8 ms·
Project Aardvark: reimagining AI weather prediction
- kgwgk 2y agoPaper available here: https://www.nature.com/articles/s41586-025-08897-0 https://www.nature.com/articles/s41586-025-08897-0
- abdullahkhalids 2y agoArxiv version here https://arxiv.org/pdf/2404.00411 https://arxiv.org/pdf/2404.00411
- foofoo55 2y agoWhy are the Arxiv and Nature versions so different, even the text?
- abdullahkhalids 2y agoIts quite common to revise papers. For example, they might have uploaded to arxiv in order to submit to a conference. Later, they revised and submitted to Nature.
- kkylin 2y agoAs another comment mentioned, papers get revised during review, usually in response to reviewer comments. Also, some journals (not sure about Nature) do not allow authors to "backport" revisions made in response to reviewer comments to preprints; I guess they view the review process as part of their "value add".
- jamala1 2y agoArxiv is mostly meant for preprints for peer review. In a Nature paper in particular, the final layout is typically done by the journal's professional production team, not the authors. Not all publishers grant permission for authors to upload the peer-reviewed and layouted postprints elsewhere.
- benob 2y agoAccording to arxiv paper, code will be made available here: https://github.com/annavaughan/aardvark-weather-public https://github.com/annavaughan/aardvark-weather-public (from 8 months ago)
- sunshinesnacks 2y agoThe Nature preprint references this zenodo archive with data and code in a big 13GB .tar file: https://doi.org/10.5281/zenodo.13158382 https://doi.org/10.5281/zenodo.13158382 I haven’t downloaded it to see what’s in it.
- dboreham 2y agoAlthough based on their bios and UK population statistics, it seems likely they're both English, the article omits this information. They do work at a British institution, located in England..
- _joel 2y agoThe name Turing and turing.ac.uk should maybe give it away. I don't think the authors' nationalities are of importance, however.
- abdullahkhalids 2y agoHow robust will ML models be in the world of rapid climate change, where the past no longer predicts the future?
- zdragnar 2y agoClimate is not weather. Weather predictions are for specific events and areas, made on the order of days- typically no more than 2-3 weeks into the future. Climate models predict future averages over large regions on the order of decades. "rapid climate change" is on the order of "within this century". Whether the climate changes or not doesn't really impact the weather models at all, because the climate's not changing on the same time scale.
- labster 2y agoWhile I really want to agree with your sound argument, I have this suspicion that climate change significantly affected forecast skill on Hurricane Otis.
- melagonster 2y agoClimate change: more typhoons in the future. Weather prediction: which path typhoons will follow?
- zdragnar 2y agoThe climate didn't change in the time from Otis forming to landing. Weather models look at current conditions and extrapolate what the next set of conditions will be. Another way to look at it is that climate change models reflect what we will experience. Weather models reflect the mechanics of humidity, temperature and air flow- given a current state, what is the next likely state. Climate change doesn't change how those mechanics interact. It only guesses what "current state" will look like in the far future.
- avianlyric 2y agoClimate change doesn’t impact the physics of weather, but our understanding of the inputs to a weather system is still very incomplete. Climate change does impact the relative importance of different physical inputs into a weather system, and inherently pushes weather systems away from the most studied and well understood aspects of weather physics. That in turn degrades the accuracy of our models, as it takes time to research and understand whats changed in a weather system, and what previously unimportant inputs have become much more important.
- lenerdenator 2y agoHmmmmmmm. I have a challenge for the model: Accurate (within 3deg F) weather predictions for the Kansas City metro more than two days out. As of 2024 these were rarely accurate.[0] [0]https://www.washingtonpost.com/climate-environment/interactive/2024/how-accurate-is-the-weather-forecast/ https://www.washingtonpost.com/climate-environment/interacti...
- lytedev 2y agoIndeed we have it tough out here. I like the excitement, though, I must say! Funny to find the sentiment here on HN, though!
- metaphor 2y ago> Accurate (within 3deg F) weather predictions Sniff test: ASTM E230 standard tolerances for the venerable Type K thermocouple is +/- 2.2°C or +/- 0.75%, whichever is greater. Expectations are in need of recalibration if anyone thinks a single number is going to meaningfully achieve that level of accuracy across a volume representing any metro area at any given point in time, let alone two days out.
- IshKebab 2y agoThe bigger the volume the more accurate temperature you should be able to get. Do you think 2C global warming is irrelevant because a single thermocouple couldn't accurately measure it?
- genewitch 2y agoI think you misunderstood the assignment. The temp predictions in my area wildly swing around, until the day of, and even then I have to adjust for my actual location. And I'm not in Kansas City. The only place that has ~3° accuracy is San Diego. And maybe Antarctica.
- kwertzzz 2y agoI am not sure if type K thermocouple are used for meteorological air temperature measurements. These sensors (based on thermal resistance) for example have an accuracy of 0.2 °C under typical conditions [1]. [1] https://www.ti.com/lit/ds/symlink/tmp1826.pdf?ts=1742799763814&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FTMP1826 https://www.ti.com/lit/ds/symlink/tmp1826.pdf?ts=17427997638...
- MostlyStable 2y agoI'm curious if some future, hypothetical AGI agent, which had been trained to have these kinds of abilities, would be akin to how most humans see a ball in flight and just instinctively know (within reason) where the ball is going to go? We don't understand, consciously and in the moment, how our brain is doing these calculations (although obviously we can get to similar results with other methods), but we still trust the outputs. Would some hypothetical future AI just "know" that tomorrow it's going to be 79 with 7 mph winds, without understanding exactly how that knowledge was arrived at?
- defrost 2y ago> would be akin to how most humans see a ball in flight and just instinctively know (within reason) where the ball is going to go? Up to a point .. and that point is more or less the same as the point where humans can no longer catch a spinning tennis raquet. We understand the gravitional rainbow arc of the centre of mass, we fail at predicting the low order chaotic spin of tennis raquet mass distributions. Other butterflies are more unpredictable, and the ones that land on a camels back breaking a dam of precariously balanced rocks are a particular problem. * https://en.wikipedia.org/wiki/Tennis_racket_theorem https://en.wikipedia.org/wiki/Tennis_racket_theorem * Dzhanibekov effect demonstration in microgravity: https://www.youtube.com/watch?v=1x5UiwEEvpQ https://www.youtube.com/watch?v=1x5UiwEEvpQ * https://en.wikipedia.org/wiki/Horseshoe_map https://en.wikipedia.org/wiki/Horseshoe_map
- MostlyStable 2y agoYes, humans are obviously limited in the things we can instinctively, intuitively predict. That's not really the point. The point is whether something that has been trained to do more complicated predictions will have the a similar feeling when doing those predictions (of being intuitive and natural), or if it will feel more explicit, like when a human is doing the calculus necessary to predict where the same ball is going to go.
- defrost 2y agoHumans have both intuition and explicit calculation, predictive calculation can be stable or inherently unstable. The point is whether a LLM has any feelings ...
- IncreasePosts 2y agoWhy would we need weather models on our desktop, instead of just consuming the output of a far more powerful model? Unless people start interacting with weather forecast differently in the future it doesn't seem like something that needs personalization
- rafram 2y agoRead the article. The point is that it can run on sub-supercomputer hardware, not specifically that it can run on consumer desktops.
- avianlyric 2y agoThe title has been editorialised. The paper is about building AI weather models on par with current state-of-the-art weather models that require super computers to run. But without the need for the super computers, just a normal desktop computer is enough. That represents a huge step-change in need compute for accurate weather models, and opens up the possibility of even more accurate models, if the accuracy of these techniques scales with available compute. If you can get state-of-the-art accuracy with a desktop computer, something that normally requires a super-computer, what happens if you run those techniques on a super-computer?
- benob 2y agoDidn't read the paper, but it would be cool to apply distillation to these big physics-driven models, in order to simulate their outcome on much smaller models.
- brigandish 2y agoAdditionally (to the other replies to your question), asking for the weather is just another data point in the huge amount of surveillance occurring. If I can get the weather for my location without accessing a remote server where someone else gets to see my query and probably location, all the better. This provides that possibility.
- rkagerer 2y agoAre all the inputs (from buoys, weather balloons, stations, etc) from decades of history stored, as well as past daily forecasts of existing weather models, so that this AI algorithm (and any future new ones) can be run across historic data and compared in terms of performance to existing methods? Is there a big Clearinghouse for this data? Kind of like how fintech algos can be run against historic stock market data to evaluate them.
- lgeorget 2y agoThe World Meteorological Organization has data that national weather services exchange to make forecast. Apart from that, its on a country-by-country basis. In France for instance Météo-France has released all of its historical data in January 2024: https://www.data.gouv.fr/fr/organizations/meteo-france/#/datasets https://www.data.gouv.fr/fr/organizations/meteo-france/#/dat...
- graemep 2y agoBut national models go beyond there borders and usually have some modelling outside so the exchanges will cover quite a lot of the world. For example, the UK's metoffice has a low resolution medium term global model: https://www.metoffice.gov.uk/research/approach/modelling-systems/unified-model/weather-forecasting https://www.metoffice.gov.uk/research/approach/modelling-sys...
- cship2 2y agoHow does windy do it? Did they just scrap all the sites or api. Would be great to have global dump every 1 minute of all the countries. Or have it have in a radio broadcast.
- jcd000 2y agoWindy uses openmeteo, which in turn aggregates many weather providers under one roof/api
- mschuster91 2y ago> Is there a big Clearinghouse for this data? Many commercial and even third party governments rely on the data from NOAA and its archives, on top of that the EU runs its own EUMETSAT fleet of data, plus a ton of national services - unfortunately, the result is there's a looooot of datasets. NOAA's dataset is public domain [1], EUMETSAT only requires attribution for most of its data [2]. On top of that you got the EU's Climate Data store [3], ECMWF [4], and ECA&D [5]. The service that many private weather services provide is to aggregate and weigh all of the publicly available datasets, and some also add in data from their own ground stations, commercially licensed "realtime" data from governmental services, and their own models as well. The interesting question is what DOGE will do regarding NOAA - it is increasingly possible that NOAA will shut down, either to be turned into a pay-for-play model, replaced by private services, or just carelessly dropped in its entirety. [1] https://www.ncei.noaa.gov/archive https://www.ncei.noaa.gov/archive [2] https://user.eumetsat.int/resources/user-guides/data-registration-and-licensing https://user.eumetsat.int/resources/user-guides/data-registr... [3] https://cds.climate.copernicus.eu/ https://cds.climate.copernicus.eu/ [4] https://www.ecmwf.int/en/forecasts/datasets https://www.ecmwf.int/en/forecasts/datasets [5] https://www.ecad.eu/dailydata/ https://www.ecad.eu/dailydata/
- westoque 2y agonot to hijack this thread but my dad did extensive research in sea breeze and rainfall modeling and he would have loved to see these AI and machine learning advancements in weather prediction. [0]: https://www.revistascca.unam.mx/atm/index.php/atm/article/download/8375/7845/8245 https://www.revistascca.unam.mx/atm/index.php/atm/article/do... [1]: https://wmo.int/about-wmo/awards/international-meteorological-organization-imo-prize https://wmo.int/about-wmo/awards/international-meteorologica...
- thenthenthen 2y agoAnother challenge I would add to the list is that weather data is strategic data.
- politelemon 2y ago> replaces all of the steps Unable to tell if this is an exaggeration or if I'm just missing nuance, how would the model replace the data gathering which is listed as step 1.
- sunshinesnacks 2y agoI think it’s replacing data assimilation. Ingesting observation data from lots of sources. Not replacing observations themselves.
- jedberg 2y agoIt's a shame they just cut funding for launching the weather balloons and other equipment needed to collect this data. https://apnews.com/article/weather-forecasts-worsen-doge-trump-cuts-tornado-da573a044916c06cebcdb92b1f1452e6 https://apnews.com/article/weather-forecasts-worsen-doge-tru...
- HumblyTossed 2y agoLet's name "they". It was Trump via DOGE.
- echelon 2y agoSuper smart to name it "Aardvark". It'll sort first in any alphabetized list. Before GFS, etc. By the same measure, it's probably smart to name your company or app A-something rather than Z-something.
- desas 2y agoPeople were doing this in the days of printed phone books. There would be an "AAA Plumbers", "ABC Taxis" and so on. Reputedly, "Apple Computers" coming ahead of "Atari" (Jobs' ex-employer) in the phonebook was one of the reasons for Apple being Apple.
- Onavo 2y agoHow does it compare to Google's https://www.nature.com/articles/s41586-024-08252-9 https://www.nature.com/articles/s41586-024-08252-9
- scellus 2y agoGoogle's is initialized with a gridded dataset, ERA5, from ECMWF. Using ERA5 is the current standard here, and ECMWF themselves build on that mostly now. Meanwhile, Aardvark tries to do the same directly from observations.
- Onavo 2y agoWhat do you mean directly from observations?
- counters 2y agoGDM's GraphCast/GenCast require an "analysis state" of the atmosphere. This is a 3D, gridded dataset with key variables like temperature, humidity, and winds. Generally speaking, an analysis is produced by a physics-based weather model through "data assimilation", an optimization process which tries to create a 3D state that is consistent with observations over some window of time. "Observations" is overloaded; it's really any "raw" weather data product, like satellite imagery, surface station measurements, or weather balloon traces. I'm handwaving away a lot of complexity here. The AardvarkWeather model is a significant new development and paradigm shift - it's one of the first models of a new class which do not require an analysis, and can directly use the "raw" weather data observations that are typically used to perform data assimilation.
- Onavo 2y agoThanks
- bazzargh 2y agoWhen I saw this I thought... "The Turing Institute? Does that still exist?" https://en.wikipedia.org/wiki/Turing_Institute https://en.wikipedia.org/wiki/Turing_Institute There was a previous Turing Institute in Glasgow doing AI research (meaning, back then rules-based systems, but IIRC my professor was doing some work with them on neural networks), which hit the end of the road in 1994. There was some interesting stuff spun out of there, but it's a whole different institute.
- twak 2y agoThe Turing had an interesting approach to naming, not only stealing the Glasgow group's name, but also choosing the initials 'ATI' (in 2015...). It's recently struggling for relevance. https://www.ft.com/content/6bfea441-e16c-499a-a887-69f735c29389 https://www.ft.com/content/6bfea441-e16c-499a-a887-69f735c29... (https://archive.ph/ujfhb https://archive.ph/ujfhb) I hope they turn it around because the UK need for AI academic coordination/leadership is so high.
- interludead 2y agoCurious to see how it handles edge cases though, like hurricanes or rare extreme weather events that aren't well-represented in historical data
- dsr_ 2y agoI note that they don't actually say "matches the accuracy of current best models". Or even "almost as accurate as the current best models".
- herodoturtle 2y agoFor a second I thought this was related to that old school FogCreek internship programme with the same name. Fun documentary at the time. Something about 12 weeks with geeks, and jumping out of windows comes to mind.
- noiv 2y agoModel data here: https://zenodo.org/records/13158382?token=eyJhbGciOiJIUzUxMiJ9.eyJpZCI6Ijk5ZWU4ZGJkLTY5NDEtNDdiNS05M2NlLTZmOGIzMzRjZDgzYSIsImRhdGEiOnt9LCJyYW5kb20iOiI0N2RkZGVlMDFiOGNjZDZhODI1MThiMDQyM2FjOTdlMiJ9.XJRSx0nAgpHBeiU6MeSuT9wjYQo9OmKv3h1gpWgf7bWguaoMbBTPJmLIPUEL8aMcLFZeLLVuspQNkERyTcqKsg https://zenodo.org/records/13158382?token=eyJhbGciOiJIUzUxMi... ~14GB, needs login, works with some federated accounts.
- timthorn 2y agoRichard Turner (one of the authors) gave a talk on how AI weather prediction works to the Cambridge Philosophical Society around 18 months ago. A recording is available here: https://www.eng.cam.ac.uk/news/quiet-ai-revolution-weather-forecasting https://www.eng.cam.ac.uk/news/quiet-ai-revolution-weather-f...
- eternityforest 2y agoI wonder if something like this could allow for local forecasts with just regular consumer sensors, or maybe with stuff like cameras or mic arrays.
- uncleyoung 2y ago[dead]
- Aeroi 2y agohows this differ from graphcast