Exposure to wind turbine noise : Perceptual responses and reported health effects

Health Canada, in collaboration with Statistics Canada, and other external experts, conducted the Community Noise and Health Study to better understand the impacts of wind turbine noise (WTN) on health and well-being. A cross-sectional epidemiological study was carried out between May and September 2013 in southwestern Ontario and Prince Edward Island on 1238 randomly selected participants (606 males, 632 females) aged 18–79 years, living between 0.25 and 11.22 km from operational wind turbines. Calculated outdoor WTN levels at the dwelling reached 46 dBA. Response rate was 78.9% and did not significantly differ across sample strata. Self-reported health effects (e.g., migraines, tinnitus, dizziness, etc.), sleep disturbance, sleep disorders, quality of life, and perceived stress were not related to WTN levels. Visual and auditory perception of wind turbines as reported by respondents increased significantly with increasing WTN levels as did high annoyance toward several wind turbine features, including the following: noise, blinking lights, shadow flicker, visual impacts, and vibrations. Concern for physical safety and closing bedroom windows to reduce WTN during sleep also increased with increasing WTN levels. Other sample characteristics are discussed in relation to WTN levels. Beyond annoyance, results do not support an association between exposure to WTN up to 46 dBA and the evaluated health-related endpoints. VC 2016 Crown in Right of Canada. All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). [http://dx.doi.org/10.1121/1.4942391]

intention to undertake a large scale epidemiological study in collaboration with Statistics Canada entitled Community Noise and Health Study (CNHS). Statistics Canada is the federal government department responsible for producing statistics relevant to Canadians.
In comparison to the scientific literature that exists for other sources of environmental noise, there are few original peer-reviewed field studies that have investigated the community response to modern wind turbines. The studies that have been conducted to date differ substantially in terrns of their design and evaluated endpoints (Krogh et al. , 201 I;Mroczek et al., 2012;Mroczek et al., 2015;Nissenbaum et al. , 2012;Pawlaczyk-Luszczynska et al., 2014;Pedersen andPersson Waye, 2004, 2007;Pedersen et al., 2009;Shepherd et al., 2011 ;Tachibana et al., 2012;Tachibana et al., 2014;Kuwano et al., 2014). Common features among these studies include reliance upon sel f-repo11e I endpoi nL~. modeled WTN exposure and/or proximity to wind turbines as the explanatory variable for the observed community response.
There are numerous health symptoms attributed to WTN exposure including, but not limited to, cardiovascular effects, vertigo, tinnitus, anxiety, depression, migraines, sleep disturbance, and annoyance. Health effects and exposure to WTN have been subjected to several reviews and the general consensus to emerge to date is that the most robust evidence is for an association between exposure to WTN and community annoyance with inconsistent support observed for subjective sleep disturbance (Bakker et al., 2012;Council of Canadian Academies, 2015;Knopper et al., 2014;MassDEP MDPH, 2012;McCunney et al., 2014;Merlin et al., 2014;Pedersen, 2011 ).
The current analysis provides an account of the sample demographics, response rates, and observed prevalence rates for the various self-reported measures as a function of the outdoor WTN levels calculated in the CNHS.

A. Sample design
Factors considered in the determination of the study sample size, including statistical power, have been described by Michaud et al. (2013), Michaud et al. (2016b, and Feder et al. (2015). TI1e target population consisted of adults, aged 18 to 79 years, living in communities within approximately 10 km of a wind turbine in southwestern Ontario (ON) and Prince Edward Island (PEI). Selected areas in both provinces were characterized by flat lands with rural/semi-rural type environments. Prior to field work, a list of addresses (i.e., potential dwellings) was developed by Statistics Canada. The list consists mostly of dwellings, but it can include industrial facilities, churches, demolished/vacant dwellings, etc. (i.e., non-dwellings), that would be classified as out-of-scope for the purposes of the CNHS. The ON and PEI sampling areas included 315 and 84 wind turbines, respectively. Wind turbine electrical power output ranged between 660kW to 3 MW (average 2.0 ::+:: 0.4 MW). All turbines were modem design with 3 pitch controlled rotor blades (~80 m diameter) upwind of the tower, and predominantly 80 m hub heights. This study was approved by the Health Canada and Public Health Agency of Canada Research Ethics Board (Protocols #2012--0065 and #2012-0072).

B. Wind turbine sound pressure levels at dwellings
A detailed description of the approach applied to sound pressure level modeling [including background nighttime sound pressure (BNTS) levels] is presented separately (Keith et al., 2016b). Briefly, sound pressure levels were estimated at each dwelling using both ISO (1993) and ISO (19%) as incorporated in the commercial software CadnaA version 4.4 (Datakustik, 2014). The calculations were based on manufacturers' octave band sound power spectra at IO m height, 8 m/s wind speed for favorable propagation conditions (Keith et al., 2016a). As described in detail by Keith et al. (2016b), BNTS levels were calculated following provincial noise regulations for Alberta, Canada (Alberta Utilities Commission, 2013). With this approach BNTS levels can range between 35 dBA to SI dBA. The possibility that BNTS levels due to highway road traffi c noise exposure may exceed the level estimated by Alberta regulations was considered. Where the upper limits of this approach were exceeded (i.e., S 1 dB), nighttime levels were derived using the US Traffic Noise Model (United States Department of Transportation, 1998) module in the CadnaA software.
Low frequency noise was estimated in the CNHS by calculating outdoor C-weighted sound pressure levels at all dwellings. There was no additional gain by analysing the data using C-weighted levels because the statistical correlation between C-weighted and A-weighted levels was very high (i.e., r = 0.81--0.97) (Keith et al., 2016a).

Questionnaire content and collection
The final questionnaire, available on the Statistics Canada website (Statistics Canada, 2014) and in the supplementary materials, 1 consisted of basic socio-demographics, modules on community noise and annoyance, health effects, lifestyle behaviors and prevalent chronic illnesses. In addition to these modules, validated psychometric scales were incorporated, without modification, to assess perceived stress (Cohen et al., 1983), quality of life (WHOQOL Group, 1998;Skevington et al., 2004) and sleep disturbance (Buysse et al., 1989).
Questionnaire data were collected through in-person home interviews by 16 Statistics Canada trained interviewers between May and September 2013. The study was introduced as the "Community Noise and Health Study" as a means of masking the true intent of the study, which was to investigate the association between health and WTN exposure. All identified dwellings within ~600 m from a wind turbine were selected. Between 600 m and 11.22 km, dwellings were randomly selected. Once a roster of adults (between the ages of I 8 and 79 years) living in the dwelling was compiled, one individual from each household was randomly invited to participate. No substitutions were permitted under any circumstances. Participants were not compensated for their participation.

Long-term high annoyance
To evaluate the prevalence of annoyance, part1c1pants were initially asked to spontaneously identify sources of noise they hear originating from outdoors while they are either inside or outside their home. The interviewer grouped the responses as road traffic, aircraft, railway/trains, wind turbine, and "other." Follow-up questions were designed to confinn the initial response where the participant may not have spontaneously identified wind turbines, rail, road and aircraft as one of the audible sources. For each audible noise source participants were asked to respond to the following question from ISO/fS (2003a): "Thinking ahout the last year or so, when you are at home, how much does noise fi'0/11 [SOURCE] bother, disturb or annoy you'!" Response categories included the following: "not at all," "slightly," "moderately," "ve,y," or "extremely." Participants who reported they did not hear a particular source of noise, were classified into a "do not hear" group and retained in analysis (to ensure that the correct sample size was accounted for in the modeling). The analysis of annoyance was performed after collapsing the response categories into two groups (i.e., "highly annoyet.f' and "11ot highly a11noyed"). As per ISO/fS (2003a), participants reporting to be either "ve,y" or "extremely" annoyed were treated as "highly annoyed" in the analysis. The "not highly a11noyecf' group wa~ composed of participants from the remaining response categories in addition to those who did not hear wind turbines. Similarly, an analysis of the percentage highly subjectively sleep disturbed, highly noise sensitive, and highly concerned about physical safety from having wind turbines in the area was carried out applying the same classification approach used for annoyance.
The use of filter questions and an assessment of annoyance using only an adjectival scale are approaches not recommended by ISO/TS (2003a). The procedures followed in the current study were chosen to minimize the possibility of participant confusion (i .e., by asking how annoyed they are toward the noise from a source that may not be audible). Although there is value in confirming the response on the adjectival scale with a numerical scale, this approach would have added length to the questionnaire, or led to the removal of other questions. Collectively, the deviations from ISO/TS (2003a) conformed to the recommendations by Statistics Canada and to the approach adopted in a large-scale study conducted by Pedersen et al. (2009).
The questionnaire assessed pru1icipant's long-term (~ l year) annoyance to WTN in general (i.e., location not specified), and specifically with respect to location (outdoors, indoors), time of day (morning, afternoon, evening, nighttime) and season (spring, summer, fall, winter). In addition, participants' long-term annoyance in general, to road, aircraft and rail noise was assessed. These evaluations of annoyance are considered to be clustered because they are derived from the same individuals (i.e., they are repeated measures) . Therefore, in order to compare the prevalence of annoyance as a function of location, time of day, season, or noise source, generalized estimating equations for repeated measures were used to account for the clustered responses (Liang and Zeger, 1986;Stokes et al., 2000).
Statistical analysis was performed using SAS version 9.2 (SAS Institute Inc., 2014). A 5% statistical significance level is implemented throughout unless otherwise stated. In addition, Bonferroni corrections are made to account for all pairwise comparisons to ensure that the overall type I (false positive) error rate is less than 0.05. In cases where cell frequencies were small (i.e., <5) in the contingency tables or logistic regression models, exact tests were used a~ described in Agresti (2002) and Stokes et al. (2000).

A. Wind turbine sound pressure levels at dwellings
Modeled sound pressure levels, and the field measurements used to support the models are presented in detail by Keith et al. (2016a,b). Calculated outdoor sound pressure levels at the dwellings reached levels as high as 46 dB. Unless otherwise stated, all decibel references are A-weighted. Calculations are likely to yield typical worst case long-term (I years) average WTN levels (Keith et al., 2016b).

B. Response rate
Of the 2004 addresses (i.e., potential dwellings) on the sample roster, 434 dwellings were coded as out-of-scope by Statistics Canada during data collection (Table I). This was consistent with previous surveys conducted in rural areas in Canada (Statistics Canada, 2008). In the current study, 26.7% and 20.4% of addresses were deemed out-of-scope in PEI and ON, respectively. No significant difference in the distribution of out-of-scope locations by distance to the nearest wind turbine was observed in PEI (/ = 3.19, p = 0.5263). In ON, a higher proportion of out-of-scope addresses was observed in the closest distance group (:S:0.55 km) compared to other distance groups (p < 0.05, in all cases). After adjusting for province, there was a The ochmn Mrmtc l-Hacnszcl chi-St1unrc 1cs1 is used m udjust for provi nce. ,~vriluc., < 1 105 11re co1tsidcrcd lO be s1111is1ic:11ly ·ignlli<:ant.
''Totnl nu111[\cr of JX)ICntinl clwcll ing,, out of sct1pc (given is a 1icrcc 11Utlle of 1owl potential th cll inlts) is broken doll'n by pmvi11L-c. tis well it is <XJUIII 10 the sum of Code A-F. 11,c percentages of dwell ings that are coded as ou1-0f-scope are based on the total num ber of fK)te 111ia l dwcl lings in the area. Code A-address was a business/du plicate/other ( 17%), address li sted in error (83%). Code B-an inhabitable dwell ing unoccupied ,It the time of the survey. newly constnictcd dwell ing 11 01 yc r inhabited. ;1 vuc1mt tntilcr in a conm1cR:ial lmilcr pur k. ode : ummc r cm 111gc, ski chulcc. or hunt ing <'mll(!S, C1Xlc 111 p:1rt icipm11s in Ille dwe ll ing were > 79 years of age. Code E-undcr construction, institution, or unavailable lo participate. Code r~dcmolished for unk nown reasons. "Chi-square test of independence. significant association between distance groups and the proportion of locations assigned a Code A (p = 0.0068) ( Table 1). A post-collection screening of interviewer notes by Statistics Canada has confirmed that of the total number of Code A locations, the vast majority (i .e., 83%) were locations listed in error. In rural areas, there is more uncertainty in developing the address list frame and this can contribute to a higher prevalence of addresses listed in error within 0.55 km of a wind turbine where the population density is lower compared to areas at greater setbacks. 2 The remaining 1570 addresses were considered to be valid dwellings, from which 1238 residents agreed to participate in the study (606 males, 632 females). This resulted in a final response rate of 78.9%, which was not statistically different between ON and PEI or by proximity to wind turbines (Table II). employment was the only variable that appeared to consistently increase within increasing WTN levels. Household income and education were unrelated to WTN levels. There was no obvious pattern to the changes observed in the other variables that were found to be statistically related to WTN level categories (i.e., age , type of dwelling, property ownership and fac ade type).

D. Perception of community noise and related variables as a function of WTN level
The prevalence of reporting to be very or extremely (i.e., highly) noise sensitive was statistically similar across all WTN categories (p = 0.8 I 75). As expected and as shown in Fig . 1, visibility and audibility of wind turbines increased with increasing WTN levels.
The overall audibility of other noise sources is shown in Table IV. Not shown in Table IV is how often the noise source was spontaneously repo11ed as opposed to being identified following a prompt by the interviewer (see Sec. II). "Potential parti cipants from locations established to be valid dwellings (equal to the diffe rence betwee n "T otal potential dwe llings" and " total number of potential dwe llings out-of-scope"; see Table I) used in the derivation of partic ipation rates . ~he C MH c hi -s4uarc tes t is used lo adjust fo r prnvincc, p -values < 0.05 arc conside red lo be statistically significant. "Chi -square test of independence. "The Cochran Mantel-Haenszel chi-square test is used to adjust fo r province unless otherwise indicated, p -values < 0.05 are considered to be statistically significant. "'rota ls may differ due to mi ssi ng data. c Amllysis of varia nce (ANOV A) mode l. ~Non-parametric two-way ANOV A model adjusted for province. "No11 -dc1achcd dwellin •s incl uded scmi/d uph:Nu1,u nmcn1 .
r hi-squurc test uf iml •111:ndcncc. Among the participants who reported hearing each specific noise source, the prevalence of spontaneously reporting road traffic, wind turbines, rail and aircraft was 84%, 71 %, 66%, and 30%, respectively. A total of l02 participants (8.2%) indicated that there were no audible noise sources around their home. These participants lived in areas where the average WTN levels were 32.4dB [standard deviation (SD)= 8.3] and the mean distance to the nearest turbine was 1.7 km (SD= 2.0) (data not shown). Table IV also provides the observed prevalence rates for high (i.e., very or extreme) annoyance toward wind turbine features. The results suggest that there was a tendency for the prevalence of annoyance to increase with increasing WTN leve ls, with the rise in annoyance becoming evident when WTN levels exceeded 35 dB. The pattern was slightly different for visual annoyance among participants drawn from the ON samp le, where there was a noticeable rise in annoyance among participants living in areas where WTN   ,, hi•N (!llill'0 tCsl of independence. <Nobody reported hearing rn il noise in PEI as 1herc is no rail activity in PEI. therefore the percent is given as a percentage of ON participants only. 'Rcfors Ill 1111yonc In the partkipunt ·~ hou~chnld ever ludgi1 1g II ronn:rl co111ph1i11t (includ ing slg11 i11g .i pccition) reganJing noise l'ru111 wind wrhincs. rRcnsons for closing lx:dr'Oom wi ndows due 11111ircrun noi. ,vns .suppres.scd due LO low cell coun~ (l.c .. 11 <5 overall). levels were between (25 and 30) dB. The prevalence of household complaints concerning wind turbines, which could include signing a petition regarding noise from wind turbines, was 2.8% overall and unrelated to WTN levels (p = 0.2578). However, complaints were found to be greater among the PEI sample (13/224 = 5.8%), compared to ON (22/ IO LO= 2.2%) (p = 0.0050).
Other notable observations from Table IV include the finding that the number of participants who self-reported to personally benefit in any way (e.g., rent, payments or indirect benefits such as community improvements) from having turbines in their area, was not equally distributed among provinces. In ON, reporting such benefits was significantly related to WTN categories (JJ < 0.000 I) and there was a gradual increase from the lowest WTN category ( <25 dB: 0.0%) to the loudest WTN category ((40-46] dB: 21.4%), whereas in PEI benefits were statistically evenly distributed across the sample (p = 0. 1700).
Closing bedroom windows to block outside noise during sleep was equally prevalent across all WTN categories (p = 0.8106); however, identifying WTs as the reason for closing the window was found to be related to WTN levels (p < 0.000 I). In the two loudest categories, [35-40) dB and (40-46] dB, 15.2% and 21.6% of participants identified WTN as the reason for closing bedroom windows, respectively, compared to :::;2. l % in the other WTN categories (Table IV). Figure 2 plots the fitted percentage highly annoyed by WTN category overall and for ON and PEI separately. WTN annoyance was observed to significantly increase when WTN levels exceeded 2:35 dB compared with lower exposure categories (p < 0.009, in all cases). Overall, observed prevalences of noise annoyance increased from less than 2. In addition to asking participants how annoyed they were toward WTN in general (i.e., without reference to their particular location), other questions were designed to assess annoyance as a function of location (i.e., indoors, outdoors). As shown in Fig. 3, the prevalence of high annoyance was significantly higher outdoors.
The prevalence of annoyance by time of day and season is provided in Table IV. For WTN levels below 30dB, the prevalence of high annoyance was very low ( < 1.2%) and similar for all times of day. Starting at 30 dB, the percentage highly annoyed during the evening and nighttime were significantly higher than the morning and afternoon; however this difference was most pronounced at WfN levels 2:35 dB. For WTN levels below 30dB, the prevalence of high annoyance was very low ( <2.2%) and similar for all seasons. At WTN levels 2:35 dB, the prevalence of high annoyance during the summer was higher compared to all other seasons.
Noise annoyance toward road, aircraft and rail noise was also assessed in the questionnaire. It was of interest to determine how annoyance to these sources compared to WTN nnnoyance. ln areas where WTN lcv Is wer• < 35 tlB the greatest source of noise annoyance was road traffic . In WTN categories 2:35 dB, annoyance toward WTN exceeded all other sources (p <0.0003, in all cases) (see Fig. 4). Table V shows that subjectively reported sleep disturbance from any source while sleeping at home over the last year, in addition to a multitude of health effects, were found Participants were asked to think about the last year or so and indicate how bothered, disturbed or annoyed they were by WTN while at home. The percentage of participants reporting to be either very or extremely (i.e., highly) bothered, disturbed or annoyed is shown as a function of calculated outdoor Aweighted WTN levels al the dwelling (<lBA). Figure 3 presents the tilted results by locution (i.e., indoors and outdoors) along with their 95% conlidence intervals. + Indoor sig nificantly different from outdoor (p < 0.001).  4. Prevalence of high annoyance toward different noise sources as a funct ion of calculated outdoo r wind turbine noise levels. lllustmtes the percentage of participants that reported to be either very or extremely (i.e ., highly) bothered, disturbed o r annoyed by road tmffic, aircraft , rail and wi nd turbine noise (WTN) wh ile at home over the last year. At home refers to either inside or outside the dwe ll ing. Resul ts represe nt fitted data along wi th their 95% confidence intervals and are shown as a function of calcu lated outdoor A-weighted WTN levels at the dwelling (dBA). 'WTN significantly different from road trnfflc and rai l noi se (p < 0.001); H WTN signitic.u11ly d ifferent from road traffic (p < 0.001 ); H-·t WTN significantly different from airc raft noise (p < 0.00 1 ). +++ 1 WTN sign ificantly different from road traflic , mi l, and ai rcraft no ise (p < 0 ,0003).

E. Self-reported health conditions and use of medication
to be unrelated to WTN levels. Similarly, medication use for high blood pressure, anxiety or depression was also fou nd to be unrelated to WTN levels. Although sleep medicati on use was significantly related to WTN levels (p = 0.0083), the prevalence was higher among the two lowest WTN categories { < 25 dB and [25-30) dB l (see Table V).

IV. DISCUSSION
The prevalence of self-reporting to be either "very" or "extremely" (i.e ., highly) annoyed with several wind turbine features increased significantly with increasing A-weighted WTN levels. When classified by the prevalence of reported annoyance overall, and in areas where WTN levels exceeded 35 dB, annoyance was highest for visual aspects of wind turbines, followed by blinking lights, shadow flicker, noise and vibrations . Consistent with Pedersen et al. (2009), the increase in WTN annoyance was clearly evident when moving from [30-35) dB to [35][36][37][38][39][40] dB, where the prevalence of WTN annoyance increased from 1 % to 10%. This conlinued to increase to 13.7% for areas where WTN levels were 140-461 dB . The prevalence of WTN annoyance wru higher outdoors, during the summer, and during evening and nighttime hours. Pedersen et al. (2009) also fou nd chat annoyance with WTN was greater outdoors compared to indoors.
Despite a similar pattern of response between the ON and PEI samples, the self-reported WTN annoyance was 3.29 times greater in ON, a difference thal was most pronounced at the two highest WTN categories. This difference is in contrast to the prevalence of household complaints related to wind turbines . Even though the overall prevalence of such complaints was low (i.e., 2.8%), complaints were more likely in PEI (5.8%) compared 10 ON (2.2%). The reasons for this difference despite greater reported annoyance in ON are unclear. Research has shown that there are several contingencies that must be met before someone that is highly annoyed will complain (Michaud et al., 2008). Such contingencies include knowing who to complain to, how to file a complaint and holding the belief that the complaint will result in positive change. The fact that the prevalence of complaints regarding wind turbines was unrelated to WTN levels is another indication that complaints do not always correlate well with changes in noise exposure (Fidell et al., 1991 ). The motives underlying household complaints were not assessed in the present study, but the disparity found with annoyance could also be related to the wording used in the questionnaire. The prevalence of complaints was the one question where the respondent answered on behalf of the entire household.
More participants reported that they were highly annoyed by the visual aspects of wind turbines than by any other feature, even at higher WTN levels. Similar to WTN annoyance, the overall prevalence of annoyance with the visual impact of wind turbines was more than twice as high in the ON sample, and more prevalent across the exposure categories when compared to PEI. In the PEI sample, no participants reported visual annoyance in areas where WTN levels were below 35 dB. This is in contrast to a clear intensification in visual annoyance among the ON sample in areas where WTN levels were [25][26][27][28][29][30] dB. Exploring the variables that may underscore provincial differences was not within the scope of the current study. The questionnaire was not designed to probe underlying factors that may explain observed provincial differences; however, reported personal benefit from having wind turbines in the area was found to be different between the ON and PEI samples (Table IV). Shepherd et al. (2011) assessed annoyance in response to WTN, but not in a manner that would pennit comparisons with the Swedish (Pedersen andPersson Waye, 2004, 2007), Dutch (Janssen et al., 2011 ;Pedersen et al., 2009) or 1he current study. Shepherd et al. (2011) reported that 59% of participants living within 2 km of a wind turbine installation spontaneously identified wind turbines as an annoying noise source, with a mean annoyance rating of 4.59 (SD, 0.65) when the 5 category adjectival scale was analyzed as a numerical scale from O to 5. No exposure-response relationship could be assessed because the authors did not provide an analysis based on precise distance or as a function of WrN levels, which they reported to be between 20 and 50 dB among participants living within 2 km of a wind turbine. This encompasses the entire WTN level range in the CNHS. As such, the only tentative comparison that can be made between the current study and the Shepherd et al. (2011) study would be that the observed prevalence of highly annoyed (i .e ., "very" or "extremely") within 2 km of the nearest wind turbine was 7 .0%. These data are not shown because the focus of the current study was on WTN levels and an analysis based solely on distance to the nearest turbine does not adequately account fo r WTN levels at any given <lweiling. WTN is a more sensitive measure of exposure level because, in addition to the distance to the turbine, it accounts for topography, presence of large bodies of water, wind turbine characteristics, the layout of the wind fatm and the number of wind turbines at any given distance. "The Cochran Mantcl-Haenszel chi-square tesl is used lo adj ust for provinces unless otherw ise indicated, p-va lues <0.05 ,ire considered lo be statistically signi fica nt. "Columns may not add 10 total due IO missing data. ' Wor.;e consists of the lwo rati ngs: "Somewhat war."' no w" and "M uch worse now." JHigh sleep disturbance consists of the two ra tings: " 1•e1)"' and "e.rtrr!mely" s leep d isturbed. "Chi-square lest of inde pendence. ,.Quality of Life (QoL) and Satisfaction with Health were assessed wi th the two staml-a lone questions on lhe W HOQOL-BREF. Reporting "poor" overall QoL reflects a response of "poor" or " very poor," and ''.~ootf' refl ects a response of "neither poor nor .~0 01/, " "good," or " l'e1:y good." Report ing "clissati.gie,f' overall Satisfaction with Hea llh reflccls a response of " l'ery dissati.gietf' or "dissati.;/ied," and "satisfiec f' rellecls a response of "neither sati.gied 11or dissatisfied," "satisfied," or " 1•ery satisjied." A detailed prese ntatio n of the resul ts re lated to Qo L is presented by Feder et al. (2015).
It was important to assess the extent to which the sample was homogenously distributed, with respect to demographics and community noise exposure. The reason for this is that the validity of the exposure-response relationship is strengthened when the primary distinction across the sample is the exposure of interest; in this case, WTN levels. Demographically, some minor differences were found with respect to age, employment, type of dwelling and dwelling ownership; however, with the possible exception of employment, these factors showed no obvious pattern with WTN levels and none were strong enough to exert an influence on the overall results. At the design stage, there was some concern that selecting participants up to 10 km might result in an unequal exposure to community noise sources other than WTN. This may have an influence on the underlying response to WTN. Limited data availability did not permit the modeling of sound pressure levels from other noi se sources as originally intended, however it was possible to model BNTS levels. Although Fields (1993) concluded that background sound levels generally do not influence community annoyance, his review did not include wind turbines as a noise source and in the current study BNTS levels were calculated to be lower in areas where WTN levels were higher. Lower BNTS could contribute to a greater expectation of peace and quiet. Therefore, a limitation in the CNHS may be that the expectation of peace and quiet was not explicitly evaluated. This factor may influence the association between long-term sound levels and annoyance by an equivalent of up to 10 dB (ANSI, 1996;ISO, 2003b). The influence this factor may have had on the exposure-response relationship found specifically between WTN levels and the prevalence of reporting high annoyance with WTN in the Cl-INS is discussed in Michaud et al. (20 l6a).
In the absence of modeling, the audibility of road traffic, aircraft and rail noise provided a crude indication of exposure to these sources. In general , road traffic noise exposure was heard by the vast majority of the sample (82. J %).
Aircraft noise was uniformly audible in ON by about half the sample; in PEI however, hearing aircraft was more common in the higher WTN exposure categories (i.e., above 35 dB) where between 61 % and 66% of the respondents indicated that they could hear aircraft. Future research may benefit from assessing the extent to which audible aircraft noise may have influenced the annoyance with WTN in PEI. Only when WTN levels were [40--46] dB was the audibility of wind turbines comparable to road traffic (i.e., both sources were audible by approximately 81 % of participants). For these community noise sources, participants were asked how bothered, disturbed, or annoyed they were while at home over the last year or so. The findings are of interest in light of the source comparisons made by Pedersen et al. (2009) andJanssen et al. (2011) , which placed WTN annoyance above all transportation noise sources when comparing them at equal sound levels. In the current study, the overall annoyance toward WTN (7.2%) was found to be higher in comparison to road (3.8%), aircraft (0.4%), and rail in ON (1.9%). Source comparisons need to be made with caution because the observed source differences in annoyance may result from an actual difference in sound pressure levels al the dwellings in this study. Modeling the sound levels from transportation noise sources in the current study would allow a more direct comparison between these sources and WT annoyance at equivalent sound exposures. Another approach is to assess the relative community tolerance level of WTN with that reported for road and aircraft noise studies. This analysis indicates that there is a lower community tolerance level for WTN when compared to both road and aircraft noise at equivalent sound levels (Michaud et al., 2016a).
The list of symptoms that are claimed to be caused by exposure to WTN is considerable (Chapman, 2013), but there is a lack of robust evidence from epidemiological studies to support these associations (Council of Canadian Academies, 2015;Knapper et al., 2014;MassDEP MDPH, 2012;McCunney et al., 2014;Merlin et al., 2014). The results from the current study did not show any statistically significant increase in the self-reported prevalence of chronic pain, asthma, arthritis, high blood pressure, bronchitis, emphysema, chronic obstructive pulmonary disease (COPD), diabetes, heart disease, migraines/headaches, dizziness, or tinnitus in relation to WTN exposure up to 46 dB. In other words, individuals with these conditions were equally distributed among WTN exposure categories. Similarly, the prevalence of reporting to be highly sleep disturbed (for any reason) and being diagnosed with a sleep disorder were unrelated to WTN exposure. These self-reported findings are consistent with the conclusions reached following an analysis of objectively measured sleep among a subsample of the current study participants (Michaud et al., 2016b). Medication use (for anxiety , depression, or high blood pressure) was unrelated to WTN levels. It is notable that the observed pievalence for many of the aforementioned health effects are remarkably consistent with large-scale national population-based studies (Innes et al., 2011 ;Kroenke and Price, 1993;Morin et al., 2011 ;O'Brien et al., 1994;Shargorodsky et al., 2010).

V. CONCLUDING REMARKS
Study findings indicate that annoyance toward all features related to wind turbines, including noise, vibrations, shadow flicker, aircraft warning lights and the visual impact, increased as WTN levels increased. The observed increase in annoyance tended to occur when WTN levels exceeded 35 dB and were undiminished between 40 and 46 dB. Beyond annoyance, the current study does not support an association between exposures to WTN up to 46 dB and the evaluated health-related endpoints. In some cases, there were clear differences between the southwestern ON and PEI participants; however, exploring the basis beh ind these differences fell outside the study scope and objectives. The CNHS supported the development of a model for community annoyance toward WTN, which identifies some of the factors that may influence this response (Michaud et al., 2016a). At the very least, the observed differences reported between ON and PEI in the current study demonstrates that even at comparable WTN levels, the community response to wind turbines is not necessarily uniform across Canada. Future studies designed to intentionally explore the factors that underscore such differences may be beneficial.