Climate models: Cannot explain reality, much less predict future

Published October 1, 2000

There is no possibility that the new method can cause any mistake. It is likely that the mistake stemmed from a mistranscription or was committed by the computers who truncated fractions too much in their calculations. Make a thorough investigation and submit a report to me.
–Emperor Kang-Xi (1654–1722)

What else can you say when the data don’t bear out your prediction?

During the Qing Dynasty, eclipse predictions were regarded as matters of state security. A false prediction meant punishment–even death–for the predictor. So strongly did the Emperor Kang-Xi heed the portents of solar eclipses that he observed one with a new tool, the telescope, brought by Jesuit missionary astronomers invited to Beijing Observatory.

Alas, imperial interference swayed the astronomers to ignore reality and instead record the authoritarian predictions, even if wrong, as observed fact.

Predicting future climate is even more complex than forecasting future flights of the moon’s shadow across the Earth.

The NASA-GISS climate simulation recently tackled the difficult task of projecting the air’s increased carbon dioxide (CO2) content from human activities to the year 2099. The authors considered several different starting conditions of the climate and various futures with greenhouse gases, sulfate aerosols, or both, added to the air. Some of the runs, they noted, have the familiar error of an artificial drift in surface temperature as large as 0.5°C in the climate without increased CO2 or aerosols, already making results uncertain.

With recent increases in the air’s CO2, how well do predictions, made both globally and regionally, test against observed recent climate change?

Two different starting points exist for the CO2-elevated experiments, whose results are then combined into an “ensemble” result. So the first thing to check is the agreement between the two model results. Averaged over June-July-August from 1960 to 1998, the model trends are inconsistent with each other in the Northern Hemisphere, and strongly opposite in the Southern Hemisphere. Averaging those two conflicting results into an ensemble may cover serious model deficiencies.

Next consider the ensemble. The predicted trend averaged over the Northern Hemisphere is inconsistent with the observed trend. In the Southern Hemisphere, the projected trend is opposite what was observed.

Consider regional disagreements between the ensemble and reality. There are three major discrepancies for the Northern Hemisphere: in Saharan Africa, the North Atlantic, and the Sea of Japan/North Pacific, the observed trends oppose the predicted ones. The projection for the Southern Hemisphere ocean is way off: A large cooling is expected where none is seen. As for this last problem, the authors correctly note the model poorly treats sea ice and ocean circulation.

The next step is to add sulfate aerosols to the air. How does the model do? The NASA-GISS opinion on the role of added aerosols–that “one danger of adding aerosols of unknown strength and location is that they can be tuned to give more accurate comparisons with current observations but cover up model deficiencies”–differs from that of the United Nations Intergovernmental Panel on Climate Change. The authors make that conclusion because adding the effect of sulfates to the increased CO2 magnifies the disagreement with observed trends. Indeed, the answer is not to be found in aerosols.

Discounting the models’ inability to represent the real world, the future climate news seems good. By 2099, if nothing else changes, the high latitudes will continue to warm in the winter and cool in the summer, a benign condition for the populated Northern Hemisphere. The large regional cooling predicted for the North Atlantic arises from a reduced heat transfer poleward in the Atlantic Ocean, delivering the possibility of strong regional cooling despite an overall global warming.

But what about other natural effects? Are forces besides the greenhouse effect changing our climate? Some Texas researchers say yes. Incredibly, they bet on a 15 to 25 percent chance of one large volcanic eruption with a cooling effect of 3 watts per square meter or more in the next 10 years, and a 37 percent chance of an eruption whose effect is greater than 1 watt per square meter. Such events would hide the human climate effects in the modelers’ vision.

Can their prediction be trusted? The researchers merely forward-project the past history of eruptions, given not by physics but by statistics of past “climatically significant” volcanic eruptions. At least it is a testable forecast, so long as we wait 10 years.

Like Emperor Kang-Xi, we may choose to believe–or even insist on–predictions at the expense of reality. But the price of ignoring reality is great. Climate in 2099 could turn out to be very far from what the models say it will be.


Hyde, W.T., and T.J. Crowley, 2000. Probability of future climatically significant volcanic eruptions. Journal of Climate, 13, 1445-1450.

Russell, G.L., et al., 2000. Comparison of model and observed regional temperature changes during the past40 years. Journal of Geophysical Research, 105, 14,891-14,898. See also

Shi, Y., 2000. Eclipse observations made by Jesuit astronomers in China: A reconsideration. Journal for the History of Astronomy, 31, 135-147.