Montag, 3. Februar 2014

Looking behind the noise: Dealing with pattern

If we recognize a pattern in a curve we can try to smoothen or eliminate it. Simply like this: If there is a ten years pattern, we just run curve with a ten years running average. Which means: A ten years running average takes the average five year left and right of every point of the original curve. Let's do it with the HADCRUT4, trying to smoothen the 11years (132 month) sunspot pattern:


It works somehow, but the pattern is still somehow zigzagging. And it produces small troughs at the top of each wave. Check at about 1880, 1910, 1940, 1970, 2000. You don't see where the top of the wave really is.

Specialists suggest a triple smoothing. Which works like this:
Divide the original smoother by 1.3371, and then again by 1.3371.

Here we go:
132 months/1.3371= 98.72 ergo 99 months
132 months/1.3371/1.3371=73.84, ergo 74 months
 And voila!


Rrrreally smoooth!

And now we can see clearly another pattern of 30 years up and down:1880 up, 1910 down, 1940 up, 1970 down, 2000 up.

What will happen, if we smoother this 60 years pattern?



Impressive thing! A wave with a low about 1885 and a high at 2003? as we know from our research in the former posts. Means this a 220 years (110 up/110 down) wave? Or is part of this man made warming? But the temperature curves are not long enough for this. For this we would need to look at proxy data. (Sediments, tree rings, etc.)

The red line was a fail. I tried triple smoothing and I cut off nearly everything. But even the green line is missing 30 years on both ends. This is the sacrifice form smoothing.

Now we go back to the original graph and do only some fine smoothing. Just lets get rid of the yearly fluctuation by triple smoothing with 12/9/7 months:



What can be seen are a lot of spikes, each about 2-4 years long. Is there something behind it? Lets take a closer look.



As you see, each sunspot cycle has three larger temperature spikes. I just try to smoother them.
132 months/3=44 months



Hmm, like expected: Having removed the three years pattern, the influence of the sunspots to the temperature is even better visible. In the 1987 to 1998 SS cycle the Mt Pinatubo Vulcan could have dampened and in the 1998-2009 SS cycle the influence of the El Nino is somehow visible.

Note: Triple smothered curves may have a certain advance or delay in the timescale.

Having done our pattern exercises, we found patterns like following: 44 months, 11 years, 60 years and possibly 220 years. We do not yet know from all of them what's behind them. But it seems that not only greenhouse gasses are related to the temperature.

Keine Kommentare:

Kommentar veröffentlichen