What Trend?

Sometimes, you go through your email and find things you’d overlooked in haste. Sometimes things with “punch”. One such was a letter pointing at a posting at Diggingintheclay on ‘trends’.


The article has a lot of Google Earth KLM files in it that let you look at trends in different parts of the world, but these two images of North America kind of tell the story nicely.

First up is how we usually see the temperature trend data displayed (though with the USA jiggered to be warmer than here, which shows it cooling in the ‘raw’ data) and with a Hot Hot Hot Canada.

North America Temperature Stations All Ages - Trend color coded

North America Temperature Stations All Ages - Trend color coded

These are color coded for stations with at least 10 years of data, so even a very new recent station can show up. Less than 10 years data gets a white dot.

But what happens if you take stations with histories shorter than a PDO cycle, those without enough data to really make a valid trend, and simply color them white? A very different story:

North America Temperature Stations Trend - 50 years of data

North America Temperature Stations Trend - 50 years of data

Stations with 50 years or more of data have a color, those with less than 50 years of data are white.

So now we see that all those “hot hot hot” stations in Canada are relative youngsters. Like making a trend line through one data point. Stations where we have no real history of climate from them, just a few years of PDO weather cycle or of El Nino years…

This emphasizes something I’ve said before. The making of “Grid / Box” anomalies to make a ‘trend’ by comparing one set of stations in one period of time to a different set of stations in another period of time is just a bad way to put lipstick over a splice.

The only anomaly that makes sense for thermometers is an anomaly made by comparing one thermometer to itself, and not to some other thermometer in some other place or with some other processing history of the data. You simply can not compare apples today with oranges in the past and say anything about the trend of fruit…

Look at the American Midwest to East Coast. Blue. Cooling. The individual thermometers are cooling, yet GIStemp finds a warming “trend” by jiggering the “grid box anomaly” and comparing different baskets of thermometers to each other. Just wrong.

So take another look at those pictures. The Canada “trend” is all from stations that may have no history during the prior 1930-40 warm phase of the PDO, nor during the cold phase. Just a nice little ‘rise’ out of a known cyclical pattern with the offsetting cold 1/2 cycle neatly ‘pruned’ by the “grid / box” spice they call an anomaly (that isn’t…)

Can you really look at that second map of long lived stations and say the trend is warming?

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 Science and Background and tagged , . Bookmark the permalink.

11 Responses to What Trend?

  1. DirkH says:

    No comments by now? So i be the first. I’m a meteorological know-nothing, but this is a very meaningful and impressive visualization. One would expect such things to be found out by highly paid climatologists if one would still live under the impression that that science is not corrupt in its entirety.

  2. E.M.Smith says:

    Yup, you get the #1 slot.

    I’ve never figure out why one posting gets comments and another doesn’t. Sometimes it’s because the posting says most of what needs saying so folks feel nothing to add; others it’s because they just didn’t care. Sometimes there is just some other topic caught folks fancy.

    In this case, as it is mostly a referral to ‘go look there’ it can easily be that they took their comments there with them (and that would be fine too…)

    I just like the very clear cut way it highlights that there is no warming trend at those stations with records long enough to have a decent part of a long term WEATHER cycle under their belt.

    Makes it pretty darned clear that there is a major problem with the data. That splicing stations together with “Grid / Box Anomalies” creates a bogus trend (as all splices can do) rather than providing a valid long record. And that playing with the thermometers is a Bad Idea.

    Oh, and that cooling chunk of the USA pretty much says there isn’t much “global” about “global warming”… It’s an artificial trend made by splicing together what ought not to be spliced, then averaging things all over the place against each other (“homogenization”) to hide the problems in the splice.

    Just like where dT/dt found that some months have cooling trends and some months have warming trends in the same place, this shows that different places can have warming or cooling trends. Basically, the whole system has a load of parts moving in different directions and there Is NO Trend overall. Just a lot of fractal bits all going their own way… on top of some major long term cycles.

  3. Verity Jones says:

    Thanks for featuring this – I haven’t made as much of the tool as I would like – yet. And note this is GHCNv3 Beta.

    It is really quite shocking to look at some of the places – e.g. Africa.

  4. BlueIce2HotSea says:

    E.M, about two years ago at skeptic blog, I tried to make the same point as in this (your) thread. My entire post was deleted with the comment that I should avoid (crackpot) theorizing.

    In one sentence, my claim was this: There is a potential for mistakes/mischief when the number and geographical distribution of temperature stations are dramatically changing and yet there is no overt consideration given as to how the temperature trend might be affected by the TIMING of these changes w.r.t. to PDO phase.

    For those who are wondering HOW it might be an issue consider that ratio of el Ninos to la Ninas change from 1/2 PDO cycle to the other. For example, it could be 2:1, then 1:2. This is a big deal depending on geographical region and season. For example, la Ninas which coincide with winter will be much more likely over the next 30 years than was in the past 30. It is therefore more likely in coming winters over the next 30 years, the US will see cooler and wetter for west coast, colder and more snow for the north and warmer and drier for the central and south.

    The when and where of temperature stations seems to matter. And the methods for in-filling missing data and the choice of grid-cell size may not necessarily negate the potential problems, especially for countries with sparse data – natural or contrived.

    I wish an expert would help clear this up.

  5. E.M.Smith says:


    You made the mistake of asking the Bishop why a vow of chastity and poverty involved so much gluttony and why the nun was pregnant…

    IMHO, you are exactly right. There are some added bits. Changes in the volatility of stations chosen (and over time with things like PDO state). Changes in the type of instrumentation and increasing concentrations over tarmac (ASOS at airports). And so on.

    But the basic problem is, IMHO, just as you stated. Constantly changing the thermometer locations introduces a boatload of “Splice Artifacts” that are not removed via the techniques of homogenizing and “grid / box anomaly”, just hidden.

    The unanswerable bit is motive. Is this an incredibly subtle deception or just a royal cock-up because folks actually believed that The Reference Station Method was infallible at fabricating missing “data” and that a “grid / box anomaly” comparing two different thermometers was able to fix the splice and was a suitable substitute for a real ‘self to self’ anomaly made on a single device in a single place.

    At this point, the ‘experts’ have their careers and reputations so wedded to the notion they’ve got it right that there is no way they can admit to having cocked it up, nor can they admit to deception. The only path is ‘tough it out’ and hope to retire or die before it falls apart.

    So they can’t “clear it up”.

    It will take a new batch of experts to clear it up. (Much like it does in most fields. Progress comes when the entrenched retires or dies… A new ‘disruptive’ technology comes along from a new direction and only then do things change.)

    My biggest concern, frankly, is not that the “warmers” will stick to their position, nor even that the government will get on board with it and screw up the economy. All those things would be fixable in time. It’s the screwing around with the actual data record that’s the biggest risk. We have a few hundred years of carefully collected data; and a bunch of clowns who are not preserving it. Too busy “improving it” with methods that are at best questionable. If the original data record is lost, then it is simply not recoverable.

  6. BlueIce2HotSea says:

    This thread reminds me of your very nice Smith’s Volatility Surmise which also mentioned PDO.

    Please tell me if I have this right. Now that California has entered a generally cooler phase, is it reasonable to say that the movement of thermometers to the more temperate coast will blunt the lower lows that would otherwise be expected?

    And was there any movement in thermometers that the emphasized higher highs beginning around thirty years ago when more el Ninos = more likely hotter summers?

    And how about the thirty years prior?

    Please correct my knowledge/understanding gaps. Thanks.

  7. Jason Calley says:

    E.M., you say: “The unanswerable bit is motive.”

    Yes, for you and me, the question of whether “climate scientists” are self-deluded or running a scam is unanswerable. We can guess, but we cannot know. For the True Believers of CAGW it is not just unanswerable, but rather unthinkable. How many times have you discussed this subject with some deeply convinced warmists, and had them respond “You CAN’T think that all those SCIENTISTS are wrong, or even worse, are lying! You CAN’T believe that!” For most supporters of the AGW theory, they literally cannot separate their faith in the character of scientists from the abstract analysis of whether AGW is, in fact, supportable.

    I have a friend — a VERY bright friend — who is, nonetheless, a true believer in AGW. He asked me point blank why I refused to believe in AGW and I explained — at some length, and he was kind enough to let me go on — my doubts about the quality of, the adjustments to, and the analysis of, global temperature data. I said, “If I had to just place a bet, I would guess that the global temperature has risen some in the last few decades, but based on how I see the data handled, I do not think anyone knows accurately how much of a rise we have had or what the cause is.”

    His response: “Why do you think they would mishandle the data that way?”

    Here is where the conversation got weird. I was extraodinarily clear: “OK, good question, but now we are not talking global climate, we are talking human psychology and my opinion of people’s motives. You understand, we are no longer discussing data, or climate, we are changing the subject now and discussing human behavior and opinions, right? Change of subjects, correct?”

    Friend: “Sure, I understand.”

    Me: “In my opinon, most of the climate science folks are just following the band wagon, but because of the recurring patterns I see in the manipulation of data, I think that the people at the highest levels, CRU, GISS, are commiting fraud.”

    Friend: “Ha! So you don’t believe in Climate Change because you think there is a conspiracy of scientists!”

    Me: “NO! The DATA drives what I think about AGW! The motives of the people involved is a whole different subject. Two subjects. My disbelief in AGW is NOT because I distrust the scientists! It’s the data!”

    Friend: “So you don’t believe in Climate Change because you think there is a conspiracy of scientists!”

    Me: “No!”

    Not a whole lot of discussion after that…and remember, this is a BRIGHT guy. Still, even after going above and beyond to talk about data with him, after being painfully, embarassingly clear to differentiate analysis of data from opinion of motive, the fact that I might doubt the honesty of the scientists involved was enough to not only short circuit his rational thought, but for him to throw me (metaphorically) into the conspiracy nut crowd. And this from a man who believes that anti-AGW bloggers are part of a big Oil propaganda conspiracy scheme.

    I can — and DO! — believe in the possibility of conspiracies, both by Big Oil and by Climate Scientists. The world is a big place and powerful people, whether businessmen or scholars, do indeed, “conspire” for their own good. I try to look at the data first, however, before deciding what conspiracies I believe in. Personally, I think that the facts certainly LOOK like climate scientists have knowingly cooked the books.

    I suspect that to my friend, his belief in the incorruptability of scientists prevents him from an honest examination of the data.

  8. E.M.Smith says:


    While not definitive, my impression from looking at the station changes is one of directed and selective ‘locking in’ of a status. So stations that were ‘cold prone’ were “in” in 1950-1980 or so, but then replaced with stations that were much less volatile (and often warmer) during the hot phase. With some tendency to have hot volatile stations in during the 1970 – 1990 period; and now some ‘low volatile’ replacements as we turn cold.

    Use of a ‘grid / box anomaly’ rather than a ‘station to itself’ anomaly (that I call a ‘self to self’ anomaly) allows this to work.

    The dT/dt graphs tend to highlight this kind of shenanigan as a clear ‘hockey blade splice’, and it’s quite visible in the graphs. “Homogenizing” and “The Reference Station” data fabrication are used to smooth out and hide those effects, IMHO.

    I probably ought to go back to a couple of interesting cases I saw in the data (where, for example, one island gets a ‘dip’; in the ‘baseline’ and thus shows a ‘warming trend’ while nearby islands show basic flat temperatures… yet the ‘warming’ islands then gets used to ‘fill in’ the ‘missing’ data from places that are not warming…. These are actually pretty quick to spot as most Island Groups get 2 or at most 3 thermometers, so the one that spikes up to 5 or 10 thermometers is usually the one being buggered. (Again, IMHO).

    Once you catch on to the idea of using differential volatility and a hidden splice via homogenizing, Reference Station Method, and hiding it via ‘grid / box anomalies’ it all makes a lot of sense. Give me choice as to what thermometers to have in the record and when, with those tools, and I can make just about any ‘trend’ you want.

    Want cooling? Put high volatility thermometers in from 1998 to about 2014. Want to hold onto it? Swap to non-volatile as the warming starts (in about 2030?). Smooth the splice with homogenizing, hide the data dropouts with RSM, hide the whole process with a ‘grid / box anomaly’ and ratify the splice with it. Easy as can be…

    Yeah, it would take the rest of my life to prove those perceptions conclusively. For now, its just what it looks like to me. An opinion. But give me a few years (or a few good folks on a team with some funding) and I’ll either prove it or show where I’ve got issues to work on.

    @Jason Calley:

    Yeah, I’ve got an engineer friend. I’ve been working on him for about 1/2 decade now. He’s FINALLY starting to think I might have something, a little. He’s very bright, but busy, and just wanted to accept authority rather than put in the time to think about it. Very depressing…

    BTW, the police bust one heck of a lot of folks for “Conspiracy to do {foo}”… even when foo doesn’t happen. There are a LOT of conspiracies in the world, and that’s why we make lots of them criminal acts. Yet to accuse some folks of conspiring makes one a ‘nut case’… unless, of course, you have a badge…

    All I can say is people are extraordinary when it comes to the ability to not think…

    So I’d not be surprised at all to find Jones and Hansen “conspiring”, but I’d expect it to be called “collaboration” in the press…

  9. BlueIce2HotSea says:

    @ E.M.Smith

    1) The Cook Is. has a land area of just 91 sq. mi. and is spread out over 3/4 million sq. mi. Are a handful of island temperatures really used to represent the SST temperatures BETWEEN islands? Why not use satellites for this?

    2) If you can make just about any trend I want, can you demonstrate a proof of concept on a smaller scale than whole planet? How about using say 10 stations or so to represent Canada? Two for each border and two in the interior? Then drop in and drop out stations creating a varying range of say one to twenty stations over time and esp. PDO phases. Could the avg temp of these stations be manipulated to show overall change in trend slope? Even .05C/decade change would be interesting.

  10. E.M.Smith says:

    In GIStemp you can choose to include SST from a Hadley set, or not. If you don’t include those SST, then land is used to fill out a box, even out to sea. There is a smidgeon of satellite data in the Hadley SST set, so the GISTemp apologists will say that it DOES have satellite data (however vanishingly small…).

    Per “why?”: “Why? Don’t ask why. Down that path lies insanity and ruin. -E.M. Smith”…

    Yes, I think I could so demonstrate. I’m working (too slowly as I need to make money enough to support this hobby and when markets are hard, I have to work more and ‘research’ less) on an example. But one problem is that the missing data in GHCN is, well, missing. So I need an alternative source for the data that isn’t there… That limits me to an example from a small geography, which is then subject to attack as ‘unrepresentative’ or ‘a fluke’.

    FWIW, you can see some of this effect, IMHO, in the Marble Bar posting. In that posting I glue together different stations near Marble Bar a couple at a time and you can see the impact of the ‘splice’ as we end up with a long term rising trend out of several disjoint segments. Yet the record set in Marble Bar has never been exceeded. So it’s “constantly warming” but has never gotten hotter… (It ought to be in the dT/dt category at the right). Oh, here it is:


    But that example will be “poo pooed” as it uses the dT/dt method (that does not seek to hide the splice artifacts) rather than using the GIStemp method. So you can demonstrate the effect more clearly that way, but it’s less authoritatively a direct critique of GIStemp.

    But I do include the GISS maps showing that area as ‘warming’… and in rough agreement with the splice artifact.

    To do the same thing and directly tie it to RSM and Grid / Box anomalies would take a fair amount more work, and time. Yeah, I’d love to get to it; but there is this ‘making a living’ thing that gets in the way of putting full time into playing scientist.

    Oh, and in Turkey the local Mets did their own study and found a cooling trend if you remove the selection bias in GHCN:


    which I think is a pretty good peer reviewed existence proof of the effect…

  11. BlueIce2HotSea says:

    Thanks. I appreciate the astonishing amount of work and unique perspective you are sharing. Hope you clean up in the markets.

Comments are closed.