Guest essay by Mark Fife
There are many times, when dealing with the analysis of real world data, looking at the average alone really doesn’t provide a complete picture of what is happening. Such is my problem with describing what the data contained in the GHCN data represents. If you look at the typical NOAA, NASA, or other graphs, including mine, you get the idea temperatures are all moving up or moving down in unison, as if temperatures everywhere were all moving together under the same set of influences. Nothing could be further from the truth.
It is true, there is some degree of “movement in unison”. However, much of the movement seen falls into one of the two categories below, often involving a combination of the two types shown. As in this illustration, the key factor in determining what is happening is the height at the apex of the curve. Each of these three curves represents the same number of subjects in the underlying population. However, the spread of the distributions is different. If you are familiar with statistics, you will understand the difference is in the standard deviation. The other factor here is the upper bound for the tail remains consistent. All three curves represent populations where 99.8% are below 3. The average does change, but the upper limit never moves.
To illustrate this, let’s look at a few charts. Each of these covers 1067 stations in the GHCN dataset reporting from 1920 to 2011. I have refined each station into 10 year rolling averages, so the charts show the years 1929 to 2011. Understand, this represents the average of ten preceding years.
Here is a video of the GHCN series:
For the complete times series from 1929 to 1911, visit my YouTube channel at the following link:
The decades represented by 1929 and 1996 are two of several which match the 1929 to 1911 average quite well.
The decades represented by 1939 and 2011 represent good examples of that unilateral spread. Notice the curve apexes are lower than the average curve apex. The upper and lower bounds on these curves are not much changed from the average curve. There is very little change in the number of stations falling at 2.25° and -1.75° for the entire period.
Finally, the decade represented by 1968 is a good example of a reduced spread in the data. Again, the lower bound on the curve is still -1.75°, however the upper bound is reduced to about 1.5°.
Understanding what this means requires understanding the basic facts. At all times there are 1067 stations being represented. The “area’ under the curve is exactly the same for each curve. In every 10-year average, 99.7% of the stations fall within 2.25° and -1.75° of their 1920 – 2011 average. These basic facts never change. Likely, all we are seeing are the affects of cold and warm periods within the US due to Atlantic and Pacific oscillations. This is unavoidable because the clear majority of long term data comes from the US. There is just not sufficient long term, unbiased data from anywhere else to make any reasonable estimate.
Below is a table of countries showing the number of stations used in this study along with the average, max, and min annual change in temperatures.
|Row Labels||Stations||Average Annual Slope||Max Annual Slope||Min Annual Slope|
|Belgium and Luxemborg||1||0.014||0.014||0.014|
|Puerto Rico [United States]||1||0.012||0.012||0.012|