[wxqc] Wview pressure calculations

[Steve]

A few years ago I put together some software code that showed how to precisely reverse the SLP algorithm in the VP console to get the sensor pressure, which could then be used to calculate the altimeter value CWOP wants. I contacted a bunch of the weather software authors, pointed them to the code and encouraged them to make these changes so CWOP would get better pressure data. For those interested, the code is here: http://www.softwx.com/weather/uwxutils.html (it's my "play around with weather formulas" file, so it has some other stuff too).

I don't recall now if wview used these algorithms or if the changes in wview were made independently.

The technique I came up with was to grab the mean temperature from the BARDATA, and use that and some other current values to precisely reverse the SLP calculation. There is a simpler way (which is what WeatherLink uses) that just uses the reduction factor in the BARDATA and applies that directly to the current pressure. I chose not to use that method because the reduction factor in BARDATA has insufficient precision to precisely calculate the sensor pressure (although it will be fairly close). 

About a year ago, someone posted on a forum an odd anomaly in the VP SLP. Being at altitude, I noticed it too, and after some testing, determined that it was an artifact of the VP's method of calculating the mean temperature (the average of the current temp and the temp 12 hours in the past). The console was using the current temp in real-time, but was only using 12 hour temps from the top of the hours. The result was that 12 hours after periods where the temperature changes rapidly (like it does after the sun comes up), the SLP would show a jog at the top of every hour as the 12 hour past temp changed from one top of the hour value to the next top of the hour value. 

We pointed this out to Davis, along with a suggestion to either store more dense historical temps, or (a simpler solution), use an interpolated value for the 12 hour past temperature. Davis was good enough to make the firmware change to address this, and it did in fact remove this anomaly. 

After I did some testing, I discovered that although they improved the mean temperature part of the SLP calculation, they did not update the code that collects the data that's placed in the BARDATA packet, which still contained the mean temperature from the old formula. I pointed this out to them also, but I have not since done testing to see if they made that change in future firmware updates so that the mean temperature in the BARDATA packet matches what the SLP calculation actually used.

The net of all this is that using the reduction factor found in the BARDATA packet to calculate the sensor pressure will have (and has always had) the potential for introducing a small error because of the insufficient precision of the reduction factor. Using the precise formula to reverse the SLP may also introduce a small error in some VP firmware versions because the mean temperature in found in the BARDATA does not exactly match the mean temperature the console actually used (except at the top of each hour, when they will match).

Normally these errors are small and not very noticeable, but at some elevations and temperatures, they can become magnified sufficiently to be noticeable if you look hard.

Steve
SoftWx
  ----- Original Message ----- 
  From: Ted Lum 
  To: Discussion of weather data quality issues 
  Sent: 09/10/2008 9:28 PM
  Subject: Re: [wxqc] Wview pressure calculations


  It may make a difference how Davis calculated SLP to begin with. Basically, SLP includes the temperature correction and altitude correction. WView is trying to back it out, but if Davis messed it up in the first place that may not be possible, or at least not easy without knowing the formula they used.

  Also, WView is using the shortcut calculation. There is a much longer form that is more accurate.

  -Ted-
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://server.gladstonefamily.net/pipermail/wxqc/attachments/20080911/c0307912/attachment.html>