With daily change evident from satellite images, the speed is most impressive, isn't it? The annual advance/retreat cycle is huge too. SharedHumanity has already pointed this out on the forum, but I also independently noticed that a large portion of ice to the northeast of the main Jakobshavin calving front appears to have come unstuck from its bed and formed a new tributary ice stream in a place where the bed depth makes it seem likely for one to appear. There are no large canyons feeding this stream, so it should not have a large flow for very long. The calving event is proceeding to the east, as much into this stream as southeast along the Jakobshavn trough. With the grounding line now down into much deeper water, even the return of 50s-70s weather wouldn't stop a further retreat of around 75km to a shallower bed topography. The main Jakobshavn trough makes a significant chicane near the current calving front, running westsouthwest, then northwest, and then west again, with the current calving front located in the portion running northwest. Further retreat of around 8-10km beyond the 2013 minimum, representing around 5 years so melting will eliminate this turn, removing back stress from the turn and essentially pulling the cork from the bottle. See A-team's link here. I expect the calving front to slow or stall here for a bit as glacier speed rises to a maximum.

Geh. The Arctic Ocean, Canada, Greenland and western Russia, although eastern Russia gradually gets a bit less cloudy too.

Here Comes the Sun After a very warm Arctic winter with very low ice formation, it's been a very cold spring with a huge near-perpetual low over the central Arctic basin, repeatedly fed by moisture injections both through the Laptev Sea and across the Bering Strait (thanks, Jai). This looks set to change immediately, as GFS is forecasting the sun to break out simultaneously on May 31 across the Arctic Ocean, Canada, Greenland and eastern Russia, and to stay that way for at least a week. They are also forecasting the development of an actual Arctic Dipole in the second week of June, with both a Greenlandic high and a low over the Russian coast and not well out in the Arctic Ocean. Given the recent persistence of the NAO index in summer (it's been the same sign in both July and August as in June for each of the last 7 years in a row), the prediction of a solidly negative NAO in early June would seem to indicate a significantly stronger melting season than I would have guessed a week ago.

Pete - Don't for get their innovation which is including realtime satellite ice color measurements in the model, which makes a large difference with the ice getting as dark as it does lately. I believe that for discharge they're just taking the difference and assuming that it's smooth, without large changes from year to year. I tend to trust the GRACE data more than trusting the discharge to be smooth more than I trust their model, but it isn't obvious which bit is off.

Pete - They graph the output of their surface mass balance model from Sept-Sept. For 2013 (Sept 2012-Sept 2013), the mass balance was slightly above the recent norm, but well below the longer-term average, about 2/3 melt of the winter accumulation. The GRACE total mass balance was near zero, and even slightly positive, where a significant drop would have been expected including discharge, at least if their model is right and discharge is about the same as other recent years. They have a similar, if opposite problem with 2012. Their surface mass balance is near zero, but the GRACE mass loss is huge. To match the GRACE mass loss requires either a huge jump in discharge or a significant negative surface mass balance. The likeliest explanation is a still problem with their model. It's done a reasonable job of predicting a lot of previous years, but it's still a model output based on weather analysis and albedo, not the received surface mass balance from God. I was just pointing out that it's still an unresolved problem and there are a lot of other things that it could conceivably be.

Pete, Michael- He's referring to this. That was written before the GRACE data became available, and their estimate clearly missed badly. Yes, they're only closely modeling surface mass balance, and assuming glacier loss model they're using is rather simple. 2013 is unique as being the NAO+ year with a low albedo, so they didn't have much data to train their model on. Presumably dark snow doesn't matter as much if there isn't much sunlight that year. There are some low probability explanations that can't be ruled out yet which would explain why a reasonable surface mass balance estimate wouldn't give the right mass loss. 2012 could have left a relatively thick ice layer, so there was extra surface refreezing, there could just have not been enough surface water to drill moulins on much of the area leading to surface refreezing, low flow could have led to deep refreeze from relic cold, and low flow could have led to flow channels not developing, so the water would still be sitting there under the ice. Probably they need a big tweak of the weather dependence in their model. Sam, D- El Nino/La Nina is mostly a short-period zonal flow phenomenon, and the greater number of events aren't much linked to the PDO. Sometimes they're quite closely linked, though. If you stop arbitrarily restricting the domain to either the Equatorial or North Pacific and look at global oscillation modes, in addition to normal ENSO, the mode which emerges (Mode 3 here) is a combination of warm PDO, northward flow Pacific Meridional Oscillation, and La Nina. That graphic is what SST looked like in 2013 and 1996. With the La Nina flipped to El Nino, it looks like this. If the El Nino flips the full EOF3 state, you get this.

That salinity change graph is interesting in that, while overall salinity does go down in the simulations, this is far from uniform. It goes down the most in the nearshore Kara and Laptev, where they big Russian rivers come in, but it goes up substantially over most of the Eurasian Basin. The amount is likely underestimated, since the models tend to underestimate the branch of the Northern Spitsbergen current that continues north past Svalbard, and there is some positive feedback, as higher salinity draws in warmer, more saline currents and this increases local evaporation.

@D: You are indeed correct that if the moisture source is local, the change in Arctic Ocean salinity will be small. It's still nonzero, because it redistributes moisture within the Arctic, but the sign isn't even that obvious. The results for precipitation change and salinity change were previously known, and only the evaporation data is new. This leads to the abstract language overemphasizing the evaporation in a way which can seem strange if you didn't know this already. Further their "mostly" local amounts to around 60%, and "lesser degree" of advected inflow amounts to around 40%. Nature puts their figures online, so this isn't hard to check out from anywhere.

@idunno: The trend since 1998 is cool in the eastern tropical Pacific and warm in western tropical Pacific. You're right, they don't do a terribly good job of explaining that. The opposite should be the case for the next year or so, which should help with the drought.

Re: Ding paper: It's interesting that fixing tropical SST to the observed is sufficient to set up the observed blocking pattern. On the other hand, the stratosphere tends to have the same number of wiggles (different in summer and winter), and it seems probable that fixing any one region as either very hot or very cold would set up a standing wave which looks very much like observed pattern. Did they do a sanity check? The sanity check would be whether the modeled heat flux is vaguely reasonable. We know that the models get the albedo of the Arctic in general and of Greenland in badly wrong, which likely sets up a blocking pattern in Greenland in reality but not in their models. If the blocking is really due to the models getting the tropical SST wrong, what is their explanation of why the models fail here? If only a few of your ensemble members match reality, it's a pretty good sign that reality is an outlier. On the other hand if none of them do, it's a pretty good sign that your model is broken. There just isn't much of a observed correlation between NAO and ENSO or PDO, and claims to the contrary will have to explain why little correlation is observed in reality. The NAO-/AMO+/Greenland blocking pattern is known to be already significantly due to natural variability, so the significant natural variability conclusion is likely correct, but I haven't heard where they've justified the conclusion of tropical SST causality. @George: Generally the top of warmer water is around 125m to 500m, so the depth you're talking about is usually cold, but multiple grounded basins have sills down to nearly 1000m. Pressure gradients will always stay small near the bottom because there's friction. If, as seems probable, the deep Southern Hemisphere oceans keep absorbing heat, this will translate to surface height changes which will start to have huge effects on ocean currents, probably pouring heat across the Equator as net northward surface currents. This already happens by the default climatology and is increasing slightly. The density-change driven geostrophic current climate change effect is still fairly small, but the deep oceans have really barely begun to respond. Compare the much greater temperature increase of the Northern Hemisphere with the much greater surface height increase of the Southern Hemisphere. Some asymmetry is predicted by current models, but this degree is really problematic. They don't do well on paleoclimate hemispheric temperature differences either.

After a very warm Arctic Ocean winter, total sea volumes on the eastern side of the Arctic are very near the record low levels of 2012, with the Laptev sector setting a new record low by quite a lot. On the other hand, sea ice on the western side remain relatively high for recent years, pretty much entirely as a carryover from the cold winter and cloud-covered melting season in 2013. One particular thing to note is the extreme ice compression against the Beaufort shore this winter. Ice levels in the nearshore Beaufort are by far the highest of any recent year, while ice levels in the northern Beaufort are at a record low by a small amount. This distribution favors extreme late-summer meltoff of the Beaufort, which will of course depend on weather as well. The near-shore Beaufort will likely melt out anyhow despite thicker ice, and the northern Beaufort is more likely to melt out with thinner ice. The GFS forecast shows a storm center basically staying put over northern Chukchi and the northern Eastern Siberian sector or close to the pole, while the Beaufort stays in the warm sector of the storm and gets torched. This is quite different from the weather of most of the latter part of the melting season in 2013, which was mostly storms centered in the southeastern Beaufort and the CAA, which allowed relatively little melting in the Beaufort. We should see relatively little melting from this, initially, because of the thick ice close to shore, but it does set up a large melt out there later in the year. It also sets up the high Chukchi and Eastern Siberian sector as being less likely to melt out completely. El Nino is here in earnest now, regardless of whether or not the CPC is putting out press releases about it yet. California and the southwest US can really use a good El Nino now, since they haven't had one in far too long and are way too dry.

@Chris I view the Siberian ridging as a preconditioning that sets up the Greenland anomalous summer ridgingI'm not sure where you're getting this view from. Your Dosbat links clearly show that in May, 500mb Eastern Siberian height is weakly correlated with 500mb height in Greenland, with what correlation there is actually running the other way, with slightly stronger Greenland ridging when Eastern Siberian ridging is weak. Both Siberian and Greenland ridging has been strong since 2005, but the year-to-year correlation over this time period is pretty much non-existent, including last year which had high Siberian ridging and and low heights over Greenland. The summer Greenland ridging pattern is mostly just a modern version of a negative NAO. With warmer Atlantic temperatures and less Arctic sea ice, the convection surrounding Greenland in a -NAO state has shifted to be closer and more evenly distributed. The classic Arctic Dipole is high Greenland ridging with low 500mb heights over Siberia. Clearly ridging in both areas as in 2012 leads to low season-low ice coverage. Last year with ridging only over Siberia saw large amounts of warm, moist Atlantic air sucked into the Arctic, where it ended up resulting in increased cloud cover and low melting. Greenland blocking reduces ice in the refuge areas north of Greenland and the Canadian high Arctic. It clearly has a much larger effect on the summer total Arctic sea ice area minimum than Siberian blocking. It's still early to predict Eastern Siberian blocking with much accuracy. Snow depth maps show way below normal snow cover in Europe, which since it's upwind and it's only March I would think would tend to indicate above normal Siberian blocking, despite the deeper than normal snow in the Dosbat "snow box". Probably things will be clearer in a few months. The flip side of a -NAO state is convection in the northern Atlantic (see the rainfall at the bottom of the page). Warm water temperatures in the far northern Atlantic including record low sea ice in the Greenland Sea and Barents Sea would seem to indicate a -NAO state and strong Greenland blocking this year, but there's still a lot of room for random atmospheric noise to have an effect. I wouldn't have predicted a strong summer +NAO last year.

Keep in mind that greenhouse gasses cool the stratosphere, which increases ozone destruction even without considering added chlorine or bromine. @Bob: In recent years some of the buoys are usually out into open water by now. The Transpolar Drift has always taken buoys from that area nearly straight out the Fram, although they can get hung up if they drift a little west. It was a speed record or nearly one this year, but other recent years haven't been much different. The Coriolis force depends strongly on ice thickness even if the drag doesn't, so the drift used to be much slower. @Wili: Chlorine will destroy some methane, but this is insignificant compared to OH. The significant effects are ozone production as a byproduct of methane destruction, ozone destruction due to stratospheric chlorine, and slowing of destruction of all photodegrading aerosols due to depletion of OH by methane. As an explanation of the currently ending methane concentration plateau, the unoxidized releases aren't really all that large yet, since most of the methane currently doesn't make it out of the ocean. I'm also convinced releases from Russia around the end of the Soviet Union are greatly underestimated, resulting in declining direct anthropogenic methane emissions in the immediate post-Soviet period. @John: The diurnal variation of stratospheric ozone is rather weak. Since the reactions catalyzed by chlorine and bromine require both UV light and low-temperature stratospheric clouds, they occur only in the spring (austral spring in Antarctica) when light returns. @Jdallen: There isn't much re-vaproizaiton as there would be in a distillation column, and the thermal energy is not anywhere sufficient for the stable contaminants we're talking about, requiring destruction by UV or biological reactions. The temperature gradient has nothing to do with the reactions in a column in any case. The greater concentration of pollutants in the Arctic much less due to higher deposition rates than to increased lifetime. All reactions just take much longer due to lower temperature and lower biological activity, so the same deposition rate leads to much higher concentrations. Unlike the effects of direct mercury emissions, the described increase in mercury deposition due to lower sea ice and increased convection is probably mostly a shift of deposition time rather than a net increase, since the tropospheric air will eventually be mixed down eventually anyhow.

Exactly, Chris. Both the HYCOM and PIOMAS models calculate the the stress and strain in the ice as a very important model parameter. They don't currently release this data, but could release it in a form accessible to casual users without a lot of additional effort. Clearly they think there isn't a whole lot of interest, and I'm not they're wrong in general. What makes modeling difficult is that there is a large hysteresis with breaking the ice. Either the ice breaks or it doesn't, and later releasing the stress won't unbreak the ice. They're now incorporating ice concentration data, but aren't incorporating data on whether the ice is fast or broken. I don't find it terribly surprising that the Beaufort is cracking up again this year. The ice thickness is still rather low and the wind stress is large at the moment. I'm wondering about the depth of the thick ice region as well. Prince Gustaf Adolf Sea fast ice broke up last year and a large amount of ice drifted south through it. You can see it just rapidly clearing out the thick ice in the HYCOM model. I'm not sure that breakup has ever happened before. I'd very much like to see a year-to-year Cryosat-2 data comparison for September and not October, so it wouldn't be skewed by snow.

@Kevin: Thanks for the link. Obviously soil chemistry would have some effect, but I hadn't known quite how large it is, particularly effect of salt causing freezing, which is obviously opposite from water ice. The strongest temperature effect in your link is for calthrate formation in the small amount of residual brine after most of the intra-pore water has frozen. Your earlier quote indicated to me only slowing of disassociation due to low initial permeability further lowered by ice formation, which will also leave some stranded calthrates, but is a different process than an equilibrium shift. Low-pressure calthrates could disassociate relatively quickly, but they do seem to be more the exception than the rule, and I wonder how much methane flux they actually constitute, and whether the are abundant to result in a flux which is greater than those from clathrates on the ice, water and methane only phase diagram boundary. For all that it is a millennial time-scale process, deep ocean calthrates can also begin disassociating relatively quickly. Any small warming will cause at least some of them to disassociate relatively quickly, and deep-ocean warming can begin within decades of the forcing, even if it takes longer to complete. I've already argued that East Siberian calthrates at the normal phase diagram boundary aren't really safe from at least partial release at shorter time scales due to pressure and hydrological change originating at the surface.

@Kevin: ~200m of water column is really the absolute limit of stability at cold temperatures. At freezing it's ~300m. Where natural gas is being produced, however, there's often a significant overpressure. At large scale, the pressure can reasonably run up to the equivalent depth of rock column, or ~3x higher. This gives minimum formation depths from the surface on land of ~66m for the cold limit and ~100m at freezing. If we're looking at above freezing at a depth of 100m which is generally representative of the East Siberian Continental Shelf, this we need at least ~66m more of rock. This gives a thermal diffusion time scale of t = 66^2 / thermal diffusion coefficient. Plugging in a normal number for rock of 10m^2/year (large variance) gives ~440 years. So, yes, waiting for the temperature change to propagate takes a long time. I think that what most people are missing here is that the normal state is an example of self-organized criticality. If gas is being produced steadily, the rock/ice/methane calthrate mass is ALWAYS close to fracturing strength. If there was no gas escaping, the pressure would just go up. State change can occur due to pressure change just as easily as due to temperature change. Melt the top 10-20m of meters of permafrost, and the overpressure in that depth disappears, lowering the pressure beneath the permafrost. This will cause fractures extending downwards from the top of the permafrost. If the temperature is above the ice melting temperature, the pressure change alone will melt the top layer of calthrates, further continuing the process. Once there is flow from deeper, geothermally warmed layers, this heat will expand the fractures. With pre-Anthropocene temperatures, the dropping bottom pressure would eventually stop the transport and cause everything to freeze up again, and the system would be mostly quiescent until the next cycle. This process goes on naturally all the time, but the eruptions of methane usually do not all occur at the same time, which has the potential to be the case with a large-scale surface melting trigger.

The conclusion of the Jennifer Francis paper that jet stream wiggles are extending farther northward seems to me to be likely correct, but the methodology is suspect. It's the slowing of meander progression which is the more robust conclusion from the paper. The ~5% reduction in progression speeds is rather small, but misses much of the effects of blocking, and persistent rainfall or drought. Unless it it large enough to make the jet stream wiggle disappear altogether, a persistent additive high pressure will just make low pressure troughs move through it quickly, doing nothing which appears in the data of the Francis paper. Over warm land, there is a major positive feedback associated with humidity. If there is water available at the surface, this will lead to high humidity, upward transfer of heat by condensation, a lower surface pressure, and net surface convergence causing higher humidity and more rain. This feedback has always existed, but since it depends on absolute water vapor concentration, it is much stronger when it is warmer, and generally increases with global warming. The opposite positive effect occurs as well, resulting in persistent drought, including fires in northern regions where the lack of previous fires indicates that this drying has not happened before in a very long time. In mid-temperate and farther poleward climates, the effect is small during the winter and the effect normally does not carry over much from year to year. Greenland blocking and the associated NAO- state as well wetness western Europe in are closely associated with deep convection in the Labrador Sea, and have seen a sharp shift in the past decade or two. This shift is not likely to reverse soon, but it should reverse eventually, although only partially because global warming has a somewhat similar effect. This year saw an extreme negative arctic dipole associated with high pressure blocking in western Siberia. The extreme northern Atlantic is warmer and saltier than ever, so this high pressure led to humidity blown over the central Arctic Ocean, and a surprising amount of low pressure and storminess there. This is a significant negative feedback which we have not seen before to anything near this extent, but it is more likely than not that next year will not have the mostly random high pressure blocking at the same place to cause it again.

The Judith Curry paper can be found here. The NAO-autoenhanced deep convection of the North Atlantic, and particularly the Labrador Sea component is the largest time-dependent internal oscillation mode (EOF) of the Northern Hemisphere, and has sizable effects throughout the Northern Hemisphere. Most Northern Hemisphere significantly covary with it, when viewed with a proper time lag and smoothing. The AMO is mostly a lagged representation of it. The "Stadium Wave" analogy consists of an unsupported assertion that, like sports spectators, these other indicies vary mostly due to each other, and that the North Atlantic overturning circulation variation is not the dominant actor in their variation We could apply the same logic to ENSO and refer to it as, "A coordinated dance of most of the wolrd's climate indicies, none of them particularly more importantly related to ENSO than any other," forgetting that ENSO is fundamentally a pattern of equatorial Pacific Ocean surface temperature gradient along the equator, and its self-reinforcing zonal equatorial wind pattern. The determination of a correlation which is nearly a 70-year period sine-wave over a time domain of 100 years is necessarily an exercise in overfitting. All possible indicies will show a high degree of covariance when evaluated by the described method, provided that they have a large amount of low-frequency noise. These smoothed indicies are never shown for comparison. Instead only the described covariant mode is pictured, without noise. This amounts to showing the conclusion of the paper without showing the actual data. The determined lags and correlation coefficients are actually rather interesting, but as described above, the short length of the observed record necessitates overfitting, so few conclusions about the significance of the correlations can reasonably be drawn. The actual MOC shift is sharply nonlinear, and observation of it provides much better predictability than a 70-year period sine wave. See here for an overview of the nearly binary-mode Labrador Sea deep convection switching on in 1995, and here for the a recent data update. The sharp switching allows causality to be determined with a very high likelihood not provided by an overfitted sine-wave. Given the lag of 10-12 years, the expected response of the Arctic sea ice to the MOC change is expected to be a nearly step-function warming near the Atlantic sectors of the Arctic Ocean around ~2005, which explains much of the recent trend acceleration. At some intrinsically unpredictable time in the future, it is likely that this convection will shut off again, resulting in a matching cooling. Despite some recent stuttering of the Labrador Sea deep convection possibly due to the strong positive NAO from 2005-2012, as of a year ago it was continuing. Given the time lag, the associated cooling cannot occur before ~2023, so it cannot reasonably explain the change from 2012 to 2013.

Really, the Antarctic response to global warming is as scary as the Arctic response. It's just totally different due to having deep grounded ice, along with deep ventilation in the Southern Ocean. It turns out that the main response is that the ocean warms at depth, this brings warm water to the deep ice and melts it, this brings fresh water to the surface, and this stops or slows convection. This makes the deep water warmer, for positive feedback. The deep water (greatest change around 110m, but it warms all the way down) warms, while the surface actually cools. The convergence zone which separates warm and cool water moves north. The scary thing is that this has already gone as far as it has, while all the simulations show that it's really a centennial time-scale response. The usual reaction is, "yeah, look, average the ice in both hemispheres and it comes out normal." Yeah, it does, but when they both have huge trends and are melting ice quickly the average surface coverage doesn't mean a whole lot.

Thanks, Jim. I somehow avoided reading that it was 2012, and yes they're very close to the same for June 2012. In fact, they're not that far apart through June 2013, and then suddenly started diverging for some unknown reason. Tying different models together can be tricky.

HYCOM has been predicting North Pole holes that are larger than observed for years. Their most recent update before this actually mentioned reducing the size of the North Pole hole to better match reality as its main motivation, but still produces one which is too large. I think it's mostly due to having an ice strength that is too weak, so it can't resist being pulled apart strongly enough. The text you're quoting is saying that basically the only correction of the modeled ice sheet to reality is correction of the edge location. Even if their analysis didn't run high, the North Pole hole still wouldn't be low enough concentration to be inserted as a correction. The model still produces the hole on its own so long as they don't arbitrarily replace their modeled weathered ice sheet with a big solid unweathered block of ice. Compare the old data to the new data, and the difference is pretty obvious. The change is mostly the initialization and not the model, since the new thick, solid ice sheet they arbitrarily inserted is weathering quite quickly in the new model.

@Jim Actually, the model evolution since the the reset is not to my eye obviously different from the old model. The big difference is the initialization. Compare the old and new thicknesses. The new initialization just isn't remotely like anything HYCOM has ever turned out as a model result in summer. Maybe it's been spun up at an extreme low resolution? Neither do the new and old fields exactly match at any time, so I really have no idea how they came up with the new initialization. Year-to-year comparison will be broken for the next year or so, and optimally they should reanalyze the whole data set with the new forcings. It's mostly turned out for operational use, so they don't prioritize year-to-year comparability. It is mostly a forcing-driven model, but it still depends a lot on the internal parameters and approximations. They don't seem to have changed those, though.

@Wayne Yep, that's the pole. The actual pole in your image is towards the bottom and near the center horizontally, and is clearly marked by converging splicing artifacts. It's the point of that exactly triangular cloud bank pointing towards the upper left.

@Kevin These are just rough calculations, but they give the basic idea. The GFS analysis in summer regularly shows at least some areas of precipitable water areas over the Arctic Ocean in the 20s of mm, and sometimes even in the 30s, although the average is lower. The heat of vaporization of water is 4460 J/g and the heat of fusion of ice = 334 J/g. 2 g/cm^2 of water vapor gives us a water vapor content difference of roughly 2 x 2260 = 4460 J/cm^2 = 14 cm ice There is a roughly a maximum 20C difference in temperature on the 850 mbar countour, and the Cp of air is roughly 1 J/K/g Assuming that this continues throughout half the 1000 g/cm^2 of atmosphere gives a dry heat content of 1 x 500 x 20 = 10000 = 30 cm ice This is probably overestimated by 1/3 or so because the temperature differential goes down with height. So the the heat content transfer as water vapor during the summer is a little smaller than the dry air heat content transfer, but the water vapor accounts for a little more of the variability. Near the surface of the ice, the saturated heat capacity is about 1/3 due to water vapor, so the condensation flux is never more than about 1/2 of the dry air heat flux. If it's gaining heat from an air mass which has more water vapor than this, the additional water vapor gets converted to heat as fog condensation.

@Tor: The recent DMI 80N average temperatures are too low to be explained by this, as they're well below -1.7C. A low pressure system centered in NW Greenland/Ellesmere had been pulling very cold air north off of Greenland, over almost completely frozen sea ice, and direclty over the well-instrumented "North Pole Webcam" area, yielding temperatures down to ~-8C. At roughly 84N, 0E, this area is only around ~320km away from Greenland. There is still a lot of open water above 80N on the eastern side, including at the pole, but not between Greenland and the "North Pole Webcam" sensors. The DMI 80N temperature has already jumped back up much closer to normal as the low pressure drifts southwest. You're right that ocean water will be colder than melt ponds. -1.7C is for ocean-normal salinity, so it's a little warmer, but not much. This effect is too small to produce the observed DMI 80N temperature changes, however. I think most of the long-term trend in DMI 80N temperature is due to increased data coverage forcing corrections which reduce errors in the modeled analysis product which it is based on. @Neven: The real opposite of a dipole pattern is an annular pattern, not a reverse dipole pattern. A reverse dipole pattern will likely have even greater volume loss than a dipole pattern, although it will cause spreading towards the Pacific, so much of this will not be apparent in the extent numbers until a later compressional phase. We have very thin ice near the pole, near Severnaya Zemlya, and in the Eastern Siberan sector. Severnaya Zemlya and the Eastern Siberan sector in particular have had persistent cool onshore winds all year due to winds sweeping aroud the Russian high pressure system, but nevertheless still have very thin ice due to their low latitude. They're forecast to finally get some warm offshore winds, and should melt out significantly. On the other hand, Beaufort melting should nearly stop, and the CAA has a good chance to even reverse. The nortwestern CAA channels (west of Ellesmere and north of the Parry Channel) have thin ice and have been breaking up very quickly, and I think we'll see most of the fast ice in them break up, leaving running ice. As usual, there's been a large bubble of warm Pacific water north of the Bering Strait all summer, and if anything it looked a little larger than normal. It has cooled way too quickly to be heat loss. The cold anomaly has to be just it being capped by fresher Arctic water earlier than normal, taking more heat down with it than usual. The normal freshwater pool north of Eastern Siberia has been getting quite big lately, and I'm wondering if we're going to see odd things start happening because of this.