At noon on January 25, the most sophisticated weather‐forecasting model in human history predicted a total snowfall for Washington, D.C., of somewhat less than an inch in the succeeding 36 hours. All the human forecasters I know went along.
The computer model, named for the Greek letter eta (pronounced “ay‐ta”), which describes its mathematical coordinate system, forecast a storm far out to sea, sparing the I-95 megalopolis that stretches from Richmond to Boston. Instead, eta confined its significant snow to a sliver of southeastern Virginia, before shoving everything north, east and offshore.
Fear in the forecasting community was palpable as minute‐by‐minute updates from our national radar network painted an army of green and yellow monsters marching north and west toward our nation’s capital. By 9 p.m. they were closer than most of the Confederates ever got, and still the forecast was for a minor dustup. It wasn’t until eta, whose major update cycle is 12 hours, was run again, that forecasters decided disaster was at hand.
Why didn’t we believe our eyes when the model was clearly busting? The truth is, as models become more sophisticated, forecasters are increasingly reluctant to abandon them, even in the face of contrary evidence. But, “the computer eta my forecast” is an insufficient excuse.
Would that this were the case merely for the 48‐hour forecast. Unfortunately, it appears that the same pathology has infected the 48‐year projection. Just like weather forecast models, our climate simulations have become increasingly sophisticated in recent years. And just like the situation with the recent snow, there has been incontrovertible and advancing evidence, this time over the course of the last two decades, that climate simulations are making a disastrous error, and it has taken forever for forecasters to acknowledge it.
Every computer model predicts that the entire troposphere, or roughly the bottom 40,000 feet of the atmosphere, should be warming rapidly. Most models even predict that much of the warming accelerates with height.
But, for over 20 years now, we have two independent measures of temperature — satellites and weather balloons — that show no net warming at all from 5,000 feet skyward. In 1996, we had 17 years of satellite data, and yet the “Policymakers Summary” (the only part that gets read) of the U. N. Intergovernmental Panel on Climate Change report contained not one mention of the word “satellite.” That is the report usually cited as the “consensus of scientists.”
If that was the 1996 consensus of my profession, it was largely a consensus of ostriches, not scientists. How long could we continue to sweep under the computer the fact that somehow we got 88 percent of the atmosphere wrong?
Not very long. Here is what the National Research Council now says about our ability to model climatic behavior: “It is clear,” they recently wrote, “that reconciling the discrepancy … is not simply a matter of deciding which [climate model] is correct.… In the long term it will require major advances in our ability to interpret and model [atmospheric behavior].” In other words, our models don’t work. Here, for once, eyeballs have triumphed over the computer.
Most people use weather forecasts to plan the future. The Kyoto Protocol on climate change was a response to a climate forecast. This onerous document would require the United States to drastically reduce the energy use that has propelled us on our gee-doesn’t-your-401 k -look‐great way for the last two decades. Everyone (except, maybe, President Clinton, judging from his State of the Union Address) believes it will cost us a fortune.
When it finally became clear that Washington was going to get buried in the last snowstorm, the federal government shut itself down. Now that we have admitted that our climate forecasts for the next century aren’t worth much at all, can we please shut down the Kyoto Protocol on climate change?