In an earlier post I showed this graph from the Berkeley Earth Surface Temperature Group (BEST).
The temperatures stop at around 2000.
By eye we can see that between 1890 and 2000 there was a temperature rise of a bit over 1 degree, say 1.1.
One of the major problems with temperature records is that they are incomplete. BEST attempt to get around this using some whizzy statistical techniques.
Michael Palmer, writing on Watts Up With That Blog has taken a different approach to the problem of incomplete temperature records. He noticed that there are 600 weather stations in US that have continuous records from 1900 to 2000 so he looked at those. The full details are in his post but here is his summary graph where he has shown rural and urban stations separately. (Note by continuous Michael means at least 90% complete, the completeness is indicated by the two dash lines at the top of the graph and the scale on the right).
And surprise surprise there is no warming over the 100 years, in fact there was a slight fall.
- -0.0073 degree C/year for rural stations
- -0.0069 degree C/year for urban stations
Was it possible to do something similar to Michael using UK weather stations?
Averaging 600 weather stations seemed more work than necessary to start with. I wondered if it would be possible to find a single weather station with a continuous record of at least 100 years?
Weather station data can be downloaded from the Met Office as a zip file containing data from over 3000 stations. As there are so many stations the files are split into a number of directories. The somewhat cryptically name directory “03” contains data from some UK weather stations and the first file in the directory (030050) is Lerwick. What’s more there is continuous data from January 1890 until May 2011. Which is nice.
The following spreadsheet shows the monthly data and calculated annual average temperature from 1890 to 2010.
It may seem a bit strange to look at 120 rows of numbers, but it never hurts to actually look at the data with your eyes before rushing off to invoke some statistical function.
Here is graph of data with trendline showing a annual rise of 0.007 degree C/year.
Unlike previous 2 graphs there is data until 2010 and it can be clearly seen that the temperature reaches a peak (8.5 in 2003) then falls so in 2010 it is lower than in 1890 (6.8 as opposed to 7.5).
Here is the same graph with the mean (7.05) shown. Temperatures clearly swing above and below the average. Yes there was a relatively high peak in 2003, but temperatures then fell back to below the average. Thus it actually makes no sense to show a trendline, as this implies there is a linear trend which there clearly isn’t.
By way of comparison if the line between the first and last post has a slope of -0.006 C/year. There is no reason for showing such a graph other than to highlight that the trendline is no more meaningful when there is oscillating data.
So can we conclude that temperatures are Lerwick rose at 0.007 degrees/year from 1890 to 2010?
Obviously not as the annual mean temperature was lower in 2010 (6.8) than in 1890 (7.5).
Should we conclude that annual mean temperature fell at -0.006 degrees/year from 1890 to 2010?
Well it did, but it is almost misleading to simply state this as to claim that temperatures have risen.
How about
temperatures oscillated around the mean of 7.05, rising to a maximum of about 1.5 higher and falling to a maximum of about 1.0 lower.
And for the future?
Who knows, but probably more of the same.
Isn’t it a bit odd how looking at raw data, without any statistical munging, can produce such a different result.
And given 601 stations show no particular rise in a century, the other weather stations that BEST used must have seen an even bigger rise than they report. Unless the rise is a result of their statistical munging.
