MegaMike

I have a hard time trusting the NAM 3km due to an issue (possibly patched or related to a vendor?) caused by its domain configuration and dynamics/physics options... If I remember correctly, if a large-scale disturbance moved too quickly, a subroutine will sporadically calculate an unrealistic wind speed (only aloft) at certain sigma levels... Can't predict atmospheric flux if a forcing field is kaput. I need to find this case study... It's pretty interesting and it happened twice from ~2016-2020. The erratic nature of NAM is off-putting too, but I'm thinking that's related to its ic/bcs... Before a system materializes, you're solely relying on ic/bcs from a regional modeling system. Those regional modeling systems cannot effectively predict, or even initialize, small-scale/convective features which leads to significant error over time. The same forecast volatility will likely occur to other mesoscale modeling systems if they ran past 48 hours. While on topic: I would like to see vendors start using the RRFS (https://rapidrefresh.noaa.gov/RRFS/; currently under development)... especially for snowfall and precipitation forecasting. It's a unified, high resolution modeling system which runs every hour. Consider it like the GEFS, but with the HRRR. The NBM/HREF are great, but its a waste of resources to post-process different mesoscale modeling systems onto a constant grid. Let's just be happy nobody mentioned the NOGAPS. Also not a fan of the ECMWF AI... That model is for data scientists, not meteorologists.

Next panel:

At this time, nobody should be relying too heavily on global (GFS, CMC, ECMWF) or regional (RGEM 10km, RAP 13km, NAM 12km, etc...) modeling systems. Stick with the mesos (HRRR 3km, WRF ARW, WRF ARW2, NAM 3km, etc...) and rely on data assimilation/nowcasting to follow any trends. That said, the 12z HREF (ensemble of mesos) increased mean snowfall for most areas in eastern MA and all of RI (compared to last night at 00z). It also expands accumulating snowfall westward.... More snow may be falling after this, but I'm too impatient to wait for the next panel... Follow it here: https://www.spc.noaa.gov/exper/href/?model=href&product=snowfall_024h_mean&sector=ne&rd=20241220&rt=1200

Good eye/question! The most recent version of the GFS has 127 hybrid sigma-pressure (terrain following at the surface and pressure aloft) vertical profile layers. In the plot you attached, the vendor is/appears to be plotting vertical profile data that has been interpolated to mandatory isobaric surfaces (1000, 925, 850, 700, 500, 400, 300, 250, 200, 150, and 100 mb). I'm assuming the same vendor uses a diagnostic categorical precipitation type/intensity field to plot precipitation type/intensity, which considers all 127 native vertical profile layers. So the Skew-T is only showing you a select few isobaric surfaces although the native model output possesses plenty more.

I published a paper for the EPA regarding an air quality modeling system/data truncation, but I haven't collaborated with NCEP on atmospheric models/simulations. I do have experience with a bunch of different modeling systems though (in order): ADCIRC (storm surge), SWAN (wave height), WRF-ARW (atmospheric and air quality), UPP (atmospheric post-processor), CMAQ (air quality), ISAM (air quality partitioning), RAMS ( atmospheric), ICLAMS (legacy atmospheric model).

Just to be clear, the HREF and NBM are great tools (I specifically use them for precipitation forecasting), however, we can do better.

No problem, friend! I run simulations fairly regularly so I thought I'd chime in

The HREF is a Frankenstein model since it requires post-processing to obtain fields on a constant grid. It's not a standard modeling system like the HRRR, NAM, GFS, etc... NWP requires a lot of static/time varying fields to run atmospheric simulations: vegetation type, elevation height, ice coverage, etc... By interpolating data onto a grid (or even transforming a horizontal datum), you degrade model accuracy since this information is lost. The HREF exists simply because we created a high-resolution ensemble with what we currently have.

Consider the RRFS an ensemble consisting of multiple (9 members + 1 deterministic) high-resolution (3km) simulations on a constant grid (unlike the HREF and NBM). In other words, it's similar to the HREF, but without any post-processing... The HREF requires post-processing since its ensemble members have various domain configurations (which is a bit taboo). Based on < https://gsl.noaa.gov/focus-areas/unified_forecast_system/rrfs >, it's set to replace the NAMnest, HRRR, HiResWindows, and HREF modeling systems. Besides what I mentioned above, the lateral boundary conditions come from the GFS (control) and the GEFS (members). All share the same core (FV3). Currently, they're testing a bunch of different options to improve the initial conditions (data assimilation) of the RRFS. Overall, it sounds like it'll be a significant improvement over the HREF and NBM once the RRFS becomes operational (and well tested). Note: I can't find much on the RRFS' performance... Based on what I've seen so far, It performs better in terms of reflectivity detection < https://www.spc.noaa.gov/publications/vancil/rrfs-hwt.pdf >

You don't think our modeling systems are trusted? Most high-resolution modeling systems perform well within 12 hours (especially once a disturbance is properly assimilated). Additionally, most global models perform reasonably well within 4 days at the synoptic level. If you expect complete accuracy for moisture/precipitation fields, you (not you specifically, just in general) don't understand the limitations of NWP... Our initial conditions/data assimilation, boundary conditions, parameterizations, and truncations (dx,dy,dz increments) leads to significant error over time which aren't necessarily related to a modeling system itself. If we could perfectly initialize a modeling system, theoretically, there would be little to no errors post-initialization. You can't say the same thing about an AI model since it's likely trained on forcing variables such as temperature/moisture (at the surface and aloft) and is not simulated using governing equations and fundamental laws which the atmosphere adheres to. I wrote, 'AI can be used to improve the accuracy of NWP output that have a known, and predictable bias...' so if we used AI to correct singular fields prior to initialization and while a modeling system is running, sure... It will likely improve the accuracy of NWP. Bottom line; use AI to assist NWP or to correct fields with known biases. At this time, I don't trust atmospheric, AI models.

I'm not sure why anyone would use AI to forecast the weather. 1) AI is a glorified analog so it won't perform well for anomalous events. 2) We don't know what's incorporated in AI models so we can't correct for known biases. 3) Why use AI when we already have trusted NWP models? AI can be used to improve the accuracy of NWP output that have a known, and predictable bias though. For example, improving wind speed: < https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2018JD029590 >.

The units are mm/hr, not inches/hr.

I totally agree with this. QPF has definitely improved over the past several decades. I just wish people would stop using mesos past 1pbl cycle. Thereafter, their reliability diminishes. It's the equivalent of using the GFS past day 7.

Snow continues to melt on contact in Albany. I'm not expecting much until after sunset. It's been like this since ~9am.

People trash a model if snowfall doesn't produce in their back yard. If someone says 'x' model sucks, I want them to cite an evaluation (post-hoc or operational) of meteorological forcing fields related to heat and moisture both at the surface and aloft. Those fields contributes significantly to the accuracy of NWP... Not snowfall at one location (which is typically calculated from a private website, excluding a couple NWP models). If someone were to run an evaluation of this event, you'll likely find that temperature performed well (sfc + aloft), moisture performed okay (Td or specific humidity at the sfc and aloft), and wind speed performed okay (directional and magnitudinal at the sfc and aloft - worse as you approach the surface). All NWP models struggle with QPF (for a number of reasons) so it's not solely used to determine a model's overall performance. Through my experience, it usually performs reasonable as a bulk statistic. IMO, ensembles should always be used for QPF forecasting. Diagnostic models should be used to determine trends.