In the first post of your thread on your calculations you posted an image showing NSIDC/IJIS/ and your work matching up, with offsets. So the summer peaking of IJIS/CT Area (inverse shown in your graph above) must be due to CT Area, not some peculiarity in Extent (the peaking is implied by monthly average NSIDC extent).
Yes, not that I wouldn't like to know all those detailed differences, extent seems rather well defined.
Yet again, the peaking seems to be telling us something about sea ice state, as August NSIDC Extent/CT Area shows an upward trend with ice recession, and a jump at 2007. And that's despite the alternate implication that CT Area is a poor metric.
Let me state again, CTArea may well be a useful index, but still be a bad metric for area. Except that my compactness metric looks a whole lot more believable than CAPIE, I am not claiming anything about the usefulness or correctness of the AMSR2-area series.
So if in this thread the question is raised "why is CTArea so high?", the correct answer does might not be that the area is higher.
Knowing the right reason is in my opinion a crucial part of using the CTArea index.
What puzzles me is that people are telling me your data shows we're in for a crash this year, that is not supported by CT Area.
I don't know what you call a crash. But what I agree with some people is that 2013 may still end up very close or below 2012. If you care to look at my regional charts it is clear that two things are required. First Beaufort needs to melt the coming 1.5 month at a rate comparable to last year one month earlier. Second the CAB will need to repeat exactly what it did last year, loosing 1M5 km2 in about one month, starting at the end of July. That is a crash, but the same as last year.
Whereas in 2012 and 2011 CT Area was indicating a new record (2011 late summer tracked 2007 very closely even including a parallel anomaly climb before the minimum - CT Area). I've noted your comparison of your method and CT Area, it is interesting that CT Area falls away from your calculation from May. This is when the Extent/CT Area curves diverge.
The divergence in May, begin of June is explained by CTArea not using the true grid cell area. It overestimates cells in low latitudes, underestimates high latitudes, is true at 70 degrees. The geography currently explains about 260k of the difference between the CTarea and AMSR2-area.
In the rest of June, the divergence grows somewhat bigger, to about 650k and that has stayed so. Comparing with 2012, the qualitative behavior is the same except that the "June bulge" is a lot bigger but also that much of that has disappeared by the beginning of September.
I know you've processed at least some of 2012, have you processed into August? I suspect from what I read previously you've not got enough years to have a baseline to form anomalies. That may make the following question hard to answer: CT Area 2012 showed level anomalies (climatological losses) until the August cyclone, when anomalies started to fall rapidly. Turning a new record into a dramatic record. Can you tell if this was the case in 2012?
Sorry if you've covered any of this elsewhere, I've been busy recently.
I have AMSR2 data only for 2013. The Uni-Hamburg people seem to have too much high priority work, so I don't ask for the 2012 data.
So I turned my attention to a SSMIS (or SSM/IS) timeline from the very same people, same algorithm. The extent/area results are very close to those of AMSR2 where the times overlap. So I use the SSMIS data for comparing with 2012.
Before 2012, there is only the SSMI (without one S) available (for 1991-2012). Unfortunately calculating area here gives results (for 2012) about 6% lower than SSMIS/AMSR2. So I would need to correct for that, which is the current state of research.
I have adjusted the scales of the area and extent graphs to include the September minimum. They are included and updated in the top post of "home brew" discussion.