As I mentioned yesterday, NSIDC had an oops moment, but with the help of skeptical blogs, was notified of the problem and responded timely and appropriately. They posted this update today:
Update, April 19, 2012: The nine-day trailing average climatology on the daily data graph has been changed to a five-day trailing average, to be consistent with the five-day trailing average for the daily data.
I verified their correction for the climatology was accurate with a new overlay, combining the unaffected graph NSIDC’s Dr. Julienne Stroeve sent me from their internal server storage Tuesday night with the corrected one published on the web today:
http://nsidc.org/data/seaice_index/images/daily_images/N_stddev_timeseries.png
The extent data and the climatology now match, whereas yesterday they did not. The x-axis offset is to be expected given that we are comparing graphs with the temporal data offset due to a trailing average they implemented.
This is what the same comparison looked like yesterday, for graphs made on the same day:
So, problem solved.
Unfortunately, somebody jumped to a conclusion and has already had to issue a correction.
April19: NSIDC graph still appears wrong
now reads:
Correction : April 19 – NSIDC Graph Now Lines Up With April 16
The maxim “haste makes waste” seems appropriate.
The way NSIDC’s Dr. Walt Meier and Dr. Julienne Stroeve handled this should be an example to other agencies that don’t bother to even respond to skeptics.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
The current and average lines are getting pretty darn close…Will NSIDC let the streams cross? That would violate GhostBusters Rule #1: Don’t Let the Streams Cross!
Increasing ice sheets are a catastrophe for the Earth’s biomass. Please refrain from referring to growing ice as a recovery. Ice sheets are massive and any decreasing trend is a recovery from our ice age infliction. Increasing ice is a relapse to our hypothermia condition… afflicting us since we came down from the trees. One day it could be fatal. GK
I’m not sure I understand your issue regarding the SD’s. I see the baseline is 1979-2000, not the entire record to date. 2007 isn’t even in this data set so obviously won’t “broaden” the SD boundaries. If there has been a general decline in the ice extent over the instrumental record, wouldn’t one expect that later years would eventually fall beyond the 2 sigma boundary if the base period remains the same? It would be a question of “when” not “if”, depending only on the rate of decline.
Sure, but this is simply cheating with statistics, especially since 1979-2000 is too few years for the standard deviation to mean much anyway. However, my objection can be formulated in perfectly reasonable terms even excluding 2007 from the data. 2007 is smoothly similar to the mean curve representing the average of the baseline years (where one has to ask — why bother plotting 2007 in particular on this curve when one has the data in hand and where adding in 2007 data would both increase the sd and increase the reliability of the sd and mean — surely a desirable and necessary thing to do for it to have any meaning in terms of the central limit theorem, presuming that one hasn’t cherrypicked a start date and end date so that it emphasizes some point one wishes to make). Given the noise, it trends along an average of around 3 sds off of the mean for the early part of the cycle, then pulls in to be 2 sds off by summer. Except for late March, where it continues perfectly smoothly to nearly parallel the mean, but the standard deviation for the mean necks in to make the March result 5 to 6 standard deviations off.
This necking in is a clear signal that the standard deviation is nearly meaningless in the curve up above. To some extent this is to be expected with such a small baseline, but this is where excluding the subsequent data is criminal. Anybody who thinks that arctic sea ice can in any way be represented by a 20 year baseline average raise your hand. What, no takers?
Of course not. We know that its natural variability on a century timescale is far larger than this.
You really ought to make at least the minimal effort to look for the data before making a claim. Start at http://nsidc.org/data/docs/noaa/g02135_seaice_index/ and note the following, since I’m sure you won’t make the effort: “Data are stored in ASCII text, Portable Network Graphics (PNG), Keyhole Markup Language (KML), and Geospatial Vector Data format and are available from November 1978 to present via FTP or the Sea Ice Index Web site.”
“They” don’t know how to make a graph? The graph you requested, of all the years, would be incomprehensible due to data overlap at any reasonable scale. Note that Bill’s doesn’t even do what you demanded and it’s pretty busy visually. You want a graph? Get off your duff and make it yourself.
You are the kind of lazy twit that gives skeptics a bad name.
Dear Dr. Meier:
You say, “If folks have other suggestions, we’d be happy to hear them and take them under consideration – post a reply here.”
I deal with these types of averaging/filtering calculation all the time in my professional capacity in feedback control systems (and I am writing this while taking a break from dealing with one such issue), and I believe you are taking a big step backwards in switching from centered to trailing averages. Especially for after-the-fact reporting, a centered average/filtered value is far superior to a trailing average, both for trends and for measurement noise. If the value you report for today, April 20, is a 5-day trailing average, you are really reporting the best estimate for April 18, not for today. This has all sorts of potential for confusion down the line.
I realize that if you want to report “real-time” data every day with an N-point centered average, the last (N-1)/2 points are necessarily tentative and subject to change. But it would be far better to show these with dotted lines and/or in a different color to emphasize their tentativeness. People will be using these plots years from now, and they will be using shifted data if you keep using the trailing average. This is simply not correct.
kbray in california says:
It plunges an ice dagger into the heart of CAGW zombieism.
I hate to bring this up, but a dagger to the heart won’t kill a zombie…
…now, back to the discussion.
I wonder why the alarm was not sounded when the “change” appeared. Credit to NSIDC that they investigated fast, gave an explanation and corrected the fail.
It seems anyway that the ice extent is close to mean value, and the alarming messages of the Arctic ice are very silent at the moment. Wonder why?
Walt Meier says:
April 19, 2012 at 11:21 am
Walt, averages mean nothing when we don’t know the >30 year variation in Arctic ice extent as the records are too short. Why not show the years with the highest and lowest summer extent along with the current year, indicate the years on the graph and let people argue from there?
Thanks for posting and thanks to WUWT for showing class in handling an honest mistake.
I agree with Curt, centered was better. A different color could be used for the 2 most recent days to show that they are in wiggle-land. Future users will not like the 2-days-offset graphs. Someday, it will be changed back — an issue has been created, not solved.
Regardless of the statistical changes, what is far more interesting is just how long the blue line will go towards black “normal” during this melt season. I see three scenarios here. 1. We have a drifting satellite, 2. We have ice up the yin yang, or 3. We have a very hard wind spreading the ice WAAYYYY out thin and it will flush-melt like an ice cream bar in the hands of a toddler.
Can’t find a trace of my posting of yesterday 🙁 Was I too numerical or technical? Or perhaps I was nit-picking! I’d like to know so that I can adjust things in future. Anyway, a pity really. I tried to put the case for being explicit about what people really understand by “standard deviation”, and to ask for a full definition of what is understood by the “SD” phrase in the current context.
Robin