Carbon Dioxide: A Satanic Gas?


Thank you for soliciting my testimony on the nature of CarbonDioxide as a “pollutant” with regard to global climate change. Iregard a “pollutant” as something that produces a demonstrable netnegative impact on climate and ecosystems.

“Negative” and “positive” impacts on climate are valuejudgements made by human beings. Within that limitation, I submitthe following:

This testimony demonstrates that the observed climate changesthat have accompanied the enhancement of the natural greenhouseeffect have been considerably smaller than they were originallyforecast to be, and that they are likely to remain similarly small.Further, they are inordinately confined into the winter, ratherthan the summer, and, within the winters, they are inordinatelyconfined to the coldest, deadliest airmasses. There is no overallstatistically significant warming in the average temperature of theUnited States, which is a record of 105 years in length. While theUnited Nations has stated that during the greenhouse enhancement,“the balance of evidence suggests a discernible human influence onglobal climate,” I cannot view what has happened as a net negative;some might easily argue that it is a net benefit. Under neitherinterpretation does this qualify carbon dioxide as a climatic“pollutant.”

In January, 1989, over ten years ago, I first testified onclimate change in this House. I argued that the computerizedclimate models from that era were dramatically overpredictingfuture warming, and that the observed history of climate projecteda much more moderate warming, of 1.0°C to 1.5°C, over thenext century. I further argued that it would eventually berecognized that this moderate climate change would be inordinatelyexpressed in the winter vs. the summer, in the night vs. the day,and that overall it was plausible to argue that these changesconferred a net benefit upon our world.

If I had the perfect vision of knowing what would have happenedto the climate in the next ten years, how the scientific literatureevolved‐​in its attempts to explain the lack of warming, and in itsrefusal to recognize persistent, damaging and pervasive errors inthe forecast that continue to this date‑I would have changednot one word.

This testimony explains why.

In the last ten years, we have learned that:

  • Observed surface warming is most consistent with a forecastbelow the lowest statistical range forecast by climate models.Recent observed changes are several times beneath what wasforecast a mere ten years ago, assuming historical changes incarbon dioxide (see Hansen, et al., 1998).
  • The postwar ratio of winter‐​to‐​summer warming is greater thantwo‐​to‐​one (Balling et al., 1998)
  • Over three‐​quarters of the cold half‐​year warming in theNorthern Hemisphere is confined to the very coldest airmasses. Thewarming outside of these airmasses is a minuscule 0.2°C percentury (Michaels et al., 1999).
  • The variation, or unpredictability, of regional temperatureshas declined significantly on a global basis while there is nochange for precipitation (Michaels et al., 1998)
  • In the United States, streamflow records show that drought hasdecreased while flooding has not increased. (Lins and Slack,1999).
  • Maximum winds in hurricanes that affect the United States havesignificantly declined (Houghton et al., 1995), and thereis no evidence for a global increase in damaging storms (Landsea etal. ‚1996).
  • The Kyoto Protocol to the United Nations Framework Conventionon Climate Change will have no discernable impact on global climatewithin any reasonable policy timeframe (Wigley, 1998).

In toto, these findings lead inescapably to the conclusionCarbon Dioxide is not a “pollutant,” and plausibly argue that it isa net benefit.

Scientific Background

It has been known since 1872 that water vapor and carbon dioxideare the principal “greenhouse” gases in the atmosphere, and thatincreasing their concentration should elevate the temperature inthe lower atmosphere. What has been a subject of contention eversince, is the amount and character of the warming.

Because of all of the atmospheric greenhouse gases emitted byhuman activity, we have progressed to roughly a 60% increase in theequivalent natural carbon dioxide greenhouse effect. The earliestclimate projections, made by Arrhenius in 1896, indicated thiswould result in a rise in mean global temperature of approximately3.25°C. Computer models that served as the basis firstScientific Assessment of Climate Change by the United NationsIntergovernmental Panel on Climate Change were around 1.8°C forcurrent greenhouse changes (Murphy and Mitchell, 1995). These werelower than original estimates largely because of the retardation ofdirect warming by the ocean.

The 1.8°C figure was typical of the range of most climatemodels, and led to the scientific bifurcation between the modellingcommunity and the more data‐​driven empiricists, who argued that theobserved 20th century warming of 0.6°C (with half of thatbefore the major greenhouse gas changes) indicated future warmingwould be around one‐​third of the mean projected value of 4.2°Cover the next century, or around 1.0 to 1.5°C.

The IPCC admitted the validity of this position in its 1995report, when it wrote that:

When increases in greenhouse gases only are taken intoaccount…most [climate models] produce a greater mean warming thanhas been observed to date, unless a lower climate sensitivity [tothe greenhouse effect] is used…There is growing evidence thatincreases in sulfate aerosols are partially counteracting the[warming] due to increases in greenhouse gases.

Are sulfate aerosols responsible for the now‐​admitted dearth ofwarming? In previous testimony I have shown how poorly thisargument stands the critical test of the data. Suffice it to saythat the record of the three dimensional atmospheric temperature inrecent decades does not appear at all consistent with thishypothesis. Instead of repeating that argument, I would simplypoint out that the southern half of the planet is virtually devoidof sulfates, and should have warmed at a prodigious and consistentrate for the last two decades. Unfortunately, we have very fewlong‐​term weather records from that half of the planet, and almostall come from the relatively uncommon landmasses. However, we dohave over two decades of satellite data (Figure 1), adjusted byJohn Christy for orbital decay and other drifts; it shows no changein temperature whatsoever, although the prominent spike and retreatfrom the 1998 El Niño is rather striking.

Figure 1. Southern Hemisphere MSU satellite temperatures,drift-adjusted, from John Christy of University of Alabama,1/1/79 – 8/31/99. The sulfate hypothesis implies this zone should bewarming rapidly.

The failure of the climate models is much more profound than anyerror that could simply be corrected by reducing the amount ofincoming surface radiation, which is what the sulfate “fix” does.Instead, it is a failure in the vertical dimension that has beenoccurring for nearly a quarter‐​century.

Figure 2 shows the entire concurrency for our three records of“global” temperature, which is limited by the beginning of thesatellite MSU data in January 1, 1979. The record is now completingits 21st year.

Our figure shows satellite temperatures, weather balloontemperatures roughly between 5,000 and 30,000 feet, and surfacetemperatures measured by thermometers. There is an increase in thesurface record of 0.15°C/decade. Research by NASA scientistsdemonstrate that about 0.02°C/decade of this is a result ofchanges in the sun (Lean and Rind, 1998), leaving a remaining0.13° C/​decade ascribable to human influence or other naturalvariation. The other two records show no change.

The disparity between the surface, satellite and weather balloonreadings is likely to have some basis in reality. The concordancebetween the satellites and balloons cannot be from chance, so theremust be some process occurring in the lowest layers (below 5,000feet) that is not being picked up in those two records.

Figure 2. Satellite, weather balloon (5,000−30,000 ft.),and surface temperatures since 1/1/79, the beginning of thesatellite record.

My research shows that this warming below 5,000 feet is largelyconfined to the winter half‐​year (October‐​March in the NorthernHemisphere, April‐​September in the Southern); as Figure 3 shows,the ratio of winter‐​to‐​summer warming is greater thantwo‐​to‐​one.

Figure 3. Winter and summer half‐​year warming since 1945 shows thedominance of winter temperature change.

Now, when we look at the cold half‐​year temperatures in theNorthern Hemisphere since World War II, it is apparent that thewarming is inordinately confined to Siberia and northwestern NorthAmerica (Figure 4). These are the two “source regions” for thecoldest continental airmasses on earth. Because of their coldness,they are very dry, and because of their dryness they have verylittle “natural” greenhouse effect and are consequently “warmed”(if changing the temperature from ‑40° C to ‑38° C can becalled a “warming”!) more rapidly than moist, summer air.

Figure 4. Winter minus Summer warming trends since 1945show the dominance of warming in Siberia and northwestern NorthAmerica in winter.

In the winter half‐​year, these airmasses occupy around 25% ofour hemisphere. Recently published research (Michaels et al., 1999;Michaels et al., accepted) shows that over three quarters of thewinter warming is confined to this very cold air. When we comparethe average postwar warming in the statistical gridcells thatcomprise these airmasses to those that don’t, the result is trulystunning. The coldest air is warming up a rate 10 timeslarger than the remainder of the hemisphere; see Figure 5. Thatresearch also proves that the warming is largely confined to thecold air masses, and that the more severely cold they are, themore they warm.

Figure 5. Average warming rate in Northern Hemispheregridcells that are cold and dry (right) vs. theremainder.

Together, these findings all prove that over the entireconcurrency of the surface, satellite and balloon records, there isa warming confined to the bottom 5,000 feet of the atmosphere, butover two‐​thirds of it is in the winter, and three‐​quarters of thatis in the most profoundly cold continental air that we know of. Ifthis is the work of carbon dioxide, carbon dioxide is not apollutant.

Together, these findings also demonstrate a persistent,damaging, and pervasive error in all climate models, includingthose that serve as the basis for the Kyoto Protocol.

Figure 6. Warming predicted for today’s change ingreenhouse gases and sulfate aerosols by the LLNL model. Note thatthe entire zone from 5,000 feet to the stratosphere is predicted tohave warmed.

Figure 6 shows the projected quarter‐​century warming from theLawrence Livermore National Laboratory (LLNL) climate modelincorporating greenhouse warming and sulfate cooling (addition ofstratospheric ozone depletion changes the result very little), asoriginally published by Santer et al. (1996). This finding, morethan any other single result, served as the basis for the 1995 IPCCstatement that “the balance of evidence suggests a discerniblehuman influence on global climate.” However, there is no reason tosingle out the LLNL model except for its wide availability; everyother one behaves in a quite similar fashion.

The LLNL model and all others are clearly making an egregiouserror that renders the magnitude of their predictions of globalwarming virtually useless: They all have dramatically failed topredict what happened between 5,000 feet and the bottom of thestratosphere. This comprises over 80% of the troposphere, or theearth’s active weather zone.

Our chart shows the observed warming in this zone (as publishedby Santer et al., 1999) for various upper atmospheric records vs.the average warming predicted the current suite of climate models.There is no statistically significant warming in the observeddata since the satellite/​balloon concurrency in 1979, whilethe models have an average warming rate of 0.23°C/decade(Figure 7).

Figure 7. Model‐​projected average tropospheric warming(left) since 1979 vs. observed values published by Santer et al.(1999).

In other words, the models have been wrong for the lastquarter‐​century‐​the period of greatest greenhouse gas increase‐​over80% of the troposphere.

The atmosphere is a mixed fluid; the behavior in one verticallevel depends in part on behavior in others. It is profoundlytroubling that, for the last quarter‐​century, that projections ofsurface warming are much closer to observed values, than what hasbeen observed in the remaining 80% of the troposphere. Thisdifferential calls into question the validity of any projection,surface or otherwise coming from these models. More important, itindicates that the “sulfate‐​greenhouse” paradigm is so inaccuratethat it misspecified almost all of the troposphere.

The Ubiquitous Nature of Observed Changes

The National Oceanic and Atmospheric Administration has analyzedpostwar temperature trends in the Unites States and found similarresults; see Figure 8. The largest warming in the last threedecades occurs in winter (January through March) which is the timeof year in which severity and presence of the cold high pressuresystems that form in northwestern North America largely determinethe winter departure from normal. Late summer and early falltemperatures actually show a slight decline.

Figure 8. Seasonal changes since 1966 in the U.S. record,according to the U.S. National Oceanic and AtmosphericAdministration.

NOAA has also analyzed U.S. temperatures back to 1895. Eventhough this record contains some large cities with artificial urbanwarming there is no statistically significant warming in theoverall record (see figure 9).

Figure 9. There is no statistically significant overallwarming trend in the record of U.S. temperatures, which is 105years long.

The record can be broken down decadally, i.e., 1895−1997,1905−1997, 1915 – 97, etc…Only one of these combinations shows asignificant warming that would exceed the Frye Rule ofdisqualification for “junk science” (i.e. at the .05 level), andthat is the period 1965‐​present. The chance that one trend (out ofthe ten possible ones) would show a warming is statisticallycommon, even at this probability level. It is noteworthy that NOAA,in the report on climate change that served as the basis for figure8, “chose” the only period (1965−97) in the entire 105‐​year recordthat showed statistically significant warming. Later decadalperiods (i.e. 1975 – 97, or 1985 – 97) do not; and neither does anyearlier period. Perhaps an appropriate question would be to ask whythe only such period out of the ten possible ones was selected foranalysis and publicity.

An Alternative Interpretation

It is hard for me to believe that the billions of dollars thatAmerican taxpayers have invested in climate modeling has produced acompletely worthless result with regard to human‐​induced climatechange. Rather more intriguing is the notion that at least themodels have the functional form of the warming right, but insteadare indecisive about its magnitude.

Figure 10 shows projected warming from a large family ofdifferent climate models. Some increase their atmospheric carbondioxide at 1% per year, effectively. Others use the U.N. standardof 0.7%, and others have been adjusted for the observed lower ratenoted by Hansen (1998). Some have sulfate aerosols in them andothers do not.

Figure 10. Output from several representative climatemodels. Once warming starts, it takes place at a constantrate.

Regardless of all of these varying assumptions about differencesin the exponential rate of greenhouse forcing or the presence orabsence of sulfates, one clear fact emerges. In general,they are all straight lines. Once greenhouse warmingstarts, it proceeds as straight line, not as an exponentialincrease.

What differs between the models is not their functionalform‐​straight lines‐​but the slope (or rate of increase) inthose lines. In fact, the mean and standard error of the warmingare 0.25°C ± 0.07°C/decade, where the confidencerange is at 67%.

Which of these models is likely to be correct? Under theassumption of linearity, nature helps to provide an answer, asglobal near‐​surface temperature has risen as a straight line, too,in the last three decades. The slope since 1968, when warmingbegan, is 0.15°C/decade (Figure 11). This is slightly below thelow confidence limit given by the ensemble of models shownhere.

Figure 11. The observed linear trend in temperature for thelast three decades is lower than the projections of most climatemodels.

However, Lean and Rind indicate that approximately0.02°C/decade of recent warming is likely to be caused by solarvariation; subtracting away this small amount gives a warming rateof 0.13°C/ten years.

Assuming linearity, this gives a rise of 1.3°C in the nextcentury‐​precisely at the midpoint of the range to which I firsttestified over ten years ago.

Models are also linear with respect to their cold and warmseason warmings. Given the differential that we have seen since1968, the expected winter and summer half‐​year warmings work out to1.45 and 1.15° C, respectively, in the next century.

During this century, we experienced a temperature rise ofapproximately half of these values. Crop yields quintupled. Lifespan doubled, in part because of better nutrition. Winters warmed.Growing seasons lengthened. The planet became greener. Increasingcarbon dioxide had something to do with each and every one ofthese. There is simply no logical reason to assume that doing thesame, this time in 50, instead of 100 years, will have anydifferent effect in kind. That kind of improvement in the qualityof human life could hardly be caused by a “pollutant.”


Balling, R.C., et al., 1998. Analysis of winter and summerwarming rates in gridded temperature timeseries. ClimateResearch, 9, 175 – 181.

Hansen, J.E., et al., 1998. A common‐​sense climate index: Isclimate changing noticeably? Proceeding of the National Academyof Sciences, 95, 4113 – 4120.

Houghton, J.T., et al., (eds.),1996. Climate Change 1995:The science of climate change, contribution of the WGI to theSecond Assessment Report of the Intergovernmental Panel on ClimateChange. Cambridge University Press, New York, 572pp.

Landsea, C.W., et al., 1996. Downward trends in the frequency ofintense Atlantic hurricanes during the past five decades.Geophysical Research Letters, 23,1697−1700.

Lean, J., and D. Rind, 1998. Climate forcing by changing solarradiation. Journal of Climate, 11,3069−3094.

Lins, H.F., and J.R. and Slack, 1999. Streamflow trends in theUnited States. Geophysical Research Letters,26, 227 – 230.

Michaels, P.J., et al., 1998. Analysis of trends in thevariability of daily and monthly historical temperaturemeasurements. Climate Research, 10,27−33.

Michaels, P.J. et al.,.1999. Greenhouse warming in coldanticyclones. To appear in Proceedings of the 1999International Congress of Biometeorology and InternationalConference on Urban Climatology, Sydney, Australia.

Michaels, P.J. et al.,. Observed warming in cold anticyclones.accepted for Climate Research.

Murphy, J.M., and J.F.B.Mitchell, 1995. Transient response ofthe Hadley Centre Coupled Model to increasing carbon dioxide. PartII. temporal and spatial evolution of patterns. Journal ofClimate, 8, 57 – 80.

Santer, B.D., et al., 1996. A search for human influences on thethermal structure of the atmosphere. Nature,382, 39 – 46.

Santer, B.D., et al., 1999. Uncertainties in “observational“estimates of temperature change in the free atmosphere. Journalof Geophysical Research, 104, 6305 – 6334.

Tinker, R., 1999. U.S. Temperature and PrecipitationTrends. Climate Prediction Center, National Atmospheric andOceanic Administration. www​.cpc​.noaa​.gov/​t​r​n​d​t​e​x​t.htm

Wigley, T.M.L., 1998. The Kyoto Protocol: CO2, CH4 and climateimplications. Geophysical Research Letters,25, 2285 – 2288.

Patrick J. Michaels

Subcommittee on National Economic Growth, Natural Resources and Regulatory Affairs
United States House of Representatives