The Holocene Deniers

[Thistlethorn]

Before starting the movie, I thought I would show everyone the inferred average temperatures of various places on the globe during the Paleocene-Eocene Thermal Maximum--a time of extreme warmth which didn't do anything to the earth in the way of destroying it.

Of course everyone today is scared to death of a very minor rise in temperature compared with this time.

Night, Movie starting.

Look, it's another straw man. Who ever said the earth would be destroyed?

 

[grmorton]

The median is 1degree F. That isn't that bad.

Physically that is bad, and it is worse when you consider the variance.

The histogram appears quite symmetrical indicating no directionality for bias.

The chi squared tests on the 20degrees, 10 degrees and 5 degrees differences also indicate no bias (there does not appear to be reason to believe that there is a significant difference in the number of days Belle Plain is hotter than Toledo versus when Toledo is hotter than Belle Plaine.[/quote]

The above is kind of silly. I can count several thousand more days of 5+ towards Belle Plaine. You say it isn't statistically significant, well it is PHYSICALLY significant.

If I have miscalculated the chi square statistic please tell me. Because as it stands now I do not see that there is any reason to claim a statistically significant bias above 5 degrees F difference.

I don't think you can take the measurements I posted and decide from that alone. At least that is my understanding of what you said you did.

As I said the histogram of differences looks quite symmetrical to me.

Belle Plaine warmer than Toledo by greater than 3 degrees-3588 times
Toledo warmer than Belle Plaine by greater than 3 degrees-1806 times.

that looks symmetrical? Not to me.Looks like there are more days that Belle Plaine is warmer than Toledo. There is also the substructure which occurs between station moves. I plotted the temperature difference histogram for 1984 to 1990. Note the decided asymmetry.

But I don't see there being a systematic bias. There are going to be errors. I've got two thermometers in my back yard that are about 3 feet apart and they often read differently. (But there is a systematic error in them and one of them, when it is above 90deg F will shoot up to 100deg F while the other stays around 95deg F.)

You just illustrated the problem. If the noise is greater than the signal, it gets very hard to see the signal. the signal in this case is 1.1 deg F of global warming, but the noise is 5 degrees (using your example above). I will never be convinced that it makes sense to say that the globe has warmed by 1.1 deg +/- 5 degrees.

A systematic bias, even if it existed, would still have to affect the TRENDS calculated based on these. If one consistently read 3 degrees higher than the other yet it still tracked with the other (ie showed an increase at the same rate as the other) it would still be effective for the present discussion.

Using linear regressions on sinusoidal data will always give a trend. The Migrant Mind: Regression Slope Voodoo

It is like an "offset" in a measurement instrument. If I were to move the little wheel on my bathroom scale so it consistently read 5 lbs heavier than the actual weight and then I proceeded to gain weight I could still tell you how much weight I gained, if not the absolute weight I was.

Not if the bias varies. From 1984 to 1990 the bias is towards Belle Plaine. The average of the temperature difference in that time is 1.68 deg F towards BP being hotter. But from 1990 to 1997 the average is -.9 deg F obviously Toledo being hotter.

So, to use your analogy, you are trying to figure out the trend when your wife keeps moving that dial on the scale and you don't know it.


You still haven't answered my question. Do you think that it is reasonable to believe that 25% of the days have a temperature gradient between these two towns which is greater than a strong cold front? Please answer the question. YEs or no--I won't stop asking so you might as well answer

 

[LifeToTheFullest!]

If everyone would get their panties out of a twist and start paying attention to data rather than to who said what when (and I am going to do that from this moment on. I will no longer reply to personal insults or to the oil issue), then people would see that the data isn't so perfect as the agw folks beleive.

Have 5 billion contributed to warming? Probably, but not necessarily only via CO2 which is the villain du jour. Albedo change due to land use is also an issue. As Carl Sagan pointed out long long ago, a change of the albedo of .01, it would account for any warming.

"Simple climate models (31) suggest
that if the global albedo changes from its
value of 0.30 by 0.01, a surface temperature
change of - 2 K will result." Carl Sagan, Owen B. Toon, James B. Pollack," Anthropogenic Albedo Changes
and the Earth's Climate", Science 206(1979),p. 1367

Note that this article is from long long before the current controversy and thus can't be claimed to be political on either side. NOte also that the 2 K is 2 C and that is greater than the 0.84 C of the IPCC.

So, what has happened to the world's albedo?

"The global CERES observations show a
small decrease of ~2 W/m^2 in shortwave
reflected flux, equal to an albedo decrease of
0.006. These results stand in stark contrast to
those of Pall2 et al. (4), which show a large increase of 6 W/ m^2 or an albedo increase of
0.017, as shown for comparison in Fig. 1."

This is 60% of Sagan's 2 degree change for a change of .01. Thus we can expect that from albedo alone we should have 1.2 deg C warming. Of course, no one discusses this they all want to blame CO2

E. Palle´, P. R. Goode, P. Montane´s-Rodrıguez, S. E. Koonin, “ Changes in Earth’s Reflectance over the Past Two Decades, Science 304(2004), p. 1299

"We correlate an overlapping period of earthshine measurements of Earth’s reflectance (from 1999 through mid-2001) with satellite observations of global cloud properties to construct from the latter a proxy measure of Earth’s global shortwave reflectance. This proxy shows a steady decrease in Earth’s reflectance from 1984 to 2000, with a strong climatologically significant drop after 1995. From 2001 to 2003, only earthshine data are available, and they indicate a complete reversal of the decline. Understanding how the causes of these decadal changes are apportioned between natural variability, direct forcing, and feedbacks is fundamental to confidently assessing and predicting climate change.”



But everyone wants to focus ONLY on CO2.




Thank you. As I said to Thau, regardless of whether I am right or wrong, such behavior is absolutely attrocious. But ignoring data and other possibilities is equally attrocious, IMO.



I am a bull in a china shop. I have a bad habit. I respond in kind (at least as I think of the kind--it is bad, I know it, but it is unlikely to change). And you are not the first one to claim that I am a bull in a china shop. Luckily, I don't care about china. I care only that people don't ignore data. It is what YECs do and it is what AGW does, IMO. YECs are not dangerous. They are very very wrong, but they won't tax me out of existence. AGW is dangerous if it is wrong, so it is important that ALL the data be examined.

Thanks for this. I'll look into the Sagan refs.

 

[Contracelsus]

The above is kind of silly. I can count several thousand more days of 5+ towards Belle Plaine. You say it isn't statistically significant, well it is PHYSICALLY significant.

Please explain to me how the chi square calculation is in error.

I don't think you can take the measurements I posted and decide from that alone. At least that is my understanding of what you said you did.

I would greatly appreciate a specific explanation as to how I have misapplied the chi square test. Only this way can I learn how to properly apply the statistics. A reference citation and possibly a clear explanation as to what part of the chi square test I messed up on would be most helpful to me. (Thanks in advance)

that looks symmetrical? Not to me.

The histogram of all the differences look symmetrical.

You just illustrated the problem. If the noise is greater than the signal, it gets very hard to see the signal.

Actually not necessarily so. I took some fake data and ran a comparison. I made up 13 "stations" and gave each of them "temperatures" such that each station has a standard deviation of about 1.3 deg F. I was able to make a trend that appears significant with a slope of 0.5degrees. So in this case each x-value on the graph has a standard deviation greater than the slope but the curve is still statistically significant.

the signal in this case is 1.1 deg F of global warming, but the noise is 5 degrees (using your example above). I will never be convinced that it makes sense to say that the globe has warmed by 1.1 deg +/- 5 degrees.

I don't believe that is how it is done. But again, I am not a scientist and I don't do this for a living.

Using linear regressions on sinusoidal data will always give a trend. The Migrant Mind: Regression Slope Voodoo

What is sinusoidal about the graphs we are talking about here? I might have missed something.

You still haven't answered my question.

I did answer that in a previous post and I think I clarified my point with the graph above.

 

[grmorton]

Look, it's another straw man. Who ever said the earth would be destroyed?

With all due respect, I must stand by my former statement that they are claiming that. Do you know what a runaway greenhouse effect is? It is where a positive feedback loop happens where rising temperature causes more greenhouse gases and those gases in turn cause higher temperatures. Such an event is why Venus is a baking lifeless desert.
Now James Hansen is saying that kind of nonsense:

Robin McKie, "President 'has four years to save Earth'" The Guardian's The Observer, Sunday 18 January 2009
Hansen said current carbon levels in the atmosphere were already too high to prevent runaway greenhouse warming. Yet the levels are still rising despite all the efforts of politicians and scientists
President Obama 'has four years to save Earth' | Environment | The Observer


And this

Christopher Flavin, “Commentary: Reconciling Poverty, Sustainability, and the Financial Crisis” WorldChanging TeamOctober 1, 2008 2:02 PM

""The following is adapted from a speech given by Worldwatch Institute President Christopher Flavin at a high-level United Nations event on September 25, 2008.
Across large areas of the Indian subcontinent, diminishing supplies of fresh water are undermining food production and leaving people with inadequate drinking water.
And from the Arctic to the Equator, the world's climate is changing rapidly - and undermining ecological systems on every continent, from forests to oceans and fresh water. Many scientists believe that a dangerous climate tipping point may be near-unleashing a runaway greenhouse effect that would feed on itself for centuries to come.
The bottom line is clear: the inefficient, carbon-intensive, throwaway economy that was so successful in an earlier era is not suited to today's world. Our planet in now in mortal danger of an ecological collapse whose human impact would dwarf the financial collapse the world is now seeking to avoid.







Now, here is why we don't need to worry about it. This concept has already been experimentally tested on earth. It is called the Paleocene-Eocene thermal Maximum. CO2 was over 1000 ppm (today we are at 385 ppm). The global average temperature of the earth was well in excess of 5 deg C above that found today.

Scott L. Wing and Elizabeth C. Lovelock," Rapid global warming and floral change at the Paleocene-Eocene boundary (in , Anonymous,), "Geological Society of America (GSA), United States, Boulder 2007

"At the onset of the Eocene the Earth warmed by 4-8 degrees C over a period of 10-20 ky."

What happened? Nothing

The term "runaway greenhouse effect" has a particular scientific meaning and by using it Hansen is saying something scary, but also ludicrous given the past history of the temperature and CO2 content of the earth. We have seen 3000+ ppm CO2 in the Cretaceous, 1000 ppm in the Eocene and we didn't have a runaway greenhouse.

Digging out an old text R. M. Goody's Atmospheres:

"The idea is that the growth of the Earth's atmosphere was halted at at moderate pressure and temperature by the formation of the oceans. Water condensed, and carbon dioxide dissolved in the water, furnishing a suitable environment of the onset of life. On Venus, however, the runaway greenhouse effect prevented oceans from ever forming. surface temperatures rose so high that atmospheric gases could react rapidly with surface rocks, leading to the chemical equilibrium between atmosphere and surfacce that we have already described. " R. M. Goody, Atmospheres, Prentice Hall, 1972, p. 134

Global warming hysteriacs modify that definition above and have the earth's temperature spiralling out of control killing life. That is why Hansen keeps talking about 'tipping points'. so, Thistlethorn, you are quite wrong that no one has predicted the end of the earth. AGW is an escatological movement.

 

[grmorton]

Please explain to me how the chi square calculation is in error.

As I understand it you didn't even download the full data set. You simply used my counts of a few randomly chosen points. Without seeing a full account of what you did, I find the above request impossible.

The histogram of all the differences look symmetrical.

So, having more on the postive side of the line and fewer on the negative side of the line is symmetrical? Are you aware that they ought to be symmetrical around ZERO?

I did answer that in a previous post and I think I clarified my point with the graph above.

Lets try again. YOu missed the point of the question. It isn't a question about the distribution of the data. Focus on physics. Cold fronts and warm fronts are regions where the temperature changes rather quickly. A strong cold or warm front will change the temperature at the rate of 0.2 deg F per mile. Anything greater than 0.2 deg per mile (1/5 of a degree per mile) is something quite remarkable. Such temperature gradients cause huge thunderstorms, huge winds, tornados etc. Yet, the temperature records show no such effects during these HUGE temperature gradients.

So, I would respectfully suggest that you not focus on distribution of the data here because the question isn't about distribution. It is about physics It is about why we don't see huge thunderstorms which are always caused by such temperature differences on 25% of the days?

Another physics issue. High temperatures create high pressure cells. Low temperatures form low pressure cells. Wind flows from the high pressure to the low pressure. That is simple physics. So, when Belle Plaine is hot it should have a relatively higher pressure than at Toledo and the wind should flow from Belle Plaine towards Toledo (subject of course to the Coriollis force) So, between 1984 and 1990 when Belle Plaine is on avereage 1.68 deg F hotter according to the temperature records we should find a predominate flow of air from Belle Plaine to Toledo. But then from 1991 to 1997 when the temperature reversed and Toledo was on average 0.9 deg F hotter than Belle Plaine, we should see a reversal of the wind direction.

Wanna bet what we find? Not that predicted by the physics of temperature.

Those who live along the coasts of the world, like me, know that when the land heats up in the day, the wind blows mostly offshore towards the cooler water. But at night when the land cools down rapidly, the winds blow from the warm sea onto the cool land. This is observational fact.

Using the above, IF the temperature record at Belle Plaine and Toledo is valid, why don't we see the same thing? I can answer that--the temperature record is not valid.

 

[Contracelsus]

As I understand it you didn't even download the full data set. You simply used my counts of a few randomly chosen points. Without seeing a full account of what you did, I find the above request impossible.

That isn't how the chi square test works. To my understanding all I need is the count data. Unless you are saying you didn't give me an accurate count.

And, clearly I did download the entire data set since I generated a 20,959 point histogram.

So, having more on the postive side of the line and fewer on the negative side of the line is symmetrical? Are you aware that they ought to be symmetrical around ZERO?

When I say symmetrical I mean in a general sense. This histogram looks reasonably symmetrical. If you wish to haggle over how much skew there is I'm sure there's a way to calculate that.

Lets try again. YOu missed the point of the question. It isn't a question about the distribution of the data.

Actually it is all about the data distribution. I thought we were talking about the data.

If 25% of the data is outside of the + 3 degrees boundary that means that 75% of the data is within that boundary. 68% is generally the boundary for 1 standard deviation. So it looks to me like more of the data is within one standard deviation than if this were a regular normal distribution.

Focus on physics.

I keep thinking we are talking about the quality of the data. And data contain errors. That's my understanding of why there's a distribution of data.

Cold fronts and warm fronts are regions where the temperature changes rather quickly. A strong cold or warm front will change the temperature at the rate of 0.2 deg F per mile. Anything greater than 0.2 deg per mile (1/5 of a degree per mile) is something quite remarkable.

So am I to assume that the only data that would pass muster with your analyses are perfectly reproduced data that is tighter than a normal distribution?

Of that there is no error in the data at all?

So, I would respectfully suggest that you not focus on distribution of the data here because the question isn't about distribution.

Seems to me you were talking about the quality of the data. If the data has errors I believe the only way to deal with the quality issues of the data is to deal with the distributions.

It is about physics

But if we ignore the statistical treatment of the data then we are lost. In your field of research (oil?) do you always have perfect agreement of data points? Do they always have perfect reproducibility?

Another physics issue. High temperatures create high pressure cells. Low temperatures form low pressure cells. Wind flows from the high pressure to the low pressure. That is simple physics. So, when Belle Plaine is hot it should have a relatively higher pressure than at Toledo and the wind should flow from Belle Plaine towards Toledo

Perhaps I am not making my point clearly. I don't believe that on one particular day there was necessarily literally a 30degree difference between the two towns, I think these are often errors. I am just interested in how many errors there are in order to ascertain if the data is as bad as you have indicated.

So far I have no real reason to believe the data is that bad. Sure there are some errors. There should be errors, right? I mean this is real data.

I can answer that--the temperature record is not valid.

So what would be the best achievable data set? Would it have a perfect zero difference between the two stations for 60 years?

 

[Contracelsus]

I took the two Iowa station difference data and ran a Skewness calculation on it.

The value I got was:
-0.006249883

This is pretty low. Anything over 1 is considered significantly skewed.

skewness =sum(X[sub]i[/sub] − aveX)[sup]3[/sup]/n(sd(X)[sup]3[/sup])
Formula from:
(www.uk.sagepub.com/upm-data/26922_Wright_Chapter_One.pdf)

I also noted that according to this page:
1.3.5.11. Measures of Skewness and Kurtosis

"any symmetric data should have a skewness near zero"

So I'd have to say this a pretty symmetrical data set.

 

[grmorton]

For those who don't think there is a bias problem I want to post two sets of two towns. Coldwater Kansas and Ashland Kansas. Below is the picture of these two cities. One can see that for most of the time Coldwater (blue curve) is hotter than Ashland (red curve), Surely people can see that there is a bias. I can tell because the blue curve is always above the red curve. But notice that it wasn't that way before 1965. (Contracelsus should remember the dial on the scale being turned by his wife and his wife not telling him about it.) Surely even Thistlethorn can see that this is a problem--in 1981 the two towns merely 24 miles apart differed by 5 deg F for the YEARLY AVERAGE!!!

Graccus said he couldn't interpret raw data. I bet he can tell that there is a bias between the two towns. Shoot, even my 4 year old grand daughter can see that one curve is above the other.

The second picture is of Waterville Washington vs Wenatchee Washington. These two towns are 24 miles apart. Yet you can clearly see the bias between the two towns. But the bias shifts around 1987. Surely Contracelsus can see a bias here. (I have adiabatically corrected the data so this is the temperature which would have been measured at the same elevation.)

After these two pictures I have to delete some pictures in this thread.

 

[grmorton]

That isn't how the chi square test works. To my understanding all I need is the count data. Unless you are saying you didn't give me an accurate count.

And, clearly I did download the entire data set since I generated a 20,959 point histogram.




When I say symmetrical I mean in a general sense. This histogram looks reasonably symmetrical. If you wish to haggle over how much skew there is I'm sure there's a way to calculate that.



Actually it is all about the data distribution. I thought we were talking about the data.

If 25% of the data is outside of the + 3 degrees boundary that means that 75% of the data is within that boundary. 68% is generally the boundary for 1 standard deviation. So it looks to me like more of the data is within one standard deviation than if this were a regular normal distribution.



I keep thinking we are talking about the quality of the data. And data contain errors. That's my understanding of why there's a distribution of data.



So am I to assume that the only data that would pass muster with your analyses are perfectly reproduced data that is tighter than a normal distribution?

Of that there is no error in the data at all?



Seems to me you were talking about the quality of the data. If the data has errors I believe the only way to deal with the quality issues of the data is to deal with the distributions.



But if we ignore the statistical treatment of the data then we are lost. In your field of research (oil?) do you always have perfect agreement of data points? Do they always have perfect reproducibility?



Perhaps I am not making my point clearly. I don't believe that on one particular day there was necessarily literally a 30degree difference between the two towns, I think these are often errors. I am just interested in how many errors there are in order to ascertain if the data is as bad as you have indicated.

So far I have no real reason to believe the data is that bad. Sure there are some errors. There should be errors, right? I mean this is real data.



So what would be the best achievable data set? Would it have a perfect zero difference between the two stations for 60 years?

The fact remains that there is a ~1 degree average difference in temperature over the entire series AND the world has warmed only 1.1 degrees. I think we are going to have to agree to disagree here cause I know that there is a mathematical bias of .83 deg for the whole series and that bias varies from 1.68 deg F for the years 1984-1990 to -.9 for the years 1991 to 1997. That is not comforting that the over all bias may or may not average out--which it doesn't. I don't know what you are doing but you are not even looking at the red curve which is a 365 day running average and It isnt' even near zero for the last few years of the BP and T record.

I guess we will agree to disagree. I don't find your number convincing in light of what I see both statistically and visually.

 

[Contracelsus]

I don't know what you are doing but you are not even looking at the red curve

Sorry I was merely looking at the data without anyone's filters on it. I thought that was the thing we were doing here.

I guess we will agree to disagree. I don't find your number convincing in light of what I see both statistically and visually.

What have you seen statistically? I've provided several assessments of the statistics even down to the skewness measurement. I think the statistics speak for themselves.

But I will agree to disagree if you like.

(By the way, this is fun! I like using statistics. I just wish I'd paid more attention to it back in school.)

 

[grmorton]

Contracelsus. I don't have a clue what you are doing wrong. Thus, I decided to plot the Belle Plaine and Toledo temperature. I don't give a rats rear end for your statistics because when I plot the two temperatures on the very same graph, I see that Belle Plaine (red) is almost always hotter (above) Toledo (light green). That is a bias. The picture below is the yearly average of the raw temperature. Belle Plaine is hotter there is a bias. the bias is greater towards Belle Plain between 1984-1990 and towards Toledo between 1991 and 1997.

Whatever you are doing wrong it can't overcome this very simple graph showing that Belle Plaine is hotter--thus has a bias.

 

[grmorton]

Sorry I was merely looking at the data without anyone's filters on it. I thought that was the thing we were doing here.



What have you seen statistically? I've provided several assessments of the statistics even down to the skewness measurement. I think the statistics speak for themselves.

But I will agree to disagree if you like.

(By the way, this is fun! I like using statistics. I just wish I'd paid more attention to it back in school.)

Yeah, but you should actually look at the data rather than the statistics. A simple plot of the two temperature series shows that Belle Plaine is hotter than toledo. Thus, you have to have a bias, thus your statistics are false.

 

[Contracelsus]

Yeah, but you should actually look at the data rather than the statistics. A simple plot of the two temperature series shows that Belle Plaine is hotter than toledo. Thus, you have to have a bias, thus your statistics are false.

Unless you can show me how I miscalculated the statistics I will have to assume that "my" statistics are not in error. (Actually I've tried to use the statistics from regular statistical manuals which I've cited.)

I might be incorrect in my statistics, but you have so far not shown me any calculations. And I was rather under the impression we were hard-nosed "follow-where-the-data-leads" types here.

I thought that was what the whole debate was about: getting away from gut feelings of what we think the data says to what the data actually says.

But I will gladly try to follow if you show me where my statistical calculations were in error.

Thanks

 

[Contracelsus]

Whatever you are doing wrong it can't overcome this very simple graph showing that Belle Plaine is hotter--thus has a bias.

I downloaded the data from the same source as you and I did a day-by-day subtration. That was my data. I can see how to get monthly data but I cannot see where you get the annualized data. Could you provide me with the link?

Here's where I get the daily data: http://cdiac.ornl.gov/epubs/ndp/ushcn/usa.html

Also, note the difference between the two stations looks like it's only a degree or two on average. But I will be glad to look at the annualized data. Please provide the link.

 

[Contracelsus]

For those who don't think there is a bias problem I want to post two sets of two towns. Coldwater Kansas and Ashland Kansas.

This is another interesting case. As in the annualized temperature data graph for the two Iowa sites there is an offset clearly visible. But when I plot the difference between the daily readings I get the following histogram:

The median difference is 1 degree F. Indeed this looks like the Iowa Station pair. There is a difference between the stations, but it's only about 1 degree median. And that's for nearly 60 years worth of data taken just about every day. That's about 20,000 data points plotted here.

Again the skewness calculated is 0.168 so it's reasonably symmetrical.

I would like to stress that I don't see any reason not to assume there is a real difference between these two stations or even the stations in the Toledo-Belle Plain example, but that difference appears to be minimal and the extremes are reasonably symmetrically distributed around it.

A median difference over 60 years of 1 degree fahrenheit doesn't really seem all that bad. And considering that the data used is the TREND data (ie how the overall system is increasing or decreasing and NOT the absolute temperature), it looks to me like both stations track each other reasonably well.

(Oh, by the way, Mr. Morton, do you have an easier way to download the data and go through finding all the places where the dates don't line up? I'm doing it by hand in Excel and it's getting quite tiresome when dealing with about 20,000 lines of data. Thanks.)

 

[grmorton]

I downloaded the data from the same source as you and I did a day-by-day subtration. That was my data. I can see how to get monthly data but I cannot see where you get the annualized data. Could you provide me with the link?

Here's where I get the daily data: U.S. Historical Climatology Network - NDP-070

Also, note the difference between the two stations looks like it's only a degree or two on average. But I will be glad to look at the annualized data. Please provide the link.

In post 192 I plotted the annual average temperatures of Belle Plain and Toledo. I don't have to do anything to prove that there is a temperature bias than have people look at the two curves. Belle Plaine is hotter than Toledo 47 times out of the last 57 years. I have no idea how you are messing up but IMO only an idiologue, or someone who hasn't actually looked at the picture can stand there and say that there isn't a temperature bias towards Belle Plaine.

As to your statistics---I have no idea what you are doing, but if it doesn't match what can be seen by a 4 year old in a simple plot of the data (that Belle Plaine's temperature is hotter than Toledo's), then some gut checks should be done.

If belief in global warming causes one to claim that Belle Plaine and Toledo have identical temperatures when the chart looks like that in post 192, then I would say it is pathalogical

 

[grmorton]

This is another interesting case. As in the annualized temperature data graph for the two Iowa sites there is an offset clearly visible. But when I plot the difference between the daily readings I get the following histogram:

The median difference is 1 degree F. Indeed this looks like the Iowa Station pair. There is a difference between the stations, but it's only about 1 degree median. And that's for nearly 60 years worth of data taken just about every day. That's about 20,000 data points plotted here.

Again the skewness calculated is 0.168 so it's reasonably symmetrical.

I would like to stress that I don't see any reason not to assume there is a real difference between these two stations or even the stations in the Toledo-Belle Plain example, but that difference appears to be minimal and the extremes are reasonably symmetrically distributed around it.

A median difference over 60 years of 1 degree fahrenheit doesn't really seem all that bad. And considering that the data used is the TREND data (ie how the overall system is increasing or decreasing and NOT the absolute temperature), it looks to me like both stations track each other reasonably well.

(Oh, by the way, Mr. Morton, do you have an easier way to download the data and go through finding all the places where the dates don't line up? I'm doing it by hand in Excel and it's getting quite tiresome when dealing with about 20,000 lines of data. Thanks.)

First mistake, you didn't down load the data until now.

No, there is no easier way which is why most people don't go look at the amount of data that I have looked at. They prefer to simply believe what they are told by the media.

Contracelsus, until you can make the temperature curve for Belle Plaine go below that of Toledo half the time, your claim that there is no bias is rather hollow. I don't care what erroneous math you crank out saying that BP is equivalent to Toledo.

More data:

I will post a picture of a work I did with Missouri data. I took the annual average temperature for each of the stations in Missouri. Then I subtracted the temperature between a town and its two nearest towns in the USHCN network. I then plotted the temperature differences (sometimes the difference was negative.)

Much like that of Belle Plaine and Toledo, the amount of difference between each Missouri town and its two nearest neighbors is much larger than the proclaimed warming for the globe of 1 deg F or so. In other words, the noise in the system is greater than the signal by many fold.

These are annual temperatures so don't think in terms of daily temperature. A 4 degree difference in annual temperature over a few miles is a huge amount of difference.

 

[Contracelsus]

First mistake, you didn't down load the data until now.

Mr. Morton, I have downloaded all the data we've discussed. There is simply no other way to generate these histograms. Believe me I have put in a goodly amount of work downloading the data and then finding the places where the days don't line up and then loading it into a statistics package and running the tests.

There simply is no other way to generate these histograms.

Contracelsus, until you can make the temperature curve for Belle Plaine go below that of Toledo half the time, your claim that there is no bias is rather hollow. I don't care what erroneous math you crank out saying that BP is equivalent to Toledo.

Actually you'll note I didn't say there was absolutely no bias. My data found 1 degree difference between the station as the median difference. However, I do not find 1 degree difference to be startling.

You will note, that I was testing your assessment that there were biases at higher degree differences which I failed to find.

 

[Contracelsus]

You just illustrated the problem. If the noise is greater than the signal, it gets very hard to see the signal. the signal in this case is 1.1 deg F of global warming, but the noise is 5 degrees (using your example above). I will never be convinced that it makes sense to say that the globe has warmed by 1.1 deg +/- 5 degrees.

I think I see what the issue is here.

If I have two data sets, let's call them Data 1 and Data 2. Data 1 has a mean of 25 deg F and a standard deviation of 2 degrees F and Data 2 has a mean of 25.5 deg F and a standard deviation of 2 degrees F, but each data set is extremely large, say like 20,000 samples.

If I want to compare the mean to see if they are truly different then I run what is called a "Z-Test" (or a t-test if I feel this is just a sample and not the full population).

The t-test is

t=[(average Data 1)-(average Data 2)]/(standard deviation of the difference between the averages)

The standard deviation of the difference between the averages is

sqrt[[(N[sub]1[/sub]s[sup]2[/sup][sub]1[/sub]+N[sub]2[/sub]s[sup]2[/sup][sub]2[/sub])/(N[sub]1[/sub]+N[sub]2[/sub]-2)]*[N[sub]1[/sub]+N[sub]2[/sub]/N[sub]1[/sub]*N[sub]2[/sub]]]

What that means is that I can test for difference between means that are smaller than the standard deviation of each data set.

So in the case of my two data sets, 1 and 2 here's the result of a t-test based on random values generated for the data to specific sd and mean:


data: data1 and data2
t = -23.7, df = 39998, p-value < 2.2e-16
alternative hypothesis: true difference in means is not equal to 0

95 percent confidence interval:
-0.5116588 -0.4334933

sample estimates:
mean of x mean of y
24.99594 25.46852


Note how both samples have standard deviations of 2 degrees F but the means only differ by about 0.5deg F. The statistics were able to parse out that difference.

Here's a couple of citations to explain the t-test:
The T-Test
1.3.5.3. Two-Sample <i>t</i>-Test for Equal Means