Tom Andersson
@tom_r_andersson
Followers
728
Following
1K
Media
72
Statuses
412
Research Engineer @GoogleDeepMind; Building AI for climate change mitigation & adaptation; @WMO Young Scientist of the Year 2022; he/him
Joined February 2021
So excited to share our new @Nature paper on GenCast, an ML-based probabilistic weather forecasting model: It represents a substantial step forward in how we predict weather and assess the risk of extreme events. 🌪️🧵.
Today in @Nature, we’re presenting GenCast: our new AI weather model which gives us the probabilities of different weather conditions up to 15 days ahead with state-of-the-art accuracy. ☁️⚡. Here’s how the technology works. 🧵
2
2
28
RT @PeterWBattaglia: Today we’re thrilled to announce that real-time and historical AI-based weather forecasts from @Google’s WeatherNext s….
0
46
0
It's been an honour to work on this study led by @IlanPrice with such a talented team✨: Alvaro Sanchez-Gonzalez, @FerranAlet, @andrew_elkadi, @dominic_masters, @TimoEwalds, Jacklynn Stott, @Shakir_za, @PeterWBattaglia, Remi Lam, & Matthew Willson.
0
0
5
Like its predecessor (GraphCast), the weights & code of GenCast have been made publicly available: We’re looking forward to seeing how the community builds on this!.
1
0
2
A GenCast ensemble member takes 8 minutes on a TPU chip, versus hours on a supercomputer for physics-based models. This opens up the possibility of large ensembles (eg 1000s of members), which could better estimate the risks of extremes. We don't yet know how much this will help.
1
0
0
Cyclone max wind speeds are still underestimated, but this performance on tracks is really promising. One recent devastating cyclone was Hurricane Milton, which caused >$85 billion in damages. GenCast predicted ~70% probability of landfall in Florida 8.5 days before it struck.
1
0
0
We also extracted cyclone tracks from GenCast and ENS and compared them with ~100 cyclones observed in 2019. GenCast's ensemble mean cyclone track has a 12-hour position error advantage over ENS out to 4 days, and more actionable track probability fields out to 7 days.
1
0
0
We created a dataset of simulated wind power data at wind farm sites across the globe, and found that GenCast outperforms ENS by 10–20% up to 4 days ahead. This is promising, because better weather forecasts can reduce renewable energy uncertainty and accelerate decarbonisation.
1
0
1
GenCast uses diffusion to generate multiple 15-day forecast trajectories for the atmosphere. It assigns more accurate probabilities to possible weather scenarios than the SoTA physics-based ensemble system from ECMWF, across a 2019 evaluation period.
1
0
6
I’m not using X anymore, find me here: (And LinkedIn).
1
0
0
DeepSensor 0.4 brings improved support for environmental forecasting and non-Gaussian likelihoods ⚡️🌪️. This is a great way to rapidly build on recent advances in ConvNP weather prediction (eg Aardvark Weather). Colab demo: Docs:
1
4
26
RT @ikwess: I’m *finally* publicly announcing my PhD thesis! It’s been over two years, but here it is:. https://t.c….
0
14
0
RT @shoyer: I'm incredibly proud to share NeuralGCM, our new AI and physics based approach to weather and climate modeling with state-of-th….
0
241
0
I love this animation showing NeuralGCM capturing the seasonal cycle of tropical cyclone trajectories 🌪️.
NeuralGCM at 1.4 degree grid spacing can accurately simulate the seasonal cycle and many features of Earth's atmosphere, including emergent phenomena like tropical cyclones (at the correct locations/seasons/numbers). 17/
0
0
2
Congratulations to the NeuralGCM team for publishing in Nature! 🎉🎉. Very exciting initial results and an open-source codebase:
New @nature paper: NeuralGCM results (all are a "first"):.1) A differentiable hybrid atmospheric model.2) Competitive with ECMWF ensemble.3) Competitive with a GCRM in a year-long simulation.4) 40-year AMIP-like runs. Smaller bias than AMIP runs. 1/
1
2
14
So proud of my GDM colleagues who won the MacRobert Award for their work on GraphCast! 🥳🥳🥳 . GraphCast is an excellent example of open-source AI research and engineering applied to a problem that affects billions of lives. Find out more:
We won the MacRobert award by the Royal Academy of Engineering for our GraphCast work on AI for weather prediction!. I'm proud to be part of this team and very excited about what's to come in AI for Science&Sustainability!
0
2
18
RT @PeterWBattaglia: We're hiring a Research Scientist in AI for Sustainability @GoogleDeepMind (in Mountain View or Cambridge MA). Seeking….
0
42
0
My team at @GoogleDeepMind are pleased to share our new GenCast results! I'm so excited about the potential for better predictions of extreme weather & renewable energy, which is urgent for humanity as climate change continues. See @IlanPrice's thread👇
🌍 GenCast update! 🌍 New highlights include 0.25° resolution, predicting extreme weather, cyclones, and wind power production. Across all our evaluations, GenCast is better than the world’s top operational medium-range weather forecast. New paper: 🧵 1/8.
2
17
105
enjoying my life of anonymity after a subtle change to my surname.
This is Tom Anderson. At 35 years old, he sold Myspace for $580M. Then he bizarrely vanished. Here's his story:
0
0
22