As noted in the September Outlook Arctic Ice, NOAA’s Sea Ice Index (SII) typically shows less ice than MASIE from National Ice Center (NIC). SII is a satellite product processed from passive microwave sensors. MASIE (Multisensor Analyzed Sea Ice Extent) adds other sources such as satellite imagery and field observations to produce high resolution ice charts for navigational purposes.

A post in 2016 NOAA Is Losing Arctic Ice showed how discrepancies between the two datasets vary considerably throughout the year, usually lower in SII except for October. Walt Meier directs the SII production and published a study in October 2015 comparing SII and MASIE, also discussed in that post.

In 2016 NOAA upgraded from SII version 1 to version 2, and later to version 2.1. The latest documentation says few datapoints were changed in v2.1, and that anomalies were unchanged. My cursory look seemed to confirm that. However, on closer inspection, there are significant differences between v1 and v2 (which carry over to v2.1). This post describes those differences.

I prepared two spreadsheet arrays for SIIv1 and SIIv2.1 and then a third array to calculate the differences. The graph below shows the results for 2006 to 2015 inclusive, being the years for which datapoints can be compared with MASIE.

It is clear that V2.1 is systematically lower than V1, on average -200k km2. The differences are less than 100k km2 the first four months, then increase May, June, July, before shrinking again in August and September. The big changes come in the last months, especially October. The October correction is not surprising. The comparison by Meier and in my post discussed large SII surpluses over MASIE in October that did not appear credible.

The graph is limited to one decade since that is the period to be compared with MASIE. The spreadsheet shows that the differences are typical of the whole dataset going back to 1979, albeit with considerable variety through the years. The graph below shows the month by month differences for all years through 2015.

As stated before, the average all years difference in green is comparable to the last decade. Differences were calculated by subtracting v1 from v2, since v2 is mostly lower. However, as the Min Diff line shows, v2 was higher for some datapoints, notably in July. The Max Diff shows that some Octobers were changed by as much as 1M km2. The dotted lines show the standard deviation for the average differences, which averaged  +/- 90k km2.

Summary

It is challenging to estimate Arctic ice extents. NOAA is to be commended for recognizing the erroneous October values, and correcting them. Clearly some of that overall diminishing of extents by 200k km2 derives from removing the bogus surpluses.

Those claiming that SII is for certain and MASIE is dubious need to reconsider. MASIE has its own challenges but is reasonably consistent in recent years. Meanwhile SII had to improve its product, resulting in changes to past values in the dataset. While error ranges are not available for these statistics, the standard deviation gives some indication of the variability in the estimates.

Fortunately, it appears that the critical months of March and September have not changed much in the new SII version.  However, it is not encouraging to see SII averages for the last two months -500k km2 below MASIE.  See September Outlook Arctic Ice

It is a good thing that several agencies and methods are involved in the effort to measure and understand Arctic ice dynamics. It is not good to claim certainty for a single record or to ignore the errors that are found along the way. It is wise to remember that measuring anything in the Arctic is difficult.