LordBaltimore

Sometimes I wonder if I'd be happier or more productive if it snowed less. I'm sitting here looking at 18z temperatures to see how long we have until it melts. It feels like when you have an older pet and you know they only have a little while left

I hope for my sake and the sake of a lot of other weather peeps, that the people in charge don't forget that these things are still dependent on physical modeling systems and will be for a very very long time. It's not "AI magically takes raw data and transforms it into forecasts", it's more like the numerical modeling system does the data assimilation - using an ensemble - and then the analysis is used as input for a ML model trained off of ERA5 (also produced with a numerical modeling system). The quality of AI forecasts directly scales based on the quality of the numerical model.

need to see summit camp once before I die

I don't care how unpopular or stupid taking over Greenland would be. I want cheap visa free flights to an ice cap and I don't care how we go about getting them

I do wonder if the physics tuning is part of the problem. Microphysics is more of an art than a science even now, and things get tuned based on what they're used for. Cold microphysics that does good for stratiform precip in MCS systems might just whiff in these cases.

Has anyone here looked at verification scores for the GFS vs 6 hour lagged Euro. Clearly the Euro is a better model but hopefully the 12z GFS has the advantage of additional obs. Please, it's critical for my happiness over the next 90ish or so minutes and I'm too sleep deprived to do it myself. We need this one

HRRR is better than probably any other operational model at nailing bow echo/MCS structure and propagation after these things have already initiated. It's pretty much what it was designed for and the way it assimilates radar reflectivity is very different from models like the NAM or globals for that matter. It has a wheelhouse

stickage has begun on mulch surfaces

Steady light snow for past 5 to 10 minutes in southern Hoco. Gonna be an all nighter for sure but telling myself it's to catch up on work

but will the snow be enough to fully cover said "short" pump is the question

Time for a real storm. None of this mid stuff

NAM nest is still the only non-suck (*cough* hi-res fv3) convective allowing model in range right now. Even if it's error prone at this range it's the only model that can pick up low end mesoscale features that could broaden the precip shield

Wouldn't put much stock into QPF from any high-res fv3 models. IIRC, it has serious problems with that on the convective scales

just got back from a jebwalk in kings contrivance. Was legit ripping snow for a few minutes. Been awhile since I've had a brain freeze from snow + wind

was trying to reactivate my account from 2011 when I was like 16 but couldn't get the password reset to work. Was just a novice weenie back then, now I code fortran for GEOS-5 so I can weenie professionally

agreed. Can't see the 12z yet, but previous euro runs want to relax the cold +tnh pattern past the 12th. We need deep snowcover now to avoid ugly above freezing days before the reload. 5" won't do it

God bless America. Euro must be out to lunch

model blends and multi-physics ensembles are the worst thing thought up by the NWP community in the last 20 years. It's the worst possible way of addressing model error

can't recall the exact event, but I remember a couple of inches of sleet in downtown baltimore around 2017ish. The texture of it was crazy

People misuse the ICON all the time. If you look at the data assimilation under the hood it's more like a global version of the RAP with 3denVAR and BC's used for regional models and such