from Hacker News

GraphCast: AI model for weather forecasting

by bretthoerner on 11/14/23, 3:42 PM with 290 comments

  • by meteo-jeff on 11/14/23, 5:28 PM

    In case someone is looking for historical weather data for ML training and prediction, I created an open-source weather API which continuously archives weather data.

    Using past and forecast data from multiple numerical weather models can be combined using ML to achieve better forecast skill than any individual model. Because each model is physically bound, the resulting ML model should be stable.

    See: https://open-meteo.com

  • by robertlagrant on 11/14/23, 4:52 PM

    This is fascinating:

    > For inputs, GraphCast requires just two sets of data: the state of the weather 6 hours ago, and the current state of the weather. The model then predicts the weather 6 hours in the future. This process can then be rolled forward in 6-hour increments to provide state-of-the-art forecasts up to 10 days in advance.

  • by xnx on 11/14/23, 4:04 PM

    I continue to be a little confused by the distinction between Google, Google Research and DeepMind. Google Research, had made this announcement about 24-hour forecasting just 2 weeks ago: https://blog.research.google/2023/11/metnet-3-state-of-art-n... (which is also mentioned in the GraphCast announcement from today)
  • by EricLeer on 11/15/23, 10:39 AM

    I am in the power forecasting domain, where weather forecasts are one of the most important inputs. What I find surprising is that with all the papers and publications from google in the past years, there seems to be no way to get access to these forecasts! We've now evaluated numerous of the ai weather forecasting startups that are popping up everywhere and so far for all of them their claims fall flat on their face when you actually start comparing their quality in a production setting next to the HRES model from ECMWF.
  • by serjester on 11/14/23, 5:24 PM

    To call this impressive is an understatement. Using a single GPU, outperforms models that run on the world's largest super computers. Completely open sourced - not just model weights. And fairly simple training / input data.

    > ... with the current version being the largest we can practically fit under current engineering constraints, but which have potential to scale much further in the future with greater compute resources and higher resolution data.

    I can't wait to see how far other people take this.

  • by dauertewigkeit on 11/15/23, 6:34 PM

    The multimesh is interesting. Still, I bet the Fourier Neural Operator approach will prove superior. Members of the same team (Sanchez-Gonzales, Battaglia) have already published multiple variations of this model, applied to other physical scenarios and lots of them proved to be dead ends. My money is on the FNO approach, anyway, which for some reason is only given a brief reference. To their credit DeepMind usually publishes extensive comparisons with previously published models. This time such a comparison is conspicuously missing.

    Full disclosure: I think DeepMind often publish these bombastic headlines about their models which often don't live up to their hype, or at least that was my personal experience. They have a good PR team, anyway.

  • by stabbles on 11/14/23, 8:28 PM

    If you live in a country where local, short-term rain / shower forecast is essential (like [1] [2]), it's funny to see how incredibly bad radar forecast is.

    There are really convenient apps that show an animated map with radar data of rain, historical data + prediction (typically).

    The prediction is always completely bonkers.

    You can eyeball it better.

    No wonder "AI" can improve that. Even linear extrapolation is better.

    Yes, local rain prediction is a different thing from global forecasting.

    [1] https://www.buienradar.nl [2] https://www.meteoschweiz.admin.ch/service-und-publikationen/...

  • by lispisok on 11/14/23, 4:54 PM

    I've been following these global ML weather models. The fact they make good forecasts at all was very impressive. What is blowing my mind is how fast they run. It takes hours on giant super computers for numerical weather prediction models to forecast the entire globe. These ML models are taking minutes or seconds. This is potentially huge for operational forecasting.

    Weather forecasting has been moving focus towards ensembles to account for uncertainty in forecasts. I see a future of large ensembles of ML models being ran hourly incorporating the latest measurements

  • by freedomben on 11/14/23, 5:17 PM

    weather prediction seems to me like a terrific use of machine learning aka statistics. The challenge I suppose is in the data. To get perfect predictions you'd need to have a mapping of what conditions were like 6 hours, 12 hours, etc before, and what the various outcomes were, which butterflies flapped their wings and where (this last one is a joke about how hard this data would be). Hard but not impossible. Maybe impossible. I know very little about weather data though. Is there already such a format?
  • by Vagantem on 11/14/23, 10:00 PM

    Related to this, I built a service that shows what day it has rained the least on in the last 10 years - for any location and month! Perfect to find your perfect wedding date. Feel free to check out :)

    https://dropory.com

  • by matsemann on 11/15/23, 3:42 PM

    How's the distribution of the errors? For instance I don't care if it's better on average by 1 Celsius each day for normal weather, if it once every month is off by 10 Celsius when there is a drastic weather event, for instance.

    I'm all for better weather data, it's quite critical up in the mountains, so that's why my question about how reliable it is in life&death situations.

  • by Gys on 11/14/23, 4:55 PM

    I live in an area which regularly has a climate differently then forecasted: often less rain and more sunny. Would be great if I can connect my local weather station (and/or its history) to some model and have more accurate forecasts.
  • by crazygringo on 11/14/23, 11:55 PM

    Making progress on weather forecasting is amazing, and it's been interesting to see the big tech companies get into this space.

    Apple moved from using The Weather Channel to their own forecasting a year ago [1].

    Using AI to produce better weather forecasts is exactly the kind of thing that is right up Google's alley -- I'm very happy to see this, and can't wait for this to get built into our weather apps.

    [1] https://en.wikipedia.org/wiki/Weather_(Apple)

  • by layoric on 11/14/23, 10:43 PM

    I can't see any citation to accuracy comparisons, or maybe I just missed them? Given the amount of data, and complexity of the domain, it would be good to see a much more detailed breakdown of their performance vs other models.

    My experience in this space is that I was first employee at Solcast building a live 'nowcast' system for 4+ years (left ~2021) targeting solar radiation and cloud opacity initially, but expanding into all aspects of weather, focusing on the use of the newer generation of satellites, but also heavily using NWP models like ECMWF. Last I knew,nowcasts were made in minutes on a decent size cluster of systems, and has been shown in various studies and comparisons to produce extremely accurate data (This article claims 'the best' without links which is weird..), be interesting on how many TPUsv4 were used to produce these forecasts and how quickly? Solcast used ML as a part of their systems, but when it comes down to it, there is a lot more operationally to producing accurate and reliable forecasts, eg it would be arrogant to say the least to switch from something like ECMWF to this black box anytime soon.

    Something I said as just before I left Solcast was that their biggest competition would come from Amazon/Google/Microsoft and not other incumbent weather companies. They have some really smart modelers, but its hard to compete with big tech resources. I believe Amazon has been acquiring power usage IoT related companies over the past few years, I can see AI heavily moving into that space as well.. for better or worse.

  • by hackitup7 on 11/15/23, 5:21 PM

    I've been really impressed at how much better weather forecasting has become already. I remember weather forecasts feeling like a total crapshoot as recently as 15 years ago or so.
  • by pyb on 11/14/23, 8:06 PM

    Curious. How can AI/ML perform on a problem that is, as far as I understand, inherently chaotic / unpredictable ? It sounds like a fundamental contradiction to me.
  • by brap on 11/14/23, 8:37 PM

    Beyond the difficulty of running calculations (or even accurately measuring the current state), is there a reason to believe weather is unpredictable?

    I would imagine we probably have a solid mathematical model of how weather behaves, so given enough resources to measure and calculate, could you, in theory, predict the daily weather going 10 years into the future? Or is there something inherently “random” there?

  • by tony_cannistra on 11/15/23, 4:25 PM

    Similar methodologies are being applied to climate modeling, too.

    The Allen Institute has worked on it for a while, and has hired quite a few PhDs (https://allenai.org/climate-modeling).

  • by thriftwy on 11/15/23, 7:05 PM

    Yandex claims to be using AI-based weather forecasting for a good part of a decade and claims it as a success. It is quite good.

    https://meteum.ai/

  • by jauntywundrkind on 11/14/23, 6:00 PM

    From what I can tell from reading & based off https://colab.research.google.com/github/deepmind/graphcast/... , one needs access to ECMWF Era5 or HRES data-sets or something similar to be able to run and use this model.

    Unknown what licensing options ECMWF offers for Era5, but to use this model in any live fashion, I think one is probably going to need a small fortune. Maybe some other dataset can be adapted (likely at great pain)...

  • by rldjbpin on 11/16/23, 9:05 AM

    > GraphCast makes forecasts at the high resolution of 0.25 degrees longitude/latitude (28km x 28km at the equator).

    the resolution, while seemingly impressive, is very imprecise compared to the SOTA in the theoretical modelling side.

    this discredits the computational claims made by the paper for me. i understand that the current simulations can go down to meter scale, but i wonder what the compuational requirements are when you calculate for this resolution.

  • by comment_ran on 11/14/23, 7:25 PM

    So for a daily user, to make it a practical usage, let's say if I have a local measurement of X, I can predict, let's say, 10 days later, or even just tomorrow, or the day after tomorrow, let's say the wind direction, is it possible to do that?

    If it is possible, then I will try using the sensor to measure my velocity at some place where I live, and I can run the model and see how the results look like. I don't know if it's going to accurately predict the future or within a 10% error bar range.

  • by joegibbs on 11/15/23, 1:27 AM

    When will we have enough data that we will be able to apply this to everything? Imagine a model that can predict all kinds of trends - what new consumer good will be the most likely to succeed, where the next war is most likely to break out, who will win the next election, which stocks are going to break out. One gigantic black box with a massive state, with input from everything - planning approvals, social media posts, solar activity, air travel numbers, seismic readings, TV feeds.
  • by user_7832 on 11/14/23, 5:59 PM

    (If someone with knowledge or experience can chime in, please feel free.)

    To the best of my knowledge, poor weather (especially wind shear/microbursts) are one of the most dangerous things possible in aviation. Is there any chance, or plans, to implement this in the current weather radars in planes?

  • by max_ on 11/14/23, 7:56 PM

    What's the difference between a "Graph Neural Network" and a deep neural network?
  • by carabiner on 11/14/23, 7:31 PM

    > GraphCast makes forecasts at the high resolution of 0.25 degrees longitude/latitude (28km x 28km at the equator).

    Any way to run this at even higher resolution, like 1 km? Could this resolve terrain forced effects like lenticular clouds on mountain tops?

  • by miserableuse on 11/14/23, 7:59 PM

    Does anybody know if its possible to initialize the model using GFS initial conditions used for the GFS HRES model? If so, where can I find this file and how can I use it? Any help would be greatly appreciated!
  • by devit on 11/15/23, 10:32 PM

    Seems like it would be much better to do conventional weather forecasting and then feed the predictions along with input data and other relevant information to a machine learning system.
  • by hammad93 on 11/15/23, 1:52 AM

    I think it's irresponsible to call first on this because it will hinder scientific collaboration. I appreciate this contribution but the journalism was sloppy.
  • by syntaxing on 11/14/23, 6:50 PM

    Maybe I missed it but does anyone know what it will take to run this model? Seems something fun to try out but not sure if 24GB of VRAM is suffice.
  • by tchvil on 11/15/23, 5:00 PM

    windguru which is in part or fully based on crowd-sourced weather stations is already surprisingly accurate few days in advance, in many regions I tried. For a few hours forecast nothing beats the rain radar. I wonder if they have already or will put some AI in their models.
  • by whoislewys_1 on 11/15/23, 3:00 AM

    Predicting weather and stock prices don't seem too far apart.

    Is it inevitable that all market alpha gets mined by AI?

  • by haolez on 11/14/23, 7:59 PM

    Are there any experts around that can chime in on the possible impacts of this technology if widely adopted?
  • by csours on 11/14/23, 10:35 PM

    Makes me wonder how much it would take to do this for a city at something like 100 meter resolution.
  • by dnlkwk on 11/14/23, 8:23 PM

    Curious how this factors in long-range shifts or patterns eg el nino. Most accurate is a bold claim
  • by supdudesupdude on 11/14/23, 9:56 PM

    I'll be impressed when it can predict rainfall better than GFS / HRRR / EURO etc
  • by knicholes on 11/14/23, 10:04 PM

    What are the similarities between weather forecasting and financial market forecasting?
  • by rottc0dd on 11/15/23, 8:31 AM

    How long does this forecasting hold, given butterfly effect et al?
  • by simonebrunozzi on 11/14/23, 8:10 PM

    Amazing. Is there an easy way to run this on a local laptop?
  • by sagarpatil on 11/14/23, 6:18 PM

    How does one go about hosting this and using this as an API?
  • by max_ on 11/14/23, 7:46 PM

    Has anyone here heard of "Numerical Forecasting" models for weather? I heard they "work so well".

    Does GraphCast come close to them?

  • by isaacfrond on 11/15/23, 3:23 PM

    I find this quite surprising actually.

    You'd think predicting the weather is mostly a matter of fast computation. The physical rules are well understood, so to get a better estimate use a finer mesh in your finite element computation and use a smaller time scale in estimating your differential equations.

    Neural networks are notoriously bad at exact approximation. I mean you can never beat a calculator when the issue is doing calculations.

    So apparently the AI found some shortcut for doing the actual computational work. That is also surprising as weather is a chaotic system. Shortcuts should not exist.

    Long story short, I don't get what's going on here.

  • by mg on 11/15/23, 3:36 PM

    It's interesting, that Google keeps publishing AI research papers. Is there a business rationale behind it?

    OpenAI has become one of the fastest growing companies of all time. And much of it is based on Google's "Attention is all you need" and other papers.

    Since Microsoft added the Dall-E 3 image creator to Bing, Bing saw a huge inflow of new users. Dall-E is also a technology rooted in Google papers.

    I wonder how Google thinks about this internally.

  • by max_ on 11/14/23, 7:44 PM

    I have far more respect for the AI team at DeepMind even thou they may be less popular than say OpenAI or "Grok".

    Why? Other AI studios seem to work on gimmicks while DeepMind seems to work on genuinely useful AI applications [0].

    Thanks for the good work!

    [0] Not to say that Chat GPT & Midjourney are not useful, I just find DeepMinds quality of research more interesting.

  • by alberth on 11/15/23, 5:33 PM

    Is this really an "AI" story?

    Aren't existing weather forecasting models, already a form of "AI"?

    I'm no AI/ML expert, but isn't the real story here is that a new model (like GPT-4.0) is better than the previous/existing model (GPT-3.5).

    It's just grabs way more attention calling the new model "AI" (vs not referring to the old as such).

  • by hexo on 11/15/23, 5:04 PM

    "AI" aka machine learning
  • by acolderentity on 11/15/23, 4:44 PM

    How could an ai, programmed with the bias of people that already suck at predicting the weather, even get close to being accurate?
  • by cryptoz on 11/14/23, 5:50 PM

    Again haha! Still no mention of using barometers in phones. Maybe some day.
  • by tokai on 11/15/23, 4:36 PM

    Just like their flu modelling outperformed conventional models right?
  • by drcongo on 11/15/23, 5:24 PM

    Is there a chance that it just made something up and got lucky like ChatGPT?
  • by amluto on 11/14/23, 5:45 PM

    I've never studied weather forecasting, but I can't say I'm surprised. All of these models, AFAICT, are based on the "state" of the weather, but "state" deserves massive scare quotes: it's a bunch of 2D fields (wind speed, pressure, etc) -- note the 2D. Actual weather dynamics happen in three dimensions, and three dimensional land features, buildings, etc as well as gnarly 2D surface phenomena (ocean surface temperature, ground surface temperature, etc) surely have strong effects.

    On top of this, surely the actual observations that feed into the model are terrible -- they come from weather stations, sounding rockets, balloons, radar, etc, none of which seem likely to be especially accurate in all locations. Except that, where a weather station exists, the output of that station is the observation that people care about -- unless you're in an airplane, you don't personally care about the geopotential, but you do care about how windy it is, what the temperature and humidity are, and how much precipitation there is.

    ISTM these dynamics ought to be better captured by learning them from actual observations than from trying to map physics both ways onto the rather limited datasets that are available. And a trained model could also learn about the idiosyncrasies of the observation and the extra bits of forcing (buildings, etc) that simply are not captured by the inputs.

    (Heck, my personal in-my-head neural network can learn a mapping from NWS forecasts to NWS observations later in the same day that seems better than what the NWS itself produces. Surely someone could train a very simple model that takes NWS forecasts as inputs and produces its estimates of NWS observations during the forecast period as outputs, thus handling things like "the NWS consistently underestimates the daily high temperature at such-and-such location during a summer heat wave.")

  • by greatpostman on 11/15/23, 3:27 PM

    OpenAI is releasing legitimate AGI, google puts out a weather prediction model lol.