The Blackboard

Polar regions do appear to be warming faster than the tropics, at least in my temp anomaly model (as well as GISSTemp, etc.):

http://i81.photobucket.com/alb...../Graph.png

Zeke– Yes. They appear to be warming faster, and it’s what modelers predict.

Zeke, Is that NH only, or both? SH high latitude does tend to get underweighed in such things, due to more empty grid boxes.

Carrot,

NH only. Should have put that in the title. The units are in degrees C per year.

“Even someone with the IQ of a vegetable..”

I think you just insulted some of your readers who don’t see the point of gridded spatial averages.

If you had used the First Difference Method, you wouldn’t get the step change in anomaly at month 28. You’d still get the wrong trend afterwards; that just can’t be helped if you don’t have the thermometers. But at least the jump down would be gone. You would have trouble with noise, though.

I wonder if EM Smith saw something like that at some grid point, and decided to use the FDM instead. Hard to tell.

Hey, I resemble that tomato.
=============

Re: carrot eater (Mar 9 17:57),

I think you just insulted some of your readers who don’t see the point of gridded spatial averages.

huh? I wrote this:

Even someone with the IQ of a vegetable would realize that if you want to know the rate at which temperature changes in the sub box, you want these thermometers distributed evenly; failing that,

Whether or not you understand the point of gridding, you would recognize that if you want to know what’s happening in the sub-box (which is not gridded any more finely), you want the thermometers to be distributed around the sub-box rather than mis-distributed so they only catch one of the trends in the sub-box.

lucia (Comment#37206) March 9th, 2010 at 8:23 pm

It’s the same concept, though. To a person who doesn’t want to grid, the entire earth is one giant subbox, and he doesn’t care that half the thermometers (if you use GHCN+USHCN) are in the US. Meaning, he doesn’t care that the thermometers are not evenly distributed around his giant subbox.

“Climate scientists predict that the northern latitudes will warm at a faster rate than southern latitutes.”

Does this imply that Northern Latitudes will COOL faster than the Southern Latitudes also??

Since warming, or cooling, can be local, what about thermometers that don’t match the overall trend of the region or the globe.

Of course with relatively flat temps…

I’ve done the full globe using CRUTEMP (this is “land only”):

Temperature trend “Tdot” from 1960-2009 inclusive.

Largely these values are consistent with Zeke’s. I need to go back and see how I computed the error bars, I agree with Zeke’s figure they should increase towards the poles.

kuhnkat:

Does this imply that Northern Latitudes will COOL faster than the Southern Latitudes also??

I think the answer is “yes”. I think it’s due in part to the difference in the climate feedbacks near the poles compared to the equator ( snow cover, increase in humidity come to mind).

Zeke and Carrick,

The higher trends for Northern high latitudes are also a much smaller area of the surface so have proportionately smaller effect on the global trend don’t they??

Same for the small high trend band in the south and the negative trend in Antarctica.

kuhnkat,

Yes. Each grid cell is weighted by its size relative to the size of all grid cells reporting for the given month. So while the arctic may be warming faster, the tropics have a much larger impact on global temp.

Carrick,

My error bars are simple 95% CIs for the regression coefficient; they probably would need to be corrected for auto-correlation and whatnot for anything formal. Actually, I’m sure there is some interesting way to propagate the variance of individual station records into the error bars of the resulting grid cell anomalies.

Kuhnkat:

The higher trends for Northern high latitudes are also a much smaller area of the surface so have proportionately smaller effect on the global trend don’t they??

To amplify on Zeke’s comment, yes this is true. Also remember that ocean is 70% of the surface of the ocean, so SST actually dominates the global trend. Similarly, with urban environments being less than 10% of the Earth’s surface, the UHI effect simply isn’t large enough to overcome the much larger weighting from oceans and rural (as long as you don’t over-weight them by not using some form of area weighting).

With respect to latitude trend variations, one thing has bothered me. I’ve heard that the reason for the SH to warm more slowly is that it has less land area, but if you compare just land areas and just ocean, the NH still seems to warm faster. Anybody got a clue why that would be?

Andrew_Fl,

Because the oceans tend to moderate the warming.

Oceans warm more slowly. They exert a cooling effect on the land. With less ocean in the NH, this cooling effect is smaller.

Lucia writes:

Yes. They appear to be warming faster, and it’s what modelers predict.

Wouldn’t “It’s what the model runs show” would be a fairer remark?

Predict is a term-of-art in modeling; discerning whether the models predict rather than merely reflect requires us to know something about the state of mind of the modelers that is essentially unknowable.

I mentioned to Zeke several days earlier that the GHCN metadata was mostly garbage. An opinion which can be formed simply by checking the consistency of the meta-data against other records. e.g., google earth. A small but random sample suggested to me that this data was not very clean. I advised Zeke at the time that it seemed unlikely a difference in trend would be identified using this meta-data–little better than if it had been a random classification.

Nonetheless, he went off and did some analysis with a result that is consistent with my claim and breathless claims to have “shown” something quite to the contrary. Does that make sense? I’m confused to why you’ve chosen to uncritically repeat and amplify his claim.

Meanwhile, you also approvingly mention an analysis by Rob Broberg. The particular analysis you link to uses the output of GISTemp based on the GHCN input. Even though GISTemp omits some of the GHCN adjustments, it still includes the SHAP adjustments, which is enough to contaminate his analysis. Although his method appears to isolate groups of stations, the subsets he uses are already combinations of stations–including the excluded stations.

Now some of Zeke’s analysis have done better than this by focusing on GHCN raw series. Thanks, Zeke, I appreciate at least that level of detail.

Nonetheless, no one seems to be setting all the ducks in order at once. Rob uses SHAP adjusted data; Zeke uses contaminated meta-data. This proves nothing because the layers of contamination are profuse and fixing only one at a time does not really resolve matters.

Meanwhile, if you notice their analysis contradicts the recent UHI effect shown by Roy Spencer’s–albeit preliminary–work. But his work is well in the mainstream; the counter-argument against UHI relies primarily on that 1990 Jones et al. paper whereas many old papers had found strong UHI effects.

Indeed, the analysis posted thus far even contradicts Hansen’s nightlights paper which purports to discredit the use of GHCN rural/urban classifications when controlling for UHI effect.

The uncritical amplification and endorsement of the half-done analysis harms rather aids the pursuit of knowledge. I hope you will post an appropriately tempered update to this post.

Thanks.

Carrick: “To amplify on Zeke’s comment, yes this is true. Also remember that ocean is 70% of the surface of the ocean, so SST actually dominates the global trend.”

Yes. And due to the many sources of contamination that mask the AGW signal in land temp records — UHI, land use changes, etc. — we should probably rely on the SST record only to quantify that signal, and ignore the (apparent) land trends.

Nonetheless, he went off and did some analysis with a result that is consistent with my claim and breathless claims to have “shown” something quite to the contrary. Does that make sense? I’m confused to why you’ve chosen to uncritically repeat and amplify his claim.

The data is what you are given to work with. Jumping up and down and complaining won’t make it any different. The satellite record seem to agree quite reasonably with what the surface record is.

http://woodfortrees.org/plot/r.....00/to:2010

They work off different base lines, which I don’t know off the top of my head, but they seem to be talking about the same planet.

Save this someone. It will be “re-adjusted”! (as it always is when it goes that way) Its not possible!
http://arctic-roos.org/observa.....e_area.png

Laura S:

I mentioned to Zeke several days earlier that the GHCN metadata was mostly garbage. An opinion which can be formed simply by checking the consistency of the meta-data against other records. e.g., google earth. A small but random sample suggested to me that this data was not very clean. I advised Zeke at the time that it seemed unlikely a difference in trend would be identified using this meta-data–little better than if it had been a random classification.

Zeke used multiple proxies, they all agree. Kudos to Zeke for his careful work, not so many to you.

If you want your argument to be taken seriously, you need to publish quantifiable, verifiable facts, with a large enough data sample that you can demonstrate than an extrapolation from your sample to the full population has any meaning statistically.

Lucia, what happens if some locations because of urbanization or airport expansion are warming faster than others, and as we prune our stations, we increase the percentage of these urban stations? What happens if the effects of urbanization are not linear but stronger in the early stages, and if as we prune, we increase the percentage of stations in the early stages of rapid urbanization?

I have to say that the hypothesis that ‘it doesn’t matter’ or gets taken care of by anomalies makes sense to me. However it has so far had one insuperable barrier to credibility, and that is, that a study by Grant Foster supports it.

Foster for me will always be the guy who defended Mannian truncated PCA with a series of posts which a guy of his supposed mathematical sophistication must have known were pure horse manure.

Now you may feel that this is a gross ad hominem argument, and it is, but I learned ad hominem arguments from a forum where its mastery has reached the level of art. So expect more, but maybe more subtle stuff next time when I’ve had coffee.

Lucia

I think your idea with a toy planet is a good but i think you miss what Smith is trying to show . Here’s an even simpler toy planet:

You have only two thermometers placed in two equaly sized boxes and here’s the temperature records for those stations:

10 10 10 10 10 10 10 10 10 10

5 5 5 5 5 5 5 5 5 5

One warm station and one cold station but none has a trend.
Now you drop the cold ones after half the time and fill in the missing values with the values from the warm ones and you have this:

10 10 10 10 10 10 10 10 10 10

5 5 5 5 5 10 10 10 10 10

Now you do the anomaly stunt and get:

0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 5 5 5 5 5

and the average will be

0 0 0 0 0 2.5 2.5 2.5 2.5 2.5

This is a trend. So if you introduce bias in the DATA by dropping thermometers the analomies will not save you. The debate should be about this question: Has the “march of thermometers” caused a bias in the DATA? If it has, what’s been done to correct for it. The anomaly process alone can’t fix it as i understand things.

Bugs: “They work off different base lines, which I don’t know off the top of my head, but they seem to be talking about the same planet.”

The same planet, perhaps, but certainly not the same climate. The land temps are outliers, esp. GIStemp. But the LT temps and SST temps are very close indeed. The NH land temps especially exxagerate the trend — and those have known sources of contamination (UHI and land use changes). For estimating the AGW signal we should just disregard the land temps.

http://woodfortrees.org/plot/r.....00/to:2010

Laura S. (Comment#37252) March 9th, 2010 at 11:29 pm

“it seemed unlikely a difference in trend would be identified using this meta-data–little better than if it had been a random classification.”

But yet some difference was found, and then again using other metadata, as well. You seem to be ignoring that he used other sources of metadata. Perhaps those aren’t perfect either, but you could acknowledge they exist.

Interesting point on the random classification, though – perhaps it would be interesting to do some more random subsampling.

” Even though GISTemp omits some of the GHCN adjustments, it still includes the SHAP adjustments, ”

For the most part, SHAP does not exist anymore. And here, we can see that you can never please everyone. Some people complain that Zeke or Tamino aren’t using the exact same code and method as GISS, so we don’t know if GISS would react the same way. So Broberg does, and then you get mad. Go figure.

“Meanwhile, if you notice their analysis contradicts the recent UHI effect shown by Roy Spencer’s–albeit preliminary–work.”

No, it doesn’t.

“Indeed, the analysis posted thus far even contradicts Hansen’s nightlights paper…”

I don’t think it even does that.

AndreasW (Comment#37268) March 10th, 2010 at 3:15 am

“Now you drop the cold ones after half the time and fill in the missing values with the values from the warm ones and you have this:”

That would indeed be a problem, if GISS actually did that. They don’t.
__
“So if you introduce bias in the DATA by dropping thermometers the analomies will not save you.”

That wasn’t a bias in the DATA. That was a bias introduced by you using some ridiculous infill procedure.

You don’t need anomalies to save you from that; simply not being silly will save you from that.

Carrot eater

“That would indeed be a problem, if GISS actually did that. They don’t.”

Okej. So what do GISS actually do when they drop a thermometer? They surely don’t rip that part out of the map. They must fill it with something but what. Of course it’s the filling rather than the dropping that could cause a problem but it’s the dropping that causes the filling.

AndreasW (Comment#37275) March 10th, 2010 at 4:46 am

Have you ever actually looked at GISS output? If a station drops out at 1990, then the GISS record for that station also stops at 1990. If there are missing months, it puts in a missing month flag, 999.

More to the point, Andreas: why do you think you have to fill in the data after a station drops? Lucia doesn’t do it in the example above. Zeke doesn’t do it. Once a station is gone, it’s just gone. You simply hope that the remaining stations in the grid box are adequate to describe that grid box. As Zeke/ccc/tamino/Broberg have shown, that appears to be the case, though there will always remain some sampling error.

Contrarian:

The same planet, perhaps, but certainly not the same climate. The land temps are outliers, esp. GIStemp. But the LT temps and SST temps are very close indeed. The NH land temps especially exxagerate the trend — and those have known sources of contamination (UHI and land use changes). For estimating the AGW signal we should just disregard the land temps.

The land effect gets systematically bigger as you go further north. (See Zeke and my plots above.)

UHI doesn’t get bigger as you go further north. Neither do land-use change.

I think this smashes your theory into 100 pieces. Make that 1000 pieces.

Laura S:

Meanwhile, if you notice their analysis contradicts the recent UHI effect shown by Roy Spencer’s–albeit preliminary–work.

I agree with Carrot Eater here.

These results are largely consistent with Spencer’s work. He would have had to find a much larger UHI effect for that to explain a significant portion of even the land temperature trend.

In fact, the often touted US-48 temperature record contradicts the notion that UHI is a dominant source of bias, if you think carefully about it (think about “1933 as warm as current decade” mantra).

Carrick (Comment#37285) March 10th, 2010 at 6:20 am

To really compare the two, you’d have to get some estimate of how population density was changing post 1975 at the different stations, see how much of a temperature slope you’d expect from the Spencer curve alone, and see what you get, comparing neighboring stations. At many stations, the expected Spencer slope could be too small to be apparent amidst the noise, even in the absence of other trends.

I appreciate that Spencer corrected for elevation, but you could correct for other things as well (see Peterson 2003). I think microsite within the city can matter more than simply being in the city, but I suppose that’s another reason for the scatter around his curve.

Carrot Eater:

To really compare the two, you’d have to get some estimate of how population density was changing post 1975 at the different stations, see how much of a temperature slope you’d expect from the Spencer curve alone, and see what you get, comparing neighboring stations

Exactly You’d use this figure to compute a UHI correction based on population density. And fortunately for the United States that sort of demographic information exists… so while Zeke couldn’t do the full planet easily, he could test it on US48.

Anyway, it’s easy to see that the slope is too small to explain very much, as you pointed out. That’s just another way of saying you need more than a few °C warming from the UHI correction to explain a significant portion of the global mean temperature trend.

Re: michel (Mar 10 01:34),

Lucia, what happens if some locations because of urbanization or airport expansion are warming faster than others, and as we prune our stations, we increase the percentage of these urban stations?

You question actually contains a hidden assumption which is that urban stations or airports warm at faster rates. We know urban stations are generally warmer than the surrounding rural areas. But, now suppose during the 90s, the population of Chicago stays fairly constant, while the population of the quite small town Peotone, Il doubles in size. Which will warm faster during this period? I don’t actually know– do you?

But to answer the other bit: If, as a result of dropping thermometers, the you drop all those with faster warming trends, you get a warm bias in the computed trend. If you drop all the ones with slowerwarming trends, you get a cold bias in the computed trend.

Re: AndreasW (Mar 10 03:15),

Now you drop the cold ones after half the time and fill in the missing values with the values from the warm ones and you have this:

10 10 10 10 10 10 10 10 10 10

5 5 5 5 5 10 10 10 10 10

If this is what EM means, then the whole issue could be resolved by finding out if the operation is or is not done. We can run the code with toy data and see what happens.

However, I have to wonder: Is this what infilling would do? It’s not mentioned in Hansen and Lebedeff 1987 and would not be done if you apply their algorithm described in their 1987 papers. Of course, maybe they added this sort of thing later on? Can you point me to the paper where Hansen/GISS describes their infilling method? Or the bit of code?

lucia (Comment#37297)
March 10th, 2010 at 8:47 am

“If this is what EM means, then the whole issue could be resolved by finding out if the operation is or is not done. We can run the code with toy data and see what happens.

However, I have to wonder: Is this what infilling would do?”

Isn’t this the whole point of the ‘Bolivia-Effect’?

I did a comparison of the last 18 years between Hohenpeißenberg (alt.: 977m) and Zugspitze (alt.: 2960m). They are just 42km apart.
Zugspitze had a trend of 0.2 K/decade LOWER than Hohenpeißenberg. And both stations are rural. Thus throwing the higher alt. station out would result in a warmer bias.

Joel:

Zugspitze had a trend of 0.2 K/decade LOWER than Hohenpeißenberg. And both stations are rural. Thus throwing the higher alt. station out would result in a warmer bias.

Did you by chance try and estimate the uncertainty in that trend?

Individual sites have so much variability, I’m not sure what comparing two trends would tell you.

Laura S.,

Oddly enough, I don’t completely trust the metadata either. That’s why I turned to two completely independent datasets in my analysis (satellite nightlights and GPW population density data) in addition to the station metadata designation. The only case I can see being a real problem is if station location changed and was not reflected in the metadata, since the nightlight and GPW lookups use station lat/lon. However, I suspect its probably not too common a problem, and if it does occur it would more likely be an urban/bright/dense station moving to a rural/dark/low pop area rather than vis-versa.

Carrick,

Once I get the 1990-2005 change in pop density for each station, that’s the next step. Well, I’ll probably try to define my own curve (e.g. increase in trend vis-a-vis stations in the same grid 0 pop density change) at stations with different starting densities (since adding 500 pop density to a 2 pop density area probably has a distinctly different effect than adding 500 in a 1000 pop density area).

Joel, you peaked my interest so I went back and looked. They have a long period of overlap from 1950-1980 (unfortunately that doesn’t overlap with the theorized period of AGW).

What I found for that period was something closer to -0.01°C/decade (Zugspitze– Hohenpeißenberg), but the uncertainty was of the same order of magnitude as the trend.

Zeke, that sounds really cool. Looking forward to seeing it!

Carrick writes:

“Zeke used multiple proxies, they all agree. Kudos to Zeke for his careful work, not so many to you.”

Wrong. He used several kinds of meta-data, so what? His work is not careful. When attempting to demonstrate something; you must start from premises that people agree on. Your ready acceptance of his meta-data choices, reflect your ignorance here. This is not to say that the conclusions wrong, but still the method is wrong, *again*.

Carrick writes:
“These results are largely consistent with Spencer’s work. He would have had to find a much larger UHI effect for that to explain a significant portion of even the land temperature trend.”

I’m not interested in changing the subject to whether UHI explains a significant portion of the land temperature trend. I did not make that claim. Your indirection to that subject is dishonest.

Zeke appears to show a UHI effect about half as much as what Spencer just published.

As an academic, I’m in the habit of expecting that if people want their work to accepted as true, they make an effort to reconcile their results with those of others. Roy’s method broke free of many of the prior strictures of known contamination. He used a much more sophisticated approach to pairing sister stations than is done for SHAP or within GISTemp.

Zeke seems to have lots of independent ways of looking at the matter, but only a few of his analysis are meaningful to consider and there are still some curious differences.

“Indeed, the analysis posted thus far even contradicts Hansen’s nightlights paper…”

I don’t think it even does that.

The nightlights work concluded that the GHCN rural/urban distinction was not effective. Again, Zeke curiously reaches the conclusion that the nightlight approach is not different from the urban/rural partitioning. He labels his graphs of the light/dark as being global but the nightlights metadata is USA48. USA48 has a different trend than the rest of the world but his graphs show about the same trend. This is not careful work.

“For the most part, SHAP does not exist anymore.”

Wrong.

So get real. We shouldn’t get into a love-fest just because someone posted some graphs–or even a few someones. These guys are too quick to post conclusions and spin a narrative. They don’t deserve so much uncritical praise.

When good, careful work gets posted; I’ll be sure to point that out too.

Re: Joel Heinrich (Mar 10 10:34),

Isn’t this the whole point of the ‘Bolivia-Effect’?

I know that Chiefio discusses Bolivia and discusses the fact that they drop out the thermometers. But where does he show that, when computing anything they would infill the way AndreasW say? I don’t believe they ever do what Andreas suggests.

Is there a difficulty presented by the fact that south america has poor coverage? Yes. If Bolivia is warming at a different rate from better instrumented countries, that data is not being included in the computation. But the fact that the mountains of bolivia are colder than near by coats is not the reason for the warm anomaly on that map.

If Chiefio stepped through what the GISS algorithm does, I suspect he would figure this out.

Laura

Zeke appears to show a UHI effect about half as much as what Spencer just published.

As an academic, I’m in the habit of expecting that if people want their work to accepted as true, they make an effort to reconcile their results with those of others. Roy’s method broke free of many of the prior strictures of known contaminatio

In this post http://www.drroyspencer.com/20.....sity-data/ Roy did not quatify the effect of UHI on trends or even global surface temperature anomalies. Roy is doing some careful work that will help him try to do that– but he will have to use the information he has to come up with a UHI correction based on changes in population over time.

I strongly suspect Roy is planning to do this, but he has not posted results on his blog. Zeke does not need to reconcile his computations of surface temperatures with Roy’s because Roy has not posted results for surface temperatures. Zeke not only doesn’t have to reconcile — he can’t.

Laura S,

Nightlight data is global. See http://www.ngdc.noaa.gov/dmsp/.....sites.html.

Where do you get the idea that my UHI results were half those that Spencer found? The absolute temps and the effects on the trend in temps aren’t really comparable unless you know the change in the relevant factor (e.g. pop density) over time for the locations in question, which the analysis I posted so far can’t really show since I’m only looking at single-point-in-time snapshots of urbanity proxies.

“These guys are too quick to post conclusions and spin a narrative. They don’t deserve so much uncritical praise.”

I agree with Laura S.

I have observed that making and looking at graphs and being impressed by them is just one of the many subjective influences that inhibit warmers from approaching climate science objectively/seriously.

Andrew

Laura S:

I’m not interested in changing the subject to whether UHI explains a significant portion of the land temperature trend. I did not make that claim. Your indirection to that subject is dishonest.

You claimed Zeke’s results aren’t consistent with Spencer. You didn’t say in what respect, I inferred you meant in a way that actually mattered…e.g., temperature trends.

It doesn’t make me dishonest to assume this is what you meant, especially when you didn’t spell it out any further.

Zeke appears to show a UHI effect about half as much as what Spencer just published.

I don’t follow you argument. How has he shown us 1/2 the UHI effect as Spencer?

Andrew_KY:

I have observed that making and looking at graphs and being impressed by them is just one of the many subjective influences that inhibit warmers from approaching climate science objectively/seriously.

These dratted scientists and engineers! Always getting persuaded by the results of carefully framed scientific studies.

Lucia:

Zeke not only doesn’t have to reconcile — he can’t.

One could do it for the US48 where there is demographic data over time, using Spencer’s published curve.

But your main point is completely valid, there are no studies done yet that directly compare the two.

Perhaps Laura S enlighten us.

Re: Carrick (Mar 10 11:33),

These dratted scientists and engineers! Always getting persuaded by the results of carefully framed scientific studies.

Heh. I still remember a high school student on one of the co-op thingies DOE sets up once marveled that everytime there were 2 engineers in a room talking, within 15 minutes, one was grabbing a marker and drawing pictures on the white board, and often, cartoon graphs.

We laughed. We knew that we communicated ideas with schematics, free body diagrams, control volumes, streamlines, isobars, maps, scatter plots with theoretical curves overlaied.. blah blah blah. They were all just scribbles to her.

Yes, Andrew. We make and look at graphs and are impressed by them. We also make and look at all sorts of other ‘picture – like -thingies’ and are impressed by those.

The fact is, nearly all Americans are used to abstract visual representations of information. For example, little kids generally understand the abstract concept of “a map”. With training, people use more complex visuals. Happens.

Andrew_KY,

If you don’t like the way the graphs display the data, I posted excel files with the full spatial model outputs for each series graphed since 1880, and you can make your own graph. If you don’t trust the model, download the source code and audit it, or make your own model like Jeff Id and Tamino have. That’s how science works

Re: Carrick (Mar 10 11:39),

One could do it for the US48 where there is demographic data over time, using Spencer’s published curve.

Yep… I suggested this to Zeke in private email a few days ago. Not so specifically, but if he had the info, it would be cool to take the info in Spencers graphs and do something. Of course, that’s a big picture suggestion and, as you know, there is always a lot of work between that and the final graph.

I strongly suspect Spencer himself is doing this. I suspect he thought of doing this before he did all the work to create the graph he revealed, and he plans to show his results. In fact, when I saw Spencer’s graph I thought “super cool, he’s finally going to do UHI in a way that I’ve always sort of wondered about. He’s going to look at how progressive evolution in population could cause a trend!”

If that’s his plan, it is a super cool idea. WAY cool.

But Roy hasn’t done that yet, and I’m sure he wants to do it carefully before he posts. When Spencer posts, or writes a paper, he will compare his results for UHI to those using previous methods. But asking Zeke to compare to Spencers results is odd because Spencer hasn’t posted any … yet!

Lucia,

Maps are only meaningful if they represent what they are supposed to represent. A map that looks nice, but doesn’t represent the area it’s supposed to represent, makes for an unimpressive map. It might be an impressive drawing, but that would make it art, not science

Andrew

Lucia,

I mention doing it in the “Next Steps” section of the UHI post.

Interestingly enough, satellite nightlight data is also available for every year since 1992, so if (and its a big if) the satellite brightness data is comparable year-to-year we might end up with two temporally-specific UHI proxies that cover the past 15 years (pop density and nightlights).

Andrew_KY–
Sure. But when you criticized people for being impressed by the graph, you didn’t explain how the map might not represent what it’s supposed to represent. You merely complained we used graphs.

To use the map metaphor: It’s as if you criticized people for the whole idea of using a map.

If those of use who hold with the general notion that maps can be useful heard you lecturing us about being impressed by the map, we’d say, “If you have a specific reason to think a particular map is flawed, state it. Describe in what way it is flawed and why you think this map doesn’t provide us at least some useful information that will help us answer the questions we actually have?”

Zeke–
Yes. I saw it mentioned in the “Next Steps”. I thought “Yippe!!!”.

Lucia,

OK. In clearer terms, what I meant to convey was that we should be aware that graphs and charts do not automatically convey meaning. You have to know how the representation was produced and why, and that kind of thing, before you can say it has meaning.

Andrew

Andrew_KY

You have to know how the representation was produced and why, and that kind of thing, before you can say it has meaning.

Of course. But I should point out that your simply making vague snide sounding cracks about people reading graphs at all fell far short of communicating the notion that you are concerned about something specific about the graphs. You certainly aren’t telling people what you think might be wrong with it.

Because one must know how a representation was producued and why, I have been explaining the methods used to create the avearge temperatures that appear in these graphs. The people who create the graphs already know these things. The papers discuss how they are created. Many of us who are impressed by the graphs already know these things.

But some people have developed misconception about how the temperature anomaly graphs were produced and what they are meant to represent, don’t trust their meaning. My goal is to explain in simple terms so that those people can understand which potential problems are real and which problems are just not problems. To a large extent, some of the wording at Chiefio has caused people who do not understand how the graphs are created or what they represent to think certain issues are problems for the graphs, when in fact these specific problems are just not problems.

If you have something specific to criticize about the way the data are handled, go ahead. But merely making snide remarks about the notion that people uses graphs is pointless. Yes. Scientists and engineers use graphs. So do economists, business people and even 13 year old students in junior high. Make all the cracks about the general notion you like, everyone who uses graphs will begin to suspect you are just a person who wants to rebut by making pointless cracks.

I’d move the toys to the side of the romper room floor and step up to the blackboard but watching the 2 year old-esque conflict is way too much fun.

I see that there are still way too many peeps who have not reviewed the literature and are spesking from a… ahem… lacking position.

EM Smiths’ mistake is that he calculated actual temps, not anomalies. If he would he would have noticed that removing northern thermometers would have the opposite effect of what he found because the stronger trends are in the higher lattitudes (I’m not really sure about southern lattitudes, not long enough observation there)

Removing more rural stations will have a very small warming effect, but not much (as has been discussed already… a weighting thing)

Altitude is somewhat like northern lattitude. The higher you go the faster it warms so you are again removing stronger trends, so the effect would be opposite to what was reported by Smith.

I’ve seen nothing in these discussions or models that address the underlying issue of the data quality itself. How can any conclusion be drawn unless and until each and every individual station is analyzed? Not by computer models, but physical inspection and measurement. There seem to be many assumptions but not much from real world observations.

From a metrology (not to be confused with the other -ology) standpoint, what industry standards for quality control are being employed to evaluate the data derived from surface stations? If it were constrained to basic quality control acceptance standards used in industry, the entire data set in my experience would be sent back to the supplier as incomplete.

As for UHI, I’m not buying the idea it is properly accounted for.

Taking two extremes of population density, one being Los Angeles, the other Barrow, Alaska.

LA population density = ~3,041.3/km2
Barrow, Alaska population density = ~84/km2

The studies are:
http://climate.jpl.nasa.gov/fi.....rticle.pdf
http://www.geography.uc.edu/~k.....A_2007.pdf

And we’re expected to believe rural is truly rural and UHI is corrected properly for urban locations?

What about microclimate, station location changes, ground level atmospheric changes from land use change, instrumentation changes etc. affecting the data? None of this has been taken into account. Who did it? When? How?

30 years experience in manufacturing engineering/QA tells me it doesn’t pass the smell test.

Lucia,

I understand what you are saying. But a general audience includes people who have never looked at this debate closely. I volunteer teach a computer course and I know from experience that people do need to be made aware of how to approach of looking at representations. They should not be taken at face value. That’s all I’m trying to say.

Andrew

Zeke,

One thing to take note of in the nightlights is blooming.
The End product is a result of the application of 3 different
gain setting for the sensor. One setting to make sure that dark is really dark ( and then the brighter sites will tend to bloom) and then a setting to get a clear urban boudary ( and some dim sites
appear dark ) and then a setting in between.
( that’s kinda how I recall what Imhoff said ( 97) ) I think in some of the later producst they may have adjusted the threshold.

FWIW

Re: DG (Mar 10 12:30),

I’ve seen nothing in these discussions or models that address the underlying issue of the data quality itself. How can any conclusion be drawn unless and until each and every individual station is analyzed?

Data quality is a real issue. Unfortunately, if a large group or people are convinced something that is a non-issue is an issue, we can’t move forward unless we can show that issue is a non-issue. Since there are some people who have a list of say 5 “issues”, which includes non-issues, we need to address the non-issues.

Joel Heinrich (Comment#37345) March 10th, 2010 at 10:34 am

No. GISS computes a temperature record at a given grid point using the method Lucia lays out above, with each station weighted by how close to the grid point it is. That’s sometimes called interpolation. It absolutely is not the in-filling that Andreas was describing. If a station drops out, then a station drops out.

Now, the USHCN does do infilling, though not as stupidly as the illustration. This is for the US, only. These values are flagged in the USHCN file. And I’m pretty sure that GISS hunts down those flags and removes the in-filled points.

In step 0, when it’s going through USHCN data:

for m in range(0,12):
temp_fahrenheit = int(line[m*7+11:m*7+17])
flag = line[m*7+17]
if ((flag in ‘EQ’) or # interpolated data
(temp_fahrenheit == -9999)) : # absent data
temp = None
else:
# tenths of degree centigrade
temp = round_to_nearest((temp_fahrenheit – 320) * 5/9.0)
valid = True
temps.append(temp)

So it reads in the month’s temperature from the file, into temp_fahrenheit.
It reads in the flag, if there is one.
If the flag says the month was infilled, it does not use the number. GISS puts in a null for that month.
If there is no such flag, then it converts from F to C, and goes on its merry way.

If EM Smith is telling people that GISS does infilling (as Andreas suggested, and I think it was suggested in the spherical cow opus), then that is one more thing he’s got wrong. GISS goes out of its way to delete in-filled data.

Lucia, you said: “You[r] question actually contains a hidden assumption which is that urban stations or airports warm at faster rates.”

Well maybe. I’d have expected it to be non linear in either case with development, probably most pronounced in the early stages of development. But the question was general, not anti-warming.

The thing I’m having trouble getting my head around is how the anomaly method possibly could compensate for changes in the characteristic of the sample as it reduces by two thirds or more. And if Tamino thinks it doesn’t matter, my instinctive bet would be, it probably does!

Michel–

The thing I’m having trouble getting my head around is how the anomaly method possibly could compensate for changes in the characteristic of the sample as it reduces by two thirds or more. And if Tamino thinks it doesn’t matter, my instinctive bet would be, it probably does!

Well… Tamino is sometimes correct. Or course, whether or not the loss of thermometers “matters” depends on what question you are asking.

Anyway, the anomaly method only compensates for certain very specific issues. It’s not a magic bullet that can solve all possible problems with the data. Reducing the number of thermometers, particularly preferentially in some reasons could potentially bias the computed trend. However, the specific mechanism for warm bias suggested at Chiefio’s is not a mechanism that we expect to affect trends based on anomalies. Other mechanisms do exist, and the loos of 2/3rds of the thermometers hypothetically could have caused a bias.

So, far, it appears that this loss did not cause any strong bias. That statement is based on people computing the surface temperature record a while bunch of different ways using raw data, and comparing trends with between the thermometers that were dropped and those that were not. It looks like dropping 2/3rds of the thermometers probably did not bias the record in any meaningful way. It probably did result in a noisier record with more variability. But that’s different.

Laura S. (Comment#37352) March 10th, 2010 at 11:11 am

As an academic, I’m sure you’ll tell us what careful work and calculation you did to reach your conclusion that Zeke and Spencer are inconsistent with each other. It certainly isn’t obvious just by looking at it.

“The nightlights work concluded that the GHCN rural/urban distinction was not effective. Again, Zeke curiously reaches the conclusion that the nightlight approach is not different from the urban/rural partitioning.”

Hansen decided that nightlights were better (which is still an open question, so far as I care). But when you switch from GHCN rural/urban to nightlights for the non-US stations, as GISS just now did, what did GISS get?

>”The effect on the global temperature trend is small, that change reduces it by about 0.005 °C per century.”

http://data.giss.nasa.gov/gistemp/updates/

Just because you think something is better doesn’t mean it’ll give you hugely different results. That said, I’m not sure if they’re using nightlights alone, or what; the paper isn’t out yet and I haven’t read the new code.

“Wrong.”

I stand by my statement. SHAP does not exist in USHCN v2.0. It’s still hanging around as a lame duck for the handful for US stations in GHCN. But that’ll probably be gone soon, too. We’ll see what they do.

Lucia: People are still struggling with the concepts of anomalies, gridding and over/undersampling.

Perhaps an educational exercise would be to take the pre-1992 data, and methodically figure out what stations you would remove, if you wanted to significantly change the results. That process would maybe illustrate the issues involved, as well as the resiliency of the database.

carrot—
I’m not going to do that! I’m just reluctant to get into computing surface trends. That’s why I’m glad Zeke is doing it and set him up to write guest posts. (Basically, one person can’t do everything, and to do that, I have to carve out time for that. I want to do other things.)

But yes, someone could go through, compute all the individual trends. then screen for the lowest 10%, drop them out. Show that new trend. Then list out which stations were dropped. A graph showing latitude vs. %dropped per latitude or average annual temp ve. %dropped would be interesting. But… not going to do it!

Also, the trend-by-latitude graphs that Carrick and I posted pretty much already do that.

Want to give the temp trend a warm bias? Drop all the warm stations! Its delightfully counter-intuitive

Again, Zeke curiously reaches the conclusion that the nightlight approach is not different from the urban/rural partitioning.

What is “curious” about it?

I strongly suspect Roy is planning to do this, but he has not posted results on his blog. Zeke does not need to reconcile his computations of surface temperatures with Roy’s because Roy has not posted results for surface temperatures. Zeke not only doesn’t have to reconcile — he can’t.

I disagree. We can still ask if the results are consistent with each other. They do not appear to be–although further rationalization (as yet not proffered by you or Zeke) may ‘explain’ the situation.

Lets start with the premise that the trends are affected–the mechanism for this that what we’d expect is that high population density stations were once low population density stations. As their surroundings developed, their temperature rose faster. Lets assume the the effects combine linearly. This is internally consistent with Zeke’s analysis as posted: rural and urban stations diverge.

Now further, we can expect that the direct CO2 warming component impacts rural and urban stations equally.

Zeke’s analysis contains a curious feature, for several decades high density and low density stations are not divergent. Zeke states as an assumption that low-density stations in the present were low-density in the past. That seems plausible, and since he assumes that himself, lets go forward with that assumption too.

The lack of divergence during the early period suggests that the high-density stations were previously low-density. That is, since L and H stations in the past are indistinguishable, and L stations in the present are low density, and L stations in the present must be L stations in the past, then H stations in the past were L stations.

We can now proceed: one of Zeke’s classifications is 10/sqkm versus 100/sqkm. Roy suggests that there is an absolute difference at similar stations of >0.6—I say greater because Zeke’s classifications are bins whereas I’m taking the closest two points from Roy that bound the bins. However, the trends have diverged by less that 0.3C in Zeke’s graph.

This is the inconsistency.

Note: Zeke claims trends of 0.209 vs 0.182 but these are artificially lowered by the long period of time during which most of the H stations were L stations in the foregoing analysis. Once the divergence behinds, the local trend difference is 0.2C/decade! But regardless, this is a secondary claim.

Now, this may be comprehensible with further analysis, but my assumptions were reasonably weak, so I don’t think its wrong to expect a bit of hesitancy and notice of the discrepancy–even as further analysis may well set all the ducks in a row after all.

It should be clear that there is a reasonable basis upon which to believe that Zeke’s conclusion is, at the moment, presumptuous–regardless of whether you believe my assumptions must be horribly wrong. That indeed, is the very business that must be shown for Zeke’s analysis to be careful and trustworthy.

… And what it takes for the analysis to be persuasive.

Before anything else, I’d suggest putting some considerably wide confidence intervals on any estimate you take from Spencer’s graph. He’s added some data.

http://www.drroyspencer.com/wp.....y-year.jpg

Not sure if I’d get take > 0.6 out of there

Laura S.,

Lets try a little thought experiment. Suppose there are 10 stations in a rural area and 10 in an adjacent urban area. All stations have been present for the past 40 years, and the relative population density of both areas has remained largely unchanged in the past 40 years. Dr. Spencer goes and measures the temperature in each station this year and finds that the urban stations are 0.8 C warmer than the rural stations on average. I look at the data for the past 40 years for each group, calculate the OLS trend for each, and determine that both have warmed at a rate of 0.1 C per decade over the period with no significant difference in the rate between the two groups. Both of us are correct in our analysis.

Now, consider a second case where we have the same 20 stations. The 10 in the rural area have remained unchanged for the last 40 years. However, the 10 in the urban area started in a rural area 40 years ago that urbanized in the interim. Dr. Spencer does his analysis of the temperature difference this year and, again, finds that its 0.8 C between rural and urban. I look at the temp trends and this time find a 0.1 C per decade trend for the rural stations but a 0.3 C per decade trend for the urban stations. Again, both of us are correct. Because all stations changed from rural to urban, the current year difference Spencer saw (of 0.8) is directly reflected in the trend (0.2 extra UHI-driven warming in urban stations per decade for 4 decades).

Now, consider a third (and perhaps more realistic case): 10 stations stay completely rural for the entire period, 5 stations switch from rural to urban over the period, and 5 stations stay urban over the entire period. Again, Spencer’s analysis for the current year remains unchanged. My analysis still shows 0.1 C per decade for the rural stations, but I now get 0.2 C per decade for the urban stations. My rate of UHI-driven warming (0.1 C per decade, or 0.4 C over the 4 decades) is half of the absolute temp difference between urban and rural that Spencer finds (0.8 C). Again, as in the first two cases, both of us are completely correct.

So as I mentioned earlier, we cannot claim that my results are inconsistent with those of Spencer without knowing how much the population density changed in the past.

I don’t know, Zeke. I think the assumption that every single urban station was rural some 20 years ago is a pretty compelling one.

Carrick: “The land effect gets systematically bigger as you go further north. (See Zeke and my plots above.) UHI doesn’t get bigger as you go further north. Neither do land-use change.”

That’s true, but as you go further north, the area involved decreases, contributing a smaller part to the hemispheric trend. Also as you go north, the ratio of coastal to inland stations increases, which means you’re detecting SSTs, which are being warmed by ocean cycles (AMO, etc.), rather than land temps. The bulk of the hemispheric trend is determined by lower-latitude stations, which are affected by the contaminations I mentioned.

There is nothing in AGW theory — that I know of — which suggests surface temps over land should be higher than over water. (If I’m wrong about that, let me know). Hence any differences detected must be spurious (as far as the AGW signal goes) and should be ignored. The easiest way to do that is to disregard the land data.

Zeke,

I agree there is some weakening of the assumption that H stations were L stations prior to time T such that your results and Roy’s would be consistent. I hope that my understanding of that is obvious by my very careful and specific statement of the assumption being made by the analysis given and my general concession that my argument can be falsified by additional data that would leave your work in tact and my concerned resolved. That is, I am not alleging that your work is wrong per se.

Nonetheless, I must reiterate that I did not pick that assumption arbitrary. Indeed, a priori, I would have considered the assumption that H stations were L stations rather excessive, but your calculations as presented suggests that the assumption is reasonable.

Which, indeed, is part of the reason that I am not convinced that as of yet you have all the ducks in a row, and so your thought experiment is consistent with my argument and does not directly respond to it.

I don’t know, Zeke. I think the assumption that every single urban station was rural some 20 years ago is a pretty compelling one.

Carrot eater: that’s a rather glib remark. I’ve been careful to not explicitly say all H were L prior to the trend divergence. That seems patently wrong. I agree there must, at a minimum be, some epsilon of stations that were always high-density; however, again, the several decade stretch of near perfect covariance during the early period implies the assumption made.

That’s part of what makes the computations as presented less than convincing.

I guess my own conclusion is, the alleged surface station record is part of the argument for getting trillions out of me and others for remedial action.

If they want money, then what I am going to have to see is a small number of stations with long records with the same equipment, no material changes to their siting that could influence readings. Proper metadata on each one. None of this crap we saw in the Jones Chinese stations, refusing to even reveal the names of the stations.

If that shows unprecedented warming, then I might start to take it seriously. Meanwhile, if we take the number of stations from 100 to 6,000 and back down to 2,000, give no account of why we’ve done it, or what the variation in the sample has been along important dimensions that this has given rise to, forget it. I’m not buying, and no-one I know is buying either.

Laura S. (Comment#37489) March 11th, 2010 at 12:48 am

“Carrot eater: that’s a rather glib remark. I’ve been careful to not explicitly say all H were L prior to the trend divergence. ”

Yet your back of the envelope calc requires it, and you initially presented your result without much of any caveat.

I don’t think this back-of-the-envelope calc is particularly insightful.
.
I do think your line of inquiry would be better directed by simply asking this question: “Why do some of the data series only seem to diverge from each other in 1985 or 1990 or so?”

Put it that way, and Zeke could poke around and see what’s going on there. Maybe there are regional effects, though there might not be enough data to really get a clean look at that.
.
Could just be sampling error, if it’s more apparent after 1990 in any of the plots. Don’t jump on me for that if it’s unlikely; I haven’t thought it carefully yet.

LauraS

I disagree. We can still ask if the results are consistent with each other. They do not appear to be–although further rationalization (as yet not proffered by you or Zeke) may ‘explain’ the situation.

If you think it is possible to show whether Roy’s graph is consistent with Zeke’s results, you do it.

We can now proceed: one of Zeke’s classifications is 10/sqkm versus 100/sqkm. Roy suggests that there is an absolute difference at similar stations of >0.6—I say greater because Zeke’s classifications are bins whereas I’m taking the closest two points from Roy that bound the bins. However, the trends have diverged by less that 0.3C in Zeke’s graph.

This is the inconsistency.

You don’t know if there is inconsistency or not. It might be or it might not be. You are speculating about the result you will get if take all the GHCN data, the population data, the gridding and the area weighting and compute the trends. Zeke said at the end of the post he (unlike you) is planning to run the actual, honest to goodness numbers and compare. When that is done, it will be possible to make an informed comparisons.

Zeke, when he makes conclusions, will be basing them on analysis he has actually done rather than these “in the air” analyses you seem to feel free to post. Not that you are not permitted to do so. But it is rather presumptuous of you to suggest that your “if I bothered to run the numbers I think I would bet this answer and so Zeke is wrong and shouldn’t report results in a blog post ” claims in comments.

Contrarian:

That’s true, but as you go further north, the area involved decreases, contributing a smaller part to the hemispheric trend

Whether this is true is irrelevant to the argument (I’m not sure why you think it matters).

If UHI and other human activity were the main biases in the trend, you shouldn’t get an increasing trend as you move away from them to the North.

But actually you’re wrong in detail anyway. The contribution to the total trend from land actually tends to increase as you go North, until you reach roughly 55°N then the contribution approximately constant until you reach 65°N. (Not that I think this is relevant, but I figured you’d like to know your claim was wrong.)

Also as you go north, the ratio of coastal to inland stations increases, which means you’re detecting SSTs, which are being warmed by ocean cycles (AMO, etc.), rather than land temps.

The coastal ABL decreases the sensitivity to temperature fluctuations not increases it. That’s a formal statement. Empirically, if you look at inland versus coastal stations you will find that the coastal stations follow this trend of lower temperature sensitivities (I have checked). Also the fraction of land to ocean increases as you go north, which decreases the marine influence on inland meteorology.

There is nothing in AGW theory — that I know of — which suggests surface temps over land should be higher than over water

Not AGW necessarily, but micrometeorology and the theory of the atmospheric boundary layer (ABL) predicts it. That’s been known since before AGW became a major research topic.

It’s the caution I’ve been giving people for a while in trying to compare global temperature trends from GCMs that don’t include ABL physics to measurements within it. Same caveat goes when comparing satellite measurements above the ABL to surface ones, because they don’t have much sensitivity to the atmospheric temperature within the ABL.

Laura S.,

The visual convergence at the start of the series is something of an artifact of the choice of a baseline (1961-1990) that covers that portion of the series. Actually, if you look closely, there is something of a inversion near the start, with H (high density) stations showing lower anomalies than L (low density stations). This is because each series is baselined relative to its own anomaly, so the difference in trends between the series manifests itself visually in the changed slope rather than the absolute magnitudes. If we try a 1989-2009 baseline, for example, the divergence visually appears in the earlier part of the series but not the latter.

http://i81.photobucket.com/alb.....ure179.png

If we look at the difference between less than 10 pop density and over 100 pop density stations, you can see that apart from some weirdness around 1992, its mostly linear.

http://i81.photobucket.com/alb.....ure177.png

If we detrend the residuals we see that while it is a lot noisier after 1992 (less grid boxes used and whatnot), the difference is only slightly larger.

http://i81.photobucket.com/alb.....ure178.png

Lucia

might be cool to do a toy model of this..

If you have a UHI effect that is non linear as Spenser purports
and you have a sampling of stations with various population growth rates, and you impose a AGW signal onto that field can you find that signal ( hmm, lots of variables to play with) must have morning coffee before thinking

Thanks Carrick.

Zeke the metadata should have a flag for coastal..
Also, when ( and if) you get to processing the whole globe you need to look at how you average gridboxs where the boc contains both land and ocean. In fact for your UHI study, you might consider dropping out those grids that have any ocean in them whatsoever.

Let me see if I get this right. Since the station drop out does not effect the trend we can randomly pick any number of stations at any point in time and remove them and the trend will be the same.

If you have a UHI effect that is non linear as Spenser purports
and you have a sampling of stations with various population growth rates, and you impose a AGW signal onto that field can you find that signal ( hmm, lots of variables to play with) must have morning coffee before thinking

Lots.

For a back of the envelope calculation, you need realistic population density numbers as a function of time. I think we also need Roy to finish up his analysis so that we know his UHI vs popdensity function is stable.

Andy Krause (Comment#37550)
To some extent, yes. If we deleted stations utterly at randome, then we would expect that the computed trend would be unaffected– within statistical uncertainty bounds.

To explain the caveat. Suppose you got a group of 2000 18 year old navy recruits from a total of 50 American cities selected at random.

You measured the height of each recruit, and computed the average height of a navy recruit. Now, you want to find out if the sample average height sensitivity to throwing away data. Using a random number generator, you throw away 1/2 the data. You will expect to get the same height using 1000 of the recruits as you got using the 2000 recruits. BUT “the same” doesn’t mean both samples will give the same average height to 12 decimal places. It means they will be “the same” means within some statistical uncertainty. (For army recruits heights, I’d guess this is around 0.3 – 0.5″)

On the other hand, if you know that ethnic group, nutrition and other factors varied widely across the US, and you dropped data from recruits who came from cities where people were known to tend to shortness, then your answer would change when you dropped those recruits from the sample.

Suppose instead, you dropped data from all blue eyed recruits. Would that matter? That’s a “bias” in the sense that you dropped based on some observable characteristic. Still, blue eyes is not the same as height. I don’t know if blue eyed ethnic groups tend to be taller or shorter than brown eyed ones– but the height of blue eyed people is, on average, different from brown eyed people, that would matter.

So, if we use the recruit thing as an analogy for thermometers, dropping more northern thermometers is like dropping recruits with blue eyes.
If you dropped samples in a biased way, that can “matter.” But we can sort of check to see if the bias seemed to matter.

Zeke Hausfather (Comment#37528) March 11th, 2010 at 9:52 am

I was wondering if it was a visual illusion like that, but then thought you pinched the graphs so they coincided in the beginning. I have to keep in mind that Zeke likes to plot everything against its own average.

Plotting also the difference is really the most convincing thing in such cases; thank you.

carrot eater,

The gridding model spits out the series data relative to each series’ own 1961-1990 anomaly, so thats what I use for convenience. Any suggestions on how to plot two anomaly series that use separate groups of stations as inputs on an objective scale, so to speak?

Zeke Hausfather (Comment#37558) March 11th, 2010 at 1:30 pm

This is a challenging question, when two series are not obviously different.

I think you just have to take it case by case. What you do is generally fine. In some cases, like comparing trends, or in looking at adjustments, I think showing the difference in a second plot is needed, as you provided above. You couldn’t do that for every single plot in your UHI article, or else it’d just be too much.

Now and then, pinching them so they coincide at one end is appropriate. I thought it was useful in the plot you made for me comparing global with different individual countries. It’s also useful when comparing raw and adjusted at single station. But when it is switched up, the reader has to know; it’s possible to play games this way, too.

It’s good to have the trend from linear regression stated each time, and you did that in the UHI post. I really appreciated that. The eyeball simply can’t linearly regress data and tell the difference between +0.16 C/decade and +0.18 C/decade. It just can’t.

An observation. Judging on the basis of the questioning of the precise methods used to produce the various temperature graphs here, how can anyone claim that the ‘science is settled’?

Dave Andrews (Comment#37575) March 11th, 2010 at 2:54 pm

An observation. Judging on the basis of the questioning of the precise methods used to produce the various temperature graphs here, how can anyone claim that the ’science is settled’?

Given the consistency of the results across different technologies, it seems to be something like that. The only argument is the precision.

I have a question, if GISS is using the methods described then why do they rewrite past temperature anomalies constantly? I looked at the GISTEMP J-D 2005 and compared it to the GISTEMP J-D 2009 and looking at only the 1880-2004 so that it was apples to apple the trend changed. The 2009 record showed 10 percent more warming than the 2005 record for the same period.

Vernon #37654
Changes to the GISS methodology are logged here.

Nick,

That would a good answer if the past record years records were not being changed monthly. Monthly changes are not address on that page. So, do you know why GISS constantly changes past years temperatures?

Lucia, you just go out further and further with drivel when you write:

But it is rather presumptuous of you to suggest that your “if I bothered to run the numbers I think I would bet this answer and so Zeke is wrong and shouldn’t report results in a blog post ” claims in comments.

I never said that. Never. I directed most of my criticism at you. Recall, my complained was this: The uncritical amplification and endorsement of the half-done analysis harms rather aids the pursuit of knowledge. I hope you will post an appropriately tempered update to this post. The rest of what I’ve written has been an exposition on what I mean by ‘half-done’.

It was you who triumphantly declared: “In the end, the proof is in the pudding. … The exact same analyses by Zeke, Tamino,, CCC, The Whiteboard etc. all suggest … ”

Now I don’t believe or advocate in the march of the thermometers but that does not stop me from being concerned about the details.

Zeke writes:

If we look at the difference between less than 10 pop density and over 100 pop density stations, you can see that apart from some weirdness around 1992, its mostly linear.

Well, you can fit a line to any data, right?

My claim amounted to the data being piece-wise linear–that is, that the data had a divergence point in trend. Attempting to fit the line is the same as rejecting my hypothesis a priori.

The model we’d like to falsify to disprove my argument is that the data is better represented by two segments for each series, one with near identical slope in period 1, and different slops in period 2. Your plot of the originals makes this point clearer, although it is visible in the original graphs too obviously: lets for the sake of simplicity say 1985. Although it might be interesting to do a formal change-point analysis.

BTW, I agree about the graphic suggestiveness, but that’s something that I thought about at the time, and I convinced myself that pinning the end of the series to a common point would have led me to the same conclusion. Still always the risk of the eyeballs going wrong.

Dave Andrews (Comment#37575)

In the real world science is never settled but in the virtual world of climate science the ‘science is always settled’ after its been through their riged peer review process and can never subsequently be questioned.

Carrick: Whether this is true is irrelevant to the argument (I’m not sure why you think it matters).

It is relevant because it progressively reduces the “latintudinal component” of the hemispheric trend, which is small anyway. E.g., the difference between the trend calculated over the bands 30-50N and >70N is negligible (CRUTEM3):

30-50:

http://www.appinsys.com/Global.....5-50N:30-2

>70N:

http://www.appinsys.com/Global.....55E%20%20x

The only latitudinal band which shows a noticeable difference from the hemispheric mean is the >70N band, which comprises only 12% of the area of the hemisphere. Most of any AGW signal will come from lower bands, which are subject to contamination.

Laura S.

It was you who triumphantly declared: “In the end, the proof is in the pudding. … The exact same analyses by Zeke, Tamino,, CCC, The Whiteboard etc. all suggest … ”

Triumphantly declared? My full paragraph is this:

In the end, the proof is in the pudding. We can ask. “Did the cold bias materialize?” It seems not– of if it did materialize, it appears small. The exact same analyses by Zeke, Tamino,, CCC, The Whiteboard etc. all suggest that any biases that might have been introduced by “the march of the thermometers” were quite small. So, while this post shows a potential issue that can arise when applying GISTemps method of computing surface temperatures, the situation where that potential actually creates a real bias does not appear to have occurred.

I hardly think my use of “seems”, “appears small” are indicators of a “triumphant declaration. And yes, I think that, not withstanding your rampant speculations, the fact that all analyses to data fail to tease out any effect of bias arising form the march of the thermometers suggests the bias is small.

Of course, further analyses may show otherwise–as I have said. But you seem to be pretty darn willing to fling around phrases like “It should be clear that there is a reasonable basis upon which to believe…” Words like “clear”, “should”, “reasonable”? In a conclusion you in a hypothetical you concoct based on the results on an analysis might be if one makes rather tenuous assumptions and someone bothered to run numbers.

Now I don’t believe or advocate in the march of the thermometers but that does not stop me from being concerned about the details.

No one is saying it should stop you from being concerned with the details.

Equally, it’s rather odd for an academic who must be familiar with conference proceedings, poster sessions, meetings with collegaues etc to suggest that someone like Zeke should not relate the conclusions one would make based on those analyses that have been done. It is even more odd to read you suggest that discussions of the existing results involving words like “suggest” and “seems” amounts to “uncritical amplification and endorsement of the half-done analysis.

’science is always settled’

Yes, (ahem), look at this squiggly line. We made it squiggle the way we wanted, so give us your money, and we’ll make it squiggle some more. We’re the experts at it and we don’t even need to tell you who we are. Just do what we say. Riiiiiiiiiight.

Andrew

Equally, it’s rather odd for an academic who must be familiar with conference proceedings, poster sessions, meetings with collegaues etc to suggest that someone like Zeke should not relate the conclusions one would make based on those analyses that have been done.

Again, no. That is not my position. I’ve read many half-done analysis from your colleagues Steve McIntrye and Jeff Id–and from you. I see nothing wrong with releasing working material and notes. Its great. Its gets people interested in your work/field, and it helps pull everyone’s cart along. I don’t complain. But when someone–a third party like yourself–says, X was demonstrated when it was not, that’s a different matter.

Zeke for his part, seems quite interested in understanding what aspects of his work I found unconvincing. Clearly he wants something more than half-done. Anyone familiar with publishing will understand that that is a great way to approach the matter. You’ve got to convince the reviewers that you’ve got your ducks in a row. So far his attitude has been great. Yours has not, but since my venom was directed you, I’ll take that as just being defensive.

So, you have your framing; I have mine. You wrote some words. You think they have a meaning. I think they have another. I suggested you clarify your meaning. You refused, saying my interpretation is wrong and odd. Well, there you have it. I hear Eugene Volokh has a nice book on writing, maybe you’ll read it one day and then have an epiphany. The light bulb will glow bright, and you’ll say oh yeah, I write this blog to change people’s minds; so I should be worried about what my words mean to them not what I think people should hear.

Thanks for listening. Have a nice day.

Laura,

But when someone–a third party like yourself–says, X was demonstrated when it was not, that’s a different matter.

Could you clarify something for me? What do you think I have claimed Zeke demonstrated?

I don’t know which words of mine you refer to when you write

“You wrote some words. You think they have a meaning. I think they have another. I suggested you clarify your meaning

Could you clarify?

Yours has not, but since my venom was directed you,

Oddly, until you posted a comment this morning, I had no idea that you had directed any venom toward me. In fact, your first comment on this thread (Laura S. (Comment#37252) March 9th, 2010 at 11:29 pm) appeared to list a litany of complaints against zeke. Thanks for clarifying that you did intend that to contain venom and that it was aimed at me.

I hear Eugene Volokh has a nice book on writing, maybe you’ll read it one day and then have an epiphany. The light bulb will glow bright, and you’ll say oh yeah, I write this blog to change people’s minds; so I should be worried about what my words mean to them not what I think people should hear.

Is that why I write my blog? With regard to the book, do you mean Academic legal writing? Or some other book?

Contraian:

It is relevant because it progressively reduces the “latintudinal component” of the hemispheric trend, which is small anyway. E.g., the difference between the trend calculated over the bands 30-50N and >70N is negligible (CRUTEM3):

You’re just wrong on this (this is using CRUTEMP3). A factor of 3 is not neglible.

See Zeke’s figure above too, which agrees with my calculation, using his own code.

Your links didn’t work.

… and as I mentioned, your argument was wrong anyway.

60° has a larger weight to global mean temperature than does 30°N, even before you fold in the in fact not negligible increase in temperature trend with latitude.

Laura,

Your comment that “The uncritical amplification and endorsement of the half-done analysis harms rather aids the pursuit of knowledge. I hope you will post an appropriately tempered update to this post.”

Followed shortly by:

“Now I don’t believe or advocate in the march of the thermometers but that does not stop me from being concerned about the details.”

Seems to be a rather ironing juxtaposition, considering that in many ways the “march of the thermometers” is a case-in-point of the uncritical amplification and endorsement of the half-done analysis.

Would I submit the modeling posts I made here to a journal in their current form? Certainly not! But I’m happy to explain my methods, post my code, show interesting results, and encourage others to improve the model and replicate it via their own analysis in the context of a blog discussing climate science. Think of it as a working paper or workshop in an academic setting, if you will.

I do share your concern that blog arguments are often cited by outsiders with an unwarranted level of certainty. E.M. Smith’s march of the thermometers arguement has appeared on cable TV, Fox News, and a number of local newspaper (in Canada, Orange County, etc) as well as political news sites (PJM, Newsbusters, American Thinker, etc.). I suspect my results won’t get similar coverage, even if I were to publish something in the peer reviewed literature to that end.

I don’t understand the objection of Laura S.

Zeke has made a couple blog posts. One is looking at UHI in various ways. Laura seems to object to its conclusions, but so far as I can tell, it doesn’t even have any clearly stated conclusions. It just says: I got my data here, and I processed it thus, and this is what you get. Since Zeke didn’t give a conclusion that I can see, my interpretation is: Looking at it different ways, it seems possible to detect a UHI signal, but it is relatively small compared to the overall warming that has been observed. Is some stone left unturned? Zeke’s been happy to look at the data in other ways, as suggested. Are there problems? Well, it seems that the population density source may need to be revisited; work in progress. Would it be better to place this in the context of previously published work, in the introduction or discussion? Absolutely.

Meanwhile, Laura (with much fanfare about being a careful academic) tells us that it’s all inconsistent with Spencer, and even gives a number to it (which itself is very much a work on progress). Upon inspection, this statement doesn’t seem to be well-grounded at all. Is this conceded by Laura S?

Meanwhile, Laura also seems bothered by Zeke’s post about the thermometer march, but hasn’t said exactly what’s wrong with it. If some calculation could be done in a different way, or interpreted in a different way, that could be suggested, but I’m not seeing such suggestions.

I am sorry, but after reading those eau e-mails I have no confidence with the loss of thermometers. If some one could do a phase change on the used location of thermometers it might put some light on this.
Question: after what has happened do YOU think if a cooling bias was added after dropping stations, would they have left it like that?
reported it? published it? “Hide the decline”, drop the northern stations would do the “trick”

Tim

Question: after what has happened do YOU think if a cooling bias was added after dropping stations, would they have left it like that?
reported it? published it? “Hide the decline”, drop the northern stations would do the “trick”

Tim

There has never been any evidence that actual temperature records have been manipulated deliberatlely. The “decline” was a well documented problem with proxies, which tree rings diverged with the actual temperature record after the 1960′s.

Carrick: “You’re just wrong on this (this is using CRUTEMP3). A factor of 3 is not neglible.”

Sorry about the links — they were very long, and probably truncated by the blog software. You can run the calcs here, using CRUTEMP (gridded or station), or GHCN, by latitude.

http://www.appinsys.com/GlobalWarming/climate.aspx

Even with your graph the differences between 30-50N and >70N are not “a factor of 3.” The difference between the means is about 0.15 degree. Also, I don’t know where you got the data >75N, since CRU has no stations above that latitude.

(Ok — Zeke’s plot uses GISStemp, not CRU. The diff is still only about 0.15C).

“60° has a larger weight to global mean temperature than does 30°N . . .”

Then it is mis-weighted. Or are you referring to greater weight from land temps at 60 v. 30? That, of course, is my point. Land temps are outliers, very likely because they are contaminated with non-AGW factors. The oceans, which do not suffer those complicating factors, comprise 70% of the surface. 70% is more than a large enough sample from which to draw global conclusions.

Oddly, until you posted a comment this morning, I had no idea that you had directed any venom toward me. In fact, your first comment on this thread (Laura S. (Comment#37252) March 9th, 2010 at 11:29 pm) appeared to list a litany of complaints against zeke. Thanks for clarifying that you did intend that to contain venom and that it was aimed at me.

I think if you reread it now, you’ll see that my criticism is about the claim that certain analysis are dispostive. My remarks against those analysis are presented as foundation. As I said then, “I’m confused to why you’ve chosen to uncritically repeat and amplify his claim.”

“Could you clarify something for me? What do you think I have claimed Zeke demonstrated? ”

“The exact same analyses by Zeke, Tamino,, CCC, The Whiteboard etc. all suggest that any biases that might have been introduced by “the march of the thermometers” were quite small. … the situation where that potential actually creates a real bias does not appear to have occurred.”

I agree that their results are consistent with there being negligible bias, but I disagree that their methods could distinguish a bias if it were present, and I say that even as I doubt that any bias exists.

To quote myself:
“We shouldn’t get into a love-fest just because someone posted some graphs–or even a few someones. These guys are too quick to post conclusions and spin a narrative. They don’t deserve so much uncritical praise.”

With regard to the book, do you mean Academic legal writing? Or some other book?

Yes, I’ve heard it has generally good advice for presenting clear and compelling arguments. Although not doubt it contains specialist material not generally relevant too.

Seems to be a rather ironing juxtaposition, considering that in many ways the “march of the thermometers” is a case-in-point of the uncritical amplification and endorsement of the half-done analysis.

Well, I’m a reader of this blog. Not theirs. So I care about what Lucia does, more than them…

I do share your concern that blog arguments are often cited by outsiders with an unwarranted level of certainty.

That’s a fair concern. “Chiefo” says he can present a clearer proof. I wait to see. Still waiting.

I don’t object to thinking out-loud! I enjoy Steve McIntyre’s posts, and they are rarely “complete”. But its nice to download his stuff and run it yourself. It was great fun to make sea ice plots for a while… and frankly is how I got started using R. Great stuff, but has he convinced me of much? not really.

Although he has done enough to cast doubt on certain things.

Laura S. (Comment#37866) March 13th, 2010 at 12:08 am

I’m confused with your writing. What test in your mind would prove
that dropping stations would not bias the estimation downward?

Laura–
Yes. I read that first comment, and reread it. I was unaware of any venom, nor that you were requesting a clarification. I did understand you to be suggesting my post was somehow intemperate and that I rewrite it. Butt precisely which bits of my post you consider intemperate or flawed remains a mystery.

I know you think you have a gripe, and I know you think you are writing clearly. But still have no idea what specifically you think I said that was incorrect. I don’t think anyone else understands either. But maybe you can clarify?

(I’ll admit that given previous comments, I suspect what you will discover is that we disagree about the temper of my post &etc. But right now, I can’t really say that because, I don’t know specifically what you think is wrong with what I wrote.)

Contrarian:

Even with your graph the differences between 30-50N and >70N are not “a factor of 3.” The difference between the means is about 0.15 degree. Also, I don’t know where you got the data >75N, since CRU has no stations above
that latitude.

Actually there are 3 stations above 80°N and 9 above 75°N in the CRUTEMP3 published station data (I used this older data set for my analyses) since 1960, with 8 of those 9 reporting data in 2009 and all providing data into the 2001-10 decade. There are also 2 more additional stations at or above 75°N if you round to the nearest degree. This of course leaves open the possibility there are more stations in the CRU full set reporting data in the interval of interest.

Politely, few of your claims seem to hold up very well when compared to the actual record.

(*coughs*)

In any case, the fact that there isn’t that much land mass up there (as you pointed out) that’s plenty of stations to measure any trend there.

Trend is a much better way to analyze any warming (including effect from UHI/land use changes) than visual inspection or absolute temperature or even local temperature anomaly. And in that respect, Zeke and I independently have done this trend analysis, with the effect of latitude is both systematic in nature—it monotonically increases—and substantial in magnitude .

Most of the trend at upper latitudes (> 60*) certainly does not come from UHI signal nor land use change—indeed it is easy to verify these high latitude stations are in wilderness-situated locations in Alaska, Canada, Greenland and Siberia, nonetheless a systematic trend remains.

Whether a station or group of stations adds much to the total global weighting though has nothing to do with whether these stations are in conflict with your original (and almost certainly false) claim that the trend originates in UHI and/or land-usage changes.

Your argument substantially ignores the reality that most of the UHI signal should be from more temperate zones:

If your attribution for the temperature trend were accurate, the “peak” in trend should be intermediate latitudes not the most northern ones (nor does your explanation provide any insight into why the trend should so systematically increase with latitude, nor by such a huge factor).

I really see no way one can argue out of the box you’ve placed yourself in here. When the analysis is done carefully and rigorously it points to latitude being a substantially more important effect than any effect observed either by Spencer or Zeke.

Laura S

I agree that their results are consistent with there being negligible bias, but I disagree that their methods could distinguish a bias if it were present, and I say that even as I doubt that any bias exists.

Instead of talking about talking, why don’t you just say why the method couldn’t detect a thermometer march bias, if it existed?

Laura S:

I agree that their results are consistent with there being negligible bias, but I disagree that their methods could distinguish a bias if it were present,

Nobody has any idea why you think this. Could you describe a mechanism by which they would miss a bias if it existed? Certainly there are biases that they are testing for and not finding, so it must be the case that a bias could exist in the data and still be detected using Zeke’s method.

Your strongest argument (as far as I could see) was potential flaws in the metadata. But since you haven’t provided any meaningful statistical analysis of this, or anything beyond a vague description, again clearly written or not, it doesn’t meet the most minimal requirements for scientific rigor.

They don’t deserve so much uncritical praise

You could directly quote us what you are talking about.

I don’t think you write nearly clearly enough as you need to, when it comes to the factual basis for your arguments. It’s pretty telling that the only person you quote is yourself. >.>

Carrot Eater:

Instead of talking about talking, why don’t you just say why the method couldn’t detect a thermometer march bias, if it existed?

Yes exactly, and in enough detail a “toy model” could replicate this flaw in Zeke’s analysis.

I just noticed my link got eaten:

CRU “ALL” data from Dec 2009. See this page for more detail and an updated data set (no additional stations above 75°N though).

Re: Laura S. (Mar 13 00:08),

Carricks response prompted me to re-read your response (37866). I still hold by my earlier response to that– but on rereading what you wrote, I conclude that you are in serious need of reading Eugene’s book and actually internalizing the advice contained therein!

I agree that their results are consistent with there being negligible bias, but I disagree that their methods could distinguish a bias if it were present, and I say that even as I doubt that any bias exists.

Maybe you should have just said that directly in your first comment?

Now that you have managed to spit out what you really meant: I think that most of your discussions touch on whether or not they can correctly identify the UHI effect. However, the very first test comparing pre/post march thermometers should reveal the effect of “the march” if it is large. This is true even if we can’t be confident of the UHI effect.

The later analyses of UHI are performed so try to see if — despite the fact that the bias cannot be detected by the simplest most obvious method– we can detect it some other way or– more importantly– if the UHI effect is biases the computed trends irrespective of “the march”.

So far, Zeke can’t detect the effect of UHI using known treatments for UHI. We can’t a bunch of other ways. But most of us agree that this is incomplete. Though you overstated what Spencer has found, most of us agree it would be interesting to see how including Spencers information when computing trends affects the assessement of UHI.

Are Zeke and Rons latter tests imperfect in figuring out whether UHI is a problem separate from the march of the thermometers? Sure. I share your reservations on that. But that has nothing to do with the rather tame — in fact tentative- conclusion I actually wrote about a different issue– what they show about “the march”. That conclusion uses words like “seems”, “appears” etc.

Echoing my previous response to the same comment: I still have no idea why you think there is any “love-fest” going on, nor why you think anything I wrote was ‘uncritical praise’. Maybe you think I am required to explain that I think any proof that the UHI effect is also small needed to be included? That’s a somewhat separate topic, and I shouldn’t think that criticizing that as incomplete in this post about the march of the thermometer effect was not required.

But, maybe that also isn’t what you thought I needed to do. As I said, I really don’t know.

lucia (Comment#37883) March 13th, 2010 at 7:39 am

I think that Laura has a problem communicating.

bugs,
As you say ‘hide the decline’ was a well documented problem. The trouble is they still carried on using the tree ring proxies to ‘prove’ AGW even though they did not understand what was happening with tree ring growth in the last 50 years.

It’s quite hard to believe that if they don’t have a clue about what is currently happening they can state with certainty what happened several hundreds of years ago.

I don’t think Zeke’s post on the March meme and his post on UHI should be seen as part of the same work. Between the two, we have different objectives, somewhat different methods, and very different level of conclusions.

So I think it makes sense to criticise them each separately.

But mostly, I’m bored about talking about talking. If we’re to be blessed by the analysis of a careful academic, let’s actually get an actual suggestion that Zeke look at something. The criticism so far on the UHI post, saying that Zeke and Spencer can already be seen to be inconsistent, turned out to be quite weak. Which is fine; we’re publicly brainstorming here and one will sometimes have a dead end.

Dave Andrews (Comment#37988) March 13th, 2010 at 2:58 pm

bugs,
As you say ‘hide the decline’ was a well documented problem. The trouble is they still carried on using the tree ring proxies to ‘prove’ AGW even though they did not understand what was happening with tree ring growth in the last 50 years.

It’s quite hard to believe that if they don’t have a clue about what is currently happening they can state with certainty what happened several hundreds of years ago.

They have reasonable evidence to suggest the proxies are valid for the majority of the measured temperature record. They are not used as the ‘proof’ of AGW, but as supporting evidence. It is the magnificent framing of McIntyre that turns the normal progress of research and evidence into a criminal conspiracy.

Carrick: “Actually there are 3 stations above 80°N and 9 above 75°N in the CRUTEMP3 published station data (I used this older data set for my analyses) since 1960, with 8 of those 9 reporting data in 2009 and all providing data into the 2001-10 decade.”

>75N was my mistake. It should have been >80N. There are 3 stations above 80, but one has not reported since 1954, another since ~1998, and the 3rd since ~2003. (These are all taken from the appinsys calculator. Can’t link URL to specific plots, ‘cuz the blog will trash them). Calculator is here:

http://www.appinsys.com/GlobalWarming/climate.aspx

“In any case, the fact that there isn’t that much land mass up there (as you pointed out) that’s plenty of stations to measure any trend there.”

You’re missing the point, which was that there is no need to consider land stations to estimate a global trend, especially if you’re looking for an AGW signal. The SST trend will suffice, and is less susceptible to confounding factors. Actually, the trend at high latitude (>70N) land stations closely tracks the SST at those latitudes, as you would expect, since all but 2 of them are islands or coastal stations. But at lower latitudes the confounding factors come into play.

“Trend is a much better way to analyze any warming (including effect from UHI/land use changes) than visual inspection or absolute temperature or even local temperature anomaly. And in that respect, Zeke and I independently have done this trend analysis, with the effect of latitude is both systematic in nature—it monotonically increases—and substantial in magnitude.”

No one is arguing the first part of that. The magnitude is the issue. My claim is that while the differences in magnitudes of the trends at different latitude bands look significant when separately plotted, they contribute declining weights to the hemispheric trend.

I tried to plot the 0-90N NH HadCRU trend against the 0-60N trend on the appinsys calculator, but it crashes with that many data points. Perhaps you or Zeke could plot them. I’m guessing that the difference (0-90N – 60-90N) will be small — perhaps 0.1 degree.

“Most of the trend at upper latitudes (> 60*) certainly does not come from UHI signal nor land use change . . .”

Agreed. But those don’t contribute much weight to the hemispheric trend either.

“Whether a station or group of stations adds much to the total global weighting though has nothing to do with whether these stations are in conflict with your original (and almost certainly false) claim that the trend originates in UHI and/or land-usage changes.”

My argument is that a portion of it does. And that that portion is roughly the difference between land and SST trends at the same latitudes. Nor will adjusting for UHI alone reconcile those trends, because UHI is not the only confounding factor in the land trends.

“If your attribution for the temperature trend were accurate, the “peak” in trend should be intermediate latitudes not the most northern ones (nor does your explanation provide any insight into why the trend should so systematically increase with latitude, nor by such a huge factor).”

It will be in the middle latitudes, once the latitudinal trend components are properly weighted.

Or so I claim.

“I’m guessing that the difference (0-90N – 60-90N) will be small — perhaps 0.1 degree.”

Difference in trend rise 1970-2009.

I find reading some of these comments exhausting. Let me propose a different explanation for you all to chew on: you all switched to rebutting my remarks per-se without attempting to comprehend what was said…

For instance, Carrick writes “It’s pretty telling that the only person you quote is yourself.” Which is definitely a bit of a dig. Take it from my perspective, I kept being told I was criticizing the dissemination of work-in-progress. I very directly each time said, “no”. But still went through a few iterations of that. Maybe you can see why I’d quote myself. Seriously.

Carrick also states: “But since you haven’t provided any meaningful statistical analysis of this, or anything beyond a vague description, again clearly written or not, it doesn’t meet the most minimal requirements for scientific rigor.”

I don’t think I’m obliged to walk you through it; you’re free to cry bullshit when I don’t convince you, but hey, that’s life. I can only show you the door. You’re the one that has to walk through it. The way to get through it is to do the leg work for yourself.

And I don’t have the time. So if you don’t want to listen any more fine that’s your choice. I work long hours; I just spent my Saturday evening grading midterms. I give what I can give, and prepare what I can prepare.

So lets cut to some substance:

Though you overstated what Spencer has found, most of us agree it would be interesting to see how including Spencers information when computing trends affects the assessement of UHI.

Well, lets write this down as another claim that you have not understood. You were quite adamant that Zeke and Spencer were not comparable. I never said Spencer computed trends, so I don’t think I overstated what he found. You assumed that–took a very narrow view on what could be compared; however, as I explained a few days ago now, if you assume that the Urban Stations of Today were Rural yesterday, then the rate of changes are irrelevant. You need only compare their temperatures at around 2005, the year for which Zeke’s population classification was defined to ascertain whether the two sets of results seemed consistent.

Now they fail that consistency test, which suggests two possibilities: 1) they are indeed inconsistent or 2) the assumption of the analysis was wrong.

I went on to explain why the assumption, while sounding extreme and excessive, in fact could be justified by Zeke’s own data as presented.

Zeke engaged this issue and suggested that I was seeing a graphical illusion. Yet the visual implication is sustained regardless of the baseline choice. Zeke then presented a more concrete counterargument based on the fit of a line to the difference between the two datasets. He wrote: “If we look at the difference between less than 10 pop density and over 100 pop density stations, you can see that apart from some weirdness around 1992, its mostly linear.”

I countered by saying that my claim amounted to saying that the residuals were better fitted by a piecewise linear function with an early period of slope zero. Since Zeke did not compare those two–he merely presented the linear data–I was not moved.

He’s gone silent on this point now. But carrot eater seems to think my claim was demolished. People here sure are moved by the mere presentation of graphs, but it would be nice if those graphs actually served the ends to which they are applied.

I don’t want to get too worked up about whether this point of concern is correct–perhaps the objection is overturned, what’s surprising is how hard I had to fight just to get consideration of the point–Zeke gracefully being the exception, although he seems to have disengaged once I made a concrete proposal.

That’s an element of why I think you’re being uncritical. Your focus is on what the analysis appears to demonstrate not why it might not be convincing. That’s a choice of focus, and its choice to not criticize.

Okay, lets step back now to the meta-question: what about the march of thermometers. You cite four different analysis. Ron’s analysis works like this: Watts et al claim that certain types of stations are dropping out, lets partition the data according to those categories and determine whether they have different trends. He then proceeds to compare the partitions and shows that the partitions are basically indistinguishable. There is a real subtle foundational problem here, which is whether the datasets have really been partitioned. Ron just assumes that his input data can be partitioned. The difficulty here is that whether we’re discussing NOAA’s SHAP and FILNET or Menne’09 or GISTemp Step 2, they all basically do the same thing which is blend distant stations together. Ron has the right idea, but he inadvertently uses blended data from the get-go. So his failure to find a difference is not surprising and not really evidence.

Zeke’s case is different. He compares trends using Watt’s station-siting data. He looks at ‘raw’ data–no blending, that’s good. One bit of trouble though is that the ‘raw’ data does show an effect on trends due to station-siting. So this does not disprove the march-of-thermometer’s claim… but the adjustments do have a purpose, so I don’t think its fair to present it as evidence in favor of Watts et al.

The more interesting part comes when Zeke fails to find a difference in the adjusted data, but again this adjusted data is blended. It cannot be partitioned into CRN12 and CRN345 in a meaningful way after the fact. One must go back to the raw- data, recreate the USHCN adjustment procedures AFTER partitioning. From what I gather this is not what was done.

Tamino takes the approach of looking GHCN raw data, and presents us the truncated curves. Tamino’s method cannot disprove the march of thermometers because it argues by counterfactual–we don’t know what those missing years would have shown.

So three of the four methods don’t even suggest anything, and their multitude is also not meaningful. Three answers found in agreement using three wrong methods is still three meaningless answers. I’m having flashbacks to MBH!

I must confess, I did not read the fourth analysis.

LauraS
Now that you’ve fleshed out your argument, I think that with respect to my conclusions limited specifically to the march of the thermometer, and involving words like “seems” and “appears” your concerns are valid to an extent but are no where near sufficient enough to support a strong claim like this:

So three of the four methods don’t even suggest anything,

.

“Most of the trend at upper latitudes (> 60*) certainly does not come from UHI signal nor land use change . . .”

Here is the history of the point barrow weather station
http://climate.gi.alaska.edu/h.....arrow.html
“Site continuity reasonable, but changes in summer exposure to ocean, and albedo changes due to growth of town, moves, snow clearance, dirt on snow, etc. could result in false long-term trends.”

Population of Barrow
1940 400
1960 1,300
1970 2,100
1980 2,200
1990 3,500
2000 4,683

Alaska Airlines has twice daily 737 service to Barrow.
Barrow has all the modern conveniences, city water and sewer, elementary,middle and high school, two year community college,
fire department, health clinic. A 6500ft x 150ft paved runway complete with state of the art Instrument Landing System.

Oh look, they’ve got their own power plant as well.
http://www.bueci.org/

Laura S. (Comment#38023),
Thanks for that (rather long) explanation.
I agree that any use of blended data makes any subsequent analysis of UHI effects extremely doubtful. (In fact, I really do not understand how temperature blending of any kind can be justified, except perhaps to in-fill some missing station data.) I also agree that the gradual change of temperature stations from more rural to more urban will inevitably lead to UHI biases that can’t be identified by any of the kinds of trend analyses I have seen so far. Roy Spencer’s preliminary work does appear to show a strong effect of population density on measured temperatures at stations that are reasonably close to each other, so perhaps incorporation of a raw temperature correction for the population trend at each station, prior to any other adjustments, would resolve the UHI question to everyone’s satisfaction.

harrywr2, Certainly compared to more temperate stations, and a modern population of only 4000, I’d worry less about UHI than I would about it’s proximity to the coast line.

Contrarian:

>75N was my mistake. It should have been >80N. There are 3 stations above 80, but one has not reported since 1954, another since ~1998, and the 3rd since ~2003. (These are all taken from the appinsys calculator. Can’t link URL to specific plots, ‘cuz the blog will trash them). Calculator is here:

In the CRU data set there are 3 three stations that have reported into this decade, one of which the most recent update was 2007, the other two have reported through 2009.

I’m not as interested in stations at a particular latitude as a I am in the overall trend south-to-north. I’m more interested in inland stations than I am island/coastal stations (to me, they should be lumped into a third category). I agree with you that the weight becomes less (after a certain point), but I believe you will have to admit that there is more weight to northern stations than you had anticipated.

You’re missing the point, which was that there is no need to consider land stations to estimate a global trend, especially if you’re looking for an AGW signal. The SST trend will suffice, and is less susceptible to confounding factors. Actually, the trend at high latitude (>70N) land stations closely tracks the SST at those latitudes, as you would expect, since all but 2 of them are islands or coastal stations

Firstly, this statement is false, at least with respect to the metric I’ve been using, which is temperature trend.

Comparison of HADSST2 to CRUTEM3

Secondly I am interested in the land precisely because I am interested in the the effects of potential contamination from human activities.

It will be in the middle latitudes, once the latitudinal trend components are properly weighted.

Or so I claim.

As I mentioned there are three factors that determine the weight.

• One is the temperature trend (increases as you go N, tending to weight higher latitudes more.

• A second is the cos(latitude) band, which over a finite band of latitudes, is given by Sin(lat_max)- Sin(lat_min). This decreases as you go N, tending to weigh equator more.

• A third is given by the increase in the fraction of land area to ocean ratio. Let’s call this Fraction_land(latitude). This also increases as you go N.

See this figure (blue line)

So what you want to compute is:

T_trend(latitude) * (Sin(latitude_max) – Sin(latitude_min)) * Fraction_land(latitude)

Latitude Dependence of Land Temperature Trend Contribution [°C/century]

The sum of the value of these points is the (correctly weight) global temperature trend for land surface area, which turns out to be 1.89°C/century.

Oh yea, the peak value is at 65°N

Laura S:

I don’t think I’m obliged to walk you through it; you’re free to cry bullshit when I don’t convince you, but hey, that’s life. I can only show you the door. You’re the one that has to walk through it. The way to get through it is to do the leg work for yourself.

I’ll call “FUD” instead (fear, uncertainty, doubt).

It’s clear to me you haven’t done any leg work yourself so I’m interpreting this as you just trying to undermine a conclusion you don’t particularly like.

I’m a bright guy, I can work out issues on my own, I don’t need you to show me the way to any door.

(That said, if you provide us with an explicit model for bias that Zeke’s analysis would have missed, I’m sure somebody here will go back and test it. I don’t put any stock in rhetoric, but unfortunately rhetoric and unsubstantiated claims is mostly all you’ve provided at this point.)

SteveF:

I agree that any use of blended data makes any subsequent analysis of UHI effects extremely doubtful.

Again a model that characterizes the effect of blended data on “any subsequent analysis of UHI effects” would be helpful.

I find it baffling that you could describe as “extremely doubtful” any analysis that uses blended data without even a back-of-the-envelope calculation to demonstrate its potential relevance.

How does a vague argument translate into anything other than “we should model the effect of blended data and characterize the effect of it on UHI.”

My prediction: Blended versus not-blended will show negligible effect (in the statistical sense) on global mean temperature trend.

Carrick (Comment#38036) March 14th, 2010 at 9:45 am

“harrywr2, Certainly compared to more temperate stations, and a modern population of only 4000, I’d worry less about UHI than I would about it’s proximity to the coast line.”

If the 4,000 population was at a middle latitude I wouldn’t even be bothered.

The 4,000 population plus a 6500′ paved, black, plowed runway significantly changes the albedo 9 months per year in the vicinity of the thermometer from pure white to black.

The same goes for the thermometer in Dead horse.

The growth of both areas was fostered by the discovery of oil in Prudoe Bay, which got going in the 1970′s. So we’ve gone from remote communities without any modern facilities to small towns with all the modern conveniences including airports capable of handling modern jet aircraft. The ‘airstrip’ in Point Barrow has gone from a 3000 foot ‘semi-improved’ airstrip to a Class I commercial airport runway. Same is true for deadhorse.

The temperature of the water surely has an effect.

So let’s go to the thermometer in fairbanks, history provide by the University of Alaska at fairbanks.

http://climate.gi.alaska.edu/h.....banks.html

“No missing data after 1929 and all observations at midnight (11 pm solar time after October 1983 due to change in time zones.) Potential major problems from site continuity and heat island effects.”

One of the problems with even the ‘remote’ thermometers in Alaska is that the Oil Royalties provided the state with a lot of cash to build a lot of infrastructure inconsistent with community size.

There are 27 Class I – Commercial Jet Capable airports in Alaska serving a population of 700,000. There are only 26 cities with populations above 1,000. Only 3 with populations above 10,000.

The runway at Point Barrow, Alaska(pop 4000) is longer then the Runway at Congonhas Airport, Sao Paolo, Brazil(pop 10+ million)

http://www.faa.gov/airports/al.....p_2007.pdf

http://en.wikipedia.org/wiki/L.....population

As far as I can see, we don’t have any ‘rural’ thermometers in Alaska, we have thermometer’s measuring the temperature trends at 6,000+ ft paved and plowed runways. Which have dramatically different Albedo then Alaska as a whole.

Carrick:
A second is the cos(latitude) band, which over a finite band of latitudes, is given by Sin(lat_max)- Sin(lat_min). This decreases as you go N, tending to weigh equator more.

What was the underlying reasoning for this weighting? It would actually be consistent with a ‘weighted’ mean of temperature anomaly that I suggested in response to DeWitt’s comment on the other thread wherein the equator had less emission than absorption and the opposite was true in the poles.

RB:

What was the underlying reasoning for this weighting? It would actually be consistent with a ‘weighted’ mean of temperature anomaly that I suggested in response to DeWitt’s comment on the other thread wherein the equator had less emission than absorption and the opposite was true in the poles.

It’s a simple area weighting.

On a sphere, if you have a band of constant latitude θ going from θ–Δθ/2 to θ+Δθ/2, the area is given by

2πR^2 * (Sin(θ+Δθ/2) – Sin(θ–Δθ/2)) = 4πR^2 Cos(θ) Sin(Δθ/2),

4πR^2 being the area of the Earth of course. If you are splitting up the Earth into N bands (which can be overlapping), you can simplify this to:

W_n = W_norm [Sin(θ_n+Δθ/2) – Sin(θ_n–Δθ/2], n = 1, …, N

where W_norm is given by the requirement that the sum over the W_n = 1.

Harryw2, thanks for the comments.

I guess my issue here is that all data has warts and rough edges, and it’s not enough to show that potential issues exist with data, but that the issues cause an observable deviation in the result.

However, in the case of a very small community, like Barrow, AK, the instrument error of the local microclime is very small (maybe hundredths of a degree C). To start, they don’t “bunch up” like they do in the lower 48, so even in town you have decent wind “fetch”, since houses and other structures have plenty of empty space around them (nothing like an urban canyon I assure you).

So even had the facility been placed inside of town, this wouldn’t have been much of an issue. But it’s not.

The Barrow station is actually a first rate facility with 10-m and 40-m towers located about 3 1/2 miles to the NE of Barrow, AK (the location is near 71° 19′ 23.73″ N, 156° 36′ 56.70″ W).

Since dominant wind direction is either to shore or away from shore, there is no issue with UHI effects being efficiently communicated from town to the measurement site. From a met standpoint the station is as ideal and first class a facility as one could ever hope for.

As I mentioned, I worry a bit about the costal location for what I want to use it for, but that was Contrarian that was pushing sites this far North. I am more interested in sites around 65*N, which corresponds to the maximum contribution to global mean temperature from land sites.

There is simply no way to explain that as “just” or even “measurably” as a UHI/land usage effect. You might as well link to Alice in Wonderland if you re going to try such an argument.

Carrick (Comment#38044),
.
The blending process automatically introduces doubt about the UHI effect, since stations of differing “urban development histories” will be blended. I think it completely reasonable to think that blending of raw station data may significantly diminish (or even hide altogether) a UHI effect. In addition, I see absolutely no advantage to blending; it only removes potentially valuable information form the record. Why do it at all?
.
I agree that the global average temperature trend will not likely be changed very much by blending of land stations, since land represents only ~ 30% of the contribution to the average. Still, the more rigorous approach is to not blend station data.

Re: Carrick (Comment#38064) March 14th, 2010 at 6:23 pm

I guess my issue here is that all data has warts and rough edges, and it’s not enough to show that potential issues exist with data, but that the issues cause an observable deviation in the result.

I would have thought that the onus was on producers/users of the data to identify these issues and show that they do NOT cause deviation in the result.

Alex Heyworth:

I would have thought that the onus was on producers/users of the data to identify these issues and show that they do NOT cause deviation in the result.

Not precisely.

They have responsibility for due diligence, but beyond that, if you raise an objection, it’s your responsibility to show that it’s a reasonable objection that can materially affect their results or conclusions. If you fail to do that, they owe you nothing.

Of course. But there is sometimes a judgment call as to whether an issue should have been canvassed in the due diligence. If the data does have “warts and rough edges”, maybe these should have been identified and either dealt with or shown to be non-issues when the data was compiled.

Still, if issues have been missed, I agree that there is no choice but to show that they should have been addressed. Not really my point, which was that perhaps they could/should have been identified earlier. Perhaps I didn’t make this as clear as I could have.

Carrick,

“Actually, the trend at high latitude (>70N) land stations closely tracks the SST at those latitudes, as you would expect, since all but 2 of them are islands or coastal stations.”

Firstly, this statement is false, at least with respect to the metric I’ve been using, which is temperature trend.

Here are the 1970-2009 trends per appinsys:

SST:

http://www.freespokane.net/wp-.....st7090.jpg

CRUTEM:

http://www.freespokane.net/wp-.....ru7090.jpg

The 3.5C step increase in CRUTEM in 2005 is mysterious.

Secondly I am interested in the land precisely because I am interested in the the effects of potential contamination from human activities.

Nothing wrong with that. But one should not factor in the land temps to estimate the AGW signal until those sources of contamination are quantified and corrected.

The sum of the value of these points is the (correctly weight) global temperature trend for land surface area, which turns out to be 1.89°C/century.

Computed over what interval? 1970-2009? The trend for SST for that interval is ~1.5C/century. The difference is probably the contamination you want investigate.

Re: SteveF (Mar 14 19:14),

I agree that the global average temperature trend will not likely be changed very much by blending of land stations, since land represents only ~ 30% of the contribution to the average. Still, the more rigorous approach is to not blend station data.

Do you have the specific reference that explains how blending is done by any group in particular? Hansen and Lebedeff (1987) does not describe anything I would associate with the word “blending”. They describe very minimal processing of raw data to detect wild outliers (like obviously displaced decimal points –200.3C when the temperature must have been something more like 20.3C), and then apply a method that does not fill in or blend anything. That method just works off the data they have.

They admit those results could be contaminated by UHI because they do nothing to deal with it at that time.

Was that idea of “blending” introduced later? I always like to read the specific description in a paper before hunting for the implementation in a code. So, if I could read the paper, that would clarify the issue for me.

Re: Alex Heyworth (Mar 14 21:20),

I would have thought that the onus was on producers/users of the data to identify these issues and show that they do NOT cause deviation in the result.

The producers of the datas job is to formulate protocols for collecting data and collect it, do reasonable QA and supply it. (They may not even the the ones to formulate the protocols.) They protocols are laid out prior to collecting the data with the intention of measuring whatever it is people funding the data collection think they want the data to reveal. In 1930, the goal was not to detect the climate change signal– so the protocols did not reflect that mission and concerns like UHI were ranged from utterly unimportant to mildly unimportant to those funding data collection. (The airport operators actually prefer temperatures at the airport. Many consumers of daily weather reports don’t care. Weather forecasters might care, but the UHI effect might be the least of the problems when trying to predict rain next tuesday.)

Move forward in time: An agency like GHCN, NCDC will be funded to *compile* the data that exists. They will compile it even if some end users might wish the data that was collected in 1880-now had warts.

Meanwhile, someone like Hansen wants to use the data that we have to learn something. He has to use it as it presents itself. If you read papers over time, there is an attempt to make a first cut– admitting issues. Document that. Start trying to address the issues, see how that changes. Document that.

So, yes, those creating products like GISSTemp have a burden of ‘proving’ that an issue does not affect the result, but the process permits them to present their current best effort– while mentioning that there may be issues like UHI. Heck, they might even overlook issues that no one recognizes at the time.

It is simply not the case that every possible flaw in an analysis must be resolve prior to publication in a peer reviewed article. This has never been the standard. If this was the standard, nothing would be published. Current results would not be shared and no one could move forwards for lack of information about what was already known.

(FWIW: There can be issues where reviewers inclination to believe the preliminary results affects their inclination to needle others about their failure to address an issue. But, that does not mean it really is the standard that one must resolve every possible objection by every possible person before publishing.)

Contrarian:

Computed over what interval? 1970-2009? The trend for SST for that interval is ~1.5C/century. The difference is probably the contamination you want investigate.

1960-2009 as documented in the figure, but it’s land only. The difference is almost certainly not “just the contamination I want to investigate” because it does not vary anywhere approaching the systematic trend one would expect from UHI/land use change, moreover I believe it is to be expected from ABL physics.

Anyway, you are still mistakenly believing that oceans and land will have the same temperature trends. You are wrong, wrong and (did I mention) wrong about that. Land should be larger, again that’s just the physics of the ABL at work.

Regarding your plot with no trend computed (/dig), I’m assuming you are plotting a non-area weighted average?

You have to use area-weighting if you’re going to get an accurate average with a nonuniform sampling. Again, that’s “just math.”

Alex Heyworth:

Still, if issues have been missed, I agree that there is no choice but to show that they should have been addressed. Not really my point, which was that perhaps they could/should have been identified earlier. Perhaps I didn’t make this as clear as I could have.

In this case, many of the problems that people are “needling” them about have been addressed by the authors, and in their opinion, fully addressed.

There is nothing “new” about area versus non-area weighting, UHI index, blending versus not blending and so forth. All of this has been around and discussed since the 1980s. Ironically many of the critics of these methods (e.g., Watts) himself appears to have not done do diligence because he makes a number of charges in that SPPI document that he did not back up at the time.

If you’re going to get angry about poor scholarship, start with him.

The only thing remotely fitting the description of ‘blending’ done by GISTEMP is in the UHI adjustment, which is the only adjustment they make.

They identify urban and rural stations by night brightness. For each urban station, they find the trends shown over time by a composite of the neighboring rural stations, weighted by distance. They then use a crude procedure to impose that trend upon the urban station. One breakpoint in trend is allowed, in case the urban station only diverged after 1950 or 1970 or whatever.

Via the USHCN, GISTEMP does import for US stations some of the homogenisation, which is based on comparing stations to their neighbors. This could also be called ‘blending’, I suppose. The USHCN also attempts to infill missing data points, but GISS removes those.

Discussions would go more smoothly if better precision were used in the terminology. I understand that it can take time to learn the specific jargon of a field, but if you’re going to make up your own jargon, make it perfectly clear what you’re talking about.

Barrow is a poster child for UHI

http://www.geography.uc.edu/~kenhinke/uhi/

carrot eater (Comment#38098) March 15th, 2010 at 10:04 am

Agreed. At some point ( we’ve said this for 2 years now ) a data flow diagram would be nice. People through around the words
blending, adjustment, smearing, infilling without paying too
much attention to what is actually being done.

The other thing I note is that people who want to duplicate
GISS using other methods do not drop the stations that Hansen
drops in “strange.py” Funny, that little step ( which is marginal
at best ) is the thing that got me going. Stations dropped ( by hand) because they are “strange” Crater Lake is one of them.

I did notice one neat thing in the barrow study I linked above and that was a comparision of 3 ways of averaging Inverse weighting, kridging
and delauney triangulation.

That would be a cool topic

Carrot Eater:

Via the USHCN, GISTEMP does import for US stations some of the homogenisation, which is based on comparing stations to their neighbors. This could also be called ‘blending’, I suppose. The USHCN also attempts to infill missing data points, but GISS removes those.

To be fair, I think they are referring to Zeke comparing raw versus adjusted USHCN data.

Steven Mosher:

Barrow is a poster child for UHI

Maybe, but it has nothing to do with the climate monitoring system, which isn’t located in town.

Oddly, your source does not seem to realize you need to compare sites that are equal-distance from coast line. They are comparing inland stations to points in town but near the coast line and assuming the difference is UHI related rather than maritime ABL.

Carrick (Comment#38095) March 15th, 2010 at 8:46 am

Alex Heyworth:

Still, if issues have been missed, I agree that there is no choice but to show that they should have been addressed. Not really my point, which was that perhaps they could/should have been identified earlier. Perhaps I didn’t make this as clear as I could have.

In this case, many of the problems that people are “needling” them about have been addressed by the authors, and in their opinion, fully addressed.

There is nothing “new” about area versus non-area weighting, UHI index, blending versus not blending and so forth. All of this has been around and discussed since the 1980s. Ironically many of the critics of these methods (e.g., Watts) himself appears to have not done do diligence because he makes a number of charges in that SPPI document that he did not back up at the time.

A ‘number of charges’? He accuses the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science. For which he provides no actual evidence.

Carrick:

Zeke has made a point of using raw data whenever and whereever somebody might wish him to do so. His UHI post was using raw data. Though if he honed in on the US, showing raw, TOB, final NCDC adjusted, and final GISS adjusted would be appropriate. The raw would show the raw difference between datasets; the TOB would show how much of the difference is due to TOB (per Peterson 2003, maybe more than you’d think), the NCDC adjusted would show how much difference remains and in which direction the gap is closed, and GISS adjusted shows what’s left after GISS basically whacks the trend off any urban stations.

Looks like SteveF is referring to the GISS UHI correction when he later discusses ‘blending’. Maybe. Seems like it.

Laura,

Ron has the right idea, but he inadvertently uses blended data from the get-go. So his failure to find a difference is not surprising and not really evidence.

Ron was showing what happens if you do exactly what GISS does. If you do use GISS’s UHI adjustment, then there is no significant bias introduced by using the smaller post-1992 subset of stations. If you don’t use any homogenisation at all, you still see practically no bias (Tamino, zeke). Either way you do it, same result, and it was useful to see it either way, so that we now know that. Perhaps you didn’t see it, but Zeke and Tamino were criticised for not doing exactly what GISS did; hence Ron and CCC.

As for your initial careful academic claim of inconsistency between Zeke’s initial UHI results and Spencer, yes, I think it’s extremely weak. First, it requires an uncritical retrieval of a number from this mess, without using confidence intervals:

http://www.drroyspencer.com/wp.....non-US.jpg

And even these graphs don’t really show how much scatter there actually is in the data. Please, put some uncertainty bounds on the numbers you extract from this.

Then, the math of your careful academic claim actually requires that every single urban station be rural within the recent past. If it’s actually just a fraction, your claim of inconsistency falls apart.

Finally, you’re interpreting a breakpoint in the the difference series; this breakpoint is not apparent to me. Try and see if there’s a statistically significant change in trend there.

As it happens, the biggest issue with Zeke’s UHI work-in-progress was found by Harry, as he found problems with the metadata on the population density. I don’t know whether Harryw2 is an academic, but I appreciate his carefulness.

Carrick (Comment#38095)

In this case, many of the problems that people are “needling” them about have been addressed by the authors, and in their opinion, fully addressed.

I agree.

There is nothing “new” about area versus non-area weighting, UHI index, blending versus not blending and so forth. All of this has been around and discussed since the 1980s.

Again, I agree. Nevertheless, it is worth revisiting these issues sometimes. Zeke’s efforts here have not been wasted.

Ironically many of the critics of these methods (e.g., Watts) himself appears to have not done do diligence because he makes a number of charges in that SPPI document that he did not back up at the time.

Comments by Lucia (maybe not on this thread) argued that D’Aleo, not Watts, was responsible for this part of the joint paper. Still, I agree someone didn’t do their homework.

If you’re going to get angry about poor scholarship, start with him.

I rarely get angry about anything these days, least of all poor scholarship. If anyone got angry about that, they’d be permanently angry!

lucia (Comment#38088)

I agree with what you say, but the last two paras were a bit off topic given that we are talking about an ongoing product.

However, those last paras bring to mind one of the causes of the recent storm in a teacup about both GISStemp and HadCRUT. Projects that started out more as one-offs for academic papers became ongoing tasks. The QC that an ongoing task needs wasn’t done initially, because it wasn’t required at that point (and presumably the resources weren’t available). If you start off that way, you will always be playing catch up. It looks at this stage as though this hasn’t had too much impact, fortunately.

It would be great to have some of the principals drop by and discuss what they see in Zeke’s and others’ work. Lucia has shown herself to be pretty open-minded, and seems willing to see that voices from different quarters can be heard.

Chiefio got an invitation, which regrettably he declined. He subsequently wrote an explanatory post on his site–but it was way too long for me to finish. Zeke managed to show a ton of analyses in a quarter as many text-inches. Maybe Cheifio will reconsider?

Concise comments highlighting what they see as the most important issues that Zeke has covered would be hugely welcome from Messers. Watts, D’Aleo, Eschenbach, or others.

Right now, these folks don’t seem to be on the field. I am unsure how to interpret that.

Comments by Lucia (maybe not on this thread) argued that D’Aleo, not Watts, was responsible for this part of the joint paper. Still, I agree someone didn’t do their homework.

Are you saying Watts didn’t even read his own paper? He has had plenty of time to retract.

Traveling for business at the moment, so my blogging is a tad limited. As Carrot mentioned, the major issue in GHCN at the moment is the quality of the metadata (e.g. some stations not being where their lat/lon indicates; see Ron’s post) and the GWPv3 pop density data not being particularly granular outside the U.S.

However, since the U.S. has both good metadata and pop density data (as well as very good and dense station coverage), I’ll do a followup at some point this week looking at UHI in the USA, with some added analysis looking at the change in pop density at each station over the last 15 years and any relationship it shows to station trends (its noisy as all hell so far…).

I’m also waiting to hear back from some NOAA satellite folks on how possible it would be to use the 18-year history of satellite nightlight images to measure changes in urbanity over time.
.
AMac,

Anthony and Steve Goddard did a UHI post last week looking at Boulder and Ft. Collins stations that I think was at least somewhat in response to my post. That said, Eschenbach took some issue with their approach in the comments.

bugs (Comment#38118)

Are you saying Watts didn’t even read his own paper? He has had plenty of time to retract.

He can’t retract what he didn’t write. Anyway, maybe he still agrees with what d’Aleo wrote. Who knows?

Besides, maybe E. M. Smith will surprise us all and come up with something of substance, although I do hear the flapping of pig wings.

The bottom line: just because you (or me, or even everyone else in the world) disagree with what the SPPI paper said, doesn’t mean that the authors are compelled to retract.

He can’t retract what he didn’t write.

Yes he can. His name is on it; he has helped publicise it. Regardless of whether he wrote the passages in question or his co-author, he can acknowledge the passages are in error.

Besides, maybe E. M. Smith will surprise us all and come up with something of substance,

Aren’t you supposed to do the analysis before you go around accusing people of fraud, not after?

The bottom line: just because you (or me, or even everyone else in the world) disagree with what the SPPI paper said, doesn’t mean that the authors are compelled to retract.

If they have zero shame, then that’s where it ends. The rest of us can continue to press them on it.

Zeke

Anthony and Steve Goddard did a UHI post last week looking at Boulder and Ft. Collins stations that I think was at least somewhat in response to my post.

Don’t flatter yourself. At least, I don’t think so. Individual shows of possible UHI are fairly routine over there.

Besides, it’s the station drop post of yours that is relevant here. Your UHI post should be considered a work in progress, and is not directly relevant to refuting the thermometer march meme.

Amac

He subsequently wrote an explanatory post on his site–but it was way too long for me to finish.

I got through a good portion of it. It explained nothing. He’s acting as if the SPPI report didn’t exist, and that Tamino refuted claims that were never made. He did not seem to understand what Tamino did; he did not even acknowledge Zeke, ccc or Broberg, and he still does not seem to understand GISTEMP. He is pressing forward with his version of the FDM.

Re carrot eater (Comment#38122)

I don’t recall the exact figure, but a reasonably recent study found an incredibly high proportion of academic papers had been refuted within (I think it was five) years. Yet retractions are rare. Most researchers seem to just get on with the game.

Another question is whether it is relevant to apply academic journal standards to a paper published by a lobby organization. I don’t see a lot of calls for retractions by the authors of papers published by the WWF or Greenpeace, though I am sure some of theirs are equally contentious.

Anyway, I have said all I am going to say on this red herring. You and bugs and carrick are welcome to have the last word if you wish.

I’m not as militant or malicious as bugs is about this, but I agree with him on this general point.

If your name is on a document entitled “Surface Temperature Records: Policy Driven Deception?” you pretty much have to be aware of what the tenor of the document was. And you are always responsible and accountable for the content of anything your name gets applied to.

That said, I think it’s extremely dishonest to reword what other people say to make it sharper than it really is. E.g.,

A ‘number of charges’? He accuses the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science.

Instead of taking the risk of misrepresenting Watts (a polite way of saying bugs outright lied about what Watts actually said), bugs should provide direct quotes of the passages in question.

Also, if you want to advocate for a position, displaying honesty and integrity is the best way to persuade to your view point.

I did a word search using Acrobat. As far as I can tell, the word fraud does not appear in the SPPI document coauthored by Watts. I haven’t checked for “unprecedented” or “conspiracy”. But I’m pretty sure bugs would find it impossible to support his claim about what Anthony said using quotes from the SPPI document.

Around 1990, NOAA began weeding out more than three-quarters of the climate measuring stations around the world. They may have been working under the auspices of the World Meteorological Organization (WMO). It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler.

It’s highlighted in it’s own box on page 6 of the pdf.

http://scienceandpublicpolicy......e_temp.pdf

Specificially.

NOAA and the WMO deliberately removed temperature readings that showed the planet was not warming. For two organisations to work together to do this with the number of people involved would required a conspiracy of hundreds.

It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler.

I don’t know how else you can interpret that claim. It’s an accusation of fraud and conspiracy.

This (bugs words)

He accuses the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science.

Is not the same as this (D’Aleo and Watts):

It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler.

What D’Aleo and Watts wrote was idiotic, to be honest, because by removing the higher-altitude and higher latitude stations they would have introduced a cooler temperature trend. Even their other argument about UHI has been pretty much discredited in studies and doesn’t even survive back-of-envelope “sanity tests”.

The origin words are arguably more damning (of Watts that is) than what bugs wrote, but in any case neither D’Aleo nor Watts accused anybody of “fraud”, “conspiracy” nor other act “unprecedented in the history of science”.

Both fraud and conspiracy imply a crime, one should always be careful about making claims of either of these, and especially about claiming that other people had said such things when they haven’t used such words. Using over the top language in a case like this just undermines your own credibility and the case you are trying to make.

There simply is no reason other than sloppiness or intellectual dishonest to not link them in a case like this.

Carrick

Read the first paragraph.

“Recent revelations from the Climategate1 emails, originating from the Climatic Research Unit at the University of East Anglia showed how all the data centers, most notably NOAA and NASA, conspired in the manipulation of global temperature records to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.”

This is an accusation of fraud and conspiracy.

Why do you have to use the word ‘fraud’ to make an accusation of fraud? That’s really poor logic.

Carrick: “Regarding your plot with no trend computed (/dig), I’m assuming you are plotting a non-area weighted average?”

Yes. Those are raw averages of the selected stations. That is what the appinsys calculator offers (it also does not compute linear trends, which is a shame). With geographical weighting both trends would be shallower, since the higher-latitude stations would be discounted.

I have not downloaded the various databases and scripts (like Zeke’s) and run the calcs myself. Since no one has yet run a weighted trend of all stations 0-90N and minus 60-90N, I might have to tackle that chore.

“Land should be larger, again that’s just the physics of the ABL at work.”

You think a difference should remain per ABL even using the anomaly method?

Nathan, what bugs said was:

He accuses the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science.

Let’s look at the word fraud, courtesy Wikipedia:

fraud is an intentional deception made for personal gain or to damage another individual; the related adjective is fraudulent. The specific legal definition varies by legal jurisdiction. Fraud is a crime, and also a civil law violation. Defrauding people or entities of money or valuables is a common purpose of fraud, but there have also been fraudulent “discoveries”, e.g. in science, to gain prestige rather than immediate monetary gain.

Does either your passage or bugs meet this standard? Interesting question, but now, let’s see what I actually said that was particularly germane to issue:

Instead of taking the risk of misrepresenting Watts (a polite way of saying bugs outright lied about what Watts actually said), bugs should provide direct quotes of the passages in question.

and

Both fraud and conspiracy imply a crime, one should always be careful about making claims of either of these, and especially about claiming that other people had said such things when they haven’t used such words. Using over the top language in a case like this just undermines your own credibility and the case you are trying to make.

The relevant issue isn’t whether one could reasonably construe W&A as saying Jones or Hansen committed acts of fraud and/or conspiracy to commit fraud. The relevant issue here is what were their own words on the matter not bugs or your interpretation of what they meant.

Surely you must have had training in technical writing at some point along the road if, as you claim, you are completing your thesis at this point.

As to the relevance of my comments, I was really suggesting was that it would improve bug’s argument to directly quote D&W rather than paraphrasing them in his own words.

An example of “really poor logic”, by the way, is making up an argument that somebody never made and labeling it as “really poor logic”.

Both fraud and conspiracy imply a crime, one should always be careful about making claims of either of these, and especially about claiming that other people had said such things when they haven’t used such words. Using over the top language in a case like this just undermines your own credibility and the case you are trying to make.

Yes, you can see where this is heading, can’t you. There are a lot of denialists out there hoping Jones will be charged with something. Inhofe is thinking along the same lines in the USA.

Carrick (Comment#38146) March 16th, 2010 at 1:12 am

Nathan, what bugs said was:

So you accept conspiracy now?

Contrarian:

I have not downloaded the various databases and scripts (like Zeke’s) and run the calcs myself. Since no one has yet run a weighted trend of all stations 0-90N and minus 60-90N, I might have to tackle that chore.

You might, because it depends a lot on how you do the math for the answer you get, at least for the land-only record.

There is a fair amount of arbitrariness to how the average gets defined (whether you stretch the land over the entire band of the latitudes, or whether you weight by the fraction of land at that latitude band).

You think a difference should remain per ABL even using the anomaly method?

Short answer is yes, and for the same reason the models predict an increase in temperature sensitivity on land as you got to more extreme latitudes.

bugs (Comment#38148) March 16th, 2010 at 1:20 am

Carrick (Comment#38146) March 16th, 2010 at 1:12 am

Nathan, what bugs said was:

So you accept conspiracy now

I’ll take that as a ‘yes’.

bugs:

Yes, you can see where this is heading, can’t you. There are a lot of denialists out there hoping Jones will be charged with something. Inhofe is thinking along the same lines in the USA.

I agree it makes it easier for you if you can make them all look like vindictive wackos.

That in turn hurts your argument because it is very polarizing (something they are counting on too).

So you accept conspiracy now?

I agree that word got used even if not in reference to anybody in specific.

I would still say though that the passage that Nathan dug up was much more effective in undermining this dreck of D&W than any paraphrase of your own would have been.

Look I pretty much agree with you here….D&W went ape-sh*t crazy in that document, it appears to have been an ill-considered, hastily cobbled together piece of crap written in response to Anthony’s feeling that he had once again been “scooped” by Mennes.

I can understand why he may feel that way, but I don’t have a lot of sympathy for him at this point. Menne’s had every right to put together and publish that manuscript, with or without Watts’ blesisng, and I don’t blame Menne’s for not wanting to work with Watts after all the prior nasty, unprofessional things he has said in public in the past.

Could I make this any clearer?

Just to put the rabbits (and or insects )who label people a nasty insulting name like “denialist” knickers in a twist I am sharing this link.

May the truth at least speak to you in your dreams at night!

http://www.timesonline.co.uk/t.....061020.ece

“Just occasionally you find yourself at an event where there is a sense of history in the air. So it was the other night at the Royal Society, when a small gathering of luminaries turned up to hear that extraordinary nonagenarian, the scientist James Lovelock.

They had all come: David MacKay, chief scientist at the Department of Energy and Climate Change; Michael Green, Lucasian professor of mathematics at Cambridge; Michael Wilson, producer of the James Bond movies; Chris Rapley, director of the Science Museum; and more. You knew why they had answered the Isaac Newton Institute’s invitation. They wanted to learn where one of the most interesting minds in science stood in the climate debate.”…

… in 1979, Lovelock published the book-length version of his Gaia theory, which postulates that the Earth functions as a kind of super-organism, with millions of species regulating its temperature. Despite initial scepticism from the Darwinists, who refused to believe that individual organisms could act in harmony, the Gaia theory has been widely accepted and now underlies most atmospheric science.

What, I wondered, would be the great man’s view on the latest twists in the atmospheric story — the Climategate emails and the sloppy science revealed in the reports of the Intergovernmental Panel on Climate Change (IPCC)? To my surprise, he immediately professed his admiration for the climate-change sceptics.

“I think you have to accept that the sceptics have kept us sane — some of them, anyway,” he said. “They have been a breath of fresh air. They have kept us from regarding the science of climate change as a religion. It had gone too far that way. There is a role for sceptics in science. They shouldn’t be brushed aside. It is clear that the angel side wasn’t without sin.”

As we were ushered in to dinner, I couldn’t help wrestling with the irony that the so-called “prophet of climate change”, whose Gaia theory is regarded in some quarters as a faith in itself, was actively cheering on those who would knock science from its pedestal.

…blah blah blah

He had. In the end, his message was that we should have more respect for uncertainties and learn to live with possibilities rather than striving for the 95% probabilities that climate scientists have been trying to provide. We don’t know what’s going to happen and we don’t know if we can avert disaster — although we should try. His sage advice: enjoy life while you can.”"

Love some the comments below the article.
(BTW no skeptic is trying to knock science from a pedestal-we just want some honest good science, like non “valued added” data !)

Carrick, try the dictionary for a definition of fraud.
From dictionary.com

”
fraud
   /frɔd/ Show Spelled[frawd] Show IPA
–noun
1.
deceit, trickery, sharp practice, or breach of confidence, perpetrated for profit or to gain some unfair or dishonest advantage.
2.
a particular instance of such deceit or trickery: mail fraud; election frauds.
3.
any deception, trickery, or humbug: That diet book is a fraud and a waste of time.
4.
a person who makes deceitful pretenses; sham; poseur.”

Quite clearly Watts and D’Aleo are calling NOAA and NASA and CRU frauds using the definition of the word ‘fraud’.
And the conpsiracy they speak of involved CRU, NOAA, and NASA and all the data centres – that would have to be one of the biggest conspiracies in history, surely.

My really poor logic comment was directed at Lucia. Why would searching for the word “fraud” in a document be particularly helpful in determining if they were claiming someone was fraudulent?

I’m sorry, but the apologists for the d’Aleo/Watts document fall flat.

Alex Heyworth (Comment#38124) March 15th, 2010 at 7:11 pm

I don’t recall the exact figure, but a reasonably recent study found an incredibly high proportion of academic papers had been refuted within (I think it was five) years. Yet retractions are rare. Most researchers seem to just get on with the game.

There’s a world of difference between that, and what we’re talking about here. We’re talking about something that’s just plainly wrong, and being used to make calls of intentional fabrication. Just read the thing.

Another question is whether it is relevant to apply academic journal standards to a paper published by a lobby organization. I don’t see a lot of calls for retractions by the authors of papers published by the WWF or Greenpeace, though I am sure some of theirs are equally contentious.

I’d like to see any standards at all applied to Watts/d’Aleo. If the WWF and Greenpeace were making easily refutable but widely publicised calls of fraud, they should be heavily criticised as well.

Anyway, I have said all I am going to say on this red herring.

Much of Watts’ output is red herrings, but he’s being taken seriously enough by some that he’s ending up getting a wider audience in the media. That media might notice that his fraud claims are baseless.

lucia (Comment#38127) March 15th, 2010 at 8:02 pm

I did a word search using Acrobat. As far as I can tell, the word fraud does not appear in the SPPI document coauthored by Watts.

This is silly. The report is clear as daylight accusing NOAA of intentionally cooking the books.

SPPI

It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler….
Calculating the average temperatures this way would ensure that the mean global surface temperature for each month and year would show a false-positive temperature anomaly – a bogus warming.

What’s bizarre is that the authors realise that the data are spatially gridded (next quote), but they show zero recognition that anomalies are used (not until p94 in a different matter, when they get all worried about the base period, without realising that choice doesn’t affect the trends).

SPPI

It is only when data from the more southerly, warmer locations is used in the interpolation to the vacant grid boxes that an artificial warming is introduced.

Carrick

Look I pretty much agree with you here….D&W went ape-sh*t crazy in that document, it appears to have been an ill-considered, hastily cobbled together piece of crap written in response to Anthony’s feeling that he had once again been “scooped” by Mennes.

Well, they did have a very eminent expert reviewer look it over. Monckton did the honors of proof-reading. That should have improved the draft, no?

Watts made an appearance over on this blog (link below) a couple days back, and touched on the issue.

Apparently Tamino’s anonymity and alleged failure to “show his work” keep him from addressing Tamino’s work.

Weird, because Tamino’s anonymity haven’t kept others from reading through his posts and understanding what he did. It didn’t keep Zeke, Broberg and ccc from confirming the results. At least he acknowledges that Zeke did something.

As for the main topic of the thread, you see Watts being at best forgetful, at worst dishonest; all the while flailing away at others for errors which are his own.

http://timpanogos.wordpress.co.....ment-96773

liza (Comment#38159) March 16th, 2010 at 4:38 am

Love some the comments below the article.
(BTW no skeptic is trying to knock science from a pedestal-we just want some honest good science, like non “valued added” data !)

I’m puzzled, you have got that.
It seems apparent you have never worked with raw data, because it is usually problematic.

Nathan–

My really poor logic comment was directed at Lucia. Why would searching for the word “fraud” in a document be particularly helpful in determining if they were claiming someone was fraudulent?

Why would it be helpful? Obviously, it’s helpful because if there are no direct accusations, you have to actually go through, find claims you think represent the equivalent of accusing fraud.

At best, what you can show is that D&O may have accused NOAA of fraud by it’s weakest definitional meaning. For example, I think it’s fair to say has accused NOAA of being “humbugs” or using “tricks”. (Heck, maybe the trick is similar to the one used to hide the decline!)

Using that definition, then committing fraud is a rather mild accusation. Do you intend a mild accusation? Or are you equivocating and using a word according to two different meanings?

bugs (Comment#38163) March 16th, 2010 at 5:29 am
“It seems apparent you have never worked with raw data, because it is usually problematic.”

Yes I have. PSI , noise, temperature, plus wear measurements -automotive break pad material testing for an R and D company. Real road tests. Los Angeles Freeway System. 200 miles per day 6 days a week; data collected ever time the breaks were used. Computer in vehicles. Worked with Ford, Toyota, GM, etc and Friction Materials Manufactures/Engineers. I down loaded, and ran the data… generated the final reports did all the charts and graphics. Handled daily records too; the engineers wanted.

Never ADDED NOR CHANGED DATA IN ANY WAY. I worked there 11yrs.

Get over yourself.

Break! Sheesh I mean BRAKES.
I need a break from people like this!

“I need a break from people like this!”

liza,

You deserve a break today. Trying to converse with ♪ One-Note ♪ Propagandists and their endless droning can be annoying.

Hang in there, kiddo.

Andrew

Andrew, thanks.

lucia (Comment#38165) March 16th, 2010 at 6:30 am

I’m sorry, but this is nonsense. There is nothing mild about this.

They are clearly saying that somebody intentionally removed stations from the record, with the express purpose of showing warming that did not actually occur.

This is as serious as a charge can get, in this field. And they knew it, so they publicised it.

It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler….
Calculating the average temperatures this way would ensure that the mean global surface temperature for each month and year would show a false-positive temperature anomaly – a bogus warming.

“They are clearly saying that somebody intentionally removed stations from the record”

Are you asserting that somebody removed the stations accidentally?

Andrew

Re: liza (Mar 16 04:38) –

So, Sunday Times reporter Charles Clover says

Despite initial scepticism from the Darwinists, who refused to believe that individual organisms could act in harmony, the Gaia theory has been widely accepted and now underlies most atmospheric science.

I hadn’t known that.

Do Denialists reject Gaia, in addition to rejecting warming and (by association) the Holocaust?

Re: carrot eater (Mar 16 05:18) –

Anthony Watts gets called out by Ed Darrell at “Millard Fillmore’s Bathtub,” heated exchanges follow in the Comments (which read, er, upside down).

This side note from Watts:

However, there’s some open source work being done elsewhere by people (at Lucia’s) that do show their work and who do put their own names to it. And in fact Joe D’Aleo and I are responding to that, because it is in fact a valid criticism, showing the work, the methods and code, where Tamino’s is just 5th grade noise.

Ed is also upset that we haven’t responded on the timeline expectation another has set.

Heh. Ed assumes that the issued is settled, just like the science according to Gore, and that it’s all an open an shut case. But you see, it’s not. We’ll publish an update soon, but we won’t rush just becuase somebody says we should.

Darrell’s rebuttals include

However, the blogger named Tamino ran the numbers and showed that with the data dropped, warming trends over the century are still clear, and that if anything, the dropouts produced a slight reduction in overall warming appearances. Tamino isn’t alone in that statement. Others also ran the data, confirming Tamino’s findings, including a data run written up at a contrarian blog in Watts’s camp.

So, Watts has clearly noted Zeke’s posts here, in any event.

Contrarian,

Here is the analysis you wanted:

http://i81.photobucket.com/alb.....ure188.png

Slope 1960-2009:

Global land – 0.202
NH 0 to 90 lat – 0.233
NH 0 to 60 lat - 0.221

AMac (Comment#38176) March 16th, 2010 at 8:33 am

the Gaia theory has been widely accepted and now underlies most atmospheric science.

I think this is plainly untrue. Gaia is something of a philosophy; in the strong form it’s basically religious. To the extent that living things and their environment are interconnected in complicated ways, a Gaia-minded person might call that evidence of their idea. I just call it the entirely unsurprising result of evolution. But there’s no Gaia underlying anything in atmospheric physics. You can sit there and write out a radiative model, or a radiative-convective model, and there’s no Gaia in sight.

This side note from Watts:

Yes, I noted that Watts acknowledged Zeke’s work. Though he didn’t address the contradiction – if Tamino’s anonymity and alleged lack of showing his work is such an obstacle, then how were so many people able to confirm the results so quickly? I also note further forgetfulness/dishonesty, take your pick, from Watts, plainly visible right there in that thread.

Andrew_KY (Comment#38175) March 16th, 2010 at 8:31 am

Are you asserting that somebody removed the stations accidentally?

The history of the GHCN has been addressed several times, such that I find it difficult to accept you’re still asking this. It was even clearly described in papers in 1997 and 1998, well before EM Smith ever started looking at this.

If I bother to type an explanation in again, will you at least in future acknowledge that you’ve heard it?

Re: carrot eater (Mar 16 09:17) –

> But there’s no Gaia underlying anything in atmospheric physics

I should have bracketed my comment on Gaia’s wide acceptance with [raised eyebrows] and [/raised eyebrows].

Zeke,
Totally up to you, how you want to waste your time on these things. But if Watts/d’Aleo keep leaning on EM Smith (I don’t know if they will), it could be useful to have the FDM programmed in, in order to quickly check whatever he comes up with. I’m not at all familiar with your platform, but I could try to help if you want.

Carrot,

I’ll put it in my long to-do list, but I need to finish the U.S. UHI stuff right now. Ron just provided me GRUMP urban designation data for all the USHCN stations, so I’m running with that right now. Plus Spencer’s recent post on the U.S. makes it timely.

Took a look at his place, and I have an update on EM Smith’s progress. He’s now saying he knew why people use anomalies, though his earlier posts show no sign of that recognition. Fine.

http://chiefio.wordpress.com/2.....anomalies/

But of greater interest, see his response to “Dougie” at March 15, 4:10 pm

He starts out OK, so far as it goes.

Then read this part. In his examples, ’0′ is the flag for missing values, so ’0′ is not an actual reading. Each row is a separate station, so there are three stations in his examples.

OK, but anomalies are not perfect either. If you had a baseline in the first 2 values and
had:

1 1 1 1 1 1
2 2 2 2 2 2
0 0 0 3 3 3

your baseline would be 1.5 (as a simple average) and you would now compare 1 to 1.5 in time zero vs 1 to 1.5 in time 6 and get ‘same as time zero’. Similarly 2 vs 1.5 and 2vs 1.5, no change. But what do you do for that “3″? Compare it to 1.5 and find 1.5 of warming? Compare it to ??? (So GIStemp makes up a basket of things it averages together and compares that to the baseline… 3+2+1 = 6 /3 = 2 … gee, that’s 0.5 C of warming… In reality, there’s an attempt to adjust that 3 to ‘what it ought to have been in time period 1′ but just how good is that ‘homogenized’ and ‘in-filled’ value?)

If you run it the other way:

1 1 1 1 1 1
2 2 2 2 2 2
3 3 3 3 3 0

We have a baseline of 6/3 = 2 (as a simple example) and we have 2+1 =3 / 2 = 1.5 at the end… so we ‘cooled’ by 0.5 C compared to the ‘baseline basket’ due only to station change. Again, you could try to ‘make up’ what that last 3 ought to have been, but then you are not measuring ‘anomalies’ so much as you are measuring ‘imaginings’.

And you can see him making a total hash of anomaly calculations. He doesn’t do the CAM correctly, nor does he understand the RSM. He then thinks the RSM infills missing values; there is no need for that. If there are missing points, you simply don’t use those points.

I’ll leave it to the reader to work out how the CAM or RSM actually work in these two situations. It’d be a nice test, after Lucia’s posts to explain the two. They simply don’t do whatever EM Smith thinks they do. Though oddly, he does the CAM OK just above these lines.

That he’s having so much trouble understanding the RSM makes me wonder if he’s implementing the FDM in any reasonable way.

He sheds some further light on his own thinking farther down the reply, here.

FWIW, I think that last line effect is how station drops warm the planet. Leaving a warmed series ‘unclosed’ by dropping at a non-random time.

I wonder whether some people do have a life (or job?) instead of spending day and night on a blog countering anyone who has the guts to doubt the data with more graphs. Lies, damned lies and climate statistics.

Spencer makes a valuable statment on WUWT:
There is a clear need for new, independent analyses of the global temperature data…the raw data, that is. As I have mentioned before, we need independent groups doing new and independent global temperature analyses — not international committees of Nobel laureates passing down opinions on tablets of stone.

Here’s is what RSM does (i.e. Lebedeff and Hansen ). The method does involve sequentially updating the pooled value for the subbox with containing those three boxes. Taking EM Smiths’ example of stations with temperature readings:

station 1: 1 1 1 1 1 1
station 2: 2 2 2 2 2 2
station 3: M M M 3 3 3

I’m using “M” for missing.

First: Find the station with the longest record. If there is a tie, pick whichever you like, it doesn’t matter.

So, we pick station 1:

The first estimate of the pooled temperature values is:
pooled Guess 1: 1 1 1 1 1 1

Now, look at the remaining list, and find the one with the longest record. That’s station 2
* Find the times when you have records for the pooled Guess 1 AND station 2: That’s every single time.
* Compute the average for the pooled guess based on the times used above: ave. poolled guess 1 = 1.
* compute the average for station 2 for those same times. That’s ave. station 2= 2.
* Take the difference (2-1) = 1. Subtract this from all station 2 temps to created a shifted value:
shifted 2 = 1 1 1 1 1 1

Average shifted 2 and pooled Guess 1 to create pooled Guess 1:
pooled Guess 2: 1 1 1 1 1 1

== Now, incorporate the last station.
* Find the times when you have records for the pooled Guess 2 AND station 3: That’s the last 3 records.
* Find the average for “pooled guess 2″ over the final 3 records: That’s 1.
* Find the average of station 3 over the last 3 records: That’s 3.
* Find the difference (3-1)=2. Subtract from all nonmissing values from station 3:

shifted 3 = M M M 1 1 1

To create pooled guess 3, averaged pooled guess 2 with shifted 3, but only where there are no missing values.

Pooled guess 3 = 1 1 1 1 1 1

Note:
A) At this point, this is not the average temperature. It isn’t really an anomaly. The RSM (i.e. hansen method) turns it into a true anomaly later and they do so in a way that is not idiotic.)
B) This is trendless. Adding guess 3 from the warmer location does not create a trend in the pooled value.

You can turn the pooled guess 3 into an anomaly defined by the average of the first two readings, by finding their average (1) and subtracting that from all values in the pooled guess. That results in

Final Anomlay = 0 0 0 0 0 0.

Someone, somewhere may use the method Chiefio describes. But GISSTemp doesn’t. NOAA doesn’t. CRUdoesn’t. People are aware that the method Chiefio describes has truly horrific difficulties. So… they don’t use it!

Carrot Eater:

Weird, because Tamino’s anonymity haven’t kept others from reading through his posts and understanding what he did. It didn’t keep Zeke, Broberg and ccc from confirming the results. At least he acknowledges that Zeke did something.

Somehow I don’t think it was Grant Foster’s supposed anonymity that stopped Watts from responding on open mind. Probably it had something more to do with Tamino acting like a little child when anybody criticizes him.

I’m thinking of RomanM having the temerity to observe issues with Tamino’s method, for which he got the usual Royal Tamino Acting the A$$ treatment.

And you actually view everybody else’s efforts as motivated only to “confirming Tamino’s results”? What a narrow word view you must have.

Ah, I’ll fill in the solutions, as well. Correct me if I make a silly error, myself.

Smith’s first example:

station A: 1 1 1 1 1 1
station B: 2 2 2 2 2 2
station C: N N N 3 3 3

I’m putting N for null, for missing data.

First, using the most simple anomaly method, the CAM. We need to know what the baseline is. EM Smith specifies the first two values. We see that station C has no data in the first two values. Therefore, we simply cannot use station C. It’s gone.

The average of Station A in the baseline is 1. The average of Station B in the baseline is 2. Subtract the appropriate baseline from each series. So the anomalies for A and B work out to

A: 0 0 0 0 0 0
B: 0 0 0 0 0 0
C: not used

So when you combine them, you just get 0 0 0 0 0 0. Not whatever mess EM Smith was suggesting.

OK, so what does GISS do, with the RSM? The RSM builds up the average, one station at a time. I’ll denote that average as R.

This time, there is no pre-defined baseline period. We simply start with the longest station. A and B are tied for length, so let’s start with A, and put it into R. We have R = 1 1 1 1 1 1.

Then we consider B. We find the period where R and B overlap. Turns out, that’s the entire data set. We find the mean value of each, over that period. That is 1 and 2, respectively, for R and B. The difference of the means is 1. This is the offset, or the bias. So we subtract 1 from each value of B.

so B with the offset becomes: 1 1 1 1 1 1

Combine R with the offset B, and you get a new
value of R: 1 1 1 1 1 1
We will now continue to build on this combination.

So, we now consider C. It’s period of overlap with the existing combination is in the last three values. The mean of C over that time is 3. The mean of the growing combination is 1. The difference is 2. So, we subtract 2 from each value of C.

C with the offset: N N N 1 1 1

So we add this to R and get a new R: 1 1 1 1 1 1

Now that I’m done adding stations, now I can subtract out whatever baseline I want. In this case, no matter what baseline I choose, I get 0 0 0 0 0 0.

So regardless of whether you use the CAM or RSM, you do not get any trend from these data here. In this toy example, neither method has any trouble with the station drop.

Now, it is certainly possible to come up with scenarios where the RSM or CAM give incorrect results. We’ve been saying this the whole time, describing how it can happen. Lucia has demonstrated it, in this very post.

But it concerns me that EM Smith has been working on this for this long, and still hasn’t worked out how these things work.

Carrick

nd you actually view everybody else’s efforts as motivated only to “confirming Tamino’s results”? What a narrow word view you must have.

It’s also very broad, since Spencer posted before Tamino. So, if confirming Tamino’s results was Spencers motive, Spencer must also be psychic!

“The history of the GHCN has been addressed several times, such that I find it difficult to accept you’re still asking this. It was even clearly described in papers in 1997 and 1998, well before EM Smith ever started looking at this.

If I bother to type an explanation in again, will you at least in future acknowledge that you’ve heard it?”

CE,

I do not know the history of the GHCN, and I haven’t had the time to read all of your posts recently. You can link me if you have a prior explanation and evidence.

Andrew

Andrew_KY, the short version is nobody at GHCN, NASA or anywhere else dropped any stations. The data is still there, except for those stations which were closed by their own agencies.

The issue is that many stations don’t automatically get updated in the GHCN network, and oddly to you I know, there isn’t anybody who is tasked to keep the thousands of stations up to date at a given moment. Further, they are doing some “house cleaning” and adding the data from a lot of those manually updated stations into the database (but once again, there will be a terminus year, until the next time this gets done manually).

You can toss as much snark as you want about the fact that the data stream isn’t kept as up to date as you would like, but the fact remains the data is there, and to the extent that it is not (in the case of station closures), it has nothing to do with any of the people at NASA or CRU.

For Watts and D’Aleo to have written this (shout-out to Nathan for finding it):

Recent revelations from the Climategate1 emails, originating from the Climatic Research Unit at the University of East Anglia showed how all the data centers, most notably NOAA and NASA, conspired in the manipulation of global temperature records to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.

is completely inexcusable on their part.

Anthony at the least must be aware that this is plainly false, must be equally aware that the intended audience would not be aware of the context I outlined above, and I’m sorry when you juxtapose those two facts together you get something very very smelly.

Carrick (Comment#38194) March 16th, 2010 at 10:48 am

Somehow I don’t think it was Grant Foster’s supposed anonymity that stopped Watts from responding on open mind. Probably it had something more to do with Tamino acting like a little child when anybody criticizes him.

Somehow I don’t think it was Watts’ possible inability to post on Open Mind that stopped Watts from responding. Watts has his own platforms where he can respond, and he has mentioned it on his own blog. Watts complains about Tamino’s anonymity and the lack of showing the work; both of which are red herrings anyway. Posting responses on Open Mind has nothing to do with Watts’ lack of a substantive response so far. What a narrow world view you must have.

Anyway, Watts tells us that something is in the works.

And you actually view everybody else’s efforts as motivated only to “confirming Tamino’s results”? What a narrow word view you must have.

Everybody’s motivated to see what effect station drop-off has. In so doing, they confirmed Tamino’s results. My point is that you don’t need Tamino’s code to do know what he did, or to do similar work for yourself. You just have to have a basic idea of how GISS, NCDC or CRU do the calculations. Something the publishers of the SPPI report should have had, before they published that mess. Yet Watts tried to make a big deal about Tamino’s code not being public.

I think Zeke was actually the first one to work on it quantitatively; he just hadn’t done spatial gridding in his first pass at it. Before that, the general response was simply, “you need to use anomalies”.

Andrew_KY (Comment#38197) March 16th, 2010 at 11:07 am

It isn’t just me who’s spelled this out. It’s been multiple people.

http://www.ncdc.noaa.gov/oa/cl.....e-1997.pdf
http://www.ncdc.noaa.gov/oa/ho.....ection.pdf

Basic gist: When the GHCN started out, the NCDC went through the various historical archives to collect whatever past data they could find. The data sources are described in the first PDF. This got them a large number of stations.

Then, to keep it up to date, they would add whatever station data were reported on a monthly basis by the different countries, electronically. The different countries were not doing this monthly updating for all of their stations. So there’s a station dropoff when you switch from the historical period to the electronic reporting period.

In the second PDF, they show which stations were reporting regularly, and see that the global coverage could be better. So they identify a network of stations they’d like to see report, and they encouraged all the countries of the world to send data for at least those stations. Some send more; some send less; some send but at irregular intervals.

At this point, you might ask why they don’t do another scrounging around for historical data. It sounds like they’re doing some of that for inclusion this year, but I don’t know how extensive it will be.

I’m amused that both Lucia and I rejected EM Smith’s convention for the missing data flag.

I’m glad we both reached the same answer.

Somebody who’s allowed to post there might helpfully tell him that in order to diagnose the weaknesses in the various anomaly methods, he has to know what those methods are in the first place.

AMac (Comment#38176) March 16th, 2010 at 8:33 am
Re: liza (Mar 16 04:38) –

“Do Denialists reject Gaia, in addition to rejecting warming and (by association) the Holocaust?”

How rude. I wasn’t interested in that part of the article. I have no idea either.

I don’t reject warming, I embrace and adapt to warming during interglacial periods like any good human who came before me would.

You know what I found more interesting?
What was said about a skeptical attitude in that article and that I got some knickers in a twist.

AND what I also find interesting is that you (or anybody else except Andrew) don’t find it interesting; that I didn’t have to tease, massage, add, subtract fix or change data that contained temperature, PSI, noise levels, and wear measurements readings 5 days a week, anywhere from 6 to 12 cars running DAS equipment daily on 200 mile routes, Los Angeles area freeways; 6 am-4pm or so a day, 1,000′s of miles at completion of tests, these were “new” cars too, not sold to the public yet; and I have 11 yrs of data processing experience. (I only left that job to be married and move to another part of California) I still have contact with my friends from there -who live all over the country-we talk on my Facebook page a lot about these things…guess how many of them believe AGW data is good science? (Yes I am a smart aleck sometimes too!)

Carrot Eater:

Somehow I don’t think it was Watts’ possible inability to post on Open Mind that stopped Watts from responding.

LOL. Possible inability?

You could write for Pravda.

I agree Watts is over the top in complaining about the lack of availability of the code (to his credit Tamino provided his method in enough details it could be replicated).

And also about complaining about the supposed anonymity of Grant Foster/Tamino.

I didn’t want to stoke Zeke’s ego too much, but it’s my impression he’s well ahead of Tamino in every respect.

Everybody’s motivated to see what effect station drop-off has. In so doing, they confirmed Tamino’s results.

For what it’s worth, I had already explained on the Air Vent (see comments, it describes the method, not necessarily all of the motivation) in January that gridding the data reduces its sensitivity to station drop out. I had even gotten to the point where I could model the way error creeps into the CRUTEMP computation of mean land temperature. In many respects, my own analysis is still advanced of what anybody to that point (or even now really) had done yet. (Not bad for two to three hours of work.)

And as Lucia points out, Spencer had also looked at this too, it appears contemporaneously with Tamino.

My only point is, youu are giving undue credit to Tamino to suggest that these other efforts were “confirmations” of what Tamino had done. I’m also not sure what Tamino has done that he thinks is actually publishable in a peer review journal at this point. It appears (last I perused there) he is uncertain to, to be fair.

Anyway, Watts tells us that something is in the works.

I hope you excuse me if I just say I couldn’t care less at this point.

Carrot–
In the excel spreadsheet, I created a column ‘B’ with 1′s and o’s for “presence/absence” then computed using “SumProduct(A,B)/Sum(B). Obviously, you need to do something to deal with missing, and using 0s for error nor “no data” flags is very poor practice.

(The spread-sheet actually has more flexibility than I documented. I figure at some point, I’ll discover whether there really is anything that I call blending or infilling. I’ve got a real work project right now, so I haven’t looked at the later Hansen papers.)

Overall, Smith is sufficiently confused about how anomaly methods are implemented, that it seems like trying to discover what he thinks “blending” and “infilling” are, and then trying to discover what they really are in order to compare is mostly a waste of time. Also, one of Smith’s main deficiencies is that he rarely provides details for what he did in words. In today’s post, there are graphs of North America. The data is from GHCN. Raw? Adjusted? Is those area averages? GISSTemp averages? Averages equally weighted over thermometers? The quantity dT is relative to what? Is it computed with his new method?

Presumably some of these things were mentioned at some time, but I really don’t know. (I can’t claim I’m always the clearest person. But one reason I include caption boxes on figures is to at least try to carry forward certain information for quick updates.)

Carrick (Comment#38207) March 16th, 2010 at 12:11 pm

LOL. Possible inability?

Regardless, it has nothing to do with it. I don’t think Watts ever mentioned trying to post there; Watts has been trying to act as if Tamino didn’t do anything worth responding to, because of anonymity and lack of code. That’s my point.

I agree Watts is over the top in complaining about the lack of availability of the code (to his credit Tamino provided his method in enough details it could be replicated).

The latter point is completely lost on Watts.

I didn’t want to stoke Zeke’s ego too much, but it’s my impression he’s well ahead of Tamino in every respect.

My impression as well, though Tamino hasn’t put out any follow up results. But Tamino admitted that his code was still a bit ramshackle at one point; I think he had to run it separately for each grid box, or somesuch.

Zeke would have been way out in front of everybody; he just thought that gridding would be harder than it actually is. Once Tamino came out with his results (NH on Feb 23), then Zeke finally got going with the gridding (presented on March 1). But he already had a non-gridded result on Jan 21.

So far as I know, Dr. Nielsen-Gammon was the first to try explaining anomalies, instead of just saying “you have to use anomalies”. That on Jan 27.

My only point is, you are giving undue credit to Tamino to suggest that these other efforts were “confirmations” of what Tamino had done. I’m also not sure what Tamino has done that he thinks is actually publishable in a peer review journal at this point.

How is that undue credit? CCC, Zeke and Broberg confirmed his result. How else would you phrase it? CCC and Zeke both mention Tamino in the first line of their posts.

http://clearclimatecode.org/th.....ng-effect/
http://rankexploits.com/musing.....-analysis/

Anyway, it’s borderline for publication. He has not publicly demonstrated whether calculating the offsets all at once is any better than doing them one by one. In principle it should be, but I don’t know if it actually is. If he can show this, then that’s a low-profile paper. Still, BAMS will publish stuff that isn’t earth-shattering.

Lucia:

That’s why I seized on his worked-out toy example. That lets you cut through all the rambling and ambiguous terminology, and see exactly what he thinks the calculations do. And after this many months, he still has it wrong.

He seems to think that the missing values have to be filled somehow. I don’t know why.

If he’s really doing the FDM, then something like this is happening (say these are data for consecutive Mays):

Station A, in Celsius: 14 16 13 15 11
The differences are: 2 -3 2 -4
The cumulative sum is: 0 2 -1 1 -3

The middle line I think he calls dT/dt, and I think he calls the last line dT.

The last line is an anomaly series; you could then subtract out whatever baseline you want.

If you had more than one station, you’d take the average of all the differences for that year, before adding to the cumsum. If a station had dropped out, you just have one less thing to put into that average.

Carrot–
But aren’t you forgetting Roy Spencer? He posted before Tamino. The proof wasn’t the focus of his post, but it was contained in something he posted.

Carrick
I think the only way Tamino can publish is if he finds a friendly editor who lowers the standard for novelty and scholarship well below the usual level for normal peer reviewed publication. That said, I think the politics of AGW will assure that such an editor exists and Tamino will find one.

For those not familiar with the norma standards of scholarship: It is not sufficient for a paper to contain things that are correct. They should teach the community of readers –who are generally researchers themselves– something new. It should answer a question researchers– not 8th graders in general science– actually have.

It’s not at all clear that Tamino’s analysis really does that. It’s much more likely that most readers of BAMS, GRL etc. would read claims at Chiefio or in the SPPI document and say “Well, of course merely loosing colder thermometers that wouldn’t result in a bias. Anomalies are specifically designed to fix that.”

Since the SPPI document presented no evidence to indicate any bias occurred ordinarily scientists would suggest that no one waste their time dis-proving something until either a) the proponents of this tenuous claim show it has even one iota of data support or b) the meme sticks around so long it causes real problems.

Here, we really don’t have either (a) or (b). So, ordinarily, why would anyone want to waste paper and ink publishing the refutation of the claim? An editor is going to have to be very indulgent to circulate this to reviewers. (That said, I suspect Tamino will find an indulgent editor.)

Andrew_KY:

The posters who are claiming that the station dropoff in GHCN is because those stations don’t report regularly are just plain wrong. Just as an example, 64 South American stations drop out of GHCN after 1992. More than 20 of them have continued to report regularly since then, and the data can be found in — wait for it — another NCDC product!

Re: carrot eater (Mar 16 12:48),
But does he doe something when he finds big jumps in dT/dt? Isn’t that sort of the point in computed dT/dt, and then summing to regain “T”?

Or is it that he computes dT/dt, averages those, then sums to regain T? That’s not going to be much different from what the other people using anomalies do. It might have slight differences at boundaries because you need two adjacent points to compute dT/dt. Does he have any discussion of details? Somewhere?

Re: liza (Mar 16 12:05) –

My apologies for unintentionally insulting you when I quipped upthread, “Do Denialists reject Gaia, in addition to rejecting warming and (by association) the Holocaust?” Note to self, use [sarc] and [/sarc] tags. I’m labeled a ‘Denialist’ often enough. The moniker’s associations offend me, so I assume the same for others, thus don’t employ the term.

Re: lucia (Mar 16 12:50) –

It seems to me that the notions that Tamino and Zeke have contested are so widely believed that a case can be made for accepting a peer-reviewed paper that places concepts like thermometer dropoff and anomaly calculation in a clean, clear context. That would be more valuable than [sarc] certain papers published in PNAS back in 2008. [/sarc]

torn8o,

Thanks. My confidence that the temperature is being measured and/or recorded meaningfully/accurately/correctly for a given site is still zero.

Andrew

But aren’t you forgetting Roy Spencer? He posted before Tamino.

Not at all, stick him in there. I don’t always list him because he took such a different approach. Tamino/Zeke/ccc/Broberg all pretty much did the same thing, but with different implementations (different grid boxes, CAM vs RSM vs tweakedRSM, etc; no UHI adjustment or yes UHI adjustment). Spencer took a rather unique approach and a different data source, so he doesn’t have a graph that looks just like everybody else’s graph, with pre- and post-cutoff stations.

As for Tamino’s newness, the main thing is in the tweaked RSM. That said, while “dropping cold stations makes it warm” is obviously wrong, “dropping the less warming stations in the subbox make it warmer” is correct. I don’t know if anybody has actually published before, showing specifically that the 1990 dropoff didn’t suffer from this defect. But seriously, page through BAMS. It’s not all hugely significant. It also need not be a full-blown paper. Notes, communications, etc are appropriate.

Or is it that he computes dT/dt, averages those, then sums to regain “T”?>

I don’t know what he does exactly, but I think his code is available. It shouldn’t be necessary to dig through somebody’s code to even get a basic idea of what they did, but this guy is a special case.

In my mind, this is how the FDM is supposed to work (mind, I’ve not studied it very carefully yet, nor do I have the details on exactly how the NCDC used it before they gave up on it):

Station A 10 12 16 M M warming rapidly, then cuts off
Station B 10 10 10 10 10 no trend

differences
A 2 4 M M
B 0 0 0 0

averaged differences
1 2 0 0

intermediate anomalies (starting at zero, by convention)
0 1 3 3 3

anomalies, subtracting out the baseline, if all values are in the baseline

-2 -1 1 1 1

You could, if you wanted, exclude outliers in the difference series.

torn8o (Comment#38215) March 16th, 2010 at 12:51 pm

The posters who are claiming that the station dropoff in GHCN is because those stations don’t report regularly are just plain wrong. Just as an example, 64 South American stations drop out of GHCN after 1992. More than 20 of them have continued to report regularly since then, and the data can be found in — wait for it — another NCDC product!

Which product?

There are multiple types of regular reports. There are SYNOP reports; there are also monthly CLIMATs. In order to get into the GHCN, you need to issue CLIMATs. These are the ones with the monthly means.

The stations that dropped at 1990 were not issuing CLIMATs at the time GHCN was put together. Now, some of those stations have started issuing CLIMATs in the last few years; not all of those have been added to the GHCN yet. This includes Bolivia, some of Canada, and scattered others. Some of these could be added this year.

To search the CLIMATs, go here
http://www.ogimet.com/gclimat.phtml.en
The JMA includes all CLIMAT data. See maps here.
http://ds.data.jma.go.jp/gmd/tcc/climatview/

Carrot Eater:

So far as I know, Dr. Nielsen-Gammon was the first to try explaining anomalies, instead of just saying “you have to use anomalies”. That on Jan 27.

Depends on the forum and context.

I certainly understood what anomalies where, why they were valuable, and further the importance of measuring trends rather than temperatures well in advance of that and I have explained this to other people on the various blogs when it was apropos to do so.

I don’t claim any special prize for this…Hansen was one of the early adopters for this area. And I give him credit for going faster and farther than anybody else in the field on this.

Watts started a brush fire and a whole new round of people became exposed to some very sound methods for data analysis. As always, rediscovery played a role and people always assume ideas that are new and novel to them that they came up with themselves must be new and novel to everybody else.

From my point it’s all good, and the only people here who lost were Watts, D’Aleo and possibly Smith.

Carrick

I certainly understood what anomalies where, why they were valuable, and further the importance of measuring trends rather than temperatures well in advance of that

That should go without saying, and you do seem to agree. Anybody who’s looked into the issue of temperature measurements should have had a grasp of that; this has been ingrained since the early 80s at least, if not earlier; I’ve not read papers on the topic from the 60s and 70s. Jones before Hansen, by the way, I think.

As it happens, EM Smith, d’Aleo and Watts did not have a grasp of it; and looking over at his worked-out toy example, EM Smith still doesn’t. We see that just about a year ago, Watts didn’t have a grasp of anomalies then, either, when he made his bizarre histograms.

I’m just saying, N-G bothered to explain anomalies in the context of the SPPI report. The guys at RC basically just said “you didn’t use anomalies, so you don’t know what you’re doing”. Which was immediately obvious to anybody who knew what was going on, but admittedly, not that isn’t the entire lay audience.

As for how it worked out: the amplification/accusation is the troublesome part. If EM Smith took a year to figure things out for himself, taking whatever trips into the weeds as he desired, that’d be fine. That’s somebody learning, the hard way. I’ve spent plenty of time in various weeds myself.

That somebody took his work and made the SPPI report out of it is where the bad judgment comes in. It wasn’t just “hey, do these equations really work? when do they break? is there some effect at 1990?”. No, it was accusations that people were intentionally cooking the books, in this manner.

torn8o:

The posters who are claiming that the station dropoff in GHCN is because those stations don’t report regularly are just plain wrong. Just as an example, 64 South American stations drop out of GHCN after 1992. More than 20 of them have continued to report regularly since then, and the data can be found in — wait for it — another NCDC product!

The actual claim I’ve seen and made was that these stations aren’t included in the automatic updates to the GHCN, and that many are part of a future update to the GHCN data base.

You see anything wrong with that?

I’m going to assume you made up the part about “posters who are claiming that the station dropoff in GHCN is because those stations don’t report regularly.”

In fact it’s a pretty good example of a form of logical fallacy that Lucia’s been hammering on here.

Carrick:

(delete if this comes up duplicate)

I meant in the context of the SPPI report, and laymen. Of course, anybody who knows how these things are done would already have the grasp of anomalies. This goes back decades, and I think Jones was doing it before Hansen. I’m not sure what was done in the 60s and 70s.

I’m just saying Dr. N-G spelled it out for the audience who might not have that background. In contrast, the RC guys just said, “you don’t know what you’re doing, because you didn’t use anomalies”. Which is fine for the audience who’s clued in, but admittedly that isn’t everybody.

Somehow, Watts, d’Aleo and EM Smith did not grasp the basic concepts of the field they were discussing. Judging from his worked-out example of a couple days back, EM Smith still doesn’t quite get it. We saw about a year ago that Watts was having trouble with the basic idea, when he made his bizarre histogram of anomalies. So this is a recurring problem spot over there. It’s more than a little weird.

As for how it all comes to pass: it’s the amplification and accusations that are the problem. If EM Smith had just taken a year to work things out, going off into the weeds in the meantime, that’s fine. I’ve spent plenty of time in various weeds myself. Understanding doesn’t always come overnight.

But instead, somebody took his work and made the SPPI report out of it. It wasn’t just questioning whether the methods work, or whether some bias could creep in at 1990. They jumped right to accusations of people intentionally cooking the books.

If Watts/d’Aleo want to be treated like responsible grown-ups at the table, they need to learn they can’t pull stunts like that.

Carrick:

There actually are a few stations that do issue CLIMATs, but don’t yet make it into GHCN. They just started doing it too recently for the NCDC guys to bother fixing their software to deal with it. They’ll do it eventually, maybe this year. They apparently didn’t think it urgently necessary to always append 2006-current data to stations that had been silent since 1990; the big gap makes that data hard to use. Maybe they’ll append the recent data when they’re able to fill in the gap.

Sort of bureaucratic inertia, I know. The JMA does include these stations.

Carrot Eater:

If Watts/d’Aleo want to be treated like responsible grown-ups at the table, they need to learn they can’t pull stunts like that.

I agree with your basic narrative on the SPPI account… the anomaly method has been around since the 19th century. I singled out Hansen because his 1987 paper (in my opinion) was basically a “home run”. I didn’t see the RC post, but it’s unfortunate that they don’t spend anytime on things like this.

Thanks for the comment on the GHCN. I’ll see if I can find the statement from NOAA on this. Also the CLIMAT link was useful. I keep losing that one and haven’t figured out the “magic words” to track it down using google.

Here’s a tale of two stations:

GHCN has data for 87596 through 2007, even though there is a gap in the data from September 1989 through May 1994.

GHCN has data for 84132 only through April 1991, and then the station is dropped. However, CLIMATs show that the data started again in June 1994.

Why was 87596 not dropped, and 84132 was? 87596 “stopped reporting” earlier and longer than 84132 did. Both stations have CLIMATs available from another NCDC product, and the CLIMATs for 84132 have been available all along. There may be a good reason for dropping 84132, but I don’t buy the lack of reporting/wait for the next big update argument, since they figured out a way to keep carrying 87596.

This is also not the only example of dropping some S. American stations while carrying recent data for others even though there is a big gap in the data around 1990-1992.

Re: torn8o (Mar 16 15:14),
…dropping 84132…
Again, they didn’t “drop” 84132. The data from that station came as part of the earlier phase of GHCN, when it was a historical project. Someone got a batch of data up to 1991 and put it into the repository. 84132 was never a regularly reporting station. Perhaps it should have been, but for whatever reason, the arrangement was never made.

Nick Stokes – you are wrong. 84132 WAS a regularly reporting station. It was not included as part of a batch of data.

Re: torn8o (Mar 16 15:40),
Well, I mean regularly reporting to GHCN, which didn’t even exist in April 1991. I guess they reported to somebody.

But as I’ve frequently said, GHCN up until mid-90s was a historical project – it collected batches of data and put them in a repository. Only later did NOAA take over the task of regular monthly updating, which required appropriate arrangements with the stations.

Do you have evidence that it was not included as a batch?

Lucia I quoted the first passage of the document, quite clearly using the dictionary definition of the word fraud they are calling a whole range of people frauds.
I find it really bizarre that a lot of people here have such a poor grasp of English.

Here it is again

“Recent revelations from the Climategate1 emails, originating from the Climatic Research Unit at the University of East Anglia showed how all the data centers, most notably NOAA and NASA, conspired in the manipulation of global temperature records to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.”

Note no use of the word fraud, but they still accuse NOAA, NASA, CRU and ‘all the data centres’ of fraud, that is they conspired in the maniuplation of global temperature records.

It’s not complicated.

Nathan–
D&W are American. They would never accuse any “centre” of anything.

You gave the dictionary definition and I pointed out that you have shown that what they say amounts to “fraud” if we use the “humbug”=”fraud” part of your dictionary definition. Your answer appears to be that what they say does match the dictionary defnition. So, am I to assume your answer to the following is yes:

Using that definition, then committing fraud is a rather mild accusation. Do you intend a mild accusation? Or are you equivocating and using a word according to two different meanings?

That is: Are you showing merely showing they are accusing others of being “humbugs”, but you wish us all to accept a word that can also be meant to convey criminal behavior?

Huh? Station 84132, complete reported history is here.

http://ds.data.jma.go.jp/gmd/t.....mp;n=84132

Nothing after 1991, until 2009. That’s exactly the sort of station that the NCDC hasn’t bothered including the recent data for, yet.

Where are you seeing post 1994 CLIMATs? I’m not seeing them on the CLIMAT search page, either.

Haven’t checked the GSN page yet (http://gosic.org/gcos/GSN-data-access.htm), but in general, if they have it, the JMA tends to have it, I think.

Let’s not play games here.

Something doesn’t have to fall under the US Code Title whatever Part whatever Chapter whatever, defining fraud as a criminal or civil offense, in order for it to fit the vernacular use of the word ‘fraud’.

A scientist intentionally cooking the books to get a desired result that otherwise would not have been observed: if you go on WUWT and asked, I’m pretty sure they’d call that fraud.

carrot–
But the question is what Watts specifically did accuse people of in the SPPI document. There is quite a bit of word parsing in that document. What’s missing in the stuff you quote in carrot eater (Comment#38174) is that the intention of “doing A” was to cause “bogus result B”.

What it does say is

a) Those working at climate centers did A — and they did A on purpose but the exact purpose is unstated.
b) The authors tell us they think doing A does result in bogus result B.

But that does not go so far as to say that those working at the climate centers knew it doing A would cause bogus result B. To make the but you quoted amount to an honest to goodness accusation of fraud in the most commonly used meaning, we need them to actually connect that dot. But they didn’t.

The reader is left to in fer that the purpose of doing A was go cause bogus result B. But is it said? No.

I know it bugs you to observe that they didn’t actually make any accusation, but this is the way of tendentious documents. They work by insinuation– that is writing in a way that causes readers to jump to their own conclusions.

In fact, if you don’t know subject matter and you see this sort of structure in writing, your spidey-sense should turn on when you are reading paragraphs that using insinuation instead of just spitting out accusations.

When people have a case, they make accusations directly.

Lucia

You seem to love word games, but please be serious here and not treat this like a rhetorical game.

By the definition of the word fraud:
“1.deceit, trickery, sharp practice, or breach of confidence, perpetrated for profit or to gain some unfair or dishonest advantage.
2.a particular instance of such deceit or trickery: mail fraud; election frauds.
3.any deception, trickery, or humbug: That diet book is a fraud and a waste of time.
4.a person who makes deceitful pretenses; sham; poseur.”

I personally believe they are basically applying every definition there, which means I think the authors believe the data centres were doing it for their own gain. At the very least it is 3 and 4. Not just ‘humbug’ but ‘deception’ and ‘trickery’.

” then committing fraud is a rather mild accusation. Do you intend a mild accusation?”

This comment is rather strange. Do you truly imagine that accusing a scientist or group of scientists or, in this case, hundreds of scientists and their organisations of deception is mild? I don’t think you can level a more serious accusation at a scientist over something science-related. You are attempting to trivialise a serious accusation. And that’s not even accounting for the fact that they were wrong.

I get the feeling you just like playing Devil’s advocate. Either that or you are taking PR lessons from Mosher.

“but you wish us all to accept a word that can also be meant to convey criminal behavior?”
??? I just think you should be honest and use the word that defines the accusation. And I don’t mean ‘accusation’ by the legal definition, rather the definition in the dictionary.

Lucia

The word ‘fraud’ does not require you to say why they did it. It can be any deception.

Lucia: I think the intent, cause and effect are in the SPPI report, clear as day. They say somebody went out of their way to remove these stations, because it would have this effect. Your argument is based on the word ‘because’ not explicitly appearing? Give me a break. That’s what a plain reading of it gives you; that’s how it’s being sold (yes, with the caveat that they may not have signed off on the final press release).
http://www.spaceref.com/news/viewpr.html?pid=30000

If you really want to dance around this, go ahead, but it’s quite the acrobatic dance that is required.

Nathan
Dictionaries provide numbered definitions when they consider the various definitions to be distinct. Accusing someone of a crime like Mail fraud — a federal crime (def 2) is not the same level of accusation as describing a book as a humbug waste of time (def 3) which is mild.

So, it appears you are saying you think you are demonstrating D&W accused others of being frauds according to entries (3 or 4) — that is the mild use. You have not shown that they commit fraud by definition (2) the criminal use, nor even that they achieved (1).

Is that how you are using fraud? Because you may. But you should be aware that some people will avoid it in these instances lest others think they intend meanings 1 or 2. It may frustrate you that they might prefer words that are not so easily misconstrued, but they will.

Lucia

“a) Those working at climate centers did A — and they did A on purpose but the exact purpose is unstated.”

Actually, they did state the purpose

“…to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.”

Carrot Eater: If you like to play the game of sticking words in other people’s mouths, just say so and be done with it.

I prefer to hang people using their own words.

You like to invent quotes and do it that way? If you think that’s the way to convince people go for it.

As I said, a career with Pravda could be in your future!

You have gone into apologist overdrive Lucia.

They use the word ‘conspiracy’, no argument there.

http://www.thefreedictionary.com/conspiracy

con·spir·a·cy (kn-spîr-s)
n. pl. con·spir·a·cies
1. An agreement to perform together an illegal, wrongful, or subversive act.
2. A group of conspirators.
3. Law An agreement between two or more persons to commit a crime or accomplish a legal purpose through illegal action.

To do what?

Around 1990, NOAA began weeding out more than three-quarters of the climate measuring stations around the world. They may have been working under the auspices of the World Meteorological Organization (WMO). It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler.

To “purposefully” manipulate the temperature record. This goes back to 1990!

They also name who is doing this. NOAA, possibly under the direction of the WMO.

It’s so wacky, I’m wondering if Monckton wasn’t a little over vigorous in his ‘peer review’, adding in his own thoughts.

Carrick: I’ve quoted the SPPI report exactly, and repeatedly. If you want to read something that isn’t there, that’s up to you.

Lucia: Seriously. Go on WUWT. Say that you have evidence that scientists at GISS intentionally manipulated the data in completely non-defensible ways, with the express purpose of creating warming where there was none. Ask the denizens there if they’d use the word ‘fraud’ for that.

Lucia

What sort of crazy world do you live in?

“1.deceit, trickery, sharp practice, or breach of confidence, perpetrated for profit or to gain some unfair or dishonest advantage.”

This does NOT imply a crime. It being a crime is NOT part of the definition. And I don’t mean ‘definition’ in the legal sense.

They are accusing them of fraud “…to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.”

” It may frustrate you that they might prefer words that are not so easily misconstrued, but they will.”
Your poor grasp of very simple concepts in English is frustrating…

Nathan:

You seem to love word games, but please be serious here and not treat this like a rhetorical game.

And you love run on argumentative conversations. I see this as nothing but a rhetorical game for you.

The point is obvious, if you can’t figure it out say your piece and move on.

Lucia

“I know it bugs you to observe that they didn’t actually make any accusation”

Actually they did make a pretty clear accusation:

“Recent revelations from the Climategate1 emails, originating from the Climatic Research Unit at the University of East Anglia showed how all the data centers, most notably NOAA and NASA, conspired in the manipulation of global temperature records to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.”

That’s an accusation… But I don’t mean that in the legal definition of accusation.

Carrot Eater, you’ve successfully painted yourself into a corner here. If the words they use are that easily interpreted, that makes the use of paraphrasing and the choice of words to ascribe to them they didn’t use even less defendable, not more so.

The point is obvious,

The obvious point seems to be that while Lucia is able to say that EM Smith is off doing irrelevant things, she is for some reason unwilling to even criticise Watts/d’Aleo at all for taking EM Smith’s material, and publicising it in ways that even he does not, on his own. Why, is beyond me. The written words are staring you in the face. ‘Purposefully’.
‘Policy-driven Deception’.

“What’s left to read in to what you say?”

Carrick, it’s the plain meaning that’s being read out of the quoted words that’s the problem.

Let’s also sample from the d’Aleo only screed, shall we?

NOAA is seriously complicit in data manipulation and fraud. After the Climategate emails were leaked, the East Anglia Hadley Centre has been the focus for data obstruction, destruction and manipulation issues and Phil Jones has temporarily stepped aside during a three year investigation as director of the Hadley Climatic Research Unit (CRU) until the completion of an independent Review resulting from allegations of inappropriate scientific conduct.
But CRU’s Director at the time Phil Jones acknowledges that CRU mirrors the NOAA data. “Almost all the data we have in the CRU archive is exactly the same as in the Global Historical Climatology Network (GHCN) archive used by the NOAA National Climatic Data Center.”
NOAA appears to play a key role as a data gatherer/gatekeeper for the global data centers at NASA and CRU. Programmer E.M. Smith’s analysis of NOAA’s GHCN found they systematically eliminated 75% of the world’s stations with a clear bias towards removing higher latitude, high altitude and rural locations, all of which had a tendency to be cooler. The thermometers in a sense marched towards the tropics, the sea and to airport tarmacs.

I think what we observe here is the inability of Carrick and Lucia to admit an error.

Quite clearly it is not a requirement of the word ‘fraud’ for it to be criminal behaviour.

Quite clearly Watts and D’Aleo accuse NASA, NOAA, CRU and others of being part of a conspiracy “to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.” and that this conspiracy is a “Policy driven agenda”.

This fulfills all the criteria for the first definition of the word fraud.

Carrot Eater:

Carrick, it’s the plain meaning that’s being read out of the quoted words that’s the problem.

Again you’ve contradicted yourself: If it’s plain meaning, you don’t need to “read out” anything, just provide the quote.

If you think it is an implication (even a “plain meaning”) of what they said, it is intellectually honest to label this as your opinion of the implications of what they said.

It is not intellectually characterize your opinion of what they said as if it where what they said.

Again this is plain obvious, I don’t see how you could argue with that, unless you have a secret love for bugs or something.

Putting it all together, there has been an accusation that.:-

NOAA did, possibly under the direction of the WMO, conspire to systematically and purposefully remove selected climate measuring stations around the world since around 1990, towards overstatement of warming, for the purposes of policy driven deception.

This is all their own words.

Carrot Eater:

Let’s also sample from the d’Aleo only screed, shall we?

Why shouldn’t we? You think I’m trying to block you from doing that?

Thanks Bugs

Now let’s see if the “humbugs” Lucia and Carrick will admit their error.

Carrick:
You missed what I’m saying. I’m saying the plain meaning of the SPPI words are staring you in the face. The d’Aleo only document, as well.

I’m saying Lucia and you are taking those statements, and somehow removing what’s there, in your acrobatic reading of it.

Why shouldn’t we? You think I’m trying to block you from doing that?

Well, you see d’Aleo’s words there. I suppose we’re to learn that ‘fraud’ doesn’t really mean fraud, or is maybe a ‘weak’ usage of the word, because it isn’t necessarily being used in the context of United States Code Title X, or something like this?

Carrick:

When I say “read out”, I mean people are taking the very words you see there, and the plain meaning therein, and trying to remove that meaning.

Why shouldn’t we? You think I’m trying to block you from doing that?

So perhaps now we’ll learn that ‘fraud’ doesn’t actually mean fraud, or maybe only in a weak sense at the very bottom of the dictionary? Perhaps because it doesn’t necessarily fit United States Code Title xyz?

The accusations here are about as serious as they get in the academic world. And d’Aleo must know it; that’s why it was pumped up so much. That’s how we get to TV shows and this http://www.spaceref.com/news/viewpr.html?pid=30000.

NOAA is seriously complicit in data manipulation and fraud.

bugs:

This is all their own words.

Do you recall my original comment?

Instead of taking the risk of misrepresenting Watts (a polite way of saying bugs outright lied about what Watts actually said), bugs should provide direct quotes of the passages in question.

And that was so difficult for you to post?

You may disagree with my representation of what you said as a “lie”, Nathan could too, Carrot Eater could 3, but doesn’t the fact there appears to be a disagreement here over literary interpretation bolster my argument that it is more effective to directly quote people than to paraphrase them, especially in the manner done by bugs?

You guys are all off on some side track about whether what bugs said was an accurate representation of a document. If his goal in writing that statement was to engage people in a debate over that, he succeeded. I’m pretty sure that wasn’t his goal, so it was a failed communication.

The conversation did get side tracked, and precisely because he choose to “make up” what they said instead of quoting them directly.

Carrick

So does Bugs’ quote above finally convince you? You didn’t make that clear.

Carrick, the real problem is that you and Lucia won’t admit the obvious point that they accuse many organisations of conspiracy and fraud.

“but doesn’t the fact there appears to be a disagreement here over literary interpretation bolster my argument that it is more effective to directly quote people than to paraphrase them, especially in the manner done by bugs?”

Actually Carrick, what it does it make you appear to be stubborn. Just because you disagree over the ‘literary interpretation’ doesn’t mean there is genuine confusion over what was meant by the words. It just indicates your inability to comprehend English.

Carrick, the real problem is that you and Lucia won’t admit the obvious point that they accuse many organisations of conspiracy and fraud.

Let alone that; I still haven’t heard Lucia even criticise the publication of the underlying substance; forget the accusations of malicious intent. Carrick has at least been clear that it’s “an ill-considered, hastily cobbled together piece of crap.”

Re: carrot eater (Mar 16 17:25),

Lucia: Seriously. Go on WUWT. Say that you have evidence that scientists at GISS

Look: I agree there are accusations of fraud in comments many places. I’m not going to google search, but I have no doubt these accusations exist in comments at WUWT. Heck, I’m sure they exist in my comments areas.

I dispute those accusations of fraud. The word is far too strong. I know the word can mean something that is not all that strong– as in the final bullets in Nathan’s defintion. But I think it’s not useful to use that word with those mild defnitions when they can be construed otherwise.

But by the same token, I am not going to take paragraphs that do not amount to the strong definition and which do not contain the word “fraud” and say the person who wrote them was making accusations of fraud. I won’t do it even if the behavior would fit the milder defnitions of fraud and not the criminal ones. There is too much possibility that people will think we are accusing the writers of much more than they did.

Let’s also sample from the d’Aleo only screed, shall we?

I’ve already said d’Aleo did accuse people of fraud. I said so long, long ago. But it seems important to a few people here to pin it on Watts who did not use that word.

he is for some reason unwilling to even criticise Watts/d’Aleo at all for taking EM Smith’s material

What are you talking about? I’ve criticized Watts. You just want me to criticize him for certain specific things using words that I are overbroad and which take on meanings that go beyond what we can criticize him for.

Nathan–

Quite clearly it is not a requirement of the word ‘fraud’ for it to be criminal behaviour.

I didn’t say an act must be criminal for someone to use the word fraud.

I am telling you that I avoid the word in contexts where we mean to describe an act that is neither criminal, nor even much above ‘humbug’ and where we think people might infer the stronger meaning contained in the dictionary.

If you want to use words that can be misunderstood to have the stronger meaning– fine. If you want to state your opinion that the stronger word applies, fine. If you want to “prove” that I must use that word because you have shown the weaker meaning applies, I am going to say I will not use the word in that circumstance. To do so would convey a meaning I do not intend to convey and which I consider untrue.

As far as I can tell, your gripe here in my comments has little to do with what Watts or D’Aleo did as your desire to try to force me to use a word whose meaning may be interpreted to mean something stronger than warranted simply because you prefer that word to statements that are more nuanced.

If you don’t like my position, fine. But don’t try to put words in my mouth.

Nathan:

Carrick, the real problem is that you and Lucia won’t admit the obvious point that they accuse many organisations of conspiracy and fraud.

No, the problem is you simply have no idea what I was actually objecting to or even trying to converse about.

You see this as a discussion over whether Watts/D’Aleo accused some unnamed scientists of fraud in their SPPI document.

But it was never about that,for me, it was about the most effective means of communicating what other people said.

If bugs wanted to directly quote them using the sentence you found (and which he likely never saw before then, knowing him):

NOAA did, possibly under the direction of the WMO, conspire to systematically and purposefully remove selected climate measuring stations around the world since around 1990, towards overstatement of warming, for the purposes of policy driven deception.

and then wanted to hop on to his band wagon (I’m rewording bug here):

What Watts and D’Aleo have done is tantamount to accusing the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science!

I’d have thought this was effective writing and possibly even agreed with it (mostly).

Didn’t you say you were actually trained in technical writing? There should be a bell ringing in there some place. hello? anyone at home?

all the data centers, most notably NOAA and NASA, conspired in the manipulation of global temperature records to suggest that temperatures in the 20th century rose faster than, in reality, they actually did.

Around 1990, NOAA began weeding out more than three-quarters of the climate measuring stations around the world. They may have been working under the auspices of the World Meteorological Organization (WMO). It can be shown that they systematically and purposefully, country by country, removed higher-latitude, higher-altitude and rural locations, all of which had a tendency to be cooler.

Interestingly, the very same stations that have been deleted from the world climate network were retained for computing the average-temperature base periods, further increasing the bias towards overstatement of warming by NOAA.

Everything I put in that statement is drawn from their own words. I think I managed to retain the meaning.

IIRC, this is wrong.

These data were then used to determine the global average temperature and to initialize climate models.

The models are not currently initialized to the current temperature.

For crying out loud. Stop talking about talking, and actually address what’s in front of you.

How on earth does this not amount to about as serious a charge of scientific misconduct as there can be, short of accusing somebody of completely making numbers up out of the blue? There’s nothing ‘weak’ about this.

The models are not currently initialized to the current temperature.

I’ve been letting that one pass. The decadal runs of Latif/Keenlyside and Smith are initialised in a sense, though it’s the oceans that matter there, I think.

bugs, we agree that it’s a very shoddy document and says very little about the understanding of climate science of anybody who wrote the document or reviewed it (searching for polite turns of phrase here >.>).

But don’t you agree that if you had blockquoted them as you did above, then followed it with your own summary “What Watts and D’Aleo have done is tantamount to accusing the scientists of helping to orchestrate a massive fraud and conspiracy unprecedented in the history of science!” it would have been vastly more effective?

Don’t look at all criticisms as argumentative.

carrot eater:

For crying out loud. Stop talking about talking, and actually address what’s in front of you.

Lucia and I had no trouble doing both.

It’s you one-track-mind types that were having trouble keeping up.

There’s the discussion of what W&A said. And then there’s the discussion about what’s the honest and effective way of presenting it.

You should be capable of doing both, and being able to delineate which topic is which. Don’t go “crying out loud” if you weren’t following it.

bugs, we agree that it’s a very shoddy document

I am not sure whether Lucia agrees with that.

Can anybody find a transcript of the TV shows they did for KUSI?

I should coin a new eponymous Internet Law: Any Blackboard thread with over 200 posts inevitably dissolves into a somewhat meaningless argument about a mostly semantic point

Zeke,

I think that that law holds way beyond the Blackboard …

Carrick

Bugs said they claimed fraud, you said they didn’t. We (bugs, Carrot Eater and I) then when about demonstrating that they did. That’s what I thought the discussion was about.

If you were just discussing the efficacy of modes of communication, then you didn’t do it very well.

Nathan (Comment#38300) March 16th, 2010 at 7:30 pm
And you told me you were a geologist! lol

Carrick

I know now what you were disucssuing…

However, just out of interest, do you agree that they accuse NASA, NOAA, CRU , and other data centers of fraud and conspiracy

Liza

Yes, I am a geologist, I edit geological work.

Re: carrot eater (Mar 16 18:13),

Let alone that; I still haven’t heard Lucia even criticise the publication of the underlying substance;

Well… no one’s dropped by comments claiming the document is good, have they?

The document strings together unsupported claims, some clearly wrong ideas (‘the march’ ) along with some correct observations whose import is not obvious.

Re: Zeke Hausfather (Mar 16 18:54),
I think this is true.

Re: carrot eater (Mar 16 17:38),
I did already say D’Aleo accused people of fraud. I’m not sure if I used screed; I’d have to google to check.

Re: carrot eater (Mar 16 18:24),

The models are not currently initialized to the current temperature.

Climate models aren’t initialized to current temperatures. But I’m not quite sure that has to do with most of the recent discussion.

Nathan-

Yes, I am a geologist, I edit geological work.

Interesting definition of “geologist”.

“Il n’ya pas de hors-texte” – “There is nothing outside the text” (Jacques Derrida).

lucia (Comment#38305) March 16th, 2010 at 8:17 pm

Climate models aren’t initialized to current temperatures. But I’m not quite sure that has to do with most of the recent discussion.

From the paper.

The thermometers were marched towards the tropics, the sea, and airports near bigger cities. These data were then used to determine the global average temperature and to initialize climate models.

I find it impressive that a model can work pretty closely by itself what the temperature should be without being told what it is.

“Interesting definition of “geologist””

I didn’t read it as being a defintion, but what he does as a geologist.

bugs–
Was the discussion about every single possible line in the SPPI article? Had me fooled!

I find it impressive that a model can work pretty closely by itself what the temperature should be without being told what it is.

I’m not sure what you mean by “pretty close”, nor why you would be impressed by the level of agreement, nor what you mean by “without being told what it is.”

It’s not entirely true that the model isn’t initialized with temperatures near the expected values. The AOGCM’s in the AR4 weren’t initialized using current 2010 temperatures. But they are initialized with estimates of historic temperatures. The modelers trying to start models with conditions in 1800 -1900 would would rather guess temperatures they think are close to 1800 or 1900 temperatures and spin up from that. Why guess 2010 temperatures?

I don’t think anyone really thinks of initializing with using the modelers guess for the right temperatures is the same as telling the models what the temperature are “supposed” to be, but if you want to use that sort of language, it seems to me that’s what’s done. They just don’t use current temperatures. Weather models may– but climate models don’t.

Lucia

“Yes, I am a geologist, I edit geological work.
Interesting definition of “geologist”.

It would have been if I was attempting to define the word ‘geologist’. However, I was attempting to show Liza there was no discrepancy between my claims of being a geologist AND an editor. And by ‘discrepancy’ I don’t mean that by the legal definition.

Hey, have you accepted that Watt’s and D’Aleo accused NASA, NOAA and CRU et al. of beings frauds and conspirators?

Nathan–

However, I was attempting to show Liza there was no discrepancy between my claims of being a geologist AND an editor.

You failed. Based on your words, you appear to be an editor, not a geologist. Geologists do or study geology. If you do either, let us know.

Zeke:

I should coin a new eponymous Internet Law: Any Blackboard thread with over 200 posts inevitably dissolves into a somewhat meaningless argument about a mostly semantic point

I call “Zeke’s Law!”

I love it.

Re: Carrick (Mar 16 21:23),
There is a reason I use a plugin that auto-closes comments after a certain amount of time. . .

Lucia

“Geologists do or study geology. If you do either, let us know.”
Well, to edit geologists work here, I must study it, no?

To be honest, I don’t care if you don’t think I am a geologist. Nor do I care if Liza does.

Nathan

Well, to edit geologists work here, I must study it, no?

No. I’m not suggesting you care if we think you are a geologist. Maybe you are an editor and a geologist. But so far, you seem to be saying that editing geology papers makes you a geologists; that’s just not so.

Lucia

“But so far, you seem to be saying that editing geology papers makes you a geologists; that’s just not so.”

No, I didn’t say that.

I made no claims about what constitutes a ‘geologist’.

Lucia

Here you go, I think you wre using the legal definition of Geologist before.

Here’s the Dictionary.com version

“ge·ol·o·gist   /dʒiˈɒlədʒɪst/ Show Spelled[jee-ol-uh-jist] Show IPA
–noun
a person who specializes in geologic research and study.”

Nathan–
I have never posted any legal definition of geologist.

Do you have a degree in geology? B.S.? M.S.? Ph.D.?
Other than editing papers, do people pay you to do geology? If yes, what do they pay you to do?

Lucia

I didn’t add a smiley so you missed the joke on the ‘legal defn’

Yes I have a BSc (hons) in geology, and a grad cert in hydrogeology. I also have a BA in Theatre Arts too.

“Other than editing papers, do people pay you to do geology? If yes, what do they pay you to do?”
BORING. You seriously want to know all this?

But yes, they pay me to edit more than papers, I edit geological maps too. And it is a requirement of my job that I keep my understanding of geology up to date. See? I am required to maintain my understanding of geology, hence I need to study it. Now I don’t know what private visions you have of what it means to edit geological manuscripts and maps, but rest assured I am required to be a geologist to qualify for my job.

Nathan–
Rest assured that there are plenty of copy editors at national labs who copy edit papers for geologists. Those papers are written by someone else. The editors are not required to have the tiniest snippet of training in geology. If you simply suggest that editing papers in geology means you must study geology, that’s wrong.

Given the weirdness of your various answers, I wanted to know what you do or have done that actually would cause people to consider you a geologist.

It seems odd to me that with a degree in geology, you edit the work of others rather than doing your own geological investigations. But if that is so, that is so.

Lucia

I told you before I wasn’t a copy editor.
Well, odd or not, that’s me

Actually I am looking to change my job – had an interview a couple of weeks back – so cross your fingers for me! If I get it I won’t have much time to bother you anymore, see we both benefit!

Zeke,

Yikes, ask and ye shall receive. Muchas gracias!

And it confirms my suspicions — remove the 60-90N stations, and the NH trend is virtually unchanged. So the diff between the NH and SH lies primarily below 60N.

Thanks again, sir.

Carrot eater writes:

Ron was showing what happens if you do exactly what GISS does. If you do use GISS’s UHI adjustment, then there is no significant bias introduced by using the smaller post-1992 subset of stations. If you don’t use any homogenisation at all, you still see practically no bias (Tamino, zeke). Either way you do it, same result, and it was useful to see it either way, so that we now know that. Perhaps you didn’t see it, but Zeke and Tamino were criticised for not doing exactly what GISS did; hence Ron and CCC.

You realize that your statement is non-responsive, right?

Then, the math of your careful academic claim actually requires that every single urban station be rural within the recent past. If it’s actually just a fraction, your claim of inconsistency falls apart.

That’s positively daft and false. Its not necessary for every single urban station to have been rural–a point I previously made very expressly. Its quite clearly a continuum.

As for your initial careful academic claim of inconsistency between Zeke’s initial UHI results and Spencer, yes, I think it’s extremely weak. First, it requires an uncritical retrieval of a number from this mess, without using confidence intervals

A 2x discrepancy is well within the range of that data; consequently I deem the idea worth discussing. I didn’t prove that it was so; I said Zeke’s analysis was not convincing. Establishing the details more closely to exclude alternatives is just one such step to making Zeke’s analysis convincing.

As it happens, the biggest issue with Zeke’s UHI work-in-progress was found by Harry, as he found problems with the metadata on the population density. I don’t know whether Harryw2 is an academic, but I appreciate his carefulness.

“found”? I TOLD Zeke about this problem. A point I reiterated on this very thread in a post you’ve previously commented on, so I cannot guess: are you just being an ass or dishonest?

Lucia asks:

Do you have the specific reference that explains how blending is done by any group in particular? Hansen and Lebedeff (1987) does not describe anything I would associate with the word “blending”.

I understand this to be an emergent property…

McIntyre discussed hansen’s later ’99 method here: http://climateaudit.org/2008/0.....l-context/

He’s posted many times on this subject, so this one link should not be regarded as exclusive.

In a similar vein:
If we look at some of the USHCN adjustments (e.g., Karl and Williams 1987): Blending is an emergent property here.

Karls method works like this:

Find a cluster of 20 near stations. Pick the two neighbor series that closely correlate, then take a weighted sum of differences between the stations–set the weights proportionate to the confidence internals; iteratively add in stations until the CI expand. Then apply the resulting difference to the series for all subsequent years.

Repeat with other stations, repeat with every discontinuity moving forward in time, using the adjusted series going forward.

Karl et al use more pages to explain, so good luck going forward on your own.

Laura

CE:
“As it happens, the biggest issue with Zeke’s UHI work-in-progress was found by Harry, as he found problems with the metadata on the population density. I don’t know whether Harryw2 is an academic, but I appreciate his carefulness.”

Laura:
“found”? I TOLD Zeke about this problem. A point I reiterated on this very thread in a post you’ve previously commented on, so I cannot guess: are you just being an ass or dishonest?

I think you may owe carrot an apology.

“As it happens, the biggest issue with Zeke’s UHI work-in-progress was found by Harry, as he [harry] found problems with the metadata on the population density. I don’t know whether Harryw2 is an academic, but I appreciate his carefulness.”

Laura : “found”? I TOLD Zeke about this problem. ”

1. Carrot is refering to harry finding the problem

I do believe that harryw2 found this problem. In fact I think
( but carrot can correct me ) zeke harry ron carrot and I were discussing this else where

http://rhinohide.wordpress.com...../#comments

but go ahead show me where

1. you told zeke about this problem
2. carrot means other than what I said

Ok, Laura

I found where you said you told Zeke.

My sense is that CE is not an ass, he’s probably just remembering our discussion over on whiteboard.

Sequence:

Laura:

“I mentioned to Zeke several days earlier that the GHCN metadata was mostly garbage. An opinion which can be formed simply by checking the consistency of the meta-data against other records. e.g., google earth. A small but random sample suggested to me that this data was not very clean. I advised Zeke at the time that it seemed unlikely a difference in trend would be identified using this meta-data–little better than if it had been a random classification.”

ZEKE:
Oddly enough, I don’t completely trust the metadata either. That’s why I turned to two completely independent datasets in my analysis (satellite nightlights and GPW population density data) in addition to the station metadata designation. The only case I can see being a real problem is if station location changed and was not reflected in the metadata, since the nightlight and GPW lookups use station lat/lon. However, I suspect its probably not too common a problem, and if it does occur it would more likely be an urban/bright/dense station moving to a rural/dark/low pop area rather than vis-versa.

Later that day over at whiteboard:

2010 March 10 at 1:18 pm | #16 Reply | Quote
Ugh, I was hoping that the lat/lon coords were reasonably accurate. Oh well, we might have to limit our site characteristics analysis to USHCN for a bit, and hopefully they will correct some of the more egregious divergences in GHCN v3…
Someone tell NOAA to hire some interns to spend a summer Google Earthing the whole GHCN networks

Looks like Zeke was on the alter for a metadata problem
and when the guys at whiteboard DETAILED that problem
he said “ugh”

So laura thinks she told told him about the exact problem
but she looks to be wrong about that given Zeke’s expression over at WHiteboard upon being informed by Oneill it appears.

Of course one needs to normalize time stamps and all.

I’m reading the 1999 paper. GISS analysis of surface temperature change, J. Hansen, R. Ruedy, J. Glascoe, and M. Sato, NASA Goddard Institute for Space Studies, New York

It uses the Hansen Lebedeff 87 method; section 5 discusses the UHI adjustment.

The adjustment is based on the assumption that human effects are smaller in rural locations. We retain the unadjusted record and make available results for both adjusted and unadjusted time series (section 10). The homogeneity adjustment for a given city is defined to change linearly with time between 1950 and the final year of data and to change linearly with a possibly different slope between 1950 and the beginning of the record. The slopes of the two straight line segments are chosen to minimize the weighted-mean root-mean-square difference of the urban station time series with the time series of nearby rural stations. An adjusted urban record is defined only if there are at least three rural neighbors for at least two thirds of the period being adjusted. All rural stations within 1000 km are used to calculate the adjustment, with a weight that decreases linearly to zero at distance 1000 km.

It’s a odd little thing, but off hand, I don’t see why it should introduce any big trends at the precise time stations are dropped. Mostly, I think that might just freeze the adjustment and possibly make past temperatures in GISSTemp updates jump around a bit less! But loss of thermometers could do some weird things if over time, we no longer had at least three rural neighbors for at least two thirds of the period being adjusted. (Seems like this could also happen if a rural area developed into a sub-urban aread.)

This still doesn’t read like “in filling” or even “blending”. (Maybe NOAA infills or blends? I haven’t read those. I’m reading one group at a time!)

Following the conversation, harryw2 deserves credit.

harryw2 found and even diagnosed the problem with the population density issue. First sniffed at it at (Comment#37126) March 9th, 2010 at 12:39 pm; got a bit closer at (Comment#37305) March 10th, 2010 at 9:03 am , and put it together at (Comment#37517) March 11th, 2010 at 8:39 am, as well as at Broberg’s.

Laura, on the other hand,

I mentioned to Zeke several days earlier that the GHCN metadata was mostly garbage.

Everybody already knew there were some issues with that; that’s why Zeke, Broberg and Mosher are bothering to come up with a new source of metadata. That’s why Hansen is using nightlights.

Lucia

This still doesn’t read like “in filling” or even “blending”.

Not in any sense of the word that I’d use “in filling”. The UHI adjustment, maybe you could call ‘blending’.

Anyway, you’re a bit outdated there. The breakpoint in trend is no longer fixed at 1950, but is allowed to move wherever it needs to go, to minimise the optimisation objective function.

Nick Stokes did a nice little poor-man’s emulation of it here
http://moyhu.blogspot.com/2010.....ments.html

(Maybe NOAA infills or blends? I haven’t read those. I’m reading one group at a time!)

Not in GHCN. But the USHCN does have infilling. Look up “FILNET”. If you look in USHCN adjusted files, there are sometimes flags next to a number. This tells you there was infilling.

Flags described here:
http://cdiac.ornl.gov/ftp/ushc.....readme.txt

GISTEMP removes anything with certain flags.

You realize that your statement is non-responsive, right?

Not even close. Some people thought it’d be important to see what happens if you do exactly what GISS does. So Ron did. Why you object to that is a mystery. Everybody else seems to see the value in it.

That’s positively daft and false. Its not necessary for every single urban station to have been rural–a point I previously made very expressly. Its quite clearly a continuum.

Your initial confident statement of inconsistency (with no caveats whatsoever), and your initial quantification absolutely require this lack of continuum. And you know this, now at least.

If you were being such a careful persuasive academic, you’d have said “making the assumption that all the urbans used to be rural, I get number X, but relaxing that assumption changes it in this direction, so I have here a limiting value”

A 2x discrepancy is well within the range of that data;

A lot of things are within the range of those data. Including perfect consistency.

ugh.

to clarify myself

Not in GHCN.

GHCN adjusted values in v2.mean.adj very much does have homogenisation based on neighboring stations, so you could perhaps call that blending. But GISS doesn’t use those GHCN adjustments.

GISS does however use USHCN adjustments, described in two papers by Menne et al (2009). GISS accepts the homogenisation, but excludes most of the in-filling of missing points.

Re: carrot eater (Mar 17 08:43),
Thanks. Laura pointed to the 1999 paper, so I was going to look at that. Looking at it was worthwhile even if Hansen no longer does that. I’ll read the 2001 paper.

It’s good to read them in order anyway. Obviously, Hansen and Lebedeff 1987 do a more thorough discussion of the reference station method than the 1999 paper. They provide a recap in the 1999 paper and do cite back. I figure the same will occur with the 2001 paper.

Yes. I guess we can call the UHI adjustment “blending”.

Yes, do read them in order.

I think I might get Laura S’s objection with Ron Broberg now. It requires a misunderstanding of what Ron did.

Ron separated out the input subsets (rural vs urban, pre drop vs post drop, high elevation vs low elevation etc) before the UHI “blending” step. Not after.

So his subsets were truly independent of each other. The data went through the UHI adjustment, but the high elevation data was only UHI adjusted using other high elevation data.

In some of the subsets, the UHI step might not have done much. When feeding rural stations only, it probably didn’t do anything, as most of the urban-flagged stations were already removed. In other subsets, there might not have been enough rural stations nearby to do the adjustment at all, in which case I think they just toss the urban station out (check on that).

Ya carrot.

It’s funny. Back in 2008 I started to look at nightlights

http://climateaudit.org/2008/0.....fantastic/

I thought I found a bug in gisstemp but I was WRONG ( its in the comments there somewhere, a dude name conrad?? corrected me )
anyways, I found a few nightlights that were wrong, had some issues with the population data

http://climateaudit.org/2008/0.....ent-138647

Well, you make a suggestion in the comments is no assurance that somebody will take up that suggestion and do the fricking work for you.

Anyways, I personally think its great that more people on the warming side ( what r you, not that it matters ) are looking at these things.

GISS and CRU do not have the time to answer every little objection. But once they post their code ( or anybody posts code that comes close to their answer) then other people, like ron, ccc, nick stokes, et al et al, can quickly do a study and put out blog fires.

That’s really why I wish tamino would share his code UP front.
It would be great to run zekes cases in tamino code. or in giss code or had cru code. Same data set different methods.

I just think back to when JohnV posted his Opentemp. John and I rarely fought about politics or mann or any of that crap. he had code. editing a data set was easy. He would duplicate my results,
I would double check his. If people attacked his code I told them to take his code and improve it.

WRT metadata: have you noticed how no one quests spenser about the metadata for his temp series? For now, I’ll leave that question alone while people pick at the method of analysis, but in the end all roads lead to the metadata and data and audits of these. That shouldn’t stop people like zeke, roman,ron, jeff nick cru et etc from building methods to estimate numbers based on exist ing metadata and data.

[...] When NASA and Hadley/CRU create a global temperature record, they aren’t estimating the average temperature of the globe per se; rather, they are calculating the global temperature anomaly, the change in temperature from a base period. This is an important distinction, as using anomalies helps avoid a number of location- and site-specific biases that will affect the absolute temperature but not the change in temperature over time (for a more technical discussion of the benefits of anomalies vis-à-vis absolute temperatures, see these three posts). [...]

[...] to demonstrate the effects of losing station loss. There is no loss of global trends, although she notes that there can be a loss of information when trends are associated with features that are related [...]