Frederik Kratzert
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Frederik Kratzert has a total of 298 co-authors. Here are the top co-authors by number of shared publications:
- Daniel Klotz
Interdisciplinary Transformation University Austria, Linz, Austria
81 shared publications - Martin Gauch
Google research
77 shared publications Grey Nearing
42 shared publicationsOren Gilon
Google Research, Electra Tower, Yigal Alon St 98, Tel Aviv-Yafo, 6789141, Israel
34 shared publicationsGuy Shalev
Google DeepMind, Electra Tower, Yigal Alon St 98, Tel Aviv-Yafo, 6789141, Israel
27 shared publicationsSepp Hochreiter
ELLIS Unit Linz and LIT AI Lab, Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
27 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-21
Everything everywhere all at once: A single-cell LSTM network unifying multi-frequency, missing data, and discharge assimilation for robust operational flood forecasting
Long short-term memory (LSTM) networks have demonstrated state-of-the-art capabilities in operational flood forecasting, with recent missing-data handling strategies further improving pipeline robustness. Building on these advancements, we introduce the multi-frequency masked-forecasting LSTM (MF2LSTM), a model architecture that combines missing-data workflows with multi-frequency approaches to generate hourly streamflow forecasts. Additionally, our framework features a flexible data assimilation strategy to include real-time discharge information. This approach enhances forecast performance when real-time observations are available, and allows the model to keep operating when the discharge signal is absent, either by gauge failure or for prediction in ungauged basins. We benchmarked the MF2LSTM against the Large Area Runoff Simulation (LARSIM) model, the current operational flood-forecasting model used by several European countries, forcing both models with ICON-D2 meteorological products. Our results indicate that the MF2LSTM yields higher predictive accuracy than LARSIM. Overall, this framework presents a robust operational pipeline that demonstrates the viability of deep learning for hourly forecasting. Furthermore, by relying on a standard single-cell LSTM architecture, the approach highlights that structurally simple deep learning architectures can achieve high operation performance.
- 3 September 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-21
Open AccessEverything everywhere all at once: A single-cell LSTM network unifying multi-frequency, missing data, and discharge assimilation for robust operational flood forecasting
Long short-term memory (LSTM) networks have demonstrated state-of-the-art capabilities in operational flood forecasting, with recent missing-data handling strategies further improving pipeline robustness. Building on these advancements, we introduce the multi-frequency masked-forecasting LSTM (MF2LSTM), a model architecture that combines missing-data workflows with multi-frequency approaches to generate hourly streamflow forecasts. Additionally, our framework features a flexible data assimilation strategy to include real-time discharge information. This approach enhances forecast performance when real-time observations are available, and allows the model to keep operating when the discharge signal is absent, either by gauge failure or for prediction in ungauged basins. We benchmarked the MF2LSTM against the Large Area Runoff Simulation (LARSIM) model, the current operational flood-forecasting model used by several European countries, forcing both models with ICON-D2 meteorological products. Our results indicate that the MF2LSTM yields higher predictive accuracy than LARSIM. Overall, this framework presents a robust operational pipeline that demonstrates the viability of deep learning for hourly forecasting. Furthermore, by relying on a standard single-cell LSTM architecture, the approach highlights that structurally simple deep learning architectures can achieve high operation performance.
- Open Access
- 26 June 2026
- Published by Copernicus GmbH
Supplementary material to "GRIT-ADB: A Global Attribute Database for the GRIT Hydrography"
- 26 June 2026
- Published by Copernicus GmbH
Open AccessSupplementary material to "GRIT-ADB: A Global Attribute Database for the GRIT Hydrography"
- Open Access
- 26 June 2026
- Published by Copernicus GmbH in EGUsphere
GRIT-ADB: A Global Attribute Database for the GRIT Hydrography
Global hydro-environmental databases provide essential information for large-scale hydrological, ecological, geomorphological, and Earth system analyses. Most existing global databases are built upon convergent river representations that do not explicitly capture bifurcating, multi-channel river systems. In addition, these databases primarily characterise long-term climatological means or static representations of environmental conditions derived from earlier-generation global datasets, limiting their applicability for time-varying analyses of hydroclimatic and geomorphological processes. Here we present GRIT-ADB, a new attribute database for the vectorised Global River Topology (GRIT), a new river hydrography dataset created from a 30 m resolution river mask and terrain data that provides a topology-explicit and physically realistic representation of river networks including divergent flow pathways. GRIT-ADB provides standardised hydro-environmental information for 19.6 million km of rivers and streams. It currently comprises 64 time-varying (multi-dimensional embeddings) and 35 static variables (>300 attributes), spanning five categories: hydrology, physiography, climate, land cover and use, and soils and geology. Attributes are derived by aggregating and harmonising data from state-of-the-art global datasets and are accumulated along the river network from headwaters to basin outlets, while preserving the topology of divergent and complex flow pathways. The attributes are linked to multiple GRIT scales, including hierarchically-nested subbasins, individual river reaches of up to 1 km long, and coarser-scale river segments of several kilometres long, providing a flexible framework that can accommodate future extensions of GRIT and additional attributes. By combining a standardised attribute framework with explicit representation of bifurcating river hydrography, GRIT-ADB enables improved large-scale yet high-resolution analyses of river connectivity, hydrological extremes, hydro-ecological processes, and climate impacts in complex river systems, supporting a wide range of global hydrological and environmental applications.
- 26 June 2026
- Published by Copernicus GmbH in EGUsphere
Open AccessGRIT-ADB: A Global Attribute Database for the GRIT Hydrography
Global hydro-environmental databases provide essential information for large-scale hydrological, ecological, geomorphological, and Earth system analyses. Most existing global databases are built upon convergent river representations that do not explicitly capture bifurcating, multi-channel river systems. In addition, these databases primarily characterise long-term climatological means or static representations of environmental conditions derived from earlier-generation global datasets, limiting their applicability for time-varying analyses of hydroclimatic and geomorphological processes. Here we present GRIT-ADB, a new attribute database for the vectorised Global River Topology (GRIT), a new river hydrography dataset created from a 30 m resolution river mask and terrain data that provides a topology-explicit and physically realistic representation of river networks including divergent flow pathways. GRIT-ADB provides standardised hydro-environmental information for 19.6 million km of rivers and streams. It currently comprises 64 time-varying (multi-dimensional embeddings) and 35 static variables (>300 attributes), spanning five categories: hydrology, physiography, climate, land cover and use, and soils and geology. Attributes are derived by aggregating and harmonising data from state-of-the-art global datasets and are accumulated along the river network from headwaters to basin outlets, while preserving the topology of divergent and complex flow pathways. The attributes are linked to multiple GRIT scales, including hierarchically-nested subbasins, individual river reaches of up to 1 km long, and coarser-scale river segments of several kilometres long, providing a flexible framework that can accommodate future extensions of GRIT and additional attributes. By combining a standardised attribute framework with explicit representation of bifurcating river hydrography, GRIT-ADB enables improved large-scale yet high-resolution analyses of river connectivity, hydrological extremes, hydro-ecological processes, and climate impacts in complex river systems, supporting a wide range of global hydrological and environmental applications.
- Open Access
- 8 June 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Comment on egusphere-2026-2283 - No compliance with the policy of the journal
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- 8 June 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Open AccessComment on egusphere-2026-2283 - No compliance with the policy of the journal
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- Open Access
- 28 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Comment on egusphere-2026-2283
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- 28 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Open AccessComment on egusphere-2026-2283
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- Open Access
- 25 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Comment on egusphere-2026-2283
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- 25 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-31
Open AccessComment on egusphere-2026-2283
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- Open Access
- 29 April 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-31
Extending Medium-Range Global Flood Forecasts: The Google Global Flood Forecasting Model Version 2
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- 29 April 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-31
Open AccessExtending Medium-Range Global Flood Forecasts: The Google Global Flood Forecasting Model Version 2
This paper evaluates an updated flood forecasting system that significantly extends reliable lead times. We evaluated this updated model (v2) against the prior system (v1) and established third-party benchmarks across 1,223 global test basins. The primary finding is that the v2 system extends the reliable predictive horizon by 6 days in gauged basins and 1 day in ungauged basins relative to the v1 nowcast, as measured by the Nash Sutcliffe Efficiency. Along with this paper, we release an open-source codebase for training both the v1 and v2 forecast models with the open-source Caravan dataset.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Introducing Caravan-Qual: A comprehensive global river water quality dataset
Here, we introduce Caravan-Qual, a new water quality dataset which represents the first global integration with large-sample hydrology. The dataset contains >70 million river water quality observations covering 100 water quality constituents, compiled from a range of national-to-global datasets covering the period of 1980-2025. By leveraging the Caravan dataset and open-source software, we have matched water water quality monitoring stations to streamflow gauges and have derived meterological variables which together provide contextual environmental data which is envisaged to aid our understanding of the water quality data and facilitate research into topics including:Spatio-temporal analysis of river water quality dynamics at local to global scales.Investigation of the relationships between (constituent-specific) river water quality responses and hydrological, meteorological and catchment characteristics.The development and evaluation of process-based, hybrid and data-driven water quality models across diverse hydrological and climatic conditions.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessIntroducing Caravan-Qual: A comprehensive global river water quality dataset
Here, we introduce Caravan-Qual, a new water quality dataset which represents the first global integration with large-sample hydrology. The dataset contains >70 million river water quality observations covering 100 water quality constituents, compiled from a range of national-to-global datasets covering the period of 1980-2025. By leveraging the Caravan dataset and open-source software, we have matched water water quality monitoring stations to streamflow gauges and have derived meterological variables which together provide contextual environmental data which is envisaged to aid our understanding of the water quality data and facilitate research into topics including:Spatio-temporal analysis of river water quality dynamics at local to global scales.Investigation of the relationships between (constituent-specific) river water quality responses and hydrological, meteorological and catchment characteristics.The development and evaluation of process-based, hybrid and data-driven water quality models across diverse hydrological and climatic conditions.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
How to deal w___ missing input data
Hydrologic deep learning models have made their way from research to applications. More and more national hydrometeorological agencies, hydro power operators, and consulting companies are building Long Short-Term Memory (LSTM) models for operational use cases. However, all of these efforts are confronted with similar sets of challenges—issues that are different from those in controlled scientific studies. One common issue is the question: how to deal with missing input data? Operational systems depend on the real-time availability of various data products—most notably, meteorological forcings. Additional forcings generally improve the model performance, but at the same time, every new dependency increases the likelihood of an outage in one of the input data products. In a recent study, we evaluated different solutions to generate predictions even when some of the meteorological input data do not arrive in time, or not arrive at all (Gauch et al., 2025). In this presentation, we will introduce these methods and discuss how they can help (1) operational forecasters to run reliable real-time flood forecasting systems, and (2) researchers and modelers to build accurate models that leverage as much data as possible. Gauch, M., Kratzert, F., Klotz, D., Nearing, G., Cohen, D., and Gilon, O.: How to deal w___ missing input data, Hydrol. Earth Syst. Sci., 29, 6221–6235, 2025.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessHow to deal w___ missing input data
Hydrologic deep learning models have made their way from research to applications. More and more national hydrometeorological agencies, hydro power operators, and consulting companies are building Long Short-Term Memory (LSTM) models for operational use cases. However, all of these efforts are confronted with similar sets of challenges—issues that are different from those in controlled scientific studies. One common issue is the question: how to deal with missing input data? Operational systems depend on the real-time availability of various data products—most notably, meteorological forcings. Additional forcings generally improve the model performance, but at the same time, every new dependency increases the likelihood of an outage in one of the input data products. In a recent study, we evaluated different solutions to generate predictions even when some of the meteorological input data do not arrive in time, or not arrive at all (Gauch et al., 2025). In this presentation, we will introduce these methods and discuss how they can help (1) operational forecasters to run reliable real-time flood forecasting systems, and (2) researchers and modelers to build accurate models that leverage as much data as possible. Gauch, M., Kratzert, F., Klotz, D., Nearing, G., Cohen, D., and Gilon, O.: How to deal w___ missing input data, Hydrol. Earth Syst. Sci., 29, 6221–6235, 2025.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Semi-Distributed Deep Learning Models at Global Scale
Hydrological modeling has reached a point where deep learning models, especially those based on the LSTM architecture, are being used operationally by national agencies, the private sector, as well as across thousands of academic publications. However, by far the most common strategy is to use these models in a lumped setup, no matter the scale of the application. For example, Nearing et al. (2024) apply LSTM based models in a lumped setup at a global scale. Unfortunately, this withholds information that is relevant to precisely predict floods, such as the location of a precipitation event relative to the prediction point. Similarly, the spatial averaging over the entire upstream area dampens the precipitation signal that is provided to the model.Classical hydrologic models use distributed or semi-distributed setups to solve this problem: they divide the basin into pixels or subpolygons and route streamflow along the river graph. There are first attempts to translate this semi-distributed modeling paradigm to end-to-end deep learning models, but so far they are typically trained only on individual river networks or select geographical regions (e.g., Kratzert et al., 2021, Kraft et al., 2025), lag behind the performance of lumped models (e.g., Kirschstein et al., 2021), cannot generalize to unseen river networks (e.g., Vischer et al., 2025), or are global and applicable ungauged basins but not trained end-to-end (e.g., Yang et al., 2025).With the learnings and experience from operating lumped LSTM models at a global scale for multiple years, we revisit semi-distributed modeling with deep learning at a global scale with a focus on end-to-end training. In this submission, we present our version of a global end-to-end semi-distributed hydrologic model. We detail the model setup, its training procedure, and compare this model to the lumped setup. Our evaluation shows that the semi-distributed model has superior performance compared to the lumped model, especially for large, ungauged rivers. Finally, we highlight how this modeling approach is a step towards a broader multi-output, multi-modal system that propagates more information than just streamflow or physical quantities in general.References:Kirschstein, N., et al. "The Merit of River Network Topology for Neural Flood Forecasting." Forty-first International Conference on Machine Learning. 2024.Kraft, B., et al. DROP: A scalable deep learning approach for runoff simulation and river routing. Authorea. November 25, 2025.Kratzert, F., et al. "Large-scale river network modeling using graph neural networks." EGU General Assembly Conference Abstracts. 2021.Nearing, G., et al. "Global prediction of extreme floods in ungauged watersheds." Nature 627.8004 (2024): 559-563.Vischer, M., et al. "Spatially Resolved Rainfall Streamflow Modeling in Central Europe." EGUsphere 2025 (2025): 1-26.Yang, Y., et al. (2025). Global daily discharge estimation based on grid long short-term memory (LSTM) model and river routing. Water Resources Research, 61.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessSemi-Distributed Deep Learning Models at Global Scale
Hydrological modeling has reached a point where deep learning models, especially those based on the LSTM architecture, are being used operationally by national agencies, the private sector, as well as across thousands of academic publications. However, by far the most common strategy is to use these models in a lumped setup, no matter the scale of the application. For example, Nearing et al. (2024) apply LSTM based models in a lumped setup at a global scale. Unfortunately, this withholds information that is relevant to precisely predict floods, such as the location of a precipitation event relative to the prediction point. Similarly, the spatial averaging over the entire upstream area dampens the precipitation signal that is provided to the model.Classical hydrologic models use distributed or semi-distributed setups to solve this problem: they divide the basin into pixels or subpolygons and route streamflow along the river graph. There are first attempts to translate this semi-distributed modeling paradigm to end-to-end deep learning models, but so far they are typically trained only on individual river networks or select geographical regions (e.g., Kratzert et al., 2021, Kraft et al., 2025), lag behind the performance of lumped models (e.g., Kirschstein et al., 2021), cannot generalize to unseen river networks (e.g., Vischer et al., 2025), or are global and applicable ungauged basins but not trained end-to-end (e.g., Yang et al., 2025).With the learnings and experience from operating lumped LSTM models at a global scale for multiple years, we revisit semi-distributed modeling with deep learning at a global scale with a focus on end-to-end training. In this submission, we present our version of a global end-to-end semi-distributed hydrologic model. We detail the model setup, its training procedure, and compare this model to the lumped setup. Our evaluation shows that the semi-distributed model has superior performance compared to the lumped model, especially for large, ungauged rivers. Finally, we highlight how this modeling approach is a step towards a broader multi-output, multi-modal system that propagates more information than just streamflow or physical quantities in general.References:Kirschstein, N., et al. "The Merit of River Network Topology for Neural Flood Forecasting." Forty-first International Conference on Machine Learning. 2024.Kraft, B., et al. DROP: A scalable deep learning approach for runoff simulation and river routing. Authorea. November 25, 2025.Kratzert, F., et al. "Large-scale river network modeling using graph neural networks." EGU General Assembly Conference Abstracts. 2021.Nearing, G., et al. "Global prediction of extreme floods in ungauged watersheds." Nature 627.8004 (2024): 559-563.Vischer, M., et al. "Spatially Resolved Rainfall Streamflow Modeling in Central Europe." EGUsphere 2025 (2025): 1-26.Yang, Y., et al. (2025). Global daily discharge estimation based on grid long short-term memory (LSTM) model and river routing. Water Resources Research, 61.