Can you rely on the weather forecast? Maybe not, at least when it comes to global warming predictions over short time periods.
That’s the upshot of a new study in the journal Nature Climate Change that compared 117 climate predictions made in the 1990's to the actual amount of warming. Out of 117 predictions, the study’s author told FoxNews.com, three were roughly accurate and 114 overestimated the amount of warming. On average, the predictions forecasted two times more global warming than actually occurred.
Some scientists say the study shows that climate modelers need to go back to the drawing board.
"It's a real problem ... it shows that there really is something that needs to be fixed in the climate models," climate scientist John Christy, a professor at the University of Alabama in Huntsville, told FoxNews.com.
'I looked at 73 climate models going back to 1979 and every single one predicted more warming than happened in the real world.'
- John Christy, a professor at the University of Alabama in Huntsville
But other scientists say that's making a mountain out of a molehill.
"This is neither surprising nor particularly troubling to me as a climate scientist," Melanie Fitzpatrick, a climate scientist with the Union of Concerned Scientists, told FoxNews.com. "The work of our community is constantly to refine our understanding of the climate system and improve models based on that," she added.
The climate models, Fitzpatrick said, will likely be correct over long periods of time. But there are too many variations in climate to expect models to be accurate over two decades.
But John Christy says that climate models have had this problem going back 35 years, to 1979, the first year for which reliable satellite temperature data exists to compare the predictions to.
"I looked at 73 climate models going back to 1979 and every single one predicted more warming than happened in the real world," Christy said.
Many of the overestimations also made their way into the popular press. In 1989, the Associate Press reported: "Using computer models, researchers concluded that global warming would raise average annual temperatures nationwide 2 degrees by 2010."
But according to NASA, global temperature has increased by less than half that -- about 0.7 degrees Fahrenheit -- from 1989 to 2010.
And in 1972, the Christian Science Monitor reported: "Arctic specialist Bernt Balchen says a general warming trend over the North Pole is melting the polar ice cap and may produce an ice-free Arctic Ocean by the year 2000." That also proved wrong.
But people should still be concerned about global warming, Fitzpatrick says.
"The paper in no way diminishes the extensive body of observations that global warming is happening and that it is largely due to human activity," she added.
"Global surface temperature is still rising ... 2012 was in the top ten warmest years on record. The period 2001-2010 was the warmest on record since instrumental measurements began," she added.
Christy agrees that there has been some warming over time, but says man-made greenhouse gasses are not as big of a driver of climate change as many think -- and that many scientists are in denial about their mistakes.
"I think in one sense the climate establishment is embarrassed by this, and so they're trying to minimize the problem," he said. "The fundamental thing a climate model is supposed to predict is temperature. And yet it gets that wrong."
The study authors did not answer questions from FoxNews.com about the policy implications of their research.
Why were the predictions off? The study authors list many possible reasons, from solar irradiation and incorrect assumptions about the number of volcanic eruptions to bad estimates about how CO2 effects cloud patterns.
Christy said he believes the models overestimate warming because of the way they handle clouds.
“Most models assume that clouds shrink when there is CO2 warming, and that lets in more sun, and that's what heats up the planet – not so much the direct effect of CO2, but the ‘feedback effect’ of having fewer clouds. In the real world, though, the clouds aren't shrinking,” he said.
The study also says that an overestimate of the power of CO2 as a greenhouse gas could be why the models over-predict, but that they do not know why the models are wrong at this point.
Christy said he is not optimistic about the models being fixed.
"The Earth system is just too complex to be represented in current climate models. I don’t think they’ll get it right for a long time."