Page 1 of 1

Dealing with Null Data freom Eastweather

Posted: Sun 13 Apr 2014 5:13 am
by Willythekorn
I just started using Cumulus yesterday. I think it's great so far, but I do have a situation that I'd like a comment on. I have a WS 2080a from Ambient Weather, and am using Cumulus to upload the data to Weather Underground (station ID KCAALTAD21).

My station is having a problem dropping the signal between the sensor unit ad the receiver, sometimes for up to two hours. In Easyweather, the dropped signals are recorded as nulls (---) in all relevant data field. When Cumulus picks the records up, it notes the error, but it stores in the database a copy of the last good signal it did receive. This leads to some very interesting graphs on Weather Underground. Clearly the best solution would be to fix the communication problem, but drops do happen from time to time even under the best circumstances.

In a perfect world, a null record is a null record, not a copy of a previous non-null record. Is there a reason Cumulus takes this solution to dealing with data drops? Is there something in the settings I can set to keep that from happening?

Re: Dealing with Null Data freom Eastweather

Posted: Sun 13 Apr 2014 6:39 am
by steve
Cumulus doesn't have a concept of 'null' readings. If there is a break in communication, it re-uses the last good value that it had, on the assumption that the break is temporary and brief, and if there is a more permanent or lengthy problem, this would be fixed at source. This also means that the data files are usable in other software for analysis without having to filter out the 'null' values. This is how I decided to approach the situation when I first wrote Cumulus for my own use, and it isn't something that is likely change in the near future.

It may be possible that in a future version there might be an option to stop uploading to WU (or just send pressure) when communication is lost with the outdoor sensors (WU has no concept of null values either). I'm not sure how WU would handle this in its graphs, but I believe that it might leave a gap in the plots, when the interval between two updates of a particular item was greater than some threshold.