Downscaling the NAM and providing precipitation probability forecasts using "smartinit" processing

In 2008, NCEP began "in-house" processing of the North American Mesoscale (NAM) model smartinit code. This code had previously been run by local National Weather Service offices to provide a first guess for NDFD elements, but the local version of the code used a coarse, isobaric NAM output grid for the processing.

Running the code at NCEP allows the use of a high-resolution grid on native levels. This detail is critical for the downscaling process which adjusts temperature, dew point, and winds to a very high-resolution terrain field. Examples of the downscaling procedure over the CONUS, Hawaii, Alaska, and Puerto Rico will be shown.

The greatest benefit to running the code at NCEP is the flexibility to adapt the probability of precipitation (PoP) products to an updated concept. The original code based its PoP primarily on high model predicted precipitation. A large amount of precipitation predicted by the model resulted in a large PoP value. The new method gives some weight to the amount of precipitation predicted by the NAM but also uses probability fields generated by the SREF to give a more realistic product.