Humans have used models to describe and predict their environment for millennia. With the advent of computers the number and sophistication of these models has taken a quantum leap. Many have proven their worth, and their impact upon our view of the universe has been profound. Unfortunately, it is almost inevitable that something with this much influence over our affairs will be misused by those whose with a self-serving agenda – much like junk science.
Dr. Nicholas Kouwen, in his study on wind turbine noise, discovered that the models used to predict that noise substantially underestimated it – a most convenient result, given Ontario’s regulatory regime, for the developer who hired the modellers. In his commentary on why this disconnect occurred he mentioned empirical models and their limitations. I thought the topic was important enough for a separate posting, and here it is.
There are basically two types of models – mathematical and empirical (also called scientific and engineering). Here’s a layman’s simplified explanation of them.
Mathematical. Let’s say you wanted to model the movement of the planets in the solar system. There exists a set of rules that determine, with a high degree of accuracy, the movement of bodies in that environment. These laws, from every observation available to us, are universally applicable over both distance and time. Conveniently for us, they can be fairly easily expressed in a mathematical form. With a computer you can accurately project the position of all the planets for a very long time into the future. Further, since those rules have always worked very well and we continue to have confirmation that they will continue to work we don’t need to keep calibrating them. We can send a spaceship on a certain route and be confident that, barring some unforeseeable incident, it will meet whatever it is supposed to.
Empirical. There are many environments that are just too “messy” for us to be able to reduce them to mathematical formulas. In Dr. Kouwen’s professional life he worked on modelling the movement of groundwater, where there is no reasonable prospect of being able to reduce that environment to a universally-applicable set of mathematical rules. Modelling in this environment is done by observing what happens in the real world and translating it as best as you can to a mathematical form that a computer can handle. It should be self-evident that such a model would apply only to the environment where the observations were made – it has bounds, unlike the mathematical model. Additionally it requires periodic calibration and testing to make sure the underlying environmental conditions still exist.
Bringing this back to wind turbines, recall that in Ontario (and many other jurisdictions) the pre-construction modelling is essentially the only protection the neighbors have from an overly noisy project. Almost needless to say, any noise model is of necessity empirically-based. The complications of buildings, noise sources, atmospheric conditions, ground cover and so on are just too complex and variable for any universal mathematical construct. The Ontario Noise Guidelines for Wind Farms defines the limits and how the models are to be constructed. Page 18, section 6.4.7, Prediction Method:
“Predictions of the total sound level at a Point of Reception or a Participating Receptor must be carried out according to the method described in the standard ISO 9613-2, Reference . The calculations are subject to the specific parameters indicated in Section 6.4.10.”
So ISO 9613-2 becomes fundamental to the process, with some of the parameters defined in a later section. And what environment does 9613 apply to? Page 2, section 1, Scope:
“This method is applicable in practice to a great variety of noise sources and environments. It is applicable, directly or indirectly, to most situations concerning road or rail traffic, industrial noise sources, construction activities, and many other ground-based noise sources. It does not apply to sound from aircraft in flight, or to blast waves from mining, military or similar operations.”
So what constitutes a ground-based noise source? They don’t say directly, but on page 32, Section 9, Accuracy and Limitations there’s Table 5 which specifies how accurate the standard is, but only for heights less than 30 m. Above that this empirical model is not calibrated. As Kouwen notes “ISO 9613‐2 is an empirical model. In general, empirical models should not be used outside the range of the data that was used in their development.”
This is especially true when human welfare and safety are at stake. Kouwen: “The writer knows of no instance in Engineering where a safety factor is not applied – especially when an empirical model is used outside its intended range. In the writer’s own field of water resources modelling, a normal requirement is for the proponent of a project to calibrate and validate any model being used that impacts the safety and well‐being of the public.”
Apparently this “normal requirement” does not apply to wind turbines. If this weren’t bad enough there are additional problems, like the plus or minus 3 dBA normal error range per Table 5, or the absorption factor of the ground (using 0.7 when frozen ground is more like 0.0 – especially from an elevated sound source), or the non-point source of wind turbine noise. Taken altogether there should be no surprise that the actual measurements are far above what the model predicted. And it should also be no surprise that the neighbors are being adversely affected by the noise.
What is quite surprising is that the Ontario government allows this to go on, even subsidizing it while belittling the victims.