Pinned
"Introduction to Physical Oceanography" by R. Stewart is now on GitHub. Free for everyone as it always was. github.com/introocean/i...
Nikolay Koldunov
@oceanographer.bsky.social
Oceanographer, climate and ocean modeler, interested in ML for climate and LLMs. Senior scientist at AWI. Working on FESOM2 unstructured ocean model, km-scale climate modelling, data visualisation. More info: https://koldunovn.github.io/
782 followers898 following86 posts
Pinned
"Introduction to Physical Oceanography" by R. Stewart is now on GitHub. Free for everyone as it always was. github.com/introocean/i...
My brothers released a new music album. Check it out! Spotify: open.spotify.com/album/3ihT8A... Bandcamp: goinsane.bandcamp.com/album/nonsense YouTube Music: music.youtube.com/playlist?lis...
Can an LLM port a complex Fortran ocean model to C++/Kokkos and make it run on GPUs? Yes. We did it with FESOM2: production runs, ~3–4× speedup vs CPU, same code on LUMI, MareNostrum5 and JUPITER. Fortran as executable specification? Maybe yes :) arxiv.org/abs/2606.11356
Nikolay Koldunov reposted
🔴 Are you in #Barcelona? Are you interested in #ClimateChange? Then this **EERIE Project public lecture** is for you! ⤵️ 💡 New Views of Climate Change - Insights from Kilometer-Scale Earth System Models 🗓️ Monday, March 9 🕕 6pm CET 🏢 @bsc-cns.bsky.social Register here: events.hifis.net/event/3767/
We posted a preprint on arXiv about PANGAEA-GPT, a multi-agent system for autonomous data discovery and analysis in geoscientific archives. Plots below are generated by the system, and you can read about the steps system made in the paper. 🧪 📄 arxiv.org/abs/2602.21351 💻 github.com/CliDyn/panga...
Built a demo agent for ERA5 on @earthmover.io Arraylake: fast slice access means you can retrieve, iterate & visualize without the coffee breaks. Eurus supports NL→retrieval+analysis, CLI/web/MCP, and a Python REPL (pandas/xarray/cartopy). Video below. Repo+HF: lnkd.in/deQMCNmH lnkd.in/dAbhnKSr
🌍 ClimSight is a system that allows users to get answers to climate-related questions based on very local climate information. Check out the short video to learn the core ideas, and dive into the paper if you’d like to explore the details: 📄 lnkd.in/eReWQ4x6 🧪
How much do LLMs “know” about present-day climate? We asked 17 models for July 1991–2020 temperatures at 15 000+ land points — and many did surprisingly well! Preprint: arxiv.org/abs/2510.08043 🧪 #LLM #Climate
DestinE’s just dropped “From Past to Future” videos: km-scale replays of real floods & hurricanes under past, present & +2 °C climates. See exactly how warming reshapes impacts—turning high-res science into must-see visuals. 🧪 destine.ecmwf.int/news/storyli... www.youtube.com/watch?v=5V62...
New opportunity in our group!
💼 Looking for a #job? We're #hiring! The @awi.de is seeking to employ a skilled and motivated Scientist with expertise in Python programming and a solid background in oceanic, atmospheric or climate dynamics. ➡️ Check all the details on our website: eerie-project.eu/job-applicat...
How will interaction with geophysical data evolve in the age of AI? 🤖🌍 In our latest preprint, we explore this question - not just conceptually, but with a hands-on prototype using the PANGAEA archive. Check it out! arxiv.org/abs/2503.05854
🚨 Last chance to submit your application! 🚨 🌊 Scientist in High-Resolution Sea Ice Modeling 🔗 jobs.awi.de/Vacancies/19... 🧠 Scientist in AI-based Emulation for High-Resolution Sea Ice Modeling 🔗 jobs.awi.de/Vacancies/19... Please share! 🧪❄️🌍 #ScienceJobs #SeaIce #AI
Nikolay Koldunov reposted
We have taken our Artificial Intelligence Forecasting System (AIFS) into operations today to run side by side with our physics-based weather forecasts. The #AIFS outperforms traditional models for many measures, at a fraction of energy use. ➡️ www.ecmwf.int/en/about/med... #AI #MachineLearning
One of the greatest things we do in #DestinE is high-resolution storylines, which show how present-day extreme events might unfold in the future. Check out this ECMWF blog post for details on how it’s done! 🧪
🌍🌐 What if we could replay recent extreme weather under different climate conditions? With storyline simulations in the EU’s #DestinE initiative, scientists are reconstructing past, present, and future extreme weather to help us adapt to climate change. Learn more ➡️ destine.ecmwf.int/news/replayi...
Alfred Wegener institute is finally here @awiexpedition.bsky.social !!!
Preprint summarizing the first results from the nextGEMS project, which prepares several coupled models to run at kilometer scale: egusphere.copernicus.org/preprints/20... 🧪
Nikolay Koldunov reposted
@carbonbrief.org lists our December paper among the most featured 2024 papers, current score 2,644 - top of the charts for @altmetric.com mentions on @bsky.app 🤯 📈 www.carbonbrief.org/analysis-the... 🔓A free-access reprint link to our paper can be found on Helges website www.awi.de/en/about-us/...
We have a new paper out in @science.org today, led by Helge Goessling from #AWI: The recent global temperature surge in 2023 was intensified by a record-low planetary #albedo 👉 doi.org/10.1126/scie... @thomasjung.bsky.social @ecmwf.bsky.social This is what we found (🧵1/8)
It's always great to visit Helsinki-one of my favorite cities in the world. This time, I caught a glimpse of sea ice, which we’ll be modeling in the @terradt.bsky.social project (by the way, we're hiring!). The project kicked off in Helsinki this week.
🚨 Job Alert! 🚨 Our team is hiring for the @terradt.bsky.social project. 🌊 Scientist in High-Resolution Sea Ice Modeling 🔗 jobs.awi.de/Vacancies/19... 🧠 Scientist in AI-based Emulation for High-Resolution Sea Ice Modeling 🔗 jobs.awi.de/Vacancies/19... Please share! 🧪❄️🌍 #ScienceJobs #SeaIce #AI
This is the description of our km-scale coupled climate model configuration. A lot of details in a thread from @trackow.bsky.social 🧪
Our article is out in @eurogeosciences.bsky.social's GMD journal, on global multi-year kilometre-scale modelling with @ecmwf.bsky.social’s IFS coupled to FESOM2.5 (finer than 5km). See the paper here: doi.org/10.5194/gmd-... This is what we found: (🧵1/n)
New IBCAO (International Bathymetric Chart of the Arctic Ocean) version is out: doi.org/10.1038/s415... Many of us who study the Arctic now need to update our Arctic Bathymetry posters on our walls! :) Figure below shows a comparison of versions 4 and 5. 🧪🌊
Great thread from @jhauck.bsky.social with more details on this study 🧪
Massive piece of work by Laurent Oziel and many #AWI colleagues, conducted within the #COMFORT project: Laurent conducted future 🌊 Arctic Ocean high-resolution simulations that account for climate change AND carbon & nutrient inputs from rivers & coastal erosion. ⬇️ doi.org/10.1038/s415...
Increased sunlight and more nutrients in the Arctic Ocean won’t boost carbon storage as expected :( 🌊🧪 #FESOM Article in Nature Climate Change: www.nature.com/articles/s41... Press release from AWI: www.awi.de/en/about-us/...
Nikolay Koldunov reposted
Nikolay Koldunov reposted
Join me and the polar team here in UCL Earth Sciences + Met Office on an exciting CASE studentship looking at Leveraging Machine Learning algorithms for improved Arctic sea-ice prediction using the Met Office suite of models ❄️ + 💻 +🛰️ www.trees-dla.ac.uk/projects/lev...
Nikolay Koldunov reposted
Exploring the Arctic Ocean? Planning to attend EGU25? Submit an abstract to our session:"Changes in the Arctic Ocean, Sea Ice, and Subarctic Sea Systems: Observations, Models, and Perspectives" (t.ly/BvM9l) Spread the word, and you see in Vienna for exciting talks, posters and discussions!
Nikolay Koldunov reposted
Our work will also be (remotely) presented at #AGU24 ! agu.confex.com/agu/agu24/me... Dec 11, 01:30pm - 02:30pm (Central Time) Dec 12, 07:30am - 08:30am (Central Time)
We have a new paper out in @science.org today, led by Helge Goessling from #AWI: The recent global temperature surge in 2023 was intensified by a record-low planetary #albedo 👉 doi.org/10.1126/scie... @thomasjung.bsky.social @ecmwf.bsky.social This is what we found (🧵1/8)
🧪 Link to the #agu24 talk today: agu.confex.com/agu/agu24/me...
🌪️ Tales from the Storm at #AGU24! Check out our km-scale storyline simulations of Hurricane Helene and other 2024 extremes. 📢 @thomasjung.bsky.social will present: How storyline simulations answer “what if” questions. 📅 Tue, 10 Dec 2024📍NH22B, 202 A ✨ Learn more: doi.org/10.22541/ess...
Nikolay Koldunov reposted
So exited to be co-hosting this #OpenScience event with @earthmoverhq.bsky.social and @developmentseed.org at #AGU24. Come join us on Tuesday!
Beyond the presentations, AGU is about community! Join us at these social + side events to connect, recharge, and collaborate. ✨ Cloud Connections: Open Science Social Dec 10, 6-8pm, co-hosted w/ @earthmoverhq.bsky.social RSVP: lu.ma/0buymhbd #AGU24
🧪 New paper on why we’ve seen an extra 0.2°C of global warming in recent years.
We have a new paper out in @science.org today, led by Helge Goessling from #AWI: The recent global temperature surge in 2023 was intensified by a record-low planetary #albedo 👉 doi.org/10.1126/scie... @thomasjung.bsky.social @ecmwf.bsky.social This is what we found (🧵1/8)
Please, spread the word! 🌊❄️🧪 Exciting opportunities in polar science! Join Lettie Roach’s new group on Southern Ocean & Antarctic Sea Ice Evolution. Multiple positions open: Postdoc, PhDs, and a Scientific Programmer. Deadline: Dec 31, 2024.
Several positions as part of Lettie Roach's new group 'Southern Ocean & Antarctic Sea ice Evolution' 🌊❄️🌍 Deadline for applications is Dec 31st 2024 Postdoc jobs.awi.de/Vacancies/18... PhD jobs.awi.de/Vacancies/18... PhD jobs.awi.de/Vacancies/18... Scientific programmer jobs.awi.de/Vacancies/18...
Wind over the Atlantic from a 5 km resolution model. Different surface types interact with wind in unique ways, which is why the continents are distinctly visible in the wind field. 🧪 Model: IFS-FESOM | U component of the wind | Project: nextGEMS | Simulations: @trackow.bsky.social #SciArt
Another mind blowing result of AI models in Numerical Weather Prediction. The next step is to train directly on observations.
A paper in Nature presents a machine learning model, called ‘GenCast,’ that can produce a reliable probabilistic weather forecast, expressed as probabilities of various outcomes, based on current and future weather. https://go.nature.com/3B8PsdD 🧪
Nikolay Koldunov reposted
📣 And the time has come to open the call for abstracts for the Liège Colloquium on Ocean Dynamics for 2025! 🌊 Topic: Marine CDR emerging views and challenges Date: 26-30 May 2025 Abstract deadline: 31 January 2025 More info? www.ocean-colloquium.uliege.be
Earth’s surface painted with 2m temperature. 🧪 Contrasts in temperature between the ocean and land, as well as between different orographic features, allow you to clearly see surface details. Model: IFS-FESOM | Shown: 2m temperature | Project: nextGEMS| Simulations @trackow.bsky.social #SciArt
AI weather forecast models will likely soon become a significant part of regular weather prediction. They already outperform “traditional” models on many skill scores. But how do these models perform in out-of-sample conditions - across past and future climates? 🧪
First post on @bsky.app 🎉: Can #AI-based weather forecasting models (trained on present-day data) provide skillful forecasts also in different colder and warmer states of the climate system? Our preprint in arxiv explores this question: doi.org/10.48550/arX... Here is what we found so far (🧵1/7)
Nikolay Koldunov reposted
🌊That’s a very rare and important paper about finding and fixing a (sea ice related) bug in a climate model. We need more papers like this so the community can learn from mistakes. 🌊🧪
New class of model bugs detected: resolution-dependent bugs emerge as we transition to kilometer-scale climate models, revealing issues hidden in lower resolutions. Preprint available, thanks to work in the EU projects nextGEMS and @eerie-project.bsky.social. doi.org/10.5194/egus...