Wind turbine noise and the abuse of technocratic regulation: Part 2

In the first article of this series, I outlined the case for the prosecution against the current approach to regulating the impact of noise from wind turbines in the UK. In this article I will examine what might be considered pleas for understanding and mitigation. I have no intention of offering a case for the defence, since any independent observer can see clearly that current arrangements are both self-interested and indefensible on purely technical grounds.

The core of any argument in mitigation is that the issues which must be considered when regulating exposure to noise are complicated. There is no well-defined measure of “noise” as (a) different sources generated different distributions of frequencies, (b) the dispersion and attenuation of noise is greatly affected by local conditions, and (c) people (noise receptors) vary hugely in their sensitivity to different types of noise such as the swish or thump of turbine blades moving through the air as well as different ranges of noise frequencies.

While noise is a complex topic, this is by no means unusual when dealing with broad areas of environmental regulation. For many years I dealt with the regulation of air and water pollution from a wide range of sources including power and industrial plants as well as household heating systems, cars and trucks, and agriculture. I am not convinced that noise issues are more complicated than the issues involved in dealing with either air or water pollution.

What does seem clear is that noise consultants have a professional interest in emphasizing the complexity of their expertise. This trait is certainly not unique to noise specialists, but on other environmental issues there is a greater pool of people willing to challenge the methods and views of the priesthood. Thus, the development of regulatory rules for air and water pollution has been subject to much greater scrutiny and debate than in the case of noise.

The lack of widespread scrutiny has allowed wind farms to obtain what amount to individual exceptions for noise exposure that would not be accepted in other environmental areas. The new guidelines proposed by DESNZ reinforce that tendency by suggesting that noise limits should vary according to number of properties affected and the capacity of the wind farm. Just imagine how environmental groups would react if the same proposition were made for emissions of sulphur dioxide or nitrogen oxides from coal and gas power plants! The asymmetry is striking.

It is also easy to over-emphasise complexity. In practice, noise regulation is based almost entirely on the use of what are called A-weights on different frequencies to construct a single measure of noise. These weights reflect the average sensitivity of the human population to sounds at different frequencies. For many purposes, A-weighted noise measurements provide a reasonable starting point for regulating noise levels. In the case of wind turbines, there is increasing concern about emissions of low frequency noise – infrasound with frequencies of less than 20 Hz. Separately, the amplitude of wind turbine noise tends to be cyclical, leading to what is called Enhanced Amplitude Modulation (EAM) under certain weather conditions and over longer distances where low frequency sounds dominate.

Exposure and sensitivity to both infrasound and EAM tends to be greatest in flat landscapes at night, leading to complaints about disturbance to sleep patterns. Concerns about the health effects of long-term exposure to moderate levels of noise focus on night-time exposure. Hence, the most stringent noise limits are usually applied at night. In contrast, noise consultants for wind farms have consistently claimed that noise limits for night periods should be significantly higher than the noise limits for evening and weekend periods. This position was challenged in a report on EAM, which also proposed what was called an AM decibel penalty.

While both regulators and noise consultants now accept that both infrasound and EAM can be an important issue, they argue that they are essentially unpredictable and, thus, allowance for such effects cannot be incorporated in regular noise limits. This puts the burden of dealing with the problem on those affected. Arrangements for adjusting and enforcing changes to noise limits are slow and desperately inefficient due to the lack of resources at local authorities, which are responsible for enforcement of noise limits. Any reasonable system of noise regulation would incorporate tonal penalties in noise limits from the outset.

Unfortunately, different responses across the range of frequencies and the cyclical nature of turbine noise are only the start of the problems involved in assessing noise exposure. Start with a simple question: what is the accuracy of reported (A-weighted) noise levels? The answer is pretty good for moderate or high noise levels but potentially poor for quiet areas. Class 1 noise meters (the type required by current standards for noise measurement) have a tolerance of about +/- 1.5 dB for quiet background noise.[1] The meters used for environmental noise measurement are rarely rated for noise levels of less than 25 dB.[2] The reason is instrument noise. For example, true noise of 22 dB and instrument noise of 22 dB would yield a reported noise level of 25 dB. The best noise meters may not have such high levels of instrument noise outside their rated range, but the technical standards do not guarantee anything better.

Such issues matter because the primary British Standard that is used to assess the assessment of industrial noise in the UK – in its current version BS 4142:2019 – sets two thresholds for the impact of increased noise.[3] An increase in noise over the background or pre-development level of 5 dB is classed as an adverse impact, while an increase of 10 dB is classed as a significant adverse impact. The default position under standard environmental procedures is that significant adverse impacts should be addressed and mitigated. In earlier versions of BS 4142 it was stated that an increase in the noise level of 10 dB was likely to give rise to noise complaints.

Treating noise from wind turbines in a manner like noise from any other industrial plant, noise assessments need to do two things. First, they should establish the level of background noise at the receptor – i.e. house or group of houses. As a matter of convention, background noise is measured by calculating what is called the LA90 value over a time-period that may vary from 5 minutes to 1 hour. The LA90 is the A-weighted noise level that is exceeded for 90% of measurements in the time-period adopted. For reliable estimates of the LA90 the number of measurements should be at least 150.[4]

Second, the assessment must estimate the potential noise exposure due to the new source (wind farm or power plant) at the receptor. This estimate will be based on the cumulative noise emitted by the source and its attenuation due to a combination of distance, ground absorption or reflection, and barriers between the source and the receptor. There are prescribed models of carrying out such calculations.[5] These are based on experimental evidence, but the formulae must be calibrated using various assumptions. For example, whether the ground between a wind turbine and a house is treated as hard (reflective) or soft (absorptive) can make a difference between an estimated exposure of 37.8 dB or one of 33.8 dB over 1 km. Such differences can be crucial when assessing whether mitigation measures are required when background noise levels are low – e.g. 25-28 dB.

This procedure is complicated by the possibility that wind speed at or near ground level may affect measurements of background noise and, separately, the role of wind speed at hub height in determining the level of noise emissions from turbines. When wind speed may affect background noise, BS 4142 requires that measurement of wind speeds should be carried out at each receptor at the same time as the measurement of background noise.

That is too complicated and expensive for the wind industry. So, instead they have attempted to link wind speed at hub height at the wind farm site to wind speed at affected properties. On its own, that is a ridiculous proposition as UK wind farms tend to be located on high ground that may be 200 or 300 meters above neighbouring buildings, so wind conditions at the separate locations may be significantly different.

The central issue concerns what is technically called “wind shear”, a feature of wind flows by which wind speeds tend to increase with height above ground level. The variable “wind speed standardised to 10 meters” is calculated assuming that there is a fixed ratio between wind speed at 10 meters and wind speed at hub height (say, 100 meters) – both measured at the wind farm site – under all wind and noise conditions. Empirically that is wrong: detailed analysis of hourly meteorological data for the UK over 4 years shows that the ratio of wind speed at 10 meters to wind speed at 100 meters varies from 0.38 (high wind shear) to 0.85 (low wind shear) with a median of 0.60.

Standardised wind speed is nothing more than a proxy for wind speed at hub height. However, it is worthless proxy if the goal is to establish equivalent noise limits across wind farms with turbines of different hub heights. It strongly favours taller turbines, since standardised wind speed is lower for taller turbines at any given wind speed at hub height – i.e. for any given level of turbine noise.

A related factor is that wind shear tends to vary systematically over the day, being, on average, highest at midnight and lowest at midday. This interacts with the conditions that lead to serious sleep disturbance due to exposure to both infrasound and EAM. The method of standardising wind speeds overestimates background noise in evening periods and at night, when this matters most.

The errors and resulting biases from using this method of standardising wind speed are potentially large if the purpose of the analysis is to express background noise as a function of wind speed close to ground level at receptors, as is conventionally assumed. Noise assessments usually report background noise for integer values of standardised wind speed, i.e. the range from 4.5 to 5.5 meters per second (m/s) is reported as 5 m/s. However, had wind speeds at 10 meters been either measured or properly calculated that range could have extended from 2.6 to 6.6 m/s for a hub height of 80 meters or from 2.8 to 7.0 for a hub height of 120 meters. And that assumes a flat plain where wind turbines and receptors are located at the same elevation.

The convention of reporting assessments of background noise and turbine noise for wind speed standardised to 10 metres is confusing in the extreme. Turbine manufacturers report noise emissions measured according to an international standard (IEC 61400-11) using wind speed at hub height. This is logical because noise emissions are the product of blade rotation and mechanical noise in the generator that depend on wind speed at hub height. Translating estimates of noise based on physical measurements to a metric that has minimal connection with reality is a variant of the flummery that accompanies stage performance of “magic”.

Such details matter when regulation is presented as a technocratic exercise in analysing and interpreting complex evidence and then applying a set of rules based on that evidence. That is what I mean in referring to the abuse of technocratic regulation. While noise assessments refer repeatedly to prescribed technical methods, the core issues are matters of judgement and debate not technocratic prescription.

One of the recurrent features of noise assessments is that, even when there are standard guidelines, the noise assessment will propose the endorsement of noise limits which exceed those guidelines. In such cases, the noise assessments are being used as a way of justifying the reluctance of developers to reduce the number of turbines and/or adopt alternative turbine layouts that would reduce noise but might also reduce potential output.

To draw a parallel with environmental regulation for power plants. Would the public and environmental groups accept that it is reasonable to increase the permitted level of emissions of sulphur dioxide from a large diesel generator because that would allow the operator to run the plant using cheaper high-sulphur diesel? Usually the answer is no, except perhaps as an emergency when regular diesel supplies are not available.

At the heart of what I regard as the abuse of technocratic regulation is an economic choice. In effect, noise consultants, wind farm developers and the Government want to impose the consequences of over-developing wind farm sites on small groups of neighbours rather than accept higher costs for meeting targets for renewable generation. They dress up the choice in technobabble and present what are important trade-offs as if the issues were entirely matters for technocratic assessment This increases the widespread suspicion of “experts”, whose judgements are hijacked to support decisions that others may question.

In the final part of the series, I will discuss a better way to set and apply noise limits.

[1] This tolerance is estimated by combining the allowed tolerance for each frequency with standard A-weights over octave bands from 31 Hz to 8 kHz.

[2] Unless otherwise stated all decibel figures in this article refer to A-weighted composite measures or dB(A).

[3] For no good reason, BS 4142 excludes noise from wind turbines. That is the result of special interest lobbying. It is absurd that, for example, properties exposed to noise from wind farms should be treated differently from those exposed to noise from gas turbines. Increasingly, hybrid developments include wind turbines, solar panels and battery storage. Noise at a receptor cannot be separated according to source, so we finish up in the ridiculous position that both ETSU and BS 4142 should be applied in some way.

[4] This is an approximation based on a limit result that the percentiles of large samples taken from a continuous distribution tend to be normally distributed – see L. Bain & M. Engelhardt – Introduction to Probability and Mathematical Statistics (Duxbury Press, 2^nd Edition, 1992). Without being too much of a statistical nerd, acousticians rarely understand the statistical foundations for their measurements and the conclusions based on them. Using LA90 values estimated over short time-periods is not good statistical practice.

[5] Currently ISO 9613:2024.