Generally, each year, GFS has one or two updates which improve its performance. On March 22nd, 2021, GFS was upgraded to version 16. This is a large update to GFS.

There is a service change notice which describes the update, in detail.

There are many technical improvements (better physics! more vertical layers! etc) which I am not qualified to comment on, but I generally put them into the category of making the model better. The service change notice is the best source for a list of the changes.

For those of you who are in a hurry, the TLDR; version is:

• From now on, please download WW3 wave data from the Ocean (WW3) parameter area in the GFS model.

Below, I’ll highlight a few of the changes that are part of this upgrade.

Wave data.

The main ocean wave model from NOAA is the WW3, Wave Watch III. WW3 used GFS for the atmosphere data in its calculation of the various wave states (wind waves and swell.) With the v16 upgrade of GFS, the WW3 model has been coupled and brought directly into GFS. NOAA is now delivering the WW3 data as part of GFS, calling it GFS Wave.

To reflect this change, the GFS model on the LuckGrib servers is now making this new wave data available as part of the GFS model.

When compared to the new GFS Wave data, the older WW3 model was limited in several ways.

• The older WW3 data had a 0.5 degree resolution. This has now been improved by being doubled to the GFS resolution of 0.25 degrees.

• The WW3 model was only providing 7 1/2 days of forecast data. With the update, wave data is now available for the full GFS forecast duration, 16 days.

• The WW3 model also used to mask its wind data so that it was only available on the oceans. Of course, GFS winds are global, including over land areas, so you can now downlaod WW3 (GFS wave) data and observe land effects on the wind and waves.

Now that waves are part of GFS, you can obtain both atmosphere and wave data in the same download. This means you can now download sea level pressure, wind and wave state in the same download, which is convenient.

Look for the new GFS wave data in the ‘Ocean (WW3)’ section of the GFS model. Along with the wave data, ice thickness and coverage is also being made available.

With these changes, the existing WW3 model has been deprecated. All customers are encouraged to move over to the GFS wave data rather than use WW3. At some point in the future, the WW3 model will be removed from the server cluster. To encourage this migration, some of the WW3 data has been removed. If you want access to wind waves or secondary swell data, look for it in the GFS model.

Wave data. Primary wave, wind wave or swell?

With the WW3 data now being incorporated into GFS, the parameters have been reorganized slightly, to express my preference for which wave data to download.

If interested in ocean waves, most customers should be downloading: combined height of wind wave and swell, along with the primary wave direction and period.

The WW3 model (now GFS Wave) produces three main sets of wave data: wind waves, primary swell and secondary swell. Depending on where you are and what is happening with the atmosphere, either the wind waves or primary swell will be more important - or primary.

The primary wave parameter chooses either the wind wave or primary swell direction and period, at each grid point, depending on which one is more important. The primary wave data, along with the height field: combined wind wave and swell should be the main wave data used by most customers.

You really only need to download the individual components of primary wave, the wind wave and swell data, if you want to study waves in much more detail than most people need.

Sea level pressure.

This change is being documented for completeness - very few people will care about this level of detail…

Back in 2018, I noticed that GFS was not reporting its sea level pressures for tropical cyclones, hurricands and storms accurately. These observations were published in a blog post and a longer article going into more detail.

As a result of this discovery, the LuckGrib servers started delivering a more accurate sea level pressure field, named MSLET, which was repackaged and delivered as PRMSL. After making this change, it was interesting to watch as shortly after the LuckGrib announcement several other weather services followed along.

In the new v16 update of GFS, the PRMSL field has been improved, and is no longer being spectrally truncated. After doing a comparison between the MSLET and PRMSL data, over the oceans, there appear to be no differences. All of the improvements that MSLET had over PRMSL have vanished with this update. There are slight differences in the two parameters over land, but they are not meaningful (in the 1/10th of mb.)

As a result, LuckGrib is now once again downloading and delivering pure PRMSL data from the GFS model.

The service change notice for this update contains the following note:

Replace filtered Shuell Sea Level Pressure with unfiltered one using the same ID (PRMSL). The heavily filtered Shuell Sea Level has caused confusion to users due to its inconsistency with the 10m wind field.

Thanks!

Thank you to all of the members of the NOAA team who contributed to this update. We are all looking forward to the future improvements.