This is Rob Dekker's Typepad Profile.
Join Typepad and start following Rob Dekker's activity
Join Now!
Already a member? Sign In
Rob Dekker
California
Recent Activity
It seems that this continued ice-dispersion due to lows in June and July has reached the limits of what the Central Arctic Basin can handle. As a consequence : Sea ice "area" is starting its nose-dive very early this year : That is not a good sign, folks..
Neven said All it takes, is relatively thin ice after a mild winter and early melt onset (similar to what we saw this year), followed by clear skies and high temperatures during May and June to get that melting momentum going I think that hits the nail right on the head. Remember that simple snow and ice albedo effect creates a stupendous amount of heat into the melting season. On the forum, Tealight has quantified that and shows that 2016 is quite far ahead of other years. And with my own SIPN work, I have noted that land snow cover, and ice concentration and ice area over June really to point at 4.0 - 4.1 M km^2 extent for September, with only 340 k km^2 SD. That is not some wild guess. It is an estimate based on past response of Arctic sea ice to the albedo effect, and the uncertainty in that estimate is so tight that there is really a 92% chance that 2016 will end up in the top 3, and a 66% chance it will beat 2007 and be second lowest on record.
Not to change the topic or so, but I would just like to express my disappointment with Cryosphere Today. http://arctic.atmos.uiuc.edu/cryosphere/ In their failure to transfer from the F-17 to the F-18 satellite data source, they are missing out on one of the most interesting melting seasons in history. And compliments to Wipneus for doing the work for them.
Toggle Commented 7 days ago on 2016 melting momentum, part 3 at Arctic Sea Ice
Wayne, let the ice decide for itself before we start judging.
Toggle Commented Jul 15, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
wayne said : Can you calculate the statistical probability of their projection being successful? 5.2 million at minima, given that it will take 40K loss of sea ice a day from now on. That depends on some factors. First off is where "average" melt will lead to. I think you made a point above that average melt from now until halfway September is 60 k/day. If that is the case, the September minimum with "average" weather from now on will lead us to something like 4.05 M km^2 minimum. That is actually remarkably close to my own projection, so I cannot disagree with that. Second is the "uncertainty" in that projection. Now, there are various calculations you can do for that. If you use a straight line decline you end up with something like a 550 k km^2 standard deviation. If that is the case, 5.2 is about 2 sigma's away from the "average" target, which suggests about 2.5 % probability of Arctic sea ice hitting 5.2 or above in September. If you take my projection method (which includes land snow cover and ice concentration and ice area) then the standard deviation (with June data) is only 340 k km^2. That suggests that 5.2 is about 3 sigmas away, which suggests it is virtually impossible (less than 0.1 %) to hit 5.2 this September. That is why I feel comfortable stating that I will eat my hat if we hit Schroder's 5.2 this year. More interesting is that given the "average weather" target (about 4.05) and the uncertainties from the statistical method I use, there is a 92 % chance that 2016 will be in the top-3. There is where I differ in opinion with Neven. It really does not take "exceptional" weather in July and August to end up in the top-3. To the contrary : It will take "exceptional" cool weather NOT to end up in the top-3. One final note on that "uncertainty" : These probabilities assume that the residual is Gaussian "normal", which is not necessarily the case. After all, I only use 24 data points (the past 24 years), which means that speculating on any probability resolution below 1/24= 4% is questionable.
Toggle Commented Jul 15, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
Regarding Dr. Schröder's model of melting ponds as a predictor of September extent, this paper by Liu et al (2015) casts doubt on their predictions : http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20160001390.pdf Specifically : Here we show that satellite observations show no evidence of predictive skill in May. and a persistent strong relationship only occurring after late July. Which suggests that Dr. Schroder's method works best late in the season. But there is more : satellite observations indicate a much higher percentage of melt pond formation in May than does the aforementioned model simulation Which suggests that Dr. Schröder still has work to do. None of this is intended to down-play the work that Dr. Schröder has done. In fact, his work is important in adjusting the IPCC models that we use in forecasting changes in the Arctic and globally. But for short-term predictions (like from June to September), the performance of statistical models like Dr. Slater's still outperforms Dr. Schröder's model. Regarding Dr. Slater's model (here) : http://cires1.colorado.edu/~aslater/SEAICE/ Neven said : just have to make some effort to try to understand how the model works and what anomaly persistence is. Dr. Slater presents two models in SIPN : a "persistence" model and a "forecast" model. Note that the 'persistence' model has no skill. Dr. Slater explains this quite clearly in his SIPN report : https://www.arcus.org/files/sio/23220/slater_persistence_july2015.pdf None of these methods have true skill at this long lead time. However, his "forecast" model DOES have skill over 50 days ahead, and even up to 85 days, so that June data is sufficient to make skill full predictions. And last year it was spot-on : https://www.arcus.org/files/sio/23220/slater_july2015.pdf And yes, the "forecast" model is appealing since it is "daily" so gets better the closer to September we get. It is not entirely clear how Dr. Slater's forecast model works, but from what I understand is that he uses "ice concentration" (a gridded 2D form) as his input variable, so that he also can make forecasts on a regional basis. Here is his gridded forecast for the end of August : http://cires1.colorado.edu/~aslater/SEAICE/prob_map_reg_2016.gif In summary, I think that Dr. Schröder's model (using modeled melting ponds only) is looking at the trees, and fails because it does not use the forest to make September predictions. Dr. Slater's model uses ice concentration (which includes melting-ponds but also leads, and ice edge fragmentation) and captures more of the overall picture than Dr. Schröder's model. Now if either one of these scientists would include land snow cover than I would be a happy camper :o) Nuf said for today.
Toggle Commented Jul 14, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
Not to mention the collective intelligence of your audience : http://forum.arctic-sea-ice.net/index.php/topic,1594.0.html
Toggle Commented Jul 13, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
Thank you for clarifying your perspective, Neven. I just think that there are other perspectives out there (like Dr. Slater's model) that deserve attention. Dr. Schröder has done marvelous work on melting ponds, and I understand that his work has been incorporated into GCMs even. But melting ponds are not the whole story of providing melting momentum. Ice concentration, and land snow cover are equally important, and as Dr. Slater's and my own work have shown in SIPN projections in the past. Maybe in the end, this melting season is a test of the influence of "albedo" feedback in general against the effect of "melting ponds". Albedo (snow cover, and ice area) are quite far in the lead for 2016, as work by Tealight and others on the forum have shown. That heat went somewhere, and despite the lows in June, statistical averaging suggest that that heat WILL come and bite us at some point.
Toggle Commented Jul 13, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
John Christensen said : I can't find the comments right now, but I believe wayne predicted the greatest melt in history in terms of extent and volume by Sept, The comment was : "Make no mistakes in judgement, this is the greatest melt in history" which wayne made on May 29 on his own blog : http://eh2r.blogspot.ca/2016/05/2nd-remarkable-retreat-front.html Which, noted, is NOT a prediction, just a statement of the state of the melting season at that time. and secondly I do not believe wayne has withdrawn or modified this prediction at this point. Since it was not a prediction, wayne has nothing to withdraw or modify. And your remark that "The greatest melt in history isn't performing too well " is still patronizing. Yet, yes. The lows in June (and continuing in July) have saved the ice from imminent demise this melting season.
Toggle Commented Jul 13, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
That should read : 2016 is still running some 600 k km^2 below 2015. That makes for a 1.3 M km^2 difference that somehow would have to be made up by significant stalls in the 2016 melting season for Dr. Schröder's projection to become reality. Personally I attach more value to Dr. Slater's prediction method. Not just because he includes all of "ice concentration" numbers and not just the "melting ponds" section of it, but also because his method has a better track record than Dr. Schröder's model. http://cires1.colorado.edu/~aslater/SEAICE/ And Dr. Slater's model suggest that we will go into September with some 4.3 M km^2 of ice left : http://cires1.colorado.edu/~aslater/SEAICE/ which is hinting at a 4.1-4.2 September minimum, which is in line with my own projection (of 4.1). In other words : I will personally eat my hat if we hit 5.2 in September.
Toggle Commented Jul 13, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
Hi Neven, I understand that you have some preference to highlight Dr. Schröder's model prediction (of 5.2 M km^2 in September). But considering that his SIPN July projection last year was about 700 k km^2 too high, and considering that according to Wipneus' AMSR2 extent ratings, 2015 is still running some 600 k km^2 below 2015, what makes you think this year his projection will be any better ?
Toggle Commented Jul 13, 2016 on 2016 melting momentum, part 3 at Arctic Sea Ice
Let me complete that sentence : If June 2016 would have shown the same highs that we observed in 2012 or 2007, thinks would have looked a LOT different, and you would not have patronized Wayne's statement from May 29. Let everyone have their opinion please. No need to patronize.
Toggle Commented Jul 12, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
John Christensen said : The greatest melt in history isn't performing too well right now after also lackluster results in June. Wayne said that in a post on his own blog on May 29. That was when 2016 was 500 k km^2 below the next contender (2015) and more than 1 M km^2 below 2012. And that was before persistent lows of June 2016 put the brakes on 2016's, at that time, race to the abyss. So do you REALLY have to rub this remark in his face ? Seriously ? Just be happy that these lows in June (and to a lesser extent in July) saved the ice and put it back into a more controlled decline. Just remember, if June 2016 would
Toggle Commented Jul 12, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
My method gives only a single number for the September ice extent, based on three variables (snow cover, ice area and ice concentration). It does not say anything about how that ice in September is distributed. Dr. Slater's method is based on only one variable AFAIK (ice concentration) but it DOES provide an indication of how ice is distributed. The latest distributed ice extent from Dr. Slater's model is here : http://cires1.colorado.edu/~aslater/SEAICE/prob_map_reg_2016.gif Now, the thing that is concerning about this projection is it suggests that a large chunk of ice by be (with a 60-70% probability) will be cut-off from the main pack by the end of August. If that really happens, we may see a repeat of 2012, with "flash-melt" events of isolated ice fields away from the pack with the lightest of storms at that time. Which is why I believe that my projection has more downside potential than upside potential...
Thanks guys ! Wayne, June was definitively dominated by low pressure (low temp) over the Central Arctic, and thus below-average melting. That resulted in an increase from 3.8 (May data) to 4.1 (June data) in my method. I am not very good with these weather maps, but are you saying that there are higher pressure systems settling over the Central Arctic in July ? Or will we continue to see low pressure systems dominating ? P-maker, I don't see that ENSO signal in my residuals. 1998 was estimated a little bit too low, and 2007 was estimated too high. The largest excursions from my estimate were in 2001 (estimated too low), 2004 (estimated too high), 2006 (estimated too low) and 2010 (estimated too low). If you know of any effect that explains 2001, 2004, 2006 and 2010, then please let me know. 2007 was estimated too high, but there I know the cause : a sustained dipole over the Arctic, causing compaction on the Pacific side and ice export on the Atlantic side. 2012 was estimated too high too, but not by much. That suggests that there was a LOT of heat absorbed in the 2012 melting season, and the GAC just topped off the losses that were largely expected based on this method of prediction.
The SIPN report shows predictions made based on May data. My entry was 3.8 as you can see in the report. Meanwhile, the June numbers are in, so I would like to share my updated projection for September sea ice extent. As many of you know, my method is based on estimating the 'albedo' effect of Arctic amplification during the melting season, and uses 3 variables to determine how much heat the Arctic is exposed to : - Land snow cover - Ice area - Ice concentration http://neven1.typepad.com/blog/2013/07/problematic-predictions-2.html Regressed over past (1992-2015) extent, I obtain between 4.0 and 4.1 M km^2 as the most likely outcome of the September sea ice extent this year. That is 200-300 k km^2 more than my May data outlook, which is mostly caused by reduced ice concentration in June, due to the persistent lows over the Arctic during the month. This is what this projection method predicted for the past 24 years : http://i1272.photobucket.com/albums/y396/RobDekker/JunePredict_zpsquedrtdc.png By itself, this projection is not so interesting. Just another guy with another projection for Sea ice extent in September. But more importantly, the standard deviation over this prediction is 340 k km^2, which is pretty darn 'tight', which means that we can say something about the probability of 2016's ranking. Based on these numbers : There is a 92% chance that 2016 will end up in the top-3 (only 2 out of 24 years showed a larger difference between 2011 and this 2016 projection). There is 66% chance that 2016 will be second place (after 2007 but shy of 2012). There is an 8 percent chance of 2016 beating the 2012 record in September. You can humiliate me if 2016 will beat all records, and if it will not make the top-3. Until then, I claim that average weather will get us to 4.0-4.1 or so, and that is bad enough.
Hi Bill, I want to thank you for your article back then, since it inspired me to post my thoughts in a reply, and thank Neven for upgrading that reply to a full article. Over the years, I finetuned the formula, which shows "skill" (a semi-official term in statistics) in predicting September sea ice extent from June on. With the most important part that the 'standard deviation' is much lower than other predictive methods. Here are the results of the method for past years : http://i1272.photobucket.com/albums/y396/RobDekker/June-14_zps7336859b.jpg And for last year, the prediction was spot-on : https://www.arcus.org/sipn/sea-ice-outlook/2015/july https://www.arcus.org/files/sio/23220/dekker_july2015.pdf I also have to admit that I probably posted my comment above about the June-data (with probabilities of the September 2016 rankings) in the wrong thread. I should have posted that June-data update to my SIPN 2016 prediction in the new SIPN thread that Neven uploaded a few hours before I posted my comment. Neven, should I re-post that comment in your SIPN thread ?
Toggle Commented Jul 7, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
There has been some speculation that 2016 will beat all records, and some speculation that it will not make the top-3. I don't want to disappoint anyone, but regression analysis suggests that it will be neither of the two. NSIDC's June 2016 "extent" and "area" numbers came in, as well as Rutger's snow lab's numbers for land snow in June. Using these numbers in a formula that estimates how "white" the Arctic was in June http://neven1.typepad.com/blog/2013/07/problematic-predictions-2.html Regressed over past extent, I obtain between 4.0 and 4.1 M km^2 as the most likely outcome of the September sea ice extent this year. More importantly, the standard deviation over this prediction is 340 k km^2, which is pretty darn 'tight', which means that we can say something about the probability of 2016's ranking. Based on these numbers : There is a 92% chance that 2016 will end up in the top-3 (only 2 out of 24 years showed a larger difference between 2011 and this 2016 projection). There is 66% chance that 2016 will be second place (after 2007 but shy of 2012). There is an 8 percent chance of 2016 beating the 2012 record in September.
Toggle Commented Jul 6, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
But, yes, you are right that I meant the Chartic site when I mentioned that these century drops don't show up yet on NSIDC's site (although the latest drop there (from 9.497 to 9.398) is as close to a century drop as you can get :o) Point is, 2016 is going down quite rapidly and not giving up its position in the leading years. It is not going to disappear in the rest of the pack.
Toggle Commented Jul 3, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
Hi Bill, No. I was looking at Wipneus' posts, who calculates a "mirror" for NSIDC, which is spot-on in following the NSIDC ftp site you mentioned. And accordingly, Wipneus reported another century drop today : Shadow NSIDC extent is now 9.1786 dropping -155.8k. http://forum.arctic-sea-ice.net/index.php/topic,1457.msg82067.html#msg82067 which matches exactly with the NSIDC ftp site.
Toggle Commented Jul 3, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
Wipneus posted a century drop today : Shadow NSIDC extent is 9.3344 a drop of -111.9. Doesn't show up on NSIDC's site yet, because of daily "averaging", I guess. But it is a sign of the times.
Toggle Commented Jul 2, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
John said : My point was that the spring snow cover probably correlates better with other variables (temperature, SLP) than with the decline in ice extent in June/July. It would make sense if land snow cover in June correlates well with temperature over the Arctic later in the melting season. But SLP ? Why would that correlate with snow cover ? Either way, do you have any (correlation) numbers that back up your statements ?
Toggle Commented Jul 2, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
John said : I will need to agree it is difficult to tell how the years of 2012, '13, '15, and '16 would turn out so different just based on the NH snow cover. Nobody said that NH snow cover is the "only" variable. Especially not if you are looking at May snow cover.
Toggle Commented Jul 1, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice
You mean that for the first time this melting season, 2016 is actually behind another year (2010).
Toggle Commented Jul 1, 2016 on 2016 melting momentum, part 2 at Arctic Sea Ice
And regarding the short-term projections : When I look at the latest image of ice concentration, like this one from Wipneus in the Beaufort : http://forum.arctic-sea-ice.net/index.php?action=dlattach;topic=382.0;attach=31733;image or this image from A-team of the fragmented ice around the 80N line : http://forum.arctic-sea-ice.net/index.php?action=dlattach;topic=1493.0;attach=31724;image I believe 2016 will not simply 'stall' its way further into the pack. We are going to see some serious decline in extent and area. And we will see these drops very soon.
Toggle Commented Jul 1, 2016 on ASI 2016 update 3: crunch time at Arctic Sea Ice