0Z Guidance Discussion 12/21/2010

[PhineasC]

It's second in verification...sorry dude.

Yeah, but how is that measured, exactly? I am just a weenie but the UKMET seems to always be jumping around at the surface.

[Ji]

Yeah, but how is that measured, exactly? I am just a weenie but the UKMET seems to always be jumping around at the surface.

Ive never seen the Ukie ever get a solution right but its 2nd in verification. Seems like we have alot of work to do to fix the modeling

[dtk]

Yeah, but how is that measured, exactly? I am just a weenie but the UKMET seems to always be jumping around at the surface.

By WMO, 500 hPa anomaly correlation is the (much overused, too heavily emphasized) standard. For the past month:

UK is statistically tied with the GFS, both of which are behind the EC in the NH (but the UK is ahead of the GFS in the SH). I can dig up some other stuff, like root mean square error for various variables and levels if you'd like.

[SteveB]

If guidance goes towards the UKIE today and it scores a coup, I will eat Ji's socks for Christmas dinner!!

[aldie 22]

If guidance goes towards the UKIE today and it scores a coup, I will eat Ji's socks for Christmas dinner!!

He's Saudi...he doesn't wear socks

[dtk]

The GFS ensembles are basically inferior runs of the GFS model.

This is rubbish....it's reduced resolution and intentionally perturbed (NOT "inferior). Do you even understand the point of ensemble systems?

[yoda]

Didn't see this posted in this thread... ECMWF ensemble mean

[A-L-E-X]

By WMO, 500 hPa anomaly correlation is the (much overused, too heavily emphasized) standard. For the past month:

UK is statistically tied with the GFS, both of which are behind the EC in the NH (but the UK is ahead of the GFS in the SH). I can dig up some other stuff, like root mean square error for various variables and levels if you'd like.

Interesting. Is the UKMET ahead of the GFS in the southern hemisphere because of the larger ratio of ocean to land in the SH?

[OSUmetstud]

Yeah, but how is that measured, exactly? I am just a weenie but the UKMET seems to always be jumping around at the surface.

the verification page is pretty comprehensive if you want to take a look.

The UKMET has not been consistently beating the GFS like it used to...it's basically on par.

I agree with others it seems to be prone to some radical solutions (just like the GFS is)

EMC model verification

[am19psu]

Guys, there are better threads for this model discussion. Let's keep this thread on track with analysis of current solutions.

[donsutherland1]

A useful link that provides a general overview of ensembles/ensemble forecasting:

http://www.hpc.ncep.noaa.gov/ensembletraining/

[thegreatdr]

This might be true across the whole hemisphere, but pressure-wise near the lower 48, once you get to day 3, if you're making a forecast issued around 1 pm each day, the 00z ECMWF ensemble mean is best, followed by the 00z ECMWF and 12z GEFS mean, followed by the 12z GFS and 00z GEFS mean, followed by the 00z GFS, with the 00z UKMET/00z Canadian/00z NOGAPS jockeying for last. The 00z bias-corrected NAEFS has rapidly lost skill since October, and it is now near the lower end of the verification spectrum. There is some concern that this may continue after NOGAPS members are added to it.

The UKMET's pressure field was superior from March into July this past year because it doesn't produce high pressure in higher elevations of the Rockies unlike the other guidance, and generally had lower pressures over the Southwest, Great Basin, and Mexico which is great in the warm season. This time of the year though, with cool air overspreading the lower 48 and a near disappearance of a thermal low over North America, those qualities significantly boost its error. The Canadian has trouble phasing systems too quickly (a problem also seen on the ECMWF beyond 120 hours), which then produces anomalously high pressure in the systems' wake. You combine this with its tropical cyclone happy/convective feedback problem in the tropics which leads to its minimal use near the United States pressure-wise. And just as a note, before the last two upgrades, the 00z ECMWF used to outperform its ensemble means on days 3-5. But over a 90 day stretch, this hasn't been true since March. That helps explain the updated pressures issued later this morning by HPC and why HPC might use the 00z ECMWF ensemble mean solution, when available and when reasonable aloft. They could just as well have gone with a slightly modified 100% EC mean pressurewise, if they wanted to beat the guidance most days.

DR

By WMO, 500 hPa anomaly correlation is the (much overused, too heavily emphasized) standard. For the past month:

UK is statistically tied with the GFS, both of which are behind the EC in the NH (but the UK is ahead of the GFS in the SH). I can dig up some other stuff, like root mean square error for various variables and levels if you'd like.

[Ender]

This is rubbish....it's reduced resolution and intentionally perturbed (NOT "inferior). Do you even understand the point of ensemble systems?

Not sure but maybe,

To mitigate/account for compromises in the physics and uncertainty in the initial conditions. Ultimately the goal is to establish the entire solution space of all possible solutions. Theoretically the solutions would cluster towards the very most likely of solutions. We lack computational power to build an "entire solution space" and we're a good five decades away. The mean of many runs of lower resolution models have been shown to be more accurate than the mean of multiple runs of the same high resolution model with slightly perturbed initial conditions.