
I recently wrote about the role of AI as it comes to forecasting, and the adaptation that will come to numerical modeling. This isn’t a one to one story relationship, but the North American Model, the NAM, which was one of the two primary models for decades (along with the GFS), ran for the last time about a week ago.
The NAM initiated off of data from the GFS at the perimeters, and because it wasn’t as concerned with the global perspective, it was able to get more granular with model output, both spatially and temporally. A drawback, then, was the lack of continuity at the periphery, since it was adopting another model’s information, which led to the accuracy of the model to degrade more quickly than the GFS, even if it was more accurate in the short term.
In the last few years, the WRF, a more local, shorter term model has become popular, particularly in severe weather forecasting, both in the summer and winter,, and this had largely taken over for a lot of the forecast scenarios that the NAM originally served. The WRF can run locally, and does so hourly.
There is still room for a regional model, however, with a higher refresh rate than the GFS, and the replacement model is already running. The RRFS (Rapid Refresh Forecast System) is online, and will serve to cover the ground formerly covered by the NAM. What is really exciting, is that as confidence grows in the models success, it will also eventually replace the WRF.
We will see how well the new forecasting system runs. It’s certainly and indicator that upgrades in compute ability are coming to the weather industry, as the RRFS will take on not only the NAM’s workload but also other models, with different run times and scopes. The NAM was the best known and longest serving, and will be the most notable departure. Undoubtedly, this is consolidation, but it is also progress.






