2016/03/09

Data Sources

I'm going to apologize for this one in advance, because it's going to be boring. But I wanted to write just a little bit more about the data that I'm using, and now is as good a time as any. So with that warning out of the way...

Environment Canada has weather data for Edmonton going back to 1880. It is broken up over 3 weather stations:

The first station was located near what would eventually become the CN Rail yards, and more recently the new downtown arena. I don't know what would have been there in 1880 though, since rail didn't reach Edmonton for another 20 years. The second and third stations were located near each other at the City Centre Airport.

There is some overlap in the times that the stations were recording: 1937-1943 for the first and the second station, and 1997-2005 for the second and the third. Since duplicate data is available for those years, I felt that I really should compare the results to see how much they vary.


For the 1937-1943 period the results of the two stations are within +/- 1°C of each other 62% of the time, and within +/- 5°C 99% of the time. And there are a handful that are way off, at +/- 15°C. For the more modern 1997-2005 period there is much less variation, with 87% within +/- 0.5°C, and 99% within +/- 1°C, with the largest gap being 4°C.

I broke the numbers into "Summer" (using the term loosely) and Winter just to see if the variation might depend on the season, and for the 1997-2005 period it really does. For those stations most of the discrepancies occur during the winter, while for the 1937-1943 period it's much more balanced. The 1937-1943 data is also nicely centered around zero, whereas looking at 1997-2005 the third weather station frequently measures about 0.5°C colder than the second station during the winter.

So with all of that said, there's really not much that I can do with this. The data is the data. The second and third stations are in pretty close agreement, even if the third station is a little colder. The original station was a little more variable. I'm not going to use that to try to calibrate the older data, but I will keep it in mind as margin of error. For what I'm doing I'm okay with a few degrees variation.

My approach is to make clean breaks when a new station is available:
  • 1880-1937: First Station #1863
  • 1937-1996: Second Station #1867
  • 1997-today: Third Station #27214
There are several other weather stations near Edmonton - most notably at the International Airport. I don't have much interest in the data from the International, because I don't live anywhere near there. I will definitely take a look at it at some point to see just how much colder it really is, but on a day-to-day basis I won't be using it for anything.

Sorry for the housekeeping, but I wanted to get this all written down somewhere. Hopefully next time will be more informative.

1 comment:

  1. It's not a boring post. It's very interesting because I've been wondering where the data source is to get the weather data. Thank you. Your site has quite a few nice charts BTW.

    ReplyDelete