Impacts of man-made greenhouse warming on rainfall would endure long after temperatures fell, a study suggests.
UK Met Office scientists constructed a hypothetical future in which carbon dioxide levels rise and then fall, and modelled what might happen to rainfall.
Their computer simulation showed temperature falling decades after CO2's decline, with changed rainfall going on for several more decades after that.
The study is published in the journal Geophysical Research Letters.
The hypothetical future saw concentrations of carbon dioxide in the atmosphere rising to four times pre-industrial levels over 70 years, then returning to the baseline over a similar period.
"This is an idealised situation," said Vicky Pope, head of climate change advice at the Met Office.
"On the other hand, we could reach atmospheric concentrations of greeenhouse gases equivalent to a quadrupling of CO2 by the end of the century.
"So if we allow emissions to increase to the end of the century and then decrease them rapidly, this is the kind of thing you'd expect to see," she told BBC News.
The computer simulation also suggested that if greenhouse gas levels remained elevated for an extended period, the rainfall changes would then endure for longer after emissions began to decline.
Rainfall changes arise largely through increased temperatures in the ocean.
More water evaporates into the atmosphere. Overall, this means the world would receive more rain - but computer simulations suggest the impacts would be very unevenly distributed.
High latitude countries such as Canada and Russia would receive more rain and snow, whereas other regions such as the Amazon basin, Australia and parts of sub-Saharan Africa would receive substantially less.
As the oceans have huge capacity to store heat, releasing the heat relating to a temporary quadrupling of the man-made greenhouse effect would take many decades.
The Met Office computer model is known to project more drying of the Amazon than most others.
"Details of exactly where the impacts fall would be different in different models because they don't agree in details," said Dr Pope.
"But the mechanism is the same in all models, because it's basic physics. As models improve, we'll be able to get more confident on this."