Regime Change in Alaska

In Hansen’s GIStemp product, we regularly see a map with the entire Arctic colored Blood Red. With “heating anomalies” in the 4 to 10 C range.

It’s a remarkably consistent feature in their graphs.  Only with carefuly selected maps can you make it go away.  (Things like changing the baseline help; as does using a very small time window).  But the basic image is one of Siberia and Alaska just melting away to become a new Tropical Paradise.

Yet there are clearly times when the reports of folks who live up there, and of the weathermen on TV, say it’s quite cold.

So “What’s the deal?”

Here is the latest map from GISS via GIStemp as an example:

GISS Anomaly Map for USA in March 2011

GISS Anomaly Map for USA in May 2011

What happens if we check in on a station in that hot band, like Barrow Alaska, on the very northern edge near the middle of Alaska, last stop before you are wading in the Arctic Ocean:

Barrow Alaska Wolframalpha

Barrow Alaska Wolframalpha

Well, that’s a bit of a surprise! Dropping by almost 1/10 F per year and taking only 10 years to drop a whole degree F. Golly! Except it looks to me more like it was a ‘step function’ down about 1975 (and with a potential step back up about 2010, though more muted). This looks more like a “Regime Change” profile with about a 30 year period and not so much like a smooth even “trend”. But, by “Climate Scientists Standards”, it’s a trend, so I’ve got to take it ;-)

Ok, so Barrow is cooling and the Arctic is On Fire Hot. Something doesn’t add up here…

Notice that The Arctic is shown as being Very Red. This is the default map you get right now at GISS. I’ve not “cherry picked” anything, just taken what the “Average Joe and Jane” will see if they say “Show Me A Map”. It is what is released by NASA each time they declare some month or other “The Hottest EVER!!!!”. Just “the latest month”.

Notice that per the legend all of The Arctic is about 4 to 9.9 C over heated. That’s like 7.2 to 18 F Hot Hot HOT!

Somehow if it was 15 to 20 F hotter up there, I think folks would notice…

Well, we can tell GIStemp to “only” smear data around by 250 km instead of by 1200 km and get a bit of an idea. On this ‘lesser smeared’ map we can see that there aren’t all that many temperature stations in The Arctic in GIStemp (or the GHCN it is based upon).

GISS GIStemp 250 km "smoothing" Mar 2011

GISS GIStemp 250 km "smoothing" Mar 2011

Here we see that the Big Red Blob extends most from Alaska into Siberia, with Honorable Mention to a stray station or two in Greenland and Canada (like the one in “The Garden Spot of the Arctic” we saw here in Eureka Canada)

So, what happens is we “dip a toe” into some temperature histories from around Alaska and Russia? Do we find that we have a nice steady 1/2 C / year rise? (To give us that 20 C of Hot Hot HOT!!! in about 40 years, as is implied to be happening in the GIStemp product, with that 1951-1980 baseline). Basically, if we “step off the reservation” of Their Method and Their Selected Stations and Their “time scope”; do we see the same thing? Is the effect in the DATA or is it in the METHODS and SELECTION BIAS?

Russian Pacific Coast

First up, Anodyr Russia. Why Anodyr? Well, it’s just across the ocean from Alaska. Heck, I bet Sara can even see it from Todd’s fishing boat if she tried real hard! ;-)

You can read up about it here:

Anadyr (Russian: Ана́дырь, Chukchi: Кагыргын, Kagyrgyn) is a port town and the administrative centre of Chukotka Autonomous Okrug, the extreme north-eastern region of Russia. It is at the mouth of the Anadyr River, on the tip of the southern promontory that sticks out into Anadyrskiy Liman. At 177°30′E, Anadyr is the easternmost town in Russia (more easterly locations, such as Provideniya and Uelen do not have town status by the Russian definition). It was founded on July 21, 1889 as Novo-Mariinsk, and renamed on August 5, 1923 following the Kamchatka Revkom. Town status was granted on January 12, 1965.

Now, GIStemp has one view of Anadyr, but Wolframalpha had another:

Anadyr Russia per GIStemp May 2011

Anadyr Russia per GIStemp May 2011

This is the “combined stations” data.

Not much happens until about 1990, then all the low going cold excursions “get clipped” and we add some more warm spikes. We’ve seen that before and it is coincident in time with when new “Quality Control” methods were started that force stations to be like those around them or they get replaced with an average of ASOS stations at airports. An average of stations, and especially those over black tarmac in the sun, will never be as cold as a single station. Averages always flatten the extremes. (To get that extreme event would require ALL the stations to have a valid low excursion extreme on the same day at the same time. A highly improbable event). So is this an actual uplift in temperatures at the end, or a result of a flawed QA “adjustment” process?

You will notice I’ve “tagged” this graph with the creation date. The structure of GIStemp is such that every month you will get different results. It rehomogenizes everything and any new temperatures change all the old temperatures, so you must know what “vintage” you have to know how it will look. But this is the view today. Who knows what it will look like next year, and it isn’t possible to know what it looked like on any given time in the past as “things change”…

And what happens after GIStemp “Cleans and homogenizes” it:

Russia Anadyr 1900 to 2011 GISS Cleaned and Homogenized May 2011

Russia Anadyr 1900 to 2011 GISS Cleaned and Homogenized May 2011

Not much difference as near as I can tell.

But what does Wolframalpha think it looks like?

Anadyr Russia per Wolframalpha

Anadyr Russia per Wolframalpha

Well, that looks rather flat and constant to me…

Wonder where all that Hot Hot HOT!!! is coming from, then, if it isn’t actually IN Anadyr?

Could it be across the water in Nome or up in the Barrow Alaska area above the Arctic Circle?

North, To Alaska!

This time we will start with the Wolframalpha data:

Nome Alaska Wolframalpha

Nome Alaska Wolframalpha

Wow! Quite a Rise! Yet it doesn’t look like a ‘trend’ so much as a ‘Regime Change’ in about 1975. The temperatures jump up in one step, then hold flat. Ten whole degrees in one go, then the “peaks” just keep hanging out about 30 F and no higher. Not “warming trend” so much as “regime change”. We can even see a hint of cyclicality in the past high points (in the part of history that gets ignored as it is before the GIStemp baseline…)

Now lets look again at that Barrow Alaska graph:

Barrow Alaska Wolframalpha

Barrow Alaska Wolframalpha

Well, here, too, we seem to have a bit of “Regime Change” but this time it’s a drop. Again in about 1975 we take a plunge of about 10 F and then hold there (though I notice that about 2010 it looks a little like a possible regime change back is in progress, just as Nome is ‘rolling down’ Barrow is ‘rolling up’ a tad. Guess it’s time for GIStemp to drop Nome and use Barrow to recreate a warming Nome with The Reference Station Method…)

What’s going on here? A big “Dig Here”, but I’d guess it is one of the normal weather cyclical patterns. AO, PDO, whatever. At this point I punt to the Weathermen of the world who know this stuff far better than I ever will. Maybe someone can get Anthony Watts to research what happened during that Regime Change in Alaska and the Arctic.

What is clear to me is pretty simple: There is not A Trend in the Arctic. There are distinct and individual trends by location. Any “trend” for the whole place is an artificial construction from too much “homogenation” and too much “adjustment”. Barrow Alaska alone says that, but Anodyr confirms it.

Some Detail on Nome

Here is a more detailed look at Nome, for completion, mostly. You can click on the graphs to get larger or more readable sizes.

Nome Alaska May 2010 2011 from Wunderground

Nome Alaska May 2010 2011 from Wunderground

So Nome has some volatility, but on average the temperature line is staying inside those temperature MIN and MAX averages. It’s just being “normal” in its range.

What does GIStemp do with the data as it moves through the digestive process?

First we have more than one station, so they get combined.

Alaska Nome GISS individual station data May 2011

Alaska Nome GISS individual station data May 2011

Notice that there is a tiny dashed “tail” hanging down about 1972 or so. Can’t have any low excursions, so that is removed in the “combining” process. (I’m sure there are more, but that one stands out):

Nome Alaska Combined Stations GISS May 2011

Nome Alaska Combined Stations GISS May 2011

OK, all nicely combined, now to “Homogenize and bake until done”:

Alaska Nome GIS cleaned homogenized May 2011

Alaska Nome GIS cleaned homogogenized May 2011

Look at that 1920s area. Notice anything? Check the scale… Yup, it’s moved DOWN from about -5.8 F to about -6.8 F. Just about one whole degree more “warming trend” baked in the cake. Baked Alaska anyone?

A bit harder to pick up, in about 1980 there is a peak. It moves up by about 1/4 to 1/3 of a degree F. I had to lay a ruler on the screen to see it. But it’s there.

More Regime Change

We see a similar shift of Regime Change in Kotzebue Alaska a bit further north.

Kotzebue Alaska GISS May 2011

Kotzebue Alaska GISS May 2011

The Wolframalpha data are a bit short, but you can see the tail of the rise out of the cold interval, then the nearly flat time since then:

Kotzebue Alaska Wolframalpha

Kotzebue Alaska Wolframalpha

Peaks hanging consistently around that 25 degree line. No trend here, just Regime Change.

Back To Russia and a Lower Alaska

So what happens if we slide down the coast just a bit on each side of the Pacific? To a place in Kamchatka Russia? (While GISS and Wolframalpha use different spellings, the lat / long matched):

Russia Kelyuchi WA

Russia Kelyuchi WA

And the GISS View:

Keljuci Russia Homogenized GISS May 2011

Keljuci Russia Homogenized GISS May 2011

I find it fascinating that even GISS does not show a warming trend for this station, yet it is Blood Red On Fire in their map… Though I do note that the low going excursions have been ‘peak clipped’ after about 1980. Mostly it has simply been truncated so that it can be “filled in” from somewhere “nearby” up to 1200 km away. Perhaps somewhere with a more suitable warming trend…

FWIW, in Juneau Alaska we get a similar cut off of data, but they have a very nice “dip” in the baseline period…

Junea Alaska GISS May 2011 Homogenized

Junea Alaska GISS May 2011 Homogenized

In Wolframalpha, you can much more clearly see that this is just a Regime Change of some sort, and things have just not warmed up since then.

Juneau Alaska Wolframalpha

Juneau Alaska Wolframalpha

But boy, do you get to “lock in” a spectacular rate of “rise”. A 2.7 years / F run rate. Why, in 27 years that’s 10 whole degrees! In 50 years, it’s 20F of Hair On Fire Trend!!! To bad it isn’t actually a trend, just a data artifact.


I was looking at the upper middle of Siberia and ran into Dudinka. It is also interesting. GISS has a dip in the baseline, and a rise out of that, with some data truncation. The actual trend is far different:

Dudinka Russia GISS Homogenized May 2011

Dudinka Russia GISS Homogenized May 2011

Dudinka Russia from Wolframalpha

Dudinka Russia from Wolframalpha

Again we see that consitent pattern of a warmer time in the 1930-1940 era. A “dip” in the 1951-1980 baseline time (though a modest one). Then the shift back up at the regime change point in 1975 or so, and basically flat since then.

Tura Russia is a similar example:

Tura Russia GISS Homogenized May 2011

Tura Russia GISS Homogenized May 2011

Cold peak clipping after the late 1980’s – 1990 period. Bit of a cool dip in the baseline period, but prior to that warmer.

Tura Russia Wolframalpha

Tura Russia Wolframalpha

Other than a1960-70 cool dip, it is basically all a slow down trend.
In that Ruby Red Russia on the GISS map…

Looks like Russia has a Regime Change problem too ;-)

In Conclusion

So now we can see what the net-net effect of GIStemp is, simply and clearly.

IMHO, the “Reference Station Method” clearly breaks down in places subject to weather Regime Changes on a 30 to 60 year cyclical basis. When combined with the peculiar kind of splicing called “homogenization” and especially after a “Quality Control” process that tosses out low temperatures, it creates an artificial (and locked in) spectacular “Hot Hot HOT!!!” reading where actual stations on the ground show one of two very different states. Either gentle cooling (or not much) or a Regime Change with a step function up in some places and down in others. Exactly the kind of thing where station drops would result in “exactly wrong” filled in data via the Reference Station Method.

The reason The Arctic is shown as Blood Red On Fire is due to processing and data handling errors masking the cyclical weather patterns and regime change in the area. IMHO, most likely an artifact of the AO Arctic Oscillation changing when the PDO flips and flops.

But at this point I must leave it as a “Dig Here!!” for the Weathermen of the world. They’ve spent a lifetime getting good at just that kind of thing and I have other places to explore ;-)

Update: Polar Projection

As it’s just a different dropdown choice at NASA GISS, here is the same map as a polar projection:

GISS May 2011 Polar Projection land

GISS May 2011 Polar Projection land

Here is the version with the HADcrut sea model included.

GISS Polar Projection w/ Ocean

GISS Polar Projection w/ Ocean

The startling thing is how empty the Southern Hemisphere is when you cut it back to 250 km “smearing”:

GISS May 2011 250 km Smear range Polar Projection Land only

GISS May 2011 250 km Smear range Polar Projection Land only

Subscribe to feed


About E.M.Smith

A technical managerial sort interested in things from Stonehenge to computer science. My present "hot buttons' are the mythology of Climate Change and ancient metrology; but things change...
This entry was posted in AGW GIStemp Specific, AGW Science and Background and tagged , , , . Bookmark the permalink.

13 Responses to Regime Change in Alaska

  1. Pascvaks says:

    Hansen’s GISS/GISTemp is beyond repair. Found it much easier to get a feel for polar temps, etc. at places like this –

  2. Pascvaks says:

    (EM – All links in “Recent Posts” have https connection.)

  3. E.M.Smith says:


    Again, I’ve “done nothing” so I think “it isn’t me”…

    I just opened this page and it is “http” no “s”…

  4. John F. Hultquist says:

    Another interesting post. My feeling is that the procedures used seemed to make sense when they began and automated the GISS output. As shown herein and elsewhere, something has gone terribly wrong.

    I also dislike the map projection. Pole ward of about 50 or 55 degrees latitude this map becomes a lesson in how to mislead with cartographic fiction. Again, this may not have been intentional as the band of Earth between about 50 North and 50 South is not too out of whack. They appear to use something like a Miller Cylindrical projection, the sixth one in the list here:

    They ought to snip these maps off and not use them to show the polar regions. On the map you post there is a great red blob in the North and a somewhat smaller blue blob in the South. Both are so large as to be laughable; or choose another word. These areas ought to be shown on a Pole-centered Lambert Azimuthal Equal Area projection. Link is at the top of the site.

    The issue becomes that of using 3 maps to cover the Earth instead of one. Is that too much to ask?

    One also ought to have a world globe beside their computer just to remind one, on occasion, just what Earth looks like. They are often available at garage sales for 2 or 3 bucks.

  5. E.M.Smith says:

    Oddly, I happen to have a 9 inch globe within arms reach most of the time. I end up using it rather often…

    I love the “tech” behind the various map projections, but most folks just glaze when I start talking about things like plotting a heading as a straight line and having it match a compass heading, or having the area accruate but wrong shape vs the right shape but wrong area vs…

    But, since you brought it up, with a click of the dropdown, the GISS site gives a polar projection (but you must ask to get it…) so I’ve added them above.


    (Amazing how empty the Southern Hemisphere is…)

  6. John F. Hultquist says:

    Not having worked with the site I did no know they provided the polar projections. Thanks!

    (Amazing how empty the Southern Hemisphere is…)

    Even with the 250 km smearing there is a blue circle (except for part of one of the white wedges) around the South Pole. This circle appears to include the area south of 80 degrees latitude. At 69 miles per degree and 10 degrees the circle has a radius of about 690 miles. As 250 km is just over 155.3 miles, even a station at the S.P. would not show as such a large area. On the second world map, this area shows as a thin rectangle at the bottom along with a small overlapping white one – also now a rectangle.

  7. Gary says:

    High temp of 122 F in 1981 in Barrow Alaska??? Wonder how that happened?

  8. E.M.Smith says:


    Maybe the sun got stuck up high for months on end and didn’t set…

    @John F. Hultquist:

    Yeah, IMHO they are a bit “loose” in their definition of 250 KM…

    the ‘smearing’ is how far an empty grid will reach to pick up a temperature from somewhere else. The may be done a couple of times in a row in some circumstances… so an individual station may find the data used even further away.

    THEN, the “grid box” gets filled in. Based on the center of the box… So if the “grid box” is a given degree size (don’t remember exactly what they are, but originally 8000 now 16000 per planet) and a station is less than 250 KM from the center of the box the whole box gets filled with that temperature….

    So lets say a station existed 200 KM from the south pole, but has not had data for a year. Being as the polar station is inside 250 KM, that station will have “homogenized” data filled in for what is missing. Then, in a later step, the “fill in the grid boxes” step may have a Grid/box that is 250 KM on center away from that station. That whole grid box gets colored in with that temperature from the fictional data from that ‘near enough’ station…

    That’s what I’ve talked about as the “recursive application of The Reference Station Method”…

    In the case of the South Pole, I think the “grid box centers” are inside the 250 km of the pole distance, so all of them always get the polar temperature over the entire length of the boxes, well beyond their centers….

    One of the more subtile “little lies” of GIStemp… I’d noticed it before, but not found a good way to comment on it so it mostly goes unmentioned…

    Now imagine what happens when you use the 1200 km recursive smearing… You could fill up the entire Arctic Ocean where there is nothing but ice with Blood Red fictional heat from a few minor stations with UHI and Regime Change issues…

  9. Earle Williams says:

    The Geophysical Institute at the University of Alaska – Fairbanks maintains a page on time series weather data statewide here:

    This page says it hasn’t been updated since 2008, but if you drill down into a site and look for example at the average temperature data it’s been updated (at least partially) through February, 2011.

    With the sole exception of Barrow, every first order site across the state shows the Great Pacific Shift of 1976. Temperatures over the last 30 years have gone nowhere.

    I suspect that the apparent warming occurring at Barrow is due in part to a heat island effect, which has been studied and published. I dug into the data available at Deadhorse from Weather Underground and recall that there was slight warming over the last 30 years. For 1998 – 2008 Barrow showed a 0.14 F/yr warming while Deadhorse, Kaktovik, and Ivotuk were all negative.

    I don’t doubt that the Great Pacific Shift isn’t manifesting along the Beaufort Sea, but I’m not convinced that the Arctic is quite the hot spot as is protrayed by GISS and others.

  10. Level_Head says:

    It seems logical, for an anomoly-based system, to simply leave out stations that do not have data. Don’t blend in anything else — just contribute a station’s anomoly from zero if known, and if not, blow it off.

    There are subtleties — how to calculation baselines when few stations have full data sets to work with, for example. But these have been worked out, and essentially could be simplified into the baseline period’s temperature reading when temperature was available, turned into the new offset value. And done by month to reduce the effect of a missing July, or December.

    No smear at all. Small gridcells — useful only because they have histories in those same geographies. But don’t take the temperature where it isn’t.

    Then we run these grinds again and see what they say.

    Even the equatorial equalizing of distribution would be helped, as a northern station no longer being tracked would simply stop contributing whatever anomalies it saw. If it never made it to the baseline period, that’s too bad. Perhaps a weighting — a proportionate reduction of anomaly based upon how much baseline period was missed — would be appropriate. A two-degree “anomaly from baseline” for a station that missed half the baseline would be one, instead. Or something like that. The read baseline timeframe was an interesting period.

    And newer stations (that were baselined) would likely have UHI issues, but would not be forced into comparison with any other station.

    ===|==============/ Level Head

  11. Edward says:

    Hansen is not attempting craft here, these ‘maps’ are purely a propaganda exercise, vintage Jim Hansen and because of his track record should be regarded as unscientific trash.

    Jim is a fully paid-up political agitator [he knows – on which side his bread is buttered] – an advocate, he dropped the science years and years ago.

  12. RACookPE1978 says:

    “You will notice I’ve “tagged” this graph with the creation date. The structure of GIStemp is such that every month you will get different results. It rehomogenizes everything and any new temperatures change all the old temperatures, so you must know what “vintage” you have to know how it will look. But this is the view today. Who knows what it will look like next year, and it isn’t possible to know what it looked like on any given time in the past as “things change”…

    So, deliberately, the trend cannot be analyzed or noticed by anyone EXCEPT those who can afford the time and talent go behind the data and recalculate it with their own program.

    He (Hansen ) re-calculates every month every past “datapoint” and llocation in his dataset – thus – gets a free pass to recalculate and replot his values to show ever higher and higher trends. Trends which suit his purpose and his pocketbook and power.

    All this from a paper he wrote 30 years ago claiming that trendlines can be extrapolated out to 1200 km from their source. On a 0.53 coorelation of data trends.

    He is fixated on “smearing” his data over as wide an area as possible and artificially creating “new” data to show as long a period as he can create of “continuous data” … Thus, rather than accepting “starts and stops” when information drops off for one month or two, he blends what used to be data from “nearby” sites and reprints the series “as if” were raw data.

  13. Pascvaks says:

    @RACook – It does make you wonder about the integrity and competence of his ‘friends and contempories’ in the field, but ‘science’ and ‘psyence’ are such vast fields that I understand that it’s now possible for someone like him to Voyage for nearly forever and not hear or see another for eons. Remember too that the Hansens are soft, lack luster, cold bodies that come and are sold by the megaton, the hard, bright, hot stuff is measured in Newtons and Einsteins.

Comments are closed.