Author Topic: Reporting of "Average" TEMP by WU  (Read 208 times)

0 Members and 1 Guest are viewing this topic.

Offline casa manana

  • Senior Contributor
  • ****
  • Posts: 215
    • Sedona, AZ
Reporting of "Average" TEMP by WU
« on: May 10, 2019, 01:07:01 PM »
I have noticed some apparent changes in the way WU reports "Average" temperatures.  Since I like to use AVE temps for analyzing some local weather trends, I would appreciate hearing any acknowledgement, corroboration or verification by others of these observations.

First, though, let me say that I recognize that there are various algorithms for computing average temps. 

Type 1: The most traditional (and, unfortunately, least accurate) method is to simply take the average of the daily HIGH and LOW temps.  (i.e., AVE = (HIGH + LOW)/2 )  This version is traditional because early analog recording thermometers used to just record daily HIGH's and LOW's, and so the crude formula was the best you could do with the limited data.

Type 2: A better (or at least more physical) way to compute the AVE is to take the integral of all of the day's observations, using the finest sampling available.  I prefer this method for my own use, and have been computing this for my own data for many years.  As an aside, I have found some interesting data biases that arise when using one form of AVE versus the other, but that is a topic for a different thread.

So, back to WU.  It appears to me that WU used to report AVE's using the Type 2 algorithm, or something similar, up to about 2016 or so.  Then it appears they switched to the Type 1 algorithm.  When I asked them about this back then, I got no response of course.  Well, looking at the data posted over the past month or two, it appears that the Type 2 algorithm (or something different still) is back.  It is these changes that I would like to hear some acknowledgement about.

I wouldn't be so interested in this if it were not for the fact that the two versions can lead to VERY different values (e.g., up to several degrees on some days -- amounts that swamp any real temp trends).  Anyone who is using the data to look at long term trends can't afford to mix the two types of AVE's.

I would be very interested in any direct feedback from WU, or from others who are similarly analyzing the data.  thanks all.

Dave G.

Offline gwwilk

  • Southeast Lincoln Weather
  • Forecaster
  • *****
  • Posts: 2249
    • SouthEast Lincoln, NE Weather
Re: Reporting of "Average" TEMP by WU
« Reply #1 on: May 10, 2019, 06:10:58 PM »
Why the Daily Average Temp?  Trying to combine apples and oranges but to what end?

More interesting to me is the Daily Average High and the Daily Average Low, the computation of which is very straightforward.

I just spent a few hours assembling an array of arrays of my local KLNK values for Daily Average High and Daily Average Low as well as Month to Date and Year to Date Average Precipitation and Snowfall.  The source was USClimateData.Com and required copying the monthly values into a preliminary Notepad++ text file before massaging it there into a PHP array using many search and replace actions on all 366 resulting members.  Yes, 366 because of February 29.  I then imported the array into a DreamWeaver php file and finalized the script.  The Precipitation and Snowfall Averages have as yet to be used on my dashboard.

The results can be seen in the 'Temperature' section of my dashboard.  And no, I don't have an easy way to generalize this rather tedious process.  How reliable is the data?  It looks about right to my eye, and that's all I can say.
Regards, Jerry Wilkins