Much ado has been made about global warming stopping since 2001, since 1998 or not increasing in the last decade. Here’s more grist to the skeptic mill. The analysis below shows the global temperature has not increased significantly since July 1979!

Data are from the TLT Satellite measurements of the Earthâ€™s lower troposphere at RSS MSU. When you calculate the global surface temperatures from July to July 1979-2008, the earth has warmed the grand amount of 0.295 degrees C. The standard deviation of the temperature changes for each July to July is 0.2522C, putting the change over 30 years at just over a non-significant one standard deviation (actually p=0.13, significant if p<0.05) of the expected change in just one year. Stated another way, every one out of eight years, global temperature changes by a similar amount to the total increase in the last 30 years.

You have to wonder what all the fuss us about. The effect of 30 years of global warming has been no greater than the change commonly seen in a single year. The data support such low estimates of CO2 doubling as the Spencer climate sensitivity.

The frequency of July-July annual differences with a normal curve (in blue) is plotted using the R language for statistics. The change in temperature from July 1979 to July 2008 is marked as a big red dot. You can see it lies well within the spread of results.

Here is another view of July temperatures over the last 30 years, with the levels of the starting and ending temperatures marked by red dashed lines.

### Like this:

Like Loading...

*Related*

Please keep up the good work. It will take years of efforts by you folks to combat the MSM propaganda machine to con the world.

Please keep up the good work. It will take years of efforts by you folks to combat the MSM propaganda machine to con the world.

IIRC, these satellite temps are measured over 4 overlapping altitudes (wavelength dependent), then deconvoluted to give an average at a nominated altitude. More recent satellites are multi band to overcome some of these problems. However, as I understand it, the conversion to satellite signal to temperature is not made with first-principles absolute maths, but is adjusted to agree with surface thermometry. I have no idea of the difference between a first principles calculation and the surface-adjusted calculation for the lower troposphere. However, it raises the devil of the choice of the surface series to use if adjustment is indeed made this way. We are mostly aware of the frequent and sometimes debatable adjustments made to surface temp. So, I wonder how robust the conclusions can be. Also, there has been a move to fewer and higher quality surface sites (= less UHI), so the surface record might have flattened from that mechanism also.

It’s such a mess. (Not yours, I mean the official record). I spent an hour reading a blog where most people were in absolute denial that temperatures had flattened over the last several years and it’s your fault because Niche Modelling led me to it.

IIRC, these satellite temps are measured over 4 overlapping altitudes (wavelength dependent), then deconvoluted to give an average at a nominated altitude. More recent satellites are multi band to overcome some of these problems. However, as I understand it, the conversion to satellite signal to temperature is not made with first-principles absolute maths, but is adjusted to agree with surface thermometry. I have no idea of the difference between a first principles calculation and the surface-adjusted calculation for the lower troposphere. However, it raises the devil of the choice of the surface series to use if adjustment is indeed made this way. We are mostly aware of the frequent and sometimes debatable adjustments made to surface temp. So, I wonder how robust the conclusions can be. Also, there has been a move to fewer and higher quality surface sites (= less UHI), so the surface record might have flattened from that mechanism also.

It’s such a mess. (Not yours, I mean the official record). I spent an hour reading a blog where most people were in absolute denial that temperatures had flattened over the last several years and it’s your fault because Niche Modelling led me to it.

Uh. Sorry. You should read what happened over at CA about this simple observation. I am going to have to explain this a bit more. This is like the article about unreliability of attribution in linear regression – post a simple observation and people go ballistic.

OK there are lots of levels (not atmospheric levels) but conceptual/abstraction levels. Lets say we are just looking at a simple null hypothesis H0: t1979.5 t2008.5. That is, is the temperature now the same as it was in July 1979?

Notably this has a lot to do with GHG as if CO2 works as expected we would expect temperature to have increased since then. So failure to reject this hypothesis reflects badly on AGW.

Note also that the hypothesis has nothing to do with trends. There is no need to fit a regression of any kind here. There is no need to talk about whether temperature trend is flat, up or down.

It is possible that temperature trended 5% up since 1979, but the final point was exactly at the start, thus wiping out all temperature gains. Would this count for or against AGW? I would say not, as it is not consistent with the physical theory for CO2 increase to cause that type of trajectory.

We need to assume that the series is consistent and reliable. I think the TLT is the best for that. We need to get rid of seasonality, so I have used July-July changes. We need to get rid of drift, so I used a mean=0.

Without intending to detract from Lucia’s analysis, one of the assumptions of linear regression is homoscedasticity (constant variance) . It is possible (in the context of LTP) that temperature series have infinite variance, and so the assumption of homoscedasticity is problematic. Probably Lucia has thought of this.

This type of analysis of series, looking at whether a series is ‘returning’ or ‘wandering away’ are more general, in that they could be applied to a series with infinite variance.

I am not saying LTP necessarily has infinite variance, or linear regression is never appropriate in with infinite variance. The taxonomy of time series analysis is rich and complex. In a very intuitive sense, the question of whether temperatures have ‘returned’ to where they were in 1979 when they were supposed to go up makes a lot of sense. Perhaps the returning or wandering distinction is more important with LTP than talking about trends, given they are ‘naturally trendy’ anyway.

David Stockwells last blog post..Linear Regression R2

Uh. Sorry. You should read what happened over at CA about this simple observation. I am going to have to explain this a bit more. This is like the article about unreliability of attribution in linear regression – post a simple observation and people go ballistic.

OK there are lots of levels (not atmospheric levels) but conceptual/abstraction levels. Lets say we are just looking at a simple null hypothesis H0: t1979.5 t2008.5. That is, is the temperature now the same as it was in July 1979?

Notably this has a lot to do with GHG as if CO2 works as expected we would expect temperature to have increased since then. So failure to reject this hypothesis reflects badly on AGW.

Note also that the hypothesis has nothing to do with trends. There is no need to fit a regression of any kind here. There is no need to talk about whether temperature trend is flat, up or down.

It is possible that temperature trended 5% up since 1979, but the final point was exactly at the start, thus wiping out all temperature gains. Would this count for or against AGW? I would say not, as it is not consistent with the physical theory for CO2 increase to cause that type of trajectory.

We need to assume that the series is consistent and reliable. I think the TLT is the best for that. We need to get rid of seasonality, so I have used July-July changes. We need to get rid of drift, so I used a mean=0.

Without intending to detract from Lucia’s analysis, one of the assumptions of linear regression is homoscedasticity (constant variance) . It is possible (in the context of LTP) that temperature series have infinite variance, and so the assumption of homoscedasticity is problematic. Probably Lucia has thought of this.

This type of analysis of series, looking at whether a series is ‘returning’ or ‘wandering away’ are more general, in that they could be applied to a series with infinite variance.

I am not saying LTP necessarily has infinite variance, or linear regression is never appropriate in with infinite variance. The taxonomy of time series analysis is rich and complex. In a very intuitive sense, the question of whether temperatures have ‘returned’ to where they were in 1979 when they were supposed to go up makes a lot of sense. Perhaps the returning or wandering distinction is more important with LTP than talking about trends, given they are ‘naturally trendy’ anyway.

David Stockwells last blog post..Linear Regression R2

David,

I’m having trouble understanding your point. If I understand you properly:

– From selected t_1 = 7/1979 to t_30 = 7/2008, the temperature increase is 0.295-C.

– The standard deviation for July measurements is 0.2522-C.

– So the temperature increase exceeds the standard deviation by a factor of 1.17. That suggests that this increase is significant.

– So then what you seem to be exercised about is that this amount is not very big. But that is a result of the small rate of change – a rate that is nonetheless about 5 times faster than “typical” temperature change rates (such as the exiting from ice ages).

– Perhaps I’m missing your point, because I’m not specially good with statistics. But suppose that I define a time series S(t_n), based on:

S(t) = alpha*t + beta*Rand(t),

where Rand(t) is a random number in the range (0,1); and alpha = beta/30.

By your argument, wouldn’t you come to the conclusion that alpha = 0, no matter what it actually was?

David,

I’m having trouble understanding your point. If I understand you properly:

– From selected t_1 = 7/1979 to t_30 = 7/2008, the temperature increase is 0.295-C.

– The standard deviation for July measurements is 0.2522-C.

– So the temperature increase exceeds the standard deviation by a factor of 1.17. That suggests that this increase is significant.

– So then what you seem to be exercised about is that this amount is not very big. But that is a result of the small rate of change – a rate that is nonetheless about 5 times faster than “typical” temperature change rates (such as the exiting from ice ages).

– Perhaps I’m missing your point, because I’m not specially good with statistics. But suppose that I define a time series S(t_n), based on:

S(t) = alpha*t + beta*Rand(t),

where Rand(t) is a random number in the range (0,1); and alpha = beta/30.

By your argument, wouldn’t you come to the conclusion that alpha = 0, no matter what it actually was?

Hi Neal. It takes almost 2 SD’s for a value to be 95% significant (1.96 in fact). On the slope, I am not coming to any conclusion about alpha when I compare the end points. It is quite possible that the alpha is non-zero, and the endpoints to be the same. To be physically realistic, consider a sawtooth generator. On one tooth, the linear model will be sloped, but the end points are the same. Yet in one cycle there has been no change.

Hi Neal. It takes almost 2 SD’s for a value to be 95% significant (1.96 in fact). On the slope, I am not coming to any conclusion about alpha when I compare the end points. It is quite possible that the alpha is non-zero, and the endpoints to be the same. To be physically realistic, consider a sawtooth generator. On one tooth, the linear model will be sloped, but the end points are the same. Yet in one cycle there has been no change.

David,

– But even a 1-SD change has 68% significance, yes? Better than 50%.

– Of course everyone involved would like any signals to be clearer. However, the value of alpha is not something that can be selected by the climate researchers: Whatever value it has is determined by the results of the measurements. If alpha/beta ~ 1/30, that’s the breaks.

David,

– But even a 1-SD change has 68% significance, yes? Better than 50%.

– Of course everyone involved would like any signals to be clearer. However, the value of alpha is not something that can be selected by the climate researchers: Whatever value it has is determined by the results of the measurements. If alpha/beta ~ 1/30, that’s the breaks.

Neal, When people say ‘significant’ they mean 95% or 99% level. 68% is only one chance in 3 of being wrong, but 95% is one chance in 20 of being wrong. So 68% is not regarded as significant.

The analysis makes no statement about alpha. There is no linear model. A linear model, trend whatever, is just a set of assumptions. No intrinsic meaning.

Neal, When people say ‘significant’ they mean 95% or 99% level. 68% is only one chance in 3 of being wrong, but 95% is one chance in 20 of being wrong. So 68% is not regarded as significant.

The analysis makes no statement about alpha. There is no linear model. A linear model, trend whatever, is just a set of assumptions. No intrinsic meaning.

David,

I guess I was misled by your statement, “You have to wonder what all the fuss us about. The effect of 30 years of global warming has been no greater than the change commonly seen in a single year.”

The concern for global warming is not for the last 30 years but for the next 200. Under a business-as-usual scenario, continued CO2 emissions for the next 200 years is expected to cause very serious and noticeable changes in temperature; unless we decide to do something about it, there is no reason to believe that CO2 emissions will cease on their own. If we assume linear growth in temperature, you would then get something like an 8-SD change over that 200-year span.

That’s the reason for the fuss.

David,

I guess I was misled by your statement, “You have to wonder what all the fuss us about. The effect of 30 years of global warming has been no greater than the change commonly seen in a single year.”

The concern for global warming is not for the last 30 years but for the next 200. Under a business-as-usual scenario, continued CO2 emissions for the next 200 years is expected to cause very serious and noticeable changes in temperature; unless we decide to do something about it, there is no reason to believe that CO2 emissions will cease on their own. If we assume linear growth in temperature, you would then get something like an 8-SD change over that 200-year span.

That’s the reason for the fuss.

Come off it, Neal. People are not going to permit an 8SD change in temp over 200 years. You know that.

Come off it, Neal. People are not going to permit an 8SD change in temp over 200 years. You know that.

#9, Geoff Sherrington:

Accepting David’s estimate of the linear trend over 30 years, this is what would happen over 200 years.

“People are not going to

permitan 8SD change in temp over 200 years.”So what are they going to

doabout it? According to the point of view you seem to espouse, nobody should attempt to cut back on CO2 emissions. So do you think people will just tell the atmosphere to “come off it”, like a new King Canute?#9, Geoff Sherrington:

Accepting David’s estimate of the linear trend over 30 years, this is what would happen over 200 years.

“People are not going to

permitan 8SD change in temp over 200 years.”So what are they going to

doabout it? According to the point of view you seem to espouse, nobody should attempt to cut back on CO2 emissions. So do you think people will just tell the atmosphere to “come off it”, like a new King Canute?Neal. What I will not do is close my mind to all but CO2. I have NEVER espoused a point of view that we should not cut back on CO2.

The 30 years of David’s article arose from the availability of data. It means nothing to me in climate theory.

Maybe the whole problem that bothers you can be solved, possibly completely, by a timely and careful adoption of large scale nuclear power generation for the next century or two. We are already half a century down that road with minimal fuss. I have already been part of a management team that has substituted nuclear fuel for CO2 emission exceeding a half billion tonnes, starting work on the matter in 1970 or so. What have you done?

A prevalent failure of doomsayer predictors like Malthus, Club of Rome, Erlichs etc is their underestimation of human ingenuity. You’d be breaking new ground by modelling that for 200 years ahead, but be warned, it also has intangible variables that can swing it, just as GCMs do. Familiar ground though, because like temperature models, it can be constrained to increase with time.

Reminder: The theme of this blog is “The Power of Numeracy”. Please cease making social comments that cause a social reaction to set the record straight.

Neal. What I will not do is close my mind to all but CO2. I have NEVER espoused a point of view that we should not cut back on CO2.

The 30 years of David’s article arose from the availability of data. It means nothing to me in climate theory.

Maybe the whole problem that bothers you can be solved, possibly completely, by a timely and careful adoption of large scale nuclear power generation for the next century or two. We are already half a century down that road with minimal fuss. I have already been part of a management team that has substituted nuclear fuel for CO2 emission exceeding a half billion tonnes, starting work on the matter in 1970 or so. What have you done?

A prevalent failure of doomsayer predictors like Malthus, Club of Rome, Erlichs etc is their underestimation of human ingenuity. You’d be breaking new ground by modelling that for 200 years ahead, but be warned, it also has intangible variables that can swing it, just as GCMs do. Familiar ground though, because like temperature models, it can be constrained to increase with time.

Reminder: The theme of this blog is “The Power of Numeracy”. Please cease making social comments that cause a social reaction to set the record straight.

#11, Geoff Sherrington:

– My question to David was in response to his question, “What’s the fuss about?”, and intended to obtain clarification on his point of view. My comment was thus exactly as “social” in nature as his. So why didn’t you jump down David’s throat?

– Maybe you should focus on maintaining a civil attitude in response, and not appoint yourself as a kind of blog policeman. I think that is clearly David’s job – and I do not detect that he took any umbrage at my responses to his postings.

#11, Geoff Sherrington:

– My question to David was in response to his question, “What’s the fuss about?”, and intended to obtain clarification on his point of view. My comment was thus exactly as “social” in nature as his. So why didn’t you jump down David’s throat?

– Maybe you should focus on maintaining a civil attitude in response, and not appoint yourself as a kind of blog policeman. I think that is clearly David’s job – and I do not detect that he took any umbrage at my responses to his postings.

A random process with 0.2522C SD has what probability to deviate only 0.295 degrees C in 30 trials? A random, not very far walk ? What leash is the temperature on ?

A random process with 0.2522C SD has what probability to deviate only 0.295 degrees C in 30 trials? A random, not very far walk ? What leash is the temperature on ?

Neal. The way to interpret a non-significant figure is more in the realms of there being nothing to explain, no evidence of an effect to attribute CO2 to.

Not so much a quantitive effect.

Please, I rather you went to a forum for general AGW discussion and stick to evidence-based discussions here. Regards.

Neal. The way to interpret a non-significant figure is more in the realms of there being nothing to explain, no evidence of an effect to attribute CO2 to.

Not so much a quantitive effect.

Please, I rather you went to a forum for general AGW discussion and stick to evidence-based discussions here. Regards.

David,

I’m simply trying to understand your original point. You say that temperature has not changed significantly in 30 years. My point is that this can only be construed as surprising if you expect a rate of increase that exceeds a particular threshold. I am trying to understand how you are interpreting the evidence.

As I see it, “general AGW discussion” began at #9.

David,

I’m simply trying to understand your original point. You say that temperature has not changed significantly in 30 years. My point is that this can only be construed as surprising if you expect a rate of increase that exceeds a particular threshold. I am trying to understand how you are interpreting the evidence.

As I see it, “general AGW discussion” began at #9.

Random error walking at sqrt(30)*SD. So you would suspect that temperature is other process probably confined ?

Random error walking at sqrt(30)*SD. So you would suspect that temperature is other process probably confined ?

Neal, There are a few provisos here. First the lack of power of a test could also give lack of significance. For the purposes of discussion lets assume the test is powerful and the change is not significant. Then you say one: “expect a rate of increase that exceeds a particular threshold”. This is the proposition AGW is it not? That it exceeds some problematic threshold. If not, sensitivity is too low for there to be a problem. Lack of increase in temps beyond noise with linear increase in CO2 is then strong evidence against AGW.

David Stockwells last blog post..Rybski Model Proof

Neal, There are a few provisos here. First the lack of power of a test could also give lack of significance. For the purposes of discussion lets assume the test is powerful and the change is not significant. Then you say one: “expect a rate of increase that exceeds a particular threshold”. This is the proposition AGW is it not? That it exceeds some problematic threshold. If not, sensitivity is too low for there to be a problem. Lack of increase in temps beyond noise with linear increase in CO2 is then strong evidence against AGW.

David Stockwells last blog post..Rybski Model Proof

David Stockwell,

But what gives you the certainty that the test is sufficiently powerful?

David Stockwell,

But what gives you the certainty that the test is sufficiently powerful?

Thats another topic that I will have to deal with later. The finding of non-significant is still valid though.

Thats another topic that I will have to deal with later. The finding of non-significant is still valid though.

David – According to the RSS web page

http://www.remss.com/msu/msu_data_description.html

the decadal trend of global TLT is 0.170 K/decade. Is that trend “significant”? What is the July-only trend? Is it different? Is the standard deviation different? Would not the estimation of “significance” be made more robust by the inclusion of all months since 1979? How would the significance be affected one considered only data from, say. the Julys of leap years since 1979?

David â€“ According to the RSS web page

http://www.remss.com/msu/msu_data_description.html

the decadal trend of global TLT is 0.170 K/decade. Is that trend â€œsignificantâ€? What is the July-only trend? Is it different? Is the standard deviation different? Would not the estimation of â€œsignificanceâ€ be made more robust by the inclusion of all months since 1979? How would the significance be affected one considered only data from, say. the Julys of leap years since 1979?

David:

– How does a test against the hypothesis of a 0.0170-K/year look, as Pat Cassen suggests?

Pat Cassen:

– There

isa reason to focus on comparisons for a given month: I recall that the expectation for GW is that the winters will warm up more than the summers; the season being defined relative to the northern hemisphere, which has more land-mass coverage.David:

– How does a test against the hypothesis of a 0.0170-K/year look, as Pat Cassen suggests?

Pat Cassen:

– There

isa reason to focus on comparisons for a given month: I recall that the expectation for GW is that the winters will warm up more than the summers; the season being defined relative to the northern hemisphere, which has more land-mass coverage.Hi Pat. Judging by the interest in this I will have to go into it much more. These are all good questions that I can’t answer right now as I am going away.

In general, this analysis is ‘different’ from a trend analysis and could well give a different result.

‘Robustness’ is another issue. Inclusion of intervening months might not necessarily make a result more robust. Trend analysis might not be robust if the trend is deceptive due to LTP. Here the intervening data are essentially duplicates, with few real points. Ultimately a more robust result would be related to the physics of the process, such as mean reversion, conservation or dissipative and the like.

It’s like saying, wouldn’t a parametric test be more powerful than a non-parametric one? Maybe, maybe not. Depends on the character of the data.

davidss last blog post..Rybski Model Proof

Hi Pat. Judging by the interest in this I will have to go into it much more. These are all good questions that I can’t answer right now as I am going away.

In general, this analysis is ‘different’ from a trend analysis and could well give a different result.

‘Robustness’ is another issue. Inclusion of intervening months might not necessarily make a result more robust. Trend analysis might not be robust if the trend is deceptive due to LTP. Here the intervening data are essentially duplicates, with few real points. Ultimately a more robust result would be related to the physics of the process, such as mean reversion, conservation or dissipative and the like.

It’s like saying, wouldn’t a parametric test be more powerful than a non-parametric one? Maybe, maybe not. Depends on the character of the data.

davidss last blog post..Rybski Model Proof

David — I guess the point of my questions in #20 is that I thought it rather imprudent to state:

and

on the basis of such a small fraction of the available data (July averages). Particularly in view of the fact that RSS gives a decadal trend which implies a temperature increase substantially larger than what you find.

You say that your analysis has nothing to do with trends, but simply the issue of “…is the temperature now the same as it was in July 1979?” The answer is apparently, yes, the average July temperature is higher by slightly more than one SD of the data you chose to look at. (You would have obtained a slightly different answer last year.) Do you really want to make a judgment on how such a conclusion supports or negates AGW? If you do, some expectation must be built into it, as Neal pointed out.

Although autocorrelation and LTP may introduce complications (for which analytic techniques exist), surely it is reasonable to expect signal/noise to improve with the inclusion of more data. Why not use it?

David — I guess the point of my questions in #20 is that I thought it rather imprudent to state:

and

on the basis of such a small fraction of the available data (July averages). Particularly in view of the fact that RSS gives a decadal trend which implies a temperature increase substantially larger than what you find.

You say that your analysis has nothing to do with trends, but simply the issue of â€œ…is the temperature now the same as it was in July 1979?â€ The answer is apparently, yes, the average July temperature is higher by slightly more than one SD of the data you chose to look at. (You would have obtained a slightly different answer last year.) Do you really want to make a judgment on how such a conclusion supports or negates AGW? If you do, some expectation must be built into it, as Neal pointed out.

Although autocorrelation and LTP may introduce complications (for which analytic techniques exist), surely it is reasonable to expect signal/noise to improve with the inclusion of more data. Why not use it?

“how such a conclusion supports or negates AGW? If you do, some expectation must be built into it, as Neal pointed out.”

Less than random walk, low pass straightjacket, not expected but suggested.

“how such a conclusion supports or negates AGW? If you do, some expectation must be built into it, as Neal pointed out.”

Less than random walk, low pass straightjacket, not expected but suggested.

So what is being said here is that if you only take a small sample of available data (30 years) rather than a larger available sample then you will loose significance on a statistical test?

OMG!

Are we allowed to ask why 30 years was chosen? Why not 29 , 20 or only 10 years? The p value would have been much less significant if you had only used the last 10 years of data!

Skeptics SAs last blog post..Guglielmucci and a reality based world…

So what is being said here is that if you only take a small sample of available data (30 years) rather than a larger available sample then you will loose significance on a statistical test?

OMG!

Are we allowed to ask why 30 years was chosen? Why not 29 , 20 or only 10 years? The p value would have been much less significant if you had only used the last 10 years of data!

Skeptics SAs last blog post..Guglielmucci and a reality based worldâ€¦

Hello Pat,

Your issue seems not be with the literal accuracy of the statement “The analysis below shows the global temperature has not increased significantly since July 1979!” but with its interpretation (or potential for misinterpretation). I admit that though accurate its a bit promotional, but I think I get a bit of latitude on a blog.

“Why not use it?” could also be asked of Rybski who based his paper on the stats of differences of temperatures. Over at CA DrK indicates that 30 years of data would have no more than 2 effective data points under LTP. Aren’t you concerned that you might be overfitting by using regression trends?

Hello Pat,

Your issue seems not be with the literal accuracy of the statement “The analysis below shows the global temperature has not increased significantly since July 1979!” but with its interpretation (or potential for misinterpretation). I admit that though accurate its a bit promotional, but I think I get a bit of latitude on a blog.

“Why not use it?” could also be asked of Rybski who based his paper on the stats of differences of temperatures. Over at CA DrK indicates that 30 years of data would have no more than 2 effective data points under LTP. Aren’t you concerned that you might be overfitting by using regression trends?

admin says

“I admit that though accurate its a bit promotional, but I think I get a bit of latitude on a blog.”

Well, ok, I won’t begrudge this latitude, but only because you run such a civil blog (smile)!

What I really had in mind did not concern the time span, but the particular selection of data over those years. Motivated by the trend given by RSS, why not do this analysis for

allthe months, not just July? This might address the point that Neal brought up, of a possible difference in summer v. winter warming.And I reiterate the point that the term “significantly” depends on an implicit expectation.

admin says

“I admit that though accurate its a bit promotional, but I think I get a bit of latitude on a blog.”

Well, ok, I won’t begrudge this latitude, but only because you run such a civil blog (smile)!

What I really had in mind did not concern the time span, but the particular selection of data over those years. Motivated by the trend given by RSS, why not do this analysis for

allthe months, not just July? This might address the point that Neal brought up, of a possible difference in summer v. winter warming.And I reiterate the point that the term “significantly” depends on an implicit expectation.

Hi Pat, Yes well that is how it was done by Rybski, concerned with the changes in climatological averages, and I will do it like that from now on.

As to significance, I am not sure where you are going with that. Sure there is an expectation and it can be different. Here the expectation is that the difference is zero — undetectable, vs some detectable change. The change between 1998 El Nino and today is detectable for example, as shown in a subsequent post.

Hi Pat, Yes well that is how it was done by Rybski, concerned with the changes in climatological averages, and I will do it like that from now on.

As to significance, I am not sure where you are going with that. Sure there is an expectation and it can be different. Here the expectation is that the difference is zero — undetectable, vs some detectable change. The change between 1998 El Nino and today is detectable for example, as shown in a subsequent post.

The “implicit expectation” has to be explicitively determined. What factors, contribute what amount. Zero cannot be found.

So for David to prove, he has to find negative ?

The “implicit expectation” has to be explicitively determined. What factors, contribute what amount. Zero cannot be found.

So, for David to prove, he has to find negative ?

The “implicit expectation” has to be explicitively determined. What factors, contribute what amount. Zero cannot be found.

So for David to prove, he has to find negative ?

So, for David to prove, he has to find negative ?

A general principle is that you propose what you expect, and see if the statistics does not match it.

In this case, you could expect a zero change, or you could expect a small change (based on the calculated impact of AGW). Both are not-incompatible with the data selected.

But only the second expectation is in-line with our understanding of atmospheric physics.

A general principle is that you propose what you expect, and see if the statistics does not match it.

In this case, you could expect a zero change, or you could expect a small change (based on the calculated impact of AGW). Both are not-incompatible with the data selected.

But only the second expectation is in-line with our understanding of atmospheric physics.

Now that “imprudent” has been identified. Kick the temperature ball some more.

Some major weather drivers are T^4 effect related. Should we not be, also, averaging T^4 to get a better climate feel ? Then, leave ((sum T^4)/n)^0.25, just the bone, to the biologists ?

Now that “imprudent” has been identified. Kick the temperature ball some more.

Some major weather drivers are T^4 effect related. Should we not be, also, averaging T^4 to get a better climate feel ? Then, leave ((sum T^4)/n)^0.25, just the bone, to the biologists ?

great job!

great job!