GISS vs CRU/RSS/UAH

[BethesdaWX]

Its alot more tha ENSO, and GISS ( )No offense....(nice work), but only a fool would use GISS, and then Argue that it is better than UAH without posting any evidence. Instead of using an outlier, how about trying something better like UAH or RSS? The error potential on UAH is +/- 0.05C according to Roy Spencer (who is not a political activist like Hansen). UAH also has Better Resolution, More Coverage, and No extrapolations that will create error.

-You also need to take out the PDO/AMOIOD warming/cooling

-Global SST's in Response to the PDO/AMO/IOD changes

-Now, the mystery of perfect correlation, the Solar Geomagnetic Flux alone...TSI (total slar irradiance) includes energies that have little to no effect on the Global Temps. Declining sunpots for example, do not represent the geomagnetif fluz from the Sun that had the correlation, as seen below.

Geomagnetic flux is really all that matters, and it peaked in the Mid 90's

The trend you will get should look like a mess of ups and downs leading to nothing.

I'm going to Predict your next response and our following conversation.

Skier: GISS has a +/- 0.03C error Bar, UAH has a +/- 0.05C error bar. (no evidence)

Mitch: You have posted no evidence of GISS and +/- 0.03C error. We had a huge debate on this issue, with several people correcting you on the meaning Behind Roy Spencers Argument.

Skier: You want evidence? Fine! *Posts a link from a warmist blog site that incorrectly analyzes UAH, and provides No Explanation behind GISS* Remember, there might be satellite drift/decay that may be affecting anomalies.

Mitch: Roy Spencer explained the issues with that review. We do not know the extent of Satellite drift/decay, if Any, and how this would affect measurements. Knowing UAH has better coverage and reso than GISS, I'd prefer to go with UAH.

Skier: STAR shows twice as much warming as UAH/RSS, and has a Better Method of Analysis (no evidence). Also UAH/RSS have to be experiencing satellite drift/decay.

Mitch: Where is your evidence? That claim is BS, we'd have no way of knowing that anyway. Why are you choosing an Outlier? We also don't know how much, if any, sattelite drift/decay UAH/RSS are experiencing, and how much effect they would have is even a bigger mystery/ STAR would have to be experiencing Drift/Decay as well.

Skier: You are a complete Joke, its obvious. UAH/RSS are not as goods as GISS, STAR, or anything else. End of Story.

Mitch:

[nzucker]

I have been working on some simple statistical models for how different variables have affected climate historically and how they might affect it in the future. So far all I have completed is models for ENSO and ENSO+CO2. The ENSO corrected model of GISS reveals that we have reached a new record high for ENSO-corrected surface air temperature in the last two months. In the future I hope to use these statistical models to show how other variables might have affected climate historically, and how they might affect it in the future.

I have only performed this for GISS so far because I believe it is the most accurate of the temperature indexes.

Once again, you continue to laud GISS's praises without responding to any of my arguments about individual incorrect extrapolations, why they are happening, and if they can affect the overall trend. You've clearly bought into GISS just to bother the other skeptics on here who consider it a biased source, which is certainly a claim with good reasoning considering Hansen's background and several mistakes/manipulations that have occurred with the GISS data. You don't explain why GISS has experienced more deviation from other sources in recent years despite vast improvements in satellite technology, why extrapolations in the arctic are universally warm when the satellites show cold anomalies in some areas, and why the satellites are incorrect considering the LT should have warmed more, not less, than the surface.

GISS has failed to fall significantly in a strong La Niña when all other sources show a big drop. Anomalies in February were basically the same as December despite a huge decline in the other sources as well as the clear presence of abnormally cold airmasses over large landmasses such as Canada. Let's just take a look at RSS/GISS for Feb 2011 and examine the problems:

What happened to the cold anomalies in Northeast Russia that RSS shows? What happened to the cold anomalies near Labrador that RSS shows? I mean it's pretty obvious, Skier, GISS extrapolates warmth anytime a majority of the area around a given point is warmer than average...there's just not enough resolution for the precise analysis the satellites give you. I've shown you countless examples of incorrect extrapolations, and yet you continue to use GISS. FAIL.

[skierinvermont]

Its alot more tha ENSO, and GISS ( )No offense....(nice work), but only a fool would use GISS, and then Argue that it is better than UAH without posting any evidence. Instead of using an outlier, how about trying something better like UAH or RSS? The error potential on UAH is +/- 0.05C according to Roy Spencer (who is not a political activist like Hansen). UAH also has Better Resolution, More Coverage, and No extrapolations that will create error.

I'm going to Predict your next response and our following conversation.

Skier: GISS has a +/- 0.03C error Bar, UAH has a +/- 0.05C error bar. (no evidence)

Mitch: You have posted no evidence of GISS and +/- 0.03C error. We had a huge debate on this issue, with several people correcting you on the meaning Behind Roy Spencers Argument.

http://www.ipcc.ch/p....html#table-3-3

The error estimates for GISS from 1901-2005 are +/-.014C. For 1979-2005 the error bars are +/-.047C. For 1979-2011 it is probably slightly smaller at about +/-.040C/decade. No matter how you slice it, this is less than the +/-.05 or .07C/decade for UAH (depending on source or time period selected).

One can see the statistical error bars for all the different tropospheric and surface sources from 1979-2004 in this chart. One can see that the error bars for UAH are among the largest on the chart. Keep in mind these are only statistical error bars and do not include methodological errors. It is also important to note that NRA, ERA, RATPAC and HADAT tropospheric and/or surface trends have all been replaced with newer versions which show more warming (ERA is a computer model reanalysis and RATPAC and HADAT are radiosonde data).

http://www.ipcc.ch/p...igure-3-18.html

Skier: You want evidence? Fine! *Posts a link from a warmist blog site that incorrectly analyzes UAH, and provides No Explanation behind GISS* Remember, there might be satellite drift/decay that may be affecting anomalies.

Mitch: Roy Spencer explained the issues with that review. We do not know the extent of Satellite drift/decay, if Any, and how this would affect measurements. Knowing UAH has better coverage and reso than GISS, I'd prefer to go with UAH.

Skier: STAR shows twice as much warming as UAH/RSS, and has a Better Method of Analysis (no evidence). Also UAH/RSS have to be experiencing satellite drift/decay.

Mitch: Where is your evidence? That claim is BS, we'd have no way of knowing that anyway. Why are you choosing an Outlier? We also don't know how much, if any, sattelite drift/decay UAH/RSS are experiencing, and how much effect they would have is even a bigger mystery/ STAR would have to be experiencing Drift/Decay as well.

Skier: You are a complete Joke, its obvious. UAH/RSS are not as goods as GISS, STAR, or anything else. End of Story.

Mitch:

You don't appear to understand the differences between UAH and STAR - which I have now explained with citations several times. STAR uses a newer superior method of calibrating for orbital drift called simultaneous nadir overpass (SNO).

http://www.star.nesd...684-02_Zou1.pdf

STAR is not the outlier. RSS, and especially UAH are becoming the outliers. Several newer radiosonde data sources agree more closely with STAR than UAH/RSS. The HadAT and RATPAC Radiosonde data sets which previously agreed with RSS and UAH have been replaced by IUK, RICH, and RAOBCORE, all of which show more like .2C/decade warming (more than UAH's .14C/decade and RSS .16C/decade)

http://onlinelibrary...1002/wcc.80/pdf

There is active debate on tropospheric trends, but RSS and UAH are rapidly becoming outliers with newer analyses including STAR, IUK, RICH, and RAOBCORE all showing substantially more tropospheric warming.

[skierinvermont]

Once again, you continue to laud GISS's praises without responding to any of my arguments about individual incorrect extrapolations, why they are happening, and if they can affect the overall trend. You've clearly bought into GISS just to bother the other skeptics on here who consider it a biased source, which is certainly a claim with good reasoning considering Hansen's background and several mistakes/manipulations that have occurred with the GISS data. You don't explain why GISS has experienced more deviation from other sources in recent years despite vast improvements in satellite technology, why extrapolations in the arctic are universally warm when the satellites show cold anomalies in some areas, and why the satellites are incorrect considering the LT should have warmed more, not less, than the surface.

GISS has failed to fall significantly in a strong La Niña when all other sources show a big drop. Anomalies in February were basically the same as December despite a huge decline in the other sources as well as the clear presence of abnormally cold airmasses over large landmasses such as Canada. Let's just take a look at RSS/GISS for Feb 2011 and examine the problems:

What happened to the cold anomalies in Northeast Russia that RSS shows? What happened to the cold anomalies near Labrador that RSS shows? I mean it's pretty obvious, Skier, GISS extrapolates warmth anytime a majority of the area around a given point is warmer than average...there's just not enough resolution for the precise analysis the satellites give you. I've shown you countless examples of incorrect extrapolations, and yet you continue to use GISS. FAIL.

I have explained that mathematically the process of extrapolation leads to a falsely cold extrapolation just as frequently as a falsely warm extrapolation multiple times, unless there is reason to think that warmer anomalies than the surrounding areas somehow congregate over GISS sensors. You have not responded to this point or provided evidence that warm anomalies are significantly more likely to fall on sensors. I have also provided countless examples of where GISS extrapolates too cold. You also have failed to respond, why if GISS extrapolations are too warm, HadCRUT+UAH in the arctic ~= GISS. And despite your lack of response to these very reasonable and universally accepted points among peer-reviewed articles, you continue to be belligerent and rude on the subject accusing me of personal vendetta just to "bother skeptics" and "FAIL."

P.S. You continue to compare GISS 1951-1980 to RSS 1979-2000 even after I have pointed out that you need to change the GISS period to 1979-2000 to make a fair comparison. Once that has been performed, we notice that GISS is vastly too cold across Mexico, SW U.S., northwest Russia, northern Scandinavia.. etc. etc. etc.

I mean christs! There are -4C anomalies across a wide area of the northwest Russian coast and offshore where RSS has positive anomalies. Clearly there is a conspiracy afoot! GISS is extrapolating brutally cold -4C anomalies where we know it was above average! That is the Law of Large Numbers for you.

[ORH_wxman]

GISS has become an outlier this decade. Their extrapolation method is a joke. There's a reason even Hadley (big AGW proponents) thinks they are too warm. I'll be interested to see what Hadley's stations near the polar boundaries show in comparison to the lesser covered GISS for the month of February when its released.

[tacoman25]

I have explained that mathematically the process of extrapolation leads to a falsely cold extrapolation just as frequently as a falsely warm extrapolation multiple times, unless there is reason to think that warmer anomalies than the surrounding areas somehow congregate over GISS sensors. You have not responded to this point or provided evidence that warm anomalies are significantly more likely to fall on sensors. I have also provided countless examples of where GISS extrapolates too cold. You also have failed to respond, why if GISS extrapolations are too warm, HadCRUT+UAH in the arctic ~= GISS. And despite your lack of response to these very reasonable and universally accepted points among peer-reviewed articles, you continue to be belligerent and rude on the subject accusing me of personal vendetta just to "bother skeptics" and "FAIL."

P.S. You continue to compare GISS 1951-1980 to RSS 1979-2000 even after I have pointed out that you need to change the GISS period to 1979-2000 to make a fair comparison. Once that has been performed, we notice that GISS is vastly too cold across Mexico, SW U.S., northwest Russia, northern Scandinavia.. etc. etc. etc.

I mean christs! There are -4C anomalies across a wide area of the northwest Russian coast and offshore where RSS has positive anomalies. Clearly there is a conspiracy afoot! GISS is extrapolating brutally cold -4C anomalies where we know it was above average! That is the Law of Large Numbers for you.

Except that GISS has consistently run warmer than other sources over most the globe, using the same baseline. Clearly, the warm extrapolations are occurring to a greater degree than the cold extrapolations. It does not balance out evenly, given what other sources show, as you claim it should.

[skierinvermont]

Except that GISS has consistently run warmer than other sources over most the globe, using the same baseline. Clearly, the warm extrapolations are occurring to a greater degree than the cold extrapolations. It does not balance out evenly, given what other sources show, as you claim it should.

As I pointed out, the vast majority of the difference between the two maps he posted was due to the fact that the GISS image used the cold 1951-1980 baseline, while the RSS map used the 1979-2000 baseline. I wish people could get their baselines straight.. it is incredibly frustrating having to always correct this as I have done dozens of times.

The rest of the difference is likely due to the fact that RSS is colder than many of the other tropospheric analyses such as STAR, RICH, and RAOBCORE.

STAR, RICH and RAOBCORE all show more warming than GISS and so if we conducted the same type of analyses Zucker is conducting, we would find based on those three sources that GISS had falsely extrapolated too cold more often than not.

You are falsely assuming that RSS is the ultimate arbiter of the correctness of GISS extrapolations. Any differences between RSS and GISS overall are likely not due to extrapolation of GISS, but rather measurement error of one or both data sources.

Nobody has yet responded to the fact that HadCRUT+UAH arctic ~= GISS (almost). You really need to be tossing HadCRUT under the bus with GISS if you are going to toss GISS under the bus.

In essence you are taking UAH/RSS over HadCRUT, GISS, STAR, RAOBCORE, and RICH even though STAR, RICH and RAOBCORE are all more modern methods.

[ORH_wxman]

Except that GISS has consistently run warmer than other sources over most the globe, using the same baseline. Clearly, the warm extrapolations are occurring to a greater degree than the cold extrapolations. It does not balance out evenly, given what other sources show, as you claim it should.

GISS also has a much warmer trend than Hadley (also sfc based) in areas that Hadley has measurements which shows that the overall difference between the two is not just the extrapolation and "Well Hadley doesn't show the areas closest to north pole, so its too cold". If you look at where Hadley actually has data, GISS is significantly warmer than them in both the arctic and antarctic regions. In fact, Hadley shows a significant cooling in the Antarctic region where it has measurements and GISS shows a warming.

There was a post last year on WUWT about the divergence occurring in the two data sets and one of the theories was that GISS was deleting SST anomalies near the poles and just extrapolating land based data.

Here was the differences between the two data sets over the past 3 decades (or close to it)....notice how much warmer GISS gets as you get near each polar region...CRU and GISS actually agree pretty good between 60N and 60S.

[tacoman25]

As I pointed out, the vast majority of the difference between the two maps he posted was due to the fact that the GISS image used the cold 1951-1980 baseline, while the RSS map used the 1979-2000 baseline. I wish people could get their baselines straight.. it is incredibly frustrating having to always correct this as I have done dozens of times.

The rest of the difference is likely due to the fact that RSS is colder than many of the other tropospheric analyses such as STAR, RICH, and RAOBCORE.

STAR, RICH and RAOBCORE all show more warming than GISS and so if we conducted the same type of analyses Zucker is conducting, we would find based on those three sources that GISS had falsely extrapolated too cold more often than not.

You are falsely assuming that RSS is the ultimate arbiter of the correctness of GISS extrapolations. Any differences between RSS and GISS overall are likely not due to extrapolation of GISS, but rather measurement error of one or both data sources.

Nobody has yet responded to the fact that HadCRUT+UAH arctic ~= GISS (almost). You really need to be tossing HadCRUT under the bus with GISS if you are going to toss GISS under the bus.

In essence you are taking UAH/RSS over HadCRUT, GISS, STAR, RAOBCORE, and RICH even though STAR, RICH and RAOBCORE are all more modern methods.

Those are not recognized, major global temperature sources like RSS and UAH are.

The baseline issue is irrelevant. If you use the same baseline, GISS still runs warmer over the majority of the globe. Little differences here and there add up, and sometimes the differences are major.

[skierinvermont]

Those are not recognized, major global temperature sources like RSS and UAH are.

The baseline issue is irrelevant. If you use the same baseline, GISS still runs warmer over the majority of the globe. Little differences here and there add up, and sometimes the differences are major.

Yes they are, in the peer reviewed literature. You just don't see them popping up on blog sites as much yet. I expect they will replace other data sources in the next IPCC report.

Zucker and Bethesda have claimed the GISS always appears to extrapolate too warm and part of their argument has been based on trying to find areas where it looks to them GISS extrapolated too warm.. but when we do the comparison using the same baseline we see many areas where GISS extrapolated far too cold. From 1979-2010 there is essentially no divergence for RSS and GISS (maybe .01C/decade). Clearly the extrapolations are not leading to a long term bias. Especially considering RSS is colder than STAR, RAOBCORE, and RICH all three of which show more warming than GISS.

[skierinvermont]

GISS also has a much warmer trend than Hadley (also sfc based) in areas that Hadley has measurements which shows that the overall difference between the two is not just the extrapolation and "Well Hadley doesn't show the areas closest to north pole, so its too cold". If you look at where Hadley actually has data, GISS is significantly warmer than them in both the arctic and antarctic regions. In fact, Hadley shows a significant cooling in the Antarctic region where it has measurements and GISS shows a warming.

There was a post last year on WUWT about the divergence occurring in the two data sets and one of the theories was that GISS was deleting SST anomalies near the poles and just extrapolating land based data.

Here was the differences between the two data sets over the past 3 decades (or close to it)....notice how much warmer GISS gets as you get near each polar region...CRU and GISS actually agree pretty good between 60N and 60S.

The good folks at WUWT could just head on over to the GISS site which says right up front that in computing the anomalies for grid cells, priority is given to nearby surface data over SST data, when surface data is available. I guess they missed this part...

Step 5 : Computation of LOTI zonal means----------------------------------------The same method as in step3 is used, except that for a particular grid boxthe anomaly or trend is computed twice, first based on surface data, thenbased on ocean data. Depending on the location of the grid box, one orthe other is used with priority given to the surface data, if available.

Clearly SSTs will warm slower than the surface air above them, so using land data when available but SST data for the rest of the oceans is probably a reasonable compromise. The fact that HadCRUT and GISS both use SST data instead of air data for the oceans could be seen as biasing them too cold, which is why many people just use the Station only data.

[ORH_wxman]

The good folks at WUWT could just head on over to the GISS site which says right up front that in computing the anomalies for grid cells, priority is given to nearby surface data over SST data, when surface data is available. I guess they missed this part...

Step 5 : Computation of LOTI zonal means----------------------------------------The same method as in step3 is used, except that for a particular grid boxthe anomaly or trend is computed twice, first based on surface data, thenbased on ocean data. Depending on the location of the grid box, one orthe other is used with priority given to the surface data, if available.

Clearly SSTs will warm slower than the surface air above them, so using land data when available but SST data for the rest of the oceans is probably a reasonable compromise. The fact that HadCRUT and GISS both use SST data instead of air data for the oceans could be seen as biasing them too cold, which is why many people just use the Station only data.

So using a land station hundreds of miles away is better?

Clearly Hadley doesn't think so. I wonder why GISS is correct and Hadley is wrong then.

Still doesn't explain the gross differences between Hadley and GISS in the predominately land based Arctic region.

[skierinvermont]

So using a land station hundreds of miles away is better?

Clearly Hadley doesn't think so. I wonder why GISS is correct and Hadley is wrong then.

Still doesn't explain the gross differences between Hadley and GISS in the predominately land based Arctic region.

The graph from WUWT is extremely misleading, most likely deliberately so. It is comparing calculated latitude bands. For HadCRUT, where portions of a latitude band have no data, they are assumed to have the global anomaly. Considering HadCRUT leaves most of the earth north of 70N and south of 70S blank... those latitude bands will be presumed to be warming essentially at the global rate. Didn't it occur to you how WUWT is comparing GISS to HadCRUT north of 70N even though HadCRUT leaves nearly everything north of 70N blank (and much of 60-70N is blank too which lowers the anomaly there artificially as well).

When WUWT goes and compares GISS vs HadCRUT in those latitude bands it is essentially comparing GISS's extrapolated data to HadCRUT's blank data (which is presumed to have the same anomaly as the globe overall).

Go look at a map of HadCRUT... it's nearly entirely blank north of 70N.... how the hell do you think WUWT is comparing GISS and HadCRUT at those latitudes? How do you compare rapid, albeit extrapolated, warming (which we know is occurring from a multitude of sources) to BLANK DATA?????

The more blanks there are in a particular latitude band.. the lower the latitudes band's anomaly. And the higher in latitude we go... the more blanks there on HadCRUT until we get to 65-70N where the anomaly on HadCRUT is essentially entirely blank and therefore assumed to be = to the global anomaly.

This is a blatant trick on the part of WUWT.

[ORH_wxman]

The graph from WUWT is extremely misleading, most likely deliberately so. It is comparing calculated latitude bands. For HadCRUT, where portions of a latitude band have no data, they are assumed to have the global anomaly. Considering HadCRUT leaves most of the earth north of 70N and south of 70S blank... those latitude bands will be presumed to be warming essentially at the global rate.

When WUWT goes and compares GISS vs HadCRUT in those latitude bands it is essentially comparing GISS's extrapolated data to HadCRUT's blank data (which is presumed to have the same anomaly as the globe overall).

Go look at a map of HadCRUT... it's nearly entirely blank north of 70N.... how the hell do you think WUWT is comparing GISS and HadCRUT at those latitudes? How do you compare rapid, albeit extrapolated, warming (which we know is occurring from a multitude of sources) to BLANK DATA?????

This is blatant trick on the part of WUWT.

CRUs data gets pretty razor thin by 75N so yes, the graph can "look worse than it really is"....but why is it nearly universally colder?...even before you get to that point, everything past 50 degrees is almost always warmer on GISS. The globe doesn't have a lot of coverage anyway by 80 degrees...its small up there relatively speaking. But those bands between 50-70N are certainly a much more significant area and GISS has run warmer in those areas too where CRU has real coverage. Regardless if we accuse WUWT of making the graph look worse, there's still a distinct difference in CRU and GISS data where CRU has good coverage.

I think the SST explanation is quite valid and something that should be looked into more for GISS. They say on their site:

“Areas covered occasionally by sea ice are masked using a time-independent mask.”

So basically any time any of the sea has any ice in it, they just ignore the SST data (which they don't on the rest of the globe) and use their "mask". CRU uses the SST data when it can to contribute to their anomaly data. It seems pretty silly to basically guess the anomaly if there is sea ice on it based from a land station which could be completely different. They obviously don't have a problem using SST data elsewhere. This is already in addition to them running warmer than CRU in predominantly land based arctic regions.

[nzucker]

I have explained that mathematically the process of extrapolation leads to a falsely cold extrapolation just as frequently as a falsely warm extrapolation multiple times, unless there is reason to think that warmer anomalies than the surrounding areas somehow congregate over GISS sensors. You have not responded to this point or provided evidence that warm anomalies are significantly more likely to fall on sensors. I have also provided countless examples of where GISS extrapolates too cold. You also have failed to respond, why if GISS extrapolations are too warm, HadCRUT+UAH in the arctic ~= GISS. And despite your lack of response to these very reasonable and universally accepted points among peer-reviewed articles, you continue to be belligerent and rude on the subject accusing me of personal vendetta just to "bother skeptics" and "FAIL."

P.S. You continue to compare GISS 1951-1980 to RSS 1979-2000 even after I have pointed out that you need to change the GISS period to 1979-2000 to make a fair comparison. Once that has been performed, we notice that GISS is vastly too cold across Mexico, SW U.S., northwest Russia, northern Scandinavia.. etc. etc. etc.

I mean christs! There are -4C anomalies across a wide area of the northwest Russian coast and offshore where RSS has positive anomalies. Clearly there is a conspiracy afoot! GISS is extrapolating brutally cold -4C anomalies where we know it was above average! That is the Law of Large Numbers for you.

After I posted the initial maps, I switched GISS to a 1979-2000 baseline to match RSS, and I still found that its anomaly exceeded RSS by .08C. So where is this difference coming from? There still appears to be some areas in which GISS extrapolated warmer than what the satellites measured. So how do we explain this? Is global warming theory flawed in expecting the lower troposphere to warm more than the surface? Or does GISS, intentionally or by historical coincidence, extrapolate too warm in low-data regions, exaggerating the warmth of the planet? It doesn't matter if GISS is too cold in some regions compared to RSS; we're only looking for regions which have been extrapolated warmer than the satellites since the recent divergence is in the direction of higher global anomalies. Why do you keep repeating the same flawed argument about GISS being colder in some areas, when you know this is not the point?

I would like to see you provide the original data for GISS stations and the method of extrapolation used; the burden of proof is on you to prove that GISS is accurate considering it has developed a substantial divergence from the other data sources. How is extrapolation done? If it's done by a "majority" method in places like the Arctic (the area which requires the most extrapolation along with Antarctica and Africa) which have largely been experiencing above-normal temperatures compared to historical baselines, then it's likely to miss small islands of normal/cool weather in a sea of warmth. For example, if there's a point in Northern Canada whose temperature is unknown, and we take the ten closest stations which average 250mi from the point desired, then wouldn't it stand to reason that these areas rate to be above average? How would you ever be able to pick up on smaller areas of normal/cool temperatures in a large area of warmer values? I've pointed out several examples of this seeming extrapolation error from the December GISS/RSS maps, and you've never refuted my theory. How can you ever find small islands of cold in a sea of warmth with the type of extrapolation used by GISS, especially considering that some claim UHI and siting biases as well?

This is not a question of the Law of Large Numbers, it's about an extrapolation technique that's biased towards warmth because it's being used in a generally warm area. Secondly, GISS has only shown a divergence in the last few years, so we're not talking about a huge sample size. I find you often resort to statistical explanations of common sense questions in order to prove a superiority that doesn't really exist. Something is either wrong with GISS extrapolations, global warming theory about the LT, or the other sources...and we need to prove what it is.

I have not been rude or belligerent at all...I just think it shows a particularly strong bias when you choose to do your climate model using only GISS data, and then present the results in a forum in which some of the meteorologists and many of the members think GISS is too warm. You're clearly trying to stir up controversy...why not just use an average of all four main global temperature measurements (RSS, UAH, GISS, Hadley) since all of them have demonstrated some biases, inaccuracies, and problems in the past? So far, none of the sources has been free of censure for improper calibrations and actual data mistakes, so why not just use a mean to reduce error? In my mind, UAH is probably the most accurate of the data sources since it measures the whole globe without surface bias; given that everyone has a different opinion of them and you're presenting the results to people who all believe in different sources, why not just use an average? I mean, you can't be serious about using GISS when I've provided countless examples of areas which have been extrapolated too warm, none of which you've been able to refute with a corroboration of GISS, and knowing that the overall anomaly comes out higher, to some degree a result of extrapolations. Also, why are you constantly so suspicious of me and my intentions to critique your posts? Just a year ago you were mostly using satellite temperatures on our radio show, and seemed to exhibit a preference for the satellites, and now you're switching to only GISS...can't you understand why someone might launch a few barbs against that?

Anyway dude just enjoy talking about the weather and stop making it such a competition. I think you take this stuff WAY too seriously, this is about having fun. No one really cares that much if GISS is too warm, I certainly don't lose sleep over it.

[nzucker]

Skier, you still haven't answered Will's question about GISS not using available SST data and instead making land-based extrapolations. Why is there a need for this, and why is GISS constantly running warmer than Hadley in the 60-75N band where the globe does contain a significant amount of area? Also, why does Hadley accuse GISS of being too warm if they are themselves huge proponents of AGW and predicting up to 4C of warming by 2060? The Hadley team must have some serious problems with GISS measurement/extrapolation techniques in order to mount such a criticism, no?

[skierinvermont]

CRUs data gets pretty razor thin by 75N so yes, the graph can "look worse than it really is"....but why is it nearly universally colder?...even before you get to that point, everything past 50 degrees is almost always warmer on GISS. The globe doesn't have a lot of coverage anyway by 80 degrees...its small up there relatively speaking. But those bands between 50-70N are certainly a much more significant area and GISS has run warmer in those areas too where CRU has real coverage. Regardless if we accuse WUWT of making the graph look worse, there's still a distinct difference in CRU and GISS data where CRU has good coverage.

I don't believe this is the case.

1) First of all, why would they? ... considering they use essentially the same data.

2) CRU has substantial blank areas 50-70N, especially 60-70N, over land in Asia, where it has warmed rapidly.

3) This can empirically be shown to be not the case. In 2005, for example, GISS LOTI had an anomaly of .56 while HadCRUT was .48. GISS with a HadCRUT mask, drops to .47, even cooler than HadCRUT.

The graph in this link shows that for the entire period 1995-present the difference between GISS and HadCRUT is entirely removed by applying a HadCRUT mask to GISS.

http://pubs.giss.nas...Hansen_etal.pdf

I believe this unambiguously disproves the notion that GISS differs from HadCRUT in regions "where CRU has good coverage." Where they both have coverage they are in agreement within .01C.

What the recent divergence really comes down to is the regions where HadCRUT leaves it blank, and GISS extrapolates. This is mostly in the arctic and Siberia where we know from numerous sources including UAH that rapid warming has occurred. And yet HadCRUT leaves these areas blank.

I think the SST explanation is quite valid and something that should be looked into more for GISS. They say on their site:

“Areas covered occasionally by sea ice are masked using a time-independent mask.”

So basically any time any of the sea has any ice in it, they just ignore the SST data (which they don't on the rest of the globe) and use their "mask". CRU uses the SST data when it can to contribute to their anomaly data. It seems pretty silly to basically guess the anomaly if there is sea ice on it based from a land station which could be completely different. They obviously don't have a problem using SST data elsewhere. This is already in addition to them running warmer than CRU in predominantly land based arctic regions.

I agree that using nearby land stations instead of SST data would lead GISS to be warmer than HadCRUT, however I am not sure that this would be incorrect. SSTs are surely warming much slower than air temperatures, especially air that may be coming off nearby landmasses. The fact that SST data is used at all anywhere on the planet probably introduces a cooling bias, and to the extent that data other than SSTs can be used, that may be a good thing. If nothing else, it is a compromise. The effect is also likely very small. We also know that this disagreement would fall under "areas covered by HadCRUT" and we know that for "areas covered by HadCRUT" there is no divergence. Thus the use of land over SSTs cannot be used to explain the divergence.

The divergence is 100% explained by the difference in coverage. Thus the only question we need ask is: Is GISS correct in those areas not covered by HadCRUT? And if we take a look at UAH.. or ERA.. or NCEP in the areas that are not covered by HadCRUT.. the answer is: by and large "Yes."

The train of logic goes:

1. HadCRUT mask to GISS eliminates 100% of the divergence.

2. Ergo, 100% of the divergence must be due to difference in coverage. (IE the areas left blank by Had, but extrapolated by GISS)

3. UAH, ERA, NCEP all confirm GISS is largely correct in areas covered by GISS but not HadCRUT.

4. Ergo, from (2) + (3), GISS is superior to HadCRUT because it is more complete.

[skierinvermont]

Skier, you still haven't answered Will's question about GISS not using available SST data and instead making land-based extrapolations. Why is there a need for this, and why is GISS constantly running warmer than Hadley in the 60-75N band where the globe does contain a significant amount of area? Also, why does Hadley accuse GISS of being too warm if they are themselves huge proponents of AGW and predicting up to 4C of warming by 2060? The Hadley team must have some serious problems with GISS measurement/extrapolation techniques in order to mount such a criticism, no?

I answered the first two questions in the above post and I reject the premise of the third. I have not seen HadCRUT people accuse GISS of being too warm. Jones had some technical criticisms of GISS in an old email, but they are not relevant to the current divergence between the two.

[skierinvermont]

After I posted the initial maps, I switched GISS to a 1979-2000 baseline to match RSS, and I still found that its anomaly exceeded RSS by .08C. So where is this difference coming from? There still appears to be some areas in which GISS extrapolated warmer than what the satellites measured. So how do we explain this? Is global warming theory flawed in expecting the lower troposphere to warm more than the surface? Or does GISS, intentionally or by historical coincidence, extrapolate too warm in low-data regions, exaggerating the warmth of the planet? It doesn't matter if GISS is too cold in some regions compared to RSS; we're only looking for regions which have been extrapolated warmer than the satellites since the recent divergence is in the direction of higher global anomalies. Why do you keep repeating the same flawed argument about GISS being colder in some areas, when you know this is not the point?

I would like to see you provide the original data for GISS stations and the method of extrapolation used; the burden of proof is on you to prove that GISS is accurate considering it has developed a substantial divergence from the other data sources. How is extrapolation done? If it's done by a "majority" method in places like the Arctic (the area which requires the most extrapolation along with Antarctica and Africa) which have largely been experiencing above-normal temperatures compared to historical baselines, then it's likely to miss small islands of normal/cool weather in a sea of warmth. For example, if there's a point in Northern Canada whose temperature is unknown, and we take the ten closest stations which average 250mi from the point desired, then wouldn't it stand to reason that these areas rate to be above average? How would you ever be able to pick up on smaller areas of normal/cool temperatures in a large area of warmer values? I've pointed out several examples of this seeming extrapolation error from the December GISS/RSS maps, and you've never refuted my theory. How can you ever find small islands of cold in a sea of warmth with the type of extrapolation used by GISS, especially considering that some claim UHI and siting biases as well?

This is not a question of the Law of Large Numbers, it's about an extrapolation technique that's biased towards warmth because it's being used in a generally warm area. Secondly, GISS has only shown a divergence in the last few years, so we're not talking about a huge sample size. I find you often resort to statistical explanations of common sense questions in order to prove a superiority that doesn't really exist. Something is either wrong with GISS extrapolations, global warming theory about the LT, or the other sources...and we need to prove what it is.

I have not been rude or belligerent at all...I just think it shows a particularly strong bias when you choose to do your climate model using only GISS data, and then present the results in a forum in which some of the meteorologists and many of the members think GISS is too warm. You're clearly trying to stir up controversy...why not just use an average of all four main global temperature measurements (RSS, UAH, GISS, Hadley) since all of them have demonstrated some biases, inaccuracies, and problems in the past? So far, none of the sources has been free of censure for improper calibrations and actual data mistakes, so why not just use a mean to reduce error? In my mind, UAH is probably the most accurate of the data sources since it measures the whole globe without surface bias; given that everyone has a different opinion of them and you're presenting the results to people who all believe in different sources, why not just use an average? I mean, you can't be serious about using GISS when I've provided countless examples of areas which have been extrapolated too warm, none of which you've been able to refute with a corroboration of GISS, and knowing that the overall anomaly comes out higher, to some degree a result of extrapolations. Also, why are you constantly so suspicious of me and my intentions to critique your posts? Just a year ago you were mostly using satellite temperatures on our radio show, and seemed to exhibit a preference for the satellites, and now you're switching to only GISS...can't you understand why someone might launch a few barbs against that?

Anyway dude just enjoy talking about the weather and stop making it such a competition. I think you take this stuff WAY too seriously, this is about having fun. No one really cares that much if GISS is too warm, I certainly don't lose sleep over it.

Actually the burden of proof is on you. I have already shown that GISS is NOT diverging from other data sources. STAR, RICH, RAOBCORE are all warmer than it. HadCRUT infilled with UAH agrees with GISS. All of this suggests that GISS is more accurate over the last 15 years, and that by missing large areas which have rapidly warmed, HadCRUT is biased too cold. HadCRUT runs colder than STAR, RICH, RAOBCORE, and HadCRUT+UAH infilling.

HadCRUT+UAH infilling ~= GISS is a definitive demonstration that the GISS extrapolations are NOT too warm. It is also a definitive demonstration that by missing areas which have warmed rapidly according to UAH, HadCRUT must according to BASIC LOGIC run too cold.

If you are interested in the technical process of extrapolation it is all explained very clearly in the relevant links on the GISTEMP homepage. Your idea that somehow extrapolating in an environment of general warmth but with pockets of cold will lead to a persistent bias is simply incorrect. I beg you to stop and think about this. If one of those pockets of cold falls on a GISS station, that pocket of cold will be extrapolated incorrectly 800+KM. When general warmth falls on a GISS station it will correctly extrapolate that general warmth over the surrounding area which is generally also warm. Occasionally it will cover up one of the cold pockets. In other words it will frequently cover up small patches of cold with warmth and occasionally cover up large patches of warmth with cold. If you cannot figure this out in your head, there are statistical studies included in the links on the GISS homepage which show that extrapolation does not create a persistent bias (which is completely redundant if you understand the theory).

There also is no difference between GISS, RSS, or HadCRUT over the last 30 years. STAR, RICH, and RAOBCORE run warmer. And yet you select UAH which runs cooler than any of the other 6.

[ORH_wxman]

I don't believe this is the case.

1) First of all, why would they? ... considering they use essentially the same data.

2) CRU has substantial blank areas 50-70N, especially 60-70N, over land in Asia, where it has warmed rapidly.

3) This can empirically be shown to be not the case. In 2005, for example, GISS LOTI had an anomaly of .56 while HadCRUT was .48. GISS with a HadCRUT mask, drops to .47, even cooler than HadCRUT.

The graph in this link shows that for the entire period 1995-present the difference between GISS and HadCRUT is entirely removed by applying a HadCRUT mask to GISS.

http://pubs.giss.nas...Hansen_etal.pdf

I believe this unambiguously disproves the notion that GISS differs from HadCRUT in regions "where CRU has good coverage." Where they both have coverage they are in agreement within .01C.

What the recent divergence really comes down to is the regions where HadCRUT leaves it blank, and GISS extrapolates. This is mostly in the arctic and Siberia where we know from numerous sources including UAH that rapid warming has occurred. And yet HadCRUT leaves these areas blank.

I agree that using nearby land stations instead of SST data would lead GISS to be warmer than HadCRUT, however I am not sure that this would be incorrect. SSTs are surely warming much slower than air temperatures, especially air that may be coming off nearby landmasses. The fact that SST data is used at all anywhere on the planet probably introduces a cooling bias, and to the extent that data other than SSTs can be used, that may be a good thing. If nothing else, it is a compromise.

1 & 2. They do not use the same data. A lot of it is close but CRU actually has better data coverage in the sparser arctic and antarctic regions (as well as Africa) where GISS has been showing significant warming. GISS has better coverage in the U.S. which is not surprising.

3. That's a nice example, but the argument is flawed in that GISS doesn't use CRU data to extrapolate, they use their own. So while the overall "hard values" might be similar over the entire globe, the critical regions are different which of course matters if that's what GISS is using as extrapolation to other areas. If you remember from the earlier graph, Hadley was actually slightly warmer than GISS in parts of the southern hemisphere, but there's not as much extrapolation going there. In the region where GISS makes most of their huge warm extrapolations, their starting points are worse in coverage and warmer to begin with.

As for SST data being actually less correct than extrapolating from a land based station sometimes hundreds of miles away....not sure how that could possibly be more accurate except out of dumb luck. It seems you might be applying a predetermined global warming argument to justify the data..."the water warms slower than the air so it probably is showing a cool bias". That's a nice argument in theory, but its still ignoring actual data. Using predetermined global warming arguments to justify the data rather than the other way around is biased science. What if we are cooling into the next colder ocean cycle?

So instead of using actual data, GISS ignores any SST data where sea ice is present at all during the year, and then masks it with their best estimate based on a few land stations...and we're supposed to believe they are correct vs other agencies? Maybe they are right and every spot they have that Hadley doesn't is heating up drastically this decade enough to cause the divergence to be that significant, but I'm not buying it. Especially since on their maps you can see some disagreement in the spots where Hadley actually does have data up in the arctic and antarctic. GISS is the only operational temp agency that thinks Antarctica is warming. At the very least, they have very questionable results whether they have some sort of "explanation" for every discrepancy vs other agencies or not.

[nzucker]

I answered the first two questions in the above post and I reject the premise of the third. I have not seen HadCRUT people accuse GISS of being too warm. Jones had some technical criticisms of GISS in an old email, but they are not relevant to the current divergence between the two.

Will, can you give a more detailed explanation of Hadley's criticism of GISS, including some quotes?

Also, can you provide some actual station data for GISS/Hadley where a divergence is shown?

It seems a lot of people are making theoretical comments without giving specific quotes or data; as a former journalist, I'd like to see some hard evidence, and this applies to everyone.

Actually the burden of proof is on you. I have already shown that GISS is NOT diverging from other data sources. STAR, RICH, RAOBCORE are all warmer than it. HadCRUT infilled with UAH agrees with GISS. All of this suggests that GISS is more accurate over the last 15 years, and that by missing large areas which have rapidly warmed, HadCRUT is biased too cold. HadCRUT runs colder than STAR, RICH, RAOBCORE, and HadCRUT+UAH infilling.

HadCRUT+UAH infilling ~= GISS is a definitive demonstration that the GISS extrapolations are NOT too warm. It is also a definitive demonstration that by missing areas which have warmed rapidly according to UAH, HadCRUT must according to BASIC LOGIC run too cold.

If you are interested in the technical process of extrapolation it is all explained very clearly in the relevant links on the GISTEMP homepage. Your idea that somehow extrapolating in an environment of general warmth but with pockets of cold will lead to a persistent bias is simply incorrect. I beg you to stop and think about this. If one of those pockets of cold falls on a GISS station, that pocket of cold will be extrapolated incorrectly 800+KM. When general warmth falls on a GISS station it will correctly extrapolate that general warmth over the surrounding area. Occasionally it will cover up one of the cold pockets. If you cannot figure this out in your head, there are statistical studies included in the links on the GISS homepage which show that extrapolation does not create a persistent bias (which is completely redundant if you understand the theory).

There also is no difference between GISS, RSS, or HadCRUT over the last 30 years. STAR, RICH, and RAOBCORE run warmer. And yet you select UAH which runs cooler than any of the other 6.

I'm just saying that if multiple stations are used to extrapolate a single data point, this might result in warmth because the probability of finding a dozen stations in the Arctic cooler than average would be much lower than one station being a "cold pocket." I don't know if this is the technique used, but I could see how it might create problems in determining the presence of colder areas in a warming region.

The problem with your explanation is that it still doesn't explain why GISS has diverged from the other sources. You can't claim satellite error because the satellites have, overall, become more accurate in the last few years with the introduction of AMSU data, and they actually used to match GISS when they were less accurate. With the satellites becoming better at measuring the global temperature, why are they showing less warming? Is there a fundamental problem with AGW theory in expecting the LT to warm more, and could this be attributed to an underestimation of the sun's role in modifying the global climate? You easily explain why GISS is warmer than Hadley, and it's a very reasonable and well-thought out answer, but you still haven't found a way to respond to the gap between the satellites and GISS, especially UAH. If the satellites matched GISS well from 1979-1998 (approximately), then both sources must have been in error if you're claiming that STAR is a better method. Was GISS too cool the entire time? And now it has suddenly warmed under the politicized Hansen's instruction? Hmmm...

I just use UAH because I think satellites are ultimately a better measure of global temperature than surface stations considering the controversy about drop-out and siting problems. WUWT has shown some stations used to measure global temperature have very dubious siting positions, and I'm extremely sensitive to these issues given that I live in a very urbanized area and have noticed the dramatic differences in temperature driving a few miles away/closer to NYC. Maybe it's a pet peeve of mine but I'm always going to be concerned about UHI since I've lived in an area where variations of 5-10F in a distance of 10 miles are not that uncommon on clear, calm nights. It's inevitable that stations in an extremely rural area like the Arctic may run a bit warmer. I also know that UAH has more coverage of the globe than RSS, and none of the sources using SNO seem to be able to explain why the satellites were able to replicate the surface data without these techniques until a few years ago. There must have been a mistake somewhere.

[skierinvermont]

Will, can you give a more detailed explanation of Hadley's criticism of GISS, including some quotes?

Also, can you provide some actual station data for GISS/Hadley where a divergence is shown?

It seems a lot of people are making theoretical comments without giving specific quotes or data; as a former journalist, I'd like to see some hard evidence, and this applies to everyone.

I'm just saying that if multiple stations are used to extrapolate a single data point, this might result in warmth because the probability of finding a dozen stations in the Arctic cooler than average would be much lower than one station being a "cold pocket." I don't know if this is the technique used, but I could see how it might create problems in determining the presence of colder areas in a warming region.

The problem with your explanation is that it still doesn't explain why GISS has diverged from the other sources. You can't claim satellite error because the satellites have, overall, become more accurate in the last few years with the introduction of AMSU data, and they actually used to match GISS when they were less accurate. With the satellites becoming better at measuring the global temperature, why are they showing less warming? Is there a fundamental problem with AGW theory in expecting the LT to warm more, and could this be attributed to an underestimation of the sun's role in modifying the global climate? You easily explain why GISS is warmer than Hadley, and it's a very reasonable and well-thought out answer, but you still haven't found a way to respond to the gap between the satellites and GISS, especially UAH. If the satellites matched GISS well from 1979-1998 (approximately), then both sources must have been in error if you're claiming that STAR is a better method. Was GISS too cool the entire time? And now it has suddenly warmed under the politicized Hansen's instruction? Hmmm...

I just use UAH because I think satellites are ultimately a better measure of global temperature than surface stations considering the controversy about drop-out and siting problems. WUWT has shown some stations used to measure global temperature have very dubious siting positions, and I'm extremely sensitive to these issues given that I live in a very urbanized area and have noticed the dramatic differences in temperature driving a few miles away/closer to NYC. Maybe it's a pet peeve of mine but I'm always going to be concerned about UHI since I've lived in an area where variations of 5-10F in a distance of 10 miles are not that uncommon on clear, calm nights. It's inevitable that stations in an extremely rural area like the Arctic may run a bit warmer. I also know that UAH has more coverage of the globe than RSS, and none of the sources using SNO seem to be able to explain why the satellites were able to replicate the surface data without these techniques until a few years ago. There must have been a mistake somewhere.

This is what I have been saying for months. If there is an unexplained gap.. it is the gap between satellites and expected surface warming based on the satellites. The gap between GISS and HadCRUT is 100% explained by the areas of no coverage (IE extrapolation) on HadCRUT, which if we infill with UAH,NCEP, or ERA data brings HadCRUT in line with GISS.

ERA and NCEP globally also agree more closely with GISS over the last 10-15 years than HadCRUT.

Having resolved the surface discrepancy between GISS and HadCRUT largely in favor of GISS, we are left wondering, why isn't the troposphere warming more?

The answer, as I have suggested, is that RSS and UAH are probably too cold. More recent analyses such as STAR, RAOBCORE, RICH all show more warming than RSS or UAH. They mostly show around .2C/decade of warming in the lower troposphere, which is in line with GISS's .17C/decade at the surface.

You asked why is this discrepancy between surface and satellite just coming to light now?

It's not... the discrepancy has been around forever. Lots of peer reviewed articles over the last 20 years have dealt with the question. It is only just now beginning to be resolved as UAH and RSS were revised upwards 5 years ago, and even newer data (STAR, RAOBCORE, RICH) shows even more warming.

Regarding your furthered question on the extrapolations... think about it some more... if we are extrapolating the value for a point based off 5 surrounding stations, and 4 of them are warm, and 1 of them is cold, when we conduct an AVERAGE it will not allow for a persistent bias. What you are describing is like we said "ok there are 5 nearby stations, 4 are warm, 1 is cold, let's toss the 1 cold one and use the 4 warm ones" .... of course that is not what is happening at all.

In fact in general what is happening is there is only 1 nearby station and we are extrapolating in a radius of 1200km around that station. Unless that station has for some reason warmed faster than the surrounding area, there is no persistent bias. Sometimes the stations we select will have warmed faster than the surrounding area, sometimes they will have warmed slower. According to the law of large numbers, it balances out in the end after a sample size of 60+. Which is why we are able to form an accurate index of global temperatures from a mere 80 stations and then extrapolating 3,000+ km from each of them.

[skierinvermont]

1 & 2. They do not use the same data. A lot of it is close but CRU actually has better data coverage in the sparser arctic and antarctic regions (as well as Africa) where GISS has been showing significant warming. GISS has better coverage in the U.S. which is not surprising.

I assume you are basing this off the WUWT post showing what they claim is the same masking for each of the sources. I have debunked this WUWT post in another thread. It only appears as if HadCRUT has more data because they are comparing Regular HadCRUT vs GISS 250km extrapolation. However, "regular" HadCRUT extrapolation is substantially more than 250km. It is often up to 500km, and they will infill any grid-squares which have a critical number of surrounding grid squares filled, which means that sometimes they are extrapolating up to 1000-1500km. By comparing this extrapolation scheme to the GISS rigid 250km scheme, WUWT successfully makes it appear as if HadCRUT has far more station data than GISS, when this is not the case.

Debunking WUWT Post on GISS Had Coverage (Link)

This is another blatant error/trick on the part of WUWT to make it appear that Had has vastly more coverage than GISS.

If you have another source showing HadCRUT to have more data than GISS I would love to see it. I wouldn't be surprised if they have a few more stations here and there, but I assume you are basing this off the WUWT post which is grossly incorrect.

Regardless, your original claim was that the two sources are different in the areas they cover. They may be regionally different, but globally, for the areas with coverage in both, they are the same.

3. That's a nice example, but the argument is flawed in that GISS doesn't use CRU data to extrapolate, they use their own. So while the overall "hard values" might be similar over the entire globe, the critical regions are different which of course matters if that's what GISS is using as extrapolation to other areas. If you remember from the earlier graph, Hadley was actually slightly warmer than GISS in parts of the southern hemisphere, but there's not as much extrapolation going there. In the region where GISS makes most of their huge warm extrapolations, their starting points are worse in coverage and warmer to begin with.

Yes, regionally they must be different. But globally, for areas covered, the trends are the same. Which means it comes down to whether or not the areas GISS covers but HadCRUT leaves blank, are correctly extrapolated by GISS. As we can see from multiple sources (UAH, NCEP, ERA) GISS is by and large correct over those areas. The fact that filling in HadCRUT with UAH satellite data in the missing regions brings it essentially into line with GISS over the last 10-15 years eliminates the divergence.

As for SST data being actually less correct than extrapolating from a land based station sometimes hundreds of miles away....not sure how that could possibly be more accurate except out of dumb luck. It seems you might be applying a predetermined global warming argument to justify the data..."the water warms slower than the air so it probably is showing a cool bias". That's a nice argument in theory, but its still ignoring actual data. Using predetermined global warming arguments to justify the data rather than the other way around is biased science. What if we are cooling into the next colder ocean cycle?

So instead of using actual data, GISS ignores any SST data where sea ice is present at all during the year, and then masks it with their best estimate based on a few land stations...and we're supposed to believe they are correct vs other agencies? Maybe they are right and every spot they have that Hadley doesn't is heating up drastically this decade enough to cause the divergence to be that significant, but I'm not buying it. Especially since on their maps you can see some disagreement in the spots where Hadley actually does have data up in the arctic and antarctic. GISS is the only operational temp agency that thinks Antarctica is warming. At the very least, they have very questionable results whether they have some sort of "explanation" for every discrepancy vs other agencies or not.

It doesn't "presume global warming." If the globe was cooling, extrapolating land based stations instead of using SSTs would yield more cooling because the land would cool more than the oceans. In ice ages, in volcanoes etc... the land always cools more than the oceans.

I am simply saying that using SSTs is not as good as using air temperature. SSTS, regardless of whether we are cooling or warming, will change slower and less than air temperature.

[skierinvermont]

Instead of infilling HadCRUT with UAH to bring it into agreement with GISS, an alternative demonstration is to apply a HadCRUT filter to UAH which would cut off everything above 70N on UAH and much of 50-70N as well.

This would probably reduce the UAH trend from .14C/decade to .something like .11 or .12C/decade... much smaller than HadCRUT's .16C/decade. Thus again we see the disagreement is between the surface and the troposphere (and the troposphere data sources amongst themselves).

The divergence between Had and GISS is largely resolved by filling in Had with UAH ... it is imply a question of coverage... they agree on the areas they cover... and when you infill the missing coverage with UAH.. they agree on the globe.

We are then left looking at trying to resolve the surface trends with the tropospheric trends which are in dissaray (.14C UAH, .16C RSS, .2C STAR, RAOBCORE, RICH).

[nzucker]

1)You asked why is this discrepancy between surface and satellite just coming to light now?

It's not... the discrepancy has been around forever. Lots of peer reviewed articles over the last 20 years have dealt with the question. It is only just now beginning to be resolved as UAH and RSS were revised upwards 5 years ago, and even newer data (STAR, RAOBCORE, RICH) shows even more warming.

2)Regarding your furthered question on the extrapolations... think about it some more... if we are extrapolating the value for a point based off 5 surrounding stations, and 4 of them are warm, and 1 of them is cold, when we conduct an AVERAGE it will not allow for a persistent bias. What you are describing is like we said "ok there are 5 nearby stations, 4 are warm, 1 is cold, let's toss the 1 cold one and use the 4 warm ones" .... of course that is not what is happening at all.

In fact in general what is happening is there is only 1 nearby station and we are extrapolating in a radius of 1200km around that station. Unless that station has for some reason warmed faster than the surrounding area, there is no persistent bias. Sometimes the stations we select will have warmed faster than the surrounding area, sometimes they will have warmed slower. According to the law of large numbers, it balances out in the end after a sample size of 60+. Which is why we are able to form an accurate index of global temperatures from a mere 80 stations and then extrapolating 3,000+ km from each of them.

1) The satellites did not diverge significantly from GISS/HadCRUT in the early stages of their existence. In fact, RSS and UAH averaged together show a warming trend of .15C/decade, which is just .03C less than GISS. However, in recent months, the difference between the GISS anomaly and the UAH/RSS anomaly is averaging .1C or even more, in some cases. Why is this so, if the satellites have generally improved in measuring global temperature? Why would the difference in the surface and LT become exaggerated in the past 10 years, if they were warming at approximately the same rate for 20 years? Also, if you claim that UAH/RSS temperatures are too cold and need to be adjusted using SNO, wouldn't this imply a far greater warmer than GISS in the 1979-1999 time-frame, and thus cast doubt on GISS for being too cold? It just seems weird that you'd see a totally opposite trend in the lower troposphere and surface between the 1979-1998 period and the more recent period.

2) But how can you ever detect small pockets of cooler anomalies if you're constantly using the average of a large number of stations? Wouldn't the fact that most stations are warming, that the stations are closer to the coast, and that the stations are sited in relatively populated areas preclude these averages from being colder than average? For example, if 90% of the arctic is warmer than average with 10% colder than average, how can I ever find the colder 10% if it's not in an area with a lot of stations? Wouldn't the fact that you're needing to use a large number of stations in a generally warm area to make an "accurate" extrapolation make it hard to see a small pocket of cold?

You can argue this mathematically, but there's clearly some examples where this extrapolation is not working properly. Look at the GISS 1979-2000 map versus RSS map for February 2011....see that area of cooler anomalies in Central South Greenland? That's been missed by GISS and picked up on the satellite analysis. How can we trust a data source if it's clearly missing smaller pockets of temperature variation by extrapolation, given that it's constantly coming in warmer than other sources? And has any study ever been done to show how much of the average .1C deviation between GISS and the satellites is due to actual tropospheric coolness compared to extrapolation mistakes? Could you link me to something of this nature?

I think your example of using 60 stations to extrapolate's the globe's temperature is different....those stations would be equally placed, with some in areas that have warmed less/more. But if you do end up with a small area of cold ringed by warm stations, as was the case in South Central Greenland during February 2011, you're going to end up with a mistake on GISS, no? I just don't see any reason to ignore the subtlety of the satellites in measuring global temperature when there is a more specific analysis available. Can you prove that the extrapolations are coming out exactly even in the short-term period of 5-10 years, as you have suggested? Can you prove that it's just a difference between LT and surface temps, especially in light of the fact that Hadley has criticized GISS for being too warm in areas that the British agency measures as well...

[skierinvermont]

1) The satellites did not diverge significantly from GISS/HadCRUT in the early stages of their existence. In fact, RSS and UAH averaged together show a warming trend of .15C/decade, which is just .03C less than GISS. However, in recent months, the difference between the GISS anomaly and the UAH/RSS anomaly is averaging .1C or even more, in some cases. Why is this so, if the satellites have generally improved in measuring global temperature? Why would the difference in the surface and LT become exaggerated in the past 10 years, if they were warming at approximately the same rate for 20 years? Also, if you claim that UAH/RSS temperatures are too cold and need to be adjusted using SNO, wouldn't this imply a far greater warmer than GISS in the 1979-1999 time-frame, and thus cast doubt on GISS for being too cold? It just seems weird that you'd see a totally opposite trend in the lower troposphere and surface between the 1979-1998 period and the more recent period.

Satellites have not generally improved. The conversion from MSU to AMSU if anything creates more room for error during this recent period.

You suggest that SNO implies GISS is too cold.. it doesn't. Troposphere is supposed to warm faster.

All I can say is that GISS warms at .17C/decade 1979-present while STAR is more like .2C/decade, which is about what you would expect. Since I don't have short time periods, I can't break it down. Moreover, I don't think such an exercise would be very useful as I wouldn't expect perfect 1.1X surface - troposphere correspondence over short periods.

GISS is in general agreement with Had +UAH infilling which is in general agreement with expectations for the the troposphere based on STAR, RAOBCORE, and RICH for the troposphere.

2) But how can you ever detect small pockets of cooler anomalies if you're constantly using the average of a large number of stations? Wouldn't the fact that most stations are warming, that the stations are closer to the coast, and that the stations are sited in relatively populated areas preclude these averages from being colder than average? For example, if 90% of the arctic is warmer than average with 10% colder than average, how can I ever find the colder 10% if it's not in an area with a lot of stations? Wouldn't the fact that you're needing to use a large number of stations in a generally warm area to make an "accurate" extrapolation make it hard to see a small pocket of cold?

No not at all.. in the arctic where there are few stations if a bubble of cold falls on a GISS station... this bubble of cold will be falsely extrapolated for 1200km in every direction. No bias.

In the U.S. where there are many stations, if 2 stations in a cell are 1C above average and 2 in the same cell are 1C below average, then the cell will be represented as 0C anomaly. No bias.

You can argue this mathematically, but there's clearly some examples where this extrapolation is not working properly. Look at the GISS 1979-2000 map versus RSS map for February 2011....see that area of cooler anomalies in Central South Greenland? That's been missed by GISS and picked up on the satellite analysis. How can we trust a data source if it's clearly missing smaller pockets of temperature variation by extrapolation, given that it's constantly coming in warmer than other sources? And has any study ever been done to show how much of the average .1C deviation between GISS and the satellites is due to actual tropospheric coolness compared to extrapolation mistakes? Could you link me to something of this nature?

There are also plenty of examples where it is extrapolating too cold. In the long run GISS and RSS are nearly in complete agreement (maybe .01C/decade divergence). STAR, RAOBCORE RICH would show more warming.

We have NCEP and ERA reanalysis which don't extrapolate at all, and confirm the GISS trend as well as GISS's faster warming than Had over the last 10-15 years. The problem is clearly not extrapolation.. anybody with a rudimentary understanding of spatial averaging understands this. ERA, NCEP and HadCRUT+UAH infilling all confirm the GISS surface trends over the last 15-30 years. The problem is that RSS somewhat, but especially UAH, are not warming as fast as expected. This discrepancy is resolved by STAR, RAOBCORE, RICH.

I think your example of using 60 stations to extrapolate's the globe's temperature is different....those stations would be equally placed, with some in areas that have warmed less/more. But if you do end up with a small area of cold ringed by warm stations, as was the case in South Central Greenland during February 2011, you're going to end up with a mistake on GISS, no? I just don't see any reason to ignore the subtlety of the satellites in measuring global temperature when there is a more specific analysis available. Can you prove that the extrapolations are coming out exactly even in the short-term period of 5-10 years, as you have suggested? Can you prove that it's just a difference between LT and surface temps, especially in light of the fact that Hadley has criticized GISS for being too warm in areas that the British agency measures as well...

Hadley has not accused GISS of that. .. please post.

I have already proven it is a surface vs LT discrepancy. ERA, NCEP, and HadCRUT+UAH infilling all confirm GISS over the last 15-30 years. And the surface vs LT discrepancy only exists if you are using UAH or RSS.

The 60 stations do not need to be equally placed.

[nzucker]

Roy Spencer claims that AMSU is a significant improvement over MSU. Why do you disagree with this and think it creates more error?

If you are using JRA/NCEP, you should read this article about mistakes made by these sources on Jan Mayen Island in their attempts to reproduce past temperatures:

http://wattsupwiththat.com/2011/03/08/models-all-the-way-down/#more-35478

Considering you believe in SNO/STAR, you are saying that when the satellites and surface data agreed, the satellites were still too cool because the troposphere should be warming faster? Also, if you go with STAR over GISS, you still get about the same rate of warming as GISS, near .2C/decade, not a significant difference as faster warming of the troposphere would imply. I believe GISS has shown a warming trend of .18C/decade since 1979, so this wouldn't even be close to statistically significant. Also, in order to consider STAR a legitimate data source, I'd like to see a table of its monthly readings, maps of its anomalies, and the overall trend for different periods. Where is this data available? How can we do a reliability check on STAR?

[skierinvermont]

Roy Spencer claims that AMSU is a significant improvement over MSU. Why do you disagree with this and think it creates more error?

If you are using JRA/NCEP, you should read this article about mistakes made by these sources on Jan Mayen Island in their attempts to reproduce past temperatures:

http://wattsupwithth...own/#more-35478

Considering you believe in SNO/STAR, you are saying that when the satellites and surface data agreed, the satellites were still too cool because the troposphere should be warming faster? Also, if you go with STAR over GISS, you still get about the same rate of warming as GISS, near .2C/decade, not a significant difference as faster warming of the troposphere would imply. I believe GISS has shown a warming trend of .18C/decade since 1979, so this wouldn't even be close to statistically significant. Also, in order to consider STAR a legitimate data source, I'd like to see a table of its monthly readings, maps of its anomalies, and the overall trend for different periods. Where is this data available? How can we do a reliability check on STAR?

I don't disaagree with him... and I think you know that.

What I said was that the merger of the two satellites creates error.

The warming rate on STAR is about what one would expect for the troposphere given GISS. You need to use the same time periods. The GISS trend only rises to .18C/decade when you extend it to present, but STAR has not been extended to present yet (I think because they are still trying to figure out the transition from MSU to AMSU).

I have posted the STAR studies/data on here several times.

[ORH_wxman]

I assume you are basing this off the WUWT post showing what they claim is the same masking for each of the sources. I have debunked this WUWT post in another thread. It only appears as if HadCRUT has more data because they are comparing Regular HadCRUT vs GISS 250km extrapolation. However, "regular" HadCRUT extrapolation is substantially more than 250km. It is often up to 500km, and they will infill any grid-squares which have a critical number of surrounding grid squares filled, which means that sometimes they are extrapolating up to 1000-1500km. By comparing this extrapolation scheme to the GISS rigid 250km scheme, WUWT successfully makes it appear as if HadCRUT has far more station data than GISS, when this is not the case.

Debunking WUWT Post on GISS Had Coverage (Link)

This is another blatant error/trick on the part of WUWT to make it appear that Had has vastly more coverage than GISS.

If you have another source showing HadCRUT to have more data than GISS I would love to see it. I wouldn't be surprised if they have a few more stations here and there, but I assume you are basing this off the WUWT post which is grossly incorrect.

Regardless, your original claim was that the two sources are different in the areas they cover. They may be regionally different, but globally, for the areas with coverage in both, they are the same.

Yes, regionally they must be different. But globally, for areas covered, the trends are the same. Which means it comes down to whether or not the areas GISS covers but HadCRUT leaves blank, are correctly extrapolated by GISS. As we can see from multiple sources (UAH, NCEP, ERA) GISS is by and large correct over those areas. The fact that filling in HadCRUT with UAH satellite data in the missing regions brings it essentially into line with GISS over the last 10-15 years eliminates the divergence.

It doesn't "presume global warming." If the globe was cooling, extrapolating land based stations instead of using SSTs would yield more cooling because the land would cool more than the oceans. In ice ages, in volcanoes etc... the land always cools more than the oceans.

I am simply saying that using SSTs is not as good as using air temperature. SSTS, regardless of whether we are cooling or warming, will change slower and less than air temperature.

None of this addresses how GISS can be more correct by extrapolating temperatures, nor does it address the potential error where GISS is consistently warmer than Hadley in the 50-70 bands. Hadley is actually slightly warmer than GISS in the roughly 20S to 50S band and 25N to 40N band, but the only spots where GISS is warmer is the polar regions, even in areas outside of significant extrapolation. Its starting to get nitpicky and slightly off tangent argument to discuss just how much data Hadley has vs GISS in the further north bands. They are still cooler up there in regions where they measure. That can be discussed in a different argument, but the bottom line is GISS shows warming in the southern polar regions where the 3 other operational agencies show cooling...and their warming in the northern polar regions is more amplified than the others.

One thing I find interesting is that from all the data we have, the polar regions cooled below normal in the 1950s-1970s time range. Yet, we don't see GISS cooling more than Hadley in that time range. If GISS is extrapolating areas that should be warmer now, then why were they not cooler back when those said regions in question were colder? It seems the arctic cooled faster than the middle of the globe.

They consistently are warmer going back to that time range as well, and if the main crux of the argument is "extrapolation", it seems to lose its consistency before the most recent years. Nevermind the actual validity of the regions in question.

I guess the main question is (to avoid nitpicking little details and going on a tangent)....why should GISS be believed over other agencies when they are trying paint over void areas and the areas closest to those void areas are already in question to begin with comparing the different sources?

[ORH_wxman]

I chopped this out of the 2011 global temp thread...it doesn't really belong in there unless its specific to 2011 global temps.