Wind turbine noise and the abuse of technocratic regulation: Part 3

A warning. I am aware that some readers find some of my articles hard going. I apologise, but that is particularly likely to be true of this article. Managing industrial pollution – including noise - is an unavoidably technical issue. When trying to change technocratic procedures, it is necessary go beyond the “it is nonsense” reaction. As I have explained, I consider that some of the current procedures for regulating turbine noise are indeed nonsense. But there are a lot of people who have their self-image and careers based on their detailed familiarity with existing rules. They may realise privately that some of the rules are hard to justify, but it is difficult to admit that and especially in the context of formal proceedings. Hence, it is very hard to change rules without demonstrating in some detail that there is a sensible and feasible alternative.

In the final article of this series, I will outline how noise limits for wind turbines could be set, monitored and enforced. The key goal is to illustrate how environmental regulations that are consistent with the principles that are applied to other industrial plants, including other power plants, can be applied to wind farm noise.

I will start with an example of how water pollution is regulated, focusing particularly on organic wastes, because it is important to emphasise that turbine noise is not a special case. Organic material discharged to rivers or lakes decay due to natural processes but the process reduces the level of dissolved oxygen, i.e. reduces the absorptive capacity of the river or lake. Water with low levels of oxygen is discoloured, has a foul smell, is generally unpleasant to be close to, and cannot support fish and other aquatic species. The parallel to noise from wind turbines is instructive, if we remember that noise pollution disperses more widely but attenuates more rapidly with distance.

Regulation of water pollution relies on two elements. First, there are limits on the maximum concentration of organic materials – such as food waste - in liquid discharges. These are measured by Biochemical (or Chemical) Oxygen Demand (BOD/COD) – the amount of dissolved oxygen required to break down the organic material per unit volume of the discharge. Second, there are water quality standards that measure the average level of dissolved oxygen in the receiving river.[1]

A new industrial plant may cause the level of dissolved oxygen in the receiving river to fall below the minimum acceptable level, even if it complies with the emission limit on the concentration of BOD in discharges. This would happen if the capacity of the river to absorb and break down BOD discharges is low. Periods of low water flow in summer are often critical. Exceeding the absorptive capacity of the receiving water body leads to episodes in which many dead or dying fish are observed. The analogy is with quiet rural areas where the background noise is very low. In such cases, turbine noise is far more obtrusive than in locations next to a busy road, especially when people have chosen to live in remote and quiet locations. These are the key sensitive receptors just as low summer flows are sensitive periods for water pollution.

When a new industrial plant whose liquid discharges exceed minimum level is proposed, it is standard to carry out a pollution assessment focusing on each of the two elements outlined above.[2] The expected flow rate and concentration of BOD discharges will be considered along with options for reducing both total discharges and the BOD concentration.

The absorptive capacity of the receiving river is assessed by measuring dissolved oxygen and BOD levels at various points – upstream of the proposed discharge, a short distance (100-400 meters) downstream, and further (2-4 km) downstream. The measurements must be taken under different flow conditions and allow for natural variations in dissolved oxygens such as heavy rainfall and flooding as well as seasonal changes – e.g. autumn leaf fall which raises the amount of organic material in the river.

Based on this information the absorption of BOD in the river can be modelled using well-validated models of chemical processes based on the Streeter-Phelps or oxygen sag equation. In this way, a safe level of BOD discharges can be established to ensure that the level of dissolved oxygen does not fall below a minimum value even under low flow or high BOD conditions.[3]

Limits on the total volume and concentration of BOD can be set to meet this goal. The permitted discharges and the quality of river water can be monitored either by analysing regular samples or by real-time monitors. With modern equipment using wireless Internet of Things (IoT) networks it is feasible to implement real-time feedback arrangements so that discharges are reduced or stopped if the level of dissolved oxygen in the river begins to approach the minimum permitted level.

Setting and applying noise limits for wind farms involves a similar process of assessing how much noise from the development will exceed existing background noise and then setting emission limits designed to ensure that the impact of noise at receptors is not excessive. Up to now, the usual option has been to require that turbines operate in what are called low noise modes, usually reducing output as well as noise emissions. However, these modes only have a limited effect in reducing total noise emissions and none on amplitude modulation effects. A better option is to switch off one or more turbines – those closest to receptors most affected by noise. It is routine for wind farms to collect real-time SCADA data including the wind speed at – or close to – hub height and the operating status of each turbine in a wind farm, so real-time control is feasible.

In some jurisdictions environmental regulators require medium and large pollution sources to make the data collected by continuous emission monitors available for public access with either no delay or a very limited delay (such as 24 hours) being permitted. This should be standard practice for all onshore wind farms with a capacity of more than [5] MW subject to noise limits. It is not difficult nor need it be costly. Many trading platforms do this for market prices, so that there is plenty of software that can be modified to establish and operate websites with such information.

The first question with respect to establishing noise limits at receptors is whether they should vary with wind speed. As explained in my previous articles, the use of wind speed standardised to 10 metres at the wind farm site is a complete nonsense. The logic of BS 4142 is that noise limits dependent on wind speed can only be allowed if wind speed is measured – and later monitored - at the receptor site. This implies that, if wind farms argue for background noise levels that vary with wind speed, they must install, maintain and monitor wind measurement equipment at each receptor – both when estimating background noise and during the operation of the wind farm.

As in the case of water quality, levels of background noise vary both randomly and due to systematic influences such as atmospheric conditions. Observations affected by nearby turbines, rain, traffic noise, and similar factors should be excluded. In the case of nearby turbines, the best way of removing their influence is to estimate background noise using measurements for periods when wind speed at hub height is less than 3 or 4 m/s, as the cut-in speed for wind turbines usually falls in this range.

The current method of estimating background noise levels (using the LA90 noise measure) is to use least squares regression to fit a polynomial over a range of wind speeds from 1 or 2 m/s to 11 or12 m/s – setting aside the standardisation issue. This is poor statistical practice which is prone to generate biased estimates. In addition, there is no reason to believe background noise varies as some simple function of wind speed. Even on its terms there are better methods – e.g. fractional polynomials. The more fundamental problem is that least squares is a method of fitting average values, which tend to be skewed upwards by occasional extreme values. That is why noise assessments often exclude high noise readings, especially at low wind speeds.

A better starting point would be to estimate the median (50^th percentile) value for ranges of wind speed near-ground level at the receptor. There are more sophisticated statistical methods available which treat noise measurements as the sum of two, three, or more distinct components with distinctive distributions. One component would be background noise. Other components might include time-dependent traffic or work-related noise.

There is no reason to estimate background noise values separately for day and night periods. The current ETSU guidelines allow much higher noise limits at night than during the day. This make little sense if the major health concern about noise exposure is sleep disturbance, especially when caused by relatively low frequency noise. In any case, other forms of environmental protection make no distinction between levels of day and night exposure, so there is no reason to treat noise differently. To be clear, this does not justify increasing noise limits during quiet day-time periods but implies reducing night-time noise limits to the same levels.

Setting and applying noise limits based on estimated values of background noise is a little more complicated. The current procedure is to set a noise limit equal to background noise + 5 dB(A), but the process of applying it is unclear and cumbersome. While rarely spelled out, the implicit assumption is that this limit applies to the average value of background plus turbine noise over a period of several weeks. This, in turn, relies on the assumption that the level and direction of turbine noise are distributed randomly over time and are independent of each other as well as of background.

Such assumptions are patently wrong. The persistence of weather systems means that wind speed and wind direction are correlated over time. This is a major factor in prompting noise complaints. Those living near to wind farms may experience extended periods with high levels of turbine noise under stable weather conditions with moderate wind speeds. From their point of view, the fact that such conditions only prevail for, say, 20% of hours in the year is irrelevant. What they experience is a high, even intolerable, level of noise for extended periods.

This is no different from other areas of environmental regulation. In the case of noise exposure, the fact that a night club is only noisy on 2 days in a week doesn’t alter the fact that it is causing a noise nuisance. Similarly, the occurrence of low levels of dissolved oxygen in a river causing fish kills may be an infrequent consequence of low river flows and other factors, but we do not ignore the possibility of such events in setting discharge limits for water pollution.

The conditions under which complaints about turbine noise are most likely will be when (a) wind speeds are high enough for turbines to operate, but (b) near-ground wind speeds are not so high that the local wind-related component of background noise is noticeable, and (c) random components of background noise are low.

Suppose a turbine noise limit of median background noise + 5 dB(A) is adopted for a specific range of near-ground level wind speeds. Based on my analyses of background noise distributions at a sample of receptor sites, the noise limit will be more 10 dB(A) above current noise levels for between 20% and 30% of evening and night periods at quiet locations. An increase in noise of 10 dB(A) has been set by BS 4142 as the threshold for a significant environment impact that is likely to prompt noise complaints. The implication is setting noise limits any higher than the baseline of median background noise + 5 dB(A) does not meet the requirement of minimising environmental impacts.

To simplify implementation by wind farms, noise limits could be re-expressed in terms of ranges of wind speed at hub height. In either case, the principle should be that LA90 noise levels at receptor sites should not exceed the relevant noise limits for more than [20%] of 10-minute time periods in any 6-hour period. This allows for occasional but not extended exceedances.

Equally important, there must be immediate and painful penalties imposed on wind farms that violate noise limits. Currently, justice delayed is justice denied. So, for example, for each 6-hour period in which noise levels breach the condition, the wind farm could be required to pay a penalty to the occupier of the receptor property within 7 days. To provide a strong incentive, the penalty should be equal to the full capacity of the wind farm for 6 hours multiplied by the official Intermittent Market Reference Price - IMRP published by the Low Carbon Contracts Company as a reference for CfD contracts. Again, to be clear the rule implies strict liability for noise exceedances if the wind farm is operating. The only defence is that all of the turbines were shut down.

Such a penalty will ensure that it is never worth a wind farm deciding to breach noise conditions to increase revenue rather than switching off a small number of turbines. Failure to pay penalties promptly should lead to the amount payable being increased by, say, 25% of the original sum for each 7-day delay. The point is that severe penalties will encourage a wind farm to develop the SCADA and management software necessary to monitor noise levels and ensure that noise conditions are met.

A word on enhanced amplitude modulation (EAM). There are noise conditions for wind farms which include a noise penalty – i.e. a reduction in noise limits – when amplitude modulation (AM) exceeds “normal” levels. These are usually considered to be up to 3 dB peak to trough variation for 3 frequency bands: 50-200 Hz, 100-400 Hz, and 200-800 Hz with peak to trough variation for 100-400 Hz treated as the reference level. With appropriate monitoring and software, AM can easily be incorporated in the implementation of noise limits discussed above.

Standing back from the details, my central point is that it is time to stop the special treatment of turbine noise and to apply the standard framework for managing other types of industrial pollution. As part of this, the use of continuous emission and exposure monitoring as a basis for real-time controls of emissions has been standard in some countries for more than 30 years. We recommended such an approach in drawing up pollution guidelines for power plants in the late 1990s. Advances in networking and monitoring equipment since then mean that implementing real-time controls to comply with noise limits is conceptually straightforward for both new and existing wind farms.

As a final point, one element in reforming the management of noise from wind farms must be to transfer responsibility for industrial sources of noise from local authorities to environmental agencies responsible for dealing with other forms of industrial pollution. Local authorities have neither the staff nor the resources to deal with a major issue of this nature. Most environmental health officials are trained to handle hygiene at restaurants and noise from night clubs. They are not equipped to handle any kind of pollution from major industrial sites, which include wind farms.

The records of, for example, the Environment Agency or the Scottish Environmental Protection Agency are far from good, but at least they have the personnel and technical resources to implement a consistent approach across their jurisdiction. They would have to be funded to carry out the necessary work. Money can be raised by requiring wind farms to pay a licence fee per MW of capacity for each turbine.

There will, of course, be loud complaints about jeopardising Net Zero from the usual suspects. This is purely self-interested. The cost of complying with a reasonable framework for regulating noise exposure from onshore wind farms will be modest – significantly less than 5% of total costs, which is in line with the costs of managing other forms of industrial pollution. So be it if a few onshore wind farms are not built, since it is clear to everyone that the bulk of wind generation in future will come from offshore wind farms.

It is outrageous that DESNZ should even consider proposals which will impose heavy costs on those who have the bad luck to be located close to new onshore wind farms. Normally, retrospective changes in regulatory rules are undesirable. However, I would sympathise with any proposal from a future government to impose licensing conditions which follow the approach outlined above for all onshore wind farms or, at least, those commissioned from 2026 onwards.

[1] I have assumed that the discharges of water pollution are made to a river, which is the most common case. In other cases, discharges may be made to lakes or, in coastal areas, the sea. More complicated situations may require special analysis – for example, when a receiving river flows into a lake that is already affected by discharges from other sources.

[2] This description covers what is regarded as good practice in managing pollution from new medium or large plants. Many episodes of water pollution are the result of accidental or deliberate spills from small sources – often agriculture or linked to agriculture – that discharge to streams or small rivers with very limited absorptive capacity. This is the equivalent of individual small wind turbines, for which no systematic noise assessment has been carried out, that are located too close to neighbouring properties. In such cases, planners or environmental regulators should impose default measures such as (generous) minimum separation distances or wetland management of discharges.

[3] The minimum level of dissolved oxygen will depend on the fish species that we want to protect. Salmon and related (salmonid) species require much higher levels of dissolved oxygen than coarse fish such as carp.