The methodology selected for this study focused on examining whether a scientific basis could be found for community concerns about noise from the wind farm, in particular, whether during the study period:
- Noise from the wind farm met established criteria, primarily the Wind Farms Environmental Noise Guidelines 2009 , or other appropriate criteria for infrasound and low frequency noise.
- Noise from the wind farm was of a sufficient level to be audible and whether any particular noise characters were present.
Noise and infrasound measurements
Regulatory documents typically require compliance measurements to be made within the audio frequency range, ie from 20Hz to 20kHz .
Additional requirements have been introduced by some authorities to address specific community concerns about low frequency and infrasound noise. In recognition of this, the current study incorporated measurements within ranges with lower limits.
For this study, two types of instruments were used, capable of recording sounds with lower frequency boundaries as described:
- Low frequency –12.5Hz (1/3 octave central frequency) for sites with deployed B&K Type 3639 noise monitoring stations;
- Infrasound – 0.25Hz (1/3 octave central frequency) for sites equipped with multi-channel Soundbook data acquisitions system.
Wind induced noise may have significant contribution to the reported values for outdoor measurements (refer to , , ). Therefore local wind speed was acquired simultaneously with the noise measurements.
Infrasound/low frequency content was measured and reported from microphones positioned approximately 1.2 m above the ground and equipped with multi layer wind shields. The wind shields have been tested and engaged in a previous comparative study on infrasound/low frequency noise (refer to the recent infrasound study  for details).
It is well understood that Infrasound measurements require specific techniques to minimise the wind noise influence; and other approaches were also considered. For example:
- Techniques utilising microphones placed on reflecting boards at ground level, as in IEC standard ), do not provide sufficient information on whether the +6dB correction for audible noise frequencies should also be applied to infrasound frequencies.
- Other studies have positioned microphones below ground level in fabric covered pits, however this is not always practical. For example the pit can be filled by water during rain periods and tends to attenuate higher frequencies of the infrasound range defined in ISO 7196 .
Wind farm operational data and shutdowns
The wind farm operator voluntarily provided data from sensors embedded into the wind turbine generator nacelle. The data are supplied for 10-minute intervals and include hub height, wind speed and wind direction, generated power, rotor rotations per minute (RPM) and other necessary parameters. They have been utilised for data analyses performed in accordance with regulatory documents and comparisons between noise levels during shutdown and operational conditions.
The wind farm operator advised that wind turbine generator on the site operated at the maximum electricity generation mode (noise mode ‘0’), which was confirmed later by operation data from the turbines.
The operator voluntarily agreed to arrange full operational shutdowns of all turbines on the site, under environmental conditions specified by the EPA; avoiding rain periods.
Shutdown periods are summarised in Table 1 and the time corresponds to Central Australian Standard Time (CAST). Shutdown periods mostly lasted for approximately up to 50 minutes of ‘core’ time. Transient periods right before and after the shutdowns were not taken into account for the comparative analysis of the shutdowns and similar operating periods.
Table 1 Periods of operational shutdowns of Waterloo Wind Farm
|Reference||Date||Start time||End time|
|Shutdown 1||1 May 2013||20:10||21:00|
|Shutdown 2||30 May 2013||19:10||20:00|
|Shutdown 3||5 June 2013||20:40||21:30|
|Shutdown 4||10 June 2013||05:10||06:00|
|Shutdown 5||12 June 2013||20:10||21:00|
|Shutdown 6||14 June 2013||20:00||20:50|
As noted previously, periods when wind speeds were less than threshold for operation of the turbines—that is, under calm conditions or very low wind speeds of less than 3.5 m/s—were not considered to be ‘shutdowns’ for this report.
Clearly, the turbines were not in energy generating mode under those conditions, so they were not useful for this analysis. The comparison needed to be undertaken under meteorological conditions that would normally allow operation of the turbines, so that the effect of the meteorological conditions on low frequency noise and infrasound could be established, in the absence of the wind turbine generators.
Residents in the Waterloo region were invited to contribute information on their experiences of noise from the wind farm throughout the monitoring study, recording their assessments in weekly noise diaries. Participation in the diary component was open to anyone who wished to volunteer, in addition to householders who had agreed to host equipment. The aim was to ensure that as wide a sample of perceptions was available for comparison with monitored information.
The EPA is strongly appreciative of the effort and time put into the diaries, given the busy lives of residents in a farming community such as Waterloo.
EPA supplied diary templates and postage paid envelopes, for volunteers to return directly to the project communications officer, on a weekly basis. Respondents were advised that EPA could not accept any diaries received after posting of periodic summary reports (monitored data and diary summaries) on the EPA website for the previous week. This precautionary approach was designed to eliminate the risk of potential contamination of records by information already published.
Respondents were requested to record when they perceived noise that they attributed to operation of the wind farm, and provide brief descriptions of what the heard, and any other factors, such as wind conditions, that they felt were important. They were asked to indicate time corresponding to beginning and end of the noise events, environmental conditions, their sensations and other relevant information.
Unattended noise monitoring was undertaken over a period of approximately two months at a total of 6 monitoring sites. Sites were chosen specifically to address the concerns raised by the residents, and attempted to make best use of EPA’s available suite of equipment by selecting:
- geographically, to obtain an appropriate spatial distribution of monitors around the wind farm
- structurally, aiming to select a mix of house construction types.
Each of the selected monitoring stations is detailed below, and Table 2 summarises their broad characteristics.
Two sites, designated as ‘Township’ and ‘North’ respectively, were selected for infrasound measurements in addition to monitoring of noise within the audio frequency range. These sites were selected for their proximity to the wind farm and severity of effects described by residents to be associated with the wind farm operation.
All stations had direct line of sight to the wind farm. Also, all sites were surrounded by trees and other vegetation. Where possible, the monitoring equipment was positioned to minimise influences of wind-induced noise from vegetation on the measured levels.
Table 2 Summary of acoustic measurements performed at the monitoring locations
|Type of monitoring||Sites||Frequency range||Weather monitoring|
|Audio noise plus infrasound
(indoors and outdoors)
|Township , North||0.25Hz to 20kHz||Yes
|Audio noise only
(indoors and outdoors)
|North East, South East, West||12.5Hz to 20kHz||Yes*|
|Outdoor noise logging only||East||12.5Hz to 20kHz||No|
*1 station monitoring with 6 weather parameters, and 2 stations monitoring with 2 weather parameters
Noise monitoring equipment at the locality included instruments indicated in Appendix B. Microphones for infrasound/low frequency measurements and standard audio frequency range measurements were deployed in the property which has a line of direct vision to the wind farm. The microphone was about 1.3 m above the ground. Microphones used for infrasound measurements were equipped with special multi-layer wind shields , and the GRAS microphone was fitted with the standard manufacturer wind shield. The outdoor equipment also comprised the weather monitoring station with a sensor positioned 2.2 m above the ground.
A microphone for infrasound/low frequency noise measurements was also positioned inside the house in a room facing the wind farm, on a tripod at approximately 1.3 m above the floor closer to the middle of the empty carpeted room measuring 4 x 4 x 3.6 m. The microphone was fitted with the manufacturer’s 90-mm wind shield to mitigate any possible influence of occasional air flows on the measurements (refer to Appendix B for details of the equipment).
The house is constructed of stone (estimated early 1900s) with small-medium single-pane wood-framed sash windows and corrugated iron roofing. The house was not permanently occupied during the monitoring period and was sporadically attended by the owners.
Monitoring equipment was similar to that installed at the Township site (Appendix B) and positioned in a similar manner. This site is closest to the wind farm, at a distance of about 1.3 km from the nearest wind turbine generator.
The house is of modern brick-veneer construction (estimated 1990s or later), with medium sized aluminium framed sash windows (panes divided into smaller lights) with corrugated iron roofing.
A set of outdoor equipment was placed in front of the garage on the gravel driveway. The microphones were mounted at the height approximately 1.2 m above the ground. A six-parameter weather monitoring station was mounted 2.2 m above the ground.
An indoor microphone was positioned at height of approximately 1.3 m in the front living room containing typical furnishings (sofas, carpeted floor, TV, etc.) directly under front window, in the corner of the room. The house has an open-plan layout, with several rooms directly off the living area. A window looks out onto a verandah. The wind farm was not clearly visible due to vegetation in the front garden. The residence was occupied during the monitoring period.
North East site
The house, situated at a significant distance from the wind farm, was estimated to be of 1900s construction, with small-medium windows and corrugated iron roofing. Two B&K Type 3639A noise monitoring stations were installed inside and outside the house (Appendix B). Vegetation was growing close to the house.
A local wind speed/direction sensor was placed on a pole at 4 m above the ground. The outdoor station was positioned at a distance from the house on the side facing the wind farm. The microphones were fitted with the manufacturer’s wind shields. The outdoor microphone was set at a height of about 1.5 m, while an indoor microphone was placed in the main hallway of the original part of the dwelling (5 x 1.5 x 3.5 m), adjacent to the former (now disused) front door of solid wood with panel inserts, flanked by sidelights. The hallway had bare floor-boards was covered with fabric. The residence was occupied during the monitoring period.
This house, of weatherboard construction with conventional large aluminium framed sliding windows and corrugated iron roofing, is located away from the wind farm by a significant distance. One side of the house (southeast corner) is enclosed by large gum trees and post-and-wire fencing. Information about instruments deployed at the site can be found in Appendix B.
The outdoor monitoring station was deployed to the side of the house with the local wind speed/direction sensor placed on top of the pole at 4 m above the ground. Height of the outdoor microphone is approximately 1.5 m.
The indoor microphone was positioned in the front bedroom of the dwelling facing the wind farm, dimensions approx. 3 x 3 x 3 m. Bedroom contains typical furnishings (bunk beds and single bed, built-in wardrobe, carpeted floor). A microphone was placed beneath the window at 1.3 m above floor level. The wind farm is visible through the window. The residence was occupied during the monitoring period.
South East site
This station was established at a house at a similar distance from the turbines as the North East and West sites. The instrumental deployment is the same as at the West and South East sites (Appendix B). A six-parameter weather sensor placed at 4 m above the ground was able to supply precipitation and other environmental information in addition to the local wind speed/direction. The height of the outdoor microphone was about 1.5 m.
The house, of double brick construction, was estimated to have been built in the 1960s to 1980s, with conventional large aluminium framed sliding windows and a tiled roof. The indoor microphone was placed in the front living room facing the wind farm, linked by cable to the main station located outside. This configuration optimised power supply and reliability of data stream via a GPRS connection. In this region, reliable connection to mobile phone services was an important consideration.
The room contained typical furnishings (sofas, floor carpeting, TV, etc). The microphone was positioned close to a corner of the room near the window, at approximately 1.3 m above the floor level. The residence was occupied during the monitoring period.
This is the most distant site from the wind farm, at almost 8 km from the nearest wind turbine generator (Appendix B). Only external monitoring was undertaken at this house, using a noise logger based on a B&K Type 2250 sound analyser positioned outside of the house to acquire the data. The microphone was set at about 1.5 m above the ground. No weather station was available for this locality during the monitoring period.
The house is of solid brick or stone construction with average to large windows and corrugated iron roofing. The residence was occupied during the monitoring period.
Data acquisition and reporting
Sound pressure is the major parameter characterising wind turbine generator as a noise source. Sound pressure levels from a turbine typically increase as the wind speed at hub hight increases. The noise level at a distant receiver is defined mainly by the wind speed and direction, so analysis of noise levels for regulatory purposes is normally made in conjunction with meteorological data.
Environmental noise levels experienced by distant receivers depend on many environmental and operational factors. In general health organisations and authorities recommend assessment of noise impacts using relatively long-term averages, varying from a few minutes to 12 months .
Noise from a wind farm is typically steady with possible amplitude modulation, if audible. Equivalent sound pressure levels are normally utilised to assess impact from such sources. There are no grounds for applying other descriptors such as maximum or peak levels. These are more appropriate in situations where abrupt changes of noise occur, such as on construction sites, shooting ranges, mining operations, or where gas guns are employed as bird scarers in agriculture.
Turbine data acquisition systems typically supply data at 10-minute intervals, and was the case for this project. Intervals of 10-minute are also recommended in many regulatory procedures for analysis of wind farm noise, so this interval was chosen as the basic interval for data analysis in this project. The time stamp in this report corresponds to Australian Central Standard Time at the beginning of the integration period.
Data with shorter averaging periods were also gathered at some sites and were utilised for more thorough analysis where required. For example ISO 7196  recommends integration time of 10 seconds for reporting G-weighting values. Ten-second data were acquired at the Township and North sites.
Wind induced noise is a potential problem for outdoor measurements [12, 35]. Acoustic data acquired outdoor at the local wind speed exceeding 5 m/s was disregarded for the current study. Periods of precipitation have also been identified from the local weather sensors and corresponding data was excluded from the data analysis.
Data also have been checked for potential presence of ambient noises, rain periods or noise from other sources, using analysis of acoustic descriptors and audio records where available. Typically, periods affected by such noises were not analysed.
Challenges of the data acquisition and data analysis
The original layout of the wind farm was designed to meet strict regulatory requirements. The noise monitoring equipment was deployed in a generally quiet rural area where the nearest monitoring site was about 1.3 km away from the nearest wind turbine generator. The low levels of noise characteristic of such areas and the potential of wind noise to overwhelm signals from the turbines under generating conditions, implies that noise from the wind farm may possibly be audible only under particular environmental conditions of low background or ambient noise levels at the houses.
Rain can be a particular interferent, and even where the rain sensor at a monitoring station does not show any rainfall, rain in adjacent areas may be heard in audio records, and may still dominate overall noise levels.
Agricultural activities may be performed at any or all times throughout the day, seven days per week during mid and late autumn. These activities need to be considered as potential contributors to the noise environment in the area.
Other noise sources may include flyovers of defence and civil aircraft, trucks and local traffic pass-bys.
Owners of the houses also utilise tools, pumps and agricultural equipment which are typical of farming activities in rural South Australia. Household activities in an occupied house can be a source of almost permanent interference for indoor noise measurements.
Audio and diary records
To ensure that the analysis was as objective as possible, acoustic descriptions recorded in diary returns were evaluated in conjunction with thorough analyses of audio records. To assist with the evaluations, audio records were amplified to facilitate identification of contributing noise sources and effects reported by the participants. The acoustic descriptors are reported as measured without amplification which may be required for process of the audio records replay.
Audio records were analysed to detect periods when wind farm noise may be audible and in conjunction with the respondents’ records. Particular attention was paid to the noise diary entries of respondents living and attending the monitoring sites and nearest neighbourhood houses.
The noise monitoring equipment at the Township and North sites were acquiring audio records continuously when the equipment was in operation (both inside and outside of the house). Audio recording at other sites were collected when the sound pressure level reached the trigger level. The trigger levels have been adjusted during the monitoring period to provide a reasonable balance between the source identification and number of the records.
The noise monitoring program included data acquisition both inside and outside of the houses (except the East site):
- Township and North sites
Multichannel sound analysers Soundbook Mk2 which are also Class 1 instruments in accordance with IEC 61672 and has 1/3 octave filters Class 0 in accordance with ISO 61260. Factory calibration charts provide data down to 0.1Hz. They show negligible deviation of instrument frequency response. Four Brüel & Kjær Type 4193 microphones have been used for simultaneous multi-channel measurements with the Soundbook data acquisition systems. The microphone calibration charts also show negligible deviation of the instrument frequency response but are only reported to a frequency of 1Hz. To expand the frequency range over which the microphones had uniform frequency response, they have each been used in combination with a Brüel & Kjær Type UC-0211 low frequency adaptor and Type 2669 (North site) or GRAS 26AK (Township site) pre-amplifier. The sensitivity of the microphone assemblies system at 0.2Hz was calculated in accordance with information obtained from the manufacturer  and found to deviate by less than 1dB for any of the microphone/preamplifier combinations. Therefore, measured levels using this measurement system could be considered across the entire frequency range required by ISO 7196 (0.25 to 315Hz).
- North East, West and South East sites
The stations included B&K Type 2250 sound analysers which are capable of reporting data at the extended low frequency range. The analyser meets requirements for Class 1 instrument in accordance with IEC 61672-1 and Class 0 for the band filters in accordance with IEC 61260 and other national standards. Data from the monitoring stations have been supplied via GPRS connection utilising B&K Sentinel system which limits the lowest reported frequency down to 12.5Hz of 1/3 octave central frequency.
- East site
Outdoor noise levels were measured with B&K noise logger based on Type 2250 sound analyser. The data have been acquired with extended low frequency range down to 6.3Hz of 1/3 octave central frequency.
Note that a side-effect of the extended low frequency range of the instruments used at Township and North Sites was an increase in the ‘self noise’ level (also known as the ‘noise floor’) at higher audio frequencies. Since typical noise levels at the monitoring sites were low, A-weighted and C-weighted levels from measurement channels engaged for infrasound/low frequency monitoring have not been reported, since their magnitudes can be compromised by this higher noise floor. For the same reason, spectral values for 1/3 octave central frequencies above 1kHz from the channels have not been included into this report.
Calibration information for acoustic instruments used is summarised in Appendix B. The equipment was periodically calibrated by B&K Type 4231 calibrator during the site visits. The results of calibrations were found to be consistent throughout the study. [k1] In addition to that, automatic Sentinel electric charge calibration was implemented on daily basis for the B&K Type 3639A stations.
Local wind speed/direction has been measured at all monitoring sites except the East site. Acoustic measurements at the North and South East sites were synchronised with data from Vaisala WXT520 sensor which also provides information about precipitation, temperature, pressure and humidity. A Davis Vantage Vue weather station deployed at the Township site also provided similar weather information in addition to local wind speed/direction. Local wind speed/direction data were acquired from Vaisala WMT50 sensors at the North East and West sites.
Conventional descriptors and criteria
Most compliance checking procedures are based on A-weighted levels as an approximation for human sensitivities to noise at relatively low levels. In accordance with the development approval conditions, wind farms in South Australia should meet requirements in the Wind Farms Environmental Noise Guidelines which define baseline noise criteria of 35dB(A) for receivers in Rural Living zones and 40dB(A) for receivers in other zones. Alternative criteria may be based on pre-construction background levels plus 5dB(A) . As noise impact have not been predicted at high levels before construction of the wind farm, pre-construction noise monitoring was not performed at the noise monitoring places. Default baseline criterion of 40dB(A) is applicable to the monitoring sites.
It is noted that regulatory documents do not require that noise from a wind farm should be inaudible; and further, the noise criteria should be met statistically, ie the approximating curve on the wind speed—sound pressure level chart should be below the applicable criteria (for the wind speeds between cut-in and speed of the rated power). Wind Farms Environmental Noise Guidelines indicate that compliance checking should be based on data collected under downwind conditions.
The Guidelines consider 5dB(A) penalty to be applied to the measured magnitudes if audible tone is present. Post-construction noise monitoring report (Marshall Day Acoustics 2011) contained tone assessment of wind turbine generator performed in accordance with IEC IEC61400-11:2006. No tones with audibility ΔLa, k above 0dB were detected. Another test report  also did not indicate the presence of audible tones in the wind turbine generator emission.
Tonal perception of sound was not a part of the respondents’ diary information. However the possible presence of tones from the wind farm and other sources (eg electric substations) was explored where analysis of 1/3 octave data or audio records showed a potential that tones may be audible.
Criteria and data analysis for infrasound and low frequencies
International Standard ISO 7196
International standard ISO 7196 specifies separate weighting (G-weighting) for measurements and reporting of infrasound levels. Previous studies  found close relationship between G-weighted noise levels and annoyance from infrasound, indicating it is an appropriate weighting to use to assess infrasound in the environment. At the moment, the ISO standard is the only widely used tool to assess infrasound. The standard itself does not clearly set acceptable criteria, but indicates that ‘weighted sound pressure levels which fall below about 90dB (meant G-weighted levels) will not normally be significant for human perception’. Multiple studies into perception threshold of infrasound indicate 85dB(A) as a conservative estimate (see report  for details) for the hearing threshold at infrasound frequencies. Typically 85dB(G) criterion is 5 to 10dB lower than the mean hearing threshold [2,15, 19, 24, 29, 39] and takes into account possible variations in individual hearing thresholds and any potential difference in the response to pure infrasonic tones and more broadband infrasonic noise.
Blade pass frequencies
Turbines may generate higher noise at the blade pass frequency, ie the frequency at which the rotating blades of a turbine passes the tower. During normal operating modes of the turbines, the blade pass frequency is expected to fall within 0.5–1Hz 1/3 octave bands. Respectively, the infrasound perception threshold can be utilised to explore whether the blade pass frequency component may be audible.
Low frequency noise
Another character that may potentially exacerbate perception of noise from a wind farm is excessive low frequency content. In spite of the fact that low frequency noise impact has been a matter of multiple research programs during last 50 years, it is difficult to indicate particular approach for assessment of low frequency noise which would have sufficient universality and worldwide applicability. Works  and  give a good overview of the low frequency and infrasound problem.
Some investigations in this area have resulted in additional sets of criteria specifically recommended for assessment of low frequency noise.
Danish EPA low frequency criteria
The Danish EPA low frequency noise criteria provide good correlation between objective and subjective assessments of low frequency noise when compared to other criteria applied in Europe [16,30]. This is a convenient way of assessing low frequency noise using a single number comparison. LpA,LF descriptor represents A-weighted level calculated for 1/3 octave frequencies between 10 and 160Hz . A limit of 20dB(A) is recommended for evening and night time. The limit should be met inside of the houses and calculated as average of the microphone indications measured at three points.
In this study LpA,LF is reported based on indication of one microphone. As for practices utilised for wind farms, the low frequency noise criteria have been included in the Danish Statutory Order on noise from wind turbines (revised 15 December 2011). An indoor low frequency noise limit for night time periods of 20dB(A) LpA,LF is applied during calculations of wind turbine noise, and only for wind speeds of 6 m/s and 8 m/s at 10 m above ground level under standard conditions.
During this study, the low frequency levels were also reported for outdoor measurements to verify that the low frequency impact is caused by an external noise source. The Danish EPA criterion is used in the study as a conservative measure to explore possible reasons for complaints in a quiet rural area. However, it is noted that the results of a South Australian comparative environmental noise study in 2012 by Evans et al  showed that this criterion could barely be achieved in windy areas, without wind farms or any other significant noise sources being present in nearby areas.
UK DEFRA low frequency criteria
The UK Department of Environment, Food and Rural Affairs (DEFRA) recommended a set of frequency dependent criteria for the assessment of internal low frequency noise from 10 to 160Hz . Low frequency criteria for non- steady noise at night time are represented in Table 2. It should be noted that limits in the table can be relaxed by 5dB for a steady noise and another 5dB increase can be applied for day time . As with the Danish EPA criterion, DEFRA 1/3 octave criteria (as presented in Table 3) were utilised conservatively in this study, to accommodate for any possible variations in human perception.
Some regulatory documents and noise guidelines recommend using the difference between C-weighted and A-weighted equivalent sound pressure levels to assist in deciding whether low frequency character is present in the measured noise.
Table 3 Proposed DEFRA reference curve
|Reference curve level in dB(Lin) at 1/3 octave band centre frequency (Hz)|
Other low frequency criteria
The Guidelines for the use of the Environment Protection (Noise) Policy 2007  suggest a difference of 15dB between the LCeq and LAeq levels may indicate a potential for a low frequency noise characteristic; however, other research indicate that the difference may need to be greater if it is to be used as an indicator of potential annoyance, although there is some controversy about this claim. Leventhall  suggests the difference should be greater than 20dB, while some authors indicate that the level may need to exceed 25dB when A-weighted noise levels are low (which is the expected situation in a quiet rural area). Given that this parameter may be useful in pointing to potential low frequency problems, it has been included in the analyses presented in this report.
Broner  has suggested a simple low frequency criterion, based on C-weighted measurements. Noise from a source which is operated 24/7 should meet a 60dB(C) criterion during night time. Apart from a desirable criterion of 60dB(C), he recommends a 65dB(C) (maximum) limit for the residential areas. However, this criterion is relatively new and does not have an extensive record of implementation.
Since previous monitoring programs could not clearly identify reasons for the residents’ feedback, additional spectral analysis was undertaken for periods when specific descriptors were recorded in noise diaries. Vasudevan and Gordon  have shown that an unbalanced spectrum with a particular roll-off rate at low frequencies (about 7–8dB/octave) may cause increased annoyance of particular group of listeners. This may well be the case in the Waterloo area.
At the moment there is not sufficient information available on the applicability of low frequency noise criteria for wind farms; and their correlation with subjective assessments from sufficiently large and diversified groups of listeners. The above criteria were considered sufficiently comprehensive and conservative to explore low frequency impact in a quiet rural area in this study.
Refer to discussion in recent reports on low frequency assessment in other references [11, 21].