The lie behind turbine noise models

The answer is simple: Herkimer County residents were lied to.

Yes, we could use softer words to explain the situation. But given what sound experts already know about turbine noise, the time for niceties has past.

Predicted turbine noise at Hardscrabble

Prior to erecting a wind facility, project owners usually engage acoustic engineers to prepare models that predict sound level increases a community can expect from an operating project at certain reference points. These engineers rely on the CADNA/A[1] software tool for their models. CADNA/A is based on ISO 9613-2, the international standard developed for sound prediction.

The CADNA/A tool generates predicted sound levels at various distances from the turbines. Developers present the sound levels as contour lines overlaid around the turbine sites. Each contour shows a sound level in decibels with the lines closest to the turbines having higher decibel levels.

The sound predictions developed for Hardscrabble showed that during periods of low wind conditions, non-participating residents closest to the turbines could expect to experience noise increases of less than 6 dBA over the presumed existing level of 35 dBA. During high wind conditions, modeled data showed property owners would experience slightly higher levels but most increases would still be under 6 dBA[2]. 

Prior to construction, Iberdrola insisted the facility would meet the New York state noise guidelines for most situations and would be in full compliance with local regulations that limited noise to 50 dBA.

CADNA/A and the ISO 9613-2 standard

Acousticians hired by the wind industry insist the ISO standard is an appropriate method for modeling wind turbine sound provided the correct input parameters are used. But what they do not admit is that the ISO 9613-2 standard, on which CADNA/A is based, was never validated for wind turbine noise. In fact, the standard is mainly applicable to situations concerning road or rail traffic, industrial noise sources, construction activities, and many ground-based noise sources. It does not apply to sound from aircraft in flight, to blast waves from mining, military, or other similar operations. And it was not designed to predict turbine noise.

The ISO Standard limits use of its methods to noise sources that are close to the ground (approximately 30 meter difference between the source and receiver height) and within 1 kilometer of the receiving location. A wind turbine with a hub height of 80+ meters exceeds the ISO height limit by 50 meters. Meteorological conditions are also limited to wind speeds of approximately 1 meter/second and 5 meters/second when measured at a height of 3 meters to 11 meters above the ground.

Only when all of these constraints are met by the situation being modeled can the predicted noise levels be assumed to be accurate within a +/- 3 dB range.

The constraints placed on the ISO standard having to do with wind speed, direction and weather conditions indicate just how limited the models are for anything other than simple weather conditions -- NOT the types of conditions that wind turbines need to operate.

The way sound spreads outdoors can be affected by temperature differences in different layers of the wind that cause sound waves to bend up or down at the boundaries just like water bends light. If a noise source is above a boundary then sound that would have gone down to the ground surface might bend up and dissipate. If the noise source is below a boundary layer then sound that might have dissipated upwards is bent down and added to the sounds that would normally be directed downwards. The current science of meteorology does not have precise ways to know what is happening right near any particular turbine.

Heinrich A. Metzen of DataKustik GmbH[3], maker of CADNA/A confirmed this fact in an e-mail where he stated:

"long range propagation including atmospheric refraction is not part of the standards used for (normal, "standard") noise calculations. It is known that atmospheric refraction may cause sound to be refracted downwards again and contributing strongly to the level at long distances. The atmosphere in the standards existing is just homogeneous above height."

Since there are no accepted algorithms to predict these refractions, sound propagation models cannot evaluate conditions that have vertical or horizontal turbulence even though we know they can add significant sound at the receiving location when present. As a result, predicted sound levels are understated.

Countries in the European Union are developing their own models for predicting turbine noise propagation because of their concerns with limitations of the ISO standard. Unlike the ISO 9613-2 standard, these newer models have been validated for turbine noise by peer-reviewed independent studies.

Iberdrola knows better

The first post-construction sound study in Herkimer revealed noise levels reaching 60 to 65 decibels, nearly 20 decibels above what was predicted for homes in the area. Iberdrola's Paul Copleman told the press the excessive noise levels were largely due to the wind rustling leaves and cannot be "attributable to the wind farm."

Seriously? Any guesses on the number of complaints filed over noisy leaves before the turbines were sited?

Use of a model that understates real-world operational sound levels is very likely the root cause of the problem at the Hardscrabble facility.

Acoustic experts who work for the wind industry, including Iberdrola, are well aware of the limitations of the ISO modeling. They are well aware that the standard is intended for ground-based sound sources and has never been validated for predicting wind turbine noise. They also know that literature on turbine noise dating back nearly a decade has shown that these models underestimate wind turbine noise levels. But here in the U.S., wind industry acousticians still use the CADNA/A tool without qualification.

Herkimer County residents are now suffering the consequences. And as stated above, the explanation is simple. Herkimer County residents were lied to.

[1] The CADNA/A software tool is written and sold by DataKustik GmbH of Munich, Germany.

[2] The 6 dBA figure comes from New York's published guidance which states "In non-industrial settings the [Sound Pressure Level] should probably not exceed ambient noise by more than 6 dB(A) at the receptor. An increase of 6 dB(A) may cause complaints."

[3] Email from H. Metzen, DataKustik GmbH, manufacturer of CADNA/A software, Nov. 16, 2006.