After a long delay, the National Centers for Environmental Prediction (NCEP) finally implemented their long awaited and likely final upgrade to the North American Mesoscale Forecast System (NAM) on March 15th.
The changes are extensive. The grid spacing of the NAM nest over the continental US decreases from 4 to 3 km. Calls to major model physics packages are more frequent, updates to cloud microphysics parameterizations (critical for the simulation of precipitation), a new data assimilation system (to create model initial conditions), assimilation of new observational datasets, and more. A complete rundown is available here.
NCEP is also providing NAM output now at higher frequency. I've updated weather.utah.edu so that it is now possible to access the NAM 3-km nest at hourly temporal resolution. The loops are great and should aid in anticipating the onset and ending of precipitation. Below is last night's NAM forecast of 1-h accumulated precipitation and winds for 2100 UTC (3 PM MDT) this afternoon showing the effects of the slow moving cold-frontal precipitation band that is spreading into the Salt Lake Valley this morning.
The older 4-km NAM nest had a major overprediction problem in the mountains. NCEP has told me that they think this new version will be better in that regard, but the proof is in the pudding. We'll have to watch it for a few events and see how it does.
Also, high-resolution model guidance like this has tremendous physical realism. It's quite alluring. However, that doesn't necessarily mean that it is skillful. You can have a very realistic precipitation pattern, but if it is in the wrong place, you have a bad forecast. This can be especially problematic at longer lead times and is a reason why one needs to be cautious using a single high-resolution model rather than an ensemble.
Nevertheless, even with those predictability limitations, I'm hoping that we will find that this upgrade is a significant step forward.