I am running automated trajectory generated using a modified pysplit python library and would like confirmation of my logic for meto file selection.
Trajectories will start a week or 2 in the past and run to a week into the future. They could be run daily. An example meteo files set is below with start/end dates for files downloaded today:
The start and end times (UTC) were obtained using Chk_file and the actual end of the current7days.t00z file by finding the end of valid data in the file.
Some observations:
1) current7days contains records for all 7 days but have missing data records for incomplete days (ie the 'end date' is misleading) and this may not align with the t00z (or other update time). Note that the 'actual end' shows there is a complete record for 3:00 (although I now realise this is a forecast from the prior 0:00 'analysis' record).
2) provided there is continuity across files (the start of one file must continue on from the actual end of the previous file) the model will run. Does it matter if the start/end times overlap (eg the t00z archive file overlaps with the gfsf file by 3 hours)? What if this overlap is greater and can I mix t00z with other run times (t12z) that will certainly have greater overlap? If so, how does hysplit choose which records to use if there are time conflicts?
3) The analysis files (gfsa suffix in ftp://anonymous@arlftp.arlhq.noaa.gov/forecast/YYYYMMDD) appear to cover 2 days rather than 1. I don't understand why as they are available daily.
4) My process will be to download daily the t00z versions of current7days.tnnz and hysplit.tnnz.gfsf and the weekly file when available and assume that the continuity will be there. t12z is closer to when I run the models (7am local Sydney time) and will be more current and I assume will also be continuous.
5) If I want to minimise the size of the downloads I suspect this can be achieved with a little more code logic by using the gfsa files for the dates between the weekly file and the current day but this would overlap if I downloaded for every day. Does the matter?
Could someone confirm my logic please as I need to embed it in my daily download code. Is there an alternative process that others use for this past->future trajectory calculation?
Thanks so much,
Andrew
Trajectories will start a week or 2 in the past and run to a week into the future. They could be run daily. An example meteo files set is below with start/end dates for files downloaded today:
CODE:
Source File Start End Actual endarchives Gdas1.aug24.w1 0:00 1/8/2421:00 7/8/24archives Current7days.t00z 0:00 8/8/2421:00 14/8/243:00 9/8/24forecast/240809. Hysplit.t00z.gfsf0:00 9/8/240:00 19/8/24The start and end times (UTC) were obtained using Chk_file and the actual end of the current7days.t00z file by finding the end of valid data in the file.
Some observations:
1) current7days contains records for all 7 days but have missing data records for incomplete days (ie the 'end date' is misleading) and this may not align with the t00z (or other update time). Note that the 'actual end' shows there is a complete record for 3:00 (although I now realise this is a forecast from the prior 0:00 'analysis' record).
2) provided there is continuity across files (the start of one file must continue on from the actual end of the previous file) the model will run. Does it matter if the start/end times overlap (eg the t00z archive file overlaps with the gfsf file by 3 hours)? What if this overlap is greater and can I mix t00z with other run times (t12z) that will certainly have greater overlap? If so, how does hysplit choose which records to use if there are time conflicts?
3) The analysis files (gfsa suffix in ftp://anonymous@arlftp.arlhq.noaa.gov/forecast/YYYYMMDD) appear to cover 2 days rather than 1. I don't understand why as they are available daily.
4) My process will be to download daily the t00z versions of current7days.tnnz and hysplit.tnnz.gfsf and the weekly file when available and assume that the continuity will be there. t12z is closer to when I run the models (7am local Sydney time) and will be more current and I assume will also be continuous.
5) If I want to minimise the size of the downloads I suspect this can be achieved with a little more code logic by using the gfsa files for the dates between the weekly file and the current day but this would overlap if I downloaded for every day. Does the matter?
Could someone confirm my logic please as I need to embed it in my daily download code. Is there an alternative process that others use for this past->future trajectory calculation?
Thanks so much,
Andrew
Statistics: Posted by Andrew Styan — August 10th, 2024, 3:26 am