TCS Daily

Hot or Not?

By Willie Soon - January 3, 2003 12:00 AM

Was 2002 one of the hottest years in history? After all, pronouncements of record warmth for 2002 were made since the beginning of 2002.

For example, in February 2002, NOAA warned that "November 2001-January 2002: Warmest on the record in the US; Global temperatures warmest on record [in the 123 years of record keeping] in January." And last April, the BBC news claimed that January, February and March of 2002 "were globally the warmest" since records began in 1860. By August 2002, announced that "Globally 2002 is likely to be warmer than 2001, and may even break the record set in 1998."

By December 11, 2002, another news blizzard dumped stories feeding off a press release from the Earth Policy Institute's Lester Brown, with headlines like "2002 heading for No. 2 spot in climate records"; "Scientists: '02 second hottest as warming speeds"; and "Scientists: Humans the main culprit in global warming."

And if the newspapers are to be believed, the new year is already gearing up to be a hot one. The New York Times writer Andrew Revkin announced on New Year's eve that "Climate experts say global temperatures in 2003 could match or beat the modern record set in 1998 ...".

All of this sounds like bad news. But how alarmed should we be?

'Truly Meaningful... to Nobody'

First of all, it's important to remember that the attempt to determine the relative ranking of a measured global surface temperature, given current limitations, is impossible. To understand why this is so, let's examine how global surface temperature is derived.

The premise is simple: Start with measurements taken from thermometer stations scattered across the globe. Now average all the good measurements within, say, a 5 degree by 5 degree box (that makes a total of 2592 boxes to cover the earth). From that, form a global mean from the values in all the 2592 grid boxes.

Sounds simple enough. The only problem is temperature has not been reliably measured in all grid boxes over time. Far from it. For example, the number of grid boxes covered by thermometers has increased by more than a factor of five since the onset of the instrumental record in the 1860s.

Moreover, local and regional temperatures are temperature readings made in absolute terms, e.g., 48 F. Those local absolute temperatures must be accurate in order to avoid systematic errors when the global average is computed.

For those reasons, the NASA Goddard Institute for Space Studies web page says that the "reported temperature [by the local news media] is truly meaningful only to a person who happens to visit the weather station at the precise moment when the reported temperature is measured, in other words, to nobody."

NASA's web page then points out that "regional absolute temperature ... cannot be determined from observations alone" because of important measurement difficulties. These difficulties arise from the lack of homogeneous sampling of areas with differing geographical features (e.g., land or ocean, changing surface vegetation or altitude of the weather station) over time or a consistent time of observation (e.g., local noon versus midnight, winter versus summer).

Model Reliability

Chasing after the absolute global surface air temperature is an enormous task. It is a task that cannot be done on the basis of averages of thermometer readings alone, but must solicit help from the output from complex computer simulations of climate. How reliable is such a model-driven process of averaging thermometer readings? The NASA web page says that "the most trusted [climate] models produce a value of roughly 14 Celsius, i.e., 57.2 F, [for the global mean temperature] but it may easily be anywhere between 56 and 58 F; and regionally, let alone locally [like those given by your nightly local weather report], the situation is even worse." In short, the global average is uncertain by about 2 F - that's an uncertainty twice as great as the 1 F surface global warming claimed for the entire 20th century!

Looking at Figure 1 below helps us better understand some of the uncertainties in the global average temperature. The top panel is the "global" surface temperature data from the United Kingdom Meteorological Office/University of East Anglia's web page. Taken at face value, the figure indicates that the global temperature was exceptionally high in the 1990s, with 1998 being the warmest and 2002 the second warmest.

Figure 1: Globally averaged surface temperature (top panel). Note that the absolute level of the global temperature is only known to within plus or minus one degree Fahrenheit (marked by the error bar, as suggested by the NASA Goddard Institute for Space Studies webpage). Area coverage of the grid boxes covered by surface thermometers (bottom panel) from 1856 to 2002. Uncertainty in the temperature record arises in part from the great change in area coverage by thermometers over the record. The global temperature also appears to be noticeably affected by the natural Great Pacific Climate Shift of 1976-1977, a widespread, natural warming of the Pacific that has endured for more than two decades.

The bottom panel of Figure 1 shows the changing area coverage by the instrumental thermometers that made up the "global" temperature presented in the top panel. The coverage has not been uniform over time. The area covered by the "global" temperature readings during 1860s extends only over 15% of the globe, 40% at 1900, peaking near 87% around 1985-1987 and declining to about 80% by 2001-2002.

How important is vastly changing area coverage of the available surface thermometer readings in affecting the relative ranking of individual years across the record? The modeled surface temperature taken from limited areas, as defined by the actual coverage by thermometer readings, can be mathematically compared to the modeled result averaged over the full globe. This gives us an estimate of the uncertainty arising from inadequate and changing area coverage.

Patrick Duffy and colleagues from the Lawrence Livermore National Laboratory reported the following results of such a study (Journal of Climate, July 1, 2001):

"In 10 out of 16 simulations, the estimate of the warmest year in the last century based on the masked model output agrees with the true answer obtained from the globally complete model data. However, estimates of the 3 (5) warmest years (including their relative ranking) in the masked and globally complete model output agree in only 3 (2) of 16 simulations. In two cases, the year that is the warmest based on globally complete model output is not among the three warmest when estimates are made from masked model output. Thus, even though the warmest years occurred during the last 1-2 decades, when observational coverage was relatively good, estimates of the relative ranking of the warmest years are still sensitive to sampling errors induced by incomplete and time-varying coverage of observational data." (Emphasis added)
And it's true that 2002's absolute "global" surface temperature has been assigned a value of 58.1 F. But the lesson is that its rank cannot be confidently distinguished from estimates of other years that are 1 F warmer or cooler. That 2002 may have been a relatively warm year can be safely established only for the past 26 years or so, since the period of natural warming that began with the Great Pacific Climate Shift of 1976-1977, and during the time of significant area coverage by thermometers. It would be much harder to confirm the ranking of the 2002 temperature compared to say, the 1900s or even 1940s, simply because of the biases and uncertainties introduced by the surface coverage of the instrumental data.

Despite these uncertainties, unsubstantiated public alarmism suggesting that "1998 and 2002 are the two warmest years ... in at least the last 1,000 years" continues. Meanwhile, scientific facts critical to the improvement of temperature records are largely unseen in the public arena, such as: "the VOS data [Volunteer Observing Ships for sea surface temperature data so important for the construction of the global surface temperature] are sparse in high southern latitudes, off major shipping lanes and in parts of the tropics, and the overall number of observations is dropping" Is it hot enough for you? We just don't know.

TCS Daily Archives