[wxqc] Local elevation effects on QC graphs

[Philip Gladstone]

The temperature calculations are all adjusted for elevation -- the 
comparison performed by MADIS is the Potential Temperature (which is a 
notional temperature at roughly sea level).

For graphical display purposes, I convert the potential temperature back 
to surface temperatures to draw the graphs. This is why drawing graphs 
with multiple stations on them do not work very well (when there are 
significant altitude differences). This is a misfeature that I intend to 
resolve (sometime soon).

I am looking at ways to use the local topography to relax the limits on 
errors (which control the red X). However, it is rather complex.

If I look at the C3833 figures, then over 56 days (8 weeks), your 
average error is essentially 2 degrees. This could be a calibration 
error at your station, or could be caused by multiple calibration errors 
at neighboring stations. Can you spot check your temperature with a 
thermometer? I'd do it on a windy evening so that there is good mixing 
and no solar radiation to be concerned about.

Philip

Hank Sniadoch wrote:
> Mark .... I have a situation like yours .... I'm at 1400 ft elevation 
> and all other spotters are at lower elevations.  My temps are lower than 
> those reported but I'm afraid to change my readings since I think 
> they're real ....
> Mark W wrote:
> 
>> Hi all,
>>     My temperature has been off by quite a bit as compared to the QC 
>> graphs, and I was wondering if elevation is taken into account for 
>> temperature deviations. Most of the surrounding stations are 700-800ft 
>> below me and as expected I have lower temps more often then not, but 
>> in the evening when cool air pools in the lower areas, I have higher 
>> temperatures.
>>  
>> In particular yesterday really showed this since we also had a local 
>> rain shower at the end of the day.
>>  
>> http://pond1.gladstonefamily.net:8080/cgi-bin/wxqchart.pl?site=C3833
>>  
>> The temperature on my hill didn't get near as cool as the elevations 
>> below me. I was wondering if in cases like this is the weather info 
>> rejected?
>>  
>> There is a reject day on my sensor array and that is due to the fact 
>> we had sun while persistent clouds were in other places, on top of the 
>> fact we started warmer right in the morning:
>>  
>> http://weather.gladstonefamily.net/site/C3833
>>  
>> I cannot imagine the daunting task of determining if the data is good 
>> or not. Soo many specific cases can cause a dataset on a particular 
>> day to fail.Why does it only analyze a day? What about a running 
>> average error rate over a period of days where the oldest time has the 
>> least impact, and the latest time has the most. Something akin to the 
>> old formula
>>  
>> Y = Y - Y / numPoints + newSample
>>  
>> And Y will eventually result in newSample times numPoints as data is 
>> passed through it. Divide by numPoints and you have a infinite 
>> averaging filter weighting the lastest samples the most. Changing the 
>> numPoints effects the amount of averaging.
>>  
>> I used a method like this to solve air-pressure changes to detect 
>> trends where the newSample was actually the difference between two 
>> adjacent points. Then I didn't have to store all of the points and sum 
>> them all to find the slope over a period of time. Then it was easy to 
>> take the output on a real-time system with very little computational 
>> overhead.
>>  
>> The reason why I bring this up is while day to day events can be 
>> highly variable from location to location, averaging over a week would 
>> make more sense I would think.
>>  
>> Just a thought.
>>  
>> p.s. I don't like those little red crosses ;-)
>>  
>> -Mark
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> wxqc mailing list
>> wxqc at lists.gladstonefamily.net
>> http://pond1.gladstonefamily.net:8080/mailman/listinfo/wxqc
>>  
>>
> 
> _______________________________________________
> wxqc mailing list
> wxqc at lists.gladstonefamily.net
> http://pond1.gladstonefamily.net:8080/mailman/listinfo/wxqc
> 
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3322 bytes
Desc: S/MIME Cryptographic Signature
Url : http://pond1.gladstonefamily.net:8080/pipermail/wxqc/attachments/20050628/959b4b60/smime.bin