Over all, the world would experience profound transformations, some potentially beneficial but many disruptive, and all at a pace rarely seen in nature, said the authors of the study, being published today in The Journal of Climate.

"The question is no longer whether we will need to address this problem, but when we will need to address the problem," said Kenneth Caldeira, an author of the study and a climate expert at the Carnegie Institution's Department of Global Ecology, based at Stanford University.

"We can either address it now, before we severely and irreversibly damage our climate, or we can wait until irreversible damage manifests itself strongly," Dr. Caldeira said. "If all we do is try to adapt, things will get worse and worse."

The paper's lead author, Bala Govindasamy of the Energy Department's Lawrence Livermore National Laboratory, said it might take 20 or 30 years before the scope of the human-caused changes becomes evident, but from then on there is likely to be no debate.

The researchers ran a computer model that simulates both the climate system and the flow of heat-trapping carbon into the air in the form of carbon dioxide, then back into soils and the ocean.

Most simulations of the potential human impact on climate have been confined to studying the next 100 years or so, but in this case the scientists started the calculations in 1870 and let the computers churn away through 2300.

The authors stressed that the uncertainties were high over such a time span, and said the study was intended to illustrate broad consequences rather than project specific ones.

They programmed the model to run as if the atmospheric concentration of carbon dioxide rose about 0.45 percent a year through 2300. That is slightly less than the current rate, about 0.5 percent.

In the simulation, the concentration of carbon dioxide doubles from pre-industrial levels in 2070, triples in 2120, and quadruples in 2160.

The results are sobering, Dr. Caldeira and other climate experts said, because the computer model used in this study tends to produce less warming from a greenhouse-gas buildup than many of the other climate simulations being run by other research teams.

It also presumes that plants and the ocean will continue to sop up carbon dioxide in the future, limiting the amount retained in the atmosphere. Many other independently developed models calculate that at some point, chemical and biological shifts caused by warming would reverse that flow and cause even more greenhouse gases to flood into the atmosphere.

Consistent with many other studies, the model showed that the Arctic would see the most warming, with average annual temperatures in many parts of Arctic Russia and northern North America rising more than 25 degrees Fahrenheit around 2100.

Antarctica would follow suit later, with temperatures there rising sharply around 2200.

The impact on vegetation and landscapes would transform large areas of the earth.

In the simulation, at least one ecosystem, the scrubby Arctic tundra largely vanishes as climate zones shift hundreds of miles north. Tundra would decline from about 8 percent of the world's land area to 1.8 percent.

Alaska, in the model, loses almost all of its evergreen boreal forests and becomes a largely temperate state.

But vast stretches of land that were once locked beneath permanent ice cover would open up. The area locked beneath ice would diminish to 4.8 percent of the planet's total land area, from 13.3 percent.

Several climate scientists not associated with the study said its main benefit was akin to the murky visions of possible futures experienced by Ebenezer Scrooge in "A Christmas Carol."

"It's a cautionary tale," said Gerald A. Meehl, a climate modeler at the National Center for Atmospheric Research in Boulder, Colo., who has conducted similar studies.