Wind Energy and Avian Mortality: Why Can't We get any Straight Numbers?

In my writing on renewable energy, I have not yet spent much time on wind energy. I did establish that a single large wind turbine (rated at about 3.5 megawatts) typically contains around 600 kilograms of rare earth metals (ref) and that the production of rare earth metals has pretty major environmental consequences. As well, I noted in my post on energy and power density that wind is strongly limited by its energy density. A study of wind farms in England, Scotland and Wales indicated that most efficient systems have trouble generating more that a maximum of about 4 W/m² (ref). My lack of commentary shouldn't be read as a negative since I cannot foresee a future, free of fossil fuels for energy production, that does not include a significant use of wind as a source of energy. Moreover, when combined with less intermittent power generating technologies (hydro, geothermal, nuclear or if necessary natural gas) wind can add nicely to a good power mix.

Wind energy does have some drawbacks. As mentioned, wind is notoriously intermittent and needs to be paired with a comparable generating capacity that can cut in when the wind chooses not to blow. In British Columbia, we also have a disadvantage in that our best wind resources are not necessarily located convenient to our population centers. This means that major wind projects will often entail transmission lines. As I wrote in my post on geothermal energy, transmission lines can be very expensive to build and maintain but are a necessary compromise in the goal to move away from fossil fuels. I will note that these two sources (geothermal and wind) seem to be obvious next options for British Columbia power production. Admittedly nothing is going to push hydro to the backseat and between run-of-the-river and Site C there may not be a lot of room on the grid for wind and geothermal, but any reasonable economy does better when it is diversified and having an excess of energy (to export?) is always better than having a shortage of energy and suffering from brown-outs.

Having dealt with the introduction, I want to consider the most repeated complaint in any discussion about wind power, that of bat and avian mortality. The topic is a very heated one, but one that is of tremendous interest to me as it hits at the heart of my research on how data is used in environmental decision-making. Depending on whose research you read the avian and bat effects of wind energy are either a major concern or a trivial non-consideration. The estimates of bird mortality in the US (I will stick with US bird data in this discussion as it is the source of much of the debate and there is not enough space to consider bats) can vary wildly with figures ranging from as few as 20,000 birds per year to over 573,000 birds per year (ref). The low number comes from a paper by Benjamin Sovacool (2013). The high number comes from a paper by Smallwood (2013). I have not had the time to deal with the high number but would like to spend a little time analyzing where that low number comes from.

The 2013 paper by Sovacool (which due to publishing schedules is titled "The avian benefits of wind energy: A 2009 update") and a second paper "The Avian and Wildlife Costs of Fossil Fuels and Nuclear Power" (which is dated 2012 (ref) but once again thanks to publishing schedules includes data from the 2013 paper) serve as a fascinating example of science on the edge. Neither paper crosses the line but both present the science with just enough of a question mark to make you wonder. The 2012 paper has a pretty inauspicious start with the following appearing in the first paragraph of the Introduction:

Use of wind turbines means less consumption and pollution of water resources – a real concern since about half of water use in the United States involves producing electricity in thermoelectric plants (US Geological Survey 2005).

Now this quotation is technically correct, but might be construed as being a tad misleading. Most people interpret "consumption" as using something up. In the case of thermoelectric plants the Geological Service points out that “[a]bout 195 Bgal/d, or 48 percent of all freshwater and saline-water withdrawals for 2000, were used for thermoelectric power. Most of this water was derived from surface water and used for once-through cooling at power plants (ref)” So yes, the water was “used”, but it was not “consumed”. Rather it was borrowed for an hour or two and then returned to the same surface water body from whence it came (albeit marginally warmer than when it went in). The paper can leave the reader with a misconception that all that water is being consumed. Later in the same paper, comes another interesting line:

“Unlike fossil fuel and nuclear power plants, which spread their avian-related impacts across an entire fuel cycle, most of a wind farm’s impact occurs in one location. (p 257)” 

This struck me as missing some very important data. You see in the paper the author considers the entire life-cycle for nuclear plants including the mining of the uranium and effects of the tailings ponds or as put at another web site:

The most dominant contribution to Sovacool’s analysis of nuclear power impacts comes from uranium mining and milling operations which he claims “can poison and kill hundreds of birds per facility per year” (ref).

Now as I pointed out in my previous post on rare earth metals, rare earths are incredibly environmentally damaging to mine, extract and refine. For the author to suggest that 600 kilograms of rare earth metals (not to mention all the incredibly energy-intensive aluminum), do not pose a risk to birds is simply inconsistent with environmental reality (for a terrifying description consider this reference). Any analysis that includes the mining, extraction, refining and decommissioning of uranium facilities should, by all measures, do the same for the aluminum and rare earth metals used in the production of wind turbines.

The meat of the two papers, however, is the calculation of an estimate for average number of birds killed per gigawatt/hour (GWh) by wind (0.279 in one paper and 0.269 in the second paper. I am not sure what serves as the basis for difference between the two numbers). This is the number upon which the entire paper builds and was the number that really perked up my ears. I immediately went to the original source material used to generate the number (Erickson 2004 in this ref). Looking at the figures from the source material I did a back-of-the-envelope calculation that resulted in a value of almost 0.700 avian fatalities per GWh. Recognizing that I may have missed something I looked around to see if I was alone in this calculation and found a reply in the same journal by Willis et. al. (ref). They did a more detailed calculation than me and came up with the following:

Sovacool’s (2009) estimate of the average number of birds killed per GWh of.wind power is incorrect and omits a large body of easily. accessible, published data. For five of the six sites listed in his Table 2, Sovacool used fatality estimates that were uncorrected for searcher efficiency and scavenger losses, despite the fact that corrected data are available. Surprisingly, he then points out that failure to correct for searcher efficiency and scavenging is a flaw in published mortality surveys. Using corrected estimates (Barclay et al., 2007 and references therein), and Sovacool’s estimated capacity factor (33%), the actual estimated number of birds killed at those 6 sites is 0.653/GWh, more than double Sovacool’s estimate of 0.269/GWh. Using all 21 sites for which data are presented in Barclay et al. (2007), the average number of birds killed is 1.46/GWh, or over five times Sovacool’s estimate. 

What this tells me is that the results from this paper have to be considered in light of other evidence. I was somewhat surprised to see that the number (0.269) has already appeared in the Wikipedia article on the topic. To be clear, the author clearly qualifies his numbers (he is entirely upfront but in a way that is hard for outsiders to understand) but as indicated by Willis (et. al.), the better numbers were all readily available. In my case, I simply went to the his referenced source (the Erickson paper) and then sought the original references that Erickson had used. It was neither taxing, nor challenging as the Erickson paper actually included many of the corrected values and virtually all of the reports were in the public domain. Moreover, the Erickson paper included additional information that was not included in the analysis. As an example, in the Sovacool paper the following was presented by the author:

Finally, the seventy-three 300 kW wind turbines in Buffalo Ridge, Minnesota, were responsible for 14 deaths per year.

But the Erickson report had the following text:

The Buffalo Ridge, Minnesota site is comprised of 73 300-kW turbines with 33-m rotor diameters and a maximum height of 52.5 m. A four-year study at this site yielded 55 fatalities representing 31 species, with 71% being migrants and just 2% raptors (Johnson et al. 2002). …Fatality rates ranged from 1 fatality per turbine per year (f/t/y) for Phase I to 4.45 f/t/y for Phase III. (The Phase III f/t/y was heavily influenced by one incident involving 14 birds at two adjacent turbines one night.) 

The reference thus had two numbers, one based on the uncorrected search data for the Phase I facility only (55 fatalities/4 years = 14) and a second  based on the corrected rate for the Phase I facility only (1 f/t/y times number of turbines = 75). Going to the original reference (Johnson et al 2002) I discovered reported data on all three phases of the facility. They reported a corrected fatality rate of 72 birds a year for Phase I (versus the 14 presented in the paper) while Phase II was reported to kill 324 birds a year and Phase II 613 birds a year for a total of 1009 birds a year. This is a pretty significant step up from the 14 birds/year used in the calculations or the 75 birds a year based on the annual rate. While you might say 14 birds/72 birds/1009 birds what is the difference, but remember, he used those numbers to extrapolate national number; this makes the potential difference a pretty substantial thing.

So what does all this mean in the end? Well I’m going to make a wildly impolitic comment here. In my mind the biggest hindrance to the advancement of wind power in Canada may, ironically, be the people pushing hardest for its implementation. They are a very ardent lot and can be aggressive in support of their technology. Unfortunately, their aggressive tendencies can result in them downplaying the necessary compromises associated with, and limitations of, the technology. When I read papers like those of Dr. Sovacool it leads me to wonder what other information is not being discussed? I know that cats kill tens of millions of birds a year in North America and that wind energy is a necessary next advancement in the goal to reduce fossil fuel use, but I am less than confident that there aren’t other skeletons hiding in the darkness waiting to pounce? When the seminal papers in the field ignore the life-cycle analysis for the wind turbines but include it for the nuclear plants, and then compare the two? what am I supposed to think about the output of the analyses?