Arctic Sea ice data collected by DMI 1893-1961

I came across a number of maps showing Arctic ice extend from 1893 to 1961 collected by DMI in “Nautisk Meteorologisk Aarbog”. Each year DMI have collected information on sea ice extend so that normally each of the months April, May, June, July and August ice extend was published.

There is much more to be said about these, but this is my summary for now.

Fig 1. 1901-1910 Arctic sea ice data collected by DMI. Click to enlarge!

Sadly, just for a few years we also have March or September available, and thus we normally can’t read the Arctic ice minimum (medio September) from these maps. The August trends will have the main focus in this writing.

First of all I would like to thank “Brunnur” in Iceland for making these maps available on the net beautifully scanned. This is a gold mine and I’m sure you know this, Brunnur.

Fig 2. August 1902.
The August data in the beginning of the century normally resembles December ice area for recent years. Year after year in the period 1901-1920 we see pretty much same picture. The sea east of the Russian island Novaja Zemlja is often frozen over even in August, and there is still sea ice between Baffin Island and Greenland.

Fig 3. 1911-1920. Click to enlarge!

Fig 4. August, 1916. The December-like August ice area continues to be observed year after year, and in 1916 most of the ocean between Baffin Island and Greenland is ice filled (- even in August!).

Fig 5. 1921-30

Fig 6.
Finally in 1923 something new happens: The ice east of Svalbard and east of Novaja Zemlja is on retreat.

Fig 7.
In 1930, the retreat has gone even further: Svalbard Is ice free, and ice free waters have been observed far east of Novaja Zemlja. In addition, the Baffin bay is now almost ice free. Puzzling is, that the ice extends on the pacific side of the Arctic remains rather constant in all these years.

Fig 8.
In 1932 we see in August open ice almost all along the Russian shore. So even though we do not see the September ice minimum here, we almost have an open NE passage.

Fig 9.
After a rather icy 1934, then 1935 again in August shows an almost open NE passage and in 1935 open waters are observed not that far from the North pole.

Fig 10.
In 1937, more open waters are observed in the Pacific and East Siberian areas.

Fig 11.
1938: Unprecedented areas of open waters.
(And again, this is not the ice minimum but just the August ice area)

Fig 12. 1931-1946
Already the year after, 1939, the ice extend resembles the pre 1923 extend.
We see that a decline in Arctic ice area from around 1921 ends possibly in 1938.

Fig 13. 1947-1956
Sadly we don’t have the Arctic warm years 1940-45, but just the colder years 1946-56.

Fig 14.
In 1952, The August sea ice area once again appears like the 1900-1920 extend. If Arctic ice areas reflects temperature well, then years around 1946-54 should be as cold as before 1923. It appears that the ice cover from 1938 to 1946 has recovered quickly.

Fig 15.
Here is an August–September comparison for 1901. For most of the Siberian shores in September we see open waters as far back as  1901.

Fig 16.
Some warm Arctic years in the 1930´ies from DMI compared to recent Chryosphere today August graphics.

It seems that ice area for 1935 and 1996 were roughly similar (and it seems that ice area for 1938 and 2000 were roughly similar etc.):

Fig 17.
However, Chryosphere Today do not show 1935 ice area similar to 1996. Instead Chryosphere has added roughly 1,9 mio km2 to the ice area 1935 compared to 1996 (- The size of Greenland is 2,1 mio km2… ).

Fig 18a. We can also illustrate the missing Chryosphere ice decline after 1921 in another way.
The Chryosphere Arctic ice area data actually suggests a little more ice in 1937 than 1921 – but as shown above DMI, suggests a strong decline after 1921.

Fig 18b – and here the ice decline 1921-38 in four stages.

Fig 19. Also in another context it appears that the ice area data on Chryosphere has added area to older data:
If we compare the Chryosphere annual sea ice extend with the IPCC SAR 1996 data, we can see that the dive in 1996 data before 1979 is not represented in Chryosphere data. The divergence is perhaps 0,9 mio km2 over just the period 1973-1979.

Fig. 20, NW Passage in DMI data.
In September 1901 we are not far from having open NW passage and in September 1907 we do have an open NW Passage. We don’t have September images later than 1907 to determine if the NW passage is open.

What have we learned according to DMI´s international compilation of sea ice data?
– That sea ice data has declined strongly even in the recent past before human CO2 outlet.
– That Sea ice from a level not far from the 2006 level has recovered very fast 1938-1946.
– That the Sea ice decline documented year after year in DMI maps after 1921 apparently is not shown in Chryosphere data for some reason.
We do not have the WW2 data, but the maps of 1957-61 ice areas EXIST!
These are the years where we had a strong Solar max and photos of US Navy submarine on a slushy North pole.
If ANYONE have these maps, I would be grateful to see them!

Environmental impacts can be dramatically reduced through the use of using environmentally friendly consumer products, electric cars, ozone generator air purification, riding bicycles etc.

Further, this series of maps as I understand it was also published by DMI for the years 1962-72 in a series called “Oceanografiske Observationer”.

The real temperature trend given by Foster and Rahmstorf 2011

Fig1. Foster and Rahmstorf recently released a writing (sponsored by Casino 41) on ”The real global warming signal”. The point from F&R is, I believe, debating to counter the “sceptic” argument that temperatures has stagnated during the last decade or more. Since this is an essential issue in the climate debate I decided to investigate if F&R did a sensible calculation using relevant parameters.


Hadcrut global temperatures do have a rather flat trend these days:

Fig2. It is possible to go back to 1 may 1997 and still see flat trend for Hadcrut temperature data, so this data set will be subject for this writing:

Can F&R´s arguments and calculations actually induce a significant warm trend even to Hadcrut 1998-2011?

F&R use three parameters for their corrections, ENSO, AOD (volcanic atmospheric dimming) and TSI (Total Solar Irradiation).

“Objection”: TSI is hardly the essential parameter when it comes to Solar influence in Earth climate.
More appropriate it would be to use the level “Solar Activity”, “Sunspot number”, “Cloud cover” “Magnetism” or “Cosmic rays”. TSI is less relevant and should not be used as label.

Fig3. FF&R has chosen MEI to represent EL Nino and La Nina impacts on global temperatures. MEI is the “raw” Nina3,4 SST that directly represents the EL Nino and La Nina, but in the MEI index, also SOI is implemented. To chose the most suited parameter I have compared NOAA´s ONI which is only Nina3.4 index and MEI to temperature graphs to evaluate which to prefer.
Both Hadcrut and RSS has a slightly better match with the pure Nina 3,4 ONI index which will therefore be used in the following. (Both sets was moved 3mth to achieve best it with temperature variations).

After correcting for Nina3,4 index (El Nino + La Nina) there is still hardly any trend in Hadcrut data 1998-2011. (If MEI is chosen, this results in a slight warming trend of approx 0,07 K/decade for the corrected Hadcrut data 1998-2011).

Fig5. I then scaled to best fit for SATO volcano data set. For the years after 1998, there is not really any impact from volcanoes, and thus we can say:

There is no heat trend in Hadcrut data after 1998 even when corrected for EL Nino/La Nina and volcanoes.

However, this changes when inducing Solar activity, I chose Sun Spot Number, SSN, to represent the Solar activity:

To best estimate the scaling of SSN I detrended the Nino3,4 and volcano corrected Hadcrut data and scaled SSN to best fit. Unlike F&R, I get the variation of SSN to equal 0,2K, not 0,1 K as F&R shows.

Now see what happens:

F&R describes the Solar activity (“TSI” as they write…) to be of smallest importance in their calculations. However, it is only the Solar activity, SSN, that ends up making even the Hadcrut years after 1998 show a warm trend when corrected. On Fig7 I have plotted the yearly results by F&R for Hadcrut and they are nearly identical to my results.

So, a smaller warming from my using Nino 3,4 combined with the larger impact of Solar activity I find cancels out each other.


For now it has been evaluated what F&R has done, now lets consider issues:

1) F&R assume that temperature change from for exaple El Nino or period of raised Solar activity etc. will dissapear fully immidiately after such an event ends. F&R assumes that heat does not accumulate from one temperature event to the next.
2) Missing corrections for PDO
3) Missing corrections for human aerosols – (supposed to be important)
4) Missing corrections for AMO
5) F&R could have mentioned the effect of their adjustments before 1979

Issue 1: F&R assume that all effect from a shorter warming or cooling period is totally gone after the effect is gone.
Fundamentally, the F&R approach demands that all effects of the three parameters they use for corrections only have here-and-now effects.


In the above approaches, the Nino3,4 peaks are removed by assuming that all effects from for example a short intense heat effect can be removed by removing heat only when the heating effect occurs, but not removing any heat after the effect it self has ended.

Now, to examine this approach I compare 2 datasets. A) Hadcrut temperatures, “corrected” for Nina3,4 , volcanoes and SSN effects as shown in the above – detrended. B) The Nino3,4 index indicating El Ninos/La Ninas and thus the timing of adjustments. (We remember, that the Nino3,4 was moved 3 months to fit temperature data before adjusting):

After for example “removing” heat caused by El Ninas during the specific El Nino periods, you see heat peaks 1 – 2 years later in the “Nino3,4” corrected detrended temperature data.
That is: After red peaks you see black peaks..

This means that the approach of systematically only removing heat when heat effect is occurring is fundamentally wrong.

Wrong to what extend? Typically, the heat not removed by correcting for Nina3,4 shows 1-2 years later than the heat effect. Could this have impact on decadal temperature trends?
Maybe so: In most cases of El Nino peaks, first we have the Nino3,4 red peak, then 1-2 years after the remaining black peak in temperature data that then dives. But notice that normally the dives in remaining heat (black) normally occurs when dives in the red Nino3,4 index starts.

This suggests, that the remaining heat from an El Nino peak is not fast disappearing by itself, but rather, is removed when colder Nino3,4 conditions induces a cold effect.

In general, we are working with noisy volcano and SSN corrected data, so to any conclusion there will be some situations where the “normal” observations is not seen strongly.

Now! What happens is we focus on periods where the Nino3,4 index for longer periods than 2 years is more neutral – no major peaks?

Now, the detrended Hadcrut temperature “corrected” for Nina3,4, Volcanoes and SSN –  black graph – has been 2 years averaged:

The impact of El Ninos and La Ninas is still clearly visible in data supposed to be corrected for these impacts. Since this correction by F&R is their “most important” correction, and it fails, then we can conclude that F&R 2011 is fundamentally flawed and useless.

Reality is complex and F&R has mostly seen the tip of the iceberg, no more.

More: Notice the periods 1976-1981 and  2002-2007. In both cases, we a period of a few years with Nino3,4 index rather neutral. In these cases, the temperature level does not change radically.
In the 1976-81 period, the La Ninas up to 1977 leaves temperatures cold, and they stay cold for years while Nino3,4 remains rather neutral. After the 2002-3 El Nino, Nino3,4 index remains rather neutral, and temperatures simply stays warm.

Issue 2: Missing corrections for PDO

Quite related to the above issue of ignoring long term effects of temperature peaks, we see no mention of the PDO.

Fig11. Don Easterbrook suggests that a general warming occurs when PDO is warm, and a general cooling occurs when PDO is cold. (PDO = Pacific Decadal Oscillation). That is, even though PDO index remains constant but warm, the heat should accumulate over the years rather than be only short term dependent strictly related to the PDO index of a given year. This is in full compliance with the long term effects of temperature peaks shown under issue 1.

Don Easterbrook suggests 0,5K of heating 1979-2000 due the PDO long term heat effect.
I think the principle is correct, I cant know if the 0,5K is correct – it is obviously debated – but certainly, you need to consider the PDO long term effect on temperatures in connection with ANY attempt to correct temperature data. F&R fails to do so, although potentially, PDO heat is suggested to explain all heat trend after 1979.

I would like to analyse temperature data for PDO effect if possible.

Fig12. PDO data taken from
To analyse PDO-effect we have to realise that PDO and Nino3,4 (not surprisingly) have a lot in common. This means, that I cant analyse PDO effects in a dataset “corrected” for Nino3,4 as it would to some degree also be “corrected” for PDO…

More, this strong resemblance between Nino3,4 and PDO has this consequence:
When Don Easterbrook says that PDO has long term effect, he’s also saying that Nino3,4 has long term effects – just as concluded in issue 1.

Fig13. Thus, I am working with PDO signal compared to Hadcrut temperatures corrected for volcanoes and SSN only. The general idea that heat can be accumulated from one period to the next (long term effects) is clearly supported in this compare. If PDO heat (like any heat!) can be expected to be accumulated, then we can se for each larger PDO-heat-peak temperatures on Earth rises to a steady higher level.

Fig14. Note: in the early 1960´ies, the correction of volcano Agung is highly questionably because different sources of data concerning the effect of Agung are not at all in agreement. Most likely I have over-adjusted for cooling effect of Agung. On the above graph from Mauna Loa it appears that hardly any adjustment should be done…

Scientists often claim that we HAVE to induce CO2 in models to explain the heat trend. Here we have heat trends corrected for volcanoes and SSN, now watch how much math it takes to explain temperature rise after 1980 using PDO:

Fig15. “Math” to explain temperature trend using PDO. Due to the uncertainty on data around 1960 (Agung + mismatch with RUTI world index/unadjusted GHCN) I have made a curve beginning before and after 1960. For each month I add a fraction of the PDO signal to the temperature of last month, that is, I assume that heat created last month “wont go away” by itself, but is regulated by impacts of present month. This approach is likely not perfect either but it shows how easy temperature trends can be explained if you accept PDO influence globally.
(In addition I made some other scenarios where temperatures would seek zero to some degree, and also where I used square root on PDO input which may work slightly better, square root to boost smaller changes near zero PDO).

Now, how can PDO all by itself impact a long steady heat on Earth?? Does heat come from deep ocean or??

Fig16. It goes without saying that SSN and PDO (and thus Nina3,4 as shown) are related.
Is it likely that PDO affects Sun Spot Numbers? No, so we can conclude that Solar activity drives temperatures PDO which again can explain temperature changes on Earth.

Suddenly this analysis has become more interesting than F&R-evaluation, but this graph also shows that F&R was wrong on yet another point: Notice on the graph that we work the temperatures “CORRECTED” for Solar activity… But AFTER each peak of SSN we see accumulation of heat on earth still there after “correcting” for solar activity. Thus, again, it is fundamentally wrong to assume no long term affects of temperature changes. This time, temperature effect can be seen in many years after the “corrected” Solar activity occurred.

Conclusion: PDO appears Solar driven and can easily explain temperature developments analysed.
Thus perhaps the most important factors to be corrected for – if you want to know about potential Co2 effects – was not corrected for by F&R 2011.

Issue 3: Missing corrections for human aerosols – that are supposed to be important

It is repeatedly claimed by the AGW side in the climate debate that human sulphates / aerosols should explain significant changes in temperatures on earth.

When you read F&R I cant stop wonder: Why don’t they speak about Human aerosols now?
Fig17. In basically all sources of sulphur emissions it appears that around 1980-90 these started to decline.
If truly these aerosols explains significant cooling, well, then a reduced cooling agent after 1980 should be accounted for when adjusting temperature data to find “the real” temperature signal.
F&R fails to do so.

Issue 4: Missing corrections for AMO
AMO appears to affect temperatures in the Arctic and also on large land areas of the NH.

Fig18. In fact, the temperatures of the AMO-affected Arctic is supposed to be an important parameter for global temperature trends, and thus correcting for AMO may be relevant.
The AMO appears to boost temperatures for years 2000-2010 , so any correction of temperatures using AMO would reduce temperature trend after 1980.
F&R do not mention AMO.
Issue 5: F&R could have mentioned the effect of their adjustments before 1979

F&R only shows impacts after 1979, possibly due to the limitations of satellite data.

Fig19. “Correcting” Hadcrut data for nino3,4 + volcanoes it turns out that the heat trend from 1950 is reduced around 0,16K or around 25%. Why not show this?
I chose 1950 as staring point because both Nina3,4 and SATO volcano index begins in 1950.


F&R appear seems to assume that temperature impacts on Earth only has impact while occurring, not after. If you heat up a glass of water, the heat wont go away instantly after removing the heat source, so to assume this for this Earth would need some documentation.

Only “correcting” for the instant fraction of a temperature impact and not impacts after ended impact gives a rather complex dataset with significant random appearing errors and thus, the resulting F&R “adjusted data” for temperatures appears useless. At least until the long term effect of temperature changes has been established in a robust manner.

Further, it seems that the PDO, Nin3,4 and Solar activities are related, and just by using the simplest mathematics (done to PDO) these can explain recent development in temperatures on Earth. The argument that “CO2 is needed to explain recent temperature trends” appears to be flat wrong.
Thus “correcting” for PDO/Nina3,4 long term effect might remove heat trend of temperature data all together.
Solar activity is shown to be an important driver PDO/Nino3,4 and thus climate.
Finally, can we then use temperature data without the above adjustment types?
Given the complexities involved with such adjustments, it is definitely better to accept the actual data than a datasets that appears to be fundamentally flawed.
Should one adjust just for Nino3,4 this lacks long time effects of Nina3,4 and more it does not remove flat trend from the recent decade of Hadcrut temperature data.

RUTI: Global land temperatures 1880-2010

First estimate of global land temperature trends from the RUTI project , recently presented at Joanne Novas for the Coastal-Noncoastal issues.

… Between 1950 and 1978, the BEST results for global land temperatures have 0,55K more warming than RUTI. Otherwise, the 2 datasets are strikingly similar ….

Fig1. First estimate of global land temperature trends. As always in the RUTI project, data are unadjusted GHCN and the main efforts in the RUTI project is to identify areas of similar temperature trend before averaging – this due to limited data periods made available from GHCN (see more). As will be the case for all data sources, older data, especially before 1900 has limited data as foundation. All RUTI data in the present article use 1961-90 as base period.


1) Temperature peak in the latest decade appears to be around 0,22 K warmer than the 1940´ies heat peak.

2) We see a strong temperature decline 1940-78 around 0,55-0,6 K.

Lets compare with the Berkeley´s BEST project:

Fig1a. Recently, Berkeley released data for land temperatures as shown. Lets compare Undajusted GHCN/RUTI with Berkeley:

Fig1b. (Red RUTI graph is 10 yr avg.)

1) Temperatures recent decade is

RUTI:      0,2-0,25 K warmer than warm peak around 1940

BEST:      0,75-0,8 K warmer than warm peak around 1940


2) Temperature decline after 1940-1978 is

RUTI:       Approx 0,55-0,6 K

BEST:       Approx 0,1-0,15 K


BEST has around 0,55 K more heat in their results than RUTI, and that this difference mostly occurs between 1950 and 1978.

Fig1c. The difference in temperature trends 1950-78 is best illustrated by setting temperatures 1940-50 for the two datasets to be equal.

Differences in temperature data 1950-78 is no news to say the least. For example is has been discussed in the following article:



If you take a closer look at fig 1c, perhaps the period of difference is mostly occurring over the 28 years 1950-78, because the first decade of the decline 1940-78 also occurs in BEST (and GISS and CRUTEM3)

If RUTI has significant errors exactly in the period 1950 to 1978 this is truly odd.

1950-78 is definitely a period with far most data available from Unadjusted GHCN, and should be just about the most solid part of the RUTI data.

Imagine that BEST was correct, and RUTI was 0,55K wrong 1950-78….

– Then IN AVERAGE I should have added 0,55K to ALL temperature sets used (around 1200) between 1950-78.

This is absurd:




About RUTI global land temperatures – First estimates.

Fig2. The first estimate of RUTI global land average is based on the above blue areas.

Primarily the central North America data are not yet included, and some central Asian areas.

Red areas: Data not available or too low quality for any scientific use.

White areas: Areas that still needs further analysis, and some of these will be included in RUTI.

Fig3a. Northern hemisphere continents show great similarity.

Remember: If BEST was correct, each area should include a 0,55K ERROR decline 1950-78.

Completely unrealistic.


Fig3b. For Africa and then the 2 SH continents, trends are somewhat more mixed. Especially Australia shows a much more flat trend than all other continents.


MORE DETAILS: See much more details on all areas from the RUTI project .

Below the large area trends and weighting shown.




Fig5. Trends for areas weighted as shown.


Fig7. Trends for areas weighted as shown.

Fig8. in the Asia data we have included the large Siberian area.

Fig9. Trends for areas weighted as shown.

Fig10. Trends for areas weighted as shown.

Fig11. This area includes parts of Afghanistan, all Pakistan, India, most of Bangladesh and a bit of Burma.

Fig11a. Temperature trends for China. (A small part of China near himalayahs highest altitudes is not yet included. I am writing on RUTI: Himalayah article for that area, coming up).

Fig11b. China, trends for areas weighted as shown.


Fig12. Australia trends.

Fig13.  Trends for areas weighted as shown.

Fig14. Europe temperature trend.

Fig15. Trends for areas weighted as shown.

Fig16. North America trend, however most central parts missing still:

Fig17. In the next estimate RUTI global land temperatures the remaining North America will be included. The trends found in the above areas are weighted as shown, and the resulting trend is used with weight as full North America in the global trend, to get closest to the correct result.

Fig18. South American temperature trends.


Fig19. Trends for areas weighted as shown.


See much more details on all areas from the RUTI project .

Where should we expect UHI in temperature data 1979-2009?

– In the largest urban areas or in areas with the fastest growing population?

[Note 17/1: UHI in limate data originates from changes of UHI over a period. So the general discussion is not if a city has UHI  – it has – but if UHI in a town has grown due to expansion of the city or other factors. A more permanent big-city UHI will not disturbe temperature readings. So when we speak of UHI in global temperature trends, then we refer to an average UHI increase around the temperature stations world wide.]

In the article I suggested:
1) Ground based ocean temperature data matches fairly the UAH TLT ocean data – but the ground based land temperature has a warmer trend than UAH TLT land data. I argued that this extra heat in the ground based land data might originate from UHI, adjustments and siting problems.


2) The UAH TLT land vs ocean data has a rather similar trend as if they seek an equilibrium in temperatures. I mentioned the problem for ground based data that the land data vs. ocean data has an ever larger difference so far not explained.

Point 1) was intensely debated which as always brings about more knowledge. Especially one argument used based on expectations to where geographically UHI should be seen strong in the satellite years 1979-2010 has been rewarding to examine closer.

Bob Tisdale has made a very useful grid-wise compare of TLT UAH data vs GISS data worldwide which enables further studies on point 1).

This could shed light on the UHI question: If Ground based GISS data warms faster than UAH TLT in areas where we expect UHI, then perhaps a UHI signature is confirmed. And UAH data in compare with GISS data an indicator of UHI.

Example: Central North America/USA.

Fig 1a.
Bob finds no significant difference between USA temperatures for UAH vs. GISS ground based for 1979-2009. If one expects that there should be UHI in USA in this period, then according to my suggestion, we should have seen GISS ground based warming faster than UAH USA temperatures.
Fig 1b.
However, for example McIntyres calculation of Petersons USA city-vs-rural data shows no UHI 1979-2009 for USA either… So perhaps we should not expect UHI 1979-2009 for USA?

Based on Tisdales results I have used the following labels on the graphic below:
“NoUHI ” :  Means that GISS ground based data often from cities and airports has a rather simlar warming trend compared to the TLT UAH (- sometimes just a little extra warm in GISS data).
“UHI” :   Means that GISS ground based temperature trend is significantly warmer than UAH TLT and thus contains something that cannot be detected by satellite, we call it UHI for now.

Heres then how we should expect UHI to be distributed world wide based on Tisdales results:

Fig2 So… We should have UHI 1979-2009 in areas with not that dense urbanization? And not in the USA and similar places? Can this be true? Well, I don’t blame anyone for saying that this is not really in favour of the UHI concept. (The North Atlantic “UHI”: See “Post Scriptum B”)

But before throwing data out in the drain and before skipping the UHI idea, do we actually know where geographically we should expect UHI 1979-2009? Lets examine it.
Where should we expect UHI in temperature data 1979-2009?

Lets take a look at the “bible” of UHI Thomas Karls analysis of UHI 1901-1984 done on temperature mostly without temperature corrections done in the period of the global warming debate – and the biggest research of its kind:


Thomas Karls data suggests that increase in population number for cities already large has much less UHI effect generally than a similar increase in population number for a small town. That is, UHI should be expected NOT necessarily in the biggest urban areas of the world, BUT in the areas with biggest relative population growth.

So where are the areas with the biggest relative population growth?
They are here:

Population growth rate: I found 2 such statistics from the 1979-2009 period, upper effective in 1990 and the lower in 2006.
Result: North America, Europe and Australia has the lowest growth while Africa incl Sahara, and Brazil and more countries in South America and also large parts of Asia has the biggest population growth rate.
So lets take a look at the UHI expectations from fig 2 from Bob Tisdales data compared to the relative population growth maps. I have pasted the “UHI”- and “No UHI”-expectaions from fig 2, Bob Tisdales data onto the map showing relative population growth:

– The match with the relative population growth 1990 and 2006 is perhaps better than we could have demanded since many factors hypothetically could have made the picture blurry.

Bob Tisdales data seems to confirm that the divergence in GISS ground based data compared to UAH TLT could possibly be due to UHI, at least partly. At least if we expect UHI mostly from areas not just with a large Urban fraction but rather UHI mostly from areas with the fastes relative growth. And this sounds to me reasonable.

Update 25/1 – Russia/Asia:

The national population growth rate of Russia is near zero in the period, but Bob Tisdale found extra heat in GISS data compared to UAH over Siberia. This further supports the population related UHI problem because population over Siberia did in fact rise in the period, here a russian statistic from 1995:

But since most Russians live in the Euopean part where population has been declining, the national population numbers are not useful to justify the warmer GISS data compared to UAH over Siberia. So, population numbers suggests UHI over Siberia, but not the European part of Russia.

Then, my “UHI” over southern asia: The Tisdale Asia covers Siberia and southern Asia, but the extra heat in GISS compared to UAH is larger for this Asia area than just the Siverian area. Therefore, the Southern Asiea has the largest divergence GISS vs. UAH.

–> I will within a short time present a new article without any reference to Bob Tisdales results, but in stead a much more detailed analysis where also regional population data from all the wolrds largest countries will be used in addition to national population data.



Fig 7
UAH ocean temperatures , UAH land temperatures and the most often used SST´s has similar warming trend. The ground based land temperature trends have warmer trends.
This may indicate that the warmer ground based land temperatures has a faulty added warm trend from UHI, adjustments and siting problems.

Further more, the long term resemblance between UAH-land and UAH-ocean appears logical and correct since temperatures over land and sea respectively should have a tendency to seek equilibrium at least on longer term. This further points to Ground the based land data type as the source of errors.

Finally, this writing: It appears, that the UHI fraction in GISS ground based data is likely if we expect that UHI should be most significant in areas with high relative population growth. The rather good match between relative population growth and divergence between GISS and UAH TLT does in fact also a support to the general idea, that UHI plays an important role for the ground based temperatures often measure from cities and airports.

Finally we must add, that many other factors than UHI can play a role – see for example “post scriptum B”.

However, notice in fig 1 how the divergence between GISS groundbased data and UAH data just vanishes all together when we focus on an area with no UHI as McIntyres Peterson data suggests. This could indicate that UHI plays a role not that tiny for the descrepancy between UAH land data and Ground based land data.
* Post Scriptum A *

Where NOT to measure UHI…

Finally a little example I have to mention concerning UHI:
When speaking of “where not to measure UHI”, the absolute best spot in the world is London.
Central Southern England is one of the areas of the world with highest population density which makes the area one big urban warmed area. In Southern central England, no matter from where the wind comes from it will come from a highly populated area, and thus there are no truly rural areas to test UHI with. On top of this, ever since 1900, London has been a multi million population city, which is extraordinary. So, any attempt to measure UHI using London of all towns compared to a non rural surrounding area is a remarkable quest indeed.

Non the less, London of all places is one of the corner stones when skepticalscience argues that there is no UHI worth mentioning:
* Post Scriptum B *
In Fig 2 above, I have inserted a “UHI”” mark in the north Atlantic.
This is not the Faroe Island or the like, no its due to a GISS (LOTI) ground based data for north Atlantic with warmer trend than AUH TLT, Bob Tisdale:

Obviously this is an example where UHI does not play a part in the difference between GISS and UAH. For the oceans, the GISS (LOTI) are hadcrut SST data sampled 2 meter under surface. In contrast, UAH mostly relects low Marine Air temperature, and these to different temperature sets are likely to induce some differences.


In the period in question, the AMO current has sent stil warmer waters to the North Atlantic. My guess is that an event like a strong AMO rise in the North Atlantic might play a role. The AMO is know for warming up the air of the North Atlantic land and sea areas, and if the heat comes from the waters, the AMO current, then perhaps its not impossible that water temperature in the period is slightly warmer than the air? This is what it take if the warm AMO current is warming up the Northern atlantic land areas etc. Just a guess, obviously.
* Post Scriptum C *

I have taken the liberty to write “No UHI” on the maps for Australia. In fact Bob Tisdale got GISS to have lower warming trend than UAH TLT on his “Australia”.

In my maps for relative population growth, I only have full countries, but Bob Tisdale chose a fraction of central Australia, not the whole Australia. but if we use all of Australia the GISS trend is much warmer.

Heres a compare of the GISS (LOTI) temperature trend for Australia 1979-2009 [little picture] compared to recent development in population [big picture] – and we see that the actually slightly falling GISS trend for central Autralia happens to be accompanied by declining population in central Australia:

However im sure that world wide one could find many areas where the match between population and heat trends will be poorer than this due to a row of other factors.

Why global mean temperature is not a valid scientific measure for global climate change

In Denmark we have an ongoing discussion on whether Global Mean Temperature  GMT is a valid proxy for global heat balance/ global warming. Professor in thermodynamics, Bjarne Andresen, explains that one cannot simply add temperatures and then divide them to get at mean temperature in a non equilibrium thermodynamic system.
Asymmetric Hemispheres

Fig 1.
Our planet is asymmetric. In the Northern Hemisphere 39% of the area is land.
In the Southern Hemisphere only 19% is land. This means twice as much continent on the NH as on the SH
Due to the heat capacity of the oceans and the huge amount of water moving vertically and horizontal oceans absorb almost 90% of all absorbed energy on Earth.
Antarctica makes the Southern Hemisphere very cold because of ice albedo , cold winds , and sea- and glacier ice cooling the air and surface of the ocean. The waste ocean area also results in a lot of cooling because of evaporation.
Whereas a reduction in ice albedo in the Northern Hemisphere when exposed to increasing temperatures makes the Northern Hemisphere more vulnerable to global warming.
This means you need a lot more energy to warm up the Southern Hemisphere by 1 degree Celsius than you need to warm up the Northern Hemisphere by 1 degree Celsius.
Maps showing actual warming and models of global warming illustrate how the Northern Hemisphere warms up much faster than the Southern Hemisphere and several degrees Celsius more if exposed to the same amount of increased infrared radiation.
Map illustrating temperatures at the end of this century caused by increased emission of greenhouse gasses.

Fig 2.
Source: DMI  Danish Meteorological Institute.
Map showing a more sensible Northern Hemispheric reaction to temperature change.

Fig 4.
Between 1910 -1940 most of the warming up took place in the Southern Hemisphere whilst most of the warming up during 1975-2005 took place in the Northern Hemisphere.
This means that globally more heat was needed to warm up the lower troposphere early in the century than later in the century. We are told by IPCC that warming up between 1910 -1940 is allegedly natural whilst warming up during 1975-2005 is said to be anthropogenic . This is very interesting. We are always told that the actual warming up is unprecedented.
To illustrate the case I will make a small calculation.
If the Northern Hemisphere warms up 1 degree and the southern atmosphere cools down 1 degree you have a temperature sum of zero. But actually you loose
energy, because it requires more energy to restore the energy in the south than you get from warming up the north.
Heat balance is a better proxy for global warming but it is very difficult to estimate.
On the other hand global mean temperature is so very biased that it has very little relevance when you wish to describe changes in global warming.

Birger Wedendahl,


– A few comments by Frank Lansner:
In the Atlantic Ocean we have a possibility to redistribute heat from South to North.

Fig 5.
As Birger Wedendahl mentions, there is much more land area on the Northern Hemisphere than on the Southern Hemisphere. Heat that is transported to the Northern Hemisphere is likely to spread out over the vast land areas. When measuring global temperatures as a simple mean of the surface area, then a simple redistribution of heat from South to North is then likely to appear as a global warming trend.
The AMO index that describes the heat of the Northern Atlantic appears oscillating, at least in the 20´ieth century:

Fig 6.
According to the AMO, one might expect extra “global” warming around year 1940 and around year 2000.
So, the mix of land area and ocean area used for simple temperature mean for the globe might enable global temperature fluctuations simply by redistribution of heat.
An area more comparable to take a simple temperature mean would be the oceans only. (or land, however, Oceans obviously have more stable temperatures). So, perhaps Ocean temperatures are a more proper indicator of temperature trend of the earth:

Fig 7.
Assuming the 70% of the planet – the Oceans – gives a better impression of the warming trend, it becomes obvious that the global warming 1940-2010 was only 0,25-0,3 K.
Further more, we see that the “global warming” was much faster 1910-40 than at any point after 1940….   !

Decline: Temperature decline 1940-78, the cold data-war

Next to the historic data war on the Medieval Warm Period, the data war on the 1940-78 perhaps stands as the most bitter and intense climate disagreement.

This is a follow up to the article: Temperature corrections of the northern hemisphere.

(Thanks to Bo Vinther – “neutral” in the climate debate – who was a great help digging up relevant data)


Changes of temperature data seems to occurs in several areas: Temperature station data, balloon temperature data, SST data, tree ring data, program fudging, cheery picking of data by scientiests, cherry picking of scientists by IPCC etc.etc.

Fig 1.

What happened to the great temperature decline 1940-78?

How did version A of 1940-78 temperatures change into version B ?

Fig 2.

This little quote is interesting because when IPCC defends cutting tree graphs from 1960 mostly, then the argument is :

“Fall in tree ring data matches real temperatures 1940-60, but after 1960 the fall in “real” temperatures cannot match tree ring data”.

But, in the quote in fig 2, 1964-74 is described as the sharpest temperature drop.
In general, the biggest differences between old pre-1980 temperature sets and new post-1980 temperature sets are located in the years 1958-78. For some reason, just when satellite data starts in 1979, the differences get smaller. No one really corrects temperature data far from satellite data, it seems… – and then we have the argument: “Land temperatures matches Satellite data”.

The present article is a follow up on an article where we presented the difference between temperature data from mid 1970´ies, Newsweek/NCAR/NOAA/NAS/National Geographic and then recent CRU temperature data, Northern Hemisphere:


To make some points below I need to mention the Raobcore temperature measurements. These are balloon based temperature measurements that begins in 1958. The Raobcore measurements thus covers around half of the decline period 1940-78. How about the quality of Raobcore surface temperature measurements? – The stunning match between Raobcore and satellite measurement is a tremendous proof of quality balloon data as well as satellite data:

Fig 4.

So it is truly defendable to regard the Raobcore measurements as a high quality product telling us about temperatures all the way back to 1958. The best illustration of Raobcore data before 1979 made easy available on the net is the Tropics 30S-30N – Steve McIntyre presented these data here:
(All temperature data sources shows more cooling in the NH than the SH for 1940-78, and therefore using tropic in stead of NH Raobcore temperature data is not increasing the cooling in data 1940-78).

Fig 5.

As usual in the climate ”science” we see that the most recent versions of adjusted data happens to show more warming trend. But still, Raobcore is dynamite. We know that all sources of temperature data confirms steady temperature decline 1940-58, but Raobcore confirms the ongoing significant fall of temperatures 1958-1978. We cannot see the whole decline 1940-1978 from Raobcore data, but for 1958-60 we see same level of temperature as in the 1990´ies.
– And just after 1958, the North pole in 1959-62 looked like this:

Fig 6.

Now lets go back to the National Geographic, “Mathews 1976” temperature set.

Fig 7.

The Mathews 1976 / national geographic temperature graph are Recorded changes of annual mean temperature of the northern hemisphere as given by Budyko (1969) (temperature data 1880-1960)

– and as updated after 1958 by H. Asakura of the Japan Meteorological Agency using 1958-75 temperature data by Angell and Korshover.

It appears that temperature data used by National geographic 1976  is based on peer reviwed data.

First lets take a look at the years 1958-75, Angel and Korshover. In fact the writing above lists temperature data sets from 4 different scientific writings. Notice the almost identical slope for all 4 Fig 8, and then the nice match with Raobcore Fig 9:

Fig 8 + Fig 9.

In all cases incl. Raobcore we see trends that confirm temperature decline 1958-75 around 0,3-0,4 K, roughly estimated.

The Korshover data is taken from 63 radiosonde stations around the globe:

Fig 10.
I dare say this work is indeed showing results from a larger absolutely serious and professional project. Impressive.

Budyko data up to 1960 is gathered monthly temperature anomalies carries out by the Main Geophysical Observatory.

Stitching Budyko and Korshover: It so happens, that the two datasets has a minor overlap period around 1958-60. Fortunately in both temperature sets around the overlap years 1958-59 appears are rather constant, and therefore the stitch appears pretty straight forward, not truly risky.

It seems that Korshover uses the same zero anomaly as Budyko, since the 2 data sets “melts” together when stitching for same zero anomaly (but I cant see what baseline years the zero anomaly is defined for).

The correctness of the stitch is further supported: We can have rather high confidence in the temperature trend of 0,3K 1958-75 as shown above from 5 data series. Therefore the resulting graf must have 1958-59 around 0,3K higher than 1975. The 1958 (and 1959) points of Koshover data is right on the trendline and therefore appears solid to use. Any significant error from the stitch should most likely come from using a wrong point at the Budyko graph.

The stitching is further supported, see fig 14. Hansen (1980) has approximately the same trend of data 1957-65 as National Geographic. Hansens trend changes significantly from NG mostly after 1965.

Fig 11.

Addition: There are 3 points where you could stitch Budyko and Korshover, 1958-59-60.

The1958 and 1959 korshover points are spot on the trend graph and thus represents the overall Korshover (and Raobcore etc.) far best. The 1960 point of Korsover is far longer from the trendline and thus represents korshover poorly. The Stitch wa carried out by the Japaneese Meteorological institute, and their apparent chocie o 1958 seems logic and correct.

UPDATE, Stitch confirmed by Yamamoto 1975 NH surface temperature:

No doubt, the stitch is where National Geographic 1976 data might be attacked by the critics, but as I´ve explained it looks like a rather reasonable stitch.

From Mitchel 1961 we find another temperature dataset that (with some minor deviations)appear to confirm the Budyko results:

Fig 12. (Global data, not NH – but SH has smaller temperature decline than NH 1940-78, so this does not explain this decline.)

Mitchel has temperature decline 1940-58 around 0,26 K.

So all in all with other data sources (Newsweek/NCAR/NOAA etc.) mentioned, ”Mathews 1976” appears to be a rather valid piece of science based mostly on peer reviewed work, in agreement with Raobcore data which eventually confirms the temperature decline roughly 0,4 K from 1940-78.

So why cut the tree graphs and thus hide the FULL 1940-78 decline?

Fig 13.

Finally. Lets checkout what Hansen did to his data set, GISS.

It appears that Hansen in the early 1980´ies did not use the above data, except for the Mitchel graph that ends in 1960.

Hansen had high quality Raobcore measurements for his disposal, The 4 series incl Korshovers 63 radiosonde stations, all data that yielded a common trend of approximately – 0,3 K 1958-75, but heres Hansens new data:

Fig 14.

For the years 1958-75 Hansen/GISS (1980) only finds a temperature decline of approximately 0,07 K (red) in stead of the 0,3 K decline (blue) from Raobcore, Korshover etc.
Hansen/GISS 1980:

Fig 15.

Hansen/GISS 2007:

Fig 16.

Fig 17. (NG 1976 is of course NH data whereas Hansen/GISS is global)

And we remember:


Here Steve McIntyres comparison of mostly GISS graphs. I added the 1980 GISS from fig 15:

See more:

Temperature corrections of the Northern Hemisphere

Most skeptics are aware, but it cannot be repeated too often: Temperature data presented before the global warming movement really started in the mid 1980´ies compared with recent official temperatures shows that the temperature trends 1940-1978 has been changed fundamentally.

I believe the best way to show this is simply to compare temperatures of the largest possible temperature area from before the 1980´ies with the same area presented today.  The area best suited for this is the entire Northern hemisphere. Temperatures in National Geographic 1976:

(Alternatively NH temperatures from Stanley 1975 )
Now compare the 1935-1975 decline for the same area – the entire Northern hemisphere – presented by CRU/Brohan 2006:

Below, the two NH temperatures datasets above (National Geographic 1976 vs 2008) are shown together:

Below, the two NH temperatures datasets above (National Geographic 1976 vs 2008) are shown together:

History has been rewritten.

Scandinavian temperatures, IPCC´s “Scandinavia-gate”

In recent years the Swedish scientist from Stockholm University, Karlén,  has tried to create attention to the fact the Scandinavian temperatures when represented by IPCC cannot be recognized in the real data from the Scandinavian temperature stations:

fig 1
Left: Karlen made a plot of 25 data series from the NordKlim database.
Right: IPCC´s temperature graph for the area does not reflect the actual Scandinavnian temperature graphs.

IPCC shows temperatures around year 2000 should be approximately 0,7 K higher than the peak around 1930-50, whereas the actual data collected by Karlen shows that year 2000 temperatures equals the 1930-50 peak, perhaps even lower.

Was Karlen wrong? To evaluate this, lets check out the National meteorological institutes of the respective Scandinavian countries:

fig 2
Only Denmark shows slightly higher temperatures around year 2000 than in year 1930-50. 0,1 – 0,3 K warmer? However, the Danish Area around  3% of the overall area. For the vast majority of the Scandinavian area shows year 2000 temperatures just like the 1930-40 peak, Sweden maybe a tiny fall, Norway a tiny increase. Denmark is also the area of Scandinavia with far highest population density, and thus Denmark is likely to show more City heat effects (UHI) than the rest of Scandinavia.
So, With good confidence, we can say that Karlens data from Nordklim matches the opinions of the highest authority on Scandinavian temperatures. The very significant temperature peak around 1930-40 has been reduced almost removed totally. And here Scandinavian Islands that to some degree also represents Sea temperatures – and due to their lower populations are more free of any potential City heat (UHI). Here data fom SMHI:

fig 3
Scandinavian Ocean temperatures indicated from Iceland, Jan Mayen and Faroe Islands actually shows a clear pattern of lower temperatures in year 2000 than in around 1930-40. In general we see: The further from population, the cooler temperature trends.
I found on the net temperature data from Kategat and the north sea (“Skagerak”). This ocean area is placed in the most populated area of Scandinavia, but still, no measurements where taken in cities, obviously:

fig 4
Again, no IPCC-warming here either. (
For Finland I also found these data, which definitely shows colder year 2000 than 1940:

Here Scandinavian temperatures 1900 – 2000 cleaned from trend lines. Still no sign of global warming in Scandinavia.

The “NEU” area.
The “NEU” area is different from the Scandinavian area so to understand the IPCC warming trend over Scandinavia, we have to examine the “NEU” area.
The NEU area is defined as -10W/40E x 48N/75N:

Fig 7.
The exact area of “NEU” is for some reason (?) not shown in the legend right under the figure 9.12 in the IPCC report AR4, but in an appendix later in AR4.
So the few IPCC AR4 readers that actually uses time to investigate the Scandinavian discrepancy will see that the actual area from the “NEU” for some reason does not stop around 55N, just between the NEU and the SEM graphic as would be expected. No, for some reason, IPCC has chosen to include areas far down in the “SEM” area on their graphic.
What is the consequence of expanding the NEU area further south than 55N? And why not stop at a round number 50N? Why 48N? Here are temperature trends for Hadrcrut 10W/40E:

fig 8
The 2 upper graphs made using hadcrut gridded data 5×5, the lower by hadcrut stations 48N-50N. So, there is more and more heat trend the more southern areas we include in “NEU”.
Second, we see that ”NEU” on the IPCC graphic includes the big Norwegian Island, Spitsbergen, but when checking out, this “NEU” area does not at all include Spitsbergen for calculating temperatures. (There is no warming trend 1940-2000 for Spitsbergen.)
The last point of IPCC´s “NEU” graphic is year 2000. However, this point represents data 1999-2005. So, in my case especially Swedish national temperatures are hard to find on the net for later years than year 2000. SMHI does show long temperature trends for 2 larger cities only(!) that shows clear warming trend. They – unlike overall Swedish temperatures that shows no warming – are easy to find. Uppsala and Stockhom SMHI shows. For Finland, I found overall national temperatures AFTER 1950 on the official national meteorological site. The peak around 1940 was not shown – “hidden” perhaps. However for Finland its rather easy to find national temperatures on the net other places than their national meteorological site.
The averaging method:
The IPCC graph is hadcrut data (Brohan 2006), so how can IPCC show hadcrutdata so surpricingly rid of the decline after 1940? It appears impossible, as some readers suggest. This is due to the averaging method. IPCC used a rather different area than shown on their graphic for the (strongly corrected) hadcrut data. On top of this, the IPCC have chosen an averaging method of ten years collected in one point. And the peak around the middle of the 1930´ies has thus been diveded by separating in 1926-1935 and 1936-1945. The peak is thus almost gone.
Here is a normal 10 yr running mean curve for -10W/40E x 50N-75N:

fig 9
And see the “magic” going on when dividing the 1930´ies peak into 2 10-year periods: It gone!

fig 10
The cold peak around 1942 seen in fig 8 is strongest in the 48N-50N band.
Where doe this bring us?
IPCC has put a warm-trend curve over a huge area of Scandinavia (that has no warming trend in original) because:
1) The NEU area does not include the full area incl Spitsbergen even though the graphic shows this.
2) The NEU area reaches down and uses warmer trends from the higher populated areas in the middle of Europe (down to paris, Wienna etc) – an area that according to their graphic should belong to their “SEM” data and not “NEU” data.
3) Their averaging method splits up the 1930´ies warm peak so we cannot see that it was just as warm around 1935-40 as it is today. Obviously, for the whole debate of global warming such an information is relevant and should appear to the reader.
Summa: IPCC does not show directly what “NEU” is by placing this information in a distant appendix. In reality, all readers that does not use hours to investigate this will get an impression of a warming Scandinavia and this is indeed misleading.

So without “doing anything wrong” IPCC misleads perhaps 99% of the readers?