by Judith Curry
The seasonal forecasts of Arctic sea ice minimum have been submitted to annual SEARCH Sea Ice Outlook
For background on the decline of Arctic sea ice, see my summary post at Climate Dialogue (and Climate Etc. references therein).
The relative recovery of the Arctic sea ice last September has pretty much put the kibosh on forecasts of an imminent ‘spiral of death’ for the Arctic sea ice.
Melt ponds as a forecast tool
One of the forecasts, by a group at the University of Reading, is getting a lot of press. They recently coauthored a paper entitled September Arctic sea-ice minimum predicted by spring melt-pond fraction. Co-author Tsemados explained this in a guest post at Ed Hawkins’ blog. Basically, they have put a prognostic model for melt ponds into the CICE model (the most sophisticated of the sea ice models used in climate models). They found find that the Arctic sea-ice minimum can be accurately forecasted from melt-pond area in spring with a strong correlation between the spring pond fraction and September sea-ice extent.
Well I find this to be pretty interesting. As the grandmother of melt pond research (well co-grandmother with Beth Ebert), I have done considerable research on this topic, including developing the first parameterization of melt ponds for climate models.
So, what are melt ponds? Well this is what they look like:
As snow and ice melts, the melt water accumulates on the surface in valleys of the sea ice topography. The area and depth of the ponds depends on sea ice age (internal structure as well as surface topography), and ponds that melt completely through are called melt holes. As can be seen from the photo, melt ponds have a much lower albedo than ice, and hence enhance the melting of the sea ice in a positive feedback loop.
The onset of the melt season, in late spring, is driven in a general sense by the local increase in solar insolation, but also by weather systems originating in the midlatitudes. The first rainfall of the season causes metamorphosis of the snow grains, whereby the snow becomes wet and individual snow grains become larger, which makes the snow less reflective. This decrease in the snow reflectivity accelerates the snow melt. Depending on the age of the ice (newer ice has more salt and hence melts at a slightly lower temperature), the ice begins melting once the snow is gone.
In late summer, the ponds freeze over, triggered in a general sense by local decrease in solar insolation, but also by weather systems from the midlatitudes. Snowfall and high winds can help initiate the melt pond freeze up (note prognostic freeze up of the melt ponds is very tricky to get right).
Curent state of the sea ice
Last winter, sea ice extent was on the low side, but the melting trajectory looks pretty average for the decade of the 2000′s. What is more telling is this figure of the air temperature at 80N:
Winter temperatures were anomalously warm – the cold air kept moving southward without building up for extensive periods over the Arctic Ocean. Now since May, temperatures have been anomalously cool.
With regards to sea ice volume, there has been some recovery relative to the values of the last 3 years:
JC’s take on seasonal sea ice forecasting
There are three things to look at:
- The evolution of the multidecadal oscillations (e.g. where you are at on the stadium wave wheel)
- Seasonal ENSO forecasts
Re #1, lets take a look at the long term anomalies:
On a (2011) Climate Etc. post Pondering the Arctic Ocean, I interpreted the record in the context of a (qualitative) change point analysis, defined by changes in trend, mean value, amplitude of the annual cycle, and interannual variability.It looks like 2013 was another change point year, characterized by low amplitude seasonal cycle. Stay tuned for a a more detailed interpretation of all this in context of the stadium wave. Key question: does the mini shift in 2013 portend flat (no trend), low amplitude anomalies, or possibly a recovery? Or continued decline?
Re #2, see a paper I coauthored Recent Arctic sea ice variability: connections to the Arctic Oscillation and the ENSO. The analysis needs updating, but the punchline is that there is not much effect of these Arctic wide (although some big regional effects). Update on the 2014 El Nino watch: the predicted El Nino is somewhat fizzling, unlikely to see very strong El Nino at least through the end of summer. In any event, the seasonal predictability of ENSO starting in June provides some regional predictability of sea ice variations.
Re #3, The meltpond predictor seems useful to some extent; a slow start to the melt season implies a shorter melt season. However, without having dug into the details of what they have done, I would caution that they need to keep track of what is going on with first year ice versus older ice. The Arctic Oscillation and North Atlantic Oscillation influence substantially the drift and deformation of the sea ice, which is of equal importance in the seasonal evolution of sea ice ice extent. There is some predictability of NAO and AO out to 30 days; CFAN’s 30 day forecasts of NAO, and AO are pretty flat (i.e. not far off neutral).
So how does all this add up? I would say that the Reading team should be fairly close – similar to last year. I disagree slightly with their reason for this, focusing more on the large-scale climate dynamics regime and change points, but I agree that the late onset to the melt season should be a contributor. I am definitely not placing any money on a spiral of death scenario.
But there are many wild cards associated with the weather, and even high latitude forest fires can play a role. It will be fun to see how the SEARCH forecasts come in, and how this plays out.