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
Recent Activity
In spite of record highs in BC and the Canadian NWT, with massive forest fires in Siberia, a strangely acting jet stream, which is causing a unique event which could be called "Fram import" (MYI from the Greenland sea blowing back into the Arctic Ocean), a Trans Polar drift going opposite direction as usual, and what appear to be sustained high density ice in the Arctic make me now think that maybe this is going to be a good year for Arctic sea ice ! I'd feel a bit more comfortable if the pattern this year can be explained as a 'return to trend' rather than a fluke weather event...
Toggle Commented Jul 22, 2014 on ASI 2014 update 5: low times at Arctic Sea Ice
Blizzard_of_Oz, thanks. I think your method (using ice concentration) is important and interesting, especially for short-term (<50 day) forecasts, but most of all for the insight it gives on the effect of melting ponds and fragmented ice in the melting ice margin. Especially, I found the lower-left graph in your poster intriguing: Your graph suggests that there is a 50% chance that a pixel with 62% ice concentration on July 27 will reduce to 15% ice concentration by September 15. I may be going our on a limb here, but I think that tells something about the ice thickness in the ice margin. 62% ice concentration on July 27 means 38% 'dark' area within the pixel reduced to 15% by September 15 means an average 'dark' area of (85+38)/2=61.5% 'dark' melting out 85-38=47% of the ice in place. From July 27 until September 15, that pixel will receive something like 300 MJ/m^2 of solar energy (ask me about that). 61.5% of 300 MJ/m^2 is 185MJ/m^2, which then melts 47% of the ice. With energy to melt 1 ton of ice set at 330MJ, the ice in the margin that melted out must have been about 185MJ/330MJ/0.47=1.2 meter thick on July 27. More importantly, your graph suggests suggests starting concentration in 2002 of 55% (instead of 62% in 2012), which implies (85+45)/2=65% 'dark' area melting 85-45=40% of the ice (0.65*300MJ)/330MJ/(0.4)=1.48 meter in 2002. With all the inaccuracies of this 'back-of-the-envelope' calculation, these numbers are consistent with PIOMAS and other estimates of ice thickness such as Neven's volume over area, so it seems to me that there is a case to be made that your graph adds evidence to reduced (mostly FYI) thickness. Specifically that ice in the margin (mostly FYI) reduced in thickness from 1.48 meters in 2002 to 1.2 meters in 2012... Or am I way off now ?
Toggle Commented Jul 16, 2014 on ASI 2014 update 5: low times at Arctic Sea Ice
Andrew, thank you for posting here. In your submission to the June SIPN report, you mention that your method (using ice concentration maps) has skill over the 50 period projection time frame, but skill drops below 0 for the September outlook. Can you please explain a bit more about the accuracy (and skill) of your method for periods shorter than 50 days ? For example, does the skill of your method (using ice concentration maps) improve for shorter periods (such as 30, 20 or 10 days) and if so, what is the period for best skill of your method ? And over that period, which 'sensitivity' do you find ? (How many km^2 of ice melt out over that period for 1 km^2 of reduction in ice concentration).
Toggle Commented Jul 15, 2014 on ASI 2014 update 5: low times at Arctic Sea Ice
Ostepop said : There will be no summer without arctic ice in our lifetimes or in the experience of several generations from now. I wish I could share your optimism (or should I say opportunism?). I'd be more comfortable with your projection (that "The bounceback is a reality" and "it will continue.") if our Arctic start showing the 6-7 million km^2 September minima that our models estimated for 2014. even though these models still project ice free summers in our lifetime.
Toggle Commented Jul 14, 2014 on ASI 2014 update 5: low times at Arctic Sea Ice
Thanks Chris, On the forum, you mention that you withdraw that projection, due to an error. Do you have a corrected projection based on PIOMAS data ? Also, you mention that due to time constraints you are "on the verge of retiring my blog". Let me just say that this would be a significant loss for us ice watchers. Your thoughtful insights here on ASI sustained by evidence on your blog are an inspiration and a valuable resource for all of us. As the Arctic never stops to amaze, I hope you can find time to continue in the discussions with your valuable insights and perspective.
Toggle Commented Jul 10, 2014 on ASI 2014 update 4: high times at Arctic Sea Ice
lodger, thanks. For starters, this correlation is not just with May snow cover. It includes March, April, May and June snow cover, as well as a factor of (June-extent minus June-area) which I believe represents melting ponds and polynia in June. I did not find any correlation between snow cover further back (such as Feb) and Sept ice extent. Correlation starts in March, and grows until June. Allow me a day or two to write down the details (which I will submit to SIPN for the July report), since I only have a few minutes right now. And thank you for the link to Lemke et al. I'll read it.
I'm not sure about you guys, but posts like the Arctic drilling plans from Rosneft and Exxon make me sick to the stomach. If these plans to drill for oil in the Arctic, just because we now can because of diminishing Arctic sea ice, is not an "insult to injury" to the Arctic and its ecosystems, then I wonder what is.
Toggle Commented Jul 5, 2014 on ASI 2014 update 4: high times at Arctic Sea Ice
Chris Reynolds, Thanks for your comprehensive assessment of PIOMAS gridded data, and your resulting projection of Sept 2014 ice extent. Your assessment is consistent with my projection based on spring snow cover but I'm surprised by the ice volume loss you report for May to June : 1980 to 1999 average loss is 1.8k km^3, from 2007 to 2009 it's 2.48k km^3, from 2010 to 2013 it's 3.63k km^3. Even 2013 had a loss of 3.06k km^3 from May to June. 2012 lost an eyewatering 4.17k km^3!! Specifically, do you have any suggestion about which physical process could explain a doubling of volume loss over May to June since the 80's ?
Toggle Commented Jul 5, 2014 on ASI 2014 update 4: high times at Arctic Sea Ice
lodger, nice to see your post. That was exactly what I was thinking. Regarding snow cover and Arctic amplification, here is a great post from Tamino that puts snow cover loss in perspective of AGW forcing In this case the net change is about 1150 TW. If spread over the entire surface of the earth, and if the difference in TOA albedo between snow/ice-covered and uncovered regions is 0.2, this accounts for a total climate forcing of about 0.45 W/m^2. Considering that AGW forcing increased about 0.93 W/m^2 over the same period, the albedo effect is significant. But here is the kicker : most of that heat is generated during the spring/summer months, and virtually nothing during the fall/winter. So spring/summer forcing is probably 2x what he reported. And since over a few months heat does not travel all around the globe, but likely stays within the Northern Hemisphere, we may have another 2x factor on our hands. So spring/summer forcing due to albedo effect may be 4x his annualized, globalized number, or in the range of 4*0.45= 1.8 W/m^2. Put that against AGW forcing over the same period (since 1980) of about 0.93 W/m^2 and you see that albedo amplification in spring/summer due to snow/ice loss quantified as being about double the CO2 GHG effect. These numbers are no small potatoes. If the GCM don't get snow cover right, then they WILL underestimate the RATE at which Arctic sea ice decline disappears, easily by a factor of 2.
The significance of the accuracy (sigma 319 k km^2) of using snow cover in spring as a predictor for Sept ice extent is that maybe summer weather is not as important as we thought it would be, or, even more interesting, that maybe summer weather may be biased by the amount of energy that spring snow cover inserted into the Northern Hemisphere atmospheric system. Either way, you guys can totally kill me if Sept 2014 shows ice extent below 4.1 or above 5.3.
Neven, sorry for posting in the wrong thread yesterday. Here is where this info should go. A quick summary : My submission ("Dekker" in the list) to SIO June report uses Northern Hemisphere snow cover in spring, as a predictor for how much ice will melt between now and September. Using May data (for the June report) I ended up with 4.6 million km^2, which was just below the median. The nice thing about using snow cover as a variable is that the standard deviation of the prediction is lower than most other entries. Now, June numbers are in (for Rutgers' snow cover, as well as NSIDC area and extent). Snow cover came in quite high at 3.6 million km^2, which is higher than it has been since 2009: The formula that gives the best correlation between ice loss (area in June to extent in September) includes June, May, April and March snow cover, as well as a factor (June_extent - June_area) which I think represents the amount of melting ponds and polynia during June. If I plug in the (snow cover, ice extent and ice area) numbers for the past 18 years, then this is the resulting prediction (using June data) versus the actual Sept extent : You may have to click the graph to see the whole thing. Also the prediction (of 4.7 million km^2) for this year is included in this graph. And I will submit that prediction to SIPN for the July report. The really important part about this method (of using mostly snow cover in spring as a predictor)is the standard deviation (319 k km^2), which is way better than the standard deviation on a linear trend (550 k km^2), and as far as I can see the smallest standard deviation of any of the methods presented in the SIO report. To put it in simple words, 4.7 with a SD of 0.319, mean that there is a virtually no chance that 2014 will turn out to be breaking the 2012 record, and more importantly, there is only a 2.5 % chance that 2014 will go above 2013's 5.3 million km^2. I'm real curious which July report SIO predictions will be above 5.3 or below 4.1, since these appear to have a probability of realization of only 2.5 % each.
Sory guys. I've never been good with creating and posting graphs. But I hope you get the point.
Toggle Commented Jul 2, 2014 on ASI 2014 update 4: high times at Arctic Sea Ice
Rutgers' snow numbers came in for June. At 3.6 million km^2, these are the highest since 2009. This is another indication that 2014 Arctic spring has been cool, and some may suggest that it was as cool as 2013. However, March, April and May snow cover suggest that there is quite a lot of energy in the system, which may still materialize. My simple formula of using snow cover (in March, April, May and June) as a predictor of September sea ice cover suggests that we are heading for a 4.7 million km^2 minimum, which is up from the May prediction of 4.6. The really important part about this graph using snow cover as a predictor is the standard deviation (319 k km^2), which is way better than the standard deviation on a linear trend (550 k km^2). In simple words, 4.7 with a SD of 0.319, mean that there is a virtually no chance that 2014 will turn out to be breaking the 2012 record, and more importantly, there is only a 2.5 % chance that 2014 will go above 2013's 5.3 million km^2.
Toggle Commented Jul 2, 2014 on ASI 2014 update 4: high times at Arctic Sea Ice
Bill Fothergill, I'm sorry that I created some confusion with my choice of words. Here is the deal. A doubling of CO2 will cause a 3.7 W/m^2 forcing (that is a log 2 curve). At a rate of 2 ppm/year, at the current 400 ppm concentration, that adds 2/400 * 3.7/ln(2) = 27 mW/m^2/year of radiative forcing. That's 67.5 TW/decade added forcing for CO2 for the Northern Hemisphere alone. A 1 million km^2 loss of snow causes some 50 TW added forcing (and that may very well be an underestimate). And the Northern Hemisphere lost some 2 million km^2 in spring over the past two decades. So the radiative FORCING from loss of snow in spring in the Northern Hemisphere over the past two decades is in the same range as the increase in CO2 GHGs over the same period. So if we assume that Arctic sea ice decline is caused by CO2 GHG emissions, then we should be able to see the signature of spring snow cover reflected in the Arctic sea ice decline. And we do. There is a strong correlation between April and May snow cover and September ice extent minimum, enough to make a skilled prediction under ARCUS SIO, which for example explains much of the difference between 2012 and 2013 : The point is that snow cover variability tells us how sensitive the Arctic (Sept sea ice minimum) is to radiative forcing, and the albedo feedback during spring.
Regarding Schroder et al my criticism of their 5.4 prediction, seattlerocks said : Rob, these are results published in Nature. They deserve some extra credit for that. You are absolutely right. I purchased their article from Nature, and read it, and found out that I was confused, incorrect (and incoherent) with my comments about the science in Schroder et al 2014. So, I would like to withdraw my somewhat chippy comments about their prediction. In fact, I think they did a great job, and even though I think they are biased high this year, their prediction (of 5.4) should be taken seriously. I think their method (using only melting pond data in May) is biased too high this year and I expect they will adjust their projection downward in the July report, more in line with other predictions which may go up.
I find it interesting that from all the 28 projections in this SIO June report, which has a median projection of 4.7 million km^2, that the only projection that obtained any significant media exposure is the 5.4 million km^2 projection from the Reading University Schroder et al team, which appears to be one of the highest predictions in the SIO report. And there seems to be some chest pounding going on there too : with remarks like "Professor Feltham says his team's system has been shown in a recent analysis to be of ‘unprecedented skill' compared to all other methods.". I have nothing bad to say about their methods (using melting ponds as a predictor) but even their own graph in Nature : shows that by the end of May, melting ponds are really not that significant yet (let alone be a good predictor for things to come). And thus, one would expect their high 5.4 prediction for Sept extent would be accompanied by a very high uncertainty. But its not. The team claims "500 k km^2" standard deviation for their prediction, which seems remarkably low for a May data prediction, especially since according to their own graph June and July are the big 'melting pond' months. Not May. So I am skeptical of the uncertainty in their widely publicized prediction. And I'm disappointed that the media singled our one prediction from the 28 in the SIO report.
With all the speculations about the extreme predictions (high and low) in this SIO report, we may almost forget about the bulk and the median, where most projections hang out. But either way, let's take this report with a grain of salt. Remember that this report is based on May data, and the variability of Arctic sea ice between May and September is about 500 k km^2 standard deviation. That means the 95 % confidence interval is +/- 1 million km^2. And then recall that according to Chris Reynolds brilliant graph : most predictions don't even obtain less than than 500 k km^2 standard deviation. That is not to say that we can't make any predictions. It just means it is hard, if you are using data available in May. The problem is not just the unpredictability of weather (in summer) to come. The problem is that in order to make a solid prediction for September, based on physical parameters in May, you FIRST need to explain the PHYSICAL parameters of why there is a down trend in the first place before your prediction can be better than just linear extrapolation of the trend. And that is where global warming and albedo feedback comes in. We know that we emit CO2 at a rate that increases the concentration in the atmosphere by 2 ppm/year. At the current level (400 ppm) and a generally accepted 3.7 Watt/m^2, this means that each square meter of our planet retains 27 mW/m^2/year. For the entire Northern Hemisphere, that is 6.75 TW added each year (see also the Hiroshima counter on this web site). That's 67.5 TW/decade for CO2 alone. Now put that against the impact of 1 million km^2 of snow or ice lost. Conservatively, if 100 W/m^2 makes it through the clouds on the edge where snow or ice melts in spring, then the albedo difference may cause half of that to be absorbed by the land or ocean. That is 50 W/m^2, or 50 TW per 1 million km^2. So a 1 million km^2 snow or ice anomaly causes our planet to warm about the same as as the CO2 forcing added over a decade. Now, look at the snow cover anomalies from May (which is kind of the average over spring) : and notice that over the past 2 decades, some 2 million km^2 of snow were lost. That means that over the past two decades, snow cover alone DOUBLED the GHG effect of CO2 over the Northern Hemisphere in spring. That's albedo feedback at work. For predictions between May and September, now that we know that snow cover has such a profound effect on the Northern Hemisphere heat uptake, and if heat (as in forcing) is the cause of the multi-decadal down trend of Arctic Sea Ice, then surely we should be able to see a correlation between spring land snow cover in spring, and Arctic sea ice extent in September. And we do. which is the basis for my prediction, but more importantly, I hope that we all understand that the Sept minimum is not just a toss of the dice, or a wild guess that depends only on the weather in summer, or an opinion based on political beliefs. Our planet and the Arctic comply with the laws of physics. Its our job to find out what matters for Arctic sea ice decline, and what does not.
ARCUS sipn June report (based on May data) is out : And yes, you seem to be on the low end of predictions, but considering the large standard deviations from all of these contributors, indeed we can't write it off.
Toggle Commented Jun 22, 2014 on PIOMAS June 2014 at Arctic Sea Ice
Also, I really like the other public outlook contribution (from Dr. Frank Bosse) : mostly because his standard deviation is nice and low (450 k km^2). What are your thoughts on that analysis ?
Toggle Commented Jun 22, 2014 on ASI 2014 update 1: melt pond May at Arctic Sea Ice
Thanks David ! Sorry, I had you confused with another contributor. Yes, it seems that 3 of the 7 lowest predictions are from readers of this blog. However, the best standard deviation of the predictions from the June report is 450 k km^2, which is only marginally better than the standard deviation for a simple linear extrapolation of the down trend in ice extent in September. With your method (using PIOMAS thickness distribution), will you predict the same for the July report ? Or if not, which June data do you need to make a new prediction next month ?
Toggle Commented Jun 22, 2014 on ASI 2014 update 1: melt pond May at Arctic Sea Ice
LRC thanks for that press release from the team of Schroeder et al. Their projection is based on model simulation of melting ponds in May and June, which has some real merit. See their scientific publication from last year here : Their 5.4 prediction is using melting ponds in May 2014 alone (without June). Since melting ponds in May are not yet well pronounced, it seems to me that they are a bit 'early' for their prediction, and thus they should have a larger standard deviation for this (May data) prediction than the 300 k km^2 they claimed for their (June data) prediction last year. That difference is not apparent in the press release. Either way, their prediction appears to be on the high side in the just release ARCUS Sea Ice Outlook June report (based on May data), which incidentally also features projections by several ASI (public) contributors, including RDallen, Chris Reynolds, and me : I hope Neven will do a post about the ARCUS June report, since there are many interesting projection methods presented there.
Toggle Commented Jun 21, 2014 on ASI 2014 update 1: melt pond May at Arctic Sea Ice
Neven, wow. That's interesting. It is totally consistent with Chris' animations of a MYI tongue in 2010. Did anyone attempt to estimate how large (in km^2) that arm was back then ? As for 2014 showing a similar tongue, I'm still on the fence. Chris' animations show a tongue of MYI, but it seems confined to the Beaufort. IIRC Cryosphere II showed a tongue of MYI from the CAB to the East Siberian shore (although I can't find that image right now; it is late). And I wonder how good PIOMAS is in simulating MYI drift. Chris' 2010 PIOMAS picture does not seem to show any arm or tongue to speak of.. Interesting stuff !
Toggle Commented Jun 12, 2014 on PIOMAS June 2014 at Arctic Sea Ice
Chris, Thank you so much for these great animations of ice thickness distribution. Especially the last two (for 2010 and 2014) are illuminating. The 2010 animation confirms your assessment that there was a large tongue of very thick (5 meter) ice across the Chukchi blocking melting into the Central Arctic Basin. Of course, 5 meter thick ice requires 5x the energy (or time) to melt out that 1 meter FYI, so it seems reasonable that this tongue caused the slowdown in extent melt in 2010 while still recording record volume losses. That is consistent with the 'anomaly' in 2010 in my "spring snow cover - Sept ice extent" correlation method. So I think your animations show that 2010 was indeed special, and maybe the PIOMAS drop in volume, while keeping extent high, was indeed caused by that tongue of thick MYI crossing the Chukchi. Did I interpret that right ?
Toggle Commented Jun 12, 2014 on PIOMAS June 2014 at Arctic Sea Ice
Oh, man, Neven, you beat me to the punch. I'd have loved to grill this guy on his fake arguments.
Toggle Commented Jun 11, 2014 on The day the ice cap died at Arctic Sea Ice
Hi Chris, I wonder if we can combine the two methods and obtain an even better Sept prediction (smaller standard deviation) using only May and earlier data. Here are my thoughts : I get good correlation between NH snow cover in April and May (a metric for the amount of ENERGY absorbed by the Arctic regions), and September extent, but that implicitly assumes unchanged thickness of ice. Specifically, my statistical model implicitly assumes that most of the ice that will melt out during the melting season will be FYI, and that that FYI thickness does not change much year-to-year, and that MYI will mostly stay in place. Now that makes perfect theoretical sense (I hope) except that the real Arctic does not work like that. There will be years when very little MYI melts out and years when MYI stays nicely in the area which will not melt that year. Keep that thought in mind, and then look at what my method would have predicted for the last couple of years (again using only land snow data available in May) : 2007: predict 4.69, final 4.30, delta -0.39 2008: predict 5.23, final 4.73, delta -0.50 2009: predict 5.76, final 5.39, delta -0.37 2010: predict 4.17, final 4.93, delta 0.76 2011: predict 4.77, final 4.63, delta -0.14 2012: predict 4.40, final 3.63, delta -0.77 2013: predict 5.78, final 5.35, delta -0.43 2014: predict 4.60, final ???, delta ??? Notice a couple of important things : (1) Snow cover in April and May alone does explain a 1.38 million km^2 difference between 2012 and 2013 Sept extent. That's 80 % of the 1.72 million km^2 that eventually separated these two years. And that is using data only up till May ! If 80 % of the variability between years is explained at the end of May, then maybe summer weather has less of an influence than we all have attached to it since the 2013 melting season.... (2) that for the past 7 years (with 2014 to be determined), my method has consistently overestimated the amount Sept ice left over, EXCEPT for 2010 ! It is almost as if in 2010, the ice was thicker than in other years, or, the thick ice was located in an area that eventually melted out. Now, you (Chris) have frequently reported about the great loss of MYI during the 2010 melting season, and the excessive volume loss during that year (possibly even shifting the Arctic into a new 'thinner' MYI state) during that year is also apparent in the Wipneus PIOMAS graphs : So, I'm wondering. What was the 2010 ice thickness distribution like during 2010 ? What caused 2010 to be a year when a lot of heat was in the system, and large volumes of ice melted out that was thicker than other years ? And I'm wondering how that thick ice in the melting margin in 2010 was related to 2009, an exceptionally cold summer year, comparable to the summer of 2013. And thus, I wonder how the ice volume spatial distribution charts from PIOMAS could possibly tell us if 2014 is going to follow 2010 with lots of thick ice in the margin, or 2011 for a more average ice thickness distribution... Any thoughts..?
Toggle Commented Jun 11, 2014 on PIOMAS June 2014 at Arctic Sea Ice