initializing forecast from future data

Questions and other topics related to UEMS 18.
Post Reply
pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

initializing forecast from future data

Post by pattim » Sat Feb 08, 2020 7:07 pm

Can anyone tell me now to initialize a forecast from future data? Such data is out there on the servers, but the --date option does not act upon future dates - it simply reverts to the current date. (This may be a bug or intentional.) I've been searching through the userguide ems_prep chapter, but no joy. Thank You!! PattiM :D

Weyand
Posts: 1
Joined: Fri Feb 21, 2020 12:18 pm

Re: initializing forecast from future data

Post by Weyand » Sat Feb 22, 2020 4:03 pm

Are you sure that's even possible?

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: initializing forecast from future data

Post by pattim » Wed Mar 04, 2020 10:23 pm

Hi Weyland - thanks for the reply! When I look on the ems2 server for datasets, I see they're broken down by day of the starting integration. I believe integrations are restarted (from some prior finished integration) every (some number) of time steps (for data assimilation, etc.).

The http://ems2.comet.ucar.edu/data/grib/na ... grib.t18z/ for instance has integrations going out to 84 hours. These are used by uEMS only for nudging and BC's (with a single integration used for IC's). The thing is, if I want to know, say, how orography affects longer-term forecasts, I'd like to use, say, use IC's several days hence. The data's there, but in the wrong subdirectory for the uEMS Perl coding; apparently because for a given day it looks only in that day's subdir on the server.

What I want to do is an experiment on the effects of the strong orography here - I need to integrate Sunday's predicted storm (say from late Wednesday) from 72 hours out, doing a 24-hour integration on my local machine but starting from the NOAA-integration 72 hours out - then rerun it using the NOAA-integration 48 hours out - and then 24 hours out (the last one is a normal uEMS run). I guess I could simply start now and integrate until Sunday, but then I'd be folding my own integration into the experiment - an uncontrolled variable; and, anyway, NOAA has been doing these integrations much longer than I and probably are better quality-controlled. The only thing I really bring to the table in my integrations is much higher resolution (0.5km to 1.4km) - which is important for resolving the orographic forcing.

So I guess I'll have to manually download the grib files and rename them, except that probably won't work since they may store the time/date internally..... :^(

I understand that global airmass tracks, etc., probably aren't terribly reliable 3 days out, but I doubt I could do a better job predicting them than NOAA/NCEP does.

EDIT: I guess I should explain that more clearly. If, on Wednesday, I want to forecast for, say, Friday using the nam forecast data started on, say, Wednesday, then in principle I could do that from the data NOAA forecasts which exist. If I do that, then I will be using NOAA's best forecast for Friday. To do that, I would use Wednesday's nam data starting at 72 hours for IC, and also subsequent times for BC's. That data (the nam Wednesday 72 hour data) is actually Friday's forecast nam data. There doesn't seem to be a way to do this in UEMS - since today is Wednesday, uEMS only accepts one of the existing cycles (cycles 00, 03, 06, 09, etc. on Wednesday) as input as the starting point of a forecast run on Wednesday.

I think I can download the 72 hour nam grib data and rename it as if it were Wednesday's data. I'll keep you posted, and if anyone has thought about this and has advice, I'd like to hear it.

Post Reply