A big upgrade to the GFS model is planned to go live in November 2014.

Read about it here:

http://www.emc.ncep.noaa.gov/GFS/impl.php

Key changes include:

Finer horizontal grid, 23km to 12km
High resolution deterministic forecast extended from 8 days to 10 days
Output datasets available to 0.25deg in addition to current 0.5deg
Better handling of gravity wave forcing and drag

What this means:
- GFS still isn't likely to verify as well as ECM because it will continue to run 4x daily with a lighter data payload. 

- Automated forecasts should be marginally better across the board and possibly noticeably better for some locations due to the increased resolution of the output datasets where they are used. In the UK locations on the south western peninsula may benefit significantly

What this doesn't mean:
- A forecasting revolution. We're at the stage and have been for a long time when finer model grids and higher resolution output datasets lead to marginal improvements

And just in time for the "silly season" too.

Finally catching up with efforts this side of the pond!

During my dissertation project I discovered just how much finer the resolution can go when working on a regional scale instead of global; the models used were at 1.5km, 0.5km and 0.2km resolutions!

Here's a snapshot of what a 200m model looks like when modelling deep convection:

The values on the y-axis are the number of grid-lengths, so 200 is 200 x 200m = 40000m or 40km.

I think that's about as much as I'm allowed to put across at the moment - though the dissertation will one day become freely available in the departmental library in Reading, and perhaps online somewhere.

Interesting news. Thanks Brian.

They seem to upgrade the GFS model every year.

I'm not entirely convinced a finer resolution leads to a more accurate outcome as a not insignificant factor is the quality of data input.
Also SC should that grid be a data cube in m3 

Improvements occur with the thermodynamics and the like, meaning better representation of convection, but there is a fundamental limit to how accurate the models can be regarding where small scale features are at specific times, so we can never realistically expect an accurate prediction for then there will be a shower at Wimbledon, for example.

Frontal features tend to verify better, but the exact timing remains a target behind bulletproof glass.

Regarding the grid, there is a vertical element - I was just referring to the distance latitudinally in my earlier post.

Having said that, the resolution is considred to be 200m2 with 140 vertical levels - twice what the operational UKV model has, as this was needed to prevent erroneous tendencies that were observed in earlier studies, such as excessive cell development and organisation - though my study found evidence to suggest that this wasn't entirely eradicated.

Cheers SC and good luck with the research - keep us updated!

Thanks, I enjoy being able to discover new things as I work - in fact with any luck, I'll some day (soon I hope) land a PhD studying convection or teleconnections, something like that, in which case I'll be able to continue studying and discovering while earning a bit of cash along the way 

Alternatively, I'll work as an operational meteorologist at the Met Office. Whatever opportunity comes my way really 

This is scheduled for December 9th 2014.

...which will of course be right after an epic cold spell handled terribly by the current model 

The GFS upgrade is now scheduled for December 17th. The 'exec summary' is:

Results are reasonable
Improved precipitation skill scores
Improved hurricane track in Atlantic and Western Pacific, but worsened in Eastern Pacific; Reduced intensity errors in all basins.
Reduced mid-latitude storm track errors.
Reduced global mean temperature bias in the upper troposphere; strengthened (improved) tropospheric winds but slightly weakened stratospheric winds.
Reduced nighttime 2m temperature cold bias over the Northern Great Plains. Large biases still exist in Northeast and Southwest.
improved 500-hPa HGT AC in both the Northern and Southern Hemispheres.

I'll start making the parallel run available on TWO next week. 

GFS GSM v12.0.0 parallel run is now available on the TWO chart viewer. Switch between Parallel and Operational then use the GFS charts as normal:

http://www.theweatheroutlook.com/twodata/datmdlout.aspx

- The parallel run isn't currently in production so may not always be available (go live is scheduled for December 17th but is subject to change)
- Most but not all of the charts available for the operational run are also available for the parallel run

The horizontal resolution of the underlying model is 13km.

The resolution of the datasets being rendered is currently:

0 to 120 hours 0.25deg
123 to 180 hours 0.5deg
183 to 384 hours 1.0deg


I may change these.

The charts on TWO will in future be stepped at 3 hour intervals out to 240 hours instead of 180 hours but the chart viewer hasn't yet been updated to give direct links to 183, 186.....237.

The official message is:

The 30-day parallel evaluation of the 2014 GFS upgrade will be restarted soon. Several users noticed issues that are being fixed by the GFS developers. Implementation will occur in January 2015.