We recently looked at trends of sea ice extent (and even forecast next year’s value) using not just the extent itself, but the latitude of the sea ice edge. To compute that, I used a 5th-degree polyonomial approximation from the original paper by Ian Eisenman (who identified the importance of geometry on Arctic sea ice).
It was pointed out that the figures computed by a different method — by actually averaging the latitude of the ice-ocean boundary based on gridded sea ice concentration — differed from the values computed from the polynomial approximation. This is to be expected, since the approximation is the value for ice which is re-arranged to be as far north as possible, whereas for actual sea ice the detailed distribution can strongly impact the average ice-ocean boundary latitude. But the difference seemed larger than expected.
So, I emailed Ian Eisenman and asked whether he could shed light on the matter. He was kind enough to respond in detail, saying this:
In the GRL paper, I focused on a measure of the ice edge latitude that was based on the average location of the contour line representing the transition from sea ice to ice-free ocean. This was computed from the full gridded daily ice concentration fields. In the Auxiliary Material, I briefly described an alternative way to account for continents based on a slightly different (and more subtle) physical justification that involves “rearranging” the ice cover. The alternative measure is perhaps more crude physically, but its advantage is that it can be computed directly from the ice extent, and I gave an approximate polynomial expression to allow others to easily compute this measure of the ice edge latitude from a time series of ice extent. The two measures – the more onerous one used in the paper and the quick rough one described at the end of the Auxiliary Material – produce different quantitative values, but both appear to do a sufficient job of accounting for the influence of coastline geography that the trend in ice edge latitude for each month (trend for 30 Januaries, etc) doesn’t have substantial seasonal structure for either measure (in contrast to ice extent).
It makes sense to me. It also confirms what I had already found, that the main result — indeed the main purpose of investigating ice edge latitude — was unchanged, namely: that most of the seasonal differences in Arctic sea ice trends are a geometric effect. When viewed in terms of ice edge latitude (whether computed from detailed gridded ice fields of the approximation formula), the trends in sea ice loss from one month to another are much more similar.