The problem of climate change models....

Our climate change models are still very, very primitive...

Two years ago, a United Nations scientific panel won the Nobel Peace Prize after concluding that global warming is "unequivocal" and is "very likely" caused by man.

Then came a development unforeseen by the U.N.'s Intergovernmental Panel on Climate Change, or IPCC: Data suggested that Earth's temperature was beginning to drop.

That has reignited debate over what has become scientific consensus: that climate change is due not to nature, but to humans burning fossil fuels. Scientists who don't believe in man-made global warming cite the cooling as evidence for their case. Those who do believe in man-made warming dismiss the cooling as a blip triggered by fleeting changes in ocean currents; they predict greenhouse gases will produce rising temperatures again soon.

The reality is more complex. A few years of cooling doesn't mean that people aren't heating up the planet over the long term. But the cooling wasn't predicted by all the computer models that underlie climate science. That has led to one point of agreement: The models are imperfect.

"There is a lot of room for improvement" in the models, says Mojib Latif, a climate scientist in Germany and co-author of a paper predicting the planet will cool for perhaps a decade before starting to warm again -- a long-term trend he attributes to greenhouse-gas emissions. "You need to know what you can believe and can't believe from the models."

The renewed discussion of inherent shortcomings in climate models comes on the cusp of potentially big financial commitments. In five weeks, diplomats from around the world will meet in Copenhagen to try to hash out a new agreement to curb global greenhouse-gas emissions. The science continues to evolve.

The goal of climate models is to project how rising greenhouse-gas emissions will interact with natural forces to affect the global temperature. The models are technological marvels. Using supercomputers, they divide the world into grids of roughly 4,000 cubic miles apiece. The grids are stacked, one on top of the other, up through the atmosphere.

It is complicated stuff. The models consist of dozens of equations written to reflect how liquids and gases move about the planet. Just as a symphony's sound is affected by the crash of symbols or the pluck of a violin string, the planet's future temperature is influenced by powerful ocean currents and tiny specks of sea salt. In between are other players, such as sunlight, clouds and rain.

Added to the equations are such measurements as past temperatures, barometric pressure and sea salinity. Calculations about the influence of sunlight are entered. Then various projections of greenhouse-gas emissions are factored in. The computers run the equations and generate projections of global temperatures.

The models are only as good as the information they are fed. One big uncertainty is ocean temperature. Oceans trap huge amounts of heat, and they process by which they release it over time affects the temperature of the planet. But there isn't a lot of actual data, because the vastness of the oceans makes gathering temperature data costly and arduous.

The success of the models also depends on the soundness of their assumptions. The effects of clouds, for example, are unclear. Depending on their shape and altitude, clouds can either trap heat, warming the earth, or reflect it, cooling the planet. The way that greenhouse gases affect cloud formation -- and how clouds in turn affect temperature -- remains a subject of debate. Different models treat these factors differently.