Anthony, Tamino and C Randals posting at Neven’s are all discussing the NSIDC September- average NH Ice Extent. Anthony is submitting his readers average to the ARCUS forecast poll. I don’t know how that is done, but I’m going to post my prediction for that specific metric just for fun.
My prediction method for the September-Average will follow the same rules as for the 7 day minimum. That is: to predict the average, I’ll restrict myself to predictors based on JAXA data. Because JAXA data exists only since 2002, I’m also going to restrict myself to predicting using 1 variable.
Applying these restrictions, fiddled around a bit trying to find a half-way decent predictor, and it turned out that if I make my prediction at this late date, using the most recent JAXA extent value happens to be a better predictor than using the most recent 7 day smooth value. The scatter plot showing the correlation between the most recent JAXA ice extent reading and the NSDIC September Ice Extent for previous years in black with the best fit line in blue, and my prediction in orange:
Using this method, my best estimate for the NH sea ice average is 4.57 million km2 which I use as my single point value for betting. My 95% uncertainty intervals are 3.78 to 5.36 million km2. (Update: Because ARCUS wants some information on skill, I added the uncertainty introduced by the uncertainty in the fitting parameter to my uncertainty estimate. This results in a range of 3.62 to 5.51 square km. I also updated the graph to show the full uncertainty intervals. )
Note that my bet corresponds to a September minimum below 4.66 km2 predicted by Tamino on July 20, 2011. (We draw from different data to make our predictions.) I’m not sure where my forecast puts me relative to C. Randal at Neven’s because (if I understand correctly) his method seems to require access to the July CT area, which is not yet available. I suspect my bet will fall well below the poll average at WUWT.
If someone can tell me how to enter an ARCUS forecast, I’ll do so if possible. Of course, prior to entering, I’ll update my forecast to base it on the freshest data possible. Assuming the deadline is near August 2, that will be near the value discussed today.
4.57 sq. km. Isn’t that much worse than we thought? 😉
I’ve given Lucia the details on how to submit to ARCUS and she advises she has done so.
j– Depends who “we” are and what “we” thought when!
In all honesty, it is likely worse than I would have predicted based on a fit to the time series for Sept NSIDC back in January. Someone using that would have expected a higher Sept average back in january. But I’m using actual melt data up to today, and the ice is currently low. So it’s reasonable to expect it to be… low….
http://ocean.dmi.dk/arctic/icecover.uk.php
Actually she is a bit on the high side. I would have guessed a bit lower but not even close to 2007. Her extrapolation would put her right in the middle of the DMI minima since records (since 2003).
Andrea/Annabelle Torres –I’m predicting NSDIC extent not OSISAF extent and i’m predicting the month average It matters.
Please stick to one name.
Lucia, isn’t there supposed to be a million in the title?
In the Spring of 2008 I famously predicted that the minimum ice extent that year would be greater than the previous year(insert link to DotEarth if you wish). I have had the wisdom to avoid predictions since then because Maria blows the ice around.
==============================================
j ferguson–
Yes. Opps! Fixed. Thanks.
The University of Bremen updates this daily:
ftp://ftp-projects.zmaw.de/seaice/prediction/estimate.png
They were predicting a record minimum last week so it’s obviously not foolproof.
It looks like they are using sub-area to predict. Do you know whether the data in their plot is available online and where?
Lucia,
Nope, sorry :'(
I found the source of that graph I’d forgotten – it’s not from Bremen – it’s by Lars Kaleschke at Hamburg University. The sub area seems to be a product of area and concentration, here’s the ARCUS submission from 2008 (.pdf):
http://www.arcus.org/search/seaiceoutlook/2008_outlook/downloads/monthly-reports/june/09_kaleschke_june_outlook.pdf
FergalR-
Thanks! It looks like he has the same general idea I did. Short term forecasts can be based on the current levels. Longer term need to be done using some other method.
We know from several studies that wind and current can have very significant effects on ice extent/area in the Arctic, specially in the so-called “peripheral seas” in the Arctic Basin, apparently at least or more so than either air or water temperature [e.g. Polyakov, et. al. 2003/2004]. We also know from fairly recent [2008] areal surveys and ongoing US Navy data that the relationship between extent/area and ice thickness is not well understood and that loss of extent/area does not necessarilly mean loss of volume/mass.
That begs the question not only of what we are “betting” on but more importantly what the area/extent numbers actually are indicative of and proof of what?
Tetris–
We are betting on the extent, not what it means. Betting has nothing to do with what the numbers mean or what they prove.
Lucia
Thx for the clarification.
It would be good if that were understood all around, because in spite of your take on the matter, it should be pretty clear at this stage of the game that should the ice extent -for whatever reason- drop below the 2007 minimum, there will be plenty of alarmists who would be all to happy to cast that particular outcome into only one interpertation, and one interpertation only…
PS: as one of my friends -a senior commander in the Canadian Coast Guard who is presently in the Arctic- has observed, the US Navy PIPPS ice data [which include thickness/volume] provide plenty of food for thought.