LWEC Report Card: A microcosm of global warming exaggeration and errors
Jun 6, 2014
Bishop Hill in Climate: Statistics, Climate: WG2

This is a guest post by reader 'Peartreefruiting'.

[Addition below by author's request 11.30am, 7.6.2014]

In the Annual Review 2013 of the British Trust for Ornithology there is an article entitled “There will be changes afoot”, which details observed and expected changes to British habitat as a result of global warming.  It contains the statement “warmer springs have also led to a trend towards many biological events becoming earlier”.  Since 2013 was the coldest British spring for 50 years, it seemed strange timing for such a statement, so I decided to probe into it.  The article on its own has no verifiable data, but it gives a link to this page at the LWEC website, which in turn links to a document entitled “Biodiversity English for Web.pdf”.   LWEC is the organization “Living with Environmental Change”, and its website states that it is a partnership of 22 major UK public sector funders and users of environmental research, including the research councils and central government departments.

On page 7 of the PDF one finds the relevant statement: “during the period 1994–2006 when average annual spring temperatures increased by 1.4°C, the diversity of the bird community increased by 8%”.  (This article will not dwell on the undoubted good news that warmer springs correlated with increased diversity of the bird community.)

The 1.4°C figure still had no provenance, but a helpful BTO ecologist referred me to Davey et al (“Rise of the generalists: evidence for climate driven homogenization in avian communities”, Global Ecol. Biogeogr., 2011), in which, at the top of the “Results” section one finds “mean breeding season temperatures were variable but showed a general upward trend with a mean increase between 1994 and 2006 of 1.39 °C +/- SE 0.004”.  This one sentence has three problems associated with it, which demonstrate the failure of peer review and basic auditing which seems so endemic to climate science.  As a result, typical statements (for example by the BTO ecologist) that “we only use peer-reviewed science” and “the temperature trends may be out of date because of delays in the peer review system” seem farcical.

The first problem is that the 1.39°C is not a trend calculated by the customary method of ordinary least squares (OLS), but is simply the first point subtracted from the last point!  I deduced this as a possibility by transcribing the graph in Figure 2 of the paper to obtain the following values in °C:

11.3,11.9,10.8,11.6,11.2,12.1,11.2,11.5,11.7,12.7,12.0,11.9,12.7

... and the only way I could see to get 1.39 was to take the difference between first and last, being 12.7-11.3 = 1.4. Even so I was surprised when the BTO ecologist confirmed this diagnosis.

The OLS trend of these data over the 12 years is 1.11 +/- 0.40°C, whereas the 1.39 figure is 25% higher.  But it could have been worse: a line drawn between the third point (10.8) and the 10th point (12.7) point would have given 1.9°C over only 7 years!

The second problem is one of Chinese whispers.  Davey et al uses the term “mean breeding season temperatures” but the BTO report talks of “spring temperatures”.  The former is defined as April to July inclusive, whereas meteorological spring is defined as March to May inclusive.

Does this matter?  Well, an independent series shows that there was not any statistically significant warming of English springs between 1994 and 2006.  Yet people will read the BTO asserting that there was.  The specifics are that the Central England Temperature (CET) mean monthly temperature series for 1994-2006 give 0.91°C for April to July (breeding season) over the 12 years and 0.49°C (not statistically significant) for the March to May data (spring).

The third problem is that the “+/- SE 0.004” is meaningless.  It is not a credible estimate of the error of the trend.  What appears to have happened (I cannot be certain and frankly it is so way out that it is hard to care) is that the temperature data used were actually interpolations (via a model) to a 1-km square resolution using more sparsely measured real data.  Then the 1-km squares were treated as independent (which novice use of a statistical tool will naturally obtain) so the vast numbers of these squares drove down the measured uncertainty.

So here is the step by step recipe to exaggerate spring warming in Britain.

Thus the BTO, and no doubt other ecological publications, continue to trumpet the warming glories of the past at a time when, in fact, using annual figures the reverse is true.  For there has been statistically significant British cooling over the last 11 years: the annual CET mean series 2002-2013 shows a statistically significant cooling of 1.07°C, which increases to 1.34°C if maximum temperatures are used instead of mean.

[Addition]

So here is the step by step recipe to exaggerate spring warming in Britain.

1. Hide behind peer review delays to avoid noting the 1994-2013 CET mean temperature spring trend of -0.003+/-0.034°C per annum.
2. Concentrating on 1994-2006 then, consider March-May CET means but reject them because the 0.49°C OLS trend over 12 years is not statistically significant.
3. Instead, use April-July (breeding season) as this gives 0.91°C over 12 years, which is statistically significant.
4. Better still, rather than using freely available CET series, use the data from Davey et al, which give 1.11°C over 12 years.
5. But why use a new fangled method like OLS when simply drawing a straight line through the end points gives 1.39°C? Yes, that value should keep the eco-troops happy, and the great British public surely won’t notice that this is not usually the way that science is done.

Article originally appeared on (http://www.bishop-hill.net/).
See website for complete article licensing information.