Email 536 An excerpt of email from David Thompson of Colorado State to Phil Jones:
…As for the dip in 1945. After iterating with John Kennedy, it appears that the dip in 1945 corresponds to a sudden drop in US measurements in Aug 1945 (the US measurements were known
to be biased warm, so the cooling is consistent with the loss of US data). But it is also now clear that the SST is fraught with many instrument changes between the 30s and 1961. So
a conclusion we’ll likely make is that the trend in SSTs between 1900 and the present is reliable, but the behavior of the time series from the 1930s to the 1960s is not.
That the data are so unreliable between the 30s and 60s means we don’t know for sure what happened in terms of global-mean temperatures during that period. In fact, if you blank out the data from the 30s to the 60s, you can actually imagine the globe warming weakly but continuously during that period…
Hence, the only real evidence we have of a midcentury about-turn in global warming comes from the land data.
Full email
date: Tue, 27 Nov 2007 10:48:58 +1100 from: David Thompson <davet@atmos.xxxxx> subject: a quick comment and a quick question to: Phil Jones <p.jones@uea.xxxx> Phil, The comment…. Thanks for the thoughts on the volcano plots. I’ve spent the last few days playing with different analyses, and I think I’m converging on the main points to make in the paper. It’s my impression that almost all aspects of the volcanic signal have been discussed in the literature, except for the longish timescale suggested by the residual data and the detrending. For sure the timescale is sensitive to the detrending, and I’ll be very careful about that in the writing. But I think using the residual data we can get folks chatting about the possibility that volcanoes impact SSTs much longer than the ~2-4 years suggested by the current literature. Anyway…. that’s how I’m leaning on the results. I should have some text ready soon… The question…. As for the dip in 1945. After iterating with John Kennedy, it appears that the dip in 1945 corresponds to a sudden drop in US measurements in Aug 1945 (the US measurements were known to be biased warm, so the cooling is consistent with the loss of US data). But it is also now clear that the SST is fraught with many instrument changes between the 30s and 1961. So a conclusion we’ll likely make is that the trend in SSTs between 1900 and the present is reliable, but the behavior of the time series from the 1930s to the 1960s is not. That the data are so unreliable between the 30s and 60s means we don’t know for sure what happened in terms of global-mean temperatures during that period. In fact, if you blank out the data from the 30s to the 60s, you can actually imagine the globe warming weakly but continuously during that period… Hence, the only real evidence we have of a midcentury about-turn in global warming comes from the land data. Are there any similar data issues in the land data during the period ~1939-1960? Thanks, Dave ——————————————————————– ——————————————————————– David W. J. Thompson www.atmos.colostate.edu/~davet Dept of Atmospheric Science Colorado State University Fort Collins, CO 80523 USA Phone: 970-xxx Fax: 970-xxx
Well, you don’t need to go into people’s emails to find this out. It’s what Thompson and Jones and others said in their paper in Nature, 2008.
If instruments changed between the 1930s and 1960s, then how can they say that the 1900-present trend is reliable either? If instrumentation was different in 1900 than now, how can we directly compare measurements between then and now? It seems that changes in the middle of the time series messes up the entire thing on some level, not just one section in the middle.
Continuing recovery from LIA.
I have a problem with sea surface temperatures anyway. Sea surface temps are often a wind speed proxy. Show me a temperature anomaly and I will generally be able to show you a wind anomaly.
Nick Stokes,
You must be getting little sleep having to defend all these scoundrels around the clock.
@crosspatch
UHI may also be a wind speed proxy. See Hinkel, et al. 2003
It took me a few reads before I realised they ment “Global Mean Temperature” rather than “Greenwich Mean Time”.
Then I was reminded that daily “average” temperatures recorded on land are actually the mid point of the highest and lowest temperatures recorded in a “day”. Rather than the the arithmetic mean of all temperature observations recorded in a “day”. Attempting to derive any sort of “mean” from the former just dosn’t make much sense either mathematically or statistically. You might just as well take the midpoint of the highest and lowest temperatures recorded anywhere on the planet between 1st January 00:00:00 and 31st December 23:59:59 and call it something like the “Annual Global Average”. (For which it probably wouldn’t matter if you had sea surface figures at all.)
With the added complication that it could easily matter if “day” is defined according to geo-political time; ditto but without any DST; using a timezone determined by longitude only; using actual local time or some other method.
Something which gets somewhat more complex with the sea considering the International Date line crosses the Equator three times. In some places appears to follow a line of latitude. As well as having changed over the last century.
I wrote a major article on SST’s, delving into their history.
http://judithcurry.com/2011/06/27/unknown-and-uncertain-sea-surface-temperatures/
Since then I’ve had correspondance with the Met office about them. The Records are mostly complete nonsense before around 1961 and I wouldn’t place too much credence on the notion of a ‘global’ record for much of the period since either.
tonyb
@Nick Stokes
The difference is the paper does not use the word ‘unreliable’- their choice of words in private emails.
The Records are mostly complete nonsense before around 1961 and I wouldn’t place too much credence on the notion of a ‘global’ record for much of the period since either.
I’ll second that.
In addition increasing SSTs contain both a global warming signal and an ocean heat lose signal. the former showing climate warming and the latter climate cooling. Separating the two requires good ocean heat content data, which we have only since 2002. And to my knowledge no one has done this and published.
A global average temperature derived from accurate SSTs (which we don’t have) would be a poor measure of GHG warming at least on a decadal timescale.
Why do they persist with it, you ask?
There isn’t an alternative for the oceans and the many problems and instrument changes are an adjusters delight.
@Mark:
The method is not quite as insane as you make out.
It is true that a ‘proper’ mean involves a continuous integration of temperature with respect to time. However, it is quite difficult to actually do that – you have to use analog computers, with their inherent inaccuracies, or you have to use discrete, sampled data and approximate the integral using some method. The question then becomes, what is your sampling rate? Obviously, in the modern age, we would choose something quite fast compared to the rate at which temperature varies, perhaps on the order of once per minute, to give us a good approximation to the true integral.
But in the age when the only way of taking a measurement was for someone to go and look at a thermometer and write down the number, before anyone was worried about climate change, before anyone knew much about forecasting weather, we are very lucky that a significant number of people thought it was worthwhile to take two measurements per day, and that they are not just the temperature at 6AM and noon, but the actual minimum and maximum.
Given that temperature varies over 24 hours in a roughly sinusoidal shape, taking the mean of the minimum and maximum as an approximation to the true mean is not so bad. And when those happen to be the two data points that we have for a considerable portion of the temperature record, it is absolutely the right thing to do.
The only question I have left is why it is still necessary to explain this to people. Yes, it might be better to have minute-resolution temperature data to construct accurate approximations to real means, but the bone-headed, bleedin’ obvious fact is that we don’t have it.
North Atlantic has probably the best coverage and since 1910 it just goes 30 years up, 30 years down, 30 years up and now reverting down again. Interesting that the periods are pretty much linear with sudden breaks into different slope. Its variability is so pronounced that it overrides not too distinctive SH record and puts its shape into the “global” record.
All these SST plays by the Team were to remove the inconvenient 1945-1975 cooling and decreasing the “1940 blip”.
I agree with Nick Stokes. (How’d you like that, Nick? We agreed on something again.) This is one of the bases of Thompson et al (2008), and is one of the things the Hadley Centre attempted to correct with HADSST3.
The reliability of the data and the certainty of the findings appear to have been grossly overestimated here. By their own lights, these findings should be hung about with so many caveats that the caveats outweigh the report. Instead we get “settled” science.
Something is seriously wrong, and I no longer find it easy to believe that it was innocent self-delusion. Some of these guys at least were deliberately out to cook the books.
Verity Jones says: November 28, 2011 at 12:35 am
“The difference is the paper does not use the word ‘unreliable’- their choice of words in private emails.”
You have to give a bit more backing in a Nature paper. So they say, in their bolded introduction:
“We argue that the abrupt temperature drop of 0.3 °C in 1945 is the apparent result of uncorrected instrumental biases in the sea surface temperature record.”
Sounds unreliable to me.
Strikes me as more a case of trying to hide/underplay the decline in SSTs mid-century. There’s patchy (Patchi?) data elsewhere that warmists are happy to use to further The Cause, such as in the high Arctic, but in this case it’s an opportunity to dump a cooling period that isn’t passed up.
Bob Tisdale says:
November 28, 2011 at 2:12 am
I agree with Nick Stokes. (How’d you like that, Nick? We agreed on something again.) This is one of the bases of Thompson et al (2008), and is one of the things the Hadley Centre attempted to correct with HADSST3.
Hi Bob,
this ongoing post on my blog may interest you. Your input is very welcome.
http://tallbloke.wordpress.com/2011/11/07/station-change-over-time-and-ghcn-v2
The ICOADS vs HADsst3 comparisons further down the thread are relevant I think.
I’ve just plotted up the difference between hadSST3 and the original ICOADS data. It’s hard to imagine this being some careful correction for buckets and other sampling issues.
http://oi44.tinypic.com/1zee6ut.jpg
key features of this “correction”.
huge warming of the whole pre 1950 record. (remove pre CO2 warming)
reduction of 1910-1920 minimum (reduce extent of natural variations)
warming adjustment falls away during 1920-1960 by almost 0.5C (remove inconvenient early warming that was greater than late century trend)
deepen Krakatoa and Mt. Agung volcano impacts (justify increasing volcanic adj means more CO2 warming needed counter balance late 20th volcanoes )
Gentle late 20th boost to help AGW (can’t get away with much here, too many other datasets)
Ironically, the only bit that does not get modified (zero difference) is the war time period that showed the most blatantly obvious anomaly.
The profile of this “correction” is a blueprint of all you’d want to do to the SST data if your were trying to get rid of some inconvenient truths that were messing up your AGW hypothesis.
>>
In fact, if you blank out the data from the 30s to the 60s, you can actually imagine the globe warming weakly but continuously during that period…
Hence, the only real evidence we have of a midcentury about-turn in global warming comes from the land data.
>>
In fact , if you blank out the same period in the land data , we don’t have *any* evidence of a midcentury about-turn in global warming. Sorted.
Next all we have to do is blank out later warming period as well and we can shut down the IPCC, sack Jones and the others and get on with our lives.
It may need some work on the wording but I think Thompson’s approach has a lot merit.
As a former collector of sea-surface temperatures, I am a little bit curious as to how this information is collected now; met. ships are probably now in very short supply, so is this data from satellite observation? If so, has there been adequate correlation with locally-collected data to verify the accuracy of these satellites?
So our Phil would like to revisit volcano adjustments. Probably a good idea, they maybe need reassessing too since they are build on simplistic approximations of how much ash gets ejected high enough into the atmosphere and what affect is has when it gets there. Has this ever been measured rather than modelled?
http://oi44.tinypic.com/16k1iiu.jpg
What can we note?
One thing I have not seen covered in the literature is that the winter following a major eruption seems to be *warmer* than the detrended average. This is logical if compared to night-time cloud. Perhaps more money is needed to study this issue. 😉
In fact despite the magnitude of Mt Agung it seems to have cause a good year of warming before the later cooling. That probably explains why this period got some extra special care in the hadSST3 corrections. The temperature record must be wrong, it does fit our model.
El Chichon is a bit tricky as well . It also seems to have caused a fair amount of warming, apart from the initial glitch that was no bigger than the short term noise in the period preceding the eruption.
Mt Pinatubo caused two warm winters that were about same as average before and after. When I say “winter” it looks like more than 6 months of each of those years was not cooled by the eruption at all. Is that part of the model.
Yes , I think this area probably needs some further adjustment. But when you look closely it may not be in the direction you were hoping for Phil. But then you’re an honest scientist aren’t you? You are surely going to publish whatever you find. This could be a significant contribution to climate science.
Land surface vs Sea Surface
Even with proper instrumentation and measurement ->
Land temperature and Sea temperature data are different animals.
The Sea measurements are dipped from the liquid, hence they represent the true ‘surface’ – not the air above the surface.
The Land measurements are made in AIR at one to two meters above the actual surface.
Why would that matter? For an energy balance or heat transfer calculation, the temperature of the actual surface (liquid or solid) is the start of the energy path out to space. The air above the surface could be either warmed or cooled by the surface below, but on a sunny day, the surface is likely warmer than the air and hence heating the air above. The air temp is not the same as the land surface temp.
At the point Land temperature readings are taken, the energy is already on the path out toward space. Air measurements one to two meters above the surface already reflect the combined convection and radiant energy transfer from surface to air. Sea measurements are the ‘surface’ itself (albeit affected by evaporation, circulation, currents etc). One should expect an ‘offset’ between land and sea measurements even with perfect data. I believe there are papers on this (Christy? Spencer?) although I don’t have a reference close to hand as I write this.
cheers, D in De
In the past, before the installation of satellite dependent weather reporting buoys, most open-ocean marine temperature data came from measurements made by ships in transit. It is my understanding that these catch-as-catch-can measurements were disrupted during World War II by changed shipping patterns. Data collected by ships in convoy would not be as extensive as measurements made by the same vessels travelling alone in peacetime. Atlantic traffic was probably diverted to the more northerly routes during the war where shore-based air cover was more likely.
Yes, it might be better to have minute-resolution temperature data to construct accurate approximations to real means, but the bone-headed, bleedin’ obvious fact is that we don’t have it.
We have fixed time temperature measurements from as far back as 1900. We have large numbers of fixed time measurements from the 1960s onwards,many stations with hourly data.
The reason they aren’t used is they show substantially less warming in the average than the min max average. Minute by minute data is a red herring.
Just a quick thought on the subject of volcano after effects. Look again at the periods following major eruptions.
http://oi44.tinypic.com/16k1iiu.jpg
About 5-6 years after each of these events there was a net warming. Could it be that climate, being intrinsically stable actually manages to compensate for the cooling effect (by adjusting cloud cover or whatever mechanism) ?
6y after El Cnichon there was a warming peak, 6y after Mt Pintubo there was the strongest El Nino in recent history (was this extra strength volcanic rebound?) , 6y after Mt Agung there was a warming peak.
Far from positive feedback and tipping points this suggests a substantial negative feedback.
This is just speculation , an hypothesis, more money is needed for further research. 😉
At least there is some empirical evidence that merits further study.
Hansen’s volcanic forcings are generally reckoned to be over stated. But since this is all pure modelling and supposition maybe this effect needs to be investigated. After all “volcanic correction” is a major player in upping the actual scientifically calculated CO2 forcing and a significant part of what underlies the positive climate feedback conjecture.
Anyone curious about the war-time variations should just look at the original ICOADS data. There was a one time increase (from one month to the next, no “trends”) in late 1941 . There was a roughly equal drop after the war. Again a step change from one month to the next.
In order to correct all that you simple need to subtract 0.24C for the duration of that interval , nothing more elaborate than what would be blindingly obvious to a school child.
http://oi43.tinypic.com/htyd61.jpg
changes in shipping caused an offset, so remove it.
If you do that the data seems to flow quite naturally across the whole period. There is NO inexplicable post war drop, just natural variation.
There is no anomaly. It was all a figment or their inept fiddling (well they would not have done that on purpose to give more post war warming would they?) . Inept fiddling that has not been corrected in last quarter century since it was introduced.
I have said it before and I’ll say it again, trying to derive a global mean temperature to look for global climate change using instruments meant to measure local weather is a fool’s errand. It is kind of as stupid as, I don’t know, using tree rings to figure out the past global climate.
The problem with sea surface temperatures, that is temperatures RIGHT AT the surface, is that if you increase wind speeds a little, it reduces the surface temperature by a lot. I would put more stock in temperatures at some depth below the surface. Actually, I would rather see temperatures measured in the abyssal deep. Those would have very little noise and a 0.1C temperature change over a decade would be significant but the logistics of measuring there is a huge problem. Sensors on undersea cables would probably be the best bet.
Tom says:
November 28, 2011 at 1:35 am
… Given that temperature varies over 24 hours in a roughly sinusoidal shape, taking the mean of the minimum and maximum as an approximation to the true mean is not so bad.
How bad is it? Quantify it. Spatially and temporally.
And when those happen to be the two data points that we have for a considerable portion of the temperature record, it is absolutely the right thing to do.
Is it? Or is is the absolute right thing to do: To admit that the data are inadequate for the new purpose – the purpose for which the data were not collected, the purpose that was not even conceived of when the data were collected.
The only question I have left is why it is still necessary to explain this to people. Yes, it might be better to have minute-resolution temperature data to construct accurate approximations to real means, but the bone-headed, bleedin’ obvious fact is that we don’t have it..
That we do not have what we need is not a justifcation for pretending that what we do happen to have is what we need. You are employing very faulty logic to avoid the necessity of having to “explain this to people”.
“That the data are so unreliable between the 30s and 60s means we don’t know for sure what happened in terms of global-mean temperatures during that period.”
In fact, I wouldn’t be surprised if it was actually a bit cooler than they’ve imagined it to be.
So now they’re trying to deny the dustbowl?
>> So now they’re trying to deny the dustbowl?
No, it’s a myth, it was all due to change from old fashioned dust sensors to modern electric ones. 😉
Tom says:
November 28, 2011 at 1:35 am
@Mark:
(((SNIP))))
The only question I have left is why it is still necessary to explain this to people. Yes, it might be better to have minute-resolution temperature data to construct accurate approximations to real means, but the bone-headed, bleedin’ obvious fact is that we don’t have it.
Do you always search for your keys under the lamp-post?
Atmospheric temperature does not equal atmospheric heat content and it is heat content we need to measure. So it might be better to have humidity and air temperature every hour at least and guess what? We actually have those metrics in many cases. At least we can _start_ gathering the data needed even if it was not appreciated in the 17th century that it was needed. Then we can estimate the average atmospheric heat content at each location every hour. It may well be that during the day temperature rises but heat content stays the same if the air becomes less humid. Why use incorrect metrics when you no longer have to?
The more we allow these individuals to adjust the records for “errors” and “step changes”, the more the temperatures over time are going to look like 4.3*ln(CO2current/CO2original).
Just leave everything the way it is.
If it continues warming, then we will know. We will not know anything if we let Mann, Jones, Karl and Kennedy continue adjusting the record. Pretty soon they will be just recording their climate model results instead of actual readings.
Dave’s email states:
This is perhaps the ignorance question of the day – but why/how were US measurement’s “biased warm” during that period, and how is that ‘known?’
Dave went on to say:
Perhaps only my cynicism speaking, but sure sounds to me like they are trying to eliminate the inconvenient mid-century cooling just as they tried to eliminate the MWP & LIA – or at least hoping to be able to come up with a way to do so.
crosspatch says
quote
I have a problem with sea surface temperatures anyway. Sea surface temps are often a wind speed proxy. Show me a temperature anomaly and I will generally be able to show you a wind anomaly.
unquote
Presumably the reasoning is higher windspeed = more evaporation and higher emissivity = lower temperature?
Why the blip? (If you haven’t seen it before, Google that phrase and Tom Wigley.)
JF
(I think i can do that as well….)
Serious discussion of the issue hasn’t even started if Earth Orientation Parameters (EOP) have not been mentioned. Trust obliterated.
“…As for the dip in 1945.”
Must have been that CO2: http://www.biomind.de/realCO2/bilder/CO2-MBL1826-2008-2n-SST-3k.jpg sarc/
There was a three-year strong El Nino from mid-1939 to the end of 1942. This was the longest El Nino on record. The AMO also spiked during the period which it often does in reponse to strong events. By 1946, they both crashed resulting in the rise then dip.
There are only two comparable periods when just as long ENSO events happened and these were both La Ninas. From 1970 to 1976 (the Great Pacific Climate Shift anyone) and the 1906 to 1911 period (1906 to 1911 was the coldest period in the 20th Century).
Just pointing out there are other explanations for large changes in the trends (and these do not have to be instrument errors).
This is a very serious issue, by far the most serious issue I have seen raised in any “climategate” thread. I will have more to say in the months & years ahead…
Tom says: November 28, 2011 at 1:35 am
Yes, it might be better to have minute-resolution temperature data to construct accurate approximations to real means, but the bone-headed, bleedin’ obvious fact is that we don’t have it.
But we do have actual MIN and MAX values to analyse for cooling/warming.
Analysing the mid-point between these two daily outliners is just crazy.
RE: P. Solar : (November 28, 2011 at 7:14 am)
“Anyone curious about the war-time variations should just look at the original ICOADS data. There was a one time increase (from one month to the next, no “trends”) in late 1941 . There was a roughly equal drop after the war. Again a step change from one month to the next.
“In order to correct all that you simple need to subtract 0.24C for the duration of that interval , nothing more elaborate than what would be blindingly obvious to a school child.”
Did they do this for the published HadCRUT and GISS data? As the war (World War II) went through several stages, I would be surprised if a single change would cover the whole period beginning in late 1939.
“…But the days grow short when you reach September.”
September Song — (used in a WW II documentary)
How well do we know the atmospheric pressures – say from the sea surface to stratosphere?
Are there any measurements by satellite?
RE: Agile Aspect: (November 29, 2011 at 3:16 pm)
“How well do we know the atmospheric pressures – say from the sea surface to stratosphere?
“Are there any measurements by satellite?”
These pressures can only be measured from aircraft. Weather balloons are also aircraft. In general, air pressure decreases exponentially with increasing altitude because of decreasing density (thinning out) with increasing altitude. That is also why a given mass of air tends to cool as it rises because it is also expanding in volume due to decreasing external pressure squeezing it together. Density is mass per unit volume.
re post by: Bill Illis says: November 29, 2011 at 4:40 am
Thanks for the explanation Bill.
Spector says:
November 30, 2011 at 1:52 am
“These pressures can only be measured from aircraft. Weather balloons are also aircraft. In general, air pressure decreases exponentially with increasing altitude because of decreasing density (thinning out) with increasing altitude. That is also why a given mass of air tends to cool as it rises because it is also expanding in volume due to decreasing external pressure squeezing it together. Density is mass per unit volume.”
;————————————————————————————–
Okay, thanks.
There was rumor NASA was developing differential radar to measure pressure.
Assuming the only forces acting on a parcel of air are the forces resulting from the gradient of the pressure and the gravitational field, then at mechanical equilibrium, the forces sum to zero and the solution the differential equation results in the pressure decreasing exponential with the height.
If the gravitational field were to vanish, then the pressure would be constant with height.