Interesting analysis of IARC-JAXA Arctic sea ice data

Interesting analysis of IARC-JAXA Arctic sea ice data

http://wattsupwiththat.com/

Analysis of the AMSR-E data on Arctic Ice

Guest post by Dr Tony Berry, 4^th October 2009

The latest value : 5,824,688 km2 (October 5, 2009)

 

The Arctic Ice extent prepared IARC-JAXA demonstrates the cyclical annual trend of freezing and thawing in the Arctic. It is clear that the cycles vary from year to year and it has been suggested that the data supports the hypothesis that AGW is responsible for a long term decline in Arctic ice although recent data has shown a recovery in the Arctic ice extent.

The raw data provide daily IARC-JAXA shows trend nicely but it is difficult to carry out comparative analysis and examine long term trends. Therefore I have calculated average monthly ice extent for each complete month from July 2002 and used this to prepare rolling averages for the data. NOTE: there are gaps in the daily data of up to 11days especially during 2002 to 2004. In these cases the gaps have been filled in using linear interpolation; I do not believe that this has affected the monthly averages in any significant way. The resulting data is as follows:-

For all graphs and tables, click them to obtain larger images.

These data contain a number of interesting observations. All the lowest ice extent figures are contained in three clusters: January-June 2006, November and December 2006 and July to September 2007. By contrast all the highest figures, bar one, are found in 2002/2003. Looking at the monthly averages over the whole period it is remarkable how little variation (Std.Dev<3%) there is apart from the period July to October. This is illustrated by the following graph

Using the above data to compute monthly 12 month rolling averages shows some very interesting trends. Considering just the period July 2003 (the first 12 month average) to September 2007 this shows the following trend:-

It is apparent that there appears to be a strong correlation between the average ice extent and time as illustrated by the high correlation coefficient (0.9232).  You might have concluded in September 2007 that this was indeed strong evidence that long term warming was taking place and you might also be concerned that it appeared to accelerating in the later months. This is just the conclusion that our AGW “friends” have reached assuming that these data can be extrapolated. However,  when you consider the later data you might change your mind as show by the next graph covering the period September 2007 to September 2009 which shows the following trend:-

These data show an entirely different picture. Whilst there is still a strong trend between the average ice extent and time (correlation coefficient 0.9185) it is precisely in the opposite direction!

These two graphs illustrate the folly of assuming that correlations are a proxy for understanding the underlying science and can be used to make predictions of the future. In fact these sorts of correlations are useful in understanding what has happened in the past and might be used as a starting point to identify the science but have little or no value in making long term projections. You can’t drive safely down the highway by looking in the rear view mirror!

The trend over all the data is as follows:-

The fitted polynomial is of no significance other than it fits the data best with a good correlation coefficient (0.8982). The curve, as one might expect, does predict the period, by differentiation, when the lowest recorded ice extent occurred – between the 16^th and 18^th of September 2007. This picture is also carried forward in a two year rolling average:-

Again I don’t believe that the polynomial itself is of any significance other than to illustrate trends. However, this analysis does highlight that there would appear to be something unusual about the period January 2006 and October 2007 when all the lowest ice extents occurred. This has quite a large effect on the graphical data which shows much larger swings during this period. It is also interesting that the graphical data seems to show short term cycles which is particularly apparent again over the period 2006-2007.This latter phenomenon might be an artifact of the analysis or the data but might be worth further investigation.