All the curves that are fit to print

Now if only the mainstream media will pick up on this…. Hah!

For a layman, could you explain the difference between 1st order and 5th order curve fit in a 30 year period? Obvoiuosly, the 1st takes a longer time frame, but is it 5 times longer?

Hansen postulated global cooling first, before big Al sucked him into the warming camp, where his investments are. Now, for the long view, why is no one saying that we are actually in a very long term cooling trend, and warming is just a temporary glitch? The opposite argument sure gets a lot of traction with nothing more than some impenetrable climate models which seem to have less and less connection to reality. The refusal to admit COO rises after warming occurs, and is not the cause of warming is just one problem.
Still no more cycle 24 sunspots. Hope everyone has their warm clothes ready.

Thanks Anthony,
As I said, I’m a layman to this. I meant no offense by asking you to be “fair”.
This helps me, I appreciate it.
Happy Independance Day!

It seems to me that the fitting of a polynomial curve to this data is inappropriate. Although you can get a polynomial to fit almost any data, it has no meaning other than it seems to fit. My preference for this kind of trending is to use an exponential moving mean that weights the most recent data heavily and earlier data less. The result is a curve that tends to follow the current change in the data which is what we want in this case whether it is going up or down.
Love your site. Read it every day.

First or fifth order curve fits have nothing to do with time frames. A first order curve fit is one of the form y=mx+b. You need at least two points to make such a fit (which would be a wast of time). A fifth order curve fit is of the form y=ax^5+bx^4+cx^3+dx^2+ex+f. You need at least six data points (ie, xy) to perform that. A first order curve fit is a linear (straight) line. A fifth order fit is useful for making a smooth fit through a lot of data, but has little physical meaning (ie, try to project it beyond the last point in Anthony’s graph and it might project down to -infinity or swing around and go to +positive infinity. Now, Anthony isn’t trying to pull the wool over anyone’s eyes by doing that – its just that making a nice smooth fit using a fifth order is easy and fast in excel.
The other “curve fits” that are often discussed are actually statistical deconstructions of wave forms, and assume that temperature follows a complex combination of long and short waves that are combined. The waves are assumed to have a basis in reality in that they could be caused by sun spots or some other periodic influence, and, unlike a fifth (or even first order curve fit) can be projected into the future. THey are used often in financial and other types of forecasting. Those types of deconstructions and take more time (and a good stats package like “R” – climateaudit.org discusses R quite extensively and is a good place to go if you want to see how stats are applied. The R CRAN site is another…
You can do the same – I believe that a few posts back someone made a link to all the data for the last few decades in excel. THere are a lot of good quickie instructions on the web as to how to do the first and other curve fits using excel. I strongly encourage those who don’t know how to fit data (or even how to graph it!) to learn…
Oh, and Neilo – don’t be lazy. It took me three seconds to find the IBD article that confirms what George M stated. Hanson has apparently been trying to find the limelight since the 70’s….
http://www.investors.com/editorial/editorialcontent.asp?secid=1501&status=article&id=275267681833290
Happy Independence Day!

Bob Zorunkle,
If memory serves, first-order and fifth-order refer to the degree of polynomial to which the data has been fitted. Both first-order and fifth-order curves provided by Anthony encompass ALL the data.
A first-order curve fitting is an assumption that the data are following a straight line and that deviations from that line are just random variations. This strikes me as a fantastically stupid assumption on something that is coupled, nonlinear and chaotic, like climate. But hey, most members of the media don’t even understand the concept of slope.
A fifth-order curve is one of the form fx5 + ex4 + dx3 + cx2 + bx + a, and is much more robust for fitting temperature trends.

Bob, in this case, first order and fifth order refer to polynomials of the form:
y=a0 + a1*x**1 + … + an*x**n
which describe a best fit curve to the data. A first order polynomial is a line.

Thanks for the graphs, Anthony. It’s pretty clear that anyone looking at the first two graphs back in 2004 would think things were getting warmer.
I sure hope all the fuss will slack off, and people will start thinking about what they will do long-term to handle climate change, both colder and warmer. It’s a good start that more people are paying attention, and more detail is being developed on the potential effects of both warming and cooling. Long range contingency planning, efforts to better inform our representatives, and acquiring more knowledge on these topics would be a lot more helpful than knee-jerk political bashing.
Tnx again – Tim (www.timprosserfuturing.wordpress.com)

Why not a 200 day moving average? That would show more of the short term moves while smoothing out the peaks and valleys. The 5th order isn’t bad, at least it shows a little bit of the variation. (just eyeballing it, it looks like about 1,000 day moving average, but I’m guessing)
With a little bit of logic, I think I can demonstrate that the first order curve fit is of no analytical value. It appears to be simply a straight line fit through the available data. Here’s the problem: the slope of that line is dependant on the starting point, which was a completely arbitrary decision by the observors. I realize that that is all of the data we have for these measurements, but the underlying temperature function stretches back to time immemorial. The choice of starting point was a human decision that has no relationship to the underlying function. When an arbitrary decision by the obersvor which has no relationship to the underlying function can dramatically skew the results of the curve fit, then that curve fit tells you next to nothing about the underlying function. The fact that the 11 year fit shows a very different slope illustrates this point.
Moving averages may not tell you much about the underlying long term function, but at least they show you relative movement over a defined time period, which seems quite appropriate when dealing with a cyclical phenomena.

neilo:

On July 9, 1971, the Post published a story headlined “U.S. Scientist Sees New Ice Age Coming.” It told of a prediction by NASA and Columbia University scientist S.I. Rasool. The culprit: man’s use of fossil fuels.
The Post reported that Rasool, writing in Science, argued that in “the next 50 years” fine dust that humans discharge into the atmosphere by burning fossil fuel will screen out so much of the sun’s rays that the Earth’s average temperature could fall by six degrees.
Sustained emissions over five to 10 years, Rasool claimed, “could be sufficient to trigger an ice age.”
Aiding Rasool’s research, the Post reported, was a “computer program developed by Dr. James Hansen,” who was, according to his resume, a Columbia University research associate at the time.
So what about those greenhouse gases that man pumps into the skies? Weren’t they worried about them causing a greenhouse effect that would heat the planet, as Hansen, Al Gore and a host of others so fervently believe today?
“They found no need to worry about the carbon dioxide fuel-burning puts in the atmosphere,” the Post said in the story, which was spotted last week by Washington resident John Lockwood, who was doing research at the Library of Congress and alerted the Washington Times to his finding.

From this article.