Hi, I’m Martin. I work on the Flood Forecasting Initiative at Google Research in Switzerland. Previously, I was a PhD student in Sepp Hochreiter’s group at Johannes Kepler University, Linz, Austria, a graduate computer science student at the University of Waterloo, Canada, and an undergraduate student at Karlsruhe Institute of Technology, Germany.
My research is at the intersection of machine learning and hydrology. We focus on developing data-driven deep learning techniques for streamflow prediction. Visit our research blog at neuralhydrology.github.io for more information on current and recent projects.
Key Publications
For a full list of publications, refer to my Google Scholar profile.
2024
HESS Opinions: Never train a Long Short-Term Memory (LSTM) network on a single basin
Frederik Kratzert, Martin Gauch, Daniel Klotz, and Grey Nearing.
Hydrology and Earth System Sciences, September 2024.
Global Prediction of Extreme Floods in Ungauged Watersheds
Grey Nearing, Deborah Cohen, Vusumuzi Dube, Martin Gauch, Oren Gilon, Shaun Harrigan, Avinatan Hassidim, Daniel Klotz, Frederik Kratzert, Asher Metzger, Sella Nevo, Florian Pappenberger, Christel Prudhomme, Guy Shalev, Shlomo Shenzis, Tadele Tekalign, Dana Weitzner, and Yossi Matias.
Nature, March 2024.
2023
Conformal Prediction for Time Series with Modern Hopfield Networks
Andreas Auer, Martin Gauch, Daniel Klotz, and Sepp Hochreiter.
Advances in Neural Information Processing Systems, December 2023.
In Defense of Metrics: Metrics Sufficiently Encode Typical Human Preferences Regarding Hydrological Model Performance
Martin Gauch, Frederik Kratzert, Oren Gilon, Hoshin Gupta, Juliane Mai, Grey Nearing, Bryan Tolson, Sepp Hochreiter, and Daniel Klotz.
Water Resources Research, May 2023.
2022
Few-Shot Learning by Dimensionality Reduction in Gradient Space
Martin Gauch, Maximilian Beck, Thomas Adler, Dmytro Kotsur, Stefan Fiel, Hamid Eghbal-zadeh, Johannes Brandstetter, Johannes Kofler, Markus Holzleitner, Werner Zellinger, Daniel Klotz, Sepp Hochreiter, and Sebastian Lehner.
Proceedings of The 1st Conference on Lifelong Learning Agents (CoLLAs), August 2022, Montréal, Canada.
The Great Lakes Runoff Intercomparison Project Phase 4: The Great Lakes (GRIP-GL)
Juliane Mai, Hongren Shen, Bryan Tolson, Étienne Gaborit, Richard Arsenault, James R. Craig, Vincent Fortin, Lauren M. Fry, Martin Gauch, Daniel Klotz, Frederik Kratzert, Nicole O’Brien, Daniel G. Princz, Sinan Rasiya Koya, Tirthankar Roy, Frank Seglenieks, Narayan K. Shrestha, André G. T. Temgoua, Vincent Vionnet, and Jonathan W. Waddell.
Hydrology and Earth System Sciences, July 2022.
NeuralHydrology — A Python Library for Deep Learning Research in Hydrology
Frederik Kratzert, Martin Gauch, Grey Nearing, and Daniel Klotz.
Journal of Open Source Software, 7 (71), 4050, March 2022.
2021
Niederschlags-Abfluss-Modellierung mit Long Short-Term Memory (LSTM) (German article)
Frederik Kratzert, Martin Gauch, Grey Nearing, Sepp Hochreiter, and Daniel Klotz.
Österreichische Wasser- und Abfallwirtschaft, May 2021.
Multi-Timescale LSTM for Rainfall–Runoff Forecasting
Martin Gauch, Frederik Kratzert, Grey Nearing, Jimmy Lin, Sepp Hochreiter, Johannes Brandstetter, and Daniel Klotz.
EGU General Assembly, April 2021, virtual event.
Rainfall–Runoff Prediction at Multiple Timescales with a Single Long Short-Term Memory Network
Martin Gauch, Frederik Kratzert, Daniel Klotz, Grey Nearing, Jimmy Lin, and Sepp Hochreiter.
Hydrology and Earth System Sciences, April 2021 (preprint published in October 2020).
The Proper Care and Feeding of CAMELS: How Limited Training Data Affects Streamflow Prediction
(open-access version published on arXiv, November 2019)
Martin Gauch, Juliane Mai, and Jimmy Lin.
Environmental Modelling & Software, Volume 135, 2021, 104926, ISSN 1364-8152.
2020
A Machine Learner’s Guide to Streamflow Prediction (revised version of June 2020 arXiv paper)
Martin Gauch, Daniel Klotz, Frederik Kratzert, Grey Nearing, Sepp Hochreiter, and Jimmy Lin.
NeurIPS 2020 Workshop on AI for Earth Sciences, December 2020, virtual event.
A Data Scientist’s Guide to Streamflow Prediction
Martin Gauch and Jimmy Lin.
arXiv:2006.12975, June 2020.
An Open-Source Interface to the Canadian Surface Prediction Archive
Martin Gauch, James Bai, Juliane Mai, and Jimmy Lin.
Proceedings of the 20th ACM/IEEE-CS Joint Conference on Digital Libraries (JCDL 2020), August 2020, virtual event.
Machine Learning for Streamflow Prediction
Martin Gauch.
Master’s Thesis, April 2020, Waterloo, Canada.
2019
Streamflow Prediction with Limited Spatially-Distributed Input Data
Martin Gauch, Shervan Gharari, Juliane Mai, and Jimmy Lin.
Proceedings of the NeurIPS 2019 Workshop on Tackling Climate Change with Machine Learning, December 2019, Vancouver, Canada.
The Runoff Model-Intercomparison Project Over Lake Erie and the Great Lakes
Juliane Mai, Bryan Tolson, Hongren Shen, Étienne Gaborit, Vincent Fortin, Milena Dimitrijevic, Nicolas Gasset, Dorothy Durnford, Young Lan Shin, Tricia Anne Stadnyk, Oyémonbadé Hervé Rodrigue Awoye, Lauren M. Fry, Emily A. Bradley, Tim Hunter, Andrew Gronewold, Joeseph Smith, Lacey Mason, Laura Read, Katelyn FitzGerald, Kevin Michael Sampson, Alan F. Hamlet, Frank Seglenieks, André G. T. Temgoua, Shervan Gharari, Saman Razavi, Amin Haghnegahdar, Mohamed Elshamy, Daniel G. Princz, Alain Pietroniro, Xiaojing Ni, Yongping Yuan, Mohammad Reza Najafi, Melika Rahimimovaghar, Martin Gauch, Jimmy Lin, and Raphael Tang.
Abstracts of the 2019 AGU Fall Meeting, December 2019, San Francisco, USA.
Machine Learning for Streamflow Prediction: Current Status and Future Prospects
Martin Gauch, Raphael Tang, Juliane Mai, Bryan Tolson, Shervan Gharari, and Jimmy Lin.
Abstracts of the 2019 AGU Fall Meeting, December 2019, San Francisco, USA.
The Cuizinart: Slice and Dice Your Environmental Datasets
Jimmy Lin, Martin Gauch, Yixin Wang, Alexander Weatherhead, Kaisong Huang, Bhaleka D. Persaud, and Juliane Mai.
Abstracts of the 2019 AGU Fall Meeting, December 2019, San Francisco, USA.
The Proper Care and Feeding of CAMELS: How Limited Training Data Affects Streamflow Prediction
(now published in Environmental Modelling & Software, Volume 135, 2021)
Martin Gauch, Juliane Mai, and Jimmy Lin.
arXiv:1911.07249, November 2019.
Data-Driven vs. Physically-Based Streamflow Prediction Models
Martin Gauch, Juliane Mai, Shervan Gharari, and Jimmy Lin.
Proceedings of the 9th International Workshop on Climate Informatics, October 2019, Paris, France.
2018
Towards Simulation-Data Science—A Case Study on Material Failures
Holger Trittenbach, Martin Gauch, Klemens Böhm, and Katrin Schulz.
2018 IEEE 5th International Conference on Data Science and Advanced Analytics, October 2018, Turin, Italy.
Education
Johannes Kepler University Linz
Machine Learning
(PhD student)
January 2021 – July 2023
At the Institute for Machine Learning, my research was at the intersection of machine learning and hydrology under the supervision of Sepp Hochreiter. We have developed data-driven deep learning techniques for environmental modeling. Visit our research blog at neuralhydrology.github.io for more information on recent projects.
University of Waterloo & Karlsruhe Institute of Technology
Computer Science
(Master's student)
September 2017 – April 2020
After two exchange terms at the University of Waterloo, I decided to continue pursuing my Master’s in Canada. This allowed me to do a lot more research throughout my degree.
Supervised by Jimmy Lin, my research was on machine learning in hydrology.
I developed neural and non-neural machine learning algorithms to predict streamflow from spatially and temporally distributed input data.
A large part of this work was in collaboration with environmental scientists.
Karlsruhe Institute of Technology
Computer Science
(Bachelor's student)
October 2014 – September 2017
Bachelor’s thesis: Data-Driven Approaches to Predict Material Failure and Analyze Material Models
In my thesis, I developed data-driven models to predict failure events of an observed material and compared traditional machine learning approaches with recurrent neural network architectures.
Work Experience
I contribute to Google’s Flood Forecasting efforts in the hydrologic modeling team, where we build Deep Learning streamflow prediction models that integrate into an operational end-to-end flood warning system. You can see the system in action at g.co/floodhub.
I developed neural and non-neural machine learning algorithms to predict streamflow from spatially and temporally distributed input data. A large part of this work was in collaboration with environmental scientists. Besides, we worked with domain specialists from the environmental sciences on developing the Cuizinart, a cloud-based interactive platform that allows researchers to “slice and dice” large environmental datasets.
Data Systems Lab, University of Waterloo
Graduate Research Assistant
January 2019 – April 2020
cs.uwaterloo.ca
Besides my main research focus on machine learning during my graduate studies, I worked as a graduate research assistant. In collaboration with domain specialists from the environmental sciences, we developed the cloud-based Cuizinart platform that provides access to large environmental datasets. Also, we worked on solutions for metadata management in the Global Water Futures program.
In parallel to my studies in Karlsruhe, I worked part-time as a working student in the cloud HR software development at SAP SuccessFactors. As part of the Rules Engine scrum team, I worked on designing and developing a rules framework that enables users to customize automatic business processes in their HR system. My work involved Java programming as well as data analytics tasks, using Python, Jupyter notebooks, SQL, and Splunk.
Between high school and university, I did a full-time internship in software development. I designed, developed, and tested on-premise HR software for the SAP ERP solution. My work mostly involved programming in ABAP. Also, I learned about SAP’s ERP software and its customizing.