Study: Forecasts for Summer Arctic Sea Ice Lack Reliability
Fairbanks, AK - Year-to-year forecasts of summer Arctic Sea Ice extent aren’t reliable. That’s according to a report out from the National Snow and Ice Data Center. But A two-day workshop that starts Tuesday in Colorado will focus on ways to improve sea ice extent predictions.
Every year various groups set out to predict summer Arctic sea ice extent. The information is useful for ship navigators, biologists who study marine mammals and scientists who consider sea ice a sensitive climate change indicator. A new study finds that the forecasts aren’t always reliable. “The wildcard really still is the summer weather patterns.” Julienne Stroeve is a Senior Research Scientist at the National Snow and Ice Data Center. She and colleagues looked at more than 300 forecasts from the last six years. She says improved summer weather predictions as well as satellite measurements of sea ice thickness and concentration could help forecasting. “We don’t predict the summer weather yet and because of that the sea ice is still sensitive to what happens in the summer time which makes these predictions difficult during those anomalous years,” says Stroeve.
Arctic sea ice reaches its minimum extent in September. Stroeve says 2012 and 2013 were anomalous years when predictions fail. She says that’s because the extent of the ice strayed from what has otherwise been accepted as a general downward trend since satellites started keeping track in 1979. “It didn’t seem to matter so much as a group what method your were employing to do the sea ice forecasting," explains Stroeve. "So, if you were using a statistical approach to forecast what the September ice extent would be, or if you used a model sophisticated modeling approach where you’re initializing sea ice atmospheric models with boundary conditions of where the ice is and what the atmosphere is, and then run those forward, those didn’t do any better.”
She says when forecasting takes place also doesn’t affect accuracy. “The forecasts for what was going to happen in September also didn’t necessarily get any better if you initialized your forecast in June, July or August and I thought that was curious because you would think as the summer progresses, you update your forecast with the current ice conditions that probably you should do a bit better forecast for what’s going to happen in September."
She says looking back at old forecasts could be helpful for future forecasts. "“You’d call that hindcast model evaluation so, go back in time and say ‘Well, would you have actually predicted the extent right if you had had all the relevant data that you needed, or is there a problem with the forecasting method itself?’”
Stroeve and colleagues are in Boulder this week for a Sea Ice Prediction workshop to discuss how to improve future forecasts.