[Cosmic_announce] Re: Change to CDAAC processing proposed

[Sean Healy]

Hi Doug,

I agree with Dave. This change could have unfortunate consequences
if bad data gets through our qc. I would ask you not to make the
change.

Sean.

Dave Offiler wrote:
> Doug,
> 
>> Hi all:  We are planning to make a small change to CDAAC processing which 
>> should be invisible to weather centers, but we would like to notify you 
>> first.
>>
>> Currently we do not create BUFR files (bfrPrf) for a certain class of bad 
>> profiles.  We are proposing to instead create these files, but mark them 
>> 'bad' in the bfrPrf file (the 'percent confidence' in the header set to 
>> zero).
>>
>> This will allow us to study this class of 'bad' profiles and perhaps 
>> understand or even fix them.
>>
>> This change should not affect any user who currently throws out 
>> bfrPrf, wetPrf or atmPrf files marked 'bad'.  Regardless of this change, 
>> weather centers should not use files marked 'bad'.
>>
>> We plan on making this change in 1-2 weeks unless this proves to be a 
>> show-stopper for anyone.
> 
> Unfortunately, our assimilation system is not currently checking the
> 'percent confidence' value[1] nor the PCD flag bits (except for
> rising/setting flag). It would be simple enough for us to make a change
> to the code, but it would not become 'live' until the next available
> operational update - which is next summer!
> 
> In short, this would be a show-stopper for us, so I would ask for the
> change not to be implemented in the bfrPrf products. Or if it is,
> not to disseminate them (if such filter is possible).
> 
> Dave
> 
> [1] Until now, no-one has been using this parameter sensibly - it's
> either always 100% or 'missing', so it's not been possible to know where
> to put a threshold. And 'bad' products have so far been very efficiently
> filtered out at source. (That's our excuse, anyway).
>