counters

11 ms·

61.
by counters 10mo ago
Machine learning _has_ been widely used in weather forecasting, but in a different way than these models. Going back to the 1970's, you never just take the output of a numerical weather model and call it a forecast. We know that limita
62.
by counters 10mo ago
... if you have access to the Intel compiler tools. gfortran does not support many of the new language features beyond Fortran 2008, even though Fortran 2018 is the defacto "current" standard.
63.
by counters 10mo ago
I wouldn't compare someone like Richard Feynman to "Big Balls."
64.
by counters 11mo ago
> Meteorology It seems like that at first glance. But in reality, GPUs have had extremely slow adoption for real-world operational meteorology applications. Because of the fundamental design and architecture of most NWP systems, it was v
65.
by counters 11mo ago
Yeah, it's underplayed in the the writeup here but the context here is important. The "sharpness" issue was a major impediment to improving the skill and utility of these models. When GDM published GenCast two years ago, ther
66.
by counters 11mo ago
Thanks for the fantastic write-up! This is a great breakdown of to lean into the strength of each of these tools.
67.
by counters 11mo ago
These models don't predict CRPS - it's just the loss that they're optimized against. They still produce a normal weather forecast with gridded variables.
68.
by counters 11mo ago
To add to the existing answers - L2 losses induce a "blurring" effect when you autoregressively roll out these models. That means you not only lose import spatial features, you also truncate the extrema of the predictions - in oth
69.
by counters 11mo ago
I mean - you don't need any ML for that. Just go grab random samples from a ~30 day window centered on your day of interest over the region of interest from a reanalysis product like ERA5. If the duration of ERA5 isn't sufficient
70.
by counters 11mo ago
Sure - you'd simply use a regional, high-resolution model. In some parts of the world, these exist for free (e.g. NOAA runs the HRRR [and soon the RRFS] over CONUS, which is re-run every hour and outputs data on a ~3km grid at up to 15
71.
by counters 11mo ago
The community has mostly abandoned SPO data. It's extraordinarily difficult to use this data because of social issues like PII and technical ones like QA/QC. But even more importantly, there's very little compelling evidence
72.
by counters 11mo ago
You're 100% correct, but there's a subtlety in what the commenter is talking about. Yes, _in aggregate_, forecasts are objectively, quantifiably better in 2025 than they were in 2005 let alone 1985. But any given, specific forecas
73.
by counters 11mo ago
> By having better dispersed ensemble forecasts, we can more quickly address observation gaps that may be needed to better solidify certain patterns or outcomes, which will lead to more accurate deterministic forecasts. Sorry - not sure
74.
by counters 11mo ago
That's not really what a weather model "does."
75.
by counters 11mo ago
Eh, it's a pretty big distinction weather-wise. Extreme Western New York and the Tug Hill plateau are all susceptible to somewhat frequent lake effect snow. Given the right time of year and wind fetch, you can see narrow convective &#x
76.
by counters 11mo ago
If it's inoperative then it shouldn't be in the language of the grant. Full stop. The language itself also overly broad. The stipulation from the grant didn't just cover activities funded by the grant itself. In the very lang
77.
by counters 1y ago
> In fact, I was under the opposite impression because of the benefit that sulfur enriched shipping exhaust had for our climate. It isn't quite accurate to state that ship tracks have/had a "benefit" on our climate. T
78.
by counters 1y ago
It depends where you are in the world. The general solution involve an app that provides two types of forecast products: (1) a short-range, high-resolution numerical weather forecast over your country / domain of interest, refreshed ra
79.
by counters 1y ago
Yes, several companies offer sophisticated aviation forecast products, which employ both modeling and observational data to capture 3D areas of turbulence, icing, etc. Similar systems are also run by NOAA and other agencies and their foreca
80.
by counters 1y ago
> Speculation: goog used the barometric sensors in many phones "near you" to increase the precision of their models, making "immediate timeframes" extremely precise. They didn't. Smartphone pressure observations
81.
by counters 1y ago
Entirely different forcing mechanism(s). The two most promising vectors are stratospheric injection and marine stratocumulus injection. Both approaches induce very different radiative and attendant circulation responses, and aren't rel
82.
by counters 1y ago
Your experience doesn't jive with what myself and many others have observed at NOAA over multiple decades.
83.
by counters 1y ago
> Governments rarely give genuine explanations for their actions and rarely need to. This is an absurdly cynical take. It certainly doesn't jive with how NOAA has historically operated - which has necessitated as much transparency a
84.
by counters 1y ago
Sure. Great. But that explanation isn't being offered by the powers-that-be. So there's no point trying to rationalize it post-hoc. There's no evidence that this is anything more than yet another round of ideologically-fueled
85.
by counters 1y ago
It's not a question of "the market." Weather stories very strongly breakthrough in our current media environment. More importantly, weather forecasting and government services related to them enjoy deep and durable bi-partisa
86.
by counters 1y ago
Well put.
87.
by counters 1y ago
> There won’t be another Andrew because the building codes were changed so that all new construction must withstand category 5 storm force I sincerely hope you're right, but there is plenty of evidence suggesting that this will not
88.
by counters 1y ago
It just isn't helpful to assume malice. Even for the most ardent, ideological Heartland Institute or Heritage Foundation conservative, there is still a path forward in discussing unintended consequences. Just look at the post-Liberatio
89.
by counters 1y ago
You're right, but I would stress that this is an over-simplification of the entangled financial interests that _might_ be at play - and there simply isn't any evidence that has been presented pointing in that direction.
90.
by counters 1y ago
We don't if, or to what extent, DOGE was involved or influential in the decision-making here. Yes, the DMSP program was aging and slated to wind down as replacements - both federal and commercial - came online in the second half of the

More ›