Skip to main content

How low will it go? Scientists to use improved forecast system to predict Arctic ice melt prediction – University of Reading

Show access keys

How low will it go? Scientists to use improved forecast system to predict Arctic ice melt prediction

Release Date 14 March 2017

Thinner ice in the Beaufort Sea past the end of the melt season in the Arctic

Scientists are using a new, more accurate forecast system to predict whether Arctic sea ice will fall to a record low level this summer.

The Arctic ice extent is already at its lowest winter level ever recorded, having shrunk over several decades as the Earth warms, ahead of the big melt that takes place every spring and summer.

Scientists have always struggled to estimate exactly how much ice they expect to have disappeared by the end of summer, but a new forecasting model based on satellite measurements will allow improved predictions to be made much earlier. The model was developed by NASA, and scientists from the University of Reading are involved in the study, published on February 27 in the journal Earth’s Future.

Dr David Schröder, in Reading’s Meteorology department and one of the study’s authors, said: “Our work will allow us to accurately predict the summer minimum levels of Arctic ice in April, rather than June as is currently the case. The new forecast model takes into account far more data, such as ice concentration, and focuses on key locations that offer clues to the bigger picture.

“But the proof is in the pudding. We have already put our theory to the test using historic data, but this will be the first time it has been used to predict future melting so far ahead. The timing is apt, too, as our forecast should give a strong indication of whether we are about to see a record low summer ice extent this September.”

An accurate forecast of the summer minimum level of Arctic sea ice is valuable for shipping companies and native people that depend on sea ice for hunting. Many animal and plant species are impacted directly by changes in the coverage of sea ice across the Arctic.

However, uncertain weather conditions through spring and summer make the forecasting of Arctic sea ice for a given year extremely challenging.

Using data from satellites, which have been measuring sea ice in the Arctic since 1979, scientists can easily calculate the downward trend in Arctic sea ice.

To make forecasts of how the Arctic sea ice cover might behave in the upcoming year, researchers have several options. The simplest approach is to assume a continuation of the long-term trend into the current year. The problem with this approach is that it will miss outliers - years when the sea ice cover will be a lot higher or lower than expected.

The new study is analysing the physical characteristics of the sea ice cover as the melt season develops, to try to more precisely estimate if the amount of sea ice come September will be more or less than expected from the long-term trend.

Back in time

In the study, the team found that the forecasts based on melt onset – the time at which sea ice starts to melt and open water appears in the Arctic Ocean – were most reliable in early spring, while sea ice coverage-based predictions were more reliable from June onwards.

The forecasts focus specifically on regions that historically corresponded with how much sea ice remains come the September minimum extent. The predictions become more accurate with each passing month, as the model integrates more near-real-time information about sea ice melt and the distribution of open water areas across the Arctic Ocean and surrounding seas.

To test whether their model produced reliable forecasts, Petty’s team went back in time and made predictions for each year of the satellite record, using historical data of the Arctic sea ice conditions. They then evaluated the results against both the actual minimum extent for that year and what the long-term trend would have predicted.

Dr Alek Petty, lead author of the new paper, and a sea ice researcher at NASA's Goddard Space Flight Center in Maryland, USA, said: “We found that our forecast model does much better than the linear trend at capturing what actually happened to the sea ice in any specific year. Our model is very good at catching the highs and the lows.

“The absolute values? Not exactly, but it tends to do very well at seeing when the sea ice extent is going to go up and when it’s going to go down compared to what we might be expecting for that year.”

The research also showed that models can produce reliable forecasts of sea ice not only for the whole Arctic, but for concrete regions; specifically, the Beaufort and Chukchi seas north of Alaska.

“The state of sea ice has a large impact on the Alaskan hunting communities,” Dr Petty said. “If they know ahead of time what the sea ice cover is going to be like that year, they might be able to infer the availability of the species they hunt.”

Future research will explore synthesising different sea ice measurements into the same model to improve the reliability of the forecasts.


Full reference:


A. Petty, D. Schröder, J. Stroeve, T. Markus, J. Miller, N. Kurtz, D. Feltham, D. Flocco (2017). ‘Skillful spring forecasts of September Arctic sea ice extent using passive microwave sea ice observations.’ Earth’s Future. DOI: 10.1002/2016EF000495


Photo credit: Alek Petty/NASA's Goddard Space Flight Center

We use Javascript to improve your experience on, but it looks like yours is turned off. Everything will still work, but it is even more beautiful with Javascript in action. Find out more about why and how to turn it back on here.
We also use cookies to improve your time on the site, for more information please see our cookie policy.