Model Resolution

[lorenzo]

With the models under some degree of scrutiny I always forget which has higher resolution here and there and at what time frames. After some digging around put this together and hope it is some use as a point of reference, this is not my area of expertise so please edit, add to, correct, grow etc etc

Data for numerical weather prediction is provided by observations from satellites, from the ground (both human and from automatic weather stations), from buoys at sea, radar, radiosonde weather baloons,, wind profilers, commercial aircraft and a background field from previous model runs. The gist of this is shown by this handy pic from a non-blacked out Wikipedia and ensembles depicted from UKMO website.

UKMO Unified Model - Global Model

25Km Resolution 70 Vertical Levels 144 Hours

UKMO North Atlantic and European Model NAE

12Km Resolution 70 Vertical Levels 48 Hours

UKMO - Other

1.5km Resolution UK 70 Vertical Levels 36 Hours

4Km Resolution Surrounding Areas

　

　

ECM / IFS

16 /31km? unsure..

Operational / Deterministic run is Twice Hortizontal Resolution of Ensembles 240 Hours

Operational 60 Vertical Levels

Ensembles 40 Vertical Levels

51 member Ensemble suite http://www.ecmwf.int...techniques.html

　

GFS

35 km / 64 vertical layers High Resolution runs to 192 Hours / 7 days - every 3 hrs

70 km / 64 vertical layers Lower Resolution runs from 192 - 384 Hrs - every 12th hr

22 Ensemble Suite 9 including Control / Operational

GEFS

using current operational GFS

-increase in horizontal resolution from 70 km to 55 km (T254) for 0-192 h, and 70 km (T190) for 192-394 hours

-increase in vertical resolution from 28 to 42 layers for 0-384 hours

http://www.emc.ncep..../GEFS/mconf.php

Edited January 24, 2012 by lorenzo

[Isolated Frost]

I am sure the ECM deterministic run is shown at a 16km resolution- and it's control at 31km, at least according to the EPS grams they give ?

[danm]

Thanks for that.

The GFS high resolution is 35km/64 vertical layers according to what you've posted - perhaps this shows why the GFS's verification stats are usually lower than the ECM & UKMO as their high resolution operational's are 16km and 25km respectively?

Edited January 25, 2012 by danm

[bluearmy]

the GEFS info is for the new parallel rather than the existing normal output.

on the parallel gefs, i think the control runs at 55km res all the way through.

[Liam J]

Afternoon all, just to add - the GFS model was upgraded from 35km to 27km resolution last year and runs at this resolution up to 192hrs. The rest of the run is produced at 70km resolution.

http://www.ncep.noaa...em_announce.pdf

Global Forecast System (GFS) Upgrade Implemented July 28, 2010

An extensive set of improvements to NCEP’s Global Forecast System will be implemented

into NCEP operations on Wednesday, July 28, 2010, following three years of development,

evaluation and testing. One of the major changes is an increase in the model’s horizontal resolution

from approximately 35 km to 27 km. Additionally, a number of improvements have been made to

the model’s ability to represent physical atmospheric processes. These improvements include an

upgraded radiation and cloud package, upgraded specification of gravity wave drag, a higher

resolution grid for hurricane relocation, an upgraded boundary layer scheme, the use of a higher

resolution snow analysis, a new mass flux shallow convection scheme and an updated deep

convection scheme. Also, the model will have a horizontal resolution of 27 km through 192 hours

in support of the GFS MOS development, which is an additional 12 hours from the previous

version. The model’s resolution from 192 to 384 hours will remain at 70 km. Finally, the

availability of three hourly forecast output will be extended out to 192 hours from 180 hours.

The result of this set of changes is improvement in the overall accuracy of the model’s

forecasts. In addition, a significant improvement in precipitation verification scores resulted

from a reduction in the appearance of spurious excessive amounts of precipitations occurring

over small geographic areas. An example of the elimination of these so-called “bulls-eyes†is

illustrated in the maps below.

Edited January 25, 2012 by Liam J

[bluearmy]

re ecm ens:

to day 10 at a resolution of 50 km

days 11 to 15 at a resolution of 80km

running at 62 levels

[Gowon]

Regarding the ECM operational - How many days are run at 16km resolution?

[bluearmy]

Regarding the ECM operational - How many days are run at 16km resolution?

the op only goes to day 10. all at 16km resolution (thats why each frame takes so long to come out)

[bluearmy]

now that the gefs parallel has gone live, it should be noted that post day 10, the gefs are running at a higher resolution than the ecm ens. also, the gefs control post T192 should be given as much weighting as the gfs op as it is running at a slightly higher resolution (though at fewer vertical layers).

the general point is that, theoretically, the GEFS post day 10 with their trending should now be seen as comparable to the ecm ens.

Edited February 16, 2012 by bluearmy

[lorenzo]

TY blue, I think the other night when I was steaming i was tryingto get my head round this

[summer blizzard]

Anybody know where the GEM model stands in resolution, it seems to be on a par with at least the GFS in my opinion.

[Zenarcher]

Anybody know where the GEM model stands in resolution, it seems to be on a par with at least the GFS in my opinion.

I believe it does 15km up to 48 hours then after that switches to 35km up to day 10. So the same as GFS but GEM's short range is run in a higher res and it does 10 days of 35km res meanwhile the GFS only does 7 days of 35km res.

Edited March 11, 2012 by weathermaster

[bluearmy]

I believe it does 15km up to 48 hours then after that switches to 35km up to day 10. So the same as GFS but GEM's short range is run in a higher res and it does 10 days of 35km res meanwhile the GFS only does 7 days of 35km res.

gfs op is 27km out to T192

thanks for posting re gem horizontal resolution. do you have info re the number of layers it operates on ??

[Zenarcher]

gfs op is 27km out to T192

thanks for posting re gem horizontal resolution. do you have info re the number of layers it operates on ??

Its hard to find any information on the GEM model but after looking around I came across something saying it uses 27 vertical sigma layers. I'm not sure what that means maybe GEM works differently to the others? I have a few PDF links about the GEM model if anyone wants to read through them I can post them here if you want.

Edited March 14, 2012 by weathermaster

[johnholmes]

sigma layers are those that, especially close to the ground are in close conformity to the actual terrain beneath the spot location

[AWD]

Apologies for bringing an old topic up but didn't know where else to put this;

http://blog.chron.com/sciguy/2012/10/in-the-model-wars-europe-kicks-north-americas-butt/

An interesting read on European & American model verification.

[Interitus]

Though resolution is obviously very important, ultimately the quality of the forecast depends upon the quality of the data fed into the model. This can be seen by looking at the northern hemisphere plots of the GFS on eg. meteociel - in particular, the model is able to resolve the temperature influences of bodies of waters such as the world's 4th largest lake, the Aral Sea in Kazakhstan, which at this time of year shows as a markedly warm area in the 2m temperature plots, tremendous.

Except the Aral Sea basically doesn't exist anymore!

[lorenzo]

ECMWF Winter newsletter contains some interesting reading on the performance of the model / ERA20C reanlysis / Ensembles.

Also many other articles for folks to get geeked out on !

http://www.ecmwf.int/publications/newsletters/pdf/134.pdf

[bluearmy]

that indicates the ecm ens members are running at 32 km horizontal res out to day 10 and 62km thereafter. thats pretty well double the resolution i thought pre day 10.

51 members running at at only 5 km behind gfs op - makes you wonder how they dont get it right all of the time!!

looking forward to the upgrades later this year. 129 vertical levels !!!

coupled with the ncep improvements to the gfs, maybe we can look forward to some better modelling next winter ??

[bluearmy]

UPDATE:

Since late june, ecm op has been running at 137 vertcial levels (increased from 91). Later in november, there is a proposed upgrade to the ens suite which will see the vertical layers increase from 64 to 91 layers. Significantly, this takes the top of the model to 0.01hpa ( currently it only goes to 5 hpa). Once, implemented, this gives the ens suite full view of the strat and will enable it to see downwelling waves from the very top which was something clearly absent last january.

Overall, ecm should be a better model this coming winter season than it was last.

[bluearmy]

The upgraded ecm ens run for the first time (not in parallel) this tuesday 12z.

The main improvement in veryical layering is in the strat although looking through the performance of the new cycle in test reveals some quite markd improvement over the current model.

Edited November 16, 2013 by bluearmy

[bluearmy]

The upgraded ecm ens run for the first time (not in parallel) this tuesday 12z.The main improvement in veryical layering is in the strat although looking through the performance of the new cycle in test reveals some quite markd improvement over the current model.

 

the links are as follows:

 

http://www.ecmwf.int/products/changes/ifs_cycle_40r1/

 

 

http://www.ecmwf.int/products/changes/ifs_cycle_40r1/scorecard.html

[Jax]

Been mentioned in our local thread, but anyone tried the FIM recently?

 

Not yet ready for release mind.

[bluearmy]

Been mentioned in our local thread, but anyone tried the FIM recently?

 

Not yet ready for release mind.

 

seems like its been going for years but i never came across it before. running at 15km, its equiv to the ecm op for horizontal resolution.  maybe i'll have a look at it in future. wonder if anyone has any other info about it ?

[Jax]

seems like its been going for years but i never came across it before. running at 15km, its equiv to the ecm op for horizontal resolution.  maybe i'll have a look at it in future. wonder if anyone has any other info about it ?

Well this is due to replace the GFS next year I think.