Zajamsek et al, Investigation: Time Dependent Nature of Infrasound Measured Near Wind Farm

Investigation of the time dependent nature of infrasound measured near a wind farm

University of Adelaide, Australia


It is well-known that wind farm noise is dominated by low-frequency energy at large distances from the wind farm, where the high frequency noise has been more attenuated than low-frequency noise.

It has also been found that wind farm noise is highly variable with time due to the influence of atmospheric factors such as atmospheric turbulence, wake turbulence from upstream turbines and wind shear, as well as effects that can be attributed to blade rotation. Nevertheless, many standards that are used to determine wind farm compliance are based on overall A-weighted levels which have been averaged over a period of time.

Therefore the aim of the work described in this paper is to investigate the time dependent nature of unweighted wind farm noise and its perceptibility, with a focus on infrasound. Measurements were carried out during shutdown and operational conditions and results show that wind farm infrasound could be detectable by the human ear although not perceived as sound.


Wind turbine noise is influenced by atmospheric effects, which cause significant variations in the sound pressure level magnitude over time. In particular, factors causing amplitude variations include wind shear (1), directivity (2) and variations in the wind speed and direction. Wind shear, wind speed variations and yaw error (deviation of the turbine blade angle from optimum with respect to wind direction) cause changes in the blade loading and in the worst case, can lead to dynamic stall (3). With regards to propagation, the wind speed and direction between the source and receiver as well as wind shear and temperature inversions can vary significantly over time as well as location. Also, wind farm noise arriving at a receptor location several kilometers away can be heavily weighted to lower frequencies due to a combination of refraction, which causes sound waves to bend towards the ground, small atmospheric absorption at low frequencies and insignificant losses on reflection from the ground at low frequencies.

When evaluating the impact of wind turbine noise on residents living near a wind farm, it is important to consider the time variability of the noise for a number of reasons. The periodic variation in the amplitude of the sound, which is known as amplitude modulation, is perceived as more annoying according to listening tests conducted by Lee et al.(4). Moreover, compliance assessment procedures often overlook peak noise levels by averaging over large sample periods and ignoring the highest 90% of the measured signal. According to Bray and James (5), wind turbine noise is characterised by high crest factor, which means that wind turbine noise is highly time variant and thus more likely to perceived as annoying. As a direct consequence, wind-turbine infrasonic and low-frequency noise can be readily audible at much lower rms levels than has been acknowledged in the literature (6).

The perceived loudness of low frequency noise can increase significantly for a corresponding small increase in the acoustic energy, which is reflected in reduced spacing of equal loudness contours at lower frequencies (7). The implication of this observation is that low frequency sounds which are only slightly above the threshold of hearing can be perceived as loud (8). Since hearing thresholds can vary between individuals, it is possible that a sound that is inaudible to some people could be perceived as loud to others (9).

Thresholds of audibility provided in the ISO 389-7 (10) standard cover the frequency range from 20 Hz to 18,000 Hz but below 20 Hz, no such international standard has been developed. Nevertheless, a considerable number of research studies have focused on human perception of low frequency noise and infrasound and a comprehensive review of this literature was conducted by Møller and Pedersen (9). Their investigation led to the development of a normal threshold of audibility curve, which is based on existing data derived from listening tests. The listening tests involved exposure to sinusoidal tones in a free-field listening environment (9). It was observed that the resulting threshold of audibility curve follows a 12 dB/octave slope. Moorhouse et al. (8) also conducted listening tests on three listening groups, namely, low frequency noise sufferers, eldery people (55-70 years old) and people of a younger age. This gives testing a more general validity.

Listening tests were conducted using three sound samples, namely, real sounds (source unspecified), pure tones and beating tones. It was observed that low frequency noise sufferers are the least sensitive in absolute1 terms and the most sensitive for beating tones and real sounds relative to the absolute hearing threshold.

The main outcome of the study done by Moorhouse et al. (8) is the development of a criterion curve for the assessment of low frequency noise complaints. The curve covers the frequency range from 10 Hz to 160 Hz and is used by the Department for Environment, Food and Rural Affairs (DEFRA) in the United Kingdom for low frequency noise assessments. A somewhat different approach for assessing hearing threshold was used by Salt and Huller et al. (11) who focused on the response to noise of the outer hair cells (OHC). The cochlea, which is the inner part of the ear, consists of inner hair cells (IHC) and the above-mentioned OHC.

However, hearing threshold measurements, as mentioned above (references (8), (9)) commonly measure the response of IHC, which are more sensitive at higher frequencies, since their response is perceived as sound (11). On the other hand, the OHC are more sensitive at lower frequencies (at levels far below the hearing threshold). Salt and Huller et al. (11) outline a sound pressure level threshold at which the OHCs respond to airborne sound stimuli. It should be noted, that the understanding of how human ear responds to low frequency sounds is based on measurements performed on animals. The comparison between “Møller”, “Moorhouse” and OHC threshold curves can be seen in Figure 1, where a large differences in noise perceptibility thresholds can be observed. . As can be seen, the difference is up to 40 dB at 10 Hz between the “Møller” and OHC threshold. The “Moorhouse” threshold, on the other hand, lies ∼32 dB above the OHC threshold. The threshold curves obtained by Moorhouse et al. (8) for pure tones, beating tones and real sounds (source unspecified) are not shown since they only extend down to the 31.5 Hz one-third octave band center frequency.

 (Download the complete paper to see Figure 1 – Comparison of threshold curves.)

Some residents have reported annoyance when the wind farm is inaudible to them. They describe such symptoms as dizziness and nausea as well as unfamiliar sensations in their ears. According to Salt and Huller et al. (11), these symptoms may be related to infrasound, which stimulates the outer hair cells of the human ear at levels below the audibility threshold. This results in information transfer via pathways that do not involve conscious hearing, which may lead to sensations of fullness, pressure or tinnitus, but also may not lead to any sensation (11). The pressure fluctuations or cyclic variations in local barometric pressure caused by wind turbine noise have also been compared to similar pressure fluctuations that are experienced by an individual on a ship in high seas (12) as a result of the up and down motion of the ship changing the atmospheric pressure experienced by people’s ears. Dooley (12) proposed that this cyclic pressure variation may be the cause of motion sickness on ships as well as nausea in the vicinity of wind farms. 

This study investigates the contribution of a wind farm to measured levels of infrasound through consid- eration of shutdown and operational conditions with comparable wind conditions at outdoor microphones. The measured levels are compared to established audibility thresholds for infrasound outlined by Møller and Pedersen (9) and response curves for the outer hair cells within the ear which have been discussed by Salt and Huller (11). The time variant nature of the sound is also investigated via the crest factor.

Download the complete paper →