A fifth of Canada’s Arctic glaciers could have melted away by the end of the century, pushing up global sea levels by 3.5cm.

Scientists made the prediction using a new climate model which simulated the shrinking of glaciers in the archipelago.

The model accurately reproduced levels of ice loss from the island group over the past 10 years.

Looking forward, it indicated that the glaciers would continue melting at an accelerating rate – and it may already be too late to reverse the trend.

One scenario showed 20 per cent of the glaciers’ ice mass disappearing by 2100. This assumed an average global temperature rise of 3C by the end of the century, half the most pessimistic international forecast.

Lead researcher Jan Lenaerts, from Utrecht University in the Netherlands, said: “Even if we assume that global warming is not happening quite so fast, it is still highly likely that the ice is going to melt at an alarming rate. The chances of it growing back are very slim.”

Global climate predictions mask local variability. If the world warms by 3C, the temperature around the Canadian ice caps is expected to rise by 8C. Should the glaciers melt completely, global average sea levels will rise by 20cm.

The new findings are published in the journal Geophysical Research Letters.

Co-author Michiel van den Broeke said: “Most attention goes out to Greenland and Antarctica which is understandable because they are the two largest ice bodies in the world. However, with this research we want to show that the Canadian ice caps should be included in the calculations.”

British expert David Vaughan, programme leader of ice2sea, the EU-funded project said: “The Canadian archipelago is an area where climate is changing rapidly, and the glaciers here contain enough ice that we should not ignore their contribution to sea-level rise.

“Added to glaciers in Alaska, the Russian Arctic and Patagonia, these apparently small contributions add up to significant sea-level rise. A key success of this study was in showing that the model performed well in reproducing recently observed changes. That success gives us confidence in how the model predicts future changes.”