GISS vs CRU/RSS/UAH

[ORH_wxman]

I'm going to look more closely at the data links tomorrow, but just first looking at them, I'm not sure they are measuring the same thing. GISS seems to start at 64N for their "arctic temps" and UAH starts at 60N...so that leaves 4 degrees of latitude in GISS's main body of temperature data in each hemisphere. So GISS minus their own arctic/antarctic will be leaving an 8 degree band (4 on each side) of sfc area that is not taken out of their anomaly. And of course, those 8 degrees will be a lot more area than 8 degrees closer to the pole.

This might be part of the discrepancy.

[skierinvermont]

I'm going to look more closely at the data links tomorrow, but just first looking at them, I'm not sure they are measuring the same thing. GISS seems to start at 64N for their "arctic temps" and UAH starts at 60N...so that leaves 4 degrees of latitude in GISS's main body of temperature data in each hemisphere. So GISS minus their own arctic/antarctic will be leaving an 8 degree band (4 on each side) of sfc area that is not taken out of their anomaly. And of course, those 8 degrees will be a lot more area than 8 degrees closer to the pole.

This might be part of the discrepancy.

Where did you find that UAH arctic = 60N? I looked for it couldn't find it defined anywhere. I assumed it was 66N (still a 2 degree difference).

That shouldn't make a difference though. I assumed for the purposes of these graphs that each was 5% of the earth's surface area. It wasn't like I subtracted off 5% of the earth's surface area and then added back on 7% of the earth's surface area.

I weighted each as 5%.

Now, insofar as the trend for UAH 60N-90N is different than the UAH trend for 64N-90N (what I would use if I had it) that could create problems. But my guess is that the trends for UAH 60N-90N and UAH 64N-90N are not significantly different (for one thing they overlap 90%, and I have no reason to believe the 10% would be much different than the other 90% and even if it were different, it's only 10%). In other words, my analysis wouldn't change much at all if I used UAH 64N-90N anoms instead of 60N-90N anoms. Remember, I pretended as if it was 5% of the earth's surface area regardless.

If anything UAH 64N-90N has probably warmed ever so slightly faster than 60N-90N, which might bump things up at the end of the graph by a hundredth.

[ORH_wxman]

Its not difference in trend for UAH between 64-90 and 60-90 that matters much, its taking another 8 degree bite out of GISS's warm anomalies and replacing them with said UAH trend. GISS's "non-polar" anomalies are going to include up to 64N or 64S. If they did not include those 8 degrees and instead only went to 60N or 60S, I'd bet you see a nice little bump downward in their "non-polar" anomaly or what you called in your graph "GISS-arctic-antarctic" since those are likely some of the warmest areas on GISS's plot.

[tacoman25]

That shouldn't matter. I could start it in 1950. If UAH warmed before GISS, and not as much as GISS, as the two of you have suggested, then we would notice the replaced GISS running warmer than the original GISS early in the graph, and then cooler than GISS late in the graph.

Ok, well even assuming your calculations/graphs are completely correct, one would have to wonder why UAH matches up so wonderfully with GISS in the Arctic (and Antarctic?), but differs so much over the rest of the globe. Because even if UAH Arctic + Hadley rest of the globe = GISS as you claim, that doesn't explain the big GISS/satellite divergence globally.

[skierinvermont]

Its not difference in trend for UAH between 64-90 and 60-90 that matters much, its taking another 8 degree bite out of GISS's warm anomalies and replacing them with said UAH trend. GISS's "non-polar" anomalies are going to include up to 64N or 64S. If they did not include those 8 degrees and instead only went to 60N or 60S, I'd bet you see a nice little bump downward in their "non-polar" anomaly or what you called in your graph "GISS-arctic-antarctic" since those are likely some of the warmest areas on GISS's plot.

Oh - I didn't use GISS's calculated non-polar anomalies for 64S to 64N. I can see how a major problem would arise if I had though, as you said that would amount to leaving in an extra 8 degrees of GISS.

I took GISS globally and then subtracted off the arctic and antarctic mannually - and I weighted each as only 5% of the earth's surface area. That is about the equivalent of 70N-90N.

EDIT: OK this is a tangent, but come to think of it, if I had used GISS's automatically calculated 64S-64N anomalies, I actually would have biased the "GISS-arctic-antarctic+UAH arctic+ UAH antarctic" (purple line) too cold. I would have been subtracting off 52 degrees worth of GISS, but then I only added 40 degrees worth of UAH back on. Remember, I weighted everything as 5% which equates to an area equivalent to 70N-90N.

[skierinvermont]

Ok, well even assuming your calculations/graphs are completely correct, one would have to wonder why UAH matches up so wonderfully with GISS in the Arctic (and Antarctic?), but differs so much over the rest of the globe. Because even if UAH Arctic + Hadley rest of the globe = GISS as you claim, that doesn't explain the big GISS/satellite divergence globally.

Bingo.. that is the more interesting issue to me and what appears to be the much more active debate. Either climate models are wrong on lapse rates, satellites have systemic global errors, or HadCRUT and GISS have systemic global errors. Or some combination.

My personal belief is UAH is biased cold.. RSS slightly so... STAR and the newer radiosonde measurements are more in line with theory. RSS is still within the lower error bound.

[tacoman25]

Bingo.. that is the more interesting issue to me and what appears to be the much more active debate. Either climate models are wrong on lapse rates, satellites have systemic global errors, or HadCRUT and GISS have systemic global errors. Or some combination.

My personal belief is UAH is biased cold.. RSS slightly so... STAR and the newer radiosonde measurements are more in line with theory. RSS is still within the lower error bound.

But only for the non-Arctic/Antarctic? Not sure how that would work...

[skierinvermont]

But only for the non-Arctic/Antarctic? Not sure how that would work...

Ehh probably a few .01Cs/decade everywhere.. but the magnitude of the trend in the arctic is so large it doesn't make a difference to my graph.

Not saying UAH is definitely wrong, but that's the way things are leaning right now given UAH<RSS<STAR/RAOBCORE/RICH, the last 3 being the most modern and theoretically superior methods

[nzucker]

Bingo.. that is the more interesting issue to me and what appears to be the much more active debate. Either climate models are wrong on lapse rates, satellites have systemic global errors, or HadCRUT and GISS have systemic global errors. Or some combination.

My personal belief is UAH is biased cold.. RSS slightly so... STAR and the newer radiosonde measurements are more in line with theory. RSS is still within the lower error bound.

I think the reality is just that the troposphere has warmed less. Can you find any papers that address this problem?

I don't think you can assume the satellites are incorrect just because they don't match "the theory." According to the common AGW theory trumpeted by the IPCC and Hadley, we should be warming at least .2C/decade (and probably close to .5C/decade if you believe Hadley's latest comments about 4C warming by 2070). Yet the trend since 1998 is about .06C/decade for the satellites and .1C/decade for the surface, so clearly the commonly accepted theory is incorrect somehow. In reality, there is no such thing as a truly verified theory since each mainstream estimate of global warming has been too high, whether you look at Hansen's 1988 predictions, Hansen's revised 1990s predictions, IPCC 2007, Hadley's comments. No agency or model has proved its ability to follow the machinations of our climate, so we just have to accept that humans haven't reached that level yet. There's clearly a pattern here, and the plateau in warming contrary to modeled estimates is showing up in UAH/RSS, GISS/Hadley, and various estimates of OHC changes.

[tacoman25]

Ehh probably a few .01Cs/decade everywhere.. but the magnitude of the trend in the arctic is so large it doesn't make a difference to my graph.

Not saying UAH is definitely wrong, but that's the way things are leaning right now given UAH<RSS<STAR/RAOBCORE/RICH, the last 3 being the most modern and theoretically superior methods

But you are proporting that UAH matches GISS nearly perfectly in the Arctic...which would seem odd given the difference in trends globally. That doesn't have anything to do with trend magnitude. All sources agree that the Arctic has warmed significantly more than the rest of the globe the last couple decades.

And if UAH is flawed in general, it should also be flawed in the Arctic.

[skierinvermont]

But you are proporting that UAH matches GISS nearly perfectly in the Arctic...which would seem odd given the difference in trends globally. That doesn't have anything to do with trend magnitude. All sources agree that the Arctic has warmed significantly more than the rest of the globe the last couple decades.

And if UAH is flawed in general, it should also be flawed in the Arctic.

Yep... UAH's trend in the arctic is probably .03C/decade too low in my opinion just like everywhere else. Even if I factored this difference into my graph, the correction would be invisible. You wouldn't have the pixelation on your screen to see it.

And oddly enough... the differences do exhibit a spatial pattern.

Would it surprise you to know UAH's TLT trend from 20S to 20N is a mere .07C/decade, while RSS is .15C/decade? Even though the two are much closer globally.

(Above figure is for 1979-2009)

[nzucker]

But you are proporting that UAH matches GISS nearly perfectly in the Arctic...which would seem odd given the difference in trends globally. That doesn't have anything to do with trend magnitude. All sources agree that the Arctic has warmed significantly more than the rest of the globe the last couple decades.

And if UAH is flawed in general, it should also be flawed in the Arctic.

It's interesting that UAH agreed perfectly with the raw radiosondes in areas that were sampled by both. The HadAT radiosonde data showed a warming trend of .202C/decade, whereas UAH measurements of the areas where radiosondes existed (mostly Northern Hemisphere landmasses) showed a warming trend of .203C/decade. If anything, this probably means that UAH is doing an excellent job in its analysis of global temperatures; the radiosondes just happen to be located in the faster warming areas such as Siberia and Canada, and the Northern Hemisphere in general which has seen more warming. Here is the graph:

[skierinvermont]

It's interesting that UAH agreed perfectly with the raw radiosondes in areas that were sampled by both. The HadAT radiosonde data showed a warming trend of .202C/decade, whereas UAH measurements of the areas where radiosondes existed (mostly Northern Hemisphere landmasses) showed a warming trend of .203C/decade. If anything, this probably means that UAH is doing an excellent job in its analysis of global temperatures; the radiosondes just happen to be located in the faster warming areas such as Siberia and Canada, and the Northern Hemisphere in general which has seen more warming. Here is the graph:

A more recent version the exact same graph shows UAH to be biased low relative to radiosonde. See below.

For TMT (second graph) UAH is a mere .08C/decade vs .12-.16C/decade for RSS, RAOBCORE, RICH, STAR.

[tacoman25]

Yep... UAH's trend in the arctic is probably .03C/decade too low in my opinion just like everywhere else. Even if I factored this difference into my graph, the correction would be invisible. You wouldn't have the pixelation on your screen to see it.

And oddly enough... the differences do exhibit a spatial pattern.

Would it surprise you to know UAH's TLT trend from 20S to 20N is a mere .07C/decade, while RSS is .15C/decade? Even though the two are much closer globally.

(Above figure is for 1979-2009)

Aren't the tropics supposed to see the least amount of warming? The UAH number makes sense, then.

[skierinvermont]

Aren't the tropics supposed to see the least amount of warming? The UAH number makes sense, then.

Yes, but in the tropics the surface to tropospheric scaling ratio (SR) is the largest at 1.4X. The surface has warmed at .11-.12C/decade since 1979 (30S-30N) (roughly the theoretical rate for the tropics).

The theoretical value for the roposphere would then be about .15C/decade. Which is what RSS, RICH, RAOBCORE all show. but UAH shows much less.

Keep in mind the below graphs are 30S-30N... not 20S-20N like the figures I provided earlier. RSS, RAOBCORE, RICH, STAR all fit the theory for tropospheric scaling much better for the globe... for the tropics... for TLT... for TMT. And they all agree with each other better than UAH.

TLT

TMT

[tacoman25]

A more recent version the exact same graph shows UAH to be biased low relative to radiosonde. See below.

For TMT (second graph) UAH is a mere .08C/decade vs .12-.16C/decade for RSS, RAOBCORE, STAR.

For the LT, which what Zucker was referring to and what really matters in this discussion, it's only a .02/decade difference. According to previous statements from you, I fail to see how that is a significant difference, or how it proves UAH is "biased cold".

[skierinvermont]

For the LT, which what Zucker was referring to and what really matters in this discussion, it's only a .02/decade difference. According to previous statements from you, I fail to see how that is a significant difference, or how it proves UAH is "biased cold".

For TLT globally, UAH is not that different than other sources (although it is .02C/decade cooler which does carry some significance). But for TLT in the tropics, UAH is quite out of agreement with other sources (and with theory):

For TMT, it is in major disagreement in all latitudes.

[tacoman25]

Yes, but in the tropics the surface to tropospheric scaling ratio (SR) is the largest at 1.4X. The surface has warmed at .11-.12C/decade since 1979 (30S-30N) (roughly the theoretical rate for the tropics).

The theoretical value for the roposphere would then be about .15C/decade. Which is what RSS, RICH, RAOBCORE all show. but UAH shows much less.

Keep in mind the below graphs are 30S-30N... not 20S-20N like the figures I provided earlier. RSS, RAOBCORE, RICH, STAR all fit the theory for tropospheric scaling much better for the globe... for the tropics... for TLT... for TMT. And they all agree with each other better than UAH.

I don't know....you are again trying to convert surface data to satellite data (and vice versa), when at other times you act like this is not a good idea. And this time you are converting all the way to the mid-troposphere, I believe? And by the way, RSS is just as close to UAH as it is to RAOBCORE, so it doesn't really seem fair to isolate UAH. RSS is still easily lower than the theoretical .15C/decade you quote. So you have some sources not matching and some matching it.

[nzucker]

A more recent version the exact same graph shows UAH to be biased low relative to radiosonde. See below.

For TMT (second graph) UAH is a mere .08C/decade vs .12-.16C/decade for RSS, RAOBCORE, STAR.

It seems the HadAT radiosonde only diverged during the 2009-2010 El Niño; HadAT and UAH agreed in areas measured by both to the "T" until 2008. It's interesting because UAH usually responds very quickly to ENSO changes, and yet the radiosonde actually showed more warming. Could this be because of a very high concentration of radiosonde data in Canada, catching some of the extreme warm anomalies there last winter? I'm not very familiar with this data so tell me if this argument makes any sense, Skier.

Canada was absurdly warm last winter, so I don't know if a higher cluster of HadAT radiosondes there could have increased the average anomaly compared to UAH which is covering Asia and North America equally...it's clear from the January 2010 RSS anomaly map that Canada blowtorched while the PV remained over Siberia and kept temps in check during the strong Niño:

In any case, I doubt a difference of .02C/decade between UAH and HadAT, which has only developed in the last two years and may go back to normality, is at all statistically significant given that the error bars for UAH are around +/-.05C.

[tacoman25]

For TLT globally, UAH is not that different than other sources (although it is .02C/decade cooler which does carry some significance). But for TLT in the tropics, UAH is quite out of agreement with other sources (and with theory):

For TMT, it is in major disagreement in all latitudes.

There was only a .02/decade difference for the tropical TLT in the HadAT graph.

[nzucker]

This was the sampling pattern for RAOBCORE in 1987; I'd guess it has improved greatly since then, but this still affects the trend. Note the tropics are poorly sampled by radiosondes because they are mostly oceans...you can also see the gaps in Asia which may have lead to divergence during 09-10 of HadAT has a similar sampling pattern:

[skierinvermont]

I don't know....you are again trying to convert surface data to satellite data (and vice versa), when at other times you act like this is not a good idea. And this time you are converting all the way to the mid-troposphere, I believe? And by the way, RSS is just as close to UAH as it is to RAOBCORE, so it doesn't really seem fair to isolate UAH. RSS is still easily lower than the .15C/decade you quote. So you have just as many sources not matching it as you have matching it.

I'm not trying to convert it as a form of observation.. I'm just saying that if the surface does this X the theory is Y for TLT and TMT.

Not sure what you looking at...

For TLT tropics RAOBCORE and RSS agree .157 and .147.. UAH outlier at .115/decade

For TMT tropics, UAH again outlier, RAOBCORE, RSS, RICH in the middle/leaning towards STAR.

[skierinvermont]

It seems the HadAT radiosonde only diverged during the 2009-2010 El Niño; HadAT and UAH agreed in areas measured by both to the "T" until 2008. It's interesting because UAH usually responds very quickly to ENSO changes, and yet the radiosonde actually showed more warming. Could this be because of a very high concentration of radiosonde data in Canada, catching some of the extreme warm anomalies there last winter? I'm not very familiar with this data so tell me if this argument makes any sense, Skier.

Canada was absurdly warm last winter, so I don't know if a higher cluster of HadAT radiosondes there could have increased the average anomaly compared to UAH which is covering Asia and North America equally...it's clear from the January 2010 RSS anomaly map that Canada blowtorched while the PV remained over Siberia and kept temps in check during the strong Niño:

In any case, I doubt a difference of .02C/decade between UAH and HadAT, which has only developed in the last two years and may go back to normality, is at all statistically significant given that the error bars for UAH are around +/-.05C.

It is somewhat surprising to me as well that a divergence of .02C was able to develop in the 4 years since your graph was updated. I don't think you can attribute it to Canada specifically, since both sources are spatially averaged.

But I agree .02C/decade isn't that big... but when model theory for TLT globe is like .19C/decade +/-.05C/decade, and UAH is on the lower edge of that, then every .01C/decade becomes important.

[tacoman25]

A more recent version the exact same graph shows UAH to be biased low relative to radiosonde. See below.

For TMT (second graph) UAH is a mere .08C/decade vs .12-.16C/decade for RSS, RAOBCORE, RICH, STAR.

These are the graphs I was looking at Skiier, the ones you originally posted. Note that RSS is actually lower than UAH in the LT, and even in the MT, it's still just as close to UAH as STAR.

So your theory of UAH being the universal outlier doesn't always work.

[tacoman25]

It is somewhat surprising to me as well that a divergence of .02C was able to develop in the 4 years since your graph was updated. I don't think you can attribute it to Canada specifically, since both sources are spatially averaged.

But I agree .02C/decade isn't that big... but when model theory for TLT globe is like .19C/decade +/-.05C/decade, and UAH is on the lower edge of that, then every .01C/decade becomes important.

Dude...you have to apply that same logic to other things as well. Global temperature trends have been on the "lower edge" of model theory in general in recent years (only .1C rise decadal rise from the 1990s to 2000s when Pinatubo is factored in, for example), so again I think you are unfairly isolating UAH here.

[skierinvermont]

These are the graphs I was looking at Skiier, the ones you originally posted. Note that RSS is actually lower than UAH in the LT, and even in the MT, it's still just as close to UAH as STAR.

So your theory of UAH being the universal outlier doesn't always work.

I didn't say it was universally a major outlier.

I said that for the TLT globally it is lower than radiosonde... which is true.. RAOBCORE and RICH are both higher than it by .02C/decade. It's also a bit lower than RSS globally (all regions sampled .14 vs .16)

I said that for TLT in the tropics it is the outlier.. RSS, RAOBCORE, RICH all .147-.157 while UAH .115.

Then for TMT globally it is an outlier at .081C/decade ... RSS is next at .123C/decade, then RICH .128, RAOBCORE .137, STAR .166.

Then for TMT tropics it is again outlier at .067 vs .120 RSS, vs .127 RAOBCORE vs .159C STAR.

[nzucker]

It is somewhat surprising to me as well that a divergence of .02C was able to develop in the 4 years since your graph was updated. I don't think you can attribute it to Canada specifically, since both sources are spatially averaged.

But I agree .02C/decade isn't that big... but when model theory for TLT globe is like .19C/decade +/-.05C/decade, and UAH is on the lower edge of that, then every .01C/decade becomes important.

What do you think caused this divergence?

BTW radar looks juicy for this system...I'm at like -1C 850s and 37F at the surface, so it might be all rain but I have been staying up as I want to get that 2.5" to pass next year. Thursday and then the 3/29 storm are probably better candidates though for a real snowfall, but wouldn't mind getting a little here too. Amazing to think about snow and highs in the 30s this late, must be the next Ice Age that RAOBCORE isn't detecting.

Dude...you have to apply that same logic to other things as well. Global temperature trends have been on the "lower edge" of model theory in general in recent years (only .1C rise decadal rise from the 1990s to 2000s when Pinatubo is factored in, for example), so again I think you are unfairly isolating UAH here.

Given that GISS and UAH were similar before this decade with the overall warming trend, and that UAH shows exactly the same warming trend as the HadAT radiosonde in places where both had coverage during the 1979-2008 period, there's a good recent to believe UAH is actually very accurate. If you are arguing for the superior accuracy of radiosonde data, Skier, then you are actually arguing in favor of UAH, which is the exact opposite of what you intended to do. If UAH can match the radiosondes to 1/1000th of a degree C over a 30-year time frame in the sampled areas, then it's likely it's measuring the globe well too.

[tacoman25]

I didn't say it was universally a major outlier.

I said that for the TLT globally it is lower than radiosonde... which is true.. RAOBCORE and RICH are both higher than it by .02C/decade. It's also a bit lower than RSS globally (all regions sampled .14 vs .16)

I said that for TLT in the tropics it is the outlier.. RSS, RAOBCORE, RICH all .147-.157 while UAH .115.

Then for TMT globally it is an outlier at .081C/decade ... RSS is next at .123C/decade, then RICH .128, RAOBCORE .137, STAR .166.

Then for TMT tropics it is again outlier at .067 vs .120 RSS, vs .127 RAOBCORE vs .159C STAR.

So it's definitely not an outlier for radiosonde.

For TMT globally, it is no more an outlier than STAR. RSS and RICH are just as close to UAH as they are to STAR.

It certainly has the lowest trend overall. However, RSS has actually been moving towards UAH over the past decade. So if UAH is wrong, RSS is getting "wronger".

[skierinvermont]

What do you think caused this divergence?

BTW radar looks juicy for this system...I'm at like -1C 850s and 37F at the surface, so it might be all rain but I have been staying up as I want to get that 2.5" to pass next year. Thursday and then the 3/29 storm are probably better candidates though for a real snowfall, but wouldn't mind getting a little here too. Amazing to think about snow and highs in the 30s this late, must be the next Ice Age that RAOBCORE isn't detecting.

Given that GISS and UAH were similar before this decade with the overall warming trend, and that UAH shows exactly the same warming trend as the HadAT radiosonde in places where both had coverage during the 1979-2008 period, there's a good recent to believe UAH is actually very accurate. If you are arguing for the superior accuracy of radiosonde data, Skier, then you are actually arguing in favor of UAH, which is the exact opposite of what you intended to do. If UAH can match the radiosondes to 1/1000th of a degree C over a 30-year time frame in the sampled areas, then it's likely it's measuring the globe well too.

Can't really tell what causes the divergence after 2006 when your graph ends... it's not that big though.

RAOBCORE, RICH, STAR don't confirm UAH over RSS.

For TLT globe, RICH/RAOBCORE are pretty close to RSS/UAH, but slightly warmer. That is of note. UAH RSS are basically the same (.006C/decade difference).

For TLT tropics, RICH/RAOBCORE/RSS all substantially higher than UAH.

For TMT globe and TMT tropics, RICH/RAOBCORE/STAR/RSS all substantially higher than UAH.

[tacoman25]

Yep... UAH's trend in the arctic is probably .03C/decade too low in my opinion just like everywhere else. Even if I factored this difference into my graph, the correction would be invisible. You wouldn't have the pixelation on your screen to see it.

By the way, going back to this...if this is the case, wouldn't this kind of make your graph meaningless? I mean, if .03C/decade isn't going to effect the trendline because of the magnitude, then the "match" with GISS doesn't mean all that much. Right?