Hi Miguel! Thank you for your answer. This is what I tested yesterday. First my time series had errors in the observation data from 3.00 to 6.00 and the forecast was to be made for 00.00. I looked into what data was used in initial.f and I saw that the data was truncated so it started at 7.00 and all the previous data was removed. When I introduced an error at 19.00-21.00 the data wasn't truncated. According to your information, the data was replaced by the forecast data. Have I understood this correctly or did I look at the wrong data?
If the surface temperature, which is not present in the forecast file, is missing in the coupling phase, is the value first removed and then interpolated? Is there a limit for the amount of data that can be missing? I saw that the model complained when there was more than 240 minutes of missing data. Before the coupling phase, when the surface temperature is used for makitp and initial, are the missing data replaced by interpolated values or is the data truncated? Thanks again! Esben --- Den mån 2008-12-08 skrev Miguel Tremblay <[EMAIL PROTECTED]>: > Från: Miguel Tremblay <[EMAIL PROTECTED]> > Ämne: Re: [Metro-developers] Error handling > Till: [EMAIL PROTECTED] > Kopia: "METRo Developer Team" <[email protected]> > Datum: måndag 8 december 2008 19.27 > Esben Almqvist wrote: > > Hi developers! > > > > I have been looking at the way errors are handled in > Metro. It seems that Metro prefers to remove data as opposed > to interpolate missing values. For ex if the data at 13.00 > and 14.00 are missing, Metro will start running from the > next proper value, in this case at 15.00. Am I interpreting > this correctly? However, it seems that the above is not > always true. Sometimes error values are interpolated. Does > it matter if the error code of Metro (9999) is used or an > out of bounds value (-999)? I suppose that if the input data > is missing 30 hours ago, it does not make a very large > difference compared to if the data is missing during the > coupling phase. Does the model distinguish between important > data and not so important data? Do you have a good > description of how errors are handled? > > Thanks! > > Esben > > > > Hi Esben, > > I am not sure that I understand what the problem is. > > Are you talking about observation or atmospheric forecast > data, or both? > > In the atmospheric forecast, the missing data are > interpolated. However, data in the atmospheric forecast > file at the beginning of period for the roadcast must be > valid. If not, the next time step of the atmospheric > forecast is used. > > For the observation, in the coupling phase, if the data is > invalid, the atmospheric forecast value is used if it covers > this period. If not, the data are in the first place removed > (this is done in metro_preprocess_qa_qc_observation.py) and > then interpolated (in > metro_preprocess_interpol_observation.py). The same is done > with forecast values (metro_preprocess_qa_qc_forecast.py and > metro_preprocess_interpol_forecast.py). > > Is this an answer to your question? > > Best regards, > > -- Miguel Tremblay > Physicien - Physicist > Centre météorologique canadien - Canadian meteorological > centre (CMC) > Environnement Canada - Environment Canada > http://www.ec.gc.ca/ > > 2121 Trans-Canada N. Suite 228 Téléphone/Phone: > 514-421-4729 > Dorval, Québec Fax: 514-421-4679 > CANADA H9P 1J3 courriel/email: > [EMAIL PROTECTED] __________________________________________________________ Går det långsamt? Skaffa dig en snabbare bredbandsuppkoppling. Sök och jämför priser hos Kelkoo. http://www.kelkoo.se/c-100015813-bredband.html?partnerId=96914325 _______________________________________________ METRo-developers mailing list [email protected] https://mail.gna.org/listinfo/metro-developers
