Kevin said It's important to note that the analysis there takes in a much wider area of the Arctic that the DMI 80N index does. (And also, that the DMI index isn't properly area-weighted, a fact that's been noted here before--I think John Christenson pointed it out to me--and which you can find on the DMI site if you dig hard enough. As I explained here : http://neven1.typepad.com/blog/2017/02/piomas-february-2017.html?cid=6a0133f03a1e37970b01b7c8df1d3f970b#comment-6a0133f03a1e37970b01b7c8df1d3f970b Indeed, the DMI 80N index attaches 40x the significance to the temperature within 0.5 deg of the NP as compared to the significance of a similar area at the 80 deg North lateral. So DMI temps are very much concentrated on the North Pole. To get a better overview of temperatures in the far North, you could use Slater's numbers (which are properly area-weighted, but unfortunately out of date now) or you could do the work yourself. by installing a GRIB2 parser and then need to figure out exactly which files (and at which index) contain the temp data, 2m or 925 hPa.

Thanks for that Tamino paper, Kevin ! His conclusion, that : Arctic warming is more like three to four times as fast as global warming. is sustained by many source references. In fact, warming in winter is even stronger, at 6X the global rate :

2018 just surpassed 2008 and dropped from 7th to 8th place : Juan Garcia just posted the latest sea ice observations on the ASIF : https://forum.arctic-sea-ice.net/index.php/topic,2223.msg170956.html#msg170956 September 2nd, 2018: 4,749,726 km2, an increase of 16,253 km2. 2018 is the 8th lowest on record.

Hi Hans. Couple of clarifications : That 6/6 prediction was based on May data, and my method does not work very well there yet. It's only sightly better than a simple linear decline. It had to be adjusted since NSIDC changed the way they calculate the monthly averages of area and extent. The July prediction (using June data) of 5.19 is much better with 380 k km^2 standard deviation over the residuals. That is decent but still not very accurate. So overall, I'm still not happy with this method, and I'm looking for ways to improve it. This will be the second year in a row that I'm predicting too high, and both years the preceding winter was very warm, which suggest that the ice was thinner than 'average'. Still looking for ways to incorporate ice thickness into the method to see if that improves the accuracy of the prediction. In short : It's a work in progress.

This kind of makes sense, since the Arctic was particularly white earlier in the melting season. Here is the prediction from June : https://forum.arctic-sea-ice.net/index.php/topic,292.msg162415.html#msg162415 Now, I don't think that we will end up with 5.19 M km^2 for the NSIDC September average, but we may get pretty close.

It looks like Arctic Sea Ice Extent indeed started to stall, as I predicted two weeks ago We are now at 7th place for extent : https://forum.arctic-sea-ice.net/index.php/topic,2223.msg170283.html#msg170283

Neven said : If there's a cliff of sea ice loss in the ESS, and the 2018 trend line can somewhat keep up with 2016 on the CAB graph, this year's minimum may end up in or close to the top 3. Personally, I think it will be a top 5 position. NSIDC puts the expected annual minimum a little higher (between 4th and 9th place) : https://nsidc.org/arcticseaicenews/ Following the 2005 to 2017 average rate of change between August 15 and the minimum, the extent is projected to drop to an annual low of 4.55 million square kilometers (1.76 million square miles), with a standard deviation range of 4.32 to 4.78 million square kilometers (1.67 to 1.85 million square miles). If sea ice extent continues at the rate of ice loss seen in 2008, the fastest recorded, the minimum at the end of summer would be 4.20 million square kilometers (1.62 million square miles), or the fourth lowest minimum in the satellite record. If sea ice extent continues with the rate for ice loss from 2006, the slowest recorded, the minimum would be 4.90 million square kilometers (1.89 million square miles), or the ninth lowest in the satellite record. We will see what the remaining weeks have in store, but considering that ice concentration is still anomalously high, and temps are already starting to drop below freezing, I expect a bit of a stall in the weeks ahead. And thus I expect a finish closer to the high end (4.9, and ninth place) than to to low end (4.2, and fourth place) of NSIDC's assessment.

I reviewed the Mueller et al paper that is the basis of this ASIB post. Here are the details : https://forum.arctic-sea-ice.net/index.php/topic,1384.msg166919.html#msg166919 In summary : overall a great paper, with the notion that maybe they overestimated the influence of aerosols on Arctic Sea Ice extent by a factor of 1.7 - 1.8.

Thank you much, Michael, for finding that free copy of the paper. I did not have enough time to read the paper in full yet, but just glancing over it, it seems to be a modeling exercise in CMIP5 models. Nothing wrong with that, but I wonder how they link it to actual observed sea ice decline (which is steeper than CMIP5 models predicted). Allow me some time to read the rest of the paper and I will report back.

Thanks Neven, and congrats on getting quoted in that article ! That's really cool ! I'd alway thought that aerosol had mostly a local cooling effect, so it is surprising that these guys (Mueller et al) find the signature of aerosol forcing in a regression model of Arctic sea ice decline. Very interesting, so it is unfortunate that their paper is under paid access. Also, do you know if there is an overview of world wide aerosol emissions over the past 4 decades ? Has it reduced ? Or increased ? or stayed the same as the West reduced emissions while the East increased them ?

Thanks for fixing the images, Neven. I'll be more careful next time. Let me note again that my gut tells me that this 5.19 prediction is probably an upper-bound. Especially since the May prediction, using the same method, came in much lower at 4.65. So maybe June was just a bad month for melting, and July and August may iron things out. Yet I have to go with the math, which suggests it is unlikely that Sept 2018 will end up below 4.5.

Argh. I messed up the formatting. Sorry, I got used to the 'edit' feature on the forum. To see the pictures, you have to click through to the forum link of this post : https://forum.arctic-sea-ice.net/index.php/topic,292.msg162415.html#msg162415

Most of you will know that since 2013, I use the "whiteness" of the Arctic in June as a predictor for how much ice will melt out between June and September. http://neven1.typepad.com/blog/2013/07/problematic-predictions-2.html Specifically, I use three variables to make this prediction : - Land snow cover in June - Ice 'area' in June - (Extent - Area) in June, which represents the amount of 'water' in the ice pack in June. A combination of these variables, each one of which affects the 'albedo' of the Northern Hemisphere, represents how much solar energy gets absorbed by the Northern Hemisphere in summer, and this correlates remarkably well with September sea ice cover. Details of this method is described in one of my entries into Arcus Sea Ice Prediction Network : https://www.arcus.org/files/sio/25738/sio-2016-july_dekker.pdf This year, land snow cover in June was quite high compared to recent years : Also, ice 'area' is quite high in June (in between 2014 and 2015) and the ice is still fairly compact. As a result, prediction for Sept 2018 September sea ice extent is quite high at 5.19 km[sup]2[/sup], with a standard deviation of 340 k km2. My gut feeling this year tells me that this an upper bound, but it's fairly clear that given the past performance of this method, it is highly unlikely (less than 2.5% chance) that Sept 2018 will end up below 4.5 M km2. Here is what this hind-cast method did for the past 26 years : Comments and suggestions on this method are welcome in the "Re: Land snow cover effect on sea ice" thread : https://forum.arctic-sea-ice.net/index.php/topic,292.msg162415.html#msg162415

The ARCUS SIPN report is out for predictions of Sept SIE based on May data : https://www.arcus.org/sipn/sea-ice-outlook/2018/june Interesting is that the median is 4.6 M km^2, pretty close to my estimate of 4.65.

the first week of July that is.

John said : Rob, Will we see your updated estimate soon? My estimate relies on Rutgers Snow Lab numbers for land snow cover, and NSIDC "area" and "extent" numbers for sea ice, as an estimate of the 'albedo' of the Northern Hemisphere. Unfortunately, most of these numbers are only available on a monthly basis, so I still can't give any daily updates. May numbers projected 4.65 M km^2 for September, and the June numbers will be clear in the first week of June.

wayne said : But John and Rob, way off with respect to minima, namely current sea ice status worse than 2012! Not only in extent, but with the presence of dark seas more prominent, and unusually long Gyre current reinforcement as with respect to anticyclone hanging about for a month and a half. I don't know about the "long Gyre current reinforcement". Can you quantify that so we can run a regression analysis over it as a predictor for Sept SIE ? But I DO know about " the presence of dark seas more prominent". It is part of my estimation formula in the form of sea ice area. In general, the smaller the 'area' the lower the albedo, which correlates with lower September SIE in my method. So are melting ponds and boundary ice (extent minus area). Snow thickness (over ice or over land) does not affect albedo, so there is no physical reason to include it in the regression formula. It would just introduce a new variable which will increase the risk of (statistical) "over-fitting". Remember Von Neumann's famous saying : "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." So we want to include enough variables to make a valid prediction, but not a single variable more. I use three variables in my prediction method (that all affect albedo and thus absorbed solar energy in the Arctic) : sea ice extent, area, and land snow cover.

Tealight at the ASIF informed me that NSIDC changed their definition of ice extent and area : https://forum.arctic-sea-ice.net/index.php/topic,103.msg157655.html#msg157655 So I re-ran my regressions using the new V3 data and ended up a bit lower than with the old V2 data : Prediction of September sea ice extent, based on sea ice area and extent, as land snow cover is now 4.65 M km^2. Standard deviation over the residuals is 470 k km^2, which is still showing significantly better skill than linear decline.

I will submit my prediction to the SIPN this year again : https://www.arcus.org/sipn

Darn. The Rutgers snow lab picture is just cut-off for the last year(s). Please open in a separate tab, and notice that snow cover in May was quite high when compared to the past 10 years. Or look at the source directly : https://climate.rutgers.edu/snowcover/chart_anom.php?ui_set=1&ui_region=nhland&ui_month=5

Most of you know that I am using Northern Hemisphere snow cover in spring and summer as a predictor for the September minimum. Here is a guest post on ASIB of my method : http://neven1.typepad.com/blog/2013/07/problematic-predictions-2.html This method of prediction works really well for June data, but even for May data it has some skill. Now that Rutgers Snow lab published the Northern Hemisphere snow cover for May, I ran my regression formula on the 1992-2015 training period, using May data for snow cover, ice concentration and ice area, and arrived at a prediction for September sea ice extent minimum of 4.84 M km2. This number makes sense, since snow cover in May was still fairly high (compared to the previous 10 years), and even though sea ice extent in May was at a record low, sea ice area is actually just 3rd or 4rth lowest. This means the ice pack is still fairly 'compact' which reduces the amount of heat the ice pack will absorb from the ever higher sun in the Arctic. Standard deviation over the prediction is 460 k km2, which is substantially better than a linear decline as a predictor (which has a standard deviation of about 550 k km2). So the method has some skill with May data, but for an accurate prediction, please wait for the start of July, when the June data is in, since that has real skill with about 300 k km2 standard deviation. Until then, the prediction of 4.84 M km2 stands.

To explain the difference between Zwally's assessment and GRACE observations, it is suggested here that Zwally assumed the increased snowfall over Eastern Antarctica to have the density of ice, while in reality of course the density of snow is about 1/3rd : https://www.scientificamerican.com/article/what-to-believe-in-antarctica-rsquo-s-great-ice-debate/ That would explain the difference between Zwally's assessment and actual observations (by GRACE) of the mass of the Antarctic continent ice shelf.

Regarding Antarctic ice loss, the GRACE satellites recorded the following : Research based on observations from NASA’s twin NASA/German Aerospace Center’s twin Gravity Recovery and Climate Experiment (GRACE) satellites indicates that between 2002 and 2016, Antarctica shed approximately 125 gigatons of ice per year, causing global sea level to rise by 0.35 millimeters per year. https://svs.gsfc.nasa.gov/30880

Thanks for the update Neven. It is indeed quite spectacular what is happening in the Bering. It would be interesting to keep an eye on how this anomaly works its way into the ice cover of the Chukchi sea going into the melting season. P.S. The Arctic Sea Ice Forum seems to be down. I get this message : Table './arcticse2/smf_messages' is marked as crashed and last (automatic?) repair failed

John, my prediction method, using NH snow cover, sea ice extent, and sea ice area as predictors, is useful with data from the end of May, but statistically significant only with data from the end of June. Sorry, I still have not found any variables that make a statistically sound prediction earlier than that.