Global warming not for Australia this june
By Joliphant Posted in Life Issues — Comments (84) / Email this page » / Leave a comment »
Funny if global warming were happening you'd expect things to be getting well warmer. Seems thats not the way its going in the land down under. Their June has just topped ? the all time record low for the month. Bureau of Meteorology
Its a funny thing about global warming you would expect the record lows to be an endangered species but somehow they aren't. You would also expect the record highs to be setting records. They aren't charts of record temperatures set by year As a matter of fact that record years for the US were the thirties.
Hrmmm, a lot of CO2 pumped into the atmosphere since then. you'd really expect more high temps.
Now the fun part of this is how the warmers explain this phenomenon. They say winter time lows are getting higher while summertime highs are getting lower. The other argument is that the CO2 forcing is occurring at night.
So let me get this straight. Our summers will be cooler, winters warmer and evenings more pleasant. I am not prepared to say that mankind is the cause of global warming but if we are, aren't we so much more lucky than smart. We did by accident what could never be accomplished by purpose.
If you want to play with the data its available here in Excel.
Crossposted at HTTP://theminorityreportblog.com
Then again my state is having both a drought and floods at the same time. Thank you army corps of engineers.
It could easily have both. Thats why this is fun though. We have an entire continent shivering, hoping for warming.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
You should know, Jolly, that Australia is in the Southern Hemisphere, where it's *Winter* now, duh! How can there be record high temperatures in the Winter?
But what this uncanny lack of record high temperatures shows that the Earth is already too warm, so we must be having global warming even worse than we ever thought. Also, this all happened after GWB took office and failed to ratify Kyoto. As far as we know, GWB is a man. Therefore, global warming is man-made.
Hmm, there seems to be some subtle error in my reasoning, but I can't find it....
--
Gone 2500 years, still not PC.
You asked, "How can there be record high temperatures in the Winter?"
There are record high temperatures for any given day or month or season or year or since records have been kept. A record high in the winter will be lower than one in the summer, but it is still a record high.
There is a minimum temperature for anything in the universe. 0 degress Kelvin, or about -272 degrees celsius IIRC. Nothing can get colder than that. However, there is not a mamimum temperature in the same sense.
A light-weight model can be used to describe that situation.
Take a bathtub with a foot of water. Make some waves. If the wave peaks don't reached down to the bottom (the floor), then they pretty much look like your standard sine wave. Equal peaks and valleys. Adding energy makes both the peak and valley of the wave more extreme.
If you keep making waves in a way that is harmonic with the size of the pool - where the waves bounce off the ends and come back - the same idea happens on a sphere, but they just go around and around instead of bouncing - then eventually your waves might overlap in a way where the bottom of the waves scrapes the floor of the tub. It has reached (approached) the theoretical minimum temperature at that location. When that happens, the top of the wave becomes very erratic in its behavior, just as a wave approaching a beach. It whitecaps.
Temperature extremes really are a reasonable part of a global warming model. Whether we are having much of an effect on weather is an entirely separate issue. Personally I think we may be having some effect, but it is likely dwarfed by natural processes - particularly Star scale processes.
Australia is very much affected by El Nino, and there has been a very strong one over the last year - driest year on record etc. Typically these end with a sudden reversal (La Nina), and that has just arrived (after being promised for months). A lot of rain, and rather cold. The temperature figures for June here mainly reflect the North, where the wet season was unusually late.
In Melbourne, where I am, the winter pattern has been perhaps similar to January in the US. Unusually warm, we wondered if winter would ever come, and then late in the month a cold gale and rain. You can see the effects of the drought and the La Nina reversal in this graph of the ups and downs of our water storage.
The usual objections - it isn't the coldest June ever, just since 1950, and from an AGW perspective, you should note the fuller Met Bureau story:
June 2007 was the coldest June recorded in Australia since at least 1950, according to the Bureau of Meteorology’s National Climate Centre. It comes close on the heels of the warmest May in many parts of Australia and the warmest autumn on record in eastern Australia.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
The claim for AGW is that it is a greenhouse phenomenon. It is supported by the increase in CO2 and other gas levels at the same time as an overall temperature increase, coupled with the physical attribute of those gases as having higher transparency to light at higher frequencies.
One thing that is never mentioned, however, is that fluids do not disperse perfectly. We should expect more warming in areas generating more greenhouse gases.
I wonder if anyone has in fact looked into that.
--
Gone 2500 years, still not PC.
And I crunched through it to look for the trends and came up with the following:
The count of max temperatures increases over time:
And the count of min temperatures increases over time:
Both red lines are first-order regression fits to the count data over the span of years included in the source.
So, insofar as it's possible to infer anything from squeezing the numbers with this much liberty, I think it suggests that temperature variation is increasing. There's a trend towards more extreme temperatures in both directions. Note that neither graph means that the earth is getting hotter or cooler; these are plots of abnormalities that may or may not reflect the mean temperatures during the same periods and places. There were some data for Hawaii mixed in the lows, too, so it's not all ConUS.
This reflects one of my concerns about GW, that not only may the earth be heating overall, but we may see more severe temperature swings and volatility in weather patterns and climates. In statistical terms, I'm worried not just about the slope of the regression, but also that the standard deviation is increasing over time.
You might want to do that. Just a suggestion.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
first temperatures in the database were recorded only on displacement and with a correction to the prior count. So if selma had an all time high of 89 degrees in june 1930, and had a new all time high of 90 in 1990 the count for 1930 would be dropped and the count for 1990 incremented.
This was mentioned on the hall of record links up at the top.
Second if you take the data set which is all over the place, and use a polynomial trend you get a different conclusion.
A polynomial fit gives us a peak for variability in the middle of the century and we are in a decline.
Using a linear trend on cyclical data will work for half a cycle (more or less). Despite the large amounts of noise this data set looks more cyclical than linear.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
That is some pretty sketchy data analysis. It makes me chuckle that you can be so adamant about tearing down climate scientists when you make comments like this one (and believe me it isn't your first time, Joliphant. If I were you, I'd consider going back to school and taking some advanced statistics and time series courses). Wow just wow, I'm somewhat glad that I'm done lurking at this site. As a scientist, I don't think I can take the mathematical incompetence anymore.
Looking over your posts. You have offered one post that was factual data and when asked to defend the methodology you cut and pasted from the article.
which can be found here
Seeing as you are so tickled by this why not share the joke. Perhaps you can explain why using a polynomial power series fit can't describe the data ? Somehow I think you can't. Somehow I think I can write a proof that shows I can use just such a series of polynomials and factors to come as close as I care to any curve.
BTW this time at least paraphrase before you cut and paste.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
This really depends on what you mean by curve.
If you mean the graph of an arbitrary function, you are incorrect. The characteristic function of the rationals is an extreme example, that is the function thats 1 on the rationals and 0 elsewhere. This function is continuous nowhere and so does not have a taylor series. Most functions are continuous nowhere.
If you mean continuous functions well you're still wrong. The absolute value function is not differentiable at 0 and so has no taylor series at 0. And most continuous functions are differentiable nowhere.
If you want to restict to smooth functions (function with derivatives of all orders everywhere), it still doesn't work. Wikipedia has the classic example of a function that is infinitely differentiable but has no taylor series at 0 since the every derivative is 0 at 0.
The class of functions you're looking for is the analytic functions that is the set of functions that have a taylor series.
But in the case at hand, we don't actually have a function but a finite series of data points. And that CAN be represented EXACTLY by a polynomial fit. To fit n points you need a polynomial whose degree is at most n-1. And that I can prove.
but in the present case, such a polynomial, while existent, would be pretty worthless for identifying a trend, yes?
An exact polynomial fit would be pretty worthless. This is one reason I am leary of most curve fits. Its very easy to get very good fits by simply raising the maximum degree of the variable or by allowing multiplications between variables. Of course the results bear no resemblance to reality.
I don't consider a sierpinski gasket a curve.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
It's a peeve of mine. Physicists and engineers, I guess, tend to think that all functions are analytic. This is simply not the case. But I guess in the realms in which they work, the assumption of analyticity will at least give good approximations.
Another peeve is the notion that dx is an infinitely small quantity. There is no such thing in modern calculus.
We are way off topic but this is very enjoyable. I haven't used these branches of mathematics in over 30 years. Brings back good memories.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
If they'd not done that, then the data would all be stacked towards the first few decades and the beginning of record keeping, when it would be much easier to set a record. So...what's your point?
As for the polynomial fit...heck, you can fit a 20th order polynomial for that matter and make it wiggle where you please. That doesn't mean that doing so is reasonable or passes the "smell test". Personally, I'd say there's just enough observability in the data to support a linear trend and not much more, and even that is pushing it. And if you haven't seen equivalent data before and after this sample, you don't know it's cyclic, let alone what its period is; it's almost certainly a superposition of a great many cycles of various magnitudes, many of them dwarfed by the noise of the data.
You need new glasses. What were the correlation numbers from your linear regressions ?
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
The data presented is in the correct form for analysis. Each data point represents that in that year, the highest temperature for the period 1864-2006 was recorded (for some particular state/month). The "methodology" you describe is just how this data was recorded over time. It's final form does not depend on this process, and would be replicated by not recording any extremes until 2006, and then picking out all the extremes at the end (so no "drops" of old data).
A polynomial fit cannot recreate cyclical data, as polynomials are not cyclical. Further, it is impossible to say data is cyclical on a period twice as big as the data range, as you appear to be claiming. There may be a cyclical trend at shorter (11 year?) time scales, but you also won't be modeling that with a polynomial.
Finally, you appear to be basically trying to obfuscate the actual question. If the question is, "Are mean global temperatures rising?" and we have measurements of mean global temperatures, that is the data to use. You have chosen instead to use USA state-by-state extreme value records to try to infer a mean global temperature. When opening a door, just use the handle, don't throw a chair at it and then claim that the other folks are the illogical ones.
-jb
Are you saying that the number of record highs in a year vs the number of all time record highs in a year isn't an entirely different animal ? Or just not interesting ?
A polynomial fit cannot recreate cyclical data, as polynomials are not cyclical.
Yes of course but whats your point. The polynomial fit captures the inflection thats clearly in the data. The linear trend line gives an impression thats not warranted. I have no idea where the first fundamental of the data is or how many cycles are superimposed. Polynomials are better. If the data were smoother Fourier analysis would be better yet. As it stands gibbs phenomena alonewould make it useless for this set.
Finally, you appear to be basically trying to obfuscate the actual question. if the question is, "Are mean global temperatures rising?"
It really helps if you read what was written. I know it must have taken some effort to misconstrue what I said so the question is was it deliberate or just plain mental rigidity
What I said was that the number of record highs had been relatively constant and the number of record lows had been relatively constant. I also said the record lows had been occurring during the summers and the record highs have been occurring during the winters.
The above implies that while the earth may be warming it does not accord with the catastrophe scenario but rather with an improving climate. This does not accord with the crop failure/desertification/ famine scenarios advanced by the IPCC.
The question is not what is the weather going to be, but can we or should we be trying to do something about it
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
"Are you saying that the number of record highs in a year vs the number of all time record highs in a year isn't an entirely different animal ? Or just not interesting ?"
As nargin pointed out, record highs in a year are not interesting, as then you have a very strong bias towards earlier years having more record highs, since they have fewer previous years to compete with. In this case 1884 would have 600 record highs simply because it is the first year in the data set.
"I have no idea where the first fundamental of the data is or how many cycles are superimposed. Polynomials are better."
This does seem to effectively capture your approach to data analysis. You don't know and simply rationalize whatever seems to back up what you want to see. It's fairly clear that there is no strong trend of any kind in this data, which is not surprising in the least. It is well known that the continental US has a very weak global warming signal, so it would be surprising to see a signal in state extreme temperature data.
"The above implies that while the earth may be warming it does not accord with the catastrophe scenario but rather with an improving climate. This does not accord with the crop failure/desertification/famine scenarios advanced by the IPCC."
The "catastrophic" impact predictions are generally not happening in the US. There is a decent probability that the Midwest will get drier and hotter, which will impact farming, but as a country we are well positioned for climate change. If the US is an island, we do in fact have less to be concerned about than other parts of the world, but that is just not the case. Our economy is very connected with much of the rest of the world. Frankly, I'm not satisfied with the "don't worry about it, it'll all be better in the end" attitude. At the very least that is not a very conservative position (in the dictionary, rather than political, meaning of the word).
"The question is not what is the weather going to be, but can we or should we be trying to do something about it."
More accurately, the question is what the weather will be, what impacts will the change in weather have and how much of the negative impacts can we mitigate without the cost being greater than the cure.
-jb
This does seem to effectively capture your approach to data analysis. You don't know and simply rationalize whatever seems to back up what you want to see. It's fairly clear that there is no strong trend of any kind in this data, which is not surprising in the least. It is well known that the continental US has a very weak global warming signal, so it would be surprising to see a signal in state extreme temperature data.
This seems to be a matter of projection on your part. Down below I asked you a simple question. What does it mean when the R^2 value for a linear fit is .005. Instead of answering you launch the above. Well that value is what you get when you try to do a linear fit to the data set above. What it means is you have no fit and shouldn't be using a linear fit.
Nargin for some reason didn't bother to publish the fit numbers for his regressions. Can't fathom why. I also fail to see how he got the weather was getting more extreme when the reverse is happening. (damage due to weather is up but thats due to population trends)
You on the other hand have been going nuts and whats more have chosen to launch an attack to distract from the fact that you have been defending rotten methodology
I make no bones about not having all the answers. nor do I shy away from the implications of the data.
As to the nature of the data points. After the initial spike you would expect both sides to track the overall temperature trends.
The "catastrophic" impact predictions are generally not happening in the US. There is a decent probability that the Midwest will get drier and hotter, which will impact farming, but as a country we are well positioned for climate change.
Funny how the catastrophic impacts never seem to happen anywhere where they can be adequately measured or where the governments take precautions against disaster.
I do agree the midwest will get drier. Nothing to do with GW though. Its aquifer of fossil water is running out.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
If you are trying to interpolate values in between data points or extrapolate values (ie predict future values), a low r-sqrared would indeed indicate that the model you had chosen was probably not a very good choice. However, this does not imply that the least-squares linear fit does not provide any information about the data.
Especially when the line has very little slope and an R^2 near zero.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
There is a substantial difference between saying "temperatures are generally rising over time" and "the temperature will be XX degrees on July 6, 2011". Low r-squared implies that the model you had chosen was not good for making the latter claim. However, this does not mean it is completely useless for making the former claim.
Furthermore, there is no reason, a priori, for choosing a 3rd order polynomial to fit to. Why not 2nd or 4th? It is a pretty arbitrary choice unless you know something about how the data is distributed.
It does. There is a combination of small slope to the line and very low value R^2 it means that what you are reading as a trend is no better than noise. And this Value is really low.
Now I am repeating myself.
As to why I chose what I did. I believe it was the default setting on my spreadsheet. The point of the chart was to show nargins chart wasn't saying much of anything.
If I remembered I had a numerical library for my fortran compiler earlier, I just would have used that first and found the value and refuted on that.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
By "r squared" do you mean the sum of the squares of the residuals, data minus fit (I speak a somewhat different statistical language)? If you've a genuine interest I can provide them, unless you've just done the equivalent of dig your old baseball mitt out of the hallway closet and arrive late to the game, eager to toss terminology around. The standard deviation of the residuals is large, sure, but I think it's reasonable to say that there's a net upwards trend in both data sets.
And good gravy, man, Fortran? IMSL library maybe? That brings back some bad memories, so I can't help needling just a bit...I've coded both linear and nonlinear parameter estimation from scratch in a half dozen languages and heartily recommend Matlab as a contemporary alternative. Two functions, "polyfit" and "polyval", meet most of one's linear (non-iterative) needs for casual data fitting; for example, my code snippet for fitting max count data:
[us_state,maxmon,maxtemp,maxyear,maxcity,...
mintemp,minyear,mincity] = ...
textread( 'c:\us_extreme_temp.csv', ...
'%s%s%s%s%s%s%s%s', 'delimiter',',' );
maxcount = hist( maxyear, max(maxyear) - min(maxyear) + 1 );
x = min(maxyear):max(maxyear);
y = maxcount;
[p,s] = polyfit( x, y, 1 );
ytmp = polyval( p, x );
rsquared = sqrt( sum( (y-ytmp).^2 ) )
Is that a first, posting source code on Redstate? My apologies if I'm trying anyone's patience here, but it's great stuff, no linking and compiling, just quick vectorized, interpreted, interactive code...
And for some reason, you keep mentioning slope of a fit being "small". This really means nothing; recall our discussion of linear scaling on one of your previous threads. Slope is a ratio, and I can give a plot any slope I want simply by scaling one of the axes independent of the other. Any statistical measurement has to be interpreted in the context of the source data set and its units.
Slope is a ratio, and I can give a plot any slope I want simply by scaling one of the axes independent of the other. Any statistical measurement has to be interpreted in the context of the source data set and its units.
I see no reason to go after such easy targets.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
On this thread alone you have:
Misunderstood record high temperature data, repeatedly.
Posted an illegitimate polynomial fit with no goodness of fit value while critiquing others for not including goodness of fit.
Misunderstood what cyclical means, or at least how to treat cyclical data and than mischaracterized your previous words (What? Did I say cyclical...?).
Compared the likelihood of a cold month in Australia in a warming climate to the entire NL East not hitting a home run for a month.
Led with a graph with record low er... record high... well it depends on if you read the title or the axis.
You aren't exactly throwing punches from a position of intellectual strength at this point. I mean, thats just one thread!
Ah heck, just cause I get such a kick out of it, an old favorite, Joliphant style papers at their best.
-jb
Is your point to distract away from the fact that Nargin just said he could make the slope of his linear approximations whatever he wanted ?
I'd like to reply to your criticisms but there only one. The graph that I didn't create is mislabeled. If thats the worst error I have in my lifetime I will live with it.
The comments can be read and read for what they are. So can yours Benj.
I must say though nice bit of coming to the aid of the guy on your team.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
Still...?
Ok, perhaps I wasn't clear...slope is rise over run, correct? So if I were to scale my data along just one axis, say by changing the units of temperature, that would change the amount of "rise" per same unit of "run". Change temperature to Celsius or microdegrees or whatever and you'll get a different slope for a fit regression line.
If the axes have units, the slope has units of y axis units over x axis units. So change the units of one quantity (scale) and you get a different slope in different units. That's why slope is only important in terms of the units in which it's expressed.
Standard deviation is also dimensional depending on data type, and this is why in more complex regression problems with disparate data types, data are often scaled by the inverse of the standard deviation, not only to weight values by precision, but also to nondimensionalize and enhance numerical conditioning of matrices.
Clear now?
You can measure in kelvin, Rankine, farenheit or celsius, the count of new highs on any given day is going to be the same.
THE Y AXIS IS DIMENSIONLESS, ITS THE COUNT OF NEW HIGHS A PURE NUMBER
Capiche ?
Now you can change the time units all you like its not going affect the Delta on the y axis. Its still going be overwhelmed in the noise.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
Sure...and I could make the units on the left be tens of counts or millicounts or whatever to place the slope where I please. The slope itself still isn't as important as its sign, the statistical properties of the data, and the context in which it's taken.
The Magnitude of the change would still be noise.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
for the concept that the volatility may be increasing? If we are witnessing an effect from the build up of gas in the atmosphere, and the gas just keeps building then we need to think very hard for a mechanism to keep cold getting colder as well as hot getting hotter.
Now I am aware that the pro-AGW arguement does allow for different effects, for example the atmosphere gas may create more clouds that can alternately reflect more heat back into space or trap more heat at the surface of the eartb. So all you need to do is switch the reflectivity from time to time to create higher highs and lower lows, but this seems to be a convenience more than an actual mechanism.
Anoter arguement is that the presense of warming puts more overall energy into the atmosphere and this extra energy can drive more wild weather, but again a benign hurricane season like last year's is hard to explain by this mechanism.
I believe it was Michael Crichton who suggested that the long length of time necessary to see the effects of AGW would lead the leaders of the movement to postulate that weather extremes are the result of GW and then point at extrema to justify the various regulations that they wish to put into practice.
I'm not a climatologist and frankly don't know enough about the subject to posit a specific mechanism. But it's a sensible expectation that a system that's had more heat added to it will generally become more energetic, including kinematically.
I think we have to get away in the climate debate from drawing trends through a single datum and declaring too small a sample to be evidence pro or con. Detecting change in a dynamic system subject to wide fluctuations requires a long view.
...to clarify matters, your original plot is the low temperatures, despite the typo in the y axis label that says "high temperatures"; the content matches my second plot of min temperature counts.
When discussing global warming it would be best to talk about the average warming trend of the planet Earth, and not a partial selection of countries or continents, don't you think?
It's like saying that a long-term increase in homeruns in the major leagues is undermined by the fact that the Chicago Cubs' homerun numbers did not increase in one given year.
Its more like saying that long term increase in home runs is undermined by NL East not hitting one in a given year.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
1) The equivalent of the NL East not recording any homeruns would have to be an extraordinary, unimaginable cold temperature for Australia; say, an average of 1000 degrees below zero.
While the baseball number in the analogy is admittedly extreme, it is nevertheless physically possible, which cannot be said about the temperature you chose: you just can't crank the ol' thermostat down much past -459.4.
soli Deo gloria
If someone asked me the question "Are home run numbers increasing?" I also would answer the question using the NL East home run total from one month. The full season full league statistic is much less accurate.
-jb
Because I would comment on the fact that the baseball season had been extended (more games per season to hit homers). I would also point out that individual batting performance had actually narrowed in its variability and the number of extremes to either side of the curve had remained relatively constant.
Predictable JB. You do realize you repeat yourself so I must ask is it thought or Memorex ?
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
You managed to seize upon one piece of data presented without refuting the rest. If I could dig up data for europe (probably would want to confine it to western europe) It would very likely echo the US data presented. The problem with using other parts of the world during the time period is that for much of it, the data is spotty. Europe by example, for about 15 years out of the twentieth century they had much on their mind other than taking a normalized temperature reading. Africa and Asia are even worse not having developed the reporting capability till later in the century.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
Europe didn't take a break from collecting measurements during the world wars. On the contrary, weather forecasting is essential for planning and executing naval and air operations, and the various powers went to great lengths to maintain weather stations in even remote and inhospitable areas towards this purpose. For example, wiki's got a nice graph of Norwegian temperature collections on Svalbard back to 1912, showing a distinct warming trend over the period, particularly in recent decades.
A lot of weather data collection in the first half of the century was performed at sea by specially equipped ships. I sailed recently on a windjammer whose original purpose was meteorological measurements in the north Atlantic.
The American Institute of Physics has an interesting article on climate change with a great deal of history included, in particular the following:
It may have been the press reports of warming that stimulated an English engineer, Guy Stewart Callendar, to take up climate study as an amateur enthusiast. He undertook a thorough and systematic effort to look for historical changes in the average global temperature. One 19th-century German had made an attempt at this in seeking a connection with sunspot cycles. Otherwise, if anyone else had thought about it, they had probably been discouraged by the scattered and irregular character of the weather records, plus the common assumption that the average climate scarcely changed over the span of a century. But meteorologists around the world had meticulously compiled weather records, and Callendar drew upon that massive international effort. After countless hours of sorting out data and penciling sums, he announced that the temperature had definitely risen between 1890 and 1935, all around the world, by close to half a degree Celsius (0.5°C, equal to 0.9°F).
There was plenty of data being taken as a result of the surge in interest in the late 19th century in geophysics and study of the earth's climate, but the effort wasn't necessarily uniform or well disseminated. Fortunately, there's still a wealth of data available, and today modern computers let us perform large-scale numerical analysis on large databases rather than "penciling sums".
From ships changed for the US during the war and after for the rest of the world. Second yes it is for war planning but consistency of methodology is important for comparison. In the case of world war 1 location and operators of monitoring stations became disrupted. The same for world war II.
Do you really want the temperature records for the Somme or the Ardenne ?
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
So study the effects of the different measurement techniques and determine the feasibility and methodology of applying correction factors. It doesn't mean the data is worthless, just that it has caveats and a context for its analysis.
Seriously Joli, other than to further tarnish your reputation as a skewer of science, why are you keeping with this argument?
Here's an idea - we're talking about global warming, right? Why not average the temperatures around the globe, not just areas that you like, and post those?
Oldest DNA ever recovered shows warmer planet: report
Scientists who probed two kilometers (1.2 miles) through a Greenland glacier to recover the oldest plant DNA on record said Thursday the planet was far warmer hundreds of thousands of years ago than is generally believed.
DNA of trees, plants and insects including butterflies and spiders from beneath the southern Greenland glacier was estimated to date to 450,000 to 900,000 years ago, according to the remnants retrieved from this long-vanished boreal forest.
That contrasts sharply with the prevailing view that a lush forest of this kind could only have existed in Greenland as recently as 2.4 million years ago, according to a summary of the study, which is published Thursday in the journal Science.
The samples suggest the temperature probably reached 10 degrees C (50 degrees Fahrenheit) in the summer and -17 C (1 F) in the winter.
They also indicated that during the last period between ice ages, 116,000-130,000 years ago, when temperatures were on average 5 C (9 F) higher than now, the glaciers on Greenland did not completely melt away.
"These findings allow us to make a more accurate environmental reconstruction of the time period from which these samples were taken," said Martin Sharp, a glaciologist at the University of Alberta, Canada, and a co-author of the paper.
"What we've learned is that this part of the world was significantly warmer than most people thought."
(emphasis mine)
___________________________________
The CIA has better politicians than it has spies - Fred Thompson
If there was ever a time when conservatives had a chance to take a "fish in a barrel" shot at environmental nonsense, it's this week. Given a serious issue like global climate change, you'd think that environmentalists would be proposing legislative solutions like carbon caps, a federal gas tax, increased subsidies for clean energy research and development, etc. But what is the self-annointed leader of the environmental movement currently doing? He's organizing a multi-continent concert series to "raise awareness" concerning the issue. I think that everyone knows what global warming is; we don't need Garth Brooks singing in the National Mall to tell us. Gore wants to be a rockstar, though, so the concerts must go on.
Just when I feel like hanging my head in shame at the antics of my team (the Libs), I read one of Joliphants posts, with his self-serving data selection (Australia? it's called GLOBAL warming), his odd way of manipulating that data (polynomial interpolation? for something as imprecise and fickle as annual temperature change?!), and his apparent obsession with proving that there is no way that global warming is happening (or that humans causing it, depending on which "new evidence" he's championing today).
It brings me solace to know that, while my team tends to go overboard and respond in silly, silly ways to serious problems, at least they aren't obsessed with proving that nearly every climatologist in the field is either a liar or too stupid to understand flaws in their system that any layman with a college level course in statistics could point out.
With environmentalists I usually know I have hit a nerve when they try and get nasty. Oh btw, it helps your case if the subject line is a proper sentence.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
By the way:
1) I'm not an "environmentalist." I think global warming is happening, and we're causing it. I'd like to see a group of climatologists, economists, and industry leaders sit down with our political leaders and suggest solutions that will work without throwing our economy into a tailspin, but I hardly think that makes me a tree-hugger.
2) You didn't "hit a nerve." On the long list of things I care about, global warming ranks pretty low. I find posts like yours amusing, not irritating. I was only offering a little advice: next time, take the easy shot and make fun of Gore's little concert.
3) What was "nasty" about my comment? In no way did I attack you personally. I criticized your assertions, your methodology, and your obsession with proving that global warming isn't a problem (for reasons that I honestly fail to grasp). I hardly think that constitutes being "nasty," though.
At any rate, if I was unintentionally nasty and I hurt your feelings, I apologize. No harm intended.
But to keep this friendly let me ask you a question.
If global warming is so low on your list of priorities why are you for carbon caps, carbon trading and an increased gas tax. These things will hurt you in ways that will be real and immediate. I'd hate to see what you would be willing to do if it were higher on your list of priorities.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
I said that if one were of the Gore-persuasion, lobbying for those types of things would be a more productive way to spend one's time than holding a concert.
Not being of such a persuasion, I'd like to see climatologists, economists, etc. perform a cost/benefit analysis of the solutions presented thus far and possibly propose other, possibly more workable, solutions.
The whole situation, like most, is best approached in a level-headed manner which recognizes both the problem and the unintended consequences that various solutions might have.
But I am not real smart on this subject, so let me ask two questions:
#1
"...they aren't obsessed with proving that nearly every climatologist in the field is either a liar or too stupid to understand flaws in their system that any layman with a college level course in statistics could point out."
So if there are flaws in the systems of climatologists that any layman with a college level course in stats could point out, why should I trust what they (and you) say?
Just asking. (And by the way, if you want to put a little cash on whether New York will be underwater in twenty years, I'm your man.)
#2
If this is such a "serious problem", why aren't your boys in congress in special session RIGHT NOW to pass a few laws to take care of it? Or is it that they also feel it is a bunch of crap?
Then I've got to get back to work.
1) I wasn't saying that there are flaws in their systems. I was saying that Joliphant's post assumes that there are such flaws that he can correct merely by using a polynomial graph to "better" analyze the data. Don't you think that people who spend their lives studying the climate would have caught such an elementary error?
2) Our "boys in congress" have managed to ignore lots of fairly obvious problems with much simpler solutions for a while now. Just look at the federal deficit.
By the way, who in the world has claimed that NYC will be underwater in 20 years? I'm not saying that no one has, but I certainly haven't heard about it.
1. The author of the work used a moving average. Nargin used a linear regression to fit the data and then claimed increasing variability. My guess is his goodness of fit/R/correlation was nearly zero certainly below .5 Either way a linear fit is almost never appropriate for this type of data. The polynomial fit I used provides a better fit. I would have used some other type of nonlinear regression but I am limited to using excel on this pc.
2. The deficit is at a negligible percentage of the economy and if current trends continue will be gone in less than a year.
3. Al gore did in 1997.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
1) I'm no statistician, and you can correct me if I'm wrong, but isn't a polynomial equation typically used either a) when you have a set of points through which your graph must run (which obviously doesn't apply here, since your graph hits very few of the points) or b) when you already know the trend line to be non-linear? Nargin's graph merely showed an upward trend. I didn't see him claim that the trend line was in fact linear. When you have points as erratic as annual temperature fluctuations, though, the best one can do is point out in which direction the trend is going. Why would one go to a polynomial graph unless one already had good reason to suspect a specific type of trend? You seem to be begging the question.
2) An argument about deficits would be a threadjack, but I'll merely point out that I was talking about both on-budget and off-budget expenditures (e.g. Social Security), with the latter set to sky-rocket in the coming years. In addition, forecasts of shrinking deficits assumes no repeal of the AMT. Not to mention that if we really are fighting a global war on terror, then we might want to start putting military projects, like Iraq, into our calculations rather than funding them through supplemental appropriations. The only reason the deficit is set to "shrink" is because the numbers are being gamed (just like they were for Clinton to get his "balanced budget" in the '90's, by the way). Every person in Congress realizes this (or at least they should), but no one wants to address the problem of federal deficits.
3) I really tried to find Gore's quote about New York under water in 20 years. I couldn't find it, but I'll take your word for it. If he did say such a thing, he was obviously wrong and (not surprisingly) overstating his case.
Both red lines are first-order regression fits to the count data over the span of years included in the source.
Thats from Nargin. I used polynomial fit because of two reasons. 1 I had it (I didn't have to write a regression program that reads data from excel to get it.) 2. We know climate cycles are non linear.
2. Yes but believe it or not the deficit is not such a bad thing as long as its kept at a manageable size.
3. NY post october 1997
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
but I assumed (perhaps he can chime in to confirm or deny this) that he was using a linear regression because a) he was only seeking a general trend line and b) we don't know exactly what the rate of change is. Linear regression is used to draw general trends of complicated systems all the time. A function doesn't have to be linear to use linear regression to identify the general direction of a system.
A polynomial, on the other hand, makes more assumptions about the data. It assumes either that the data fit to the equation precisely or that we already know the shape of the trend line.
In other words, Nargin's linear regression (not his standard deviation argument, which I'm not sure we have enough data to warrant) is a looser fit, but it makes fewer assumptions about the data. Your polynomial equation (which is but one of many that could be applied to the data to achieve a variety of tighter or looser "fits" to the data) makes more assumptions about the data and its trends than are warranted.
I hope that I'm explaining my thoughts clearly. As I said, though, I'm not a statistician, and I may be missing something quite obvious.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
hey benjneb you are representating yourself as an expert on this
On a linear fit what does an r^2 value of .005 indicate ?
Its amazing what you can do with a csv file and fortran.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
Low r-squared suggests that temperature as a function of time is probably not linear. This does not mean that the function is not generally increasing.
If I start the trend line at 1930 it shows a negative trend.
What the R^2 number is telling you, is you can't use the trendline.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
It means that the slope provides very little additional information over the null hypothesis of no slope. Effectively it indicates there is no appreciable linear trend in the data. Which is pretty clear just from looking at it.
It certainly does not mean that you should add higher power polynomials to your regression. As nilram mentions above you can get perfect fit by using an n-1 factor polynomial fit, but that is an artifact of the finiteness of the data set and the additional terms are not providing any real insight or predictive power. Every added poly term will necessarily increase goodness of fit, at the cost of complexifying your underlying model, which is a bad thing.
-jb
Its good to know. Then again I wasn't the guy who started trying to use the data that way.
I'd agree about the increasing the number of polynomial terms, thats why I used an order 3 fit. Not an order 120. It pretty much caught the visible trends (wasn't going to mention this but since you have allowed that the eyeball can serve as a judge of data why not). We had lows at the end of the last century peaked in the 30s dropped off through middle part and have started up again.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
The r^2 on your 3rd order poly fit is 0.0354
Which means, your polynomial terms provide very little information over the null hypothesis of no trend...
Additionally, it is not legitimate to do a poly fit on this data, poly fits depend on where the origin is, and you cannot define an origin on this data. Linear fits do not depend on the location of the origin, which is why they can still be used.
Just letting you know.
-jb
You would mention the only reason I used the polynomial was to impeach the linear fit If you recall I said from the beginning there was no substantial trend. Thats up there in the very top.
Let me know if you find me using that fit to make a prediction.
______________________________
"Those who expect to reap the blessings of freedom must, like men, undergo the fatigue of supporting it."
-Thomas Paine: The American Crisis, No. 4, 1777
Yes, thank you, I only wanted to illustrate the general trend of both data sets. Then I suggested that the implication of the upward trend might be increasing volatility of temperature extremes; I readily admit that this is a combination of hypothesis and impression, not a strong assertion.
To what degree you can "fit" a set of data and draw conclusions from that fit depends on the observability of the data. Granted there are numerical tests in specific circumstances, but over time in engineering and science you develop an intuition for how much information can be practically extracted from a dataset, the "smell test" I mentioned. I gauge the smell test for these data to be a first order fit.
He would have forced his victims to read arguments by engineers/mathmeticians/statisticians regarding temperature variables, charts, graphs and global warming.
I meant what I said and I said what I meant. An elephant's faithful 100 percent.
A goverment that is big enough to give you all you want is big enough to take it all away.
Barry Goldwater
I know Australia has had droughts before , but this is crazy.