The Kyoto December deadline is fast approaching and the Greenhouse Industry is now into overdrive, pulling out all the stops to convince other scientists, the public, bureaucrats and governments that Global Warming is for real and that we must all plunge ourselves into economic depression to cut down on our energy use.

There is an air of desperation about all this as several governments, including Australia, are now baulking at the horrendous costs involved. Coupled with this is the stubborn refusal of the Earth itself to unambiguously demonstrate there is even anything to worry about. The polar regions are as cold as ever, in spite of model predictions to the contrary. The satellite global temperatures, measured since January 1979, show a slight cooling trend, and all the Global Warming theorists have to show for their hysterical calls for `action now' are some computer models, models which predict the very things the Earth is failing to do.

Even though the models are failures as predictive tools, we are asked to act on their results anyway and plunge ourselves into poverty.

To `prove' that Global Warming is a real threat, there is now increasing resort to `proxy' studies. Since ordinary thermometers fail to show any warming going on, attempts are being made to show warming by indirect evidence, called proxies. There are already examples of such proxies featured on this web site (Arctic Springs, the Bering Glacier, Tree Rings etc.), all of them lame attempts to show that there really is warming around in spite of what the thermometers or older people say about present and past temperatures.

The latest `proxy' is "Granger Causality", a statistical technique dredged out of Economic modelling and grafted onto a climate context. It is unfamiliar to climatology, but it's use in econometrics is controversial. A statistical study by Kaufmann & Stern, based on Granger Causality was published in Nature (reference above) and immediately trumpeted by GreenPeace and others as more `proof' of Global warming.

According to the environmentalist Union of Concerned Scientists, "This study is thorough and elegantly designed science. It is an excellent compliment to Santer et al. (Nature [1996] vol 382 p39), which found changes in the observed spacial patterns of atmosphere temperature were GHG-related."   While there was certainly an `elegant design' to the paper, it was certainly not `science' for reasons which will become clearer.   It was also comparable with the Santer paper, but for quite different reasons from those implied by the UCS.

`Granger Causality', the statistical technique in question, has been used to compare stock market prices with changes in GDP, allowing phase correlations between the two to predict future GDP based on prior stock market trends (1987 being just a mistake?). It is thus primarily a creature of econometric models, and should perhaps have remained so.

An example of `statistical causality' would be this. I drive to work on the highway around 8 am every morning, Monday to Friday, but not Saturday or Sunday. On exactly the same days, a torrent of traffic hits the highway about 15 minutes after I drive on it, but not on the days that I don't drive on it. Therefore there is statistical causality here, that I cause the tide of traffic to follow me 15 minutes after my passage. That's the kind of nonsense that Granger Causality can get you into. Even economists use it with caution as it has many hidden traps, such as the quality of input data and ignorance or oversight of other external causal variables.

As to Kaufmann and Stern's use of Granger Causality in a climate proxy study, few people (not even GreenPeace) would even understand what the paper was about as it was structured in a highly algebraic form (a guaranteed way to avoid public accountability).  Coupled with that was a `technobabble' style of language to ensure even mathematically-inclined people would have difficulty following it.  After all, where in any dictionary this side of Hell would you find the words "counterfactual", "trend-stationary", "teleconnections", or even my favourite - "spatiotemporal".  What on earth were the Nature editors thinking of when they allowed such linguistic abortions to be printed in their journal?  When authors resort to technobabble like that, it's a sure sign they are selling a lemon dressed up as a pineapple.

Here's how an enthusiastic environmental media presented it all to the public.

(Boston, Mass.) - Scientists have found the strongest evidence to date that human activity-burning fossil fuel and cutting down forests-causes global warming. Robert Kaufmann, associate professor of geography at Boston University and David Stern, research fellow at Australian National University, uncovered the evidence using statistical analysis. Their full report, "Evidence for Human Influence on Climate from Hemisphere Temperature Relations," will appear in `Nature' on July 3.

Kaufmann and Stern examined the northern and southern hemispheres' historical temperature record from 1865 to 1994. Using a statistical technique known as the Granger Causality test, they found that there was a "causal order" from the southern hemisphere to the northern hemisphere. "What causal order means is that past values for temperature in the southern hemisphere help us predict temperature in the northern hemisphere better than just looking at past values for temperature in the northern hemisphere," says Kaufmann. Temperature in the northern hemisphere, he adds, where most human activity takes place, is in a statistical-but not physical-sense, dependent on temperature in the southern hemisphere. The reverse is not true.

Kaufmann and Stern then try to account for the causal order with variables that represent natural forces and variables that represent the effects of human activity. Their results showed that changes in solar and volcanic activity alone could not account for the pattern of temperature change. The best explanation for this causal order pattern, says Kaufmann, is human activity-burning coal, oil and natural gas; cutting down forests; and emitting chloro-flouro carbons (gases used in air conditioners and refrigerators).

"The south-north causal order is generated by human activities that increase the warming effect of greenhouse gases globally, but increase the cooling effects of sulfate aerosols mainly in the northern hemisphere," says Kaufmann. Sulfate aerosols, formed from particles when coal and oil is burned, reflect sunlight and produce a cooling effect on the earth's temperature.

"Southern hemisphere temperature, therefore, reflects the increase in greenhouse gases," says Kaufmann. In the northern hemisphere, this "greenhouse signal" is partly obscured by the reflective sulfate aerosols in the lower atmosphere.

The differential effect of the greenhouse gases and sulfates on temperature in the northern and southern hemispheres has increased over time. "Eventually, the differential effects became strong enough to generate the causal order," says Kaufmann. "Only since the early 1970's has the greenhouse signal that we uncovered in our study emerged from the background noise of natural climate fluctuations."

"The fact that the pattern of causal order changes with the concentration of greenhouse gases and sulfates in the atmosphere implies that the order of temperature change in the northern and southern hemisphere can be used as a fingerprint to identify the effects of human activity on global temperature."

Perhaps the gem quote from the above piece is - "Temperature in the northern hemisphere ... is in a statistical-but not physical-sense, dependent on temperature in the southern hemisphere".  In other words, something can be statistically dependent on something else - but not physically dependent on it.  Only in the strange esoteric world of statistics could such an absurd contradiction be uttered, in print, and with a straight face.

But the paper's primary faults has nothing to do with the statistical operations described or with the grotesque language style, but it's prior assumptions, the key premises upon which the whole exercise is based. To re-quote the BU media story -

Their paper implies, but does not state, that the hemispheric temperature series (1865 to 1994) is actually suitable for this kind of statistical analysis. They cite references mainly from the works of P.D. Jones and J. Hansen. While disputes do exist about the presence of heat island distortions in the data series published by Jones and Hansen, it is nonetheless possible, in theory, to develop a generally acceptable temperature series for the northern hemisphere (NH), even back to the mid-19th century.

Not so with the Southern Hemisphere (SH).

Comprehensive station data from the SH only really dates from the 1950s, particularly beginning with the 1957 International Geophysical Year. Prior to that, data collection was fragmentary, with vast ocean areas not covered at all.

Oceans cover some 83% of the SH, and ice a further 5%. The only reasonable source of pre-1950s data for the SH comes from Australasia and South Africa, but even in these cases, the Australian Bureau of Meteorology has been quite emphatic that anything prior to 1910 is useless for historical comparison purposes. Indeed, many Australian stations do show quite inconsistent and erratic records pre-1910, and virtually impossible to correct accurately.

The oceans have been poorly covered throughout the 1865-1994 period, especially the southeast Pacific, while many records from South America and Africa are erratic even up to the present day. In other words, SH surface data is essentially meaningless from 1865 to 1910, highly suspect from 1910 to 1957, and probably flawed even since 1957.

Ironically, the paper claims the greatest statistical significance between NH and SH during the very period (post-1979) when satellite data was available to give some substance to the SH, but which apparently was not used by Kaufmann and Stern.  Had they done so, they may have come up with a very different result.

Uncritical acceptance of the validity of the Jones/Hansen SH numbers was the unstated premise of the whole paper, and it is this premise which is flawed. By accepting them at face value, any statistical operations or conclusions made from such data inherits all the defects of the original numbers.

A second problem with the paper was the process of elimation of the various `third variables' which might explain the observed causal order. Variations in solar irradiance was mentioned, as was El-Nino. However, the most recent finding of solar-climatic effects is the tight correlation between solar cycle length and global temperature (Friis-Christensen/Lassen, Science 254, p.698), an effect which was not mentioned at all in either the paper or in the list of references.

Also not mentioned among 3rd variables was volcanic effects on temperature. Volcanoes were mentioned in the BU press release above, but not explicitly in the paper itself. (However, the authors stated in personal communication that volcanic effects were included). Such effects are not uniform between the hemispheres, as shown by the recent Mount Pinatubo eruption which had a much more profound effect on the NH than on the SH. Climate-impact eruptions occur about once per decade and have differential effects between the hemispheres requiring quite separate screening treatment in any statistical analysis.

In other words, the search for the third variable(s) was limited in scope, and predictably focused on greenhouse gases to the detriment of other likely third variables.

Finally, the concluding sentence reads - "Combined with other results, our results provide further evidence for the conclusion that `the observed trend in global mean temperature over the past 100 years is unlikely to be entirely natural in origin'."

This last quote given in the concluding sentence comes directly from Chapter 8 of the IPCC 1995 Report, and was one of the altered items of text inserted into the IPCC Report after the meeting of panel scientists, becoming the subject of considerable controversy in the USA. From the public perspective, any quoted text from Chapter 8 (especially the altered bits) is tainted by that affair. It was a highly political quote to conclude the paper on.

But then, the pre-Kyoto timing of the paper is also significant.