included to mimic the recordings. The sound was played to create an indoor sound level in a typical Swedish house resulting from a constant outdoor WT sound of 45 dBA. The timeaveraged frequency spectrum was chosen to be constant in all WT sound files.

Outcome measures included self-reported sleep quality and physiological measures, including polysomnography, wrist actigraphy, heart rate, blood pressure, cortisol level after awakening and long-term cortisol level from hair samples. Key confounders were subjective stress and noise sensitivity. Results showed that a single night of WT sound exposure affected the duration of the REM sleep. No effects on other measured physiologic outcomes could be detected. The findings show that continuous WT sound with AM may impact sleep according to both objective and subjective measures, leading to lower sleep quality and less restoration in the morning. This was the case for people who both were and were not habitually exposed to WT sound. When compared to the reference group in both the control and WT-night, the habitually exposed group gave a more negative rating of sleep quality and tiredness, and indicated that they slept worse than usual. They also reported higher noise-induced sleep disturbance overall in both the control and exposure night, when compared to the reference group. The Chinese study by Song et al (2016) with 227 participants (response rate 77%) included measured WT sound levels, categorised into 5 sound levels (<40 dB up to >47.5 dB), and self-reported sleep disturbance.

Gender, age, residence time, visibility, noise sensitivity, attitude and general opinion about WTs were included as important confounders. The association between LAeq from WT sound and subjective sleep was significant but weak with a Spearman correlation of .21. The risk of bias of this study was estimated as being medium. The cross-sectional study of Kageyama et al (2016), published at the end of 2016 in Japan, among 1079 residents (response rate 47%) included field measurements during the survey on a limited set of addresses. Based on these measurements WT sound levels were estimated for each address. Sleep symptoms (self-reported) and insomnia (self-reported) were the two key outcomes. The risk of bias of this study was estimated to be medium. Sound from road traffic, noise sensitivity and attitudes towards wind turbines as well as demographic variables were accounted for. No evidence was found for an adverse Page 33 van 73

RIVM report 2020-0150

effects of WT sound on physical/mental health, self-reported sleep disturbance and insomnia based on self-reported symptoms. Insomnia was found to be more prevalent in areas with WT sound levels above 40 dBA at night. Insomnia and other symptoms seemed to be primarily affected by personal features such as noise sensitivity and visual annoyance with wind turbines. The Canadian Noise and Health (CNHS) study was discussed in our 2017 review as far as publications up to 2016 included sleep disturbance. The sleep study (Michaud et al, 2016c) among 742 participants included both subjective and objective measures and concluded that there was no effect of WT sound level on any of the sleep indicators, after adjustment for confounding such as age, caffeine use, BMI and health condition. The risk of bias was estimated to be low. No new studies on sleep were published based on the CNHS. 3.3

Cardiovascular effects For cardiovascular effects there was insufficient evidence for a direct relation with wind turbine sound level at the time of our review in 2017. The WHO evidence review on cardiovascular and metabolic effects prepared by van Kempen et al (2018) yielded three cross-sectional studies investigating the association between wind turbine sound and self-reported cardiovascular disease. No studies were available on the association between WT sound and diagnosed hypertension, ischaemic heart disease or stroke. An update to this review (van Kamp et al, 2020a) yielded three publications pertaining to two studies investigating the association between wind turbine noise and hypertension: one cross-sectional study (Michaud, 2018c) and one cohort study (Poulsen, 2018a). The authors of the cohort study (The Danish Wind turbine Study or DWS) concluded that their study does not support an association between wind turbine sound level and redemption of antihypertensive medication. Note that redemption of antihypertensive medication is used as an indicator of having hypertension. We further identified and selected two new studies investigating the association between wind turbine noise and ischemic heart disease (IHD). Both were cohort studies (Poulsen 2019b; Bräuner et al, 2018).

Also, the search yielded one study (Poulsen, 2019b) that investigated the association between wind turbine sound level and the incidence of stroke in a cohort study carried out in Denmark (n = 712.402). The studies by Poulsen (2019b) did not provide conclusive evidence of an association between outdoor WT sound and IHD or stroke. This finding was confirmed by the study of Bräuner et al (2018) in the so called Danish Nurse Cohort, which lend little support to a causal association between outdoor long-term WT sound exposure and IHD. In the same Danish Nurse Cohort study, Bräuner et al (2019a) found suggestive evidence of an association between long-term exposure to WT sound and atrial fibrillation (AF) amongst female nurses. Of the 28731 nurses involved in the cohort, 1413 developed AF. They were exposed to slightly higher levels of WT sound than the controls in this study. A 30% statistically significant increased risk (95% CI: 1.05–1.61) Page 34 of 73

RIVM report 2020-0150

of AF was found amongst nurses exposed to long-term (11-year running mean) indoor WT sound levels ≥20 dBA at night compared to nurses exposed to levels <20 dBA. According to the authors, AF as a result from chronic annoyance may be a plausible explanation. However, interpretation of these findings should be cautious as exposure levels were very low. Analysis on the same cohort (Bräuner et al, 2019b) did not yield convincing evidence for an association between long‐term WT sound exposure and stroke risk. 3.4

Metabolic effects For metabolic effects (diabetes and obesity) there was insufficient evidence for a direct relation with wind turbine sound level at the time of our review in 2017. In the WHO evidence review on metabolic effects of environmental noise (van Kempen et al, 2018) three studies were identified and selected that investigated the association between wind turbine noise during the night and self-reported incidence of diabetes. These studies were published in Persson Waye and Pederson et al (2007), Bakker et al, (2012), and Pedersen et al (2011). The results of the WHO review do not support an association between night time wind turbine sound level and higher risk of diabetes. The update for DEFRA published in 2020 (van Kamp et al, 2020a) yielded two new studies investigating the association between wind turbine noise and the incidence of diabetes: one cross-sectional study (Michaud, 2016b) and one cohort study (Poulsen, 2018b). Neither in the WHO review nor in the updates nor in our literature search any studies were identified that investigated the impact of wind turbine noise on obesity.

3.5

Cognitive and mental health effects of WT sound In the framework of the WHO Guidelines also the effects of WT sound and cognitive and mental health effects were reviewed (Clark and Paunovic, 2018). No original studies were identified on the theme of WT sound, but five systematic reviews of the evidence specifically regarding the association of wind turbine sound with quality of life, wellbeing and mental health were identified. Therefore, for wind turbine noise a review of these existing systematic reviews rather than non-existing primary research papers was undertaken. It was concluded that the number of studies was very limited and of poor quality, using a range of different outcome measures. In particular the exposure characterisation was evaluated as poor and distance was often used as a proxy for sound exposure. Based on study limitations, inconsistency and qualitative comparisons across studies the authors conclude, on very low-quality evidence, there is no substantial effect of wind turbine sound on quality of life, wellbeing or mental health. One recommendation is to also consider the spectral levels in future studies (see chapter 4). An update of this review on cognitive and mental health effects and wellbeing, cancer, self-reported health and birth effects (Clark et al, 2020), prepared for DEFRA, showed additional evidence to conclude that Page 35 van 73

RIVM report 2020-0150

there is very low-quality evidence for an absence of effects of wind turbine sound level on self-reported quality of life or health as well as very low quality evidence of an effect on mental disorders (anxiety, depression) and birth outcomes (see also Poulsen, 2018c) . In this review no evidence was confirmed of a WT sound effect on cancer. In the review of Freiberg et al (2019) it was concluded that no relationship was found between WT sound and stress effects and biophysiological variables of sleep; and inconsistent findings concerning sleep disturbance, quality of life, as well as mental health problems (depression and anxiety). 3.6

Social and physical aspects other than noise; influence of contextual and personal factors Based on the evidence available about non-acoustic factors we concluded in 2017 that reducing the impact of wind turbine sound could profit from considering a large range of other factors associated with levels of annoyance due to living near a WT and the moderate effects of sound level. The influence of these factors is not necessarily unique for wind turbines, but some are. The new search yielded some 36 new articles on the effect of situational, contextual and personal factors on annoyance and other health effects of which 25 the full text was read. Two of the articles identified were reviews on determinants of annoyance, one in general (Simos, 2019) and one on visual aspects only (Freiberg et al, 2019). Selection criteria included that the article needs to concern a study (so no commentary, editorials or opinion articles, or letters to the editor), have some reference to WT sound as well as to annoyance or other relevant outcomes. In total 24 papers were accepted for review and 12 articles were rejected because they did not meet these criteria or were already were already discussed above. In the description a distinction is made between studies addressing visual aspects, social aspects and other aspects such as personal aspects and context.

Review The narrative review by Simos et al (2019), which was also discussed in section 3.1, included 104 articles (on 67 studies) and the results are discussed along a range of determinants. Aspects considered were WT sound, infrasound and low-frequency sounds, shadow flicker, safety, landscape impacts and real estate. At the outcome side, annoyance and the so-called wind turbine syndrome were included. Shadow flicker has been shown to have a weak association with annoyance and health indicators. Prevalence rates due to accidents around WTs are very low but only a few systematic studies on this have been performed. Results on the effect of visibility of the turbines are inconsistent. Social networks, risk perception and social acceptability have been found to be important when analysing the social aspects of wind farm development. In general, the authors conclude that there is evidence that community participation at an early phase can prevent negative perceptions associated with wind energy projects. It has been shown in several projects that housing prices drop considerably as a result of plans for WT projects. In some cases, prices returned to normal when the park

Page 36 of 73

RIVM report 2020-0150

became operational. Based on the findings a set of recommendations was formulated. 3.6.1

Visual aspects It has been well documented that visual aspects play an important role in the response to wind turbines. These aspects are strongly intertwined with the auditory aspect: annoyance from visual aspects may add to, or perhaps even reinforce annoyance from noise (and vice versa). Visual aspects are related not only to the visibility of the wind turbines and perceived landscape pollution, but also to other characteristics of the turbines such as blinking lights, shadow flicker and rotor blade movements. Review In 2019 a high-quality review was published by Freiberg et al including the literature regarding the influence of visual aspects on annoyance and sleep disturbance next to acoustical aspects. It concerns a systematic review in which the PRISMA approach was applied. The study protocol was published on the International prospective register of systematic reviews prior to the study. Seventeen studies published between 2000 and mid 2018 were included. The pooled prevalence of high annoyance due to altered views and shadow flicker was 6% each. The results of other health effects were inconsistent, with some evidence that wind turbine visibility directly increases sleep disturbance. Other studies showed that generally annoyance by visibility, shadow flicker, and blinking lights was significantly and directly associated with an increased risk for sleep disturbance.

Only one study indicated an interaction effect of visual and auditory stimuli, meaning that a combination of visual and auditory aspects together leads to more sleep disturbance than sound or visual aspects separately. The authors conclude that direct and indirect wind turbine visibility may affect residents’ health, and annoyance may differ between individuals. Original studies Several studies published since 2017 were identified on the impact of visual aspects of wind turbines in relation to their acceptability (Delicado et al, 2017; Grima-Murcia et al, 2017; Lamy et al, 2017; Frantál et al, 2017; Sklenicka and Zouhar, 2018; Landeta-Manzano et al, 2018). It concerns surveys, experiments, document analysis as well as consultation of stakeholders and experts, all aiming at mapping the role of visual aspects in the planning and decision process and exploring ways to mitigate the negative environmental and social impacts of the use of wind energy. Delicado et al (2017) carried out media analysis, analysis of Environmental Impact Assessment reports and analysis of official positions on the issues. Visual pollution is often used in the debate and brought forward as an important argument against wind farms. Either framed negatively as a risk of damaging the image of an area or positively as indicative of a landscape presenting technological progress.

Media analysis showed that the word landscape was hardly ever used, but rather the rural-urban divide came forward in opinion articles. Analysis of the EIA reports showed that objections against wind farms Page 37 van 73

RIVM report 2020-0150

on grounds of landscape pollution was used more often by NGO’s and in some cases by the tourism industry. Grima-Murcia et al (2017) performed a laboratory study among 14 respondents. The respondents would see pictures of landscapes with different energy saving measures and with different durations of exposure (so a picture was shown for different lengths of time). Effects were measured by way of questionnaires and electroencephalographic recordings (EEG’s). No differences were found in EEG reactions between the different stimuli including WTs. But for nuclear plants a reaction of the brain indicative of processing negative emotions was registered in the 400 msec time frame, indicating that EEG recordings can be a useful procedure for measuring visual impact. Lamy et al (2017) held semi-structured interviews among 15 residents living at varying distances from a wind farm and visual impact came forward as one of the main aspects that influenced their perception, next to economic benefits, safety issues, noise, renewable energy benefits. In a survey among 474 adults by Frantal et al (2017) the influence of visual aspects of the landscape itself on the impact of wind farms was studied. Results showed that the contribution was highly dependent on the local environmental and socioeconomic context. These include noise annoyance, economic benefits as well as educational level.

In a survey among 400 participants in four different central European countries (Sklenicka and Zouhar, 2018), participants were presented with a range of pictures including photoshopped wind turbines and related to three landscape planning indicators (relief, land cover and landscape pattern). The aim was to obtain an objective method, based on general principles, for predicting the visual impact of onshore wind farms. However, none of the indices showed a significant association with the acceptability of the turbines. Landeta-Manzano et al (2018) evaluated the intervention of a leading WT manufacturer to secure acceptance of wind energy projects by local communities. This involved 47 stakeholders and 6 experts (n=53) in a qualitative study using semi-structured interviews. The focus of interventions was on the visual impact of the developments, health and safety issues, community involvement and social investment in the community. With respect to visual impact, results of the consultation showed that the most negative contributions to community acceptance are related to the location of the WT: when deciding on locations the opinion of the communities was ignored. More technical aspects in relation to visibility, which received ample attention at an earlier stage (and were described in our 2017 review), were not identified in the new search.

In the new studies aspects of safety and electromagnetic fields due to wind turbines were also not studied or only briefly mentioned and do not add new information. We refer here to the descriptions in our 2017

Page 38 of 73

RIVM report 2020-0150

review and the study of Lubner et al (2020) which is described in chapter 4. 3.6.2

Contextual, situational and personal factors Research in the past decade has shed some light on the question why some people are more disturbed by wind turbines than others (Ansensio, et al 2017; Lercher et al, 2017; Haubrich, 2020). Next to physical aspects, personal and contextual aspects influence the level of annoyance. Often these aspects are referred to as non-acoustic factors, complementary to the acoustic factors expressed in decibels and Hertz. Because the term "non-acoustic" refers to a broad range of aspects, which as a result are very unspecific, we prefer the term personal and contextual factors. Thus, non-acoustic factors can be further subdivided in the following sub-categories: • Demographic and socio-economic factors (age, gender, income, level of education); • Personal factors (fear or worry in relation to source, noise sensitivity, economic benefit from the source); • Situational factors (frequency of sound events, meteorological circumstances, other sound sources, distance to amenities, attractiveness of the area). • Social and economic factors (expectation, attitudes towards producers or government, media coverage, willingness to pay and accept); Without pretending to be exhaustive some of these aspects relevant in the framework of wind turbines are discussed in more detail below, based on the new evidence.

Note that no papers were identified on situational factors. Demographic aspects Age, gender and educational level have not been identified as crucial predictors of noise annoyance in general. Although demographics are usually accounted for in annoyance and sleep studies and studies addressing other health outcomes, they are usually merely treated as confounders rather than as important determinants of annoyance. However, in the controversial and highly political domain of wind energy and wind turbines there is some evidence that gender and educational level do play a role. Personal factors Fear and noise sensitivity (NS) keep coming forward as important predictors of annoyance and stress-related effects. Noise sensitivity refers to an internal state (physiological, psychological, attitude, lifestyle and activities) of a person that increases the reactivity to sound in general. Noise sensitivity has a strong genetic component (i.e. is hereditary) but can also be a consequence of a disease (e.g. migraine) or trauma. Also, serious anxiety disorders can go together with an extreme sensitivity to sound which can in turn lead to a feeling of panic (Van Kamp and Davies, 2013). As it was in the previous period, only very few studies focus on these aspects of sensitivity and fear as the main issues related to WT sound.

Several studies (Smith, 2020; Klaeboe, 2016; Michaud, 2016a; Song 2016; Kageyama, 2016), Page 39 van 73

RIVM report 2020-0150

reviewed in the previous sections, included NS and sometimes fear as confounder in their statistical analysis. These studies confirm the independent role of noise sensitivity on reaction to wind turbines and this is a well-established notion also found with other noise sources. Fear and concern seem to play an increasing role in relation to low frequency noise and infrasound and the assumed link to vibroacoustic disease (see section 4.3), a disorder which is not acknowledged in the medical world as diagnosis but the notion of which contributes to fear and concern. Social and economic aspects Earlier studies and reviews have emphasized that for the social acceptance of wind turbine projects by a local community it is crucial how the community evaluates the consequences for their future quality of life. Next to physical aspects, situational, social and political aspects play an important role on acceptance as well as on feelings of fairness. The communication and relation between the key parties (residents, (local) authorities, project developer) is crucial in this regard. In the past 10 years we can observe an increasing polarisation between these key parties. Since our previous review in 2017, we first noticed an increasing number of studies investigating the social acceptance of wind projects by local communities in a number of countries.

Many of these studies stress the relevance of a fair planning process and local involvement and participation (Beuret, 2016) (Brennan, 2017) (Langer, 2017) (Liebe, 2017) (MacDonald, 2017) (Scherhaufer, 2017) (Sonnberger, 2017) (Kongprasit , 2018) (Clark and Botterill, 2018) Janhunen, 2018) (Kim, 2018) (Gölz, 2018)(Langer, 2018), (Landeta-Manzano, 2018) (Scherhaufer, 2018) (Sæþórsdóttir 2018). Remarkably the number of articles lately seems to decline: no new articles were identified after 2018. In the studies on social acceptance of wind projects by local communities, it is concluded overall that people are more willing to accept new turbines in their vicinity if they can participate in decision making, or the turbines are owned by a group of citizens, and/or if the generated electricity is consumed in the region instead of being exported. Also, people who are already exposed to WT sound near their homes are more willing to accept the wind farm than newly exposed groups. Several researchers emphasise that local circumstances may differ considerably and should be accounted for in a study on acceptance. They emphasise that a complex set of individual and collective values and preferences should be considered, and the perspectives of scientists, policymakers and citizens should hereby be integrated.

An interesting paper by Clark et al (2018) argues that different stakeholders with different interests contribute different “facts” about the legitimacy of health complaints. The wind turbine syndrome is an example of an illness that is controversial in the medical world and therefore leads to ample debate. This finding shows how the interests and legitimacy of arguments are particularly relevant for competing descriptions of the ‘facts’ of wind turbine health effects. Earlier studies concluded that economic aspects can also affect annoyance from wind turbines. Co-ownership and benefit came forward Page 40 of 73

RIVM report 2020-0150

as important predictors in counterbalancing adverse responses to WT. It was emphasised that not only benefits are important in this context, but also a sense of control (van den Berg et al, 2008). This was confirmed in the CNHS and to a lesser degree to studies in Japan. More recent literature into the social and economic aspects is focussed more on willingness to pay and willingness to accept and the perceived reduction in housing values related to wind farms (Wen et al, 2018; Thomson, 2018). Thomson concluded that residents near a WT would, on average, be willing to pay for a windfarm to stay, whereas people living near a coal plant were willing to pay for the coal plant to be removed. Demographics did not have a significant effect on these results. In contrast to these findings and based on a systematic review Wen et al (2018) concluded that respondents in different studies consistently showed increasing willingness to pay for moving wind farms to greater distances from their dwellings. But this also strongly depended on the number and height of WT and this might indicate that we are not dealing with a linear association.

Page 41 van 73

RIVM report 2020-0150

Page 42 of 73

RIVM report 2020-0150

4

Health effects specific for low frequency sound and infrasound

As stated in our earlier review, a range of residential health effects are attributed to the presence of wind turbines and infrasound and/or lowfrequency sound is sometimes mentioned as an important cause of these effects, also when the infrasound levels must be very low or are unknown. It was noted in section 2.1 that acoustically low-frequency sound and infrasound is different from sound at higher frequencies. When compared to sound at higher frequencies low-frequency sound and infrasound is attenuated less over larger distances and through building façades. However, here we want to know about the effects: can infrasound or low-frequency sound have effects on people that are different from effects of normal sound? In this chapter the question is addressed whether infrasound or low-frequency sound deserves special consideration with respect to the effects of wind turbine sound. Most of the new studies involved measuring brain activity in response to infrasound, often in comparison to low-frequency or ‘normal’ sound. So, in contrast to chapter 3 the focus here is more on mechanisms that play a role and less so on long term health effects of low-frequency and infrasound. We will use the term audio sound for sounds at intermediate and higher frequencies, that is: at frequencies above about 100 Hz to 200 Hz.

Thus audio sound is similar to what we mean by ‘normal sound’ except for an overlap where low-frequency sound merges with audio sound –with no clear boundary between the two sound ranges. Below we will describe the results from these studies. 4.1

Audibility of infrasound and low-frequency sound In our 2017 review (Van den Berg and Van Kamp, 2017) we noted that low-frequency sound can be heard daily from road and air traffic and many other sources. Low-frequency sound is usually included as part of the normal frequency range of everyday sounds. Less is known about infrasound and the perception of infrasound is not as common as it is for low-frequency or ‘normal’ sound. However, infrasound is not unique for wind turbines; it is produced by natural sources (storm, surf), big animals, and transport and man-made machinery at levels comparable to what wind farms produce. Due to the high threshold of hearing, we are usually not aware of most of this infrasound. The new literature search yielded 24 publications, of which 17 papers were relevant for this update, two of them dating before 2017. They include two reviews, thirteen original experimental studies, one crosssectional field study and one desk study. Two newly published experimental studies (Maijala et al, 2020 and Krahé et al, 2020) were included after closing the search, thus resulting in 19 relevant papers.

There are several new studies on the possible effects of sounds of low frequency on persons: how do people respond when exposed to a lowpitched sound or even to an inaudible infrasound? Immediate or shortterm reactions can be investigated in an experimental study with short sound samples in a laboratory setting. Longer lasting or long-term Page 43 van 73

RIVM report 2020-0150

reactions are usually investigated in field studies where many people in their own environment are exposed to a number of environmental sources including the one under study, such as a wind farm. Several of the studies reviewed here are part of the European EARS II project on infrasound and ultrasound. This was a follow-up to the first EARS project that expired in 2015. With respect to infrasound the conclusion of this first project was that below about 20 Hz the perception seems to change, and possibly other sensory processes give input to the auditory cortex. Also, for sound levels slightly below the hearing threshold regions of the brain were active that process emotional activity (EARS Communiqué, 2015). An important part of the EARS II project was to investigate brain activity in persons exposed to infrasound, including inaudible infrasound. The EARS II project closed in 2019. Koch (2017) presented an overview of results of the project, including studies mentioned below. Behler and Uhlenkamp (2020) report on the loudness and unpleasantness of sound of either 8 Hz or 32 Hz when presented over 1.5 seconds at the right ear of 19 young, normal hearing persons. The maximum sound level used was 140 dB (presumably to prevent any hearing damage). They furthermore measured brain activity when the same persons were exposed to the sounds in an MRI-scanner.

The individual hearing thresholds at 32 Hz varied from about 60 to 80 dB, at 8 Hz from 90 to 115 dB; these threshold values are comparable to thresholds known from literature. On average the unpleasantness of each sound changed linearly with the perceived loudness, but individually there were large variations. In the MRI-scanner the sounds were again presented at the right ear, now at either low or medium loudness according to each person’s individual loudness scaling. At low loudness (5 out of the maximum 50 loudness units) for both tones there was some activity in the auditory cortex at both sides of the brain. At medium loudness (35 out of 50 loudness units) there was significant activity at the same brain locations. The auditory cortex is the location in the brain were sound is known to be processed and this study shows that this is also true for infrasound. Activation of the auditory cortex in this study was found to correlate better with perceived loudness than with the actual sound level. In another laboratory study Burke et al (2019) investigated whether the ability to hear a sound was influenced by the presence of another sound.

For example: if an infrasound tone without any other sound is just audible, will it still be audible when audio-sound is added? This was tested with 13 young, normal hearing participants using two infrasound tones (5 Hz and 12 Hz) two audio-sound tones (100 Hz and 1000 Hz), and pink noise between 250 and 4000 Hz. First, for each individual the threshold was measured three times and generally for each participant the three outcomes varied over a 5-dB range. Between participants the hearing thresholds varied over 20 dB or more. All thresholds were in the range of thresholds known from literature. Second, a soft audio-sound (5 dB over the threshold of that sound) was added to either the 5 or 12 Hz infrasound. This did not significantly influence the detection threshold for the infrasound. However, the presence of a louder audio-sound (50 dB over the threshold of that sound) on average did lead to a raise of the detection threshold of the infrasound. This shift was significant and Page 44 of 73

RIVM report 2020-0150

ranged from 1 to 9 dB, depending on frequency. In a third experiment it was tested if adding infrasound would change the detection threshold for audio-sound. The detection threshold of audio-sound was tested in the presence of 5 or 12 Hz infrasound. The infrasound levels were very low: either sub-audible (10 dB below the hearing threshold) or barely audible (5 or 10 dB above the hearing threshold) There was no significant effect of the infrasound on the detection threshold of the audio-sound. Thus, a medium loud audio-sound raises the detection threshold for infrasound. But a (very) soft audio-sound has no effect on the audibility of infrasound and vice versa. Burke et al (2019) remark that sound is known to mask other sound of comparable or higher frequency, making the other sound less or not audible. It was therefore unexpected that the audio-sounds could partially mask sound of a much lower frequency. They conclude that apparently masking of infrasound is different from masking audio sounds. However, they also remark that the presence of audio-sound may draw attention to that sound, away from the soft infrasound. In that case, the shift in detection threshold would be les a consequence of audibility and more an effect of attention.

The authors of this study also note that there is a correlation between the individual detection thresholds of the two infrasound tones and between the individual detection thresholds of the audio-sounds, but there is no correlation between the infrasound and audio-sound thresholds. They suggest that this could be a result from different detection mechanisms for infrasound and audio-sound. Weichenberger et al (2017) investigated the effect of infrasound and low-frequency sound at discrete frequencies on brain activity. The hearing thresholds were determined for the right ear of 14 young, normal hearing participants at eight frequencies ranging from 8 to 125 Hz. All thresholds were in the range of thresholds known from literature. After that the participants estimated the loudness at each frequency for sound levels up to the maximum of 124 dB. From these estimates for every participant a medium loud level was determined for the 12 Hz infrasound. In the MRI-scanner participants were exposed to three different conditions: the medium-loud 12 Hz tone was presented to the right ear at either the medium-loud level or 2 dB below the individual threshold, or no sound was presented. Each condition lasted 200 seconds, which is relatively long. When exposed to the medium loud infrasound no corresponding brain activity showed up.

The authors speculate that this may be due to adaptation of the neurons: with a constant stimulus the activation decreases over time. As a result, averaged over the 200 seconds of exposure, the brain activation was not strong enough to show up in the measurements. This contrasts with results of an earlier study with almost the same test persons and the same instrumentation (Weichenberger et al, 2015). In that study the participants were exposed to short bursts (3 seconds) of 12 Hz and medium loudness infrasound, which resulted in significant brain activity in the auditory cortex. Exposure to the 12 Hz infrasound at a level just beneath their individual hearing threshold elicited brain activity not found in the other two conditions. This activity occurred in the auditory cortex and two other Page 45 van 73

RIVM report 2020-0150

brain areas associated with conflict regulation and emotional processing. According to the authors the brain activity at sub-audible level shows there is an unconscious reaction of the body and they speculate that for prolonged exposure there might be a ‘potential link’ with ‘the emergence of various physiological as well as psychological health effects’. Krahé et al (2020) did an extensive experimental study with 39 participants of whom 16 were ‘predisposed’: they had requested authorities or an engineering firm to investigate a problem with infrasound at their home. All participants were exposed to four different infrasounds and complete silence, each for half an hour, in a very quiet, home-like room in a remote building. The infrasound was presented at three frequencies (3 Hz modulated, 5 and 10 Hz unmodulated) at levels close to a standard threshold used in Germany and at 10 dB above this threshold at a frequency of 18 Hz. It was expected that some participants would be able to hear the sound and some not. The results show that, on average, the participants perceived the silence period as not annoying, the period with lower frequencies (3 and 5 Hz) as somewhat annoying, with higher frequencies as moderately annoying. However, for most sounds (3, 10, 18 Hz) individual scores covered the entire scale from not annoying to very annoying.

For the 5 Hz and the silent period scores ranged from not to rather annoying. Participants perceived the 18 Hz sound as rumbling and humming, the others sounds as rumbling and pulsating. Predisposed participants were not found to react differently from the other participants. The authors conclude that ‘essentially’ perception is sensed by the ears, even when there is not always a hearing sensation. Jurado and Marquardt (2020a) investigated the use of EEG as a means to measure the perceived loudness of a very low-frequency sound. With a technique called Frequency Following Response (FFR), electrodes on the head picked up neural activity that was measured as a function of the loudness of the sound. With this, they measured the brain response to a constant sound of either 11 or 38 Hz in 11 young, normal hearing participants. The general trend was that at zero loudness (sound at hearing threshold) the measured signal was close to the background of electric noise in the brain. Generally, with increasing loudness the signal at first increased relatively steeply and above a low to medium loudness remained constant. However, there were large individual differences from this trend and the authors concluded that the FFR signal that was measured here did not correlate with the loudness perceived by the participants. They conclude that FFR is not a useful method to measure loudness.

In another study Marquardt and Jurado (2018) investigated the perception of amplitude modulation in wind turbine sound: if we hear swishing or beating in wind turbine sound, is that just the sound level variation in the WT audio-sound with the frequency of blades passing the tower, or is there an (added) effect of the infrasound peak at that same frequency? The perception of the (simplified) phenomenon was investigated for two sounds at discrete frequencies and either a modulation of the tone at 8 Hz or an 8 Hz infrasound tone added to the unmodulated sound. The variation in the amplitude (strength) of the modulated 63 or 125 Hz sound was 25% or 37.5% of the original tone Page 46 of 73

RIVM report 2020-0150

amplitude. With a separate test the level of the 8 Hz infrasound was determined that would give the same loudness to the combination of infrasound plus 63/125 Hz compared to the amplitude-modulated 63/125 Hz tone. Then, all the different sounds of each 1.2 second duration were played many times in random order, with a total of 400 samples. 12 normal hearing, young participants had to decide for every sample if it contained infrasound. For each of the different sounds the percentage of correct answers was not significantly different from what one can expect from pure guessing. The authors conclude that a combination of a tone together with a constant 8 Hz infrasound is similar to our ears to the tone amplitude modulated at 8 Hz (without the infrasound): we cannot hear a difference between both. Jurado and Gordillo (2019) investigated if fluctuations in the level of a low-frequency sound influenced the perceived loudness of that sound. This was tested with 24 young, normal hearing participants who matched the loudness of three simple low-frequency tones (40, 63 or 80 Hz) and one 1000 Hz tone with a number of tone combinations. Each combination consisted of two tones close in frequency to one of the three simple tones. The combination produces fluctuations (variations in amplitude) at a frequency equal to the difference in the frequencies of both tones.

The differences in frequency were 1, 2, 5 and 12 Hz. The results show that the effect of fluctuation at the lower frequencies on loudness was modest and corresponded to 2 dB or less. The results were in agreement with what was already known in literature and loudness models. Based on a number of publications, including some mentioned above, in their review Carlile et al (2018) summarize that small differences in the infrasound hearing threshold can result in large differences in sensation because of the steep increase of loudness at infrasound frequencies. This was already reported by Moller and Pedersen (2004) and the more recent studies (Koch et al, 2017) have confirmed this. 4.2

Effect of lower frequencies compared to ‘normal’ sound Infrasound and low-frequency sound from wind turbines have been suggested to affect the health of residents in a way other than audiosound does, but at the time of our 2017 review there was little scientific evidence to support this hypothesis. Other effects, such as vibration of the body, nausea or dizziness, have been shown to occur in laboratory experiments, but only at higher levels of infrasound compared to those from wind turbines. Here we will also review studies that may reveal specific effects of infrasound and low-frequency sound, regardless whether they include wind turbines or not. In his narrative review of the possible effects of infrasound from wind turbines Tonin (2017) concludes that a large number of measurements have shown that wind turbine infrasound is below the threshold of hearing. There is disagreement about whether this means it cannot have an effect. He notes a lack of studies in which persons are ‘intentionally subjected to infrasound and a response is measured’. However, in recent years some relevant laboratory studies have been published and these will be summarized below. Page 47 van 73

RIVM report 2020-0150

In a large survey in the Netherlands Van Kamp et al (2017) asked residents in three cities if they were annoyed by a low-frequency ‘humming’ noise, for example from ventilators. 7% of the almost 4000 participants in this study were highly annoyed by such noise. Other noise sources (road traffic, construction works, mopeds and neighbours) each led to higher numbers of annoyed persons, ranging from 13% to 22% of the participants. Some sources (rail traffic and industry) led to less annoyance (each about 4%). Persons dissatisfied with their residential situation as well as noise-sensitive persons reported more annoyance compared to people scoring high on residential satisfaction and/or low on noise sensitivity. In the daytime the percentage of persons highly annoyed by humming sounds was higher when background sound levels from road traffic were higher. At night the reverse was true: a higher background level was related to somewhat less annoyance from humming sounds. There was no correlation between annoyance from humming sounds and sound insulation at the façade (double glazing, wall cavity wall filling, absorbing ventilation grille). As, according to a conference paper by Krahe (2019) “low-frequency noise is strongly annoying”, he devised a method to rate the annoyance from WT low-frequency and infrasound.

This was done by giving a penalty when the sound had one or more annoying features such as pulsating or a rhythmic variation in level or a tonal quality. From this, penalties were calculated based on knowledge of the annoyance of each feature. The calculated penalty values were compared to the assessments of a number of sounds by a panel of 23 experts. They each had to rate the annoyance of the sounds by giving it a penalty of either 3 or 6 dB, or zero (no penalty). After modifying the original calculations, a correlation was apparent between the calculated penalty values and the experts’ assessments. This yet must be extended to infrasound and tested in field studies. The literature search yielded two laboratory studies of possible adverse effects typical for infrasound and/or low-frequency sound. Stevens and Martens (2018) investigated the effect when either identical or somewhat different stimuli are presented to each ear. In the everyday environment sound from a sound source often reaches one ear earlier than the other ear, and at high frequencies, where a sound wave is shorter than the dimensions of the head, this means the ears do not receive quite the same sound at the same time. This is less so for low frequencies, where the wave length is much larger than the head and the difference between the left and right ear is very small.

However, the authors suggest that when outdoor sound is audible indoors, due to reflections also at low frequencies the ears do not receive the same sounds. The basis of each stimulus they used in their study was one of three similar sounds with most audible energy at frequencies between 100 and 200 Hz. Each sound could be modulated at either 2, 5 or 10 Hz; that is, the sound varied in strength in a rhythm of 2, 5 or 10 times per second. Then the sound was presented through one loudspeaker at each side of the participant in three combinations: 1) the basic, unmodulated sound at both ears; 2) the modulated sound at both ears; and 3) the unmodulated sounds at one ear and the modulated one at the other. On average the 21 participants rated incoherent sound (stimulus 3) as most Page 48 of 73

RIVM report 2020-0150

annoying. Also, the annoyance was higher when the modulation frequency (at one of the ears) was 2 Hz, when compared to 5 and 10 Hz. The authors remark that these results are in line with earlier studies which showed that greater differences between the ears create a sense of spaciousness and this is usually experienced as unpleasant. In a Finnish report, not yet published in a scientific journal, Maijala et al (2020) describe a survey, sound measurements and laboratory experiments that were set up in order to investigate the role of infrasound in health complaints related to wind farms. 70 out of 1351 survey respondents (5%) reported symptoms they attributed to infrasound from a wind farm. On average these ‘symptomatic respondents’ lived closer to a wind farm than those without symptoms. For these respondents a number of factors were significantly associated with their symptoms: having chronic diseases, being annoyed by different aspects of wind turbines and considering wind turbines as a health risk. Of all respondents, 10% considered WT infrasound a high risk to their personal health; 18% considered it as a high risk to health in general. The sound measurements were performed in two uninhabited countryside dwellings at 1.5 km from a wind farm.

From the measurements it was concluded that in these dwellings the infrasound levels were like levels occurring typically in an urban environment. The highest levels were at 0.1 to 1 Hz, but at these extreme low frequencies the levels were well below the hearing threshold. Recordings with the highest infrasound and amplitude modulation levels were selected for the laboratory experiments. The results of these experiments were: 1) those that had reported WT (infra)sound related symptoms did not exhibit an increased sensitivity for WT infrasound; 2) total WT sound level and amplitude modulation were a cause for increased annoyance, not infrasound; 3) WT infrasound or WT sound annoyance had no association with heart rate or heart rate variability, nor with skin conductance (as physiological measures of stress). In the report of Krahé et al (2020), already mentioned in 4.1, participants were submitted to several physiological tests during exposure. Tests involved blood pressure, heart rate and EEG. Tests showed no differences between the different exposures and no differences between predisposed participants and others. 4.3

Sub-audible including vestibular effects In our 2017 review we concluded that residential levels of infrasound from wind turbines are not strong enough to affect the sense of balance. Also, at the present levels of wind turbine sound, the occurrence of syndromes (not medically accepted as diagnosis) vibroacoustic disease (VAD) or the ‘visceral vibratory vestibular disease’ (VVVD) causing the wind turbine syndrome (WTS) is unproven and unlikely. However, symptoms associated with WT sound could result from stress, possibly in relation to the presence of a wind farm. In recent years no studies were published that support the existence of VAD or the VVVD. Jurado and Marquardt (2020b) investigated the effect of airborne infrasound on the vestibular system. They used a clinical method to assess the functioning of the vestibular system by measuring the electric Page 49 van 73