Skip navigation.
Home
uk.sci.weather resources

Information on the various models

(NB: information presented with no warranty!)

Although a nominal data initialisation time is shown (00, 06 UTC etc.), modern NWP analysis routines 'assimilate' data over a period either side of these times. With integration of asynoptic data (i.e., not at standard times), plus the use of previous operational (or special pre-analysis) 'runs' to provide background fields, the use of such DT (data time) indicators is more a convenience for us mortals!

A computer completes its calculations quite quickly with the modern 'super numbercrunchers' in use, and the computations are in any case performed using code that is written to optimise speed: however, always bear in mind that there will be a delay in availability to us on the Internet due to the post-processing of the raw NWP data into a format that can be squirted onto data-networks.

 Model  UKMO- GLOBAL
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00 & 12 (but other runs are performed to 'update' the analysis)
 Number of full 'runs' per 24hr:  2
 Out to T + ..... (hr)  144
 Usually available:
(on the Internet .. the output will be available in a more timely fashion on 'closed' networks. )
 0500 - 0800; 1800 - 2100
 Other information:  Currently (winter 2005 / 06) the UK Global model has the following specifications:
~40 km horizontal resolution & 50 levels vertical resolution: a change in late 2005 put several additional levels into the stratosphere (and extended the 'top' of the model too) as well as reducing the grid-length from 60km: I understand that it is intended to go to 70 levels (twenty more in the troposphere) sometime in 2006.
 Note that a limited-area, finer-resolution (~12km) model [ North Atlantic European / NAE ] is run 4 times daily, with an earlier cut-off time, out to T+48. However, this is not (generally) available on the Internet. It currently takes it's boundary conditions from the previous main Global model run, updated by a 're-analysis' phase.
Other models, at 'mesoscale' resolution (~11km or less) are also run, using the above for 'boundary' conditions - these are also NOT available publicly. Of particular note is the future operational running of the UK 4KM model, expected late 2005 or early 2006. (see HERE for more on UKMO models.)

 

 Model  NOGAPS
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00, 06, 12 & 18
 Number of 'runs' per 24hr:  4 (but 2 via some Internet portals)
 Out to T + ..... (hr)  144
 Usually available:
(on the Internet .. the output will be available in a more timely fashion on 'closed' networks. )
 On the US Navy site .. 0500 - 0600 & 1700 - 1800; On other sites, circa 0800 & 2000.
 Other information:  On the US Navy site (if you get past the various security checks etc.), you can get intermediate hour runs as well.

 

 Model  NCEP - GFS (formerly: AVN)
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00, 06, 12 & 18 (not all 'runs' are carried by all Internet sites.)
 Number of 'runs' per 24hr:  4 (but 2 only to some Internet sites)
 Out to T + ..... (hr)  384 (high resolution to T+180/7.5 days; lesser resolution thereafter .. to 16 days) [ see note below re: resolution ]
 Usually available:
(on the Internet .. the output will be available in a more timely fashion on 'closed' networks. )
 about 4.5 to 5 hr after initial time (e.g. 00Z issue usually available by 05Z), but perhaps earlier on the US Navy site. Note that the first 180hr will be available fairly quickly; the remaining files (at courser spatial and temporal resolution), a little way behind.
 Other information:  NB: Now (since April 2002) the primary model run from NCEP - anything labelled 'MRF' are essentially 'look-alike' output based on the full Global AVN run as above.
[ Historical: Up to April, 2002, the 'MRF' (medium-range forecast) run was output once daily, to T+384, with the same physics, mathematics etc., as the (then) AVN, (now) GFS, but with a later cut-off time to allow extra data to be used - supposedly to achieve a better end result. Tests showed that it was no longer necessary to do this, hence the standardisation on the 4-times daily GFS (AVN) as above. (to emphasise, although these models run out for over 2 weeks, don't take each time-step slavishly; major errors can occur beyond even 3 or 4 days lead time.)
Resolution: Currently (Winter 2005/06), the GFS has an approximate horizontal resolution of ~40km up to T+180h, and ~80km thereafter; there are 64 vertical levels. However, bear in mind that model development may render these figures obsolete: visit this site for updates.

 

 Model  ECMWF
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00 & 12
 Number of 'runs' per 24hr:  2
 Out to T + ..... (hr)  240 (i.e. 10 days) ... but limited output to Internet users .. either 6 or 7 days according to site, fields etc.
 Usually available:
(on the Internet .. the output will be available in a more timely fashion on 'closed' networks. )
 On the EC site, and those carrying such data (e.g. Wetterzentrale), then about 6 to 7 hours after data-time (e.g., 00Z run available between 06 and 07Z). However, where additional processing is required, then a considerable delay from this time may be encountered.
 Other information:  This model runs out to 10 days, but only the first 6 days (7 days for mslp fields) are freely available via these sites.

 

 Other models  DWD: GME (Global Modell Ersatz) (or Global model replacement)
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00, 12 (& a short update run at 18)
 Number of global 'runs' per 24hr:  2 (for full output)
 Out to T + ..... (hr)  174 (i.e. just over 7 days) ... but limited output to Internet users.

 

 Other models  MétéoFrance: "ARPEGE" (Arpege is the acronym for Action de Recherche Petite Echelle Grande Echelle: Research Project on Small and Large Scales. This is the successor model to the EMERAUDE baseline model.)
 Data times: (UTC)
(i.e. the nominal time for initialisation, raw data ingest etc.)
 00, 12
 Number of global 'runs' per 24hr:  2
 Out to T + ..... (hr)  240 (i.e. up to 10 days) : NOTE: French model output to Internet users very limited.