Models Update (2022 Chase Season Prep)

Jul 5, 2009
Newtown, Pennsylvania
In preparation for the 2022 chase season, I would like to update my understanding of the “status” of the various global and CAM models that are most commonly used for chasing. I know that many of the models are upgraded each year, that certain upgraded versions may be experimental or operational, and that even as some biases are corrected, others remain and new ones become apparent. Without being a professional meteorologist, it’s impossible to stay on top of all the latest in the field of NWP (heck, I can barely keep up with stuff in my own profession 😏), so I (and I assume many other ST members) would appreciate any meteorologists willing to share a summarized update of major changes since last season, and/or any useful resources or papers about the topic (as long as not too complex for a relative layman). The objective is to get up to speed on the latest and to provide context for forecasting this spring, i.e. what should we be keeping in mind as we look at various models, based on recent upgrades and/or newly discovered strengths, weaknesses, and biases?


Feb 16, 2014
Norman, Oklahoma
All, weve got our yearly convective training coming up at SPC in March. We use that time a refresher to gear into convective mode after the off-season. It usually contains a summary of CAM changes. Ill try and post an update after that. In the meantime here is a non-exhaustive list of the biases of common models.

  1. HRRR/RAP - Overmixing bias in the boundary layer. Typically see increased drying in low levels on days with strong moisture gradients. its got significantly better in recent years but still there occasionally. Usually manifests as veered low-level winds and decreased surface dewpoints. Can negatively impact MLand SBCAPE calculations. Keep in mind the SPC mesoanalysis is RAP based and can be misleading in certain situations.
  2. NAM/NAM4k/NNMB- Opposite of the HRRR/RAP the boundary-layer biases are usually cool and wet. This can result in increased surface dewpoints and unrealistically backed low-level flow. Also, tend to boost sb/mlcape and can struggle to initiate when strong capping is present given the cooler surface temperatures. Sometimes more useful for cold pool propagation and shallow cold fronts when checking for undercutting.
  3. GFS/GEFS/HRWFV3- Can suffer significant dry bias which negatively impacts CAPE calculations. Be aware of dryline positioning using the FV3 core it can be off. The HRWFV3 also loves to spinup storm-scale vorticity. The HREF was recalibrated when it was added for this very reason. It does often do a decent job at initiation but suggests more supercells than may be realistic.
  4. HREF- Best all-around given it uses a dispersed set of dynamical cores and BL schemes. I highly recommend using the HREF for daily chase forecasting it very rarely leads me astray especially if you know the biases of the input members. The machine learning and calibrated guidance is also very good.
  5. Honorable Mention HRRR SCRAMM- Bit different as this is more statistical guidance but it can be very useful for quick and dirty forecasting with the latest runs of the HRRR. John hart and Ariel cohen put a ton of work into finding the best statistical predictors of severe weather. It doesnt have any true biases per say, but it can highlight corridors of more or less intense activity fairly well.
Again ill try and update more as time goes on. But that's the basics I keep in mind while Im at work or chasing.

Jeff Duda

EF6+, PhD
Staff member
Oct 7, 2008
Broomfield, CO’s impossible to stay on top of all the latest in the field of NWP (heck, I can barely keep up with stuff in my own profession 😏)...
I work at the NOAA Global Systems Laboratory (where RAP/HRRR come from) and even I cannot keep up with all the latest developments. So you're far from alone.

As far as I recall there have been no upgrades/updates from any major modeling center since last year. There have been some changes in graphics offering, mainly with your favorite sites now offering more ECMWF data (since ECMWF made it cheaper/free to access much more of its products).

The most interesting experimental modeling systems of use to chasers is WoFS and whatever else NSSL is messing around with. WoFS is really gaining traction in the world of experimental, rapidly updating, high-resolution, high-impact, high-(whatever other buzzwords will get lots of NSF funding headed your way) forecasting.
  • The experimental FV3 runs at NSSL can be viewed at
  • The classic WoFS viewer is available here: WoFS - Realtime Viewer
  • NSSL is messing with a cloud-based version of WoFS which runs less often, but can be even cooler. Peruse this site for specifics: Cb-WoFS -
Note: for those used to viewing convection-allowing model output from the WRF era, know that WRF is being phased out in favor of FV3. FV3 will ultimately serve as the dynamical core for all NOAA NWP models by ~2024, including the HRRR, which will become known as RRFS. Experimental RRFS forecasts are already running at GSL and you'll see more of this over the next couple of years. The main thing to note with FV3 is how it resolves convective storms -- they won't look the same as they do in WRF. You'll notice a smoother and more circular appearance to the reflectivity field in FV3 compared to WRF. You'll also notice reflectivity and precipitation amounts significantly increase. We are still investigating parameter selections such as horizontal diffusion that impact the texture of these fields, but the above tendencies have been pervasive for the past few years' of experimental forecasts from the FV3, so get used to seeing it.