[wxqc] Software DP calculations

[Dave Helms]

Hi Sam,

Dew point/humidity (humidity is what is sent to CWOP in the APRS 
message)  is one of the more "noisy" parameters reported by CWOP 
members.  QCMS uses the surrounding 8 stations for the analysis 
calculations, which stations getting into the mix depends on who is 
reporting in the time interval of the sample, which will vary (as Philip 
says, this can be the cause of the "jaggies").  The analysis assumes 
most of the stations are generally functioning well.  If several 
stations have bad reports all in the same direction, then the analysis 
will be negatively impacted.  The best way to test if you are producing 
good data is to monitor a nearby ASOS station, like KAGS in your area, 
overlayed on Philips daily QC graphs. 

Here is the 2 week mean error and standard deviation plot of your 
station (*) against other CWOP stations:
http://pond1.gladstonefamily.net:8080/cgi-bin/wxqual.pl?site=AR471

For dew point, it appears you may be 1F-2F too high on the dew point 
measurements.  This is only slightly out of the target accuracy "box" 
Philip has placed on the graphs (you want to be inside the boxes).  I 
would use the 2 week mean over daily graphs to draw conclusions about 
sensor performance as the daily info can be pretty choppy.  My 
assessment of your data is that it is pretty good right now and I 
wouldn't suggest any action on your part.

I don't think the folks at Forecast Systems Lab are going to make any 
significant changes to the QCMS program as they are very busy expanding 
the station database to include additional State Road Weather networks.  
Mike can respond if I am off-base on this call.

What type of weather station do you have?  I think it is a Davis VP, but 
I can't remember right now.

Dave
CW0351

Sam Drinkard wrote:

>    I'd like to know for sure the DP calculations being done by the 
> various wx software packages is correct.  I'd think it would be, but 
> perhaps somehow something gets trashed a little causing wild 
> fluctuations in outputted dewpoints.  Given the fact that some flakey 
> data can throw off the analysis for a "cluster" of stations, does the 
> QC not have a feature or function that looks at all the peer stations 
> and throw out the "offending" one?  Is it in fact working?  Maybe I am 
> paranoid a bit about minor things, but if I'm going to send wx data, I 
> want to know what I send is of the highest accuracy that can be had 
> from a "consumer grade" wx station.  Watching the trends via Phill's 
> graphs, it appears that my DP is going from bad to worse, and daily 
> abberations are increasing.  (Dave/Sandy, can you peek at what I've 
> been producing and tell me if I need to look for some problem, or is 
> there some ragged data causing the increasing Std Dev's ?) If I 
> recall, Dave mentioned something to the effect that software chnages 
> were forthcoming in the WXQC to better handle flakey station data.  
> Any timeframe for implementation?
>