GISS vs CRU/RSS/UAH

[skierinvermont]

Actually, I believe it was suggested that comparisons to RSS (not UAH) maps indicated GISS was too warm in some areas, but you rejected that at the time because you said one couldn't accurately compare surface maps to satellite maps. So I don't see how you think it is ok to fill in some surface areas with UAH. Your base premise is that satellite temps cannot be directly applied to the surface.

Indeed that was suggested, and I have no problem with simply making that observation. Indeed some of the extrapolations will be too cold, others too warm. There's no problem with simply making that observation. What I objected to was the assertion that these 'errors' should be corrected on the basis of satellite data. The assertion was that anywhere where GISS differed from RSS, RSS should be substituted. Zucker went around the globe and said "ooo look at all these areas where GISS is different than RSS.. we need to fix all these 'errors' and replace them with RSS." That would essentially turn GISS into RSS. The basic premise is that anywhere GISS differs from RSS, GISS is wrong. THAT is what I objected to, on the basis that "GISS and RSS are not the same."

This is obviously substantially different from what I am suggesting here.

1) The use of satellite data would not be global wherever the surface differed from the satellites. It would be exclusively in the arctic where there is no surface data. It would form a rough approximation that was superior to the null hypothesis of no change (blank data).

2) I am not suggesting the 'correction' of GISS in any way. GISS would continue to exist as an independent measure of surface temperatures. I am suggesting an alternative and complementary measure of global temperatures (namely Had+UAH infilling).

[nzucker]

Indeed that was suggested, and I have no problem with simply making that observation. Indeed some of the extrapolations will be too cold, others too warm. There's no problem with simply making that observation. What I objected to was the assertion that these 'errors' should be corrected on the basis of satellite data. The assertion was that anywhere where GISS differed from RSS, RSS should be substituted. Zucker went around the globe and said "ooo look at all these areas where GISS is different than RSS.. we need to fix all these 'errors' and replace them with RSS." That would essentially turn GISS into RSS. The basic premise is that anywhere GISS differs from RSS, GISS is wrong. THAT is what I objected to, on the basis that "GISS and RSS are not the same."

I am not saying that GISS should be corrected in every area that disagrees with satellite analysis, but there should be a system for cross checking highly/frequently extrapolated areas so you don't always have the same mistakes. I shouldn't be able to point to a GISS map and say "there, that's an obvious huge divergence."

[skierinvermont]

I am not saying that GISS should be corrected in every area that disagrees with satellite analysis, but there should be a system for cross checking highly/frequently extrapolated areas so you don't always have the same mistakes. I shouldn't be able to point to a GISS map and say "there, that's an obvious huge divergence."

That is not what you were doing. You pointed out ALL of the areas around the world where RSS looked different and said those 'mistakes' needed to be fixed, not just those in the high arctic where there are no stations nearby.

Moreover, the other critical difference between what you are saying is that you were saying we should correct GISS. What I am saying is that one can form a complementary and corroborating measure of temperature (HadCRUT +UAH). GISS with all of its beautiful extrapolations would continue to exist as an independent measure.

Your viewpoint is founded in a failure to understand the mathematical implications of long distance extrapolations and a desire to "correct" them. My viewpoint is founded in an understanding of the mathematical validity of long distance extrapolations, and a desire to double check them with an alternative and complementary measure (HadCRUT + UAH infilling). As we see, and unsurprisingly if you understand the basic math behind spatial averaging, it makes very little difference whether we use long-distance extrapolations or UAH infilling. The end result is the same. HadCRUT is too low over the last 15 years because the areas it leaves blank have rapidly warmed.

[tacoman25]

Indeed that was suggested, and I have no problem with simply making that observation. Indeed some of the extrapolations will be too cold, others too warm. There's no problem with simply making that observation. What I objected to was the assertion that these 'errors' should be corrected on the basis of satellite data. The assertion was that anywhere where GISS differed from RSS, RSS should be substituted. Zucker went around the globe and said "ooo look at all these areas where GISS is different than RSS.. we need to fix all these 'errors' and replace them with RSS." That would essentially turn GISS into RSS. The basic premise is that anywhere GISS differs from RSS, GISS is wrong. THAT is what I objected to, on the basis that "GISS and RSS are not the same."

This is obviously substantially different from what I am suggesting here.

1) The use of satellite data would not be global wherever the surface differed from the satellites. It would be exclusively in the arctic where there is no surface data. It would form a rough approximation that was superior to the null hypothesis of no change (blank data).

2) I am not suggesting the 'correction' of GISS in any way. GISS would continue to exist as an independent measure of surface temperatures. I am suggesting an alternative and complementary measure of global temperatures (namely Had+UAH infilling).

1) The Arctic/sub-Arctic is not the only place globally where there is sparse station coverage, and GISS leans heavily on extrapolation. Regions of Asia, Africa, and South America in particular stand out, and these are regions where GISS is often warmer than other sources would indicate. This is why I think it's pretty clear the Arctic warming is not solely responsible for the GISS divergence over the past decade.

[ORH_wxman]

1) The Arctic/sub-Arctic is not the only place globally where there is sparse station coverage, and GISS leans heavily on extrapolation. Regions of Asia, Africa, and South America in particular stand out, and these are regions where GISS is often warmer than other sources would indicate. This is why I think it's pretty clear the Arctic warming is not solely responsible for the GISS divergence over the past decade.

I disagree...Hadley has actually been slightly warmer than GISS in the middle latitudes...its the higher latitudes where the breakdown occurs and occurs heavily.

The argument of replacing UAH in areas of blank Hadley temps still doesn't add up though during this period of significantly increasing divergence. Arctic and antarctic have basically canceled eachother out in that period. Also, Hadley has been colder anyway than GISS in those polar regions where they measure. The warmer measurements in the mid latitudes are what offset that and make GISS/Hadley similar in the non-extrapolated regions.

[tacoman25]

Actually it looks like both the arctic and antarctic regions were cooling late this past decade

Yup, if you look at both the Arctic and Antarctic (which for some reason skiier never seems to mention), there is nothing that indicates 2006-09 should have seen the greatest period of divergence.

[ORH_wxman]

I deleted the personal posts...keep that out of here.

[tacoman25]

I disagree...Hadley has actually been slightly warmer than GISS in the middle latitudes...its the higher latitudes where the breakdown occurs and occurs heavily.

The argument of replacing UAH in areas of blank Hadley temps still doesn't add up though during this period of significantly increasing divergence. Arctic and antarctic have basically canceled eachother out in that period. Also, Hadley has been colder anyway than GISS in those polar regions where they measure. The warmer measurements in the mid latitudes are what offset that and make GISS/Hadley similar in the non-extrapolated regions.

Hadley is only one other source, though. The satellites (which to me are more reliable sources in areas with little real surface data) often tend to show these other regions cooler than extrapolated surface data.

[ORH_wxman]

Hadley is only one other source, though. The satellites (which to me are more reliable sources in areas with little real surface data) often tend to show these other regions cooler than extrapolated surface data.

I agree with that..I was just talking the Hadley vs GISS direct comparison in the non-polar regions. GISS extrapolates in Africa quite a bit and Hadley still come out a bit warmer than GISS in the mid-latitudes between 50N and 50S. Its the 50-70 latitude bands where Hadley makes up that warmth by being cooler than GISS. Then of course GISS extrapolates beyond 70 degrees quite warm which creates all the divergence.

[skierinvermont]

I disagree...Hadley has actually been slightly warmer than GISS in the middle latitudes...its the higher latitudes where the breakdown occurs and occurs heavily.

The argument of replacing UAH in areas of blank Hadley temps still doesn't add up though during this period of significantly increasing divergence. Arctic and antarctic have basically canceled eachother out in that period. Also, Hadley has been colder anyway than GISS in those polar regions where they measure. The warmer measurements in the mid latitudes are what offset that and make GISS/Hadley similar in the non-extrapolated regions.

Agree with your first statement but disagree with the bolded.

If you substitute the slight Antarctic warming of around .1C/decade from 1995-present into HadCRUT the effect would be minimal. I am not even sure whether the effect would be positive or negative. You are substituting slight warming for blank data over a small area. The effect on the HadCRUT anomaly would be tiny.

On the other hand, in the arctic, you are substituting over 1C over warming since 1995 into cells that were left blank. This would have a pronounced positive effect on the HadCRUT anomaly over the last 15 years, and especially the last 5-8 years.

[tacoman25]

Agree with your first statement but disagree with the bolded.

If you substitute the slight Antarctic warming of around .1C/decade from 1995-present into HadCRUT the effect would be minimal. I am not even sure whether the effect would be positive or negative. You are substituting slight warming for blank data over a small area. The effect on the HadCRUT anomaly would be tiny.

On the other hand, in the arctic, you are substituting over 1C over warming since 1995 into cells that were left blank. This would have a pronounced positive effect on the HadCRUT anomaly over the last 15 years, and especially the last 5-8 years.

Why do you insist on using 1995 as a starting point? As has been pointed out to you several times, the main divergence started in the early 2000s. And the Antarctic has seen slight cooling since then, not warming.

[skierinvermont]

I agree with that..I was just talking the Hadley vs GISS direct comparison in the non-polar regions. GISS extrapolates in Africa quite a bit and Hadley still come out a bit warmer than GISS in the mid-latitudes between 50N and 50S. Its the 50-70 latitude bands where Hadley makes up that warmth by being cooler than GISS. Then of course GISS extrapolates beyond 70 degrees quite warm which creates all the divergence.

I'm not entirely sure that GISS is warmer than HadCRUT 50-70N. HadCRUT has a lot of blank cells in this area, especially 60-70N. That would significantly effect the graphic you posted from WUWT.

What we really need is global comparison of trends of the areas both cover. I am surprised that I can't find one.

[skierinvermont]

Why do you insist on using 1995 as a starting point? As has been pointed out to you several times, the main divergence started in the early 2000s. And the Antarctic has seen slight cooling since then, not warming.

Because there was also a big divergence in 1995 and 1996, and then again in 2001. Simple chance could have covered up the divergence 1997-2000, especially given arctic temperatures were still not as high as they were 2001-present.

Also, Antarctic temperatures jumped .5C 2001-2003.

[tacoman25]

I'm not entirely sure that GISS is warmer than HadCRUT 50-70N. HadCRUT has a lot of blank cells in this area, especially 60-70N. That would significantly effect the graphic you posted from WUWT.

What we really need is global comparison of trends of the areas both cover. I am surprised that I can't find one.

Why would blank cells cause it to be cooler or warmer? It's just less data, correct? Not a positive or negative in terms of temperature.

[tacoman25]

Because there was also a big divergence in 1995 and 1996, and then again in 2001. Simple chance could have covered up the divergence 1997-2000, especially given arctic temperatures were still not as high as they were 2001-present.

Also, Antarctic temperatures jumped .5C 2001-2003.

And they dropped about the same amount from 1992-1995. What kind of divergence did that create?

[skierinvermont]

Why would blank cells cause it to be cooler or warmer? It's just less data, correct? Not a positive or negative in terms of temperature.

The blank cells were assumed to be = no trend. ORH and I agreed on this the other day. The WUWT graphic he posted was a comparison of GISS latitude band trends to HadCRUT latitutde band trends but the HadCRUT trends factored in blank data as = no trend. For example, at the 80N latitude band GISS showed vastly more warming than HadCRUT. How do they even perform a comparison at 80N if HadCRUT is blank there? GISS's trend of .4C/decade at 80N was vastly warmer than HadCRUT's assumed trend for the blank data.

This problem is obviously most extreme when you are comparing GISS to entirely blank latitude banks on HadCRUT (IE 70N, 80N, 90N) but the problem still exists at say 60N where HadCRUT still has 1/3 of the cells blank.

(I should make a technical correction, blank cells on HadCRUT are assumed to have the background trend of the hemisphere).

[skierinvermont]

And they dropped about the same amount from 1992-1995. What kind of divergence did that create?

Probably none that was noticeable, given it's too small in magnitude to have much impact and the arctic was cooling which would cover up any effect.

[tacoman25]

Probably none that was noticeable, given it's too small in magnitude to have much impact and the arctic was cooling which would cover up any effect.

Wait...if it was too small in magnitude to have much impact, why did you bring up the fact that the Antarctic warmed the same amount in the early 2000s? Also, it appears to me that the Arctic had a pretty flat trend 1992-95. Actually, the Arctic saw a big jump in 1995 while the Antarctic remained cool.

[skierinvermont]

Wait...if it was too small in magnitude to have much impact, why did you bring up the fact that the Antarctic warmed the same amount in the early 2000s? Also, it appears to me that the Arctic had a pretty flat trend 1992-95.

It might have contributed to a small divergence. The point is one can't expect a perfect 1:1 relationship between Arctic and Antarctic temps and GISS VS HAD divergence. The general pattern of higher arctic temps and a higher divergence is observed. The Antarctic plays a relatively minor role given the magnitude of the trends is so much smaller.

[LakeEffectKing]

Why don't we all compromise, and someone go back to when the satellites started up, and just average the 4 anomolies through to the present (I haven't looked, but I bet it's been done already somewhere...but I haven't seen it) then we can turn our attentions to something other than bickering over what method is better or not or yada yada???....Unless everyone here likes to debate minutiae.......well then carry on....

[tacoman25]

The blank cells were assumed to be = no trend. ORH and I agreed on this the other day. The WUWT graphic he posted was a comparison of GISS latitude band trends to HadCRUT latitutde band trends but the HadCRUT trends factored in blank data as = no trend. For example, at the 80N latitude band GISS showed vastly more warming than HadCRUT. How do they even perform a comparison at 80N if HadCRUT is blank there? GISS's trend of .4C/decade at 80N was vastly warmer than HadCRUT's assumed trend for the blank data.

This problem is obviously most extreme when you are comparing GISS to entirely blank latitude banks on HadCRUT (IE 70N, 80N, 90N) but the problem still exists at say 60N where HadCRUT still has 1/3 of the cells blank.

(I should make a technical correction, blank cells on HadCRUT are assumed to have the background trend of the hemisphere).

Ok, thanks for the explanation.

Given all the evidence, I agree with you that Hadley is probably too cold over the last decade, and GISS is probably too warm. The correct trend is probably somewhere in between...which, believe it or not, would actually be pretty close to the UAH trend over the same period.

Too bad that data source is so horribly flawed.

[skierinvermont]

Why don't we all compromise, and someone go back to when the satellites started up, and just average the 4 anomolies through to the present (I haven't looked, but I bet it's been done already somewhere...but I haven't seen it) then we can turn our attentions to something other than bickering over what method is better or not or yada yada???....Unless everyone here likes to debate minutiae.......well then carry on....

Well given the large magnitude of warming in the arctic the last 15 years, and the even larger magnitude of projected arctic warming, using an index which leaves 90% of the area north of 60N blank seems sort of obsolete to me when we can approximate what is going on using extrapolations from stations at 60-70N, or by using satellite data as a reasonable approximation. If the arctic continues to warm as projected, the problem is only going to grow.

[tacoman25]

It might have contributed to a small divergence. The point is one can't expect a perfect 1:1 relationship between Arctic and Antarctic temps and GISS VS HAD divergence. The general pattern of higher arctic temps and a higher divergence is observed. The Antarctic plays a relatively minor role given the magnitude of the trends is so much smaller.

Except that the Arctic does NOT have a much greater magnitude trend since the early 2000s. Since 2003, the Arctic has seen slight warming, and the Antarctic slight cooling. Yet we saw the greatest divergence between GISS and other sources 2006-09.

[ORH_wxman]

Except that the Arctic does NOT have a much greater magnitude trend since the early 2000s. Since 2003, the Arctic has seen slight warming, and the Antarctic slight cooling. Yet we saw the greatest divergence between GISS and other sources 2006-09.

Yes that is my biggest beef with the divergence in the last 8-9 years or so. Post-2002 the two poles have virtually canceled eachother on UAH yet the divergence got a lot worse. GISS is likely over exuberant in their warming of areas with 0 data.

[skierinvermont]

Ok, thanks for the explanation.

Given all the evidence, I agree with you that Hadley is probably too cold over the last decade, and GISS is probably too warm. The correct trend is probably somewhere in between...which, believe it or not, would actually be pretty close to the UAH trend over the same period.

Too bad that data source is so horribly flawed.

I probably would agree as well that GISS is too warm the last 15 years.. the arctic warming is somewhat more than UAH or other estimates. But it's much better than the null hypothesis of zero warming on Had.. which is why I'd go with 25% Had 75% GISS.

I think there are numerous more interesting differences between GISS and HadCRUT than the divergence over the last 15 years (which to me is obviously mostly attributable to HadCRUT missing the rapid arctic warming).

For example, GISS uses the HadISST2+OISSTv2 SST reconstruction which shows nearly .1C less warming than the HadSST2 reconstruction used by HadCRUTv3. So HadCRUT tends to show more warming over the oceans.

But GISS, on the other hand, uses nearby land data and extrapolates that over the oceans when possible. HadCRUT doesn't do that.

It seems the two effects largely canceled each other out 1880-present, although HadCRUT still shows slightly more warming due to the warmer SST reconstruction.

So GISS and HadCRUT arrive at a similar result in significantly different ways 1880-present. I think that's more interesting than the recent divergence which I think is fairly obviously primarily due to HadCRUT excluding the arctic which has warmed rapidly. I'd like to be able to agree that the primary reason that GISS and HadCRUT have diverged is HadCRUT missing the rapid warming of the arctic this decade. And that GISS represents this warming fairly well. When discussing surface trends the last 10-15 years it is probably best to use a GISS 50 or 75% compromise, or HadCRUT +UAH.

[tacoman25]

Yes that is my biggest beef with the divergence in the last 8-9 years or so. Post-2002 the two poles have virtually canceled eachother on UAH yet the divergence got a lot worse. GISS is likely over exuberant in their warming of areas with 0 data.

A couple other points to consider in regards to how much of a role Arctic warming plays in global temperature trends, and how it relates to larger-scale AGW.

1) The area north of the Arctic circle (66 degrees north, roughly the same area lacking HadCRU data) comprises about 6% of the earth's surface, obviously a small percentage. And it's mostly water. So even if the Arctic is seeing significantly greater warming than the rest of the earth, it shouldn't be enough to significantly effect the greater global trend, and certainly the NH land temp trends...which have diverged quite a bit with GISS in recent years.

2) Hadley has records going back hundreds of years. Why should we now alter the area covered in their records? They have seen roughly the same amount of warming (AGW percentage?) over the past 100 years as GISS (or 30 years), so why would the Arctic only now matter so much?

3) The Arctic has seen huge occillations in temperature historically. It could definitely be a mistake to attribute too much of the recent Arctic warming to AGW.

[skierinvermont]

Yes that is my biggest beef with the divergence in the last 8-9 years or so. Post-2002 the two poles have virtually canceled eachother on UAH yet the divergence got a lot worse. GISS is likely over exuberant in their warming of areas with 0 data.

Indeed since 2002/2003 the trend at the two poles nearly cancels (slight net warming). As one might expect, the divergend hasn't really grown since 2002-2003 which had very large divergences.

[tacoman25]

Indeed since 2002/2003 the trend at the two poles nearly cancels (slight net warming). As one might expect, the divergend hasn't really grown since 2002-2003 which had very large divergences.

Not true. You are singling out two years, but when you look at a period twice as long, 2006-09 has the greatest divergence. And I'm talking about GISS vs. HadCRU/UAH/RSS...all the other major sources.

[skierinvermont]

A couple other points to consider in regards to how much of a role Arctic warming plays in global temperature trends, and how it relates to larger-scale AGW.

1) The area north of the Arctic circle (66 degrees north, roughly the same area lacking HadCRU data) comprises about 6% of the earth's surface, obviously a small percentage. And it's mostly water. So even if the Arctic is seeing significantly greater warming than the rest of the earth, it shouldn't be enough to significantly effect the greater global trend, and certainly the NH land temp trends...which have diverged quite a bit with GISS in recent years.

2) Hadley has records going back hundreds of years. Why should we now alter the area covered in their records? They have seen roughly the same amount of warming (AGW percentage?) over the past 100 years as GISS (or 30 years), so why would the Arctic only now matter so much?

3) The Arctic has seen huge occillations in temperature historically. It could definitely be a mistake to attribute too much of the recent Arctic warming to AGW.

1) The entire area of the earth missed by HadCRUT is quite small. And yet we KNOW for a 100% FACT that this "missing area" is the cause of the divergence. Because if you use a HadCRUT mask on GISS, they agree completely.

2) For one thing the magnitude of the arctic temperature changes the last 15 years is larger than anything before. For another, there are some big differences in SST reconstructions used by HadCRUT and GISS earlier in the century. If they used the same SST data, GISS would show more warming than HadCRUT, primarily because GISS includes the faster warming in the arctic, and HadCRUT does not.

[skierinvermont]

Not true. You are singling out two years, but when you look at a period twice as long, 2006-09 has the greatest divergence. And I'm talking about GISS vs. HadCRU/UAH/RSS...all the other major sources.

2006-2009 is 4 years. 2002 and 2003 had bigger divergences than half of them (2006 and 2008 had smaller divergences).

If I drew a trend line from 2002-present, the divergence would have only grown slightly.

Again, I have no expectation of a perfect correlation between arctic temperatures and the divergence. I don't know how you can continue to deny it's the arctic. If you remove the arctic, GISS nearly agrees with HadCRUT.