DMI polar data shows cooler Arctic temperature since 1958

” In general DMI´s data (if correct) reveals a cooling from the mid 1990´ies till today.”
If you plot a trend from 1958, you get a steep decline.

Shouldn’t all temperatures be denominated in C instead of K?

marek says:
August 5, 2010 at 2:47 pm
Shouldn’t all temperatures be denominated in C instead of K?
==============
Yes…
Unless the next sound you hear is the atmosphere solidifying… Which you wouldn’t actually hear, since there would be no atmosphere…

Wouldn’t the warm ocean currents coming from the south have more impact on the arctic ice melting than the surrounding air?

Anthony writes:
“Im not sure why DMI shows such a cooling trend for the 80N-90N area, but it could appear as if the ice covered areas of the Arctic has its own history of temperatures? And how should GISS data from distant land stations account for this?”
Sounds like real science. The truth is in the details and they are messy as all get out.

“There’s far more data in DMI than you might think. – Anthony”
Hi Anthony,
A minor point of clarification. DMI don’t produce this data. The data comes from the initial conditions of the ECMWF weather forecast model after assimilation of observational data. DMI plotted the graph, but weren’t involved in the production of the data.
See my previous comments here:
http://wattsupwiththat.com/2010/07/28/giss-arctic-vs-dmi-arctic-differences-in-method/#comment-442773

Looks like the hockey stick is pointing down.

Does James Hansen have the satelite readings from 1958 to 1980 to refute this?

Steven Goddard: Please disregard my earlier (August 5, 2010 at 3:29 pm) comment. I saw the discussion of the Arctic and the use of the DMI data and assumed, incorrectly, the post was written by you. I’ve redirected the comment to Frank Lansner.
Sorry for the confusion on my part.

Bob,
Do you feel that the usage of NCDC or Hadley data would have any impact on Frank’s conclusions?

Data from the North Pole webcam is at:
http://psc.apl.washington.edu/northpole/POPS13_atmos_recent.html
The last temperature recorded was 1.9 C.

Ed Caryl
-1.9C

Frank,
Well done. It is a unfortunate that some official sites release graphs without the digital data which they were plotted from.

Ed Caryl says:
August 5, 2010 at 4:27 pm
Nice link. The current DMI 80N reading is generous, being that the webcam is at 85N.

Frank Lansner says: “The differences i show are the differences GISS shows themselves. I just focus on the areas where both SST and GISS land temperatures for oceans are available to make it easier to see how huge the differences are.”
But GISS does not represent that the dTs data where it extends out over the oceans is SST data or that it would resemble SST data.
You wrote, “And yes im aware that GISS has some of their stations on Islands, but non the less, the differences GISS show between their land projections and the SST they use appear considerable.”
They should have considerable differences. They are different data, illustrating different things.
You wrote, “And even though you think the GISS meteorological index from GISS in only for “sentimental” reasons, these data are still on the net for the public to see and use, and thus should be correct.”
But you have not shown that it is incorrect. In the areas where the dTs data extends out over the oceans, do you have any proof that is not extrapolated and interpolated correctly from the land surface data? As I wrote above, GISS is not representing that will equal SST data. You assume the two should be equal and that is your error.

Steven Goddard: You wrote, “Frank’s post is about the discrepancy between DMI and GISS.”
Actually, Frank’s shifts to another subject.
You wrote, “Regardless of which of the two data sets the GISS site allows you to choose from (Hadley or NOAA) you get the same answer.” and continued, “So, I would suggest that you take your complaint to GISS for claiming to use a data set which you think they don’t use.”
I have no compaint. Unlike some others, I understand what datasets they use on the mapmaking page, because they state what data is used. And I also understand which of the two is the official dataset. Their official SST dataset is made up of HADISST from Jan 1880 to Nov 1981 and Reynolds OI.v2 SST data from December 1981 to present.
GISS added the ERSST.v3b SST data to their mapmaking webpage back in February. And they refer to it in the draft of the Hansen et al (2010) paper:
http://data.giss.nasa.gov/gistemp/paper/gistemp2010_draft0803.pdf
And in Hansen et al (2010) they also confirm their official SST dataset. Refer to page 26. They write, “The standard GISS global analysis uses the concatenated HadISST1+OISST data set, as described in the main text. Any of the alternative ocean data sets described here would yield slightly greater global warming, both in recent decades and on the century time scale.
“Until improved assessments of the alternative SST data sets exist, the GISS global
analysis will be made available for both HadISST1+OISST and ERSST+OISST. The HadISST1 concatenation with OISST will continue to be based on equating means for 1982-1992, as it always has been in the standard GISS analysis. The ERSST+OISST concatenation will be as described in the main text of this paper. HadISST1+OISST will continue to be our standard product unless and until verifications show ERSST+OISST to be superior.”
So it appears they’re planning to shift to the ERSST.v3b data and in the following post, I showed how that would increase their 20th century trend, as they acknowledge:
http://bobtisdale.blogspot.com/2010/02/when-did-giss-add-ersstv3b-data-to.html
And I followed up on that post here:
http://bobtisdale.blogspot.com/2010/03/giss-acknowledges-addition-of-ersstv3b.html

Jay says:
August 5, 2010 at 5:34 pm
Oh no, its worse than we thought.
…And if I do my eyeball of that trend within 20 years, the temp in melt season may be 0C or lower.
I make it ten years before there is no melt below 90N (or is that above 90N – which way is up?)
It even worster than we thought!

Bob Tisdale says:
August 5, 2010 at 5:43 pm

With the greatest of genuine respect (as this is very much your area) when you say;

“…do you have any proof that is not extrapolated and interpolated correctly from the land surface data?”
…isn’t the point that it shouldn’t be done at all (even if technically handled correctly)
The coast is an important/significant meteorological boundary and ignoring that in the global temperature picture surely gives an incorrect overall reading.
I am way outside my comfort zone here but I think I see what Frank is driving at.

Although Bob Tisdale has made many valid criticisms of Frank’s procedures, I think the central theme of his post still bears careful consideration, as Steven Goddard pointed out.
Certainly GISS does not cover over the SST data with land data in their final product, but they do more or less that very thing in the artic. With little station data, they use Hansen’s 1200km smoothing (a peer-reviewed method)to ‘fill in’ the artic with a trend that diverges from that of DMI .
Now who is right? I don’t know; but they can’t both be right. Anyways, Willis Eschenbach posted a while back dismantiling the 1200km correlation thesis (IMHO).
_MichaelM

R. de Haan says:
August 5, 2010 at 3:22 pm
Another interesting insight:
Researchers discover irrefutable proof that Arctic Sea Ice had disappeared in the mid 1800′s, at the end of the Little Ice Age.
***
Irrefutable? What’s your definition of irrefutable. Then again, whatever it is, it will be ‘irrefutable’.

MichaelM wrote: “Certainly GISS does not cover over the SST data with land data in their final product, but they do more or less that very thing in the artic.”
Just to clarify, it’s really not a “more or less” thing. They purposely delete SST data for much of the Arctic so they can replace it with land surface data, which has risen at a higher rate.

Bob, you write: “Frank had written “Hadley/Reynolds” I probably would not have commented about his post.”
Its true, its a typo, it should have been “Hadley/Reynolds” as this is the data I show and use.

Frank Lansner: You wrote, “Bob: I think its good that you want the sceptic forums to be of highest possible quality, but i cant help wondering why I dont see any excitement in your writings of the central element in my findings? Honestly dont you find it interesting that DMI actualy shows a cooling 80N-90N ??”
Not particularly. The majority of the DMI data is from the ECMWF ERA40 computer model reanalysis, and if you’re not aware of it, ERA40 has a higher monthly Arctic surface temperature trend (80N-90N) from 1958 to 2002 than GISTEMP.
http://i33.tinypic.com/2wgt92d.jpg
You wrote, “In april (i think) I made a story about La Nina ‘shark’ at joanne Novas site indicating that La Nina would come soon, and PANG the day after you wrote a whole article that nobody should take notice of this ‘fuss’ that one should not expect La Nina. But what happened… ?
“So we can all make mistakes, we should just help each other in a good way, motivate each rather than the opposite.”
Are you implying that I made a mistake in that post? Because there is none. Here’s a link:
http://bobtisdale.blogspot.com/2010/05/typical-average-el-nino-traditional-el.html
The intent of the post was to illustrate that very few El Nino events are followed by La Nina events. I did not make a prediction that there would be no La Nina. In fact, my closing comment was, “Will a La Nina follow the 2009/10 El Nino? Considering that only 2 of 10 El Nino Modoki events since 1950 were followed by La Nina events, the odds are against it. But nature does provide surprises.”
And with respect to your thinking I wrote that post in response to your “The La Nina shark rises to bite” at Jo Nova’s, the date on your post at Jo Nova’s is May 2, 2010, but my post was dated May 1, 2010. And I wrote it months before it was posted. I even noted that in the second paragraph, “A number of months ago I noticed some of my visitors arrived from Google searches of ‘typical El Niño’ or ‘average El Niño’. I prepared this post for them back then but got sidetracked and never posted it.”
So don’t think my posts are written to prove that yours are wrong.

Nylo wrote, “GISS deals with temperature anomalies.”
Sorry if the comment you’re referring to did not include the word anomaly. My mistake.
You continued, “What GISS says is, no matter what the true temperatures at the airport and at the sea are, if the airport temperature is 0,5K higher than normal for the airport, the temperature at the sea will probably be 0,5K higher than normal for that place at the sea.”
GISS in its standard product (LOTI) does not extend land surface out over the oceans for 1200km. They only do it with their dTs product and as I noted above, that dataset was prepared in the 1980s when SST data coverage was poor. GISS also has a disclaimer on its Surface Temperature Analysis webpage:
http://data.giss.nasa.gov/gistemp/
It reads, “Note: LOTI provides a more realistic representation of the global mean trends than dTs below; it slightly underestimates warming or cooling trends, since the much larger heat capacity of water compared to air causes a slower and diminished reaction to changes; dTs on the other hand overestimates trends, since it disregards most of the dampening effects of the oceans that cover about two thirds of the earth’s surface.”
Also, the land surface station data (GHCN) used by GISS does not all come from airports.

Bob, I too have considered that ERA-40 data ends in 2002 and this might be part of the explanation.
Fact is, DMI today shows Arctic temperatures in melt season around 0,5 K and they show higher temperatures for earlier years like i illustrated. So if this is not interesting, then you are saying that DMI is presenting data wrongly? DMI did not consider the shift from one dataset to the next and therefore shows colder temperatures today for the Arctic?

This discussion is a storm in a tea cup. It is a graph averaging above zero temperatures taken from DMI graphs. I’m not sure how these averages were derived, from real data or guestimates from the graphs. Whichever way, to me it seem quite irrelevant, because all it shows is surface temperatures above ice ranging between 1.25C and .04C. I’d consider that normal.
Apparently it did not prevent the large melt in 2007 or the general decline. This is well within the -0.5C to +0.5C range you considered as ‘normal’ in the SST map colour scheme thread. The latest image from http://www.iup.uni-bremen.de:8084/amsr/arctic_AMSRE_nic.png up to 55N looks like an assaulting army stalking 80N to 90N.

Can someone explain why the buoys shown on the NPEO’s Bouy Drift Track Map took a a 90 degree turn recently after drifting almost in a straight line for months?
http://psc.apl.washington.edu/northpole/index.html

Fuzzylogic19
There was very little melt in 2007 north of 80N. Your assumption is incorrect.

Frank Lansner asked, “So if this is not interesting, then you are saying that DMI is presenting data wrongly?”
No. You are reading too much into what I wrote.
My interests are SST, OHC, ENSO, and things of that nature. And occassionally I’ll write about other topics like the differences between the Hadley, GISS, and NCDC data. I’m not necessarily interested in sea ice or the Arctic, but you are. We have different interests.
The only reason I commented this thread was because of the reference to the wrong SST dataset in the post.

Hi RACookPE1978!
Thanks for input!
You write: “So. The question is raised: What is the yearly trend of the winter Arctic temperatures since 1958?”
Its true, after this the Winter trends, the annual trends become interesting to know as well…
K.R, Frank
PS: when Bob Tisdale writes above that there is a reference to a wrong SST dataset, its just the text under fig 5, it should have been “Hadley/Reynolds” and not just “Hadley”. (However, these 2 datasets are really very similar.)

RACookPE1978 says:
August 6, 2010 at 12:25 pm
Frank Lansner says:
August 6, 2010 at 9:36 am
Fuzzylogic19 says:
” to me it seem quite irrelevant, because all it shows is surface temperatures above ice ranging between 1.25C and .04C. I’d consider that normal.”
For the AGW crowd these Arctic temperatures appears rather important… Ask the AGW why Arctic temperatures are so important.
The temperatures does however give the impression, that the ice retreat has indeed been strongly influenced by the special winds that occurred mostly in the period 2000 – 2007.
***
Arctic temperatures are important, but in the context of this thread, I’ll point out that (1) the 2010 melt season is not yet over so the claim about 2010 is pre-mature. If you look back at 2009, you’ll see that the temperature at the beginning of the melt season stayed closer to the melt point and then rose. In calculationg the average melt season temperature was it based on a fixed number of melt days, or the actual number of days? What is considered the start of the melt season? Is it the first rise above the melt point, or the second point if it drops below again for a few days. Identical argument for the end point of the melt season. Have you looked at the other end of the graph? Just look at say day 0-150 and 250-360. In 1958 this showed variation between 233K (-40C) and 253K (-20C) compare with say 2009 242K (-31C) and 259K (-14C). Look at the other end; days 250-360. From 2000 to 2009 nearly all temperatures are above average. Take 1958 to 1967 and only 50% or less of the temperatures are below average.
Steve’s graph remains deceptive trying to make an issue where there isn’t. Temperature measurements withing 80N and 90N will vary with cloud cover, snow on ice and age of ice exposed. When I express concerns about .5C temperature rise it relates to sea and land temperatures over the longer term. To raise sea temperature by that much requires gigantic amounts of energy. From the DMI graphs I conclude that the melt season temperature is relatively stable (for now) but that the freezing season temperature average has risen a lot, meaning 6-9C. The impression is created of talking about the arctic in general on the premise of small temperature changes in an area of continuous ice, where temperature fluctuations are limited to begin with. It remains a storm in a tea cup.

Hi Fuzzy!
its true the melt season is not over yet. That why I put a conservative estimate for 2010 at 0,5K. Everthing points to something nearer 0,4K perhaps even an all time record low for 2010. But my graph shows a trend 1991 – 2010 that most certainly is not prematurely illustrated.
So, I had to add a consevative value for 2010 as 10-12 days where missing, and so everything points to that i have underestimated this years cold in the Arctic 80N-90N – i will update in weeks time.
When i alowed my self to put a conservative value for 2010 its because the rest of the melting season should behave like a natural disaster + an atomic bomb to really change things radically now 🙂
Already by now my consevative value of 0,5K is too high.
K.R. Frank
K.R. Frank

RE: Query by RACookPE1978 about temp trends outside melt season.
Tom Moriarty at Climate Sanity produced a highly commendable “reverse engineered” version of the temperature trend graphs that appear on the DMI site.
Mr Moriarty’s dataset is available at…
http://climatesanity.wordpress.com/2009/09/09/dmi-arctic-temperature-data-does-show-increasing-temperature-trend/
In this article, he describes in some detail the extraction technique employed.
As a newcomer to WUWT, I do not have first hand information, but, from what I can glean from the article, it appears that an animation of the DMI figures has already been posted on this site. Whilst such animations can undoubtedly be very illuminating, they are not necessarily the optimum technique when faced with noisy data. My approach to the analysis was therefore somewhat more traditional.
Mr Moriarty’s dataset is basically a flat file extending from 1st Jan 1958 to 6th Sep 2009 that can be easily downloaded into Excel. As one can see from the end date, this was 116 days shy of a full 52 year dataset, roughly equivalent to about 0.6% missing. (N.B. Mr Moriarty’s data actually has two entries per day, presumably at the 00z at 12z measurement windows.)
In a fit of creative accountancy, I “in-filled” the missing days by taking the corresponding unweighted average temperature for each of the 10 preceding years. (i.e. The temp used for (say) 7th Sep 2009 was the arithmetic mean of the temps on that date between 1999 and 2008 inclusive.)
As noisy daily data spread over a period of 52 years is a little difficult to interpret, I simplified things by aggregating the data into units we are all familiar with – calendar months. Linear regression on the ensuing 12 x 52 dataset gave the following decadal trends… Jan +0.76C/decade, Feb +0.71C, Mar +0.24C, Apr +0.31C, May +0.18C, Jun -0.08C, Jul -0.03C, Aug -0.07C, Sep +0.28C, Oct +0.68C, Nov +0.53C, Dec +0.63C/decade.
The daily mean temperatures shown on the DMI site are derived from the ERA 40 product, and this uses a 45 year baseline extending from 1958 – 2002. As the summer (Jun – Aug) trends are negative, it is hardly surprising that the current temperature anomalies are mirroring this behaviour.
If these figures are plotted on a graph it still looks like a dog’s dinner, so I took two additional steps to filter out some more of the extraneous noise signal. Firstly, I aggregated the data further into seasons, rather than months. These, unsurprisingly, were Winter (Dec – Feb), Spring (Mar – May), Summer (Jun – Aug) and Autumn (Sep – Nov). The additional step was to further smooth the figures by using an unweighted rolling 5 year average. (Obviously, this step reduces the notional length of the dataset from 52 to 48 years.)
The resulting decadal trends came out as follows… Spring +0.3C/decade, Summer -0.05C, Autumn +0.47C, Winter +0.68C, Annual +0.35C/decade
When I look at the graphs of these 5 data series, my interpretation is that it looks like a multi-decadal oscillation superimposed upon a pronounced warming trend. (Except, of course, for summer.) Obviously many people will disagree. Alternative interpretations could certainly involve very long term oscillations, Little Ice Age Rebound hypothesis (shame about the acronym) or simple disputation of the original data.
As I may well have made some stupid mistake whilst number crunching, perhaps someone on this site might care to repeat my reanalysis and post their findings, including, if possible, the graphical representation of the monthly, seasonal and annual trends.

PN, you write: “What is the special significance of daily average temperatures above 0 degrees C? Sea ice melts at -1.8C…”
PN – (why anonymous?) – I showed that high quality data, perhaps the best available, shows a cooling trend 1991 and foreward till today in for the melting season in the area 80N-90N. This is a piece of inforation in the scientific puzzle and then its up to you and everybody else to interpret and use this information. If you think this is an irrelevant piece of information, fine with me – I dont.
Method: its described earlier in the comments – i think i will add it to the article.
In general ERA-40 data globally shows much less warming than CRU, GISS etc. See:
http://hidethedecline.eu/pages/posts/the-perplexing-temperature-data-published-1974-84-and-recent-temperature-data-180.php
Then you say that the rest of the period (outside the melt season) shows something quite different. PN, please show me the trend for the non-melt season that you refer too – can you provide a link?
Anyway, I think its very interesting that melting season temperatures in the ice covered area have been cooling in the same time that other Arctic areas have been melting more and more until 2007. This supports that this “melting” perhaps is not just a melting but rather the results of the unusual winds and currents in the period 2000-2007 just as NASA them selves declared in october 2007.
And honestly, GISS temperature data maps of glowing red Arctic 80N-90N should not allways be that red for the Arcitic – but thats irrelevant too i suppose.

Method & attempt at recreating;
I asked about your method in producing the graph, as I was unable to reproduce the results. Using the data provided by Moriarty and simply calculating for each year the sum of all temperatures above 273.15 K divided by the number of such data points provides this outcome;
http://i25.photobucket.com/albums/c69/Nuveena/Strategy/Recreation1.jpg
As you can see, the graph tracks yours quite closely with the same “dips” and “rises”. However, it does not show such a strong downward trend towards the end as yours. I was unable to find the exact method you used, I fear, so can’t draw any real conclusions about this difference. As you can see, I also calculated the best linear fit to the data I had available and it shows a very marginal cooling for the so called “melting season”, only 0,1K over 50 years. This is likely not a strong enough trend to be statistically significant.
The 0C choice;
In order to show why I think the choice of a fairly arbitrary temperature cutoff point followed by a calculation of an average I constructed a synthetic (exaggerated) data set to illustrate my point;
http://i25.photobucket.com/albums/c69/Nuveena/Strategy/Illustration1.jpg
Clearly, the “fast” curve would end up producing a far higher average daily temperature above the 54C cutoffline compared to the “slow” curve.
The picture clearly shows the weakness of the method you have chosen. If we maintain the temperature of our sample above the ambient (0C) temperature in my synthetic example, we very clearly expend a lot more energy than if we bring the sample to a brief, high, temperature.
So, unlike the poster above maintains, the full curve is what matters and the “slopes” aren’t simply lines that connect the dots. The total energy of the system is determined by the full area under the curve and not by some semi-arbitrarily chosen segment of it. Now, let as have a look at the data provided by Moriarty to display this in practice.
Decade binning;
I chose a simple method, calculating the average half daily temperature over 5 10 year bins for each half day in Moriarty’s data. (Sum temperatures for day 0.5 over 1958-1967, divide by 10. Repeat for all days as well as for 1968-1977 etc). I’m not an expert in the fields of meteorology or polar research so the choice of 10 year bins is simply based on; the data set is approx. 50 years long and it is hard to try and understand a graph with too many curves, so a limit of 5 X 10 years seemed logical. Plotting all the 50+ yearly graphs on top of each other would just have been confusing so some sort of binning seemed to be in order. The result should be illustrative, though it certainly lacks scientific power of proof.
http://i25.photobucket.com/albums/c69/Nuveena/Strategy/Illustration2.jpg
Even though my method is a bit crude, it goes to show how minute the changes are in the summertime over the 1958-2007 period and how noticeable the warming in the wintertime is.
“Melting season”;
Finally, you are making the claim in the original post and the replies later on that the melting of arctic sea ice towards the end of the priod is not due to warming of the weather and submit your, as I have tried to show, flawed method to show it would allegedly have become cooler in the artic during what you term the “melting season”. As I’m not an expert in the field I cannot offer a full dissection of this topic but there are a couple of fairly obvious points which quite clearly argue against your claims;
1) As I stated before, you have not chosen the correct melting/freezing point for saline sea water which is -1.8C and not 0C.
2) The temperature data from DMI is an average over the latitudes from 80 degrees to 90 degrees. However, the melting of the ice does not take place evenly, it mainly takes place at the edges of the ice cap. If you would want a sort of an analogy, you could consider Miami and Washington DC, which are about 10 degrees apart. The actual melting would take place in “Miami”, but the average temperature between the two towns would be somewhere in “North/South Carolina” when in fact you would be much more interested in the temperature data from “Miami”. The effect of this is that the actual melting of sea ice begins at a far lower mean temperature than even #1 implies, in fact as soon as temperatures and sunlight at the edge of the ice cap allows for melting.
Both of these effects means the actual melting season, if we define it as the time of he year the polar ice cap begins to shrink is quite certainly much longer than what you limited your graph to express. How much longer, I just can’t say. What is clear though, is that knowing the real length of this season would be important to draw any conclusions of the data.
And that’s another important point, as my synthetic data illustration displays, the number of melting days is very important. One day at +1 C simply does not melt as much ice as, say, 50 days at +0.5 C, something you just chose to ignore in your analysis by removing the time dimension and only focusing on the temperature dimension.
I’m really sorry, but nothing I have seen here so far can convince me that the analysis is to be taken seriously. There are a number of very basic flaws which any first year university student would be able to catch in your methodology and your reasoning. And the claims you make are completely beyond the scope of your analysis.
Oh, and I don’t maintain a website.

Petter Nikula
Thanks for your work.
I found a rather significant downtrend in these Acrtic data from 1993 and foreward. Sod do you:
http://i25.photobucket.com/albums/c69/Nuveena/Strategy/Recreation1.jpg
Then you speak of “only 0,1K downtrend from 1958 and foreward. This is not in contrast to my findings as I dont mention quantitatively a specific trend 1958 and foreward. So how can this be showing me wrong? I just said that some recent values are colder than anything seen in the beginning. Thats why i got puzzled and checked things out.
By the way, 2010 may give an all time low.
I can see that the results from some of the latest years are different, and it would be nice if DMI just showed the exact numbers in stead of reading pixel for pixel as I did.
But Petter: You seem much more interested in finding some flaws in what I did rather than focus on the really really interesting here: We have heard that the Arctic is warning faster than any where else, largely based on (STUPID) measurements taken from land stations. And then both your and my reading shows this to be far from the whole picture (!)
But all you care about is your claimed “faults” by me ?
The climate debate is truly odd.
And then your last comment “Oh, and I don’t maintain a website.”
Whats that all about, Petter? Whats going on?

Update:
http://hidethedecline.eu/pages/posts/colder-arctic-melt-season-temperatures—prognose-2010—6-days-to-go-189.php
includes an example of how I obtain data, quite simple.
K.R. Frank Lansner

UPDATE: 14/8 Estimate now 0,365 C after DMI has updated. Coldest (by far) on record so far.
3 days to go for 2010 DMI melting period average temperature.