The best numerical weather prediction model on the planet just got a bit better this morning. As of this morning’s run the ECMWF global model has been upgraded to an ~9 kilometer resolution. That’s a higher resolution than the GFS (I think 12 or 13km?). Resolution refers to grid point spacing in the model. Generally the closer together the grid points the more features the model can see. Of course that’s assuming the initialization is correct…a big assumption at times. You can see how ECMWF regularly beats the American model (GFS) in skill scores; I found this on Ryan Maue’s Twitter page:
Note the comparison is for 500mb height in the mid-latitudes. A score of “1” would be perfect, so the higher the better. See the parallel (new) model run for the past few months has been slightly better than the operational (regular ECMWF) run. Today the parallel run IS the operational run with the slightly higher resolution.
But what does it mean???? Well, it’s not quite as remarkable as that famous double rainbow video and I don’t want to spend 3 minutes on it, but…
The model will just tend to do slightly better as it catches a few more details. Also, with slightly better terrain you may see a bit more detail. Here is an example with two images:
- Last night’s ECMWF run for peak wind gusts with Wednesday night storm
- This morning’s higher-res ECMWF run for the same time
You can see a little more detail in the 2nd image in the Coast Range and maybe in the Cascades. It’s going to be a crazy windstorm at the Coast tomorrow evening! Could see gusts in the 65-80 mph range! Both runs show that.
Check out estimated snowdepth for next Monday; same timing with last night’s run and then this morning’s upgraded run…
Notice less “overspill” of contours into the valleys east and west of the Cascades? Similar totals in the mountains but you don’t see as much down below. Especially obvious in Jefferson and Wasco counties in north-central Oregon.
This is all minor stuff, but the point is that global models keep improving at a slow pace. Compare this to 10-15 years ago and it’s a huge improvement!
By the way, I don’t generally show these images on the blog because they are not supposed to be redistributed. They are available through a paid service…WeatherBELL. About $185/year. I think a one-time showing for the purposes of the model upgrade is alright though. You can always give WeatherBELL a try…I think they have a free 7 day trial.
Chief Meteorologist Mark Nelsen