"Now let me look at..." "Now I have a thorough understanding of.." "Now I have all the context I need..." "Now I have a comprehensive understanding of..." These sentences... 😅
Frederik Kratzert
@kratzert.bsky.social
Research Scientist @ Google Research. Working on the intersection of machine learning and hydrology. Located in Vienna, Austria.
If the Moon were only 1 pixel - A tediously accurate scale model of the solar system. One of the best sites ever. Click. Scroll. Enjoy! joshworth.com/dev/pixelspa...
If only a learning from the current situation would be to force push towards electrifying the grit and expanding renewables...something we could have learned already from the Russian invasion into Ukraine. But not in this country...
xLSTM Distillation: arxiv.org/abs/2603.15590 Near-lossless distillation of quadratic Transformer LLMs into linear-time xLSTM architectures enables cost- and energy-efficient alternatives without sacrificing performance. Efficient xLSTM variants of instruction-tuned Llama, Qwen, and Olmo models.
People definitely prefer my dataset publication posts over anything else I post on this homepage^^
Excited to announce Groundsource - an open-source dataset of historic flood events! This has easily been one of the coolest projects I've worked on recently! Thread 🧵 for details and all relevant links. 1/n
Excited to announce Groundsource - an open-source dataset of historic flood events! This has easily been one of the coolest projects I've worked on recently! Thread 🧵 for details and all relevant links. 1/n
I present the most interesting graph ever made. HUMAN ON BICYCLE beats every other living thing. www.scientificamerican.com/article/a-hu...
Interesting read and something I have been thinking about recently. How fit is our education system for the new era and what does this mean for _learning_ (especially learning to learn), if students/pupils over rely on LLMs. What does this mean for my kids (2nd grade and kindergarten)...?
Last term I tried an experiment: I walked into my Tech and Design Ethics class, admitted that I had *no idea* what to do about ChatGPT - so I would let them figure it out. As in: their first project was to decide and write the ChatGPT policy for the class. Here's what happened:
Some updates on NeuralHydrology and I'll also use this opportunity to give a shout-out to @gauchm.bsky.social for his continuous support and maintenance of NH
New #NeuralHydrology release 🎉 Some news from v1.13.0: * CAMELS-IND & CAMELS-DE support * AORC hourly forcing support * xLSTM supportSupport for embedding layers in MTS-LSTMs ...and various other improvements and fixes. The full release notes: github.com/neuralhydrol... Thanks to all contributors!
Are you working in large-sample hydrology? If so, we invite you to submit an abstract to our session for EGU 2026: HS2.4.2 "Large-Sample Hydrology: Advancing dataset developments, enhancing process understanding, and unifying insights through catchment modeling"
For those of you who are still undecided.
Back by popular demand: At #EGU26 we'll organize another BUGS session: Blunders, Unexpected Glitches, and Surprises! Submit abstracts on ideas that seemed great but didn't work, errors and bugs that led to new insights (or funny stories), or any other unexpected results. www.egu26.eu/session/56997
We (wife, 7 y/o, myself) spent New Years Eve playing "Stuffed Fables" boardgamegeek.com/boardgame/23... My wife's first roleplay game and also the most complex game we played so far with our oldest. Time was flying and we only made it to the start of the second chapter 😅
Stuffed Fables
Live the story of a child's stuffed toys, saving her from monsters under the bed.
boardgamegeek.com
Hydrology Paper of the Day @kratzert.bsky.social on integrating water quality observations into the Caravan dataset: the need for proper metadata and processing of observations; a new river network with water quality stations; catchment delineation; and validation and flagging of suspect data.
📦 Full Data: doi.org/10.24416/UU0... 📂 Lite (CSV): zenodo.org/records/1778... 💻 Code: github.com/SustainableW... 📖Preprint: eartharxiv.org/repository/v... #OpenScience #Hydrology #WaterQuality
Interesting point. I've been thinking about this as well in the past and I think, especially with ML, we have the opportunity to optimize for what we really want to know from the model, instead of training it on some proxy task
Check out the end-of-the-year blogpost about model design by Jonathan Frame! deepgroundwater.com/blog/what-do...
since everyone in Brazil's lost Twitter, I'm gonna try to make sure anything I post there is posted here too 🥲 here's a free building game I shared there a few weeks ago! mattstark.itch.io/tramstertram
the success of the user-created For You feed while the official Discover feed flops is actually a pretty major win for Bluesky — it literally is the embodiment of their goal to decentralize feed generation and give users more control/autonomy
Kid #2 got sick in time for Christmas and it seems I caught whatever he has, too. For the last three years we had at least one kid sick at Christmas. Happy Holidays 🎄
Excited to unveil Caravan-Qual! 🌊 We've expanded the Caravan framework to bridge the gap between water quantity and quality. This new open dataset features: 🔹 ~70M observations 🔹137k stations 🔹 100 constituents 🔹 Linked streamflow, forcing & attributes Links in 🧵
@rarakihydro.bsky.social leads an exiting new preprint that maps dominant hydrologic processes across >14k US watersheds using streamflow signatures + ML. It’s been such fun following Ryoko and Hilary’s leadership alongside Admin, Anne and Gemma. egusphere.copernicus.org/preprints/20...
It is so wild how all academics seem to have decided that LinkedIn is the new platform. Essentially the same announcement receives 100x the engagement.. On a job market platform. In the good old days of Twitter, this was different and it's kinda sad that not more researchers decided to join BlueSky
Excited to unveil Caravan-Qual! 🌊 We've expanded the Caravan framework to bridge the gap between water quantity and quality. This new open dataset features: 🔹 ~70M observations 🔹137k stations 🔹 100 constituents 🔹 Linked streamflow, forcing & attributes Links in 🧵
Excited to unveil Caravan-Qual! 🌊 We've expanded the Caravan framework to bridge the gap between water quantity and quality. This new open dataset features: 🔹 ~70M observations 🔹137k stations 🔹 100 constituents 🔹 Linked streamflow, forcing & attributes Links in 🧵
First time in years that I won't attend #AGU Inspiring week for everyone who attends and I hope you all stay safe.
Back by popular demand: At #EGU26 we'll organize another BUGS session: Blunders, Unexpected Glitches, and Surprises! Submit abstracts on ideas that seemed great but didn't work, errors and bugs that led to new insights (or funny stories), or any other unexpected results. www.egu26.eu/session/56997
Quick shout-out for an #EGU26 session that I'm not involved in but think is worth sharing: EOS4.4 BUGS: Blunders, Unexpected Glitches, and Surprises For a detailed session description, see www.egu26.eu/session/56997
Session EOS4.4
egu26.eu
What do people do when their taxi driver is a) speeding like crazy and b) interacting with his phone all the time while driving. Not sure if I'm unlucky but this happens 4/5 times to me when taking the airport taxi back home...
It has been a while since the last news from Caravan but I'm happy to announce that a new extension has just been published. Caravan-CZ, a dataset curated by CHMI itself in cooperation with Grey Nearing. For details check: github.com/kratzert/Car... All extensions: github.com/kratzert/Car...
[DATA CONTRIBUTION] CAMELS-CZ]: Catchment Attributes and Meteorology for Large-Sample Studies -- Czechia · Issue #49 · kratzert/Caravan
Basin prefix camelscz Zenodo DOI https://zenodo.org/records/17593968 Number of catchments 249 Location of catchments Czechia For which periods are streamflow records available in your dataset? 1950...
github.com
The paper is finally published and we were able to keep the original title. hess.copernicus.org/articles/29/...
How to deal w___ missing input data
Abstract. Deep learning hydrologic models have made their way from research to applications. More and more national hydrometeorological agencies, hydro power operators, and engineering consulting comp...
hess.copernicus.org
Guess the "joke" we were trying to make here was not as obvious than we thought.
jax.config.update("jax_debug_nans", True) saved my day.