While not hugely significant by itself, it is interesting to note that the DMI 30% Arctic extent has reached its highest number for this date, exceeding 2006. The refreeze has been very fast:
Here’s the zoom:
The JAXA 15% plot show it equal with 2006, and a steepening slope:
JAXA AMSR-E Sea Ice Extent -15% or greater – click to enlarge
Scott,
Here is a link to my previous post;
http://wattsupwiththat.com/2010/10/04/sea-ice-news-25-nsidc-says-2010-3rd-lowest-for-arctic-sea-ice/#comment-500490
and a link to my correction of (2);
http://wattsupwiththat.com/2010/10/04/sea-ice-news-25-nsidc-says-2010-3rd-lowest-for-arctic-sea-ice/#comment-500642
So, one basic question is;
Can we mix and match between the six known datasets: JAXA (15%), NSIDC (15%, monthly, or daily from chart), Bremen (15%, daily from chart), UIUC (100%, areas), Norsex (15%/100%), and DMI (30%)?
The answer seems to be yes, maybe, and no.
For example, I’ve captured the daily values from NSIDC (15% from graph), JAXA (15% digital), and Bremen (15% from graph) for the past 52 days and produced three graphs;
http://picasaweb.google.com/117077348819630829996/ArcticSeaIce#
From each original raw time series the means and standard deviations (10E6 km^2 for both) are;
NSIDC, Bremen, JAXA, Bremen – NSIDC, JAXA – NSIDC
5.295, 5.338, 5.468, 0.043, 0.173
0.638, 0.613, 0.576, -0.025, -0.062
NSIDC is the lowest, followed by Bremen, then JAXA.
The three graphs (in order) are; (1) Raw time series, (2) demeaned time series with the common mean from all three added back in, and (3) all three combined into an ensemble dataset, with a 5-day composite moving average added in (NSIDC as is, JAXA with a 0.5,1,1,1,0.5 weighting, and Bremen with a 0.2,0.2,0.2,0.2,0.2 weighting).
This appears to work reasonably well for short duration time series, but may not hold true if several years of data were used as the differences in standard deviation, min, max, mean, and median may have some temporal seasonal differences.
And I’ll leave it just at that point for possible discussion.
In my next post, I’ll compare NSIDC and JAXA monthly values for their common time period of 2002-2010.
EFS_Junior says:
October 16, 2010 at 1:06 pm
Looks like you’ve put a lot of work into this. One question – where do you get your daily values for NSIDC (I have only monthly averages) and Bremen (I only see images here: http://www.iup.uni-bremen.de:8084/amsr/)?
Also, why do you weight the 5-day average differently for JAXA and Bremen?
-Scott
Scott says:
October 16, 2010 at 2:43 pm
EFS_Junior says:
October 16, 2010 at 1:06 pm
Looks like you’ve put a lot of work into this. One question – where do you get your daily values for NSIDC (I have only monthly averages) and Bremen (I only see images here: http://www.iup.uni-bremen.de:8084/amsr/)?
Also, why do you weight the 5-day average differently for JAXA and Bremen?
-Scott
_____________________________________________________________
I’m retired.
I’ve always done a lot of data analyses, it keeps me occupied, call it a life long hobby.
The NSIDC and Bremen dailies are pulled straight from their daily graphs.
NSIDC is fairly easy to read, but you must always check the previous two days and adjust accordingly due to their 5-day moving average.
I scale from two known y-axis values from their graph, and believe that I can read their daily value to a quarter pixel, assuming that I’m within half a pixel of the true value, gives me an error estimate of ~10K km^2, based on the pixel resolution.
NSIDC is stated to be a 5-day moving average, so I wanted to adjust the other two datasets to the same 5-day NSIDC moving average, thus NSIDC is given a weight of one for that current daily average.
JAXA is a 2-day moving average, using current and previous days. I went with 0.5,1,1,1,0.5 to get their moving average. But there is no “right” way to weigh JAXA. For example, since JAXA uses current/previous days, I could backshift the date axis by 0.5, and go with the following 6-day centered moving average, 0.5,1,1,1,1,0.5.
Bremen, to be honest, I’d have to look into their description of their graphical time series, I have assumed that it is a 1-day value, as their graph appears to be more jagged (it also doesn’t help that their current year is ~6 pixels in width (other years are ~3 pixels in width) and it can be fairly hard to see when it zigs when it actually zags), as the previous days are never updated (I’ve flipped through a series and only new pixels are added for the newest date), and their current date on their graph is today’s date (e. g. today’s graph is labeled “Version 2010.10.16”), so I’ve backdated their entire timeline by one day.
My error estimate for Bremen is one pixel or ~20K km^2.
Lots of details, I know.
I’d really like to see more transparancy from all sea ice data websites, just like JAXA and UIUC, daily numerical values.
For example, you can find NSIDC daily’s up through 2007, but after that, nothing that I’ve been able to find, and I’m not up to downloading their gridded dataset, as I don’t know if I’d be able to reproduce their time series as is.
Oops, I forgot to mention that my assumption of one day for Bremen, means that I use a “boxcar” moving average (e. g. 5-day moving average is 1/5 for each day’s sea ice value, or 0.2,0.2,0.2,0.2,0.2).
Oops Part Deux!
For JAXA 5-day window it’s divided by the sum of the individual weights, in that case 4.
For JAXA 6-day window it’s divided by the sum of the individual weights, in that case 5.
All weighting is divided by the sum of the individual weights, NSIDC, Bremen, and JAXA are summed (weight =1), then the total is divided by N, in that case N = 3.
So, your point being what, Mr. Watt? This “fast refreeze” from the 3rd lowest ice extent recorded will trend exactly how? Nice sleight of hand. It does have the desired effect on your adoring followers, but the reality of the situation is…….
The trend for Arctic sea ice over the past 30 years of satellite scrutiny and possibly longer (we’ll see what Tamino comes up with), has been going down.
David W. Walters says:
October 19, 2010 at 6:39 pm
Nice trick yourself to post so late on a thread no one was posting on anymore.
Anyway, Anthony never said that the trend wasn’t going down, and he even started off the post with saying it was likely insignificant. Why can’t people (e.g. Tamino) understand this?
As for the downward trend, want to guess how many consecutive years of record high summer minimum ice areas, starting in 2011, it would take to make the long-term trend positive? Try 12 (assuming they just barely beat the previous record). I would think that two in a row would indicate a very strong recovery, and almost a complete one (implies large amounts of thick ice by that point), so looking for the long-term trend to reverse is a silly argument, and anyone plotting a line for a fit of a sine wave when it starts at a value of 1 can tell you that. A positive trend line would imply that the ice has more than recovered and that the recovery began long before then.
-Scott
Scott, just stumbled upon this, sorry i didn’t post sooner.
David,
Sorry for the aggressive post and misunderstanding.
I just saw your post and assumed you were just trying to score a cheap shot. If you’re interested, Tamino did post on the pre-satellite era here:
http://tamino.wordpress.com/2010/10/16/history-of-arctic-and-antarctic-sea-ice-part-1/
-Scott