New Zealand, like many countries around the world, has been encouraging the installation of wind energy projects. Unfortunately, also like many countries, the residents who end up living next to these projects have been complaining about the noise, after having been assured that it would be minimal. NZ has noise regulations that predate wind projects, but like many countries they have developed special noise regulations uniquely customized for the wind industry - something called NZS 6808. The previous version of this Standard (1998) didn’t protect the neighbors – there were hundreds of complains while the developer could plausibly claim to be following the standards. So they formed a committee and came up with 6808:2010. One person on that committee, Philip Dickinson, refused to agree to the new standard and went public about his reasons.
Originally I intended to write a typical posting about the nature of the disagreements, and which side had the strongest evidence for their position. Not to worry, I’ll still cover that further below, in eye-glazing detail. But in his writings Dickinson departs from the usual technical issues to strike closer to the heart of the problem with industry standards in general, and wind industry standards in particular. In his words (from Pragmatic View):
This initially was intended as a discussion paper on the sound propagated from wind turbines and what seems to be accepted practice in national and international standards. We tend to forget that someone [in this case, the developer] has to bear the cost of such standards, and that almost always this will be someone with a vested interest in the subject – someone or some organisation that can afford the cost and to whom the way the standard is written has a direct bearing on what they want it to achieve for them. Those that review the standard, in its public comment stage, often do not see the hidden implications – particularly if mathematics is involved in any way.
Industries will always do whatever they legally can to increase their profits, and controlling the regulatory environment is just part of that. They more than anyone else have the money, experience and incentive to help the government draw up regulations that look effective at protecting the public but don’t get in the way of their business. In practice these resulting regulations have subtleties that are not immediately apparent to the general public (and often not to the politicians as well) that allow them to proceed with their business unhindered by rules.
It is Dickinson’s view that the new version of 6808 is exactly that – a set of rules that on its surface appears to be protective, but is not. The new standards haven’t been in force long enough to have a history in NZ, but they are largely based upon standards from other countries. This was in fact used as a selling point for them. But how have those other standards worked in their countries? Aside from the large number (now in the multiple thousands) of significant complaints from around the world, people (now in the hundreds) have essentially abandoned their homes. Any standard that claims that it protects the neighbors amenity and then results in abandonments is a failure, no matter its appearance or how much the industry tries to spin it.
Before I get started on the merits of Dickinson’s concerns, a little history. In 2008 Standards New Zealand put together a committee, consisting of 11 experts, to come up with a new standard. That committee produced a series of drafts throughout 2009 and finally a consensus of the committee was reached later in 2009. Except not quite. Dickinson, the delegate from Massey University, refused to go along with the new standard. The committee “…could not reconcile the arguments he advanced against the Standard…“ They then went back and reviewed the evidence and concluded the standard was in their opinion valid and it was published in March 2010. At least the committee’s Chairman, Stephen Chiles, had the grace to publicly state that there was a dissenting vote.
Dickinson has not been quiet about his disagreements. He stated his major concerns late in 2009 in a submission to the committee and a submission to Acoustics 2009 entitled Nonsense on Stilts. Later in March 2010 he again presented his concerns to the Turitea Board of Inquiry.
Stephen Chiles works for a large multinational engineering consulting company, URS. In proving Dickinson’s earlier point about regulatory capture, URS is not a disinterested party. Wind developers are keen to hire acousticians who can provide cover for their projects, and Chiles is now right on up there. Chiles took the time to author a rebuttal to Dickinson’s criticisms and Dickinson basically reissued his Nonsense paper, now retitled a more gentile A pragmatic view of a wind turbine noise standard.
As for the standard itself, I couldn’t find the final version online, except to buy it at considerable cost from SNZ. The best I could do was a draft version from September 2009.
Enough history, now about the merits. Unfortunately, as Dickinson mentions, the problems are subtle enough that most people won’t bother digging into them to find out where their ineffectiveness originates. I’m going to try to present them quickly, hopefully suitable for a general audience, but the subtleties are just that. We will use Dickinson’s presentation to the Turitea Board, as it is succinctly written for a more general audience and conveniently Chiles uses that presentation as the basis for his rebuttal. Dickinson states that the 6808 methodology makes six underlying assumptions about the noise from wind farms and these six are all incorrect, and in a direction that uniformly favors the wind developers. These six are:
1) The background sound level at any residence is directly proportional to the wind speed at the hub of the nearest wind turbine.
Dickinson wraps 4 different problems under this one assumption.
(a) The methodology the standard uses to test for violations is both unworkable and invalid, almost certainly intentionally so. It calls for taking 10-minute samples over a 10-day period (1440 samples) and then correlating them with the wind speed at the hub of the nearest turbine. The linear regression line created by the procedure would then be compared with an equivalent line generated when the turbines were not operational, presumably created before the project went into operation. Many jurisdictions will not have the capability to install and operate the necessarily complex equipment. It will necessarily be unattended, so it won’t be certain what is being measured. To add to Dickinson’s concerns I’d add that obviously this procedure requires a fair amount of cooperation from the project operator. First, there’s the use of the wind speed data produced by the operator. Worse, what happens when there are no pre-project measurements, which would be the case in most situations? The operator shuts down the turbines for 10 days so a new series can be run to establish the no-turbine background levels? Don’t forget the operator has a large financial interest in the results, and expecting them to not protect those interests is dreaming.
In his rebuttal Chiles does not address these concerns.
(b) Each of the 10-minute samples are reduced to one number to represent the noise during that period. The standard calls for using the La90 value, which captures just the softest 10% of the noise, throwing away the loudest 90%. La90 is typically used to represent background noise, and the “logic” used by the standard is that since turbine noise is broadband and constant this would be a good way to eliminate “transient” noises. The first problem is that wind turbine noise is anything but broadband and constant, as the charts that Dickinson presents demonstrate. The second problem is that you take the bottom 10% of a logarithmic function (the individual dBa measurements that make up the La90 number) and then use a linear regression that produces an arithmetic average of that. By the time you’ve finished all these calculations you’re quite far removed from representing what the neighbors are hearing, and that distance favors the developer.
Chiles retorts that “The benefit of analysis using the hub height wind speed is the direct relationship with wind farm sound.“ Except that measurements indicate that this is not so. The sound varies all over the place depending on a large number of variables – for example in a stable atmosphere the wind could be blowing very hard at hub height while the calmness at the surface produces no noise at all.
(c) A large portion of the noise level meters in use do not measure below 30 dBa. Thus in rural areas, where the background is often in the 20′s, these values get raised to the floor of the equipment, thus artificially raising the background level.
Chiles responds that adequate equipment is readily available. The actual verbiage from the standard is: “In some cases a sound level meter with a low noise floor may be necessary so that sound levels at low wind speeds can be accurately measured.“ Left out is how the necessity is determined. Recall that the acousticians doing the measurements are hired by the developer.
(d) There’s no restriction on how much wind strikes the microphone/windscreen itself during the 10-days, and it is well-known (as shown by Dickinson’s charts) that just the wind causes the noise readings to increase, thus artificially raising the background level as well as masking the noise from the turbines.
Chiles responds that taking lots of measurements when the turbines are both productive and non-productive should allow you to cancel out the effects of wind on the microphone itself. To me, it is likely that the noise from the turbines will also disappear during this cancellation. Only if the turbine noise were a direct function of wind speed (and nothing else) would this be a valid argument.
2)The 90 centile level (LA90 ) is the most suitable measure to use as wind turbine sound is continuous and will be captured whereas transient events such as passing road traffic, aircraft etc., the sounds from which would last only a small percentage of the time, will be ignored.
The use of La90 might be valid if in fact wind turbine noise mimicked natural background noises. It doesn’t, as witnessed by thousands of complaining neighbors, and measured by people like Rand and Ambrose.
Chiles responds that L90 is valid because “wind farm sound can only be measured in the presence of wind.” Perhaps that is true if you mean wind at the hub, but too often that has little to do with wind at the home. As a practical matter, only a human observer can differentiate between wind noise and turbine noise, and having a human observer is what Dickinson suggests.
3)Any sound outside will be attenuated by 15 dB through an open window.
The 6808 standard is predicated on keeping the average noise level in a bedroom below 30 dB. To get to that point, the assumption is made that homes provide a 15-dB reduction. Dickinson presents an unsourced chart with measurements that indicate that is only true with the windows pretty much completely closed, while most rural residents open them widely to get the best ventilation during the summer.
Chiles responds that Dickinson has no measurements to buttress his point and references what is probably be Bellhouse(2000), which is unpublished and which I could not find. Chiles ignores the ability of low frequency noise to penetrate home construction almost without attenuation and measurements that show the noise inside can be louder than the noise outside. Dickinson mentions this while Chiles ignores it completely.
4)The greater of the background sound + 5 dB, or 40 dB will be a good design level and will meet the World Health Organization recommendations of 30 dB in a bedroom to avoid sleep disturbance.
Dickinson points out that 6808 calls for 40 dB or background +5, so there’s no assurance that a bedroom will stay at 30 dB. One could argue that the extra 5 dB won’t be noticed very much, but the way in which the background level is set in the first place is very problematical, with the actual increase likely being much larger than the method would indicate.
Chiles doesn’t mention any of this in his rebuttal.
5) Wind turbine sound can be considered as from a point source (at the hub of the wind turbine) and ISO 9613 is a good prediction methodology. Only properties inside the predicted 35 dB contour need be considered.
The picture on the cover of the 8608:2010 Facts Sheet puts the lie to the point source assumption. Rand among others has measured the noise and it most commonly decreases in a cylindrical fashion, indicative of a line source. I explain point and line sources in an earlier posting.
Chiles opines that modern simulations can handle multiple point sources so why should they use a line source? The quick answer is because that’s how wind turbines have been shown to behave. Even one turbine behaves like a line source (Rand, page 36).
6)Wind turbines emit little or no low frequency sound, so an LA90 of 40 dB is quite sufficient to protect from low frequency sound.
I have to wonder if Chiles is being intentionally ignorant. Modern wind turbines emit a great deal of LFN, as shown by Rogers, Rand etc etc. He presents a reference, a study by the Danish Energy Authority, that “Wind farm sound like most other sounds does include low frequencies, but these are within appropriate limits.” And how were these limits set? “Even close to the wind turbines the sound pressure level is much below the hearing threshold. Thus infrasound is not considered a problem.” Apparently the Danes have never read Salt.