Guess what? The much anticipated Hadley monthly surface temperature anomalies are now available. I always use the NH+SH simple average.
Guess what else? According to this metric, the global surface temperature anomaly September 2009 cooled relative to August 2009 dropping from0.548C to 0.457C. In contrast, GISSTemp, NOAA/NCDC, UAH and RSS all reported distinctly warmer anomalies in September relative to August. This divergence is a pit surprising– though I’d have to plough through numbers to see if this sort of mismatch is unprecedented in the record.
One of the interesting happenings this month was Hadley’s decision to delay processing because they considered the some data they received to be obviously wrong. We don’t have details on precisely what was wrong about it, but I noticed large blanked out areas on their map:
The blanked out areas do seem to be surrounded by warm regions. Maybe the computed value for September’s monthly average will rise when that region reports data Hadley trusts. In the meantime, Hadley’s September temperature is low relative to the other metrics.
Since we anticipate October temperature will be reported soon, and I suspect some revisions for September, I’ll just show the trends based on reported temperatures since both 2000 and 2001, and also compare them anomalies to the multi-model mean anomalies from the AR4 climate models driven by the A1B SRES.
As you can see, EL Nino has caused temperatures to rise; the anomalies for individual months values are currently approaching the mean value projected by the models. As El Nino warms further, the observations for individual months may finally catch and surpass the models, as the do from time to time. However, it’s going to take sustained warming for the trends since either 2001 or 2000 to catch up with the projections. Will it happen? We’ll wait and see.
Lucia, I don’t think this is anything out of the ordinary to have so may data holes. Look at GISS for September:
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2009&month_last=09&sat=4&sst=0&type=anoms&mean_gen=09&year1=2009&year2=2009&base1=1951&base2=1980&radius=250&pol=reg
The trend of missing stations in GHCN continues. It appears that Hadley actually has more stations than GISS. Maybe the delay was to allow more trickle in of late reporters.
Anthony–
I think Hadley usually waits for more data than GISS. Their monthly estimates don’t revise as much the following month. (I know this is partly due to GISS’s interesting method. But I suspect it’s partly due to data updates.)
How do we lowly common simple folk determine this is the ‘correct’ data and not still the ‘obviously incorrect’ data?
And for those who are sensitive to me not ‘knowing my place’ or are upset by me violating sacred rights by asking ‘obviously incorrect’ questions… I’m just askin’.
If anyone would like to answer my question, I would be appreciative.
Andrew
Andrew_KY
Ok, I’ll offer a response.
I don’ t think it matters for a particular month. What will that tell us about climate change (or not)? Nor, indeed, do I think it matters so much if there is some uncertainty over the whole record (and, of course, GISS etc. express margins of error in their analyses), it is a matter of degree (that is, obviously, if the running error is large then we might question the usefulness of the data).
What would matter, of course, is ‘error’ persistently in one direction, and what would matter even more would be if such persistence were being deliberately engineered. It seems that some people think that either one or both of those things is true. Personally, I think such people are likely to be deluded by their own prejudices. If not, I’d like to see their evidence that any of the temperature records shows increasing divergence in comparison to the full set. If even that can’t be shown then I have to think that such suspicious people imagine that these scientists are manipulating the data but incompetently failing to make any difference by their manipulations ;-).
I think, btw, that we could do with better systems for measuring global temperatures – satellites designed in the first place to measure temperature would be good.
Simon,
Thank you for responding, but you didn’t answer my question. The question isn’t ‘does it matter to person X that he or she can determine what data is correct and what data is obviously incorrect?’
Suppose it does matter to person X. HOW do they determine the data they are looking at is correct or incorrect, if they wanted to make that determination?
Andrew
I’ve no idea, Andrew. I imagine that in any given month some of the data isn’t correct. Errors happen. I don’t suppose anyone (including GISS or Hadley) could get back to every source to be totally sure that there is no error in the data, so I wouldn’t fancy my chances of finding out. The best we can expect, I think, is that they should screen for ‘obviously incorrect’ data. It’s a constant process of checking out the data (for the satellite records too, of course), and there will remain some margin of error (from memory, GISS estimates something like 1%?).
Andrew_KY
Why don’t you attend this conference and attend the presentation on
“Quality Control of Meteorological Data for the Chemical Stockpile Emergency Preparedness Program
James C. Liljegren, ANL, Lemont, IL; and S. Tschopp, K. Rogers, F. Wasmer, L. Liljegren, and M. Myirski”
The section written by by Jim discusses algorithms for automating the process of detecting some data that is “obviously” wrong. The fact is, that it is sometimes possible to look at data and detect when it is “obviously” wrong. Then, you can send someone out to inspect and repair equipment. I’m not going to explain what we do to you because a) the specific details are not specifically relevant to a different data set and and b) I already gave an example related to last Octobers GISS Temp data.
Specific tests for detecting obviously wrong data do have to be tailored to the properties of the data you collect. If you want to know the details of what Hadley does you will have to ask them not us. But the fact is that if data are “obviously” wrong, those who actually look at the parts of the data that are obviously wrong can generally tell it is wrong. Anyone who has any experience doing experiments knows this.
Lucia,
You are cracking me up! This reminds me of a Seinfeld skit. (Everything does, BTW)
Jerry: So, we’re going to make the post office pay for my new stereo, now?
Kramer: It’s a write-off for them.
Jerry: How is it a write-off?
Kramer: They just write it off.
Jerry: Write it off what?
Kramer: Jerry all these big companies they write off everything.
Jerry: You don’t even know what a write-off is.
Kramer: Do you?
Jerry: No, I don’t.
Kramer: But they do – and they are the ones writing it off.
ob·vi·ous (bv-s)
adj.
1. Easily perceived or understood; quite apparent. See Synonyms at apparent.
2. Easily seen through because of a lack of subtlety; transparent
As has been aptly demostrated here, you, I, and Simon do not know what is ‘obviously incorrect or correct’ about what we are looking at. It’s not ‘obvious’ to us. We don’t know if it’s correct or incorrect or what it means.
Like Kramer in the skit you just surrender your own critical thinking to ‘them.’
Andrew_KY–
Huh? Of course people who haven’t looked at the data can’t see obvious flaws. The flaws were obvious to Hadley. That’s why Hadley waited to process it.
If you, a lowly person, wished to see the obvious flaws for yourself, it is rather obvious that you would need to request the data from whoever has it. This is why I said
The obvious fact that if you want more details you will need to ask Hadley keeps eluding you despite my repeating the notion.
If you just want to know whether errors can be obvious: They can. They often are.
Andrew_Ky
there’s loads of stuff that we don’t really know (immediately) whether it’s correct or not. Bank statements? Electricity bills? Garage charges for replacement parts? Tax assessments? The stated temperature in your home town? Some things are easier to check out than others. I don’t bother myself unless I have reason to be suspicious. Do you have reason to be suspicious of the Hadley data, presenting, as it does, September as a cooler anomaly against August, in contrast to every other temperature record? Do you suspect it ‘should have been’ warmer? Yeah, good point – go check it out!
There is no such thing as correct data. Only useful data and not so useful data.
Some useful questions in assessing data usefullness:
What processes does the data collection agency have in place to ensure accurate measurements?
What processes do they have to spot and correct errors?
How consistent are they in applying these processes?
If something is supsicious, what is the worst possible case? For instance as many as 20% of data points may be up to 10 measurement units out, so the maximum error is 2 units.
Does anyone else measure the same thing? If the measurment is exactly the same, or very different that is suspicious. If it is similar, then that is good. If it is different, what are the reasons for the difference?
Is the measurement consistent with other data? If temperatures are going up in a location where measurements of wind direction or cloudiness suggest the temperature should be going down, or where ice is increasing, or wildlife showing behaviours typical of cold weather then be supsicious.
Simon–
Hadley briefly posted a paragraph explaining that they delayed processing the September data because the data the initially received was “obviously wrong”. They also told us they were trying to locate reliable data and would publish when the good data appeared.
They have now published. So presumably, they believe they obtained data that is suffiently good to process.
For some reason, Hadley’s decision to not process data Hadley thought was wrong has put Andrew_KY’s knickers in twist. This is causing him to try to advance some sort of argument by posting what appear to be pointless rhetorical questions. I’m not entirely sure what:
a) He actually wishes to know or
b) What point he is trying to make.
But, in my opinion:
a) If Hadley thinks they data the receive from the stations is wrong, they shouldn’t process it. This appears to be what happened in October. So, they appear to have acted correct.
b) It is sometimes possible for data to be “obviously wrong”. So, if the guys at Hadley who saw the data thought it was “obviously wrong”, I’m perfectly satisfied that they acted correctly. (I even think they acted correctly if it turns out they were mistaken and, on further investigation, discovered that the data were right. One should never waste time processing data one thinks is “obviously wrong”.
c) In the end, the delay was slight.
So, all in all, it looks like Hadley’s behavior with respect to processing September 2009 data seems to have been appropriate.
“There is no such thing as correct data.”
Tell that to Lucia and Hadley.
Andrew
Lemme simplify into a step-by-step parable if this is too difficult for some: 😉
1) Andrew’s Weather Service (AWS) normally issues monthy reports
2) AWS is late with their report
3) AWS’s excuse for the lateness is that they got Obviously Incorrect data X from Al Gore’s Temperature Network
4) AWS published their report with data Y
Q: How does Lucia know that data Y is different from data X? In other words- what is the difference between data Y and data X?
A: Lucia doesn’t know
Big Picture: Lucia doesn’t know what is the right data and what is the wrong data. AWS’s report is meaningless to her.
Andrew
Lucia,
Ok so the Hadley people thought the data was wrong. Have they said why they thought it was? If it was a unique event why not describe it? Have there been previous occassions where the data was perhaps a little suspect?
These questions flow from the delay and the apparently, as you report it, lack of explanation for the aforementioned.
Hmm. HadCRUT for August says .518. Was .548 a typo?
All I can say is:
http://www.youtube.com/watch?v=YcRGzjE_xcA
No. They only said they thought it was obviously wrong, and so consequently did not process it to provide a global product.
I think this was the right thing to do. I don’t think we need to know the answers to the other questions to recognize that if they thought the data provided to them were wrong, they should not process and provide additional data.
If you (or Andrew) want the answers to the other questions, those should be directed to Hadley. But those answers are utterly irrelevant to what I have said: If Hadley thought the data provided them was obviously incorrect, it made sense for them to not process to provide a monthly data product.
kuhnkat,
Actually, this one is better I think :
http://www.youtube.com/watch?v=WGiV7mxzdGQ
Andrew
Andrew_KY,
Nope.
Andrew_KY: First, a question: If Hadcrut had posted data on time, would you have worried about your inability to detect “obviously incorrect” data?
Second: As Lucia has mentioned, there are a number of algorithms you yourself can use to detect obviously wrong data. 1) Suspicious number of data points exactly equal to last month. 2) Data points that are more than, say, 10 standard deviations from the previously recorded data. 3) Do a “last digit” analysis and make sure the distribution is fairly even (eg, see FiveThirtyEight.com and their expose of a probably fraudulent pollster).
(note that none of these are absolute proof of wrongness, but are certainly flags that the data deserves careful scrutiny)
Of course, this doesn’t tell you whether the “obviously wrong” Hadley data fell into the above categories, but it shows that you, too, can help prevent forest fires.
Marcus,
“If Hadcrut had posted data on time, would you have worried about your inability to detect “obviously incorrect†data?”
Yes. This part of the foundation of my Denialism. I have had doubts about the validity of data ever since I heard the phrase ‘Global Warming.’ So, I have been worried for years.
“As Lucia has mentioned, there are a number of algorithms you yourself can use to detect obviously wrong data.”
What algorithm should I use in this particular case?
Andrew
Andrew_KY
Why should we care what algorithm you should use? Are you planning to code it? Or get the data? Or start your won data processing method? Will you document it down to the gnats ass?
If you sincerely want to know the answer to your questions and the endless other pointless questions you are asking, contact Hadley. Then use the information in anyway you wish. If you want to provide the world updates feel free to tell us. But, quite frankly, I don’t care what the algo is. Moreover, the specific algorithm is utterly irrelevant to my conclusion that if Hadley thought the data was obviously incorrect, their decision to not process it was entirely proper.
“Why should we care what algorithm you should use?”
We should know what algorithm so we can check and see if it’s ‘correct’.
“Are you planning to code it?”
No
“Or get the data?”
No
“Or start your won data processing method?”
No. I’m not the alleged scientist here.
“Will you document it down to the gnats ass?”
Whomever publishes the report should document it down to the gnats ass.
Who’s asking the irrelevant questions, again? 😉
Andrew
Andrew_KY-
I have no inclination to check under the current circumstances. If you do, you should contact Hadley.
Quite honestly, given your answers to my questions, it’s clear you don’t intend to check and don’t really give a hoot about the answers to the questions you ask.
Clearly, you are. You are asking questions whose answers you don’t give a hoot about.
If you gave a hoot about the answers, you would be asking Hadley not me, and you would be planning to do something other than whine in response to any answers.
Going forward, on this topic, stop with the “rhetorical questions” format. If you have positive points to make, make them without using this idiotic imitation socratic method.
“Yes. This part of the foundation of my Denialism. I have had doubts about the validity of data ever since I heard the phrase ‘Global Warming.’ So, I have been worried for years.”
Ok. In which case, I think you might be asking the wrong question. Maybe it is not “how can one detect obviously incorrect data in the Hadley dataset” it is “how can one believe that the data shows the globe is warming”. The answer to the latter is that you have 3 somewhat independent surface temperature datasets, 2 somewhat independent satellite datasets, some ocean heat content data, and a whole footload of observational data on glacial retreat, sea level rise, earlier springtime greening, alpine plant migration, melting permafrost, etc. which are all consistent directionally with a warming planet. There is more uncertainty about the consistency in magnitude, even between the satellites + surface datasets, and determining consistency in magnitude between, for example, sea level rise and temperature requires complex modeling to do attribution of SLR to glacial melt vs. thermal expansion, but they don’t seem to be too far apart. In this context, Hadley’s data is just a part of the larger picture, so one can spot check it as best you can, but if everything is pointing the same way, one can make a reasonable inference that the errors in any given dataset are unlikely to be totally screwing the overall picture.
Lucia,
Did somebody get up on the wrong side of the bed this morning?
It’s Friday, dude. Why the bully tactics? I still like you and love your blog, so I will do as you request. 😉
Andrew
“one can make a reasonable inference that the errors in any given dataset are unlikely to be totally screwing the overall picture.”
Marcus,
Yes, I have heard the It Doesn’t Matter Defense many times.
Still leaves all my illegal questions unanswered.
Andrew
Andrew_KY
It would have been preferable if did not resort to trying to argue by rhetorical questions as in 23140 while simultaneously agreeing not to do it.
We’ve discussed this irritating pointless habit of yours several times before not just today. If you post any more “arguments by rhetorical question” I will moderate you.
Lucia,
“If you post any more “arguments by rhetorical question†I will moderate you.”
Can you make a list of what questions I’m allowed to ask? lol
Andrew
Marcus,
” In this context, Hadley’s data is just a part of the larger picture, so one can spot check it as best you can, but if everything is pointing the same way, one can make a reasonable inference that the errors in any given dataset are unlikely to be totally screwing the overall picture.”
Sorry, all you listed still does not tell us whether the slight temp change is AGW, natural variation, or a mixture, and what proportion that is.
Kuhnkat,
“Sorry, all you listed still does not tell us whether the slight temp change is AGW, natural variation, or a mixture, and what proportion that is.”
Um. I wasn’t trying to. The attribution question is completely separate from the ““how can one believe that the data shows the globe is warming†question which is what I thought would be the useful question for Andrew_KY to be asking. (And in response to his last comment, the Defense is not that It Doesn’t Matter but rather that multiple independent lines of evidence give a lot more confidence than any individual line, though one tries to make every individual line as trustworthy as possible)
Marcus,
the datasets are not independent.
Kuhnkat: Yes, which is why I specifically said “somewhat” independent not “completely” independent. And I’d argue that the “footload” of other measures I listed _are_ completely independent of the surface and satellite temperature records (well, except that they all probably depend on global average temperatures, which was kind of the point of my post).
It does seem, though, as if taking on the biggest conceivable global challenge has helped heal the wound, and perhaps even provided him with a satisfaction that being vice-president didn’t. “It’s a blessing to have work that feels fulfilling,” he says. “There’s a passage in the Bible – not that I wear religion on my sleeve; I do not – but there’s a passage that’s long had meaning for me: ‘Whatsoever thy hand findeth to do, do it with thy might’… There’s that wonderful old English movie, Chariots Of Fire, when the runner says at one point, When I run, I feel God’s pleasure. He was expressing a universal human emotion that I think is applicable.”
http://m.guardian.co.uk/?id=102202&story=http://www.guardian.co.uk/world/2009/nov/07/al-gore-interview-climate-change
Leave it to Big Preacher Al (can you say AGW is religious now?) to totally misrepresent a quote from one of my favorite movies.
The runner wasn’t expressing a universal human emotion. He was explaining to his sister that God had made him uniquely and purposefully for two specific things: Being a Christian missionary in China (where he would later be killed), and running. He was trying to make her understand that he was justified in spending his time running because God designed him for that. He is running at God’s pleasure, not his own.
Is Big Al finally admitting that he is playing The Slick Reverend and AGW is a religious scam by quoting the Bible and comparing himself to a fictionalized Eric Liddell?
Andrew
Lucia –
It seems to me the claims that current temperatures are unprecedented are now in the unproven category. This rather weakens the case for attributing the recent warming to CO2.
The radiation explanation is not affected by this but the physics and the models now play a greater role in the question of attribution.
You have shown that the models are not doing too well with regard to trends in global temperature and I wondered if it would be helpful to repeat your analysis for the NH and SH separately.
This would give us two opportunites to see how well the models are predicting the real earth trends. Clearly this could be done for all latitudes but I suspect the NH/SH data is more readily available.
If this is not too difficult and interests you it would be nice to see the results.
Jorge– Hemm.. Yes. I could add NH and SH separately. But I’m looking at a different contentious matter first!
Lucia – Something contentious in climate science? You heretic, you!