Guest Post by Willis Eschenbach
My theory is that the BEST folks must have eaten at a Hollywood Chinese restaurant. You can tell because when you eat there, an hour later you find you’re hungry for stardom.
Now that the BEST folks have demanded and received their fifteen minutes of fame before their results have gone through peer review, now that they have succeeded in deceiving many people into thinking that Muller is a skeptic and that somehow BEST has ‘proven the skeptics wrong’, now that they’ve returned to the wilds of their natural scientific habitat far from the reach of National Geographic photographers and people asking real questions, I thought I might take a look at the data itself. Media whores are always predictable and boring, but data always contains surprises. It can be downloaded from the bottom of this page, but please note that they do not show the actual results on that page, they show smoothed results. Here’s their actual un-smoothed monthly data:
Figure 1. BEST global surface temperature estimates. Gray bars show what BEST says are the 95% confidence intervals (95%CI) for each datapoint.
I don’t know about you, but Figure 1 immediately made me think of the repeated claim by Michael Mann that the temperatures of the 1990s were the warmest in a thousand years.
WHAT I FIND IN THE BEST DATA
Uncertainty
I agree with William Briggs and Doug Keenan that “the uncertainty bands are too narrow”. Please read the two authors to see why.
I thought of Mann’s claim because, even with BEST’s narrow uncertainty figures, their results show we know very little about relative temperatures over the last two centuries. For example, we certainly cannot say that the current temperatures are greater than anything before about 1945. The uncertainty bands overlap, and so we simply don’t know if e.g. 2010 was warmer than 1910. Seems likely, to be sure … but we do not have the evidence to back that up.
And that, of course, means that Mann’s claims of ‘warmest in a mill-yun years’ or whatever he has ramped it up to by now are not sustainable. We can’t tell, using actual thermometer records, if we’re warmer than a mere century ago. How can a few trees and clamshells tell us more than dozens of thermometers?
Disagreement with satellite observations
The BEST folks say that there is no urban heat island (UHI) effect detectable in their analysis. Their actual claim is that “urban warming does not unduly bias estimates of recent global temperature change”. Here’s a comment from NASA, which indicates that, well, there might be a bias. Emphasis mine.
The compact city of Providence, R.I., for example, has surface temperatures that are about 12.2 °C (21.9 °F) warmer than the surrounding countryside, while similarly-sized but spread-out Buffalo, N.Y., produces a heat island of only about 7.2 °C (12.9 °F), according to satellite data. SOURCE
A 22°F (12°C) UHI warming in Providence, and BEST says no UHI effect … and that’s just a couple cities.
If there were no UHI, then (per the generally accepted theories) the atmosphere should be warming more than the ground. If there is UHI, on the other hand, the ground station records would have an upwards bias and might even indicate more warming than the atmosphere.
After a number of adjustments, the two satellite records, from RSS and UAH, are pretty similar. Figure 2 shows their records for global land-only lower tropospheric temperatures:
Figure 2. UAH and RSS satellite temperature records. Anomaly period 1979-1984 = 0.
Since they are so close, I have averaged them together in Figure 3 to avoid disputes. You can substitute either one if you wish. Figure three shows a three-year centered Gaussian average of the data. The final 1.5 years are truncated to avoid end effects.
Remember what we would expect to find if all of the ground records were correct. They’d all lie on or near the same line, and the satellite temperatures would be rising faster than the ground temperatures. Here are the actual results, showing BEST, satellite, GISS, CRUTEM, and GHCN land temperatures:
Figure 3. BEST, average satellite, and other estimates of the global land temperature over the satellite era. Anomaly period 1979-1984 = 0.
In Figure 3, we find the opposite of what we expected. The land temperatures are rising faster than the atmospheric temperatures, contrary to theory. In addition, the BEST data is the worst of the lot in this regard.
Disagreement with other ground-based records.
The disagreement between the four ground-based results also begs for an explanation. Note that the records diverge at the rate of about 0.2°C in thirty years, which is 0.7° per century. Since this is the approximate amount of the last century’s warming, this is by no means a trivial difference.
My conclusion? We still have not resolved the UHI issue, in any of the land datasets. I’m happy to discuss other alternative explanations for what we find in Figure 3. I just can’t think of too many. With the ground records, nobody has looked at the other guys’ analysis and algorithms harshly, aggressively, and critically. They’ve all taken their own paths, and they haven’t disputed much with each other. The satellite data algorithms, on the other hand, has been examined minutely by two very competitive groups, UAH and RSS, in a strongly adversarial scientific manner. As is common in science, the two groups have each found errors in the other’s work, and when corrected the two records agree quite well. It’s possible they’re both wrong, but that doesn’t seem likely. If the ground-based folks did that, we might get better agreement. But as with the climate models and modelers, they’re all far too well-mannered to critically examine each other’s work in any serious fashion. Because heck, if they did that to the other guy, he might return the favor and point out flaws in their work, don’t want that kind of ugliness to intrude on their genteel, collegiate relationship, can’t we just be friends and not look too deeply? …
w.
PS—I remind folks again that the hype about BEST showing skeptics are wrong is just that. Most folks knew already that the world has been generally warming for hundreds of years, and BEST’s results in that regard were no surprise. BEST showed nothing about whether humans are affecting the climate, nor could it have done so. There are still large unresolved issues in the land temperature record which BEST has not clarified or solved. The jury is out on the BEST results, and it is only in part because they haven’t even gone through peer review.
PPS—
Oh, yeah, one more thing. At the top of the BEST dataset there’s a note that says:
Estimated 1950-1980 absolute temperature: 7.11 +/- 0.50
Seven degrees C? The GISS folks don’t even give an average, they just say it’s globally about 14°C.
The HadCRUT data gives a global temperature about the same, 13.9°C, using a gridded absolute temperature dataset. Finally, the Kiehl/Trenberth global budget gives a black-body radiation value of 390 W/m2, which converts to 14.8°. So I figured that was kind of settled, that the earth’s average temperature (an elusive concept to be sure) was around fourteen or fifteen degrees C.
Now, without a single word of comment that I can find, BEST says it’s only 7.1 degrees … say what? Anyone have an explanation for that? I know that the BEST figure is just the land. But if the globe is at say 14° to make it easy, and the land is at 7°, that means that on average the ocean is at 17°.
And I’m just not buying that on a global average the ocean is ten degrees C, or 18 degrees F, warmer than the land. It sets off my bad number detector.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
steven mosher,
You were doing fine until #4:
Since the planet has warmed up ≈0.7°C over the past century and a half, and since the LIA was considerably cooler by all accounts, and also according to ice core evidence from both hemispheres, what I suspect happened is that somehow a decimal point got inserted in the wrong place.
Of course, that would make the LIA 3° colder than now, which may be a little bit too much. But it’s certainly closer to reality than 0.3°C, which looks like a Michael Mannian guesstimate.
@Willis
“In Figure 3, we find the opposite of what we expected. The land temperatures are rising faster than the atmospheric temperatures, contrary to theory.”
WTF is up with “we find”? Do you have a mouse in your pocket?
I expected no such thing. This is exactly what I’ve been harping about, dopey. GHG effect is primarily a land-based phenomenon. Got it? Write that down!
Perhaps if you hadn’t petulantly dismissed me when I explained the physics to you about why and how the ocean rejects downwelling FIR you wouldn’t be surprised by this finding.
Get a clue, Eschenbach. Buy one if you must.
[COMMENT: Drop the attitude, it makes you look like a rabid beaver on crack. I disagreed with your claims. That’s science, doesn’t make you right. Also, take your never-ending obsession with how DLR can’t heat the ocean elsewhere. This blog is for science, not science fiction. I’ll snip it next time, along with any wingeing about my policy. There’s lot’s of places to discuss your fantasy about DLR, but here is not one of them. -w.
PS—no, that’s not a mouse in my pocket, I’m just glad to see you …]
Robert says:
October 25, 2011 at 12:04 pm
This is supposed to be a surprise? You can’t see anything on their chart at all, it’s totally jumbled together. And yes … mine doesn’t look like that. Are we surprised?
Oh, can the attitude, for goodness sake. You are the jerk who proved above that you didn’t do your due diligence, you couldn’t tell GHCN from NOAA. You have no standing for your snark.
In addition, you offer up the explanation of Reto Ruedy (of NASA GISS) for why GISS differs from GHCN/NOAA, from Lucia’s site, emphasis mine:
Seems clear. If you give more weight in the average to the northern hemisphere, you will show more warming, just as he says. I’m not clear why you’d want to do that, but yes, Robert, you can change the averaging method and get a different result … but you are arguing that that is the only correct result, when clearly it is only a result of the choice of averaging method.
However, it gets worse. Ruedy’s claim is in total contradiction to another explanation for the exact same phenomenon, viz:
Two explanations for the same phenomenon. First one is from Reto Ruedy, of NASA/GISS, so it’s likely the right one, it’s likely the zonal averaging method that is the reason. Oh, except that the second explanation is from James Hansen of NASA/GISS, and he says the land mask is the reason … and you proudly announce that you know the real reason for the disparity, when even the GISS guys can’t agree on it? Hubris, anyone?
You may be right about the satellites. You may be right about GISS, and James Hansen may be wrong. You may be right about the STAR satellite analysis.
But since we don’t know if you are right, for today, for this discussion, I have used the databases as they are. Not the latitude-band averaged version described (but not archived) by Ruedy. Not the land area masked version described (but not archived) by Hansen. Not the “STAR analysis” you are on about, which doesn’t even exist yet. I used the records that I had as they are.
I used the existing ground-station based datasets. If you don’t like it, sorry, it’s what I have. I said if you sent me a citation to a land-masked version of the GISS dataset, I’d toss it in the mix. You have not yet done so.
Finally, note that according to Hansen, when the GISS dataset uses the Hadley missing data mask, GISS agrees with Hadley. If that is true, then why are you claiming that Hadley the outlier? If I apply the Hadley mask to the GISS data and post it up there, GISS doesn’t move towards BEST and NCDC. It becomes more like Hadley … which means they’re both outliers. Or that the BEST/GHCN data are outliers. The jury is still out, my friend, despite your attempt to claim that the science is settled.
And in either case, we still have about 0.7°C per century between either the high and low ground stations, or between the high ground stations and the satellites. That’s true even if you take GISS out of the equation entirely.
w.
Oh, I see now. #4 changed direction and was about UHI. Sorry, I got hung up on the LIA references. My bad [but actually my SWAG would be a higher UHI, based on personal experience and common sense].
Too busy to search at the moment, but didn’t Anthony do an experiment with a datalogger showing several degrees warming driving from the country into the city?
Oh, yeah, one more thing. At the top of the BEST dataset there’s a note that says:
Seven degrees C? The GISS folks don’t even give an average, they just say it’s globally about 14°C.
The HadCRUT data gives a global temperature about the same, 13.9°C, using a gridded absolute temperature dataset. Finally, the Kiehl/Trenberth global budget gives a black-body radiation value of 390 W/m2, which converts to 14.8°. So I figured that was kind of settled, that the earth’s average temperature (an elusive concept to be sure) was around fourteen or fifteen degrees C.
Now, without a single word of comment that I can find, BEST says it’s only 7.1 degrees … say what? Anyone have an explanation for that?
w.
yes smokey your math does not make sense.
1. you believe in a LIA
2. according to best the last 50 or so years of the LIA were 1.5c colder than today
3. You believe that some of this 1.5C is due to UHI.
4. How much
############
or look at willis shows. willis shows that the troposphere above the land is .3C warmer
( anomaly) than the land below it. And, the suggestion is that this difference perhaps is due to UHI
That is, if there was no UHI the curves would match better.. So, if you believe Willis, you have an upper bound on the UHI bias.. .3C
think carefully now..
####
if SST warms by .5C over decades and we know that land warms faster than water.. would
you expect the land to warm.. the same as SST? more than SST.. less than SST?
steven mosher says:
October 25, 2011 at 2:52 pm
While that’s true, advection also affects the ground level measurements … when the wind off the Canadian Plains hits the US, ground surface temperature measurements plummet …
That’s certainly valid as a first order cut.
I would agree with those four statements. I’m still curious why BEST finds a global average temperature of 7°·C, though.
w.
Rob Honeycutt says:
October 25, 2011 at 2:59 pm
Yes, it makes it easy to see how far the records have diverged over that time. You can mash them all together as you suggest, but that hides the decline, metaphorically …
w.
@Willis
“In Figure 3, we find the opposite of what we expected. The land temperatures are rising faster than the atmospheric temperatures, contrary to theory.”
Actually you screwed up that graph in Figure 3. The satellite temperature track is global, land + sea. BEST & GHCN are land-only. GISS & CRUTEM are land-ocean but stitch in satellite data for ocean where they can get it and use whatever data they can find or manufacture for ocean surface temp in pre-satellite period.
The caption and text indicate you believe those are all land-only records.
What it shows is that surface temperature is rising a lot more over land than it is over water.
“Too busy to search at the moment, but didn’t Anthony do an experiment with a datalogger showing several degrees warming driving from the country into the city?”
Yes smokey, but thats ONE CITY.. when you average all cities and rural together…
you get a curve, like the BEST curve.
Willis shows the temperature at the troposphere where there is no UHI.
That anomaly is at MOST .3C lower than the average of all the sites, both good and bad.
So, since the land from BEST is at MOST .3C higher ( anomaly) versus the troposphere..
right there you have a boundary for the maximum bias that UHI could have in the record.
Now, one city could be 1C.. another could be 0C..
one DAY could be 8C.. bt you dont get UHI every day. not when it rains or when the wind blows, or when its cloudy.. we are talking about the average bias.. BEST says its 1.5C warmer now than in the LIA.. and you think there is UHI bias in that 1.5C number.. how much? willis chart hints that the UPPER boundary is .3C bias.
Willis Eschenbach says:
October 25, 2011 at 9:17 am
David says:
October 25, 2011 at 4:47 am
I also would love to see an oceans only graph layered into the five data sets chart to see which data set follows the ocean air temp chart the most accurately.
The Berkeley dataset doesn’t cover the oceans, just the land.
w.
Thanks Willis, and yes I know, but it would still give a comparison and perhaps some insight into the proper UHI adjustment, and perhaps some insight into the unexplained increasing divergence here; David says: October 25, 2011 at 4:02 am, which no one else has commented on.
From Willis Eschenbach on October 25, 2011 at 3:32 pm:
I noticed that here. BEST is giving the land figure, their baseline for the anomalies. NCDC here says they use 8.5°C for their land baseline, 1901-2000 period. For the period the annual databases overlap (1880 to 2009 inclusive) the average absolute difference is 1.339 lower for BEST, NCDC is always higher.
Not much of an absolute difference, as long as you’re not worried about permafrost thawing or ice melting if the temp goes up even just another degree or two. If you are, BEST has yielded more breathing room before the “catastrophic” of CAGW kicks in. Surprising that doesn’t receive more press, no?
Thanks Willis. I would trust that smokey and others will see the implications of your chart.
UHI is bounded from above. we know that its not zero and your chart suggests .3C as an upper bound.
Next I direct people to look at BEST error bars.. 2 sigma is .19C
Then do some thinking
Next. Look at SST.. The land is.. what? maybe .2C warmer.. Does that help you bound the
BIAS..
remember. if we had 1000000 rural stations and only 1 urban with a huge 12C bias.. that one
cities effect would be smothered. When we look at all stations good and bad, rural and not..
the effect, the overall bias, is somewhere between 0 and .3C, lower if we consider SST and how close it is to the land.
Then look at the 2 sigma width..
and think about “power” in a statistical sense
steven “one trick pony” mosher says:
October 25, 2011 at 3:43 pm
“if SST warms by .5C over decades and we know that land warms faster than water.. would
you expect the land to warm.. the same as SST? more than SST.. less than SST?”
Land warms faster and also cools faster so average temperature no different. The big difference is that the ocean has an albedo close to zero and land has an albedo of about 15% so the SST should go up more because it absorbs more energy. In point of fact land surface temp is warming faster than ocean surface temp and the numbers do not add up indicating that someone’s got some mistakes in their physics. Unable to admit that but able to calculate what the ocean temperature should hypothetically be they’re scrambling around looking for the rise in temperature below the thermocline. They’re not finding it there either for the simple reason that GHGs have little effect over the ocean. That’s because GHGs operate by slowing down radiative cooling and while land surfaces cool primarily by radiation the ocean cools primarily by evaporation and GHG’s don’t do jack diddly squat to slow down evaporate cooling. Got it? Write the down.
Steve Mosher states, “So, since the land from BEST is at MOST .3C higher ( anomaly) versus the troposphere..right there you have a boundary for the maximum bias that UHI could have in the record.”
Your logic is logical, except I see close to .4 divergence in the chart Willis provided.
Willis, the previous post was made before I saw the discussion on the diverging data sets. I see the reasons are somewhat contradictory and do not really explain clearly why it is increasing through 2007.
Dave Springer says:
October 25, 2011 at 3:47 pm
@Willis
“In Figure 3, we find the opposite of what we expected. The land temperatures are rising faster than the atmospheric temperatures, contrary to theory.”
Actually you screwed up that graph in Figure 3. The satellite temperature track is global, land + sea. BEST & GHCN are land-only. GISS & CRUTEM are land-ocean but stitch in satellite data for ocean where they can get it and use whatever data they can find or manufacture for ocean surface temp in pre-satellite period.
The caption and text indicate you believe those are all land-only records.
What it shows is that surface temperature is rising a lot more over land than it is over water.
Thanks, Dave. Citations to both your claims and to the land-only you believe were used would be greatly appreciated. In the meantime, I find this:
Doesn’t look like a lot ocean in CRUTEM, certainly not enough to affect a global average much. Let me see what the CRUTEM site says:
SOURCE
I also find here:
I find nothing about satellite. The grid-boxes containing islands are used, but they are given the values from the “land air temperature anomalies”.
GISS, as I’ve been discussing with someone else, extends out land temps into the ocean, a crazy procedure in my book. But if you remove that, according to both Ruedy and Hansen, it moves it closer to CRUTEM. In any case, the dataset I’m using is identified as:
SOURCE: GISS
The word “only” in there should give you a clue that your claims about GISS using satellite data are simply not true.
SUMMARY: No indication that they “stitch in satellite data” in either one. No indication that they “use whatever data they can find or manufacture for ocean surface temp in pre-satellite period.” Neither one uses either satellite data or ocean data in any form.
You seem to be confusing HadCRUT for CRUTEM, and mistaking the GISS LOTI (land-ocean temperature index) dataset for the GISS land-only dataset that I actually used.
Thanks for playing, though, and next time, save the snark. In Spanish there’s a lovely line from a traditional song. It goes, “Dí les que no batan al aqua, que al cabo lo han de beber”. Loosely translated, that means “Tell them not to muddy the water … they might end up having to drink it.”
Regards,
w.
kadaka (KD Knoebel) says:
October 25, 2011 at 4:06 pm
Thanks, KD. I know that the BEST figure is just the land. But if the globe is at say 14° to make it easy, and the land is at 7°, that means that on average the ocean is at 17°.
And I’m just not buying that on a global average the ocean is ten degrees C, or 18 degrees F, warmer than the land. It sets off my bad number detector.
I need a land mask. Someone suggested the HADISST, which has N/A values where there’s land. I need to think some about this.
w.
From Willis Eschenbach on October 25, 2011 at 5:15 pm:
Well I’m not just writing for you and me. 😉
NCDC says 8.5°C land, sea surface 16.1°, combined 13.9° for baseline.
I find that believable. The sea retains the heat gained from the sun when the sun goes away, the land retains little heat and cools quickly (cloud heat retention effect at night being ignored) which leads to plummeting night temps. Using the commonly-accepted 70% sea figure, the combined NCDC figure with simple math is 13.8, close enough.
A mere 7.6°C land to sea difference, roughly corresponding to twice that as a day/night temperature swing on land of 15.2°C (27.4°F), daily max to min, doesn’t appear that bad. Arid tropical and extra-tropical areas see far more, islands surrounded by moderating sea and the polar regions will see far less but they make up a smaller bit of the total area. Here in Pennsylvania I expect 10 to 20°F swings, at the same latitude in the Midwest they can see far more.
Hard number crunching may reveal that difference as unrealistic, but for now at a glance I can accept it.
kadaka (KD Knoebel) says:
October 25, 2011 at 6:38 pm
Appreciated, KD. Here’s my problem. I’ve gone swimming in the sea in a lot of places around the planet, both at dawn and at dusk. At that time, the air temperature will be somewhere around average. But the sea temperature is never 18°F (10°C) above that.
That means the difference must be due to the cooler elevated land in the interior. Hmm. Could be, 10°C is about a kilometer in elevation … whoa, cool, here’s a scientific paper about that. Or wait, there a better way … OK, the median elevation of the stations in the BEST database is just over 200 metres. So the station elevation only explains about 2 degrees of the difference … where’s the other 8°??
Still unclear to me, I need funding for a major inquiry here …
w.
Willis, why use a Gaussian mean. Makes no sense to me. I would prefer an exponential mean where the previous data affects the current readings.
Willis, you (and others) use plots with ranges that are all different. It would make sense to use the same upper and lower bounds on plots so they can be compared visually. I suggest +- 2 degrees typically for anomalies. Also the graphs should be centered on zero. I know the software does it, but you shouldn’t.
Keep up the good work. I, for one, read all the stuff you post.
David.
“Your logic is logical, except I see close to .4 divergence in the chart Willis provided.”
That’s fine. Any number will do if you are willing to follow along and be logical.
So, here is what you have. You believe you have evidence that the 1.5C increase from the LIA
has UHI bias. and you believe its on the order of .4C
Which means that we are actually closer to LIA temps than we think. I don’t know what effect you think the Sun has on the climate, but your argument effectively diminishes it. However, Lets leave that aside.
You have evidence that .4C of the warming in the land record is due to UHI bias. Lets use that as your starting premise and do some math and more logic. ok? Lets’ ignore that this would make the
land cooler than SST, you could always adopt a different argument to handle that issue. Again,
Lets accept your .4C premise.
Now, BEST categorizes stations into rural and non rural. 16K rural and 23K not rural. do you think that percentage of rural is high? or is it low? Lets suppose you thought there were NO rural stations. none at all. what does .4C now represent? It represents the total UHI effect. Let me know if you get that. And lets swing it the other way, suppose you thought ALL the stations or lets say
99% are Rural.. What’s that mean? That means a few urban stations have Huge UHI.
so, understanding that.. what fraction of stations do you think are actually rural… remember if you pick a low number… then the UHI bias after averaging urban with rural approaches the entire
UHI effect.. and if you pick a high number of rural, then you are arguing that a few urban stations do all the biasing.
Lets say that BEST argues that 33% of the sites are rural.. is that low? or high? or just about right?
Springer.
I think I’ll pass on writing down anything you say. especially the part about GHGs over the ocean.
that was funny.
Dear Mr Willis Eschenbach
This is just to say Thank You !
You see, for people like myself, being able to understand al the statistic measurements in the reviews of BEST statement, is really difficult. But you make a hard job easy. What you say, even a “brutita” like me can read through, and understand.
“Gracias de nuevo, de su rendida admiradora española ”
María Maestre
Someone up above asked why BEST is rushing this to the media before it’s been reviewed…there’s an answer.
Someone else asked why they’re ignoring the (very real, you can see it yourself this winter) impact of Urban Heat Islands. There is an answer there, too.
The answer is, they’re not doing science, they’re doing POLITICS, and they’re counting on having their results measured with a thumb on the scales marked “good intentions”, because, as we all OUGHT to know, polluting is bad, so is inefficient use of power and resources. The whole Global Warming scare is really about scaring people into doing what the experts think they OUGHT to do, whether it’s Kyoto (giving the 3rd world a leg up by ‘sharing the wealth’), or funding otherwise economically nonviable work (Solyndra), the whole thing is like the preacher telling the people they need to be nicer to their neighbours and give a little to charity because god-is-coming-and-he-is-pissed.
In a way, a lot of the ‘solutions’ resemble the old Church’s sales of ‘indulgences’, there’s a reason for that-in the modern world, a PhD has replaced the Clerical Collar-so few people know, or understand even the actual SCIENTIFIC METHOD, that they will simply roll over on the word of an “expert’ with a doctorate.
‘taint science, ’tis Theology and Politics.