Climate Audit

Wind speed trends over the contiguous USA by Pryor et al. (2009, in press, JGR)

Some (read on to see who) would say that this particular wind farm energy reduction study is speculative, inconclusive, preliminary, and premature, and with the authors’ hesitant equivocation in press interviews, even they may agree with that particular straw man. [Read comment 1 for more about bad science reporting]

The Associated Press story heralds this as a “first-of-its-kind study suggests that average and peak wind speeds have been noticeably slowing since 1973, especially in the Midwest and the East.”

Quotes from various scientists:

“It’s a very large effect,” said study co-author Eugene Takle, a professor of atmospheric science at Iowa State University. In some places in the Midwest, the trend shows a 10 percent drop or more over a decade. That adds up when the average wind speed in the region is about 10 to 12 miles per hour.

The new study “demonstrates, rather conclusively in my mind, that average and peak wind speeds have decreased over the U.S. in recent decades,” said Michael Mann, director of the Earth System Science Center at Penn State University.

Jeff Freedman, an atmospheric scientist with AWS Truewind, an Albany, N.Y., renewable energy consulting firm, has studied the same topic, but hasn’t published in a scientific journal yet. He said his research has found no definitive trend of reduced surface wind speed.

Robert Gramlich, policy director at the American Wind Energy Association, said the idea of reduced winds was new to him. He wants to see verification from other studies before he worries too much about it.

Gavin Schmidt, a climate scientist at NASA’s Goddard Institute of Space Studies, told the Guardian the study had yet to establish a clear pattern of declining winds, and that it was too soon to be thinking of the effects on wind energy industry.
“It’s still very preliminary. My feeling is that it is way too premature to be talking about the impact that this makes.”

Now, this is not an example of consensus, especially when Mann and Schmidt have contrary views, which were marvelously printed one after another in the AP story. Mann’s gleaming endorsement seems a bit premature and some would wonder if he actually read the same study that his fellow Team Member did.

Now, to the study. What does the actual text, meaning the abstract, body, and conclusions actually say? Do the remarks in the press releases by the authors actually represent the research they submitted to the Journal, underwent peer-review, and was accepted for publication? The following equivocation in the AP story by the authors suggests that some hyperbole is at work here:

Still, the study, which will be published in August in the peer-reviewed Journal of Geophysical Research, is preliminary. There are enough questions that even the authors say it’s too early to know if this is a real trend or not. But it raises a new side effect of global warming that hasn’t been looked into before.

The ambiguity of the results is due to changes in wind-measuring instruments over the years, according to Pryor. And while actual measurements found diminished winds, some climate computer models — which are not direct observations — did not, she said.

Yet, a couple of earlier studies also found wind reductions in Australia and Europe, offering more comfort that the U.S. findings are real, Pryor and Takle said.

It also makes sense based on how weather and climate work, Takle said. In global warming, the poles warm more and faster than the rest of the globe, and temperature records, especially in the Arctic, show this. That means the temperature difference between the poles and the equator shrinks and with it the difference in air pressure in the two regions. Differences in barometric pressure are a main driver in strong winds. Lower pressure difference means less wind.

Even so, that information doesn’t provide the definitive proof that science requires to connect reduced wind speeds to global warming, the authors said. In climate change science, there is a rigorous and specific method — which looks at all possible causes and charts their specific effects — to attribute an effect to global warming. That should be done eventually with wind, scientists say.

Let’s get this straight: the study is inconclusive and has many outstanding questions with ambiguous results, but it is consistent with what you would expect with global warming. Presto. But, in climate science, there is a rigorous and specific method of attribution — which the authors did not do — but suggest should be eventually done with wind measurements.

Figure 2: (a) Annual percentiles for 1200 UTC observations from site 724320 (5th, 10th, 20th …
90th, 95th percentile, where the 50th and 90th percentiles are shown in the blue and red,
respectively). Despite considerable inter-annual variability, data from this station exhibit
a significant downward trend in both the 50th percentile (of approximately 0.7%/year) and
the 90th percentile (of approximately 0.6%/year) wind speed. Output from the other data
sources used herein for the grid cell containing Evansville are shown in frames (b) – (g).

The 10-meter wind speeds from the various climate model and/or reanalysis data sources are condensed in this figure to one-grid point or station location, which is a risky way to validate the observations — in this case Evansville. Originally, I thought that these medians were an area-average of the entire United States, which would be more representative of the grandiose claims in the press releases. The reanalysis data sets range from grid-spacing of 0.33 (NARR) to 2.5 degrees (ERA-40 and NCEP) and the climate models are at 50 km grid spacing, all too small to resolve the topography to be representative of a given station location. The authors do point out, importantly:

Observational data – due to instrumentation changes, station moves, changes in land-use or obstacles, and observational sites may not be regionally representative

[Note: the land surface data are described as follows]
1. 00UTC and 12UTC NCDC land-based near-surface wind speeds NCDC-6421 [Groisman, 2002] 336-stations chosen out of 1655 1973-2000
2. DS3505 surface data, global hourly, 193 stations available all stations are airports and military installations

Reanalysis products ensure the data sets are homogeneous and complete, but the near-surface wind speeds are strongly influenced by model physics and data that are assimilated.

Also, it must be noted that 10-meter wind speed is not an assimilated quantity in the models, but it is extrapolated from the lowest model-level (often 50-100 meters height) by use of Monin-Obukhov similarity theory.

With the work of Anthony Watts on the Surface Station observational records with regards to temperature, it is perhaps a good idea to investigate and pay more attention to the type and location of the anemometer. The authors helpfully point out that the “Observational data – due to instrumentation changes, station moves, changes in land-use or obstacles, and observational sites may not be regionally representative.”

They go on to say:

Studies that have analyzed wind speed data from terrestrial anemometers have generally found declines over the last 30-50 years (see summary in McVicar et al. [2008] and Brazdil et al.[2009]), the cause of which is currently uncertain. In part because of the difficulties in developing long, homogeneous records of observed near-surface wind speeds, reanalysis data have also been used to quantify historical trends and variability in near-surface wind speeds either in conjunction with in situ observations or independent thereof [Hundecha et al., 2008; McVicar et al., 2008; Pryor and Barthelmie, 2003; Trigo et al., 2008].

So, with these data caveats in mind, the authors continue on in their research avenue to address the following points (from the paper):

Herein we analyze 10-m wind speeds from a variety of observational data sets, reanalysis
products and Regional Climate Model (RCM) simulations of the historical period in order to:

Quantify the magnitude and statistical significance of historical trends in wind speeds and the consistency (or not) of trends derived using different data sets; direct observations, reanalysis products and output from RCMs. As a component of this analysis we provide preliminary diagnoses of possible causes of temporal trends in the in situ observations. Specifically, we examine trends in terms of their temporal and spatial signatures, and the role that major instrumentation changes may have played in dictating those trends.

The methodology is fairly straightforward: time series analysis of the 50th and 90th percentiles of the wind speed distributions and the annual mean wind speed using the 00 and 12 UTC observations or model output from each day of the year.

Conclusions: (quoted from Pryor et al. 2009)

1. Magnitudes of trends in observed wind speed records for 1973-2000 and 1973-2005 are substantial – up to 1% per year at multiple stations.

2. Trends in reanalysis data sets and RCM output where present are generally of lesser magnitude and no other data source is as dominated by negative trends as the in situ observations.

3. Temporal trends in the data sets from in situ measurements are of largest magnitude over the eastern US, but negative at the overwhelming majority of stations across the entire contiguous USA. The trends in wind speed percentiles from in situ observations do not exhibit strong seasonality (Figure 6) or a clear signature from the introduction of the ASOS instrumentation (Figure 7). Hence the cause(s) of the declines remains uncertain.

4. Output from NARR for 1979-2006 indicate contrary trends in the 0000 UTC and 1200 UTC output with declining trends over much of the western US in the 0000 UTC wind speeds but increases in the 1200 UTC output (c.f. Figures 4 and 5).

As expected, the period of observation used in the trend analysis has a profound impact on the presence and absence of temporal trends and indeed the sign of trends.

And the most relevant for last:

Based on the analyses presented herein we conclude there are substantial differences between trends derived from carefully quality controlled observational wind speed data, reanalysis products and RCMs, and indeed between wind speeds from different reanalysis data sets and RCMs.

A few comments:

The quality of reanalysis data sets is not yet sufficient for this type of attribution study of high-resolution, regional climate change. This problem is exacerbated when looking at derived variables such as surface temperature and winds, which are not assimilated observations nor representative of station locations. Also, while the authors chose to use ERA-40, NCEP-Reanalysis, and the NARR (a regional reanalysis over North America), there are other sources of reanalysis data available that may be of significantly higher quality. A recently completed reanalysis is called the ERA-interim, which begins in 1989 and continues on to today, based upon a very recent version of the ECMWF operational forecasting numerical model, which is the best on the planet.

While it is no more the final authority on the subject, I downloaded the data and conducted a few simple experiments to verify the rather ambiguous results of Pryor and company. With access to my university servers, and a program called GrADS, it is very easy to replicate Pryor’s work with minimal effort. It is just a few lines of code and a lot of processing to sort out the distributions of 10-meter wind at each grid point over the past 20-years.

Figure: Running calculations of the distribution of 10-meter winds over the United States, domain averaged: 22.5N-51N; 232.5E-294E). So, the median each month is calculated from the ERA-interim reanalysis data 4 times a day [00,06,12,18] UTC, including a total of 1460 model data points.

Figure analysis: no trend.. Statistically and physically speaking, there is no decreasing trend in the distribution of 10-meter winds over the contiguous United States according to the new ERA-interim reanalysis. (Technical note: This also matches the results from the JRA-25 and NARR reanalysis, which extend from 1979-2008)

What about globally? A fair way to determine where downward trends in wind speed energy are occurring is to simply take the first 10-year period [1989-1998] and subtract it from the second 10-year period [1999-2008] in the reanalysis. The annual medians only are averaged to create the 10-year period. The units are meters per second [m/s]. The second figure is the difference in the 90th percentile wind speed.

The relatively large tropical differences are associated with the trade wind modulation by ENSO. The absolute differences are less than 0.2 m/s over the USA, which is well less than a knot. This is in accord with Pryor et al. (2009) who stated in their conclusions that the reanalysis data did not show a decrease like the observations.

So, what does this all mean?
First of all, it is baffling that a press release would produced for this journal manuscript, which can be described as nothing more than a simple observational vs. reanalysis/climate model comparison study, with ambiguous results, which only leads to more questions about the quality of the data used. There is no attempt to even attribute climate change or global warming to the results in the paper, so the suggestions in the press accounts is a rather egregious fear mongering or perhaps simply an example of journalistic malpractice.

This type of study is critical in the incremental improvement in understanding of micrometeorology, which is modulated by large-scale climate. It is a good question why the observations show one thing and all of the reanalysis datasets show the opposite. Thus, in this case, Gavin Schmidt’s assessment is right on the money. “It’s still very preliminary. My feeling is that it is way too premature to be talking about the impact that this makes.”

Or you can simply read the authors own equivocations in the press — in between the crazy, amateur hour so-called journalism by environmental correspondents.

Note: ECMWF ERA-Interim data used in this study/project have been provided by ECMWF/have been obtained from the ECMWF data server.