OSLO (Reuters) - Scientists are trying to improve predictions about the impact of global warming this century by pooling estimates about the risk of floods or desertification.

"We feel certain about some of the aspects of future climate change, like that it is going to get warmer," said Matthew Collins of the British Met Office. "But on many of the details it's very difficult to say."

"The way we can deal with this is a new technique of expressing the predictions in terms of probabilities," Collins told Reuters of climate research published in the journal Philosophical Transactions of the Royal Society A.

Scientists in the U.N. climate panel, for instance, rely on several complex computer models to forecast the impacts of warming this century, ranging from changing rainfall patterns over Africa to rising global sea levels.

But these have flaws because of a lack of understanding about how clouds form, for instance, or how Antarctica's ice will react to less cold. And reliable temperature records in most nations stretch back only about 150 years.

Under new techniques looking at probabilities, "predictions from different models are pooled to produces estimates of future climate change, together with their associated uncertainties", the Royal Society said in a statement.

The approach might help quantify risks for a construction firm building homes in a flood-prone valley, for instance, or an insurance company wanting to work out what premiums to charge.

Collins said uncertainties include how natural disasters out of human control affect the climate. A volcanic eruption, such of Mount Pinatubo in the Philippines in 1991, can temporarily cool the earth because the dust blocks sunlight.

"Climate science is a very new science and we have only just begun to explore the uncertainties," said David Stainforth of Oxford University in England who contributed research to the Royal Society.

"We should expect the uncertainty to increase rather than decrease" in coming years as scientists work to understand the climate, he said. That would complicate the chances of assigning probabilities.

As an example, he said designers of schools in Europe wanted to know if there would be more heatwaves like one in 2003 when children were sometimes barred from playing outside because of the risks of sunburn and possible skin cancers.

If so, they might design schools with a lot of shaded outdoor play areas.

"But it might be the case that warmer temperatures mean more cloudiness, so then you won't get the risk of skin cancers," Stainforth told Reuters. "Non-temperature factors are the hardest to predict."