“Clearest Evidence” For Human “Fingerprint?”
Results clouded if more complete data used

More fireworks than usual were detonated on July 4, when Nature magazine published a lengthy article by B.D. Santer and 12 coauthors that claimed “The observed patterns of temperature change in [the atmosphere away from the earth’s surface] are similar to those predicted by state-of-the-art climate models.”  Newspaper articles featured Australia’s Neville Nicholls, one of the lead authors on the new U.N. report of the Intergovernmental Panel on Climate Change (IPCC), stating this is the “clearest evidence yet that humans may have affected global climate.”

There can be little doubt that the timing of this report—released four days before the signatories of the Rio Treaty sit down to discuss limiting greenhouse emissions—was for maximum political effect.  Apparently Nature, which bills itself as “the world’s most prestigious weekly journal of science” is not immune to being toyed with.

More important, though, is the science.  Santer’s paper correlates temperature changes predicted by climate models to those observed in the atmosphere above the surface, as measured by weather balloons.  In general, he found an increasing correlation over time between the models (which include warming from increasing greenhouse gases and compensatory cooling of sulfate aerosols) and atmospheric temperatures.  There was a slight additional improvement from inclusion of the effects of stratospheric ozone changes, but much of which resulted from an “idealized altitudinal profile of ozone loss.”  All in all, it’s fairest to say that the best-claimed match-up lies in the sulfate and greenhouse combinations.

Essentially, the models say that sulfates work to mute warming primarily in the Northern Hemisphere (even creating some cooling), while the middle atmosphere, from 5,000 to 30,000 feet, warms dramatically in the sulfate-free Southern Hemisphere.

Santer used two models.  One, developed at Lawrence Livermore National Laboratory (LLNL), has the wrong greenhouse effect.  Nowhere does the paper explicitly state this problem; instead it repeatedly says it is modeling the “atmospheric CO₂,” which, in fact, underestimates human-caused changes to the greenhouse effect by almost 50 percent.  The other half (actually 48 percent) comes from non-CO₂ greenhouse enhancers such as methane and CFCs.*

In Figure 1, we adjusted the output of the LLNL model using the correct greenhouse forcing.  After making standard allowances for the fact that the earth’s oceans will retard some of the warming, observed temperature changes should be around 60 percent of what the map shows.  Nonetheless, the overprediction of current temperature is enormous.  While ground-based records show a warming of about 0.6°C in the last 100 years, this simulation heats up more than 2°C globally—and the massive warmings of the Southern Ocean and the North Atlantic, respectively, represent the largest regional errors that we know of.

Figure 1.  Temperature change estimated for the Santer et al. sulfate + greenhouse model when the accepted greenhouse change of 2.45 watts/square meter is put in the model.

The other sulfate + greenhouse model that Santer et al. used was published by Mitchell et al. last year and has been repeatedly cited in this Report because it predicts so little warming.  Because of the sulfates, this model has just 67 percent of the warming energy of a greenhouse-only model, resulting in a net warming for a doubling of CO₂ of 1.7°C.  But since 0.4°C of this warming has purportedly occurred, this leaves a measly 1.3°C.  If the observed change matches up best with this model, herein lies strong evidence for people who championed the idea that climate change would be much less than feared.

Correlation Studies vs. Reality

Ironically, if the full greenhouse effect changes were incorporated into the model, the correlation between the models and reality would still increase over time.  That’s because of the mathematics involved in calculating the correlation coefficients that quantify the relative patterns of temperature change.  The global error does not matter as long as more warming (or less cooling!) is predicted in the Southern Hemisphere troposphere (lower atmosphere) than in the Northern Hemisphere.  Thus, this methodology can take a perfectly absurd model and still claim that reality increasingly resembles something that everyone—including even the most strident greenhouse alarmist—knows to be wrong.

Another reason for the observed behavior is that much of the correspondence that Santer et al. found between models and reality is way up in the stratosphere, which has experienced a strong cooling that also appears in the climate models.

Figure 2 shows their most important correlation graphs.  The chart on the top shows the increasing correspondence between the LLNL models and the mean temperature recorded between 5,000 and 65,000 feet.  The one on the bottom is between 18,000 and 65,000 feet.

Figure 2.  Correlation pattern changes over time for 5,000–65,000 feet (top) and 5,000–15,000 feet (bottom).  Solid lines are the results for the CO₂ only model, while the dotted lines are the results of including the effects of sulfate aerosols with CO₂.

Technically, these are graphs of the correlation coefficient between the model and observed temperature fields.  Squaring the correlation coefficient gives the percentage of year-to-year variation that is “explained,” or related, between the model and reality.  In the recent years, the explained variance between the layer between 5,000 and 18,000 feet is a mere 16 percent (0.4 x 0.4).  But when they looked at the layer between 5,000 and 65,000 feet, the explained variance shoots up to 64 percent (0.8 x 0.8).  Thus, 75 percent of the correspondence  [(64%–16%/)/64%] is because of conformity above 18,000 feet.

Which gets us back to the old problem that has dogged this issue for decades, namely that greenhouse warming has been very hard to find in the bottom of the atmosphere, which just happens to be where it matters the most.

Comparisons to Satellite

Everyone, from the editor of this Report to Vice President Gore, acknowledges that the hemispheric mean temperatures measured by satellite correlate perfectly with mean temperatures measured by weather balloons between 5,000 and 30,000 feet.  Because the satellite record shows no warming since 1979 (near the midpoint of the study period of 1962 to 1987 Santer used), most of the observed correspondence between the model and reality in recent years must be almost entirely above 30,000 feet, not just 18,000 feet.

Need proof?  Figure 3, taken from Santer et al., shows that the warming predicted to have occurred in their combined sulfate-greenhouse model is obviously greatest in the layer that runs roughly from 5,000 to 30,000 feet in the Southern Hemisphere.  Compare this with the significant decline in the satellite temperatures for that hemisphere as can be found in our Planet Watch section.  We’re afraid that much of the correspondence in Santer et al. stems from the period chosen for study.

Figure 3.  Warming predicted in Santer et al., 1996.  Hatched areas represent cooling.

We compared Santer’s results with the tropospheric and stratospheric temperatures measured by satellites during their period of overlap.  While that period is not long (eight years), one thing jumped out of our analysis:   The contribution of satellite-measured tropospheric temperature to the increasing correlation pattern during that period is statistically insignificant.  The contribution of stratospheric cooling, which is forecast by virtually every climate model, was exceedingly important.

Is the Result an Artifact of the Data Sample?

Much of the increased correlation between the observed temperatures in the lower atmosphere and the model results from selection of the weather balloon data.  Figure 4 (top), from Santer et al., shows the trend in temperature over the period of his study, from 1963 to 1987.  Note the warming shown in the Southern Hemisphere around 30° to 60° between 5,000 and 30,000 feet.  We’ve plotted out the longer temperature record of 5,000 to 30,000 feet, from Angell et al., which runs from 1957 through 1995, and included it as Figure 4 (bottom).

Figure 4.  Observed warming in Santer et al. (1996) from 1963 to 1987 (top).  We have highlighted the region in the Southern Hemisphere that shows the strong observed warming.  The entire temperature history over the same region from 1957 to 1995 shows no significant warming trend (bottom).  However, the period that was chosen for study by Santer et al. (filled circles) warms dramatically.

The period that Santer et al. studied corresponds precisely with a profound warming trend in this region.  But when all of the data (1957 to 1995) are included, there’s no trend whatsoever!  We don’t know what to call this, but we believe that at least one of the 13 prestigious authors on this paper must have known this to be the case.

The Bottom Line

We are frankly rather amazed that this paper could have emerged into the refereed literature in its present state; that is not to say that the work is bad, but that there are serious questions—similar to ours—that the reviewers should have asked.

The inescapable conclusions:

1. The vast majority of the “fingerprints” of the greenhouse effect are found way up  in the atmosphere, especially in the stratosphere.

2. The “detection” models that were used either don’t predict very much future warming or were run with the wrong greenhouse effect and produce absurd results when the right numbers are put in.

3. And finally, down here in the lower atmosphere, the evidence is much more smudged and is based upon a highly selected set of data that, when viewed in toto, shows something dramatically different than what the paper purports.

Two weeks ago, in a series of email exchanges with Dr. Santer, he requested that we print his recent correspondence to the Wall Street Journal, and we agreed to place it in this issue.  We were not informed at the time of his upcoming Nature article, which we discuss here.  We simply don’t have the space for Santer’s letter in this issue, but we will print it in the future.

It may have been more appropriate to delay anyway, as three more letters that relate to this issue appeared in the July 11 Wall Street Journal.

*The change in warming energy due to the greenhouse effect is 1.26 Watts/square meter in Santer et al.  The IPCC (1995), cited as the “consensus of scientists,” gives the figure as 2.45.  The atmosphere responds quite linearly to changes at this level; i.e., 2.45 Watts will produce about 1.9 times as much warming as 1.26 Watts.

References:

Angell, J.K., (1994 and updates)  From Trends 93, U.S. Department of Energy, 636–672.

Nicholls, N., (1996)  An Incriminating Fingerprint.  Nature 382, 27–28.

Santer, B.D. et al., (1996)  A Search for Human Influences on the Thermal Structure of the Atmosphere.  Nature 382, 39–45.