Upgrades to US forecasting including the FV3 forecast model

[Bill Hark]

Nice Time magazine article. It will be interesting to see how this new upgraded model verifies when it comes online in 2019.

http://time.com/longform/better-storm-prediction/

Bill Hark

 

[rdale]

It's about as bad as the existing GFS so far... Has its good times - but on an overall scale it's at the bottom of the ladder.

 

[Mark Blue]

It sounds promising but I have no knowledge of what makes one model better than another, just end user experience which means little, if anything. I’ll be curious to hear what @Jeff Duda thinks about the FV3 since he now works with model development for a living.

I’ve often wondered why the NWS hasn’t created more reporting stations across the CONUS grid to obtain more raw data to feed into our models. I presume money is at the top of the list as to why it hasn’t happened, but if they budgeted for a few new stations to be built per year it would soften the monetary blow. It could also be stretched out over several years until complete. Not full blown NWS stations but scaled down and automated as much as possible. We have to eventually catch up to the Europeans and the only way as I see it is to invest in our infrastructure. Easier said than done I suppose...

 

[MClarkson]

According to my verification stats for surface temp and wind, the regular flavor of the GFS already outperforms NAM, NAM-hires, RGEM, and GEM. This is quite robust... it is rare to find a station where the GFS doesnt win, and often wins by a lot. I do not have UKMET/EURO grib files to include unfortunately. Interested to see how FV3 stacks up... havent bothered with the exp. runs yet.

 

[Jeff House]

I have been looking at the FV3 for a couple months with a work subscription. Frankly I feel like the old GFS is better mid-term.

 

[Jeff Duda]

I don't work with the FV3, so I generally do not know how it performs. I don't get the impression that it is particularly bad or anything, either, though.