Tag Archives: FV3-GFS


There are widely scattered showers around most of the area right now (4:49 PM Sunday)

NexRad Radar at 4:49 PM (source- https://www.weathertap.com )

Most of today’s models missed forecasting these showers entirely or forecast much less coverage, somewhat later, about 6-7 PM.  The NAM and NAM-NEST were particularly disappointing today.  The GFS and ARW2 were close, but no cigar.   Even this morning’s HRRR didn’t predict showers.

However NOAA weather scientists are on top of it, developing the next generation of weather models.

Here’s a forecast from one of their still-experimental SAR-FV3 models  (SAR = “Stand Alone Regional”)

(FV3 = Finite­ Volume Cubed-Sphere dynamical core)

SAR-FV3 Model forecast

I’d love to use this SAR model all the time, but it’s run once a day at 8 PM and takes much of the next day before it becomes available.  It has limited available parameters and no raw data feed available yet.

Over the next two years, some of these new models under development will become available.

As mentioned yesterday, new improvements and time extensions in the HRRR and RAP models are to be released next week.

The process of rewriting the HRRR with the new FV3 core is planned to start shortly.   An FV3 version of the HIRESW is currently running experimentally as well. So I think you can expect large improvements in forecasts over the next few years.

Speaking of FV3 core models, the GFS went to a FV3 core last year and here’s what it’s forecasting for this week’s weather-

GFS Forecast 1AM Tuesday  (Click on image for a larger view.)

The weather setup is complex with an “omega block”, closed upper low and surface low in the Midwest and a coastal low with some tropical characteristics moving northeastward.

The current general consensus for this week’s weather is cloudy, windy and somewhat chilly, but most if not all of the rain will be suppressed to our south and east by high pressure in eastern Canada. 

Things should settle down by Memorial Day Weekend.


Today, with its morning model run, the National Center for Environmental Prediction (NCEP) will have upgraded the GFS model (version 14) to the new and improved FV3-GFS (GFS version 15.1.1). The FV3-GFS becomes the main global model of the National Weather Service today.

The FV3-GFS (Finite Volume Cubed Sphere) has been in development for well over 10 years and has been in testing for the past three. It uses a different three dimensional geometry to reduce errors inherent in all numeric weather models. It includes different modeling physics and new parameters.   Significant information on the FV3-GFS can be found here.  Background information is here.

The model data has been available on university web sites in a very limited fashion for awhile.  Over the past two weeks, the NCEP made the data available on the main weather model download server, “NOMADS”.

The FV3-GFS has been heavily evaluated and is considered “equal or better” than the current GFS.


This weekend’s weather will be the nicest we’ve had in many weeks and the forecast  will be easier as well —  a  far cry from the boundary conditions with a stalled front  we’ve had to deal with over past weekends.

Both Saturday and Sunday will be sunny.  High on Saturday 84 and high on Sunday 90!

[su_note note_color=”#d9f2da”]Sat AM Update: High temperatures have shifted a bit higher for Saturday 86 and Sunday 91.[/su_note]

We’re finally entering a summer-like weather stretch. Monday’s high is expected to be 96-98 with increasing dew points.

By the way, the NWS announced today an expected major update in the GFS model. The new GFS model, to be referred to as the FV3-GFS,  will become operational in mid January 2019, in time to improve snow forecasts.

From what I could infer from reading about the new model, its improvement is in part due to changes in the geometry of how the globe’s atmosphere  is broken down into a 3 dimensional mathematical grid.

The new geometry, a Finite Volume Cubed Sphere, (hence FV3) allows for better computation at the edges of each cubed grid, and reduces approximations where some of the physics would ordinarily result in infinite mathematical series or unsolvable equations.  This should result in better forecasts .

The pre-release version of this new model was used for this forecast’s high temperatures. We’ll see if the highs on Monday really reach 96+.