Dr. Judith Curry writes about the Cowtan and Way paper which (according to some pundits) purports to “bust” the temperature pause of the last 17 years by claiming we just didn’t pay enough attention to the Arctic and Antarctic where there is no data. They do this by infilling data where there is none, such as NASA GISS tries to do by infilling temperatures from stations far away with their smoothing algorithm.
GISS station data with 250km smoothing:
GISS station data with 1200km smoothing:
Breathless interpreters of Cowtan & Way claim that by doing the same with satellite data instead of tortured surface data, Voilà “the pause” disappears.
Cowtan & Way are trying to address this lack of surface station data in these regions by doing infill from satellite data. At first glance, this seems an admirable and reasonable goal, but one should always be wary of trying to create data where there is none, something we learned about in Steig et al’s discredited paper on the supposed Antarctic warming. Plus, as some WUWT readers know, there’s a reason that satellite temperature data coverage doesn’t fully cover the poles. See the information on the UAH data at the bottom of this post.
A video of their methodology follows.
WUWT readers will note the before and after HadCRUT imagery from Cowtan & Way below. Take special note of the Arctic.
A discussion on that Arctic temperature infilling addition at high latitude follows Dr. Curry’s analysis.
Dr Judith Curry writes:
=============================================================
Let’s take a look at the 3 methods they use to fill in missing data, primarily in Africa, Arctic, and Antarctic.
- 1. Kriging
- 2. UAH satellite analyses of surface air temperature
- 3. NCAR NCEP reanalysis
The state that most of the difference in their reconstructed global average comes from the Arctic, so I focus on the Arctic (which is where I have special expertise in any event).
First, Kriging. Kriging across land/ocean/sea ice boundaries makes no physical sense. While the paper cites Rigor et al. (2000) that shows ‘some’ correlation in winter between land and sea ice temps at up to 1000 km, I would expect no correlation in other seasons.
Second, UAH satellite analyses. Not useful at high latitudes in the presence of temperature inversions and not useful over sea ice (which has a very complex spatially varying microwave emission signature). Hopefully John Christy will chime in on this.
Third, re reanalyses in the Arctic. See Fig 1 from this paper, which gives you a sense of the magnitude of grid point errors for one point over an annual cycle. Some potential utility here, but reanalyses are not useful for trends owing to temporal inhomogeneities in the datasets that are assimilated.
So I don’t think Cowtan and Wray’s [sic] analysis adds anything to our understanding of the global surface temperature field and the ‘pause.’
The bottom line remains Ed Hawkins’ figure that compares climate model simulations for regions where the surface observations exist. This is the appropriate way to compare climate models to surface observations, and the outstanding issue is that the climate models and observations disagree.
Is there anything useful from Cowtan and Wray? Well, they raise the issue that we should try to figure out some way obtain the variations of surface temperature over the Arctic Ocean. This is an active topic of research.
===============================================================
More from the same post at Dr. Curry’s site here
What is really funny is how Dana Nuccitelli has done an about-face since the satellite data now supports his argument. In his Guardian 97% piece [cited in Dr. Curry’s article] he’s all for this method.
But, just two years ago he was trashing the UAH satellite data on SKS as “misinformation”.
[http://www.skepticalscience.com/uah-misrepresentation-anniversary-part1.html]
But Dana thinks UAH data is apparently OK today. What a plonker.
I will give Dr. Cowtan props though for realizing what the hypers don’t. He says this in the Guardian article:
“No difficult scientific problem is ever solved in a single paper. I don’t expect our paper to be the last word on this, but I hope we have advanced the discussion.
I give him props for having a sense of reality, something sorely lacking in climate science today.
Here’s why trying to use the satellite data to infill surface data at the poles is problematic:
Take a look at this latest image for 1000mb (near the surface) from the polar orbiting satellite NOAA-18, one of the satellites UAH now uses for temperature data:
Source: NOAA/NESDIS Office of Satellite Data Processing and Distribution (OSDPD)
Note how the data near the poles starts to get spotty with coverage? Note also how the plot doesn’t go to 90N or 90S?
NOAA doesn’t even try to plot data from there, for the reasons that Dr. Curry has given:
Second, UAH satellite analyses. Not useful at high latitudes in the presence of temperature inversions and not useful over sea ice (which has a very complex spatially varying microwave emission signature).
NOAA knows high latitude near-pole data will be noisy and not representative, so they don’t even try to display it. UAH is the same way. Between the look-angle problem and the noise generated by sea ice, their data analysis stops short of the pole. RSS does the same due to the same physical constraints of orbit and look angle.
As you can see, the polar orbit isn’t truly polar. Here are maps from UCAR that helps to visualize the problem:
As you can see, the orbit path never reaches 90N or 90S.
Source: http://www.rap.ucar.edu/~djohnson/satellite/coverage.html#polar
They write:
Note that the orbit is slightly tilted towards the northwest and does not actually go over the poles. While the red path follows the earth track of the satellite, the transparent overlay indicates the coverage area for the AVHRR imaging instrument carried by NOAA/POES satellites. This instrument scans a roughly 3000 km wide swath. The map projection used in this illustration, a cylindrical equidistant projection, becomes increasingly distorted near the poles, as can be seen by the seeming explosion of the viewing area as the satellite nears its northern and southern most orbital limits.
So, not only is the satellite coverage distorted at the poles due to the look angle, the look angle issue actually causes the satellite to image a wider swath of an area known to produce noisy and highly uncertain microwave data. Basically, the higher the latitude of the satellite imaging past about 60N/60S, the more uncertain the data gets.
It seems to me that all that Cowtan & Wray have done is swapped one type of highly uncertain data infilling with another. The claim that the addition of this highly uncertain data to HadCRUT4 seems to contradict ‘the pause’ most certainly isn’t proven yet, as even Dr. Cowtan admits to in his caveat.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
I’m guessing this is another of those papers that will be used by politicians as if it was written on tablets of stone and will be quoted as the reason why we must continue to de-industrialise the West.
JQuip,
My view is that sure, it would be lovely if we had surface readings everywhere. The fact is that we don’t. The question becomes, how best do we deal with that?
Steve and Nick seem to be arguing that ignoring this is worse than using this approximation – and I don’t think anyone is pretending this is anything but an approximation or estimation method. At least I hope they aren’t. But I don’t seem to see anything insane in what Steve or Nick are saying.
The question for me is, did they do a reasonable job when they worked it out in the paper? The fact that they published accessible methods code & data looks promising. My impression is that they worked pretty hard to validate the method as best they could. If I had to bet I’d bet this is solid.
~shrug~
Just my two cents.
Davidmhoffer,
I suspect the ‘pausebuster’ hype is just that; a bunch o’ hype without much substance. Over in the comments at Climate Etc. I think someone showed the statistics demonstrating this really doesn’t mean much of anything, not for the pause and not for the models either.
The sad thing here appears to be using data which is not understood.
The even sadder thing, is that a (supposedly) respected journal and its expert reviewers didn’t catch that.
How to be a “Climate Scientist”
Step 1: Take a dataset that superficially shows the effect you like, despite being thoroughly riddled with errors and biases, leading to error bands larger than that effect.
Step 2: Pretend the enormous error bands do not exist, and declare that the dataset is wonderful as is, and perfectly good for informing thousand trillion dollar decisions with high certainty.
Step 3: Selectively eliminate some of the errors in the dataset, with corrections that drive the results toward your pre-selected “finding”. Hysterically declare: It’s worse than we thought!
Step 4: Declare the newly improved dataset is wonderful as is, and perfectly good for informing thousand trillion dollar decisions. Decry the ‘uncorrected’ dataset as hopelessly biased and useless for any decision making, especially if the recent data in that version is trending in inconvenient directions. Belligerently assert that even though you are constantly ginning up changes that as much as double the severity of your estimates from one instance to the next, your ongoing confidence in the accuracy of your estimates remains beyond question.
Rinse, repeat.
David,
Yeah, the comment I was referring to about the lack of impact on the pause was:
http://judithcurry.com/2013/11/13/uncertainty-in-sst-measurements-and-data-sets/#comment-413386
The discussion of how this doesn’t save the models is over at Lucia’s.
Mosher, good points. Except your GISS claim that their trend at higher latitudes is equal to the trend at 70N is not supported by GISS data. The trend by latitude graph from the giss.website run from 1998-present (or any other recent time period) shows latitudes north of 70N having a significantly higher positive temperature trend.
http://s21.postimg.org/eej70vqnb/nmaps_zonal.gif
Source site: http://data.giss.nasa.gov/gistemp/maps/
“The new research suggests, however, that the addition of the ‘missing’ data indicates that the rate of warming since 1997 has been two and a half times greater than shown in the Met Office studies.”
But if the warming since 1997 has been essentially zero, then 2.5 X 0 = 0, doesn’t it?
Mark Bofill: “The fact is that we don’t. The question becomes, how best do we deal with that?”
Set up new instruments so we can get the data we don’t have. Ostensibly (Caveat Latitude) we have spacecraft now and shouldn’t be interpolating anything. But then no one seems to like the satellites since they don’t seem to be terribly reliable.
But as for interpolating? Sure, if you already know what the empty space *should* look like as a statistical normalcy. But to know that, you’ve had to have measured it, a bunch, previously. Otherwise it’s just made up numbers and you’d be just as close, or as far, to treat it like Dungeons and Dragons, and roll dice for every empty mile.
David Hoffer
The average does not reflect the temperature gradient. In fact, the average you are talking about reflects an observed cooling in the south pacific (17 year pause). If the interpolation was taken strictly from that region and applied over the arctic it would show decided cooling over the last 20 years.
of course, they did not apply normalized temperature changes from the tropical pacific to the arctic, they used regions approaching the northern limit of the UAH data and compared them with observed surface temperatures.
Since the polar regions show a much higher average temperature increase than the global average, the interpolated temperatures show higher rates of warming when applied.
Of course, this is a methodology that so many here decried when applied to the GISS data. As if not leaving the arctic warming out of the average is somehow manipulating the data so that it is no longer honest, or, that the only “honest” measurements don’t include the arctic where the majority of the observed warming has been occurring.
Whenever you calculate a space average from sampled data, there is an implied assumption that the samples are representative of data in between.
Has the assumption been tested? Are the sample intervals small enough to detect discontinuities? Because we know from experience that temps can vary greatly over just a few miles. So what is the data grid? What is the interpolation grid?
the arctic where the majority of the observed warming has been occurring.
Shouldn’t we see low sea ice extent as we did in 2012?
I still don’t know that I’m happy with comparing simulated-GMST to actual-GMST as a measure of ‘model error’ in the first place.
If the model underpredicts somewhere (say, Seattle) by 1C, and overpredicts somewhere else of vaguely equal size (say, near Houston) by 1C, the error should be determined -before- the averaging. If you do it after the averaging you get ‘zero error’, which is incorrect. This is like least-squares, where an error in either direction is always added to the cumulative error – and here performed in a spatial fashion.
Interpreting the deviations in this fashion should also highlight any specific regional difficulties or successes. (Hey, Model X doesn’t work on coasts, Model Y does non-tropical oceans very well, etc.)
jai mitchell;
Since the polar regions show a much higher average temperature increase than the global average, the interpolated temperatures show higher rates of warming when applied.
>>>>>>>>>>>>>>>>>>>>
You can’t interpolate rising temps from something that isn’t rising. Thanks for playing, but I will be patient and wait for someone who actually knows what they are talking about to answer.
I love the smell of desperation in the morning. So, basically, they’re saying all their previous handwaving was just so much bluster. Otherwise, they would reject this alternative, claiming they already had the reasons for the “pause” in hand.
So those 1070 hPa polar anticyclones never existed since it’s all warm in the Arctic… LOL
In an effort to be scientifically correct we seem to be hyper-focused on an area of the planet where nobody lives.
Moreover, IF these guys are right, then what should we think of the IPCC “scientists” who explained the pause? Once again, this shows how biased the HadCRUt/GISS stuff is and how o.5C can be generated by statistical artefacts. Fortunately pressure readings are not subjected to that kind of BS…
UAH already covers up to 85N.
The total Earth surface area above 85N is 0.0019 of the total area or 0.19%
It takes a lot of warming in a 0.0019 area to change the global temperature anomaly from UAH by 0.01C. 5C I guess.
It doesn’t matter, all the other temperature datasets are now so contaminated that we should just give up on them and quit quoting the numbers.
If the warmers are still in business by the year 2100, their warming metric is going to show +3.0C no matter what the actual temperatures have done. In the year 2100, everyone will know it was too cold in 1900 to grow corn in Iowa.
davidmhoffer says: “Let’s say these places where nothing lives, nothing grows, and everything is frozen, warm by several degrees. The result will be that these are now places where…. nothing lives, nothing grows, and everything is frozen.”
O, the humidity!
Steve from Rockwood: “In an effort to be scientifically correct we seem to be hyper-focused on an area of the planet where nobody lives.”
That’s exactly why we need to be hysteric about it. If the global warming thesis is true, they will be. In a few hundred years or so.
Jquip says:
Disclaimer – to make use of the distinction David advanced, I don’t know what I’m talking about. Haven’t read the paywalled paper, probably wouldn’t be competent to opine even if I had. Effectively I’m repeating hearsay from people I presume are more competent than myself; caveat emptor.
This said, my understanding is that the method figures out a relationship between the satellite data and the surface readings, and uses that relationship to deduce what the surface readings ought to be from the satellite readings.
RSS shows no warming for 17 years. And the major focus has been on the Arctic and not the Antarctic. RSS covers up to latitude 82.5 degrees.
With the circumference of Earth being about 40000 km, the distance from 82.5 to 90 would be 7.5/90 x 10000 = 830 km. So the area in the north NOT covered is pir^2 = 2.16 x 10^6 km2. Dividing this by the area of the earth, 5.1 x 10^8 km2, we get about 0.42% NOT covered by RSS for the portion relevant to our discussion. As a fraction, this is about 1/230 of the area of the earth. So for argument sake, let us assume that this 1/230 warmed up more than other parts over the last 17 years. How would that change the length of the pause according to RSS? Would it become 16 years and 10 months perhaps?
At best this would a stopgap measure. Since the satellite data shows cooling outside of the polar areas the most likely situation is that the poles are simply lagging the rest of the planet and will also start to cool. And, they will likely cool faster than the rest of the planet. So, while this might provide some short term relief to the pause, it will eventually lead to a faster rate of cooling.
But it is even worse than you think!
A little more accurately, for a radius of 6371 km, you can show that:
Between 81 degrees and the pole, there is a total of 3,139,000 sq km
Between 82 degrees and the pole, there is a total of 2,481,000 sq km
Between 83 degrees and the pole, there is a total of 1,900,000 sq km
Between 84 degrees and the pole, there is a total of 1,397,000 sq km
Between 85 degrees and the pole, there is only 970,000 sq km
So this entire earth-equally temperature INCREASE has to occur within a maximum limit of 2.5 million sq km’s ……
At the same time that the DMI ENTIRE summertime record from 1959 through 2013 for latitude 80 north shows 0.0 difference in temperature between days 150 and 250 of the year.
So you actually need a very hot Canadian and Siberian Arctic land temperature, a stable (not increasing!) temperature band at 80 north latitude, and THEN a sudden (but unmeasured!) temperature increase in a very small area across the pole!