Hansen, Zajamsek, Hansen, Noise Monitoring, Waterloo Wind Farm

Noise Monitoring in the Vicinity of the Waterloo Wind Farm

Kristy Hansen, Branko Zajamsek and Colin Hansen, School of Mechanical Engineering
University of Adelaide May 26, 2014

This report by the above authors describes the results of their concurrent full spectrum acoustic monitoring conducted at a number of homes located between 2 km out to nearly 10km from the Waterloo Wind Development. This monitoring was independent of the South Australian Environment Protection Authority (SA EPA) and was requested by Mrs Mary Morris and other concerned residents in the Waterloo district. The monitoring occurred during the period of the South Australian EPA Acoustic Survey, conducted in mid 2013.

The results in this independent survey as well as the conclusions are in marked contrast to the results and conclusions of the SA EPA Acoustic Survey report, and reinforce the Waubra Foundation’s opinion expressed at the time the initial SA EPA report was released that there were serious problems with the methodology used by the SA EPA in its acoustic survey at Waterloo. This report provides further evidence that the current SA EPA Wind Farm Noise Guidelines do not protect the health and sleep of the neighbours to these wind developments, out to nearly 10km from the closest wind turbine, because they do not regulate the acoustic emissions to protect health, and most importantly, the sleep of the neighbours.

Emeritus Professor Colin Hansen has advised that he sent the report to the EPA, requesting their comment. To date, three months later (19th August, 2014) no comment or feedback has been received by the Adelaide University researchers from the SA EPA responsible public officials.

Extract from the Conclusions:

“Therefore, the results show that there is a low frequency noise problem associated with the Waterloo wind farm. Therefore, it is extremely important that further investigation is carried out at this wind farm in order to determine the source of the low frequency noise and to develop mitigation technologies. In addition, further research is necessary to establish the long‐term effects of low frequency noise and infrasound on the residents at Waterloo. This research should include health monitoring and sleep studies with simultaneous noise and vibration measurements.”

Key Extracts from the report are reproduced below, and the report is downloadable from the link beneath.

1 Introduction

This report details independent noise measurements and their analysis taken in the vicinity of the Waterloo Wind Farm during the period 9/4 – 22/6, which is the same period as the study undertaken by the EPA and reported in EPA (2013). Measurements were taken outside of as well as inside a number of residences. Due to the potential for data contamination by background noise during the day, only data measured between midnight and 5am are reported here, as during those hours, the dominant noise source was generally the wind farm. The following sections of this report detail the measurement equipment, measurement procedures, data analysis and data interpretation, followed by a conclusion summarising the results detailed in the rest of the report.

The data analysis and interpretation comprises four sections:

• overall levels averaged over 10‐minutes for all night‐time data collected at each residence;
• unweighted third‐octave spectra and overall levels for the shutdown periods;
• unweighted third‐octave spectra and narrowband spectra for measurement times corresponding to noise diary entries; and
• unweighted and A‐weighted third‐octave spectra for measurements which exceeded 40 dB(A).

2 Measurement Details

Three B&K 4955 microphones were used for the indoor measurements. These microphones have a low noise floor of 6.5 dB(A) and a flat frequency response down to 6 Hz. While these microphones do not have a flat frequency response below 6 Hz, they are still capable of measuring the blade‐pass frequency and harmonics (Hansen, 2013). The microphones were connected toLAN‐XI hardware and continuous 10‐minute recordings were made using Pulse software. The average sound pressure level of the three microphones was calculated in accordance with the Danish guidelines for indoor low‐ frequency noise measurements (Jakobsen, 2001). This average includes one microphone positioned in the room corner. In this position, the maximum sound pressure level would be measured since this is an anti‐node for all room response modes. A singleGRAS 40AZ / SV 17 microphone was connected to a SVAN 979 sound level meter. This microphone was used as a back‐up and check for the indoor measurements made with the Pulse system.

The outdoor measurements were made using GRAS 40AZ / SV 12L microphones connected to aSVAN 958 sound level meter, which measured continuously over 10‐minute intervals. The microphones have a noise floor of 17 dB(A) and a flat frequency response down to 0.8 Hz. Hemispherical secondary windshields were used to minimise wind‐induced noise experienced by the outdoor microphones, and they were designed to be consistent with the IEC 61400‐11 standard, which specifies the use of these secondary windshields for measurements close to a wind turbine. A spherical secondary windshield and box windshield with specifications described in Hansen (2013) were also used for comparison but these results are only presented in the narrowband analysis in Section 6.2. Wind speed and direction were measured at heights of 1.5 m and 10 m using Davis Vantage Vue and Vantage Pro weather stations, respectively. The weather measurements were collected in 5‐minute intervals and then the 10‐minute average was calculated during post‐ processing.

3 Guidelines

It is well known that wind farm noise is dominated by low‐frequency energy (Moller & Pedersen, 2011), particularly at large distances from the wind farm, where the high‐frequency noise has been more attenuated than the low‐frequency noise. As such, a number of different weighting functions have been applied to the data in Section 4 to highlight different characteristics of the noise. A detailed description of these weightings and their applications is given in the report by the EPA(2013). This section provides a brief analysis of the limitations of some of these weighting functions in the context of wind turbine noise. Additionally, some drawbacks of the current SA EPA guidelines (EPA, 2009) are discussed and recommendations for improvements are suggested.

In South Australia, compliance of a wind farm is determined based on the applicable outdoor limit specified in the SA EPA guidelines (2009). Most of the measurement locations detailed in this report correspond to “rural industrial” zones where the allowable limit is 40 dB(A). One of the township locations is situated in an area which has been zoned “township” according to the Clare and Gilbert Valleys Council regulations. For lack of additional information, it is assumed that this translates to “rural living” in the context of the SA EPA guidelines (2009), which has a corresponding outdoor limit of 35 dB(A). According to the EPA guidelines (2009), a compliance analysis requires collection of over 2,000 data points, with 500 data points recorded for the worst‐case wind direction. For the measurements outlined in this report, such a large amount of data were not collected at any one location, however it was still considered valuable to plot a regression curve to illustrate the degree of compliance over short periods as well. In any case, the use of night‐time data is expected to reduce the degree to which data are contaminated by extraneous sources, thus giving a reasonable estimate of the degree to which the wind farm is compliant.

The SA EPA guidelines also specify use of the LA90 metric, which is the A‐weighted noise level that is exceeded 90% of the time. It should be noted that wind farm noise can be significantly underestimated using LA90 levels due to the unsteady nature of the noise. Hence, the LAeq, which is the energy average of the noise, is considered to be a more realistic representation of the actual noise level attributed to the wind farm, particularly between midnight and 5am when there are very few other noise sources of a similar level to the wind farm noise.

Despite the fact that low‐frequency noise has been identified as a potential issue associated with wind turbine operation, the SA EPA guidelines (2009) do not provide guidance for acceptable levels of low‐frequency noise and infrasound, even though there are several recommendations available in the literature. For example, the C‐weighting can be used to provide an indicator of the presence of low‐frequency noise. According to Broner (2010), a night‐time limit of 60 dB© is recommended, and this limit was included in the NSW draft guidelines (2011). Low‐frequency noise can also be identified by finding the difference between the overall C‐weighted and A‐weighted levels. When LCeq – LAeq > 20, a potential low‐frequency noise problem is indicated, and Broner and Leventhall (1983) and DIN 45680 (1997) would recommend further investigation into the time‐dependent low‐ frequency noise characteristics including noise fluctuations, spectral balance and amplitude modulation.

The G‐weighting is used to indicate the level of infrasound. According to ISO 7196 (1995) and DIN45680 (1997), the audible threshold for the overall G‐weighted noise level is 85 dB(G). On the other hand, this does not preclude the possibility that lower levels of infrasound will have an effect on people (Salt & Lichtenhan, 2014).

The SA EPA guidelines (2009) suggest that the indoor A‐weighted noise level should not exceed 30 dB(A). According to the World Health Organisation night‐time guidelines (WHO, 2009), the no observed effect limit for outdoor noise is 30 dB(A). To quantify the low‐frequency contribution to the indoor noise, it is useful to refer to the Danish guidelines for indoor low‐frequency noise (DanishEPA, 1997) and the UK Department of Food and Rural Affairs criteria (DEFRA, 2005). The Danish limit considers A‐weighted levels in the frequencies from 10 Hz to 160 Hz and the limit is the calculated average of the sound pressure level measured at three different locations in a room. According to the Danish guidelines, the indoor noise level, LpA,lf in the frequency range from 10 Hz to 160 Hz should not exceed 20 dB(A). The DEFRA criteria are frequency dependent and also span the frequency range from 10 Hz – 160 Hz. The allowable limits for each third‐octave frequency bin in this range are specified in the relevant report (DEFRA, 2005). The specified limits can be relaxed for steady noise and for daytime measurements but the measurements in this report did not fall into either of these categories. It is well‐known that wind farm noise is an unsteady noise source due to sudden changes in wind speed/direction, inflow turbulence, wind shear (van den Berg, 2005) and directivity (Oerlemans & Shepers, 2009). It has also been found that wind farm noise is modulated at the blade‐pass frequency (Hansen et al., 2013), which causes a periodic variation in the loudness of the sound.

It is worth noting that wind farm compliance according to the SA EPA guidelines is based on a regression line fitted to 2000 or more data points plotted on a graph of noise level (dBA) (y‐axis) vs hub height wind speed (x‐axis). Each data point is a 10‐minute average, which means that the influence on people of a noise source that is highly variable in nature will be underestimated. In addition, many 10‐minute average data points are above the acceptable 35 or 40 dB(A) requirement and as compliance is based on the regression line only, these times of relatively high noise level are ignored. In other words, compliance with the EPA guidelines does not mean that noise levels will never exceed the recommended limits – in fact, they can exceed the recommended limits many times as can be seen by the graphs shown in this report. Furthermore, the 10‐minute average values are lower than the peak values, which means that the wind farm could generate high levels of intermittent noise and still be compliant.

It is also important to recognise that thresholds of audibility are not dependent on the 10‐minute average of the root mean square (rms) value of the noise signal alone. This type of analysis ignores any difference in character between the measured noise and the noise used in the laboratory to determine threshold levels. The main differences in character that are important include the presence of multiple harmonics of the blade passage frequency and the crest factor of the noise. The crest factor is the ratio of the peak noise level to the average (or rms) noise level. The measured average noise levels for wind farm noise have been shown to contain peaks that are up to 20 dB above the reported average level. Even for “compliant” wind farms, such peaks are well above the levels required to disturb sleep (according the 2009 WHO document, “Night Noise Guidelines for Europe”). It is also worth noting that traffic noise, on which the WHO document on night noise levels is based, is not characterised by such high crest factors and thus has less potential for disturbing sleep. Nevertheless, this is an area of future work for our research group and the purpose of this report is to provide an analysis similar to that carried out in the EPA study so that comparison can be made between the two sets of results.

4 Overall Noise Levels

The following section presents data that were measured at the same residences as the EPA study (EPA, 2013), as well as three additional residences. The North East residence is not included in the analysis as we were unable to measure inside at this location. A number of weighting functions have been applied to the data and where applicable, a linear regression curve has been included. The DEFRA criteria and the Danish guidelines for indoor low‐frequency noise have only been applied to the indoor data, as they are not considered relevant for outdoor noise.

The figures presented in this section show data plotted against the wind speed at a height of 10 m in the left hand column and data plotted against the wind speed at hub height in the right hand column. Data points shown in red correspond to times when the residence was downwind from the proposed wind farm, according to the definition that downwind is ±45from the direction of the residence relative to the wind farm. Data points shown in green indicate times where the wind speed at a height of 1.5 m was greater than 5 m/s. For such wind speeds, noise measurements can be contaminated by wind‐induced noise. The sources of wind‐induced noise are pseudo‐noise and acoustic noise. Pseudo‐noise is caused by turbulent pressure fluctuations and vortex shedding incident on the microphone which lead to false indications of the sound pressure level whereas wind‐induced acoustic noise arises when objects such as tree branches and leaves are put in motion by the wind. Pseudo‐noise is only relevant for outdoor measurements but wind‐induced acoustic noise is relevant to both indoor and outdoor measurements. Both indoor and outdoor measurements taken during periods of rain have been discarded from the analysis.

In this section, all plotted data corresponds to night‐time measurements made between 12 am and 5 am. During the night, people are trying to sleep and this time also represents the greatest contrast between ambient noise and wind turbine noise, due to the absence of other sources such as traffic and farming machinery. These times were also selected to minimise contamination from noise sources other than the wind farm.

8 Conclusions

Based on the findings in this report, the following conclusions can be drawn:

• For the 50 Hz third‐octave band, the sound pressure level difference between shutdown and operational conditions can be higher than 25 dB for both outdoor and indoor measurements.
• The noise level in the 50 Hz third‐octave band is often above the audibility threshold (ISO389‐7, 2005) when the wind farm is operating.
• The peak in the 50 Hz third‐octave band would be classified as a tone according to some standards (NZS 6808:2010, 2010; ANSI S12.9 ‐ Part 4, 2005).
• The allowable limits should be reduced by 5 dB(A) to account for such tonal noise.
• The outdoor and indoor noise levels measured during the shutdown cases were consistently lower than those measured when the wind farm was operating.
• The most significant differences between shutdown and operational conditions can be observed when the residence is downwind from the nearest wind turbine and the hub height wind speed is greater than 8 m/s.
• The shutdown periods should have occurred during 12 am – 5 am when the contribution from extraneous sources would be minimised and the contribution from the wind farm more able to be quantified.
• For all shutdowns reported here, the closest wind turbine to the residence did not reach its rated speed of 15 m/s. In most cases, the wind speed at hub height was significantly lower than rated speed for the shutdown and adjacent times.
• The peak in the 50 Hz third‐octave band is a consistent feature of the noise diary results and is often above the audibility threshold (ISO 389‐7, 2005).
• A narrow‐band analysis with frequency resolution of 0.1 Hz reveals distinct peaks at the blade‐pass frequency and harmonics for many of the results corresponding to noise diary entries.
• The narrow‐band analysis also shows the existence of tones, which occur at 23 Hz, 28 Hz, 46 Hz, 56 Hz and 69 Hz.
• These tones have several sidebands which are spaced at the blade‐pass frequency and allude to the occurrence of amplitude modulation.
• There is a good correlation between low frequency noise events and complaints registered in noise diaries.
• At many of the residences, there were many occasions during the hours of 12 am and 5 am where the outdoor noise level exceeded the SA EPA (EPA, 2009) criteria of 40 dB(A).
• The indoor limit for wind turbine hosts of 30 dB(A) recommended by the SA EPA (EPA, 2009) was exceeded on many occasions between 12 am and 5 am. This is also the no observed health effect limit for outdoor noise according to the WHO (2009).
• The range in the overall A‐weighted levels was noticeably large indoors and could be as low as 5 dB(A) and as high as 38 dB(A). The lower value highlights that the night‐time noise levels in this rural environment are sometimes so low that even low levels of wind turbine noise would be noticeable. It is plausible that the upper value is related to the presence of wind turbine noise.
• It has been shown that there can be a large variation in the results obtained by considering the LAeq as opposed to the LA90, between the hours of 12 am and 5 am.
• Since the number of extraneous noise sources is expected to be low during these night‐time hours and wind turbine noise can be highly variable with time, it does not seem justified to only consider noise levels which were exceeded 90 % of the time.
• The C‐weighted level was often higher for downwind conditions and hub height wind speeds greater than 8 m/s. However, consideration of the overall level with respect to recommended limits did not prove useful in identifying any low frequency noise issues.
• The LCeq ‐ LAeq criteria was often exceeded and there was a large scatter in the data.
• The overall G‐weighted level of 85 dB(G) was never exceeded however this does not preclude the possibility that infrasound was not detectable.
• The Danish low frequency noise guidelines were exceeded on a number of occasions. In general, the exceedences occurred for downwind conditions and hub height wind speeds greater than 8 m/s.
• The DEFRA criteria were exceeded on multiple occasions, usually corresponding to downwind conditions and hub height wind speeds greater than 8 m/s.

Therefore, the results show that there is a low frequency noise problem associated with the Waterloo wind farm. Therefore, it is extremely important that further investigation is carried out at this wind farm in order to determine the source of the low frequency noise and to develop mitigation technologies. In addition, further research is necessary to establish the long‐term effects of low frequency noise and infrasound on the residents at Waterloo. This research should include health monitoring and sleep studies with simultaneous noise and vibration measurements.

Download the complete report of the acoustic survey by Hansen, Zajamsek and Hansen →

For access to the SA EPA Waterloo Acoustic Survey 2013, and documents expressing concerns which have been raised about that acoustic survey, please seehttp://waubrafoundation.org.au/resources/waterloo-wind-farm-environmental-noise-study-sa-epa/

Careful reading of the UN Convention Against Torture makes it very plain that public officials who know of the damage to human beings and allow it to continue could be facing criminal charges (sleep deprivation is acknowledged as torture by a number of bodies including the Committee Against Torture http://waubrafoundation.org.au/resources/un-convention-against-torture/ and also the Physicians for Human Rights — see pp 22 — 26 of their report called “Leave no Marks” http://physiciansforhumanrights.org/library/reports/leave-no-marks-report-2007.html

The Waubra Foundation advised the Clean Energy Regulator (CER) Board members over a year ago about the damage to human health from proximity to wind turbines — and their response was that they preferred the explanation that it was a Nocebo effect:http://waubrafoundation.org.au/resources/letter-notice-clean-energy-regulator-5-april-2013/

Consequently in 2013 the Waubra Foundation advised both South Australian Premier, Jay Weatherill and the Chair of the CER, Ms Chloe Munro of the Foundation’s concerns about the Waterloo Acoustic Survey — neither of whom responded to the letter:http://waubrafoundation.org.au/resources/open-letter-premier-south-australia-clean-energy-regulator-concerning-sa-epa-acoustic-survey-2/