These models actually outperform traditional methods on many fronts, including accuracy a lot of the time. They are technically generative AI models, but they're definitely not LLMs.
Really exciting to see NOAA finally make some progress on this front, but the AIGFS suite likely won't outperform ECMWF's AIFS suite any time soon. The underlying architecture between AIFS and GraphCast/AIGFS is pretty similar (both GNNs), so there won't likely be a model-level improvement. And most of ECMWF's edge lies in its superior 4DVar data assimilation process. AIGFS is still being initialized on NOAA's hybrid 4DEnVar assimilation process as far as I understand it, which is still not as good as straight up 4DVar unfortunately.
There are multiple efforts and a good number of VC working on AI DA system. DA is fundamentally a hand-crafted optimization process just like NN. I once reimplemented an EnKF in pytorch and it works amazingly fast. But our observations are so dirty and sparse. ECMWF tuned their system so well. NOAA definitely has potential being even better, but no hope any soon future IMHO.
Things are changing quickly in this area. Several projects/companies working on AI data assimilation (an alternative route to creating analysis data like ERA5)[0].
Also a lot of companies working on the data collection side, replacing/augmenting government data collection. Spire's an example of this in the space domain, and Windborne and Sorcerer (my company) do weather balloons.
Very interesting, I'd love to hear more about it! In short, yes, we plan to do descents near urban areas if there's a route where we can go down to a safe height and stay away from any airspace. What cities are you looking at right now?
ARM is a DOE program that ships top-tier instrumentation to various sites around the world. Loads of university researchers will follow, and you end up with a massive open source data pool. Houston in particular was focused on Aerosol effects on precipitation in the Coastal-Urban environment. There were loads of balloon launches from sites all over the city during the campaign, from large ozonesondes to the tiny sparv embedded foam cup ones (https://sparvembedded.com/products/windsond)
I'm in NY, and my university NOAA department has a focus on PBL Ozone measurements lately. My work in particular is focused on low cost UAV profiling up to about 150m, with a pipe dream of doing 0-3km.
I'm just a grad student, but if anything there sounds interesting feel free to email and I can try and get you in touch with more knowledgeable people.
One of the first ideas we explored was putting an electrolyzer on the balloon to replenish hydrogen over time. Unfortunately right now, for balloons our size and power budget it's just not feasible. And actually, we can get a pretty low leakage rate with our materials which lets us stay aloft for a really long time, but eventually the UV degradation becomes to extreme.
We're actually just going to have kids launch our balloons, it's no problem:
A person who is 6 years of age or younger who
intentionally releases, organizes the release of, or
intentionally causes to be released balloons as
prohibited by s.
379.233 does not violate subsection (4) and is not.
subject to
the penalties specified in subparagraph 1.
I don’t know much about WAPS at all, but generally speaking, we’re only interested in doing weather sensing. As of now, we’re not planning to license the vehicle at all. And generally speaking our platform doesn’t work for payloads that are heavier than ~0.25lb and have more than minimal power requirements.
