Home > Uncategorized > Climate: The classical period is 30 years

## Climate: The classical period is 30 years

2012 May 15

What is Climate?

Climate in a narrow sense is usually defined as the “average weather,” or more rigorously, as the statistical description in terms of the mean and variability of relevant quantities over a period of time ranging from months to thousands or millions of years. The classical period is 30 years, as defined by the World Meteorological Organization (WMO). These quantities are most often surface variables such as temperature, precipitation, and wind. Climate in a wider sense is the state, including a statistical description, of the climate system.

— WMO / World Climate Programme / Commission for Climatology / FAQ
http://www.wmo.int/pages/prog/wcp/ccl/faqs.html

While thinking about appropriate windows for the lowess smoothing, I recalled the commonly used definition of climate as a 30 year average. That would seem to lead to a lowess smooth of roughly f=.26 (given a time span of 117 years). Now, I don’t understand the actual windowing used by the R lowess, but if the above is about right to get what is effectively a 30 year window, then we have this …

1. 2012 May 15 at 8:28 am

Yesterday, I was fiddling around a bit to try to see how contingent the appearance of endpoints is to new data. So, I did a few what ifs, by freezing the future at April 2012’s temperature, at the average for March 2011-April 2012 and at the average for 5 years. I then applied lowess, but chopped the Lowess smooth at April 2012– to see how the current period might look after future data came in.

Comparing those three, it’s quite clear the Lowess smooth through about 1/2 the width of the window will be influenced by future data. All three Lowess smooths ever very similar to each other in the far past, but the smooths for the recent path diverged. (Admittedly, this should be obvious based on the math. The weighting function becomes assymatric near the end points. But I wanted something concrete.)

I’d have posted some images, but I also saw a puzzling thing. Using the same number of data points in the window, I also created a fourth smooth: one using only real data ending in April 2012. Not surprisingly, the recent history for this one differed from my other three. But for some reason I have not been able to grasp, the distant past for my Lowess using only real data also differed from the other three. I’m trying to figure out what it is I don’t know about LOWESS or possibly R’s LOWESS package that causes this. (Or, failing that, finding my bug!)

Do you know any reason why the shape of the LOWESS smooth during the time period 1895-1920 or so would change when additional data are added? As I wrote, I’m looking for a bug– but maybe there is some a reason this might happen. If you know it I can stop looking for the bug!

2. 2012 May 15 at 8:57 am

The issue I found is related to the choice of ‘number of iterations’. I need to read what that does– it’s evidentlly something done to deal with outliers.