GISS vs CRU/RSS/UAH

[tacoman25]

I think you are vastly underestimating the area that RSS has +4C anomalies.. +4 on RSS is anything orange.. the bright yellows are +5-6C. The oranges and yellows occupy all of France, Germany, the UK, Denmark, and most of Spain Sweden and Norway. GISS has these same areas at +1-2 and +2-4. RSS has them at >+4 and plenty of +6. This is a discrepancy of 2-4C over a large area.

I only see a few pixels that are bright yellow. Most of western Europe is in the orange, the same area that GISS has 2-4C.

We could argue back and forth about pixels and where one shade ends and the other begins...but the bottom line is that both sources show significant warmth over much of Europe. Again, you are ignoring the fact that RSS has a large area of colder anomalies than GISS to the west of Greenland. I don't think that matters. The anomaly patterns are very similar throughout the globe, with minor differences here and there. The one significant exception is in South America.

2) Even when we remove most of the extrapolation, and just look at HadCRUT 60-60 vs UAH over the last 30 years, the divergence remains. The surface is simply warming more, that is all there is to it.

Once again, the 30 year trend is not the divergence we are referring to.

3) Even when we remove the extrapolation from GISS, by using GISS 250km instead of 1250km, the divergence with HadCRUT remains from 2004-2010.

#2 and #3 empirically prove the extrapolations do NOT cause the divergence. Removing the extrapolations doesn't remove the divergence. It is the PHYSICAL data used, not the methodology. The methodology is sound.

You specifically claimed the extrapolations caused GISS to diverge from HadCRUT since the early to mid 2000s. I have performed a very straightforward test of this hypothesis which unequivocally proves it false. Removing the extrapolations from GISS since 2004 does not remove the divergence.

Perhaps you've forgetten, but in a discussion a week or so ago you agreed with me that GISS extrapolations were the reason they had diverged. I can bump the post to show you, if I can find it...I think in this thread.

Changing the extrapolation resolution doesn't prove that the extrapolations aren't part of the issue. Low data regions are still low data regions, and those are the main areas where GISS tends to be warmer. And to infill some of those low data regions, they have to extrapolate.

Just look at maps from previous months. You will find, as I have, that the GISS U.S. maps are almost always excellent matches to the RSS ones. Areas with lots of ground data/little extrapolation have a definite tendency to match the satellite data better than areas with little data. You are saying this shouldn't matter...but when GISS has had a tendency to extrapolate warmer (at least in recent years) than other sources in areas with little data, the overall result is warmer. Which is why GISS consistently comes in warmer than all the other sources year after year, and why their trend over the past decade is the warmest.

[tacoman25]

The extraps obviously ARE the reason for the divergence, but that doesn't mean the extraps are wrong. In fact, we know they're right, since when we infill HadCRUT with UAH we get the same result as GISS with only minor differences over the last 30 years.

[BethesdaWX]

Oh and BethesdaBoy...

in addition to my last post, need I remind you again that not all error is quantifiable?

+/-.05C in error is fairly large. It means that in reality the troposphere could be warming as much as .2C/decade and still fall within UAH's error bounds.

And as I said, this doesn't include all error. Some error is not quantifiable.

Even Spencer and Christy say the error may be significantly larger than .05C. You have just ignored this:

I QUOTE the ALMIGHTY SPENCER AND CHRISTY:

Error ranges of these estimates, if we do not apply information that indicates some data sets contain noticeable trend problems, are at least ±0.05°C decade^-1, which needs reduction to characterize forcing and response in the climate system accurately.

http://www.informawo...tent=a934129296

They also estimate the LT trend as .15C/decade 1979-2009 which is higher than UAH. So they're 1) saying that given all the available evidence, our best estimate is +.15C/decade 1979-2009, which is higher than UAH, and 2) even considering all the data, the error is still AT LEAST +/-.05C/decade.

Please refirain from the "BB", "BethesdaBoy", etc name calling.. just a suggestion. For all the times you've cried and screamed about people disrespecting you........you do the same thing consistantly. If you do it again, don't expect any respect from me. You Post Threads that have to be Deleted by Mods, calling people "stupid", or "wingnuts", Harass via PM..... frankly it gets old bro. You need to Stop...Now.

DENIALINVERMONT: We're only talking the post 2002 period here. Old Outdated Paper Bro ..., read Spencer's 2010 paper on the Issue of "verifyable and "non-verifyable" methods... Mr denialinvermont. Since 2002 (AQUA), the devation in the trend is no larger than +/- 0.05C, and is likely smaller. Your problem is using the Entire 1979-2011 dataset. THAT IS IRRELAVENT!.....We're Only Talking 2002 and later

[skierinvermont]

I only see a few pixels that are bright yellow. Most of western Europe is in the orange, the same area that GISS has 2-4C.

We could argue back and forth about pixels and where one shade ends and the other begins...but the bottom line is that both sources show significant warmth over much of Europe. Again, you are ignoring the fact that RSS has a large area of colder anomalies than GISS to the west of Greenland. I don't think that matters. The anomaly patterns are very similar throughout the globe, with minor differences here and there. The one significant exception is in South America.

Once again, the 30 year trend is not the divergence we are referring to.

Perhaps you've forgetten, but in a discussion a week or so ago you agreed with me that GISS extrapolations were the reason they had diverged. I can bump the post to show you, if I can find it...I think in this thread.

Changing the extrapolation resolution doesn't prove that the extrapolations aren't part of the issue. Low data regions are still low data regions, and those are the main areas where GISS tends to be warmer. And to infill some of those low data regions, they have to extrapolate.

Just look at maps from previous months. You will find, as I have, that the GISS U.S. maps are almost always excellent matches to the RSS ones. Areas with lots of ground data/little extrapolation have a definite tendency to match the satellite data better than areas with little data. You are saying this shouldn't matter...but when GISS has had a tendency to extrapolate warmer (at least in recent years) than other sources in areas with little data, the overall result is warmer. Which is why GISS consistently comes in warmer than all the other sources year after year, and why their trend over the past decade is the warmest.

GISS doesn't extrapolate warmer in areas with little data. The difference between GISS in the areas with and without a lot of data are the same. Once again showing that the difference has nothing to do with extrapolation. GISS runs warmer than UAH in both the highly and minimally extrapolated areas. If extrapolation were creating the warmth, that would not be the case.

Yes it does prove extrapolations are not the issue. If extrapolations were the issue, then removing them would remove the divergence. Instead, removing the extrapolations has almost no effect on the divergences with HadCRUT or with the satellites. It's very straightforward. There is no way around this. If extrapolation causes the divergences, then removing the extrapolations would remove the divergences. It doesn't. It has not effect. The physical data and other aspects of the methodology create the divergences.

[skierinvermont]

Also, I suppose I should have been clearer. GISS's extrapolations between 60S-60N have no effect. When we include or disclude them, GISS remains divergent from HadCRUT. And both HadCRUT and GISS are divergent from UAH.

Clearly GISS's extrapolation over the arctic is one of the major reasons it has warmed more over the last 10+ years than HadCRUT. That is all I was saying in that post. Those extrapolations are largely verified by other data.

[tacoman25]

GISS doesn't extrapolate warmer in areas with little data. The difference between GISS in the areas with and without a lot of data are the same. Once again showing that the difference has nothing to do with extrapolation. GISS runs warmer than UAH in both the highly and minimally extrapolated areas. If extrapolation were creating the warmth, that would not be the case.

Then why are the places where GISS shows up strangely warmer in the maps ALMOST ALWAYS IN LOW DATA AREAS? You still have not been able to explain this. In fact, you basically deny it, even though I could point to any number of map comparisons to prove it to you. At which point you will deny that the RSS maps should even match up with GISS...EVEN THOUGH THEY USUALLY DO OVER THE VAST MAJORITY OF THE GLOBE.

[tacoman25]

Also, I suppose I should have been clearer. GISS's extrapolations between 60S-60N have no effect. When we include or disclude them, GISS remains divergent from HadCRUT. And both HadCRUT and GISS are divergent from UAH.

Clearly GISS's extrapolation over the arctic is one of the major reasons it has warmed more over the last 10+ years than HadCRUT. That is all I was saying in that post. Those extrapolations are largely verified by other data.

Over the majority of the past decade, GISS has run warmer than all sources 60/60, and also in the Arctic. I realize your point about using smaller resolution with GISS results in very similar divergence...but this doesn't really matter because GISS uses small amounts of data to cover huge areas (in certain places) no matter what. So if a few stations in certain areas have a tendency to run warm, that data will be used over a large area.

You just don't ever see the glaring differences between RSS and GISS in the U.S. Why is that?

[skierinvermont]

GISS contains a UHI adjustment. It's right there in their methodology if you ever bothered to read it. All urban stations are calibrated to rural stations. The surfacestations project just proved again what we already knew... UHI is fully accounted for in GISS. Surfacestations found no difference between the most pristine stations and the rest of the stations.

[tacoman25]

GISS contains a UHI adjustment. It's right there in their methodology if you ever bothered to read it. All urban stations are calibrated to rural stations. The surfacestations project just proved again what we already knew... UHI is fully accounted for in GISS. Surfacestations found no difference between the most pristine stations and the rest of the stations.

Of course, that is just for the U.S. As I've pointed out in above posts, GISS tends to agree with satellites quite well in the U.S. It's other places with considerably less data like parts of Africa, South America, Asia, and the Arctic that there tends to be glaring differences.

[skierinvermont]

Of course, that is just for the U.S. As I've pointed out in above posts, GISS tends to agree with satellites quite well in the U.S. It's other places with considerably less data like parts of Africa, South America, Asia, and the Arctic that there tends to be glaring differences.

Not true.. GISS does it globally now. And other analyses just omit the urban stations and get the same result. See Parker 2010.

And as I've explained many many times before, GISS is often cooler than the satellites in regions as well because there is no expectation that they will correspond spatially. Moreover, it makes little sense to be cross-checking actual surface observations with satellite derived estimates which continue to contain a much larger degree of uncertainty.

[tacoman25]

Not true.. GISS does it globally now. And other analyses just omit the urban stations and get the same result. See Parker 2010.

And as I've explained many many times before, GISS is often cooler than the satellites in regions as well because there is no expectation that they will correspond spatially. Moreover, it makes little sense to be cross-checking actual surface observations with satellite derived estimates which continue to contain a much larger degree of uncertainty.

Nothing I said was untrue. I was pointing out that the Surfacestations.org analysis you were talking about was just for the U.S.

And you can make those claims about "no expectation to correspond spatially" all you want, but that notion goes out the window when the maps are actually compared. The vast majority of the globe usually matches up (including the U.S., which is almost always a great spatial match)...but places it doesn't, GISS tends to be warmer. And those tend to be regions with less surface data.

[skierinvermont]

Nothing I said was untrue. I was pointing out that the Surfacestations.org analysis you were talking about was just for the U.S.

And you can make those claims about "no expectation to correspond spatially" all you want, but that notion goes out the window when the maps are actually compared. The vast majority of the globe usually matches up (including the U.S., which is almost always a great spatial match)...but places it doesn't, GISS tends to be warmer. And those tend to be regions with less surface data.

Let's test this hypothesis. If this were true, then we would expect that the difference between UAH and GISS largely comes from less populated areas with less data, primarily the poles. A comparison of the mid-latitudes should show them fairly similar, while GISS will have warmed more in the polar regions.

Except this is not what we find. HadCRUT (and GISS) have warmed much more than UAH in the mid-latitudes 60-60. And when we replace the low-data areas with UAH... it makes nearly zero difference as to the global anomaly trend.

The differences between UAH and GISS (and HadCRUT) come from exactly where we have the MOST data (the mid-latitudes).

Don't you just love the scientific method?

[tacoman25]

Let's test this hypothesis. If this were true, then we would expect that the difference between UAH and GISS largely comes from less populated areas with less data, primarily the poles. A comparison of the mid-latitudes should show them fairly similar, while GISS will have warmed more in the polar regions.

Except this is not what we find. HadCRUT (and GISS) have warmed much more than UAH in the mid-latitudes 60-60. And when we replace the low-data areas with UAH... it makes nearly zero difference as to the global anomaly trend.

The differences between UAH and GISS (and HadCRUT) come from exactly where we have the MOST data (the mid-latitudes).

Don't you just love the scientific method?

I never said primarly the poles. In fact, I mentioned parts of South America, Africa, and Asia as areas that most frequently diverge when comparing the maps. But of course, over the past 8-10 years, as we have discussed already, GISS has run warmer than UAH in the Arctic.

Furthermore, you have failed to address my point: if there is no expectation of spatial matches between GISS and AMSU, why do the maps consistently match up very well over the vast majority of the globe?

[skierinvermont]

I never said primarly the poles. In fact, I mentioned parts of South America, Africa, and Asia as areas that most frequently diverge when comparing the maps.

Furthermore, you have failed to address my point: if there is no expectation of spatial matches between GISS and AMSU, why do the maps consistently match up very well over the vast majority of the globe?

Much of Asia is north of 60N.

And the portions of S. American and Africa which have sparse data are tiny. By replacing the poles I have removed a large portion (most) of the area with poor surface coverage. And yet it yields absolutely NO convergence between satellite and surface data.

The fact is... GISS ... AND... HadCRUT have been diverging from UAH for 30 years even in the places we have good data. Why? Probably because UAH is only a highly uncertain estimate of temperature trends. Various other satellite based estimates agree with GISS and HadCRUT.

[tacoman25]

Much of Asia is north of 60N.

And the portions of S. American and Africa which have sparse data are tiny. By replacing the poles I have removed a large portion (most) of the area with poor surface coverage. And yet it yields absolutely NO convergence between satellite and surface data.

The fact is... GISS ... AND... HadCRUT have been diverging from UAH for 30 years even in the places we have good data. Why? Probably because UAH is only a highly uncertain estimate of temperature trends. Various other satellite based estimates agree with GISS and HadCRUT.

The vast majority of Asia is south of 60N.

And how do you know the portions of South American and Africa that have sparse data are tiny? Why do those areas diverge much more often in the maps than the U.S., for instance?

Apparently you missed this question, so I'll pose it to you again: if there is no expectation of spatial matches between GISS and AMSU, why do the maps consistently match up very well over the vast majority of the globe?

[skierinvermont]

The vast majority of Asia is south of 60N.

And how do you know the portions of South American and Africa that have sparse data are tiny? Why do those areas diverge much more often in the maps than the U.S., for instance?

Apparently you missed this question, so I'll pose it to you again: if there is no expectation of spatial matches between GISS and AMSU, why do the maps consistently match up very well over the vast majority of the globe?

That is true on a monthly basis, not for the trendlines. For the long-term trend lines... the disagreement between GISS/HadCRUT and UAH is where we have good data.

On a monthly basis obviously you will find some spots extrapolated too cold and too warm in areas of sparse data. But in the long-run this yields to no bias.. and GISS's extrapolation of areas of sparse data are corroborated by UAH in the long-run.

[tacoman25]

That is true on a monthly basis, not for the trendlines. For the long-term trend lines... the disagreement between GISS/HadCRUT and UAH is where we have good data.

On a monthly basis obviously you will find some spots extrapolated too cold and too warm in areas of sparse data. But in the long-run this yields to no bias.. and GISS's extrapolation of areas of sparse data are corroborated by UAH in the long-run.

I can't compare maps from 20 years ago. But I can compare maps from recent years, which is when GISS has diverged warm. And it therefore makes perfect sense that in the areas of divergence, there are more where GISS is warmer. I have ZERO doubt of this, I looked at 4-5 years of monthly maps previously and it was obvious that GISS was warmer in areas of disagreement. And it only makes sense that they are. Again, you bring up longterm trends, and ignore the fact that GISS has diverged warm over the last 5-10 years.

You still have yet to offer any defense of your erroneous statement that there should be no expectation of spatial matching on the maps...and yet they consistently match most places.

[skierinvermont]

I can't compare maps from 20 years ago. But I can compare maps from recent years, which is when GISS has diverged warm. And it therefore makes perfect sense that in the areas of divergence, there are more where GISS is warmer. I have ZERO doubt of this, I looked at 4-5 years of monthly maps previously and it was obvious that GISS was warmer in areas of disagreement. And it only makes sense that they are. Again, you bring up longterm trends, and ignore the fact that GISS has diverged warm over the last 5-10 years.

You still have yet to offer any defense of your erroneous statement that there should be no expectation of spatial matching on the maps...and yet they consistently match most places.

There is no expectation in the sort term (IE monthly maps) that areas of poor data will match. Of course high data areas will match.

In the long-run, the expectation is that both high and low density data areas match.

Of course, that's not what we find. We find that they don't match in high or low density data areas. Probably because UAH is biased cold.

[tacoman25]

There is no expectation in the sort term (IE monthly maps) that areas of poor data will match. Of course high data areas will match.

In the long-run, the expectation is that both high and low density data areas match.

Of course, that's not what we find. We find that they don't match in high or low density data areas. Probably because UAH is biased cold.

The maps I'm referring to are RSS.

[skierinvermont]

The maps I'm referring to are RSS.

Well then in the long-run the two agree globally so there's nothing to discuss. Any monthly or yearly deviations are caused by short-term variability and greater measurement uncertainty for shorter periods.

[tacoman25]

Well then in the long-run the two agree globally so there's nothing to discuss. Any monthly or yearly deviations are caused by short-term variability and greater measurement uncertainty for shorter periods.

Well, you're in the camp of "all that matters are 30 year trends". Maybe that's the case, but I don't think it's fair to completely dismiss trends over the past 5-10 years, either. I guess time will tell. Arctic ice alarmists sure care a whole lot about the last 5-10 years.

[skierinvermont]

Well, you're in the camp of "all that matters are 30 year trends". Maybe that's the case, but I don't think it's fair to completely dismiss trends over the past 5-10 years, either. I guess time will tell. Arctic ice alarmists sure care a whole lot about the last 5-10 years.

Yes except with the arctic sea ice we can measure the ice fairly accurately for even just one year.

For global temperature for single years, the surface and especially satellite sources have substantial error bars attached, so that even when the annual values are substantially different, it may not be statistically significantly different.

So that's not exactly a fair analogy.