August 29, 2012

How low is low?

Data graph showing sea ice extent Sea ice extent reached a new record low on August 27, 2012 and continued to decline. The last six years have seen minimum sea ice extents below the two standard deviation range of the data. The graph above shows Arctic sea ice extent as of August 13, 2012, along with daily ice extent data for the previous five years. 2012 is shown in blue, 2011 in orange, 2010 in pink, 2009 in navy, 2008 in purple, and 2007 in green. The gray area around the average line shows the two standard deviation range of the data.

Satellite observations since 1979 show that sea ice melted to its lowest extent in the satellite record, during August 2012 . As of this post date, the ice continues to melt, with two to three weeks left before the days shorten enough for the ice extent to begin to expand through the winter. Readers often write to us asking what such records really mean. How far from normal is this year’s record low, and how do scientists decide what is normal?

Scientists begin to answer this question by looking at the longest, most consistent time series of data that are available. In the case of sea ice, that means data from passive microwave sensors on a series of satellites, which have been collecting data on sea ice extent since 1979. They know that in any given year, variable weather conditions can result in more or less sea ice. To understand climate, which is a more persistent, average set of conditions, looking at trends over many years tells the story beyond the short-term ups and downs of particular years.

How do you define normal?

Statistically, the average Arctic sea ice extent in September is a single value. But how much variation occurs close to that average? A statistical method called standard deviation helps show how closely most of the data is clustered around the average. Using this method, scientists can see a sort of “average of the average,” a wider set of values that might still be considered typical, or within the range of natural variability.

NSIDC calculates two standard deviations for sea ice extent data, which means that 95% of the sea ice extent values will fall into this range. The two standard deviation range is considered, in statistical terms, to be where you would expect to find most of the data. The 5% of values that fall outside of this range are considered outliers, values that deviate most markedly from the rest of the data. Looking at the data this way helps bring some perspective. As an analogy, swimmer Michael Phelps has won a total of 18 Olympic gold medals in his career, as of the 2012 London Olympics. Is that a lot of medals, or just a few more than other top athletes? In comparison, the next most winning athlete won 8 gold medals, and the average number of gold medals for athletes who have won 3 or more medals is just 3.7 medals. The two standard deviation range is plus or minus 2.5 medals; so up to 6.2 medals is within the typical range. Clearly Phelps is an outlier.

In the case of sea ice, a graph of sea ice extent shows that during spring, sea ice extent was in the two standard deviation range for sea ice from 1979 to 2000, as shown by the gray shaded area of the graph. But once the melt season began, extent quickly dropped out of this statistically normal zone and plummeted to set a new record low. But was this just an extreme year probably not to be repeated soon, like Phelps’ medals? A graph of the last six years shows that the minimum sea ice extent for 2007, 2008, 2009, 2010, 2011, and 2012 were all far below the two standard deviations, suggesting that what was once normal may no longer be normal.