wxsniss

Agree, could be both at play AIs are fuzzy/probabilistic at H5 (potentially a weakness or a strength depending on context), but here's a comparison of 12z runs today, for Monday 0z timepoint... AIGFS does have significantly better tilt that could be responsible for northwest extent of QPF... and it also seems to have an overdone spread of QPF (also to eastern extent) in comparison to legacy GFS:

Same. In case people missed it and are interested... below is a thread for general discussion on AI guidance beyond the Jan 18 storm: https://www.americanwx.com/bb/topic/62517-skynet-rise-of-the-forecast-machines-general-discussions-on-ai-guidance/ While I "do not have a lot of experience utilizing a lot of the AI guidance" (quoting Box AFD from this morning), some things I highlight in other thread relevant to Jan 18: • AI models tend to have broad QPF swaths that are overdone in areal coverage and not supported by physics • major weakness that is on display here: we have no idea why AI models do what they do... Why is AIGFS so robust here? Is it seeing something that physics models are not? Are the legacy models too sensitive to smaller features (e.g. the other day I suggested feedback between vorticity and convection sweeping off southeast coast hampering cyclogenesis), or is AI guidance not sensitive enough? If AI guidance changes, why did it change? Will that change be abrupt? These models are totally nebulous. Jan 18 is shaping up to be a great showdown.

Most recently exemplified by this imminent showdown for Jan 18 where AI-GFS has been consistently and substantially more impactful than legacy GFS… …which is correct TBD... and because lots of people are interested in this topic, I thought I’d start a thread to encompass general debates, updates, and verification tally of AI vs. legacy non-AI / physics-based forecasting guidance. AI forecasting guidance has become operational and easily accessible in the past year. This feels like the first winter we are routinely considering these solutions. Some things this thread can encompass: • rolling tally of AI vs. non-AI performance • situations where AI does better / worse • news of model upgrades and rollouts • Ex Machina “You shouldn’t feel sorry for her. Feel sorry for yourself.” reflection: Do we want perfect AI guidance? So much of the joy in this hobby / profession is the suspense of uncertainty, dissecting vorticity and trends and pattern recognition, the thrill of the chase leading up to a storm that sometimes matches the actual storm. Will we lose this as AI improves? I summarized some relevant information here (yes, ironically with the help of ChatGPT). Major models today: • ECMWF AIFS (Artificial Intelligence Forecasting System) - Became operational early 2025, running alongside the traditional ECMWF - Supposed to be very fast and computationally cheap relative to ECMWF - Recently updated to incorporate “ physical consistency constraints through bounding layers, an updated training schedule, and an expanded set of variables”: https://arxiv.org/abs/2509.18994 • AIGFS (NOAA / experimental guidance) • GraphCast (Google DeepMind) - Claim skill competitive with global models at medium range: https://www.science.org/stoken/author-tokens/ST-1550/full) https://deepmind.google/blog/graphcast-ai-model-for-faster-and-more-accurate-global-weather-forecasting/ • FourCastNet (NVIDIA) https://build.nvidia.com/nvidia/fourcastnet/modelcard What’s driving the rollout? • Speed - AI global forecasts run in seconds to minutes, not hours • Cost & compute efficiency - Orders of magnitude cheaper than traditional NWP - Lower barrier for ensembles, re-runs, sensitivity testing • An arms race --- @Typhoon Tip articulated this nicely in the Jan 18 storm thread - ECMWF, NOAA, tech companies—everyone wants to be first (or not be left behind) • Decent skill at medium range - AI models are particularly strong at large-scale pattern evolution (trough/ridge placement, TC tracks, etc.) How have they performed so far? Strengths • Synoptic pattern recognition in mid-range, like days 3–7 • TC track forecasts • Consistency run-to-run (less chaotic noise) --- we’re seeing this already with AI-GFS / EC for the Jan 18 system Weaknesses 1) QPF magnitude and placement (especially winter storms, and Jan 18 threat might demonstrate this) • AI models often: - Smooth precipitation fields - Struggle with sharp gradients (e.g., coastal fronts, deformation bands) - Miss mesoscale enhancement tied to frontogenesis or CSI • They can produce: - Broad, confident-looking swaths that lack realistic structure --- we’re already seeing this with AI-GFS for Jan 18 - QPF maxima that are too spatially uniform • In winter storms, this shows up as: - Overdone areal coverage of moderate QPF --- again, see Jan 18 where this may be happening - Underrepresentation of narrow heavy bands - Difficulty separating rain/snow/ice impacts 2) Mesoscale & convective-scale processes: • AI models excel at large-scale pattern evolution, but:they do not explicitly resolve: - Convection initiation - Storm-scale feedbacks - Boundary interactions (outflows, lake effects, terrain-driven circulations) • As a result: - Severe weather ingredients can be misrepresented - Convective QPF tends to be overly smeared - Lake-effect and terrain-enhanced precipitation is inconsistent So, AI global models can look great at Day 5 synoptics but shaky inside Day 2–3 details. 3) Sensitivity to initial-condition uncertainty: AI models less good at knife-edge marginal setups where small differences have big impacts. 4) What’s behind the curtain is nebulous and not physically intuitive: When a physics model is wrong, you can often diagnose why (e.g. poor phasing, weak cold air damming, overmixed boundary layer) But with AI models: - Errors and biases can be systematic but reasons opaque - Forecasters will not know what to mentally discount This makes them harder to weigh or adjust the way we do with known biases (e.g. NAM is overamped, GFS southeast bias) 5) AI models are not optimized for rare events. They may overly weigh climatology and as a result miss extreme and rare events. What might we expect in the next few years? • Hybrid systems (physics + AI) becoming the norm • Continuous learning: Unlike traditional NWP (which evolves slowly via code and physics algorithm upgrades), AI models can be retrained frequently as new data become available, and incorporate recent busts or rare events more rapidly. Machine learning can incorporate more physics constraints and verification feedback. This theoretically allows self-correction and reduction in biases over time.

EC a definite improvement though still a graze

0z EC AIFS ticked up, warnings to south shore Excellent trends at 0z so far

Just catching up... Holy AI GFS... 6z-12z-18z-0z today have been steady ticks to increase QPF in SNE... at H5 looks like it progressively downplays the lagging vorticity that shears out off the southeast coast. We get better tilt of the trough by 0z Monday. Will be a memorable GFS vs. AIGFS showdown. Can we rely on the age-old RGEM+AIGFS combo?

Increasingly interesting how these perform (at least as a distraction from a middling winter) 12z-18-0-6 GFS AI steadily ticked up each cycle 12z GFS AI stabilized / ticked back but would still be warning southeast SNE 12z EC AI has been fairly steady, advisory most of southeast SNE

You're right. I was proposing more the convection -> latent heat release -> amplified PV, so a feedback amplification of the remnant vorticity. But you'd expect that more downstream rather than upstream of the convection popping up. Anyway this is academic, not the biggest impediment to this system.

Probably both true... it is indeed remnant of a trailing S/W (circled in green below), but it also pulses off the southeast coast (green circle in 2nd frame) and I wonder if it's convection-related feedback... in any case, that might be interfering with more proficient cyclogenesis Variation of theme of too many chefs and lacking a discrete enough shortwave I don't think the 0z Euro or 6z models close this potential. AIs support advisory and probably what I'd favor at this point, but everything from a graze to warning in parts of SNE still realistic possibilities 84h out.

Euro AI stable 12z-18z-0z, just outside BM, widespread advisory

Lagging vorticity I was referring to was here... this could be feedback from the line of convection sweeping off southeast coast, I'm not sure as 6z GFS hinted at this too, but it disrupts much better cyclogenesis. Otherwise, we'd have a huge hit.

CMC 0z ~100-150 mile shift northwest over 12z... So far 0z major league guidance is 2/2 with a substantial hit

Yeah there's some lagging vorticity at 0z (? feedback with the line of convection sweeping off southeast coast), so cyclogenesis not as clean as 12z This verbatim solution less important than adding support for a subtantial SNE hit, 12z was not just a fluke GFS run

Definite improvement over 18z, thru 90h more similar to 12z with shortwave digging over Georgia

I'm in awe of the absolute clinic (interference and otherwise) by which SNE is missing significant snowfall, especially now in a BN cold stretch... For 15-16th and 18-19th in particular, strength of the vort on guidance a few days ago seemed like it would overcome any inteference issues to at least graze (15-16th) or cut (18-19th) A corollary: there are many more routes to having no event than there are for an event to materialize. I did this last year and was curious how it extrapolates this year... I charted 4-year rolling average snowfall at KBOS. Obviously 2025-2026 is not over, so for February and March I just used historic means (* indicates extrapolation from historic means for those 2 months). Look how off the charts this 4 year stretch has been... well below 2 standard deviations:

Just catching up, good analysis all, great blog post Ray To add to all the flies in the ointment, I think the ULL over St. Lawrence River is hindering trough from tilting more negative, scoots this east just as it looks ready to explode. Diminish that and we’d have a huge hit on this 0z GFS.

Agree, best look in weeks... Lots of chefs in kitchen in a more favorable longwave pattern = potential, even if there is interference. At moment (and will obviously evolve this far out), EPS and EPS-AI hinting at a SNE grazer 16th followed by at least CNE/NNE hit 18th, but tons of scatter around those focal points with huge range in ceiling as we're seeing seeing on OP runs. I too am hoping to at least get the 6" monkey off our back, nearing 4 years now... and in the shorter term, let's get <32F by kickoff Sunday.

Great snow growth and grass whitening now in Boston Fenway area... And forecasts definitely busted a few degrees too warm this morning with continuing northerly drain in place As much as we continue to be starved for a meaty storm, at least this winter has given decent vibes with persistent cold and some white

https://xmacis.rcc-acis.org/ Specifically for KBOS (noting the unrepresentative wastewater facility caveats): 1936-37 9.0" 2011-12 9.3" 2023-24 9.8" 2025-26 5.0" to date Mean 42.2"

Landed in Logan at 5:45am this morning… one of the most turbulent landings I’ve ever experienced. Looks like KBOS 59 mph gust. Multiple people throwing up before even we got to gate. Took me hours to settle my nausea. Missed the wounded duck storm… congrats to the region especially Connecticut and south shore areas.

18z AI-GFS (I have no idea if this is worth anything) is also ticking better than 12z

If there's any takeaway from 18z suite, it's that this is not continuing to slip away into nothing (which was a distinct possibility on yesterday's guidance). It will be meager, but there will be accumulation.

On this 18z NAM: some decent fronto scrapes the south coast... could see upwards of 3-5" right at the water if that occurs. Let's hold or continue these ticks in the next 12 hours...

Yeah it's been a tough stretch for the region and this hobby. I posted last night... I remember the forum days of all-hands-on-deck war-room analysis for an impending KU, and now I'm micro-dissecting 1-2" vs. 2-3". We're famished for a region-wide larger event. Dec 20 2024 was the last positive bust for MBY (Boston suburbs) in years, and it was ~5.5". But easy to forget (maybe because the unseasonal cold) that it's very early and there's lots of winter to go.

Box update As noted above, upside potential at this point IF the vort comes in stronger, could add 1-2" south and east zones, with maybe spots 4-5" total in outer Cape...