Sow the Wind Turbine, Reap the Wind Turbine Noise – Watts Up With That?
It is not the purpose of this essay to bash wind energy projects, but rather to illuminate some of the controversy about wind turbine noise, and to point out the surprisingly primitive state of making wind turbine noise calculations for purposes of permit applications. In the past year I have been drawn slowly into a controversy over the siting of a large wind turbine project. Over the past two months I have spent more than one full day in hearings of one sort or another regarding the permitting of this project and actually gave about 75 minutes of testimony in one case not as a technical expert, but as a concerned citizen with a deeper than average understanding of particular concerns.
There are many wind farms currently operating in my state, so many in fact that Wyoming is near the top of states ranked by installed wind generating capacity per capita. This is the first one to my knowledge that has drawn significant public opposition. The reason for this opposition is that the wind farm is proposed to be built in an area of the Laramie Range mountains with a beautiful view of the Colorado Rocky Mountains to the south. It has also become a residential neighborhood of sorts with quite a large collection of ranches and ranchettes. The windfarm proposes to place an industrial facility, with towers extending to possibly 675 feet at blade tip, intermixed with residences and ranches. As more and more wind turbines are needed to meet renewable energy demands, the problem of industrial wind farms encroaching on residences, farms and ranches will only become more acute. One common problem, acoustics, is poorly understood by most of the public. Thus was born this essay.
The approval process
No renewable energy facility may operate in our state without its full complement of permits — one from local authority, a county commission generally, and then one from the industrial siting council (ISC) at least. As part of the review and approval process, the ISC requires submittal of an application that outlines the evaluation of potential project impacts and mitigation measures related to environmental, social, and economic resources. One section of the submittal is devoted to acoustics, and justifiably so as excessive noise is considered a health hazard in many workplace settings, and is regulated by any number of governmental agencies. Though there are different opinions about the health effects of wind turbine noise, low frequency noise, and infrasound, there is no reason to discount the health effects of nuisances that can disturb people’s rest.
Beyond health concerns, however, industrial wind turbines are considered a nuisance if placed too close to residences or noise sensitive facilities. They interfere with a person’s quiet enjoyment of their property. It is interesting how people who wish to avoid the nuisances presented by wind projects are now called NIMBYs by environmentalists. Such is our upside down world.
One ISC mandate is to mitigate such nuisances (Wyo. Stat. § 35-12-109). Rather than investigate this matter independently, however, the council will depend on input from other agencies such as a group within the Department of Environmental Quality, or perhaps county regulations, and the stamp of a professional engineer licensed to practice in the state on the permit application. A rather expensive and complex project like a wind farm cannot build a turbine and then decide if the noise it produces exceeds a local limit. No measurements are involved. Instead, the approach is one of modeling the propagation of noise from the turbines to local residences or property lines. Here is where trouble begins.
The noise scale
The range of natural noise energy is exceptionally broad, extending from the just barely detectable sound of a rustling leaf (at one picowatt per meter squared) to the startling sound of nearby thunderclap (one watt per meter squared). In order to handle such range, human hearing evolved to detect ratios of sound energy. Thus the ear responds somewhat logarithmically to sound level, and sound or noise level measurements use a logarithmic scale. The just barely detectable change of sound loudness is approximately one decibel (one tenth of a Bel) abbreviated as dB. It is a factor of 1.26 in the ratio of sound power level. The just barely audible sound is the reference point of zero dB. From this basis we derive a plethora of additional particular measures involving different schemes of weighting by frequency, or averaging over different time periods, to attempt to convey more and more information about sound in very specific instances. For example, the most common weighting scheme to describe impacts on human hearing is the “A-weighting” which is applied to frequencies over seven octaves centered from 63Hz to 8kHz, and which covers most of the range of human hearing.
Unfortunately a plethora of sound statistics combined with a logarithmic scale is beyond the ability of most people to comprehend reasonably. During my testimony one of the council members asked me why, if the hearing room sound level was 60 dB as measured by his smart phone, we should be concerned with a 3dB uncertainty. I explained that the scale is logarithmic and to add 3dB would make this room twice as loud. I don’t know if this helped explain the matter, but it is apparent that he was interpreting the decibel scale as linear, and on that basis 60 and 63 are nearly the same value.
Modeling acoustic propagation is a complex and technical subject. In order to do it from first principles would require a fully 3-D model of the atmosphere, land surface, and possibly the local geology. One would need realistic temperature and wind profiles, a model of the local terrain and estimates of the sound absorptive properties of soils and vegetation. Instead of taking this route consultants opt instead for a far simpler approach of using very idealized models of propagation combined with factors that model the attenuation of sound in vegetation, upon reflection from objects, through moist atmosphere, and so forth. Upon reading numerous environmental impact statements (EIS) regarding sound propagation, I realized these estimates are made commonly using a method described in an international standard, ISO9613-2. This propelled me to fetch the standard and read it.
ISO9613-2 is a quarter of a century old. Wind turbines have evolved from a typical 500kW unit at the time this was written to a more typical 3MW plus now. They have grown from a few tens of meters in height at the hub to more than a hundred meters. Rotor diameters are now pushing 220m.
Moreover ISO9613-2 Is not meant specifically for wind turbines, but rather is a general approach to any sort of noise source. In fact, it doesn’t specify its expected accuracy for noise sources and receivers higher than 30m (mean height) and more distant than 1000m. The standard spells out explicitly its other limitations. For instance, it contemplates sound diverging spherically from point sources declining at 6dB per doubling of distance. The attenuation with distance can be much less than 6dB per doubling of distance in many situations. To the extent it is applicable to extended (non-point) sources of noise, these can be broken down into a collection of point sources so long as the collection of points representing the extended source have the same intensity and height above ground, and the same conditions of propagation from source point to receiver. This would seem to preclude its use in noise emanating from the turbine blades.
Under meteorological conditions the method is limited to propagation within 45 degrees of downwind direction, wind speed between 1 and 5m/s between 3 and 11 m height. It presumes a well developed, moderate, ground based inversion, such as develops on clear, calm nights. It fails to define what a moderate inversion means. To more fully understand this term I searched and found on Cliff Mass’s blog that a moderate inversion is one between 10 and 15C temperature gain per 1000 feet of elevation. My calculations indicate an inversion of 10C would bring a sound ray propagating horizontally at its source at a height of 100m to the ground surface in less than 2km. Strong inversions might do the same in a distance less than 1km. What this means is that a strong inversion maps the acoustic output of the entire lower hemisphere of what ISO9613-2 imagines a wind turbine to be (an isotropic radiator) into a small surface patch. It concentrates sound energy near surface in a way similar to how inversions trap pollutants.
Of particular interest is Table 5 in ISO9613-2 which lists the uncertainties of the method. These are generally ±3dB (SPL), but the table lists nothing concerning uncertainties for high sources, nor for distant field points. Footnotes indicate further limitations. For instance, as Note 24 indicates, variation in sound level at a certain site and on a certain day can be expected to be considerably larger than Table 5. Unfortunately the term “considerably” is not quantified.
Finally the standard does not contemplate many other potential effects from wind turbines: Amplitude modulation, tonal noises, ground vibrations, low frequency noise, infrasound, reflections, and the actual directivity of wind turbine noise. Meteorological conditions may make the divergence cylindrical with attenuation of only 3dB per doubling of distance, rather than 6dB per doubling, as propagation becomes confined to near surface ducts. Ground reflections from a low source, which occur at near grazing angles serve to quiet the direct waves through destructive interference; while, reflection at larger angles from a high noise source are coherent and interfere both destructively and constructively. A list of the deficiencies of ISO9613-2 for purposes of modelling sound propagation from modern wind turbines seems pretty long.
Beyond these many hypothetical concerns, though, actual field measurements have shown that the method can produce substantial underestimates of wind turbine noise of 0-5dB at close distances to 10dB beyond 10 km. 
“Why use a standard which appears to apply and perform so poorly?” one may ask. As Keith, et al, have pointed out
“It is not currently feasible to use more sophisticated methods than ISO (1996) as those methods require data that are usually not available: a sound speed profile, or wind speed and temperature as a function of height.” 
Thus, we seem stuck with this customary method and so we should contemplate how to ameliorate its deficiencies.
Margins and Factors of Safety
What we are faced with is a method of estimation that does not pertain to some circumstances at all, and pertains to others with a large and unquantified degree of uncertainty. This may be an unusual situation, but not completely unlike problems that engineers face commonly. What a sensible engineer would do in such a situation is to construct an adequate margin for uncertainty, and add this margin to estimates of noise. Indeed, factors of safety and margins are a core concept in engineering design and system operations. The question is, “How large should this margin be?”
Considering factors not even addressed by ISO9613-2, things like amplitude modulation, infrasound, ground vibrations and reflections, all we can say is these must be addressed eventually by appropriate research. There is very little to go on at present and are not a focus of this essay. The wind farm that has been the focus of my concerns is being built on crystalline bedrock which may make ground vibrations a particular problem. Homes are practically defenseless against vibrations entering through their foundation, which then interact with resonances in the structure. Because this project now has its permits, time will tell if this is a problem or not.
On the other hand, for noise as we commonly think of it, measured by the A-scale of weighting, we can calculate an uncertainty budget. Consultants for the wind turbine entities do two things in this regard that I view as improper. First, they add a 2dB margin that is nothing more than a single standard deviation of noise pertaining to manufacturing and installation variations of the turbine/generators themselves. This is a very minimal margin. Obviously the departure of field measurements from calculated estimates shows that there are additional sources of uncertainty that arise from meteorology, terrain, wind turbine height, and from the rotor not being well-approximated as a point source. What we need are standard deviations of as many independent sources of noise variation as we can find, and combine them in a manner consistent with the Guide to Estimation of Uncertainty in Measurements (GUM).
Second, in a number of acoustics reports I have read the wind energy consultants claim that by not including provisions for attenuation of sound they have produced “conservative” estimates of noise. Unfortunately there is no way of knowing if one can trade unknown biases of this sort for unknown uncertainties of the random kind, and to rely on it is not a good engineering practice.
If we assume the 2dB standard deviation of noise in wind turbine manufacture and installation, the 3dB standard deviation of uncertainty mentioned in Table 5 of ISO9613-2 (the warning in Note 24 notwithstanding), and one more standard deviation of 1dB for distance uncertainty between the source of noise on the wind turbine and the receiver location, and a coverage factor of k=2 , then our uncertainty budget looks like that in Table 1, nearby.
Table 1. Uncertainty Budget
|Source of Uncertainty||Sound Pressure Uncertainty relative to zero dB||Variance||Variance and coverage factor (k2=4)|
|Wind Turbine||(2dB): 10(2/20)-1 = 0.259||0.067||0.268|
|Propagation factors||(3dB): 10(3/20)-1 = 0.413||0.171||0.684|
|Distance||(1dB): 10(2/20)-1 = 0.122||0.015||0.06|
|Totals||1.006 (square root of total variance)||1.012|
Since the uncertainty should be expressed in dB which we can add directly to the modeled noise value, we convert back to dB using 20 Log (1.006+1) = 6.1dB. The complication here is that the uncertainty of the propagation factors is not independent of the turbine-contributed uncertainty — the turbine noise and distance uncertainties are independent issues, but a more noisy turbine will contribute to noisier contributions from terrain and meteorology as these are merely a channel conveying the noise from source to receiver. Moreover turbines become more noisy with rising wind speed. However, we are at least working our way toward a defensible and meaningful value of design margin.
What is the pertinent regulated noise limit?
More potential conflict arises from imprecise regulation language. From our county regulations regarding noise for wind energy conversion (WEC) projects we see
“Noise associated with WECS operation shall not exceed fifty-five (55) dBA as measured at any point along the common property lines between a non-participating property and a participating property.”
Note that the regulation makes reference only to a weighted scheme (A weighting for example) but with no explanation about any averaging over a measurement period of any kind. We might assume it is a very short period measurement — an instantaneous value practically. Otherwise we would expect some notation more specific such as Leq, the equivalent continuous sound level, or Lnight, which is the A-weighted sound pressure level averaged over the 8-hour long night duration. The wind turbine operators do not want to adhere to a near instantaneous limit as this becomes a more stringent standard to meet when a facility is prone to occasional loud bursts in an otherwise long, quiet averaging period. Obviously regulator would do better to consult someone with relevant knowledge of acoustics to help them compose more useful regulations.
As wind farms proliferate they are bound to invade residential areas and noise sensitive facilities with increasing frequency, making arguments over nuisance and health concerns increasingly acute and acrimonious. Yet, the methods used to estimate noise levels are surprisingly primitive and don’t appear at times to follow good engineering practice. The public including agency regulators are generally baffled by the technology surrounding noise specifications and measurements. Hopefully this essay helps illuminate some of the important issues.
-See for example Simon Carlile, John L. Davy, David Hillman, and Kym Burgemeister, A Review of the Possible Perceptual and Physiological Effects of Wind Turbine Noise, Trends in Hearing Volume 22: 1–10 2018. Or, Jianghong Liu, Lea Ghastine, Phoebe Um, Elizabeth Rovit, Tina Wu, Environmental exposures and sleep outcomes: A review of evidence, potential mechanisms, and implications, Environ Res. 196:110406. May, 2021; doi: 10.1016/j.envres.2020.110406.
-Sound Power Level (PWL) is defined as dB=10LogW+120; where W is the sound intensity in watts/m2. The reference level of 0dB occurs equates to one picowatt per meter squared. An alternative equivalent definition uses Sound Pressure Level (SPL) and is dB=20LogP+94 with P measured in Pascals. Reference level of 0dB occurs at 20 micropascals.
-Other weighting schemes are C, Z, and G which may provide important utility in various contexts. For example Z is an unweighted scale useful for peak loudness measurements; G is useful for characterizing low frequency sound.
-Keith et al, J. Acoust. Soc. Am. 144 (2), 2018;: Keith et al, J. Acoust. Soc. Am.139, 1436 (2016)
-From Cliff Mass’s blog site I found that moderate inversion are those with temperature at 1000 feet height 10-15 C warmer than the ground surface. Strong inversions, beyond 20C occur on clear nights with snow covered ground. This definition applies strictly to the entrapment of pollutants, which was the topic of Mass’s blog and not to sound propagation. However, the calculations of ray curvature suggest such inversions are also not a bad guide to concentration of sound energy near the ground per the sorts of distances involved in ISO9613-2.
-For information on calculating the curvature of acoustic rays see William W. Seto, Acoustics, Schaum’s Outline Series, Chapter 8, 1970.
-Recent Advances in Wind Turbine Noise Research, Colin Hansen, and Kristy Hansen,
Acoustics 2020, 2, 171–206; doi:10.3390/acoustics2010013; Keith et al, J. Acoust. Soc. Am. 139 (3), March 2016.
-Keith et al, J. Acoust. Soc. Am. 139 (3), March 2016.
-GUM: Guide to the evaluation of uncertainty in measurements: https://www.bipm.org/en/committees/jc/jcgm/publications
-The location uncertainty largely results from what part of a large turbine is currently contributing noise and how it is oriented with respect to a receiver. Regarding coverage factor, if a coverage factor of 99% (k=2.57) is more appropriate than a coverage factor of 2 (95%) then the corresponding margin rises to 7.2dB, and so on. This coverage factor seems to be a point of confusion in some acoustics reports. The acoustic modelers appear to think of “k” as the margin itself, whereas in standard metrological terminology “k” is a coverage factor. That a typical k-value is 2 and a typical manufacturer reported noise uncertainty is 2dB seems an unfortunate coincidence.