The GISS Temperature Record Divergence Problem
Guest post by Tilo Reber
“Areas covered occasionally by sea ice are masked using a time-independent mask.”
So if there is sea ice coverage for any part of the year, GISS will not use SST values to cover those cells for the entire year. Those cells must be covered by extrapolations from land for that year. This means that when the area is cover with ice or with water or with part ice part water, it will have it’s anomaly extrapolated from land, regardless. HadCRUT, on the other hand, does not extrapolate their coverage. But they will use SST values for a cell when SST values are available for part of the year. If the area is covered with ice for the entire year, HadCRUT will not assign it a value. Therefore we get polar areas that are covered by extrapolation by GISS and not covered at all by HadCRUT.
When we look at the HadSST2 record, we see that the cool cells that show up above Svalbard in 2005 are consistent with the numbers in that record. And these then go into creating the sea surface portion of the HadCRUT3 temperature record. So, obviously, how cells are filled with data can have a profound effect on the anomaly value that those cells have. This leads one to wonder if extrapolations at the pole are legitimate. I decided to look at some of the northern Russian stations, at the GISS site, that show up as being so hot in the 2005 version of the GISS chart when compared with the 1998 version of the chart. I found that those big changes are in fact represented in the individual records – especially for the coastal stations. Here are three of them.
1998 Annual Mean – -3.39
2005 Annual Mean – 0.60 1998 – 2005 delta 3.99 C
1998 Annual Mean – -14.99
2005 Annual Mean – -10.79 1998 – 2005 delta 4.2 C
Gmo Im.E.K F: 77.7 N, 104.3 E.
1998 Annual Mean – -15.96
2005 Annual Mean – -12.67 1998 – 2005 delta 3.29 C
For comparison, let’s look across the Arctic ocean and see what was happening in Canada and Alaska at the same time.
Eureka, N.W.T.: 80.0 N, 85.9 W.
1998 Annual Mean – -17.38
2005 Annual Mean – -17.34 1998 – 2005 delta 0.04 C
1998 Annual Mean – -8.80
2005 Annual Mean – -10.44 1998 – 2005 delta -1.64 C
So it seems that the North American side of the Arctic changed little, or even got cooler between 98 and 05, the Russian side warmed considerably. Why is that? I think that this ice cover map gives us the answer. As is immediately apparent, the coastal ice cleared out far earlier in 2005 in northern Russia than it did in 1998. This is even though the rest of the globe was slightly warmer in 1998 than in 2005. When dealing with coastal stations, removing the ice and exposing the water is like taking the hatch off a heating source for the coastal thermometers. For stations that are in areas where the temperature is well below zero, exposing the immediate area of that thermometer to a surface that is above zero, changes everything. Looking at Ostrov Vize, we see that it is a small island, and therefore even more subject to changes in coastal sea ice. And when we compare 1998 months on this island with 2005 months we can see that there are differences in some of the monthly means that are larger than 10C. Even a partial ice cover as opposed to a complete ice cover will supply the stations with more heat.
So I think that we can safely say that the huge change in the anomalies of Russian coastal stations is mostly due to coastal sea ice changes. In fact, if we look at stations further inland in Russia, the coastal effect begins to decline. With this in mind, we need to ask if the GISS extrapolations of land based stations, particularly coastal stations, to the poles is appropriate.
The answer would seem to be that it is not, and the Svalbard case makes this perfectly clear. There we had a case where the SST anomaly was actually cool, and yet the land based extrapolation actually turned those sea based cells more than 3C hotter. Reaching across the Arctic Ocean with temperatures that are the result of a coastal sea ice effect cannot give valid answers for what the temperature anomalies away from those coastal stations should be. In fact, taking the variation that is represented by those coastal stations and extrapolating into the interior of Russia is also not appropriate, because the interior areas did not undergo the magnitude of temperature change of the coastal stations.
Looking at the SST temperature anomalies that NOAA uses for 1998 and 2005 it again looks like nothing exceptional was happening in the Arctic (Note, the chart will not retain the months that I selected; so use your own sample months and they will plot). It seems, from this analysis, that GISS polar extrapolations and interpolations are likely to simulate large variations away from the Arctic coasts that are really only present as changes at the Arctic coasts. And the GISS divergence from HadCRUT, as well as from UAH and RSS are likely to be errors instead of enhancements.
By the way, Friday’s Fresh Aire on National Public Radio here in the US featured a guy talking about a book he’s publishing on the Carbon Credit Market. What i’d interested to see is how a ‘Railway Engineer’ ends up at the top of the IPCC. Who decides who is to be the top guy? do they take applications for the job? Did Puschari stand to profit by ‘insider trading’ of construction contracts to Indian
contractors? Who are these people that are supposed to be verifying Carbon Emissions? How much is the carbon market worth today? How big is the Carbon Trading lobby in DC? Who is lobbying in DC? Dadyady yada. When can we expect the debate on climate being driven by AGW Vs. Sun/Solar particle to be settled?
barry:
“I averaged the monthly UAH anomalies for the North Pole for 1998 and 2005. The area is roughly what HadCRUT leave out. The results are:”
barry, you are showing about .7C difference. Now go back up to the article and look at the difference for the stations that I list. All the way from Svalbard to the far northeastern coast of Russia the anomalies for coastal stations were on the order of 2.5 to 3 C. A few, like Ostrov Vize, had more than 4C. And given the far nothern position of stations like Ostrov Vize they are going to play the largest role in extrapolating to the pole.
Warmer UAH and warmer GISS are not the same thing. You have to look at the magnitude.
You can try another experiment if you like. This is very crude, but it will help to make the point. Look at the top row of cells in the 2005 HadCrut and GISS charts. HadCrut has about 30% coverage in that top row. Now take the average value for the cell color for each cell, add them all together, and divide by the number of cells available for that top row of the chart. Both GISS and HadCrut will show warming. But I think that you will find that the warming anomaly that GISS shows is nearly twice as large as that of HadCrut. The reason for the difference is that the HadCrut cells are mostly SST cells, while the same cells in GISS have been extrapolated from shore stations.
Oh please. Here we go again with your troll word play. He makes a personal opinion. By your logic should I then snip any personal opinions you don’t like? Shall I snip yours?
Everyone is entitled to their own opinions, but no one is entitled to their own facts. I did not ask for snipping of content, merely correcting factual errors. This is not word play. Whatever the original poster’s intent, it is presented as fact, and the fact is what he wrote is wrong. When someone presents something as fact and it is wrong, I reserve the right to call out the error. The fact that you did not point the error out is what troubles me, because it suggests (opinion here) that you do not care about correcting obvious factual errors. We already have way too much public blurring of the line between fact and opinion (just turn on TV news for daily evidence). My opinion is that we will all be better off if we try to keep opinion out of discussions of science and deal with the facts.
He is right though, Hansen IS an activist.
A fact which I do not challenge, so no need for me to address.
Quick correction. Where I say “anomalies for coastal stations” I mean to say “differences between 1998 and 2005 for coastal stations”.
What all this tells me is that (as I believe one of the Pielkes once stated) there is no such thing as an ‘average global temperature’, or, at the least, there is not now any valid method to measure/determine one. This leaves us with ocean heat content as the only metric for determining whether the earth is getting warmer, colder or staying constant. Perhaps our research dollars would be better spent enhancing the ARGO buoy system (and possibly the satellite remote sensing instrumentation) and leave the historical temperature data and associated ‘weather stations’ for what they were originally intended (i.e., determining regional weather patterns).
Right is might!
Janurary 28th, 2010. Hundreds of cold related deaths and snow drifts all the way down to Instanbul… Ships stuck in ice in the Baltic…. the Danube freezing over… Temps 14deg colder in Berlin than average. …(Wood TV Grand Rapids)
http://blogs.woodtv.com/2010/01/28/bitter-cold-in-europe/comment-page-1/
I think the divergence problem is about to disappear.
Weather stations are profoundly affected by local weather-related conditions (which are many). Good thing too. They are designed to measure exactly that: The local temperature response to weather-related conditions. If you want to measure warming due to CO2 and you believe that warming is buried in weather temperatures, you will not be able to extract it from ground sensors currently in use. They were NEVER meant to measure the subtleties of climate change!
So the next question, if Hansen truly wants to measure climate change as a result of CO2, what kind of sensor would he want to use and where would he put them? What part of our Earth would show warming (or proxies of warming) as a result of increasing CO2, given that this gas is a VERY tiny fraction of our atmosphere’s other greenhouse gasses?
I have some ideas but what do you all think? Remember, CO2 is a globby substance that gets caught in the jet streams.
The UAH doesn’t do these extrapolations but uses measurement values.
This doesn’t make sense. The satellites only cover to 82N and 82S. Of course they extrapolate from ‘measurements’. That is what GISS do, although the data and techniques are different.
This leads them to different conclusions. Where’s the problem with that? We could now argue whose conclusions are better. I tend to go with the UAH, others might be more inclined to listen to Hansen.
Well, the obvious question is – why? I asked if there has been as much scrutiny on the satellite records as surface, particularly UAH as it is favoured here for some reason. Having searched around the site, it would appear that there are no vigorous appeals for processing algorithms, methodologies – all the things that people demand of GISS, NCDC and HadCRU to replicate and ‘audit’ their products. And there is at least one excellent reason for questioning UAH: of the three surface and two major satellite records, UAH is the outlier. It should be first in line for scrutiny. Is there a reason for ‘tending to go with UAH’ without the rigorous questioning of that source?
Thank you for the links to explanations on the UAH (and RSS) measuring systems. I note no one demanded the algorithms and processing methodologies (although someone did ask how often algorithms are used).
Tilo,
Now go back up to the article and look at the difference for the stations that I list.
I find it curious that you would extrapolate from a few examples – crudely, the same method applied by GISS to discern polar temps. I see that there is a difference in coastal between GISS and HadCRU, but to know for sure if GISS have overestimated, you’d need to crunch all the numbers. As A Watts says, it’s not enough to show pictures and graphs of a few questionable data points. The work needs to be done before conclusions can be drawn. There is already way too much innuendo based on scant info, which, as you must see, is accepted as gospel by many of the readers hereas it is on warmist sites, rather than challenged as vigorously as information form the ‘other side’. The proper antidote to propaganda is hard facts and maths, not more speculation, or we’re simply contributing to the politics.
Tilo, I do see that HadCRU shows cooler anomalies for SS areas at the NP. However, there are more degrees of measurement error going on here. You wrote:
As we know, ice is blown around by the wind and can open up along the edge of the Arctic (and the interior) at any time. We can see from the anomaly maps that HadCRUT has measured different areas at high latitudes between 1998 and 2005. It could be that for the periods when ice covered areas that were measured at one point in the year, it was warmer. That might sound counterintuitive, but we know it can be so. This is why a quantitative analysis is necessary to say anything approaching conclusive.
Also, we’re dealing with blocks of colour here. There may well be bleed simply because of the gridding, and that bleed may average out faithfully for all we know. However, it is clear that there is a significant area of warming that HadCRUT do not include in their global temperature product, and even if GISS extrapolate overly from land for some coastal areas, the corroborating satellite record on the warming leaves the question hanging. The only thing left to do is crunch thje numbers.
Tilo Reber (09:50:43) :
barry:
“I averaged the monthly UAH anomalies for the North Pole for 1998 and 2005. The area is roughly what HadCRUT leave out. The results are:”
barry, you are showing about .7C difference. Now go back up to the article and look at the difference for the stations that I list. All the way from Svalbard to the far northeastern coast of Russia the anomalies for coastal stations were on the order of 2.5 to 3 C. A few, like Ostrov Vize, had more than 4C. And given the far nothern position of stations like Ostrov Vize they are going to play the largest role in extrapolating to the pole.
Warmer UAH and warmer GISS are not the same thing. You have to look at the magnitude.
Tilo
Compare 2005 UAH & GISS ‘arctic’ anomalies using a standard base period to determine the true “magnitude”.
UAH use the 1979-1998 period which, as was shown above, gives an anomaly of ~1.3 deg.
Using the same base period for GISS gives a GISS anomaly of ~1.9 deg.
So – yes – GISS is relatively warmer but its not ‘way out’ and there may be good reason for the discrepancy. GISS and UAH are not measuring the same things. But let’s say Hadley used the UAH figure as their estimate for the arctic. How would the 2005 anomalies look then?
Let’s remind ourselves of the respective 2005 anomalies:
GISS +0.63
Hadley +0.48
But these are relative to different base periods (GISS 1951-80; Hadlet 1961-90) so we need to find the GISS anomaly relative to Hadley (1961-90) base period. This show it here:
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2009&month_last=12&sat=4&sst=1&type=anoms&mean_gen=0112&year1=2005&year2=2005&base1=1961&base2=1990&radius=1200&pol=reg
GISS 2005 Anomaly relative to 1961-90 is +0.56
Now, instead of Hadley ignoring the arctic, let’s use the UAH No Pol anomaly to estimate the arctic. I’m going to assume the arctic is ~7% of the earth’s surface so the calcuation becomes
0.48 x 0.93 + 0.07 x 1.3 = ~0.54
Hadley 2005 anomaly with UAH arctic adjustment is +0.54
The GISS anomaly is 0.02 deg higher – some fiddle!
“The results may be come out of the computer with 2 decimals, but it’s garbage, nevertheless. Expensive garbage. ”
That’s my main beef here. The margins of error for measurement are LARGE compared to the described temperature increase (.1 C?) being sited, and the margins of error for interpolated “guesstimates” and proxy data are HUGE. Astronomical in fact.
When variance falls within the margin of error you can’t say anything at all with confidence. You’ll notice that no one seems to perform a signifigant figures analysis. It would be laughed out of real science.
http://www.chem.tamu.edu/class/fyp/mathrev/mr-sigfg.html
John Finn:
Unfortunately, John, what you are trying to compute is not what I was addressing. You are saying, “in an absolute sense, how close are Hadley and GISS at one point. Namely 2005. I’m talking about a divergence of Hadley and GISS which is clearly shown here.
http://4.bp.blogspot.com/__VkzVMn3cHA/SQkAxK2k6CI/AAAAAAAAADs/F4NlhqTzFgM/s1600-h/U+11+Year+Temp+Data.bmp
In order to show why that divergence is happening I show the difference between 1998 and 2005 as an example. Basically using Hansen’s own example. When comparing those differences, the base period doesn’t matter. I can compare 1998 GISS against 2005 GISS to get one difference and then I can use 1998 HadCrut to 2005 HadCrut to get the other differences. Each of them can be compared to their own baseline. It’s the change that counts.
Hansen already shows that you can get GISS to look like HadCrut by masking GISS cells off that are not represented in HadCrut. You simply go the other way. But that is not what my article is about. It’s about the GISS extrapolations being correct. If you want to use UAH as a point of comparison, fine. But you need to do this. Compare the difference at the N. Pole between UAH 1998 and UAH 2005. Then compare the difference of GISS 1998 to 2005 at the N pol. Again, when you are looking for the amount of change in the data within a dataset, the baseline doesn’t matter. Also, comparing against a base period doesn’t give you the complete picture when you are looking for a divergence, because both 98 and 05 contribute to the divergence, not just 05 against the base period.
So, try 98 UAH against 05 UAH and 98 GISS against 05 GISS, then compare. At the N. Pole of course.
barry:
Also, we’re dealing with blocks of colour here. There may well be bleed simply because of the gridding, and that bleed may average out faithfully for all we know.
We are only using blocks of color to get the initial impression. The differences that I’m talking about between 98 and 05 come from coastal stations for GISS. The Arctic SST anomalies that you can see in the link to the chart that I gave you are colors, but the difference between 98 HadCrut Arctic colors and 05 HadCrut Arctic colors is small enough that you can tell that the HadCrut changes are much lower than the GISS changes, even if you allow for bleed.
barry:
“It could be that for the periods when ice covered areas that were measured at one point in the year, it was warmer. ”
I agree that the SST of ice covered water may be warmer than the SST of open water. But that is not particularly relevant. You would have to make the case that the SST of ice covered water is warmer than the SST of open water in general, and by quite a significant amount. I would probably disagree with that. In any case the significance of the open water is in it’s ability to warm coastal stations where surface temperatures are much lower than the water temperatures. This is an effect that you cannot get with ice covered water.
Thanks for posting this excellent piece of work, Tilo. Looks like Hansen has been getting up to his old tricks again with the GISS data. Wouldn’t trust the guy as far as I could throw him.
Please keep up the good work debunking the NASA crew – it’s good to see a real scientist doing his job, rather than someone who wants the data to show what he believes.
Tilo Reber (11:21:38) :
John Finn:
Unfortunately, John, what you are trying to compute is not what I was addressing. You are saying, “in an absolute sense, how close are Hadley and GISS at one point. Namely 2005. I’m talking about a divergence of Hadley and GISS which is clearly shown here.
Tilo
Are you saying that if we made a similar UAH adjustment to the Hadley record for every year between, say, 2002 and 2009 the divergence would remain. The arctic has been warm in recent years. The UAH record confirms this.
Tilo Reber (11:21:38) :
John Finn:
…
So, try 98 UAH against 05 UAH and 98 GISS against 05 GISS, then compare. At the N. Pole of course.
This does just compare 2 years and as GISS 98 is relatively low this seems a bit like cherry picking (well – a lot actually). But if you insist; GISS is ~1.4 deg higher than GISS 98 while UAH is ~0.7 deg higher than UAH 98. This tells us very little, though.
The difference in recent trends between GISS and Hadley can be explained by the GISS inclusion of the arctic.
Well, we have a lot of maybes here, Tilo.
I was going to try and rebaseline UAH NPo to fit with GISS NPo for 1998 and 2005 – but John Finn’s analysis goes a step further.
I think you’re a bit stuck on a handful of stations. Until you crunch all the data, this is speculation. John Finn’s analysis is likewise provisional, and shows a different conclusion to yours. This is another analysis that should go in the ‘not sure’ basket at best.
John Finn
“The difference in recent trends between GISS and Hadley can be explained by the GISS inclusion of the arctic.”
Yes, we know that. In fact we start with that. The question to be answered is if the GISS inclusion of the Arctic was done correctly. I contend that it was not. The SSTs that one would normally use over water were not used by GISS. Also, the variation caused by coastal sea ice coverage on the stations that are used for the extrapolation causes a local wide variation to be spread to other areas who’s variation is not that wide.
“GISS is ~1.4 deg higher than GISS 98 while UAH is ~0.7 deg higher than UAH 98. This tells us very little, though. ”
It tells me quite a bit. The divergence of the GISS and UAH trend lines since 98 is about .1C per decade. So 7 years worth (98 to 05) would be about .07C. The coastal extrapolations to the Arctic, and the extrapolation of those same stations further inland would account for most of that divergence.
barry:
“I think you’re a bit stuck on a handful of stations. Until you crunch all the data, this is speculation. ”
While an exact quantification would be better, the only way that I could make it exact is by running the GISS code. But there are not that many land stations on the polar rim. And I looked at about 80% of the ones that covered both periods. The 5 that I gave in the post are examples only. The stations from Svalbard all the way across norther Norway to the far Northeastern Arctic coast display the same 2.5 to 3 C variation between 98 and 05. That covers about 3/5th of the Arctic rim. On the North American side 98 is very slightly warmer or about the same as 2005. So the net effect all the way around the polar edge will still give a huge positive variation between 98 and 05.
John Finn:
“Are you saying that if we made a similar UAH adjustment to the Hadley record for every year between, say, 2002 and 2009 the divergence would remain.”
Yes, I’m saying that the divergence will still remain. It will be a little smaller, but it will remain. A simple look at the covered polar Hadley cells and the same GISS cells should tell you that.
“This does just compare 2 years and as GISS 98 is relatively low this seems a bit like cherry picking (well – a lot actually). ”
You’ll have to talk to James Hansen about that. He selected those two years to compare as examples for his post at RealClimate.
GISS data presented together with Hadley or any other data is meaningless by itself.
GISS can always claim that “the other guy got it wrong,” and it becomes nothing more than an academic discussion of “who’s right” and whoever has the more true believers “wins.”
To gain an accurate picture of how GISS data compare with others, the ERRORS on GISS data should be derived from independent sources. Then if there is a systematic bias on the part of GISS, the data will be consistently higher than the means of the independently derived errors. Right now GISS errors ore only estimates derived by themselves to make their data track the means of the errors.