Webberweather53
Meteorologist
The NAM cycles early so and is nested within the GFS global domain which means that it may miss some potentially important information that the 18z GFS cycle can use, is not updated w/ our ROAB sounding network data (which is only launched over the contiguous US at 0z and 12z barring there being a special need for more data), and because it's nested within the GFS data near the edge of the NAM's domain may be too similar to the GFS forecast and as the run progresses more of this information from the parent GFS model is communicated towards the center of the NAM's domain. Not to mention the NAM's convective parameterization scheme sucks for the mid-latitudes, but is used anyway because it's computationally less expensive. The convective parameterization scheme the NAM utilizes is called the Betts-Miller-Janic (BMJ) scheme that was developed in 1986 and refined by Janic in 1994. In order for this scheme to activate, there has to be some convective available potential energy (CAPE) and it needs to be at least 200 hPa deep, and it finds the portion of the sounding that is most unstable to check for this. While this scheme is not used in the 3km NAM (which has other issues in its own right namely boundary layer problems wherein it often overdeepens extratropical and tropical cyclones, among other things), the BMJ scheme isn't applicable in the mid-latitudes because the convection here is sheared and has cold pools and is well beyond the parameter space of where the scheme was developed (& initially intended for)... in the tropics. The BMJ scheme also doesnt account for entrainment and detrainment like the Kain-Fritsch scheme and also isn't able to resolve compensating, large-scale subsidence that occurs outside the domain of convection. I'm not saying the BMJ scheme is bad in every situation, it's just probably not the right choice to make when you're in the mid-latitudes. Dr. Lackmann here at NC State has performed a host of numerical weather prediction (NWP) simulations interchanging the KF and BMJ schemes in the NAM and noticed when they inserted the KF scheme into the NAM it significantly reduced the number of spurious, oversimplified, dibatically induced coastal vortices near the SE US and eastern seaboards. Long story short, there's a reason the NAM sucks beyond 48 hours and a lot of it has to do with the convective parameterization scheme in the model (3km NAM doesnt use the convective parameterization because its grid spacing is sufficient to partially resolve convection, but it has major boundary layer problems). While these microphysical cloud processes don't seem like a big deal to most and should be confined locally to where the convection occurs, that's not what actually happens in the real atmosphere. The errors actually non-linearly grow upscale with time and in a few days you can end up with a completely different synoptic scale forecast because a model is using a different convective parameterization scheme... Not to mention convective precipitation (which is what we'll be dealing with the next few days and will be primary player in determining precipitation type over much of the southeast) is traditionally the hardest parameter for the models to forecast so the uncertainty is well above what's climatologically "normal" for a 3-5 day forecast...
