Articles | Volume 26, issue 6
https://doi.org/10.5194/hess-26-1673-2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/hess-26-1673-2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Research article
| Highlight paper
| 
31 Mar 2022
Research article | Highlight paper |
| 31 Mar 2022
| 31 Mar 2022
Uncertainty estimation with deep learning for rainfall–runoff modeling
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Frederik Kratzert
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Martin Gauch
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Alden Keefe Sampson
Upstream Tech, Natel Energy Inc., Alameda, CA, USA
Johannes Brandstetter
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Günter Klambauer
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Sepp Hochreiter
Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
Grey Nearing
Google Research, Mountain View, CA, USA
Related authors
Frederik Kratzert, Martin Gauch, Daniel Klotz, and Grey Nearing
Hydrol. Earth Syst. Sci., 28, 4187–4201, https://doi.org/10.5194/hess-28-4187-2024, https://doi.org/10.5194/hess-28-4187-2024, 2024
Short summary
Short summary
Recently, a special type of neural-network architecture became increasingly popular in hydrology literature. However, in most applications, this model was applied as a one-to-one replacement for hydrology models without adapting or rethinking the experimental setup. In this opinion paper, we show how this is almost always a bad decision and how using these kinds of models requires the use of large-sample hydrology data sets.
Andreas Auer, Martin Gauch, Frederik Kratzert, Grey Nearing, Sepp Hochreiter, and Daniel Klotz
Hydrol. Earth Syst. Sci., 28, 4099–4126, https://doi.org/10.5194/hess-28-4099-2024, https://doi.org/10.5194/hess-28-4099-2024, 2024
Short summary
Short summary
This work examines the impact of temporal and spatial information on the uncertainty estimation of streamflow forecasts. The study emphasizes the importance of data updates and global information for precise uncertainty estimates. We use conformal prediction to show that recent data enhance the estimates, even if only available infrequently. Local data yield reasonable average estimations but fall short for peak-flow events. The use of global data significantly improves these predictions.
Daniel Klotz, Martin Gauch, Frederik Kratzert, Grey Nearing, and Jakob Zscheischler
Hydrol. Earth Syst. Sci., 28, 3665–3673, https://doi.org/10.5194/hess-28-3665-2024, https://doi.org/10.5194/hess-28-3665-2024, 2024
Short summary
Short summary
The evaluation of model performance is essential for hydrological modeling. Using performance criteria requires a deep understanding of their properties. We focus on a counterintuitive aspect of the Nash–Sutcliffe efficiency (NSE) and show that if we divide the data into multiple parts, the overall performance can be higher than all the evaluations of the subsets. Although this follows from the definition of the NSE, the resulting behavior can have unintended consequences in practice.
Eduardo Acuna Espinoza, Ralf Loritz, Frederik Kratzert, Daniel Klotz, Martin Gauch, Manuel Álvarez Chaves, Nicole Bäuerle, and Uwe Ehret
EGUsphere, https://doi.org/10.5194/egusphere-2024-2147, https://doi.org/10.5194/egusphere-2024-2147, 2024
Short summary
Short summary
Data-driven techniques have shown the potential to outperform process-based models in rainfall-runoff simulations. Hybrid models, combining both approaches, aim to enhance accuracy and maintain interpretability. Expanding the set of test cases to evaluate hybrid models under different conditions we test their generalization capabilities for extreme hydrological events.
Grey S. Nearing, Daniel Klotz, Jonathan M. Frame, Martin Gauch, Oren Gilon, Frederik Kratzert, Alden Keefe Sampson, Guy Shalev, and Sella Nevo
Hydrol. Earth Syst. Sci., 26, 5493–5513, https://doi.org/10.5194/hess-26-5493-2022, https://doi.org/10.5194/hess-26-5493-2022, 2022
Short summary
Short summary
When designing flood forecasting models, it is necessary to use all available data to achieve the most accurate predictions possible. This manuscript explores two basic ways of ingesting near-real-time streamflow data into machine learning streamflow models. The point we want to make is that when working in the context of machine learning (instead of traditional hydrology models that are based on
bio-geophysics), it is not necessary to use complex statistical methods for injecting sparse data.
Juliane Mai, Hongren Shen, Bryan A. 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
Hydrol. Earth Syst. Sci., 26, 3537–3572, https://doi.org/10.5194/hess-26-3537-2022, https://doi.org/10.5194/hess-26-3537-2022, 2022
Short summary
Short summary
Model intercomparison studies are carried out to test various models and compare the quality of their outputs over the same domain. In this study, 13 diverse model setups using the same input data are evaluated over the Great Lakes region. Various model outputs – such as streamflow, evaporation, soil moisture, and amount of snow on the ground – are compared using standardized methods and metrics. The basin-wise model outputs and observations are made available through an interactive website.
Thomas Lees, Steven Reece, Frederik Kratzert, Daniel Klotz, Martin Gauch, Jens De Bruijn, Reetik Kumar Sahu, Peter Greve, Louise Slater, and Simon J. Dadson
Hydrol. Earth Syst. Sci., 26, 3079–3101, https://doi.org/10.5194/hess-26-3079-2022, https://doi.org/10.5194/hess-26-3079-2022, 2022
Short summary
Short summary
Despite the accuracy of deep learning rainfall-runoff models, we are currently uncertain of what these models have learned. In this study we explore the internals of one deep learning architecture and demonstrate that the model learns about intermediate hydrological stores of soil moisture and snow water, despite never having seen data about these processes during training. Therefore, we find evidence that the deep learning approach learns a physically realistic mapping from inputs to outputs.
Frederik Kratzert, Daniel Klotz, Sepp Hochreiter, and Grey S. Nearing
Hydrol. Earth Syst. Sci., 25, 2685–2703, https://doi.org/10.5194/hess-25-2685-2021, https://doi.org/10.5194/hess-25-2685-2021, 2021
Short summary
Short summary
We investigate how deep learning models use different meteorological data sets in the task of (regional) rainfall–runoff modeling. We show that performance can be significantly improved when using different data products as input and further show how the model learns to combine those meteorological input differently across time and space. The results are carefully benchmarked against classical approaches, showing the supremacy of the presented approach.
Martin Gauch, Frederik Kratzert, Daniel Klotz, Grey Nearing, Jimmy Lin, and Sepp Hochreiter
Hydrol. Earth Syst. Sci., 25, 2045–2062, https://doi.org/10.5194/hess-25-2045-2021, https://doi.org/10.5194/hess-25-2045-2021, 2021
Short summary
Short summary
We present multi-timescale Short-Term Memory (MTS-LSTM), a machine learning approach that predicts discharge at multiple timescales within one model. MTS-LSTM is significantly more accurate than the US National Water Model and computationally more efficient than an individual LSTM model per timescale. Further, MTS-LSTM can process different input variables at different timescales, which is important as the lead time of meteorological forecasts often depends on their temporal resolution.
Frederik Kratzert, Daniel Klotz, Guy Shalev, Günter Klambauer, Sepp Hochreiter, and Grey Nearing
Hydrol. Earth Syst. Sci., 23, 5089–5110, https://doi.org/10.5194/hess-23-5089-2019, https://doi.org/10.5194/hess-23-5089-2019, 2019
Short summary
Short summary
A new approach for regional rainfall–runoff modeling using long short-term memory (LSTM)-based models is presented and benchmarked against a range of well-known hydrological models. The approach significantly outperforms regionally calibrated hydrological models but also basin-wise calibrated models. Furthermore, we propose an adaption of the LSTM that allows us to extract the learned catchment understanding of the model and show that it matches our hydrology expert knowledge.
Frederik Kratzert, Daniel Klotz, Claire Brenner, Karsten Schulz, and Mathew Herrnegger
Hydrol. Earth Syst. Sci., 22, 6005–6022, https://doi.org/10.5194/hess-22-6005-2018, https://doi.org/10.5194/hess-22-6005-2018, 2018
Short summary
Short summary
In this paper, we propose a novel data-driven approach for
rainfall–runoff modelling, using the long short-term memory (LSTM) network, a special type of recurrent neural network. We show in three different experiments that this network is able to learn to predict the discharge purely from meteorological input parameters (such as precipitation or temperature) as accurately as (or better than) the well-established Sacramento Soil Moisture Accounting model, coupled with the Snow-17 snow model.
Karsten Schulz, Reinhard Burgholzer, Daniel Klotz, Johannes Wesemann, and Mathew Herrnegger
Hydrol. Earth Syst. Sci., 22, 2607–2613, https://doi.org/10.5194/hess-22-2607-2018, https://doi.org/10.5194/hess-22-2607-2018, 2018
Short summary
Short summary
The unit hydrograph has been one of the most widely employed modelling techniques to predict rainfall-runoff behaviour of hydrological catchments. We developed a lecture theatre experiment including some student involvement to illustrate the principles behind this modelling technique. The experiment only uses very simple and cheap material including a set of plastic balls (representing rainfall), magnetic stripes (tacking the balls to the white board) and sieves (for ball/water gauging).
Claudia Färber, Henning Plessow, Simon Mischel, Frederik Kratzert, Nans Addor, Guy Shalev, and Ulrich Looser
Earth Syst. Sci. Data Discuss., https://doi.org/10.5194/essd-2024-427, https://doi.org/10.5194/essd-2024-427, 2024
Preprint under review for ESSD
Short summary
Short summary
Large-sample datasets are essential in hydrological science to support modelling studies and advance process understanding. Caravan is a community initiative to create a large-sample hydrology dataset of meteorological forcing data, catchment attributes, and discharge data for catchments around the world. This dataset is a subset of hydrological discharge data and station-based watersheds from the Global Runoff Data Centre (GRDC), which are covered by an open data policy.
Frederik Kratzert, Martin Gauch, Daniel Klotz, and Grey Nearing
Hydrol. Earth Syst. Sci., 28, 4187–4201, https://doi.org/10.5194/hess-28-4187-2024, https://doi.org/10.5194/hess-28-4187-2024, 2024
Short summary
Short summary
Recently, a special type of neural-network architecture became increasingly popular in hydrology literature. However, in most applications, this model was applied as a one-to-one replacement for hydrology models without adapting or rethinking the experimental setup. In this opinion paper, we show how this is almost always a bad decision and how using these kinds of models requires the use of large-sample hydrology data sets.
Andreas Auer, Martin Gauch, Frederik Kratzert, Grey Nearing, Sepp Hochreiter, and Daniel Klotz
Hydrol. Earth Syst. Sci., 28, 4099–4126, https://doi.org/10.5194/hess-28-4099-2024, https://doi.org/10.5194/hess-28-4099-2024, 2024
Short summary
Short summary
This work examines the impact of temporal and spatial information on the uncertainty estimation of streamflow forecasts. The study emphasizes the importance of data updates and global information for precise uncertainty estimates. We use conformal prediction to show that recent data enhance the estimates, even if only available infrequently. Local data yield reasonable average estimations but fall short for peak-flow events. The use of global data significantly improves these predictions.
Daniel Klotz, Martin Gauch, Frederik Kratzert, Grey Nearing, and Jakob Zscheischler
Hydrol. Earth Syst. Sci., 28, 3665–3673, https://doi.org/10.5194/hess-28-3665-2024, https://doi.org/10.5194/hess-28-3665-2024, 2024
Short summary
Short summary
The evaluation of model performance is essential for hydrological modeling. Using performance criteria requires a deep understanding of their properties. We focus on a counterintuitive aspect of the Nash–Sutcliffe efficiency (NSE) and show that if we divide the data into multiple parts, the overall performance can be higher than all the evaluations of the subsets. Although this follows from the definition of the NSE, the resulting behavior can have unintended consequences in practice.
Eduardo Acuna Espinoza, Ralf Loritz, Frederik Kratzert, Daniel Klotz, Martin Gauch, Manuel Álvarez Chaves, Nicole Bäuerle, and Uwe Ehret
EGUsphere, https://doi.org/10.5194/egusphere-2024-2147, https://doi.org/10.5194/egusphere-2024-2147, 2024
Short summary
Short summary
Data-driven techniques have shown the potential to outperform process-based models in rainfall-runoff simulations. Hybrid models, combining both approaches, aim to enhance accuracy and maintain interpretability. Expanding the set of test cases to evaluate hybrid models under different conditions we test their generalization capabilities for extreme hydrological events.
Gab Abramowitz, Anna Ukkola, Sanaa Hobeichi, Jon Cranko Page, Mathew Lipson, Martin De Kauwe, Sam Green, Claire Brenner, Jonathan Frame, Grey Nearing, Martyn Clark, Martin Best, Peter Anthoni, Gabriele Arduini, Souhail Boussetta, Silvia Caldararu, Kyeungwoo Cho, Matthias Cuntz, David Fairbairn, Craig Ferguson, Hyungjun Kim, Yeonjoo Kim, Jürgen Knauer, David Lawrence, Xiangzhong Luo, Sergey Malyshev, Tomoko Nitta, Jerome Ogee, Keith Oleson, Catherine Ottlé, Phillipe Peylin, Patricia de Rosnay, Heather Rumbold, Bob Su, Nicolas Vuichard, Anthony Walker, Xiaoni Wang-Faivre, Yunfei Wang, and Yijian Zeng
EGUsphere, https://doi.org/10.5194/egusphere-2023-3084, https://doi.org/10.5194/egusphere-2023-3084, 2024
Short summary
Short summary
This paper evaluates land models – computer based models that simulate ecosystem dynamics, the land carbon, water and energy cycles and the role of land in the climate system. It uses machine learning / AI approaches to show that despite the complexity of land models, they do not perform nearly as well as they could, given the amount of information they are provided with about the prediction problem.
Louise J. Slater, Louise Arnal, Marie-Amélie Boucher, Annie Y.-Y. Chang, Simon Moulds, Conor Murphy, Grey Nearing, Guy Shalev, Chaopeng Shen, Linda Speight, Gabriele Villarini, Robert L. Wilby, Andrew Wood, and Massimiliano Zappa
Hydrol. Earth Syst. Sci., 27, 1865–1889, https://doi.org/10.5194/hess-27-1865-2023, https://doi.org/10.5194/hess-27-1865-2023, 2023
Short summary
Short summary
Hybrid forecasting systems combine data-driven methods with physics-based weather and climate models to improve the accuracy of predictions for meteorological and hydroclimatic events such as rainfall, temperature, streamflow, floods, droughts, tropical cyclones, or atmospheric rivers. We review recent developments in hybrid forecasting and outline key challenges and opportunities in the field.
Grey S. Nearing, Daniel Klotz, Jonathan M. Frame, Martin Gauch, Oren Gilon, Frederik Kratzert, Alden Keefe Sampson, Guy Shalev, and Sella Nevo
Hydrol. Earth Syst. Sci., 26, 5493–5513, https://doi.org/10.5194/hess-26-5493-2022, https://doi.org/10.5194/hess-26-5493-2022, 2022
Short summary
Short summary
When designing flood forecasting models, it is necessary to use all available data to achieve the most accurate predictions possible. This manuscript explores two basic ways of ingesting near-real-time streamflow data into machine learning streamflow models. The point we want to make is that when working in the context of machine learning (instead of traditional hydrology models that are based on
bio-geophysics), it is not necessary to use complex statistical methods for injecting sparse data.
Sella Nevo, Efrat Morin, Adi Gerzi Rosenthal, Asher Metzger, Chen Barshai, Dana Weitzner, Dafi Voloshin, Frederik Kratzert, Gal Elidan, Gideon Dror, Gregory Begelman, Grey Nearing, Guy Shalev, Hila Noga, Ira Shavitt, Liora Yuklea, Moriah Royz, Niv Giladi, Nofar Peled Levi, Ofir Reich, Oren Gilon, Ronnie Maor, Shahar Timnat, Tal Shechter, Vladimir Anisimov, Yotam Gigi, Yuval Levin, Zach Moshe, Zvika Ben-Haim, Avinatan Hassidim, and Yossi Matias
Hydrol. Earth Syst. Sci., 26, 4013–4032, https://doi.org/10.5194/hess-26-4013-2022, https://doi.org/10.5194/hess-26-4013-2022, 2022
Short summary
Short summary
Early flood warnings are one of the most effective tools to save lives and goods. Machine learning (ML) models can improve flood prediction accuracy but their use in operational frameworks is limited. The paper presents a flood warning system, operational in India and Bangladesh, that uses ML models for forecasting river stage and flood inundation maps and discusses the models' performances. In 2021, more than 100 million flood alerts were sent to people near rivers over an area of 470 000 km2.
Juliane Mai, Hongren Shen, Bryan A. 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
Hydrol. Earth Syst. Sci., 26, 3537–3572, https://doi.org/10.5194/hess-26-3537-2022, https://doi.org/10.5194/hess-26-3537-2022, 2022
Short summary
Short summary
Model intercomparison studies are carried out to test various models and compare the quality of their outputs over the same domain. In this study, 13 diverse model setups using the same input data are evaluated over the Great Lakes region. Various model outputs – such as streamflow, evaporation, soil moisture, and amount of snow on the ground – are compared using standardized methods and metrics. The basin-wise model outputs and observations are made available through an interactive website.
Jonathan M. Frame, Frederik Kratzert, Daniel Klotz, Martin Gauch, Guy Shalev, Oren Gilon, Logan M. Qualls, Hoshin V. Gupta, and Grey S. Nearing
Hydrol. Earth Syst. Sci., 26, 3377–3392, https://doi.org/10.5194/hess-26-3377-2022, https://doi.org/10.5194/hess-26-3377-2022, 2022
Short summary
Short summary
The most accurate rainfall–runoff predictions are currently based on deep learning. There is a concern among hydrologists that deep learning models may not be reliable in extrapolation or for predicting extreme events. This study tests that hypothesis. The deep learning models remained relatively accurate in predicting extreme events compared with traditional models, even when extreme events were not included in the training set.
Thomas Lees, Steven Reece, Frederik Kratzert, Daniel Klotz, Martin Gauch, Jens De Bruijn, Reetik Kumar Sahu, Peter Greve, Louise Slater, and Simon J. Dadson
Hydrol. Earth Syst. Sci., 26, 3079–3101, https://doi.org/10.5194/hess-26-3079-2022, https://doi.org/10.5194/hess-26-3079-2022, 2022
Short summary
Short summary
Despite the accuracy of deep learning rainfall-runoff models, we are currently uncertain of what these models have learned. In this study we explore the internals of one deep learning architecture and demonstrate that the model learns about intermediate hydrological stores of soil moisture and snow water, despite never having seen data about these processes during training. Therefore, we find evidence that the deep learning approach learns a physically realistic mapping from inputs to outputs.
Frederik Kratzert, Daniel Klotz, Sepp Hochreiter, and Grey S. Nearing
Hydrol. Earth Syst. Sci., 25, 2685–2703, https://doi.org/10.5194/hess-25-2685-2021, https://doi.org/10.5194/hess-25-2685-2021, 2021
Short summary
Short summary
We investigate how deep learning models use different meteorological data sets in the task of (regional) rainfall–runoff modeling. We show that performance can be significantly improved when using different data products as input and further show how the model learns to combine those meteorological input differently across time and space. The results are carefully benchmarked against classical approaches, showing the supremacy of the presented approach.
Martin Gauch, Frederik Kratzert, Daniel Klotz, Grey Nearing, Jimmy Lin, and Sepp Hochreiter
Hydrol. Earth Syst. Sci., 25, 2045–2062, https://doi.org/10.5194/hess-25-2045-2021, https://doi.org/10.5194/hess-25-2045-2021, 2021
Short summary
Short summary
We present multi-timescale Short-Term Memory (MTS-LSTM), a machine learning approach that predicts discharge at multiple timescales within one model. MTS-LSTM is significantly more accurate than the US National Water Model and computationally more efficient than an individual LSTM model per timescale. Further, MTS-LSTM can process different input variables at different timescales, which is important as the lead time of meteorological forecasts often depends on their temporal resolution.
Frederik Kratzert, Daniel Klotz, Guy Shalev, Günter Klambauer, Sepp Hochreiter, and Grey Nearing
Hydrol. Earth Syst. Sci., 23, 5089–5110, https://doi.org/10.5194/hess-23-5089-2019, https://doi.org/10.5194/hess-23-5089-2019, 2019
Short summary
Short summary
A new approach for regional rainfall–runoff modeling using long short-term memory (LSTM)-based models is presented and benchmarked against a range of well-known hydrological models. The approach significantly outperforms regionally calibrated hydrological models but also basin-wise calibrated models. Furthermore, we propose an adaption of the LSTM that allows us to extract the learned catchment understanding of the model and show that it matches our hydrology expert knowledge.
Frederik Kratzert, Daniel Klotz, Claire Brenner, Karsten Schulz, and Mathew Herrnegger
Hydrol. Earth Syst. Sci., 22, 6005–6022, https://doi.org/10.5194/hess-22-6005-2018, https://doi.org/10.5194/hess-22-6005-2018, 2018
Short summary
Short summary
In this paper, we propose a novel data-driven approach for
rainfall–runoff modelling, using the long short-term memory (LSTM) network, a special type of recurrent neural network. We show in three different experiments that this network is able to learn to predict the discharge purely from meteorological input parameters (such as precipitation or temperature) as accurately as (or better than) the well-established Sacramento Soil Moisture Accounting model, coupled with the Snow-17 snow model.
Karsten Schulz, Reinhard Burgholzer, Daniel Klotz, Johannes Wesemann, and Mathew Herrnegger
Hydrol. Earth Syst. Sci., 22, 2607–2613, https://doi.org/10.5194/hess-22-2607-2018, https://doi.org/10.5194/hess-22-2607-2018, 2018
Short summary
Short summary
The unit hydrograph has been one of the most widely employed modelling techniques to predict rainfall-runoff behaviour of hydrological catchments. We developed a lecture theatre experiment including some student involvement to illustrate the principles behind this modelling technique. The experiment only uses very simple and cheap material including a set of plastic balls (representing rainfall), magnetic stripes (tacking the balls to the white board) and sieves (for ball/water gauging).
Related subject area
Subject: Catchment hydrology | Techniques and Approaches: Modelling approaches
On the use of streamflow transformations for hydrological model calibration
Simulation-based inference for parameter estimation of complex watershed simulators
Multi-scale soil moisture data and process-based modeling reveal the importance of lateral groundwater flow in a subarctic catchment
Catchment response to climatic variability: implications for root zone storage and streamflow predictions
Hybrid hydrological modeling for large alpine basins: a semi-distributed approach
Karst aquifer discharge response to rainfall interpreted as anomalous transport
HESS Opinions: Never train a Long Short-Term Memory (LSTM) network on a single basin
Large-sample hydrology – a few camels or a whole caravan?
Comment on “Are soils overrated in hydrology?” by Gao et al. (2023)
Multi-decadal fluctuations in root zone storage capacity through vegetation adaptation to hydro-climatic variability have minor effects on the hydrological response in the Neckar River basin, Germany
Projected future changes in the cryosphere and hydrology of a mountainous catchment in the upper Heihe River, China
On the importance of plant phenology in the evaporative process of a semi-arid woodland: could it be why satellite-based evaporation estimates in the miombo differ?
Regionalization of GR4J model parameters for river flow prediction in Paraná, Brazil
Evolution of river regimes in the Mekong River basin over 8 decades and the role of dams in recent hydrological extremes
Skill of seasonal flow forecasts at catchment scale: an assessment across South Korea
To what extent do flood-inducing storm events change future flood hazards?
When ancient numerical demons meet physics-informed machine learning: adjoint-based gradients for implicit differentiable modeling
Assessing the impact of climate change on high return levels of peak flows in Bavaria applying the CRCM5 large ensemble
Impacts of climate and land surface change on catchment evapotranspiration and runoff from 1951 to 2020 in Saxony, Germany
Quantifying and reducing flood forecast uncertainty by the CHUP-BMA method
Developing a tile drainage module for the Cold Regions Hydrological Model: lessons from a farm in southern Ontario, Canada
To bucket or not to bucket? Analyzing the performance and interpretability of hybrid hydrological models with dynamic parameterization
Widespread flooding dynamics under climate change: characterising floods using grid-based hydrological modelling and regional climate projections
HESS Opinions: The sword of Damocles of the impossible flood
Metamorphic testing of machine learning and conceptual hydrologic models
The influence of human activities on streamflow reductions during the megadrought in central Chile
Elevational control of isotopic composition and application in understanding hydrologic processes in the mid Merced River catchment, Sierra Nevada, California, USA
Lack of robustness of hydrological models: A large-sample diagnosis and an attempt to identify the hydrological and climatic drivers
The Significance of the Leaf-Area-Index on the Evapotranspiration Estimation in SWAT-T for Characteristic Land Cover Types of Western Africa
Enhancing long short-term memory (LSTM)-based streamflow prediction with a spatially distributed approach
Broadleaf afforestation impacts on terrestrial hydrology insignificant compared to climate change in Great Britain
Impacts of spatiotemporal resolutions of precipitation on flood event simulation based on multimodel structures – a case study over the Xiang River basin in China
A network approach for multiscale catchment classification using traits
Multi-model approach in a variable spatial framework for streamflow simulation
Advancing understanding of lake–watershed hydrology: a fully coupled numerical model illustrated by Qinghai Lake
Technical note: Testing the connection between hillslope-scale runoff fluctuations and streamflow hydrographs at the outlet of large river basins
Empirical stream thermal sensitivity cluster on the landscape according to geology and climate
Deep learning for monthly rainfall–runoff modelling: a large-sample comparison with conceptual models across Australia
A large-sample modelling approach towards integrating streamflow and evaporation data for the Spanish catchments
On optimization of calibrations of a distributed hydrological model with spatially distributed information on snow
Toward interpretable LSTM-based modeling of hydrological systems
Flow intermittence prediction using a hybrid hydrological modelling approach: influence of observed intermittence data on the training of a random forest model
What controls the tail behaviour of flood series: rainfall or runoff generation?
Learning Landscape Features from Streamflow with Autoencoders
Seasonal prediction of end-of-dry-season watershed behavior in a highly interconnected alluvial watershed in northern California
Glaciers determine the sensitivity of hydrological processes to perturbed climate in a large mountainous basin on the Tibetan Plateau
Leveraging gauge networks and strategic discharge measurements to aid the development of continuous streamflow records
On the need for physical constraints in deep learning rainfall–runoff projections under climate change: a sensitivity analysis to warming and shifts in potential evapotranspiration
Evaluation of hydrological models on small mountainous catchments: impact of the meteorological forcings
Projecting sediment export from two highly glacierized alpine catchments under climate change: exploring non-parametric regression as an analysis tool
Guillaume Thirel, Léonard Santos, Olivier Delaigue, and Charles Perrin
Hydrol. Earth Syst. Sci., 28, 4837–4860, https://doi.org/10.5194/hess-28-4837-2024, https://doi.org/10.5194/hess-28-4837-2024, 2024
Short summary
Short summary
We discuss how mathematical transformations impact calibrated hydrological model simulations. We assess how 11 transformations behave over the complete range of streamflows. Extreme transformations lead to models that are specialized for extreme streamflows but show poor performance outside the range of targeted streamflows and are less robust. We show that no a priori assumption about transformations can be taken as warranted.
Robert Hull, Elena Leonarduzzi, Luis De La Fuente, Hoang Viet Tran, Andrew Bennett, Peter Melchior, Reed M. Maxwell, and Laura E. Condon
Hydrol. Earth Syst. Sci., 28, 4685–4713, https://doi.org/10.5194/hess-28-4685-2024, https://doi.org/10.5194/hess-28-4685-2024, 2024
Short summary
Short summary
Large-scale hydrologic simulators are a needed tool to explore complex watershed processes and how they may evolve with a changing climate. However, calibrating them can be difficult because they are costly to run and have many unknown parameters. We implement a state-of-the-art approach to model calibration using neural networks with a set of experiments based on streamflow in the upper Colorado River basin.
Jari-Pekka Nousu, Kersti Leppä, Hannu Marttila, Pertti Ala-aho, Giulia Mazzotti, Terhikki Manninen, Mika Korkiakoski, Mika Aurela, Annalea Lohila, and Samuli Launiainen
Hydrol. Earth Syst. Sci., 28, 4643–4666, https://doi.org/10.5194/hess-28-4643-2024, https://doi.org/10.5194/hess-28-4643-2024, 2024
Short summary
Short summary
We used hydrological models, field measurements, and satellite-based data to study the soil moisture dynamics in a subarctic catchment. The role of groundwater was studied with different ways to model the groundwater dynamics and via comparisons to the observational data. The choice of groundwater model was shown to have a strong impact, and representation of lateral flow was important to capture wet soil conditions. Our results provide insights for ecohydrological studies in boreal regions.
Nienke Tempel, Laurène Bouaziz, Riccardo Taormina, Ellis van Noppen, Jasper Stam, Eric Sprokkereef, and Markus Hrachowitz
Hydrol. Earth Syst. Sci., 28, 4577–4597, https://doi.org/10.5194/hess-28-4577-2024, https://doi.org/10.5194/hess-28-4577-2024, 2024
Short summary
Short summary
This study explores the impact of climatic variability on root zone water storage capacities and, thus, on hydrological predictions. Analysing data from 286 areas in Europe and the US, we found that, despite some variations in root zone storage capacity due to changing climatic conditions over multiple decades, these changes are generally minor and have a limited effect on water storage and river flow predictions.
Bu Li, Ting Sun, Fuqiang Tian, Mahmut Tudaji, Li Qin, and Guangheng Ni
Hydrol. Earth Syst. Sci., 28, 4521–4538, https://doi.org/10.5194/hess-28-4521-2024, https://doi.org/10.5194/hess-28-4521-2024, 2024
Short summary
Short summary
This paper developed hybrid semi-distributed hydrological models by employing a process-based model as the backbone and utilizing deep learning to parameterize and replace internal modules. The main contribution is to provide a high-performance tool enriched with explicit hydrological knowledge for hydrological prediction and to improve understanding about the hydrological sensitivities to climate change in large alpine basins.
Dan Elhanati, Nadine Goeppert, and Brian Berkowitz
Hydrol. Earth Syst. Sci., 28, 4239–4249, https://doi.org/10.5194/hess-28-4239-2024, https://doi.org/10.5194/hess-28-4239-2024, 2024
Short summary
Short summary
A continuous time random walk framework was developed to allow modeling of a karst aquifer discharge response to measured rainfall. The application of the numerical model yielded robust fits between modeled and measured discharge values, especially for the distinctive long tails found during recession times. The findings shed light on the interplay of slow and fast flow in the karst system and establish the application of the model for simulating flow and transport in such systems.
Frederik Kratzert, Martin Gauch, Daniel Klotz, and Grey Nearing
Hydrol. Earth Syst. Sci., 28, 4187–4201, https://doi.org/10.5194/hess-28-4187-2024, https://doi.org/10.5194/hess-28-4187-2024, 2024
Short summary
Short summary
Recently, a special type of neural-network architecture became increasingly popular in hydrology literature. However, in most applications, this model was applied as a one-to-one replacement for hydrology models without adapting or rethinking the experimental setup. In this opinion paper, we show how this is almost always a bad decision and how using these kinds of models requires the use of large-sample hydrology data sets.
Franziska Clerc-Schwarzenbach, Giovanni Selleri, Mattia Neri, Elena Toth, Ilja van Meerveld, and Jan Seibert
Hydrol. Earth Syst. Sci., 28, 4219–4237, https://doi.org/10.5194/hess-28-4219-2024, https://doi.org/10.5194/hess-28-4219-2024, 2024
Short summary
Short summary
We show that the differences between the forcing data included in three CAMELS datasets (US, BR, GB) and the forcing data included for the same catchments in the Caravan dataset affect model calibration considerably. The model performance dropped when the data from the Caravan dataset were used instead of the original data. Most of the model performance drop could be attributed to the differences in precipitation data. However, differences were largest for the potential evapotranspiration data.
Ying Zhao, Mehdi Rahmati, Harry Vereecken, and Dani Or
Hydrol. Earth Syst. Sci., 28, 4059–4063, https://doi.org/10.5194/hess-28-4059-2024, https://doi.org/10.5194/hess-28-4059-2024, 2024
Short summary
Short summary
Gao et al. (2023) question the importance of soil in hydrology, sparking debate. We acknowledge some valid points but critique their broad, unsubstantiated views on soil's role. Our response highlights three key areas: (1) the false divide between ecosystem-centric and soil-centric approaches, (2) the vital yet varied impact of soil properties, and (3) the call for a scale-aware framework. We aim to unify these perspectives, enhancing hydrology's comprehensive understanding.
Siyuan Wang, Markus Hrachowitz, and Gerrit Schoups
Hydrol. Earth Syst. Sci., 28, 4011–4033, https://doi.org/10.5194/hess-28-4011-2024, https://doi.org/10.5194/hess-28-4011-2024, 2024
Short summary
Short summary
Root zone storage capacity (Sumax) changes significantly over multiple decades, reflecting vegetation adaptation to climatic variability. However, this temporal evolution of Sumax cannot explain long-term fluctuations in the partitioning of water fluxes as expressed by deviations ΔIE from the parametric Budyko curve over time with different climatic conditions, and it does not have any significant effects on shorter-term hydrological response characteristics of the upper Neckar catchment.
Zehua Chang, Hongkai Gao, Leilei Yong, Kang Wang, Rensheng Chen, Chuntan Han, Otgonbayar Demberel, Batsuren Dorjsuren, Shugui Hou, and Zheng Duan
Hydrol. Earth Syst. Sci., 28, 3897–3917, https://doi.org/10.5194/hess-28-3897-2024, https://doi.org/10.5194/hess-28-3897-2024, 2024
Short summary
Short summary
An integrated cryospheric–hydrologic model, FLEX-Cryo, was developed that considers glaciers, snow cover, and frozen soil and their dynamic impacts on hydrology. We utilized it to simulate future changes in cryosphere and hydrology in the Hulu catchment. Our projections showed the two glaciers will melt completely around 2050, snow cover will reduce, and permafrost will degrade. For hydrology, runoff will decrease after the glacier has melted, and permafrost degradation will increase baseflow.
Henry M. Zimba, Miriam Coenders-Gerrits, Kawawa E. Banda, Petra Hulsman, Nick van de Giesen, Imasiku A. Nyambe, and Hubert H. G. Savenije
Hydrol. Earth Syst. Sci., 28, 3633–3663, https://doi.org/10.5194/hess-28-3633-2024, https://doi.org/10.5194/hess-28-3633-2024, 2024
Short summary
Short summary
The fall and flushing of new leaves in the miombo woodlands co-occur in the dry season before the commencement of seasonal rainfall. The miombo species are also said to have access to soil moisture in deep soils, including groundwater in the dry season. Satellite-based evaporation estimates, temporal trends, and magnitudes differ the most in the dry season, most likely due to inadequate understanding and representation of the highlighted miombo species attributes in simulations.
Louise Akemi Kuana, Arlan Scortegagna Almeida, Emílio Graciliano Ferreira Mercuri, and Steffen Manfred Noe
Hydrol. Earth Syst. Sci., 28, 3367–3390, https://doi.org/10.5194/hess-28-3367-2024, https://doi.org/10.5194/hess-28-3367-2024, 2024
Short summary
Short summary
The authors compared regionalization methods for river flow prediction in 126 catchments from the south of Brazil, a region with humid subtropical and hot temperate climate. The regionalization method based on physiographic–climatic similarity had the best performance for predicting daily and Q95 reference flow. We showed that basins without flow monitoring can have a good approximation of streamflow using machine learning and physiographic–climatic information as inputs.
Huy Dang and Yadu Pokhrel
Hydrol. Earth Syst. Sci., 28, 3347–3365, https://doi.org/10.5194/hess-28-3347-2024, https://doi.org/10.5194/hess-28-3347-2024, 2024
Short summary
Short summary
By examining basin-wide simulations of a river regime over 83 years with and without dams, we present evidence that climate variation was a key driver of hydrologic variabilities in the Mekong River basin (MRB) over the long term; however, dams have largely altered the seasonality of the Mekong’s flow regime and annual flooding patterns in major downstream areas in recent years. These findings could help us rethink the planning of future dams and water resource management in the MRB.
Yongshin Lee, Francesca Pianosi, Andres Peñuela, and Miguel Angel Rico-Ramirez
Hydrol. Earth Syst. Sci., 28, 3261–3279, https://doi.org/10.5194/hess-28-3261-2024, https://doi.org/10.5194/hess-28-3261-2024, 2024
Short summary
Short summary
Following recent advancements in weather prediction technology, we explored how seasonal weather forecasts (1 or more months ahead) could benefit practical water management in South Korea. Our findings highlight that using seasonal weather forecasts for predicting flow patterns 1 to 3 months ahead is effective, especially during dry years. This suggest that seasonal weather forecasts can be helpful in improving the management of water resources.
Mariam Khanam, Giulia Sofia, and Emmanouil N. Anagnostou
Hydrol. Earth Syst. Sci., 28, 3161–3190, https://doi.org/10.5194/hess-28-3161-2024, https://doi.org/10.5194/hess-28-3161-2024, 2024
Short summary
Short summary
Flooding worsens due to climate change, with river dynamics being a key in local flood control. Predicting post-storm geomorphic changes is challenging. Using self-organizing maps and machine learning, this study forecasts post-storm alterations in stage–discharge relationships across 3101 US stream gages. The provided framework can aid in updating hazard assessments by identifying rivers prone to change, integrating channel adjustments into flood hazard assessment.
Yalan Song, Wouter J. M. Knoben, Martyn P. Clark, Dapeng Feng, Kathryn Lawson, Kamlesh Sawadekar, and Chaopeng Shen
Hydrol. Earth Syst. Sci., 28, 3051–3077, https://doi.org/10.5194/hess-28-3051-2024, https://doi.org/10.5194/hess-28-3051-2024, 2024
Short summary
Short summary
Differentiable models (DMs) integrate neural networks and physical equations for accuracy, interpretability, and knowledge discovery. We developed an adjoint-based DM for ordinary differential equations (ODEs) for hydrological modeling, reducing distorted fluxes and physical parameters from errors in models that use explicit and operation-splitting schemes. With a better numerical scheme and improved structure, the adjoint-based DM matches or surpasses long short-term memory (LSTM) performance.
Florian Willkofer, Raul R. Wood, and Ralf Ludwig
Hydrol. Earth Syst. Sci., 28, 2969–2989, https://doi.org/10.5194/hess-28-2969-2024, https://doi.org/10.5194/hess-28-2969-2024, 2024
Short summary
Short summary
Severe flood events pose a threat to riverine areas, yet robust estimates of the dynamics of these events in the future due to climate change are rarely available. Hence, this study uses data from a regional climate model, SMILE, to drive a high-resolution hydrological model for 98 catchments of hydrological Bavaria and exploits the large database to derive robust values for the 100-year flood events. Results indicate an increase in frequency and intensity for most catchments in the future.
Maik Renner and Corina Hauffe
Hydrol. Earth Syst. Sci., 28, 2849–2869, https://doi.org/10.5194/hess-28-2849-2024, https://doi.org/10.5194/hess-28-2849-2024, 2024
Short summary
Short summary
Climate and land surface changes influence the partitioning of water balance components decisively. Their impact is quantified for 71 catchments in Saxony. Germany. Distinct signatures in the joint water and energy budgets are found: (i) past forest dieback caused a decrease in and subsequent recovery of evapotranspiration in the affected regions, and (ii) the recent shift towards higher aridity imposed a large decline in runoff that has not been seen in the observation records before.
Zhen Cui, Shenglian Guo, Hua Chen, Dedi Liu, Yanlai Zhou, and Chong-Yu Xu
Hydrol. Earth Syst. Sci., 28, 2809–2829, https://doi.org/10.5194/hess-28-2809-2024, https://doi.org/10.5194/hess-28-2809-2024, 2024
Short summary
Short summary
Ensemble forecasting facilitates reliable flood forecasting and warning. This study couples the copula-based hydrologic uncertainty processor (CHUP) with Bayesian model averaging (BMA) and proposes the novel CHUP-BMA method of reducing inflow forecasting uncertainty of the Three Gorges Reservoir. The CHUP-BMA avoids the normal distribution assumption in the HUP-BMA and considers the constraint of initial conditions, which can improve the deterministic and probabilistic forecast performance.
Mazda Kompanizare, Diogo Costa, Merrin L. Macrae, John W. Pomeroy, and Richard M. Petrone
Hydrol. Earth Syst. Sci., 28, 2785–2807, https://doi.org/10.5194/hess-28-2785-2024, https://doi.org/10.5194/hess-28-2785-2024, 2024
Short summary
Short summary
A new agricultural tile drainage module was developed in the Cold Region Hydrological Model platform. Tile flow and water levels are simulated by considering the effect of capillary fringe thickness, drainable water and seasonal regional groundwater dynamics. The model was applied to a small well-instrumented farm in southern Ontario, Canada, where there are concerns about the impacts of agricultural drainage into Lake Erie.
Eduardo Acuña Espinoza, Ralf Loritz, Manuel Álvarez Chaves, Nicole Bäuerle, and Uwe Ehret
Hydrol. Earth Syst. Sci., 28, 2705–2719, https://doi.org/10.5194/hess-28-2705-2024, https://doi.org/10.5194/hess-28-2705-2024, 2024
Short summary
Short summary
Hydrological hybrid models promise to merge the performance of deep learning methods with the interpretability of process-based models. One hybrid approach is the dynamic parameterization of conceptual models using long short-term memory (LSTM) networks. We explored this method to evaluate the effect of the flexibility given by LSTMs on the process-based part.
Adam Griffin, Alison L. Kay, Paul Sayers, Victoria Bell, Elizabeth Stewart, and Sam Carr
Hydrol. Earth Syst. Sci., 28, 2635–2650, https://doi.org/10.5194/hess-28-2635-2024, https://doi.org/10.5194/hess-28-2635-2024, 2024
Short summary
Short summary
Widespread flooding is a major problem in the UK and is greatly affected by climate change and land-use change. To look at how widespread flooding changes in the future, climate model data (UKCP18) were used with a hydrological model (Grid-to-Grid) across the UK, and 14 400 events were identified between two time slices: 1980–2010 and 2050–2080. There was a strong increase in the number of winter events in the future time slice and in the peak return periods.
Alberto Montanari, Bruno Merz, and Günter Blöschl
Hydrol. Earth Syst. Sci., 28, 2603–2615, https://doi.org/10.5194/hess-28-2603-2024, https://doi.org/10.5194/hess-28-2603-2024, 2024
Short summary
Short summary
Floods often take communities by surprise, as they are often considered virtually
impossibleyet are an ever-present threat similar to the sword suspended over the head of Damocles in the classical Greek anecdote. We discuss four reasons why extremely large floods carry a risk that is often larger than expected. We provide suggestions for managing the risk of megafloods by calling for a creative exploration of hazard scenarios and communicating the unknown corners of the reality of floods.
Peter Reichert, Kai Ma, Marvin Höge, Fabrizio Fenicia, Marco Baity-Jesi, Dapeng Feng, and Chaopeng Shen
Hydrol. Earth Syst. Sci., 28, 2505–2529, https://doi.org/10.5194/hess-28-2505-2024, https://doi.org/10.5194/hess-28-2505-2024, 2024
Short summary
Short summary
We compared the predicted change in catchment outlet discharge to precipitation and temperature change for conceptual and machine learning hydrological models. We found that machine learning models, despite providing excellent fit and prediction capabilities, can be unreliable regarding the prediction of the effect of temperature change for low-elevation catchments. This indicates the need for caution when applying them for the prediction of the effect of climate change.
Nicolás Álamos, Camila Alvarez-Garreton, Ariel Muñoz, and Álvaro González-Reyes
Hydrol. Earth Syst. Sci., 28, 2483–2503, https://doi.org/10.5194/hess-28-2483-2024, https://doi.org/10.5194/hess-28-2483-2024, 2024
Short summary
Short summary
In this study, we assess the effects of climate and water use on streamflow reductions and drought intensification during the last 3 decades in central Chile. We address this by contrasting streamflow observations with near-natural streamflow simulations. We conclude that while the lack of precipitation dominates streamflow reductions in the megadrought, water uses have not diminished during this time, causing a worsening of the hydrological drought conditions and maladaptation conditions.
Fengjing Liu, Martha H. Conklin, and Glenn D. Shaw
Hydrol. Earth Syst. Sci., 28, 2239–2258, https://doi.org/10.5194/hess-28-2239-2024, https://doi.org/10.5194/hess-28-2239-2024, 2024
Short summary
Short summary
Mountain snowpack has been declining and more precipitation falls as rain than snow. Using stable isotopes, we found flows and flow duration in Yosemite Creek are most sensitive to climate warming due to strong evaporation of waterfalls, potentially lengthening the dry-up period of waterfalls in summer and negatively affecting tourism. Groundwater recharge in Yosemite Valley is primarily from the upper snow–rain transition (2000–2500 m) and very vulnerable to a reduction in the snow–rain ratio.
Léonard Santos, Vazken Andréassian, Torben O. Sonnenborg, Göran Lindström, Alban de Lavenne, Charles Perrin, Lila Collet, and Guillaume Thirel
Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2024-80, https://doi.org/10.5194/hess-2024-80, 2024
Revised manuscript accepted for HESS
Short summary
Short summary
This work aims at investigating how hydrological models can be transferred to a period in which climatic conditions are different to the ones of the period in which it was set up. The RAT method, built to detect dependencies between model error and climatic drivers, was applied to 3 different hydrological models on 352 catchments in Denmark, France and Sweden. Potential issues are detected for a significant number of catchments for the 3 models even though these catchments differ for each model.
Fabian Merk, Timo Schaffhauser, Faizan Anwar, Ye Tuo, Jean-Martial Cohard, and Markus Disse
Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2024-131, https://doi.org/10.5194/hess-2024-131, 2024
Revised manuscript accepted for HESS
Short summary
Short summary
ET is computed from vegetation (plant transpiration) and soil (soil evaporation). In Western Africa, plant transpiration correlates with vegetation growth. Vegetation is often represented with the leaf-area-index (LAI). In this study, we evaluate the importance of LAI for the ET calculation. We take a close look at the LAI-ET interaction and show the relevance to consider both, LAI and ET. Our work contributes to the understanding of the processes of the terrestrial water cycle.
Qiutong Yu, Bryan A. Tolson, Hongren Shen, Ming Han, Juliane Mai, and Jimmy Lin
Hydrol. Earth Syst. Sci., 28, 2107–2122, https://doi.org/10.5194/hess-28-2107-2024, https://doi.org/10.5194/hess-28-2107-2024, 2024
Short summary
Short summary
It is challenging to incorporate input variables' spatial distribution information when implementing long short-term memory (LSTM) models for streamflow prediction. This work presents a novel hybrid modelling approach to predict streamflow while accounting for spatial variability. We evaluated the performance against lumped LSTM predictions in 224 basins across the Great Lakes region in North America. This approach shows promise for predicting streamflow in large, ungauged basin.
Marcus Buechel, Louise Slater, and Simon Dadson
Hydrol. Earth Syst. Sci., 28, 2081–2105, https://doi.org/10.5194/hess-28-2081-2024, https://doi.org/10.5194/hess-28-2081-2024, 2024
Short summary
Short summary
Afforestation has been proposed internationally, but the hydrological implications of such large increases in the spatial extent of woodland are not fully understood. In this study, we use a land surface model to simulate hydrology across Great Britain with realistic afforestation scenarios and potential climate changes. Countrywide afforestation minimally influences hydrology, when compared to climate change, and reduces low streamflow whilst not lowering the highest flows.
Qian Zhu, Xiaodong Qin, Dongyang Zhou, Tiantian Yang, and Xinyi Song
Hydrol. Earth Syst. Sci., 28, 1665–1686, https://doi.org/10.5194/hess-28-1665-2024, https://doi.org/10.5194/hess-28-1665-2024, 2024
Short summary
Short summary
Input data, model and calibration strategy can affect the accuracy of flood event simulation and prediction. Satellite-based precipitation with different spatiotemporal resolutions is an important input source. Data-driven models are sometimes proven to be more accurate than hydrological models. Event-based calibration and conventional strategy are two options adopted for flood simulation. This study targets the three concerns for accurate flood event simulation and prediction.
Fabio Ciulla and Charuleka Varadharajan
Hydrol. Earth Syst. Sci., 28, 1617–1651, https://doi.org/10.5194/hess-28-1617-2024, https://doi.org/10.5194/hess-28-1617-2024, 2024
Short summary
Short summary
We present a new method based on network science for unsupervised classification of large datasets and apply it to classify 9067 US catchments and 274 biophysical traits at multiple scales. We find that our trait-based approach produces catchment classes with distinct streamflow behavior and that spatial patterns emerge amongst pristine and human-impacted catchments. This method can be widely used beyond hydrology to identify patterns, reduce trait redundancy, and select representative sites.
Cyril Thébault, Charles Perrin, Vazken Andréassian, Guillaume Thirel, Sébastien Legrand, and Olivier Delaigue
Hydrol. Earth Syst. Sci., 28, 1539–1566, https://doi.org/10.5194/hess-28-1539-2024, https://doi.org/10.5194/hess-28-1539-2024, 2024
Short summary
Short summary
Streamflow forecasting is useful for many applications, ranging from population safety (e.g. floods) to water resource management (e.g. agriculture or hydropower). To this end, hydrological models must be optimized. However, a model is inherently wrong. This study aims to analyse the contribution of a multi-model approach within a variable spatial framework to improve streamflow simulations. The underlying idea is to take advantage of the strength of each modelling framework tested.
Lele Shu, Xiaodong Li, Yan Chang, Xianhong Meng, Hao Chen, Yuan Qi, Hongwei Wang, Zhaoguo Li, and Shihua Lyu
Hydrol. Earth Syst. Sci., 28, 1477–1491, https://doi.org/10.5194/hess-28-1477-2024, https://doi.org/10.5194/hess-28-1477-2024, 2024
Short summary
Short summary
We developed a new model to better understand how water moves in a lake basin. Our model improves upon previous methods by accurately capturing the complexity of water movement, both on the surface and subsurface. Our model, tested using data from China's Qinghai Lake, accurately replicates complex water movements and identifies contributing factors of the lake's water balance. The findings provide a robust tool for predicting hydrological processes, aiding water resource planning.
Ricardo Mantilla, Morgan Fonley, and Nicolás Velásquez
Hydrol. Earth Syst. Sci., 28, 1373–1382, https://doi.org/10.5194/hess-28-1373-2024, https://doi.org/10.5194/hess-28-1373-2024, 2024
Short summary
Short summary
Hydrologists strive to “Be right for the right reasons” when modeling the hydrologic cycle; however, the datasets available to validate hydrological models are sparse, and in many cases, they comprise streamflow observations at the outlets of large catchments. In this work, we show that matching streamflow observations at the outlet of a large basin is not a reliable indicator of a correct description of the small-scale runoff processes.
Lillian M. McGill, E. Ashley Steel, and Aimee H. Fullerton
Hydrol. Earth Syst. Sci., 28, 1351–1371, https://doi.org/10.5194/hess-28-1351-2024, https://doi.org/10.5194/hess-28-1351-2024, 2024
Short summary
Short summary
This study examines the relationship between air and river temperatures in Washington's Snoqualmie and Wenatchee basins. We used classification and regression approaches to show that the sensitivity of river temperature to air temperature is variable across basins and controlled largely by geology and snowmelt. Findings can be used to inform strategies for river basin restoration and conservation, such as identifying climate-insensitive areas of the basin that should be preserved and protected.
Stephanie R. Clark, Julien Lerat, Jean-Michel Perraud, and Peter Fitch
Hydrol. Earth Syst. Sci., 28, 1191–1213, https://doi.org/10.5194/hess-28-1191-2024, https://doi.org/10.5194/hess-28-1191-2024, 2024
Short summary
Short summary
To determine if deep learning models are in general a viable alternative to traditional hydrologic modelling techniques in Australian catchments, a comparison of river–runoff predictions is made between traditional conceptual models and deep learning models in almost 500 catchments spread over the continent. It is found that the deep learning models match or outperform the traditional models in over two-thirds of the river catchments, indicating feasibility in a wide variety of conditions.
Patricio Yeste, Matilde García-Valdecasas Ojeda, Sonia R. Gámiz-Fortis, Yolanda Castro-Díez, Axel Bronstert, and María Jesús Esteban-Parra
Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2024-57, https://doi.org/10.5194/hess-2024-57, 2024
Revised manuscript accepted for HESS
Short summary
Short summary
Integrating streamflow and evaporation data can help improve the physical realism of hydrologic models. In this work we investigate the capabilities of the Variable Infiltration Capacity (VIC) to reproduce both hydrologic variables for 189 headwater located in Spain. Results from sensitivity analysis indicate that adding two vegetation is enough to improve the representation of evaporation, and the performance of VIC exceeded that of the largest modelling effort currently available in Spain.
Dipti Tiwari, Mélanie Trudel, and Robert Leconte
Hydrol. Earth Syst. Sci., 28, 1127–1146, https://doi.org/10.5194/hess-28-1127-2024, https://doi.org/10.5194/hess-28-1127-2024, 2024
Short summary
Short summary
Calibrating hydrological models with multi-objective functions enhances model robustness. By using spatially distributed snow information in the calibration, the model performance can be enhanced without compromising the outputs. In this study the HYDROTEL model was calibrated in seven different experiments, incorporating the SPAEF (spatial efficiency) metric alongside Nash–Sutcliffe efficiency (NSE) and root-mean-square error (RMSE), with the aim of identifying the optimal calibration strategy.
Luis Andres De la Fuente, Mohammad Reza Ehsani, Hoshin Vijai Gupta, and Laura Elizabeth Condon
Hydrol. Earth Syst. Sci., 28, 945–971, https://doi.org/10.5194/hess-28-945-2024, https://doi.org/10.5194/hess-28-945-2024, 2024
Short summary
Short summary
Long short-term memory (LSTM) is a widely used machine-learning model in hydrology, but it is difficult to extract knowledge from it. We propose HydroLSTM, which represents processes like a hydrological reservoir. Models based on HydroLSTM perform similarly to LSTM while requiring fewer cell states. The learned parameters are informative about the dominant hydrology of a catchment. Our results show how parsimony and hydrological knowledge extraction can be achieved by using the new structure.
Louise Mimeau, Annika Künne, Flora Branger, Sven Kralisch, Alexandre Devers, and Jean-Philippe Vidal
Hydrol. Earth Syst. Sci., 28, 851–871, https://doi.org/10.5194/hess-28-851-2024, https://doi.org/10.5194/hess-28-851-2024, 2024
Short summary
Short summary
Modelling flow intermittence is essential for predicting the future evolution of drying in river networks and better understanding the ecological and socio-economic impacts. However, modelling flow intermittence is challenging, and observed data on temporary rivers are scarce. This study presents a new modelling approach for predicting flow intermittence in river networks and shows that combining different sources of observed data reduces the model uncertainty.
Elena Macdonald, Bruno Merz, Björn Guse, Viet Dung Nguyen, Xiaoxiang Guan, and Sergiy Vorogushyn
Hydrol. Earth Syst. Sci., 28, 833–850, https://doi.org/10.5194/hess-28-833-2024, https://doi.org/10.5194/hess-28-833-2024, 2024
Short summary
Short summary
In some rivers, the occurrence of extreme flood events is more likely than in other rivers – they have heavy-tailed distributions. We find that threshold processes in the runoff generation lead to such a relatively high occurrence probability of extremes. Further, we find that beyond a certain return period, i.e. for rare events, rainfall is often the dominant control compared to runoff generation. Our results can help to improve the estimation of the occurrence probability of extreme floods.
Alberto Bassi, Marvin Höge, Antonietta Mira, Fabrizio Fenicia, and Carlo Albert
Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2024-47, https://doi.org/10.5194/hess-2024-47, 2024
Revised manuscript accepted for HESS
Short summary
Short summary
The goal is to remove the impact of meteorological drivers in order to uncover the unique landscape fingerprints of a catchment from streamflow data. Our results reveal an optimal two-feature summary for most catchments, with a third feature needed for challenging cases, associated with aridity and intermittent flow. Baseflow index, aridity, and soil/vegetation attributes strongly correlate with learned features, indicating their importance for streamflow prediction.
Claire Kouba and Thomas Harter
Hydrol. Earth Syst. Sci., 28, 691–718, https://doi.org/10.5194/hess-28-691-2024, https://doi.org/10.5194/hess-28-691-2024, 2024
Short summary
Short summary
In some watersheds, the severity of the dry season has a large impact on aquatic ecosystems. In this study, we design a way to predict, 5–6 months in advance, how severe the dry season will be in a rural watershed in northern California. This early warning can support seasonal adaptive management. To predict these two values, we assess data about snow, rain, groundwater, and river flows. We find that maximum snowpack and total wet season rainfall best predict dry season severity.
Yi Nan and Fuqiang Tian
Hydrol. Earth Syst. Sci., 28, 669–689, https://doi.org/10.5194/hess-28-669-2024, https://doi.org/10.5194/hess-28-669-2024, 2024
Short summary
Short summary
This paper utilized a tracer-aided model validated by multiple datasets in a large mountainous basin on the Tibetan Plateau to analyze hydrological sensitivity to climate change. The spatial pattern of the local hydrological sensitivities and the influence factors were analyzed in particular. The main finding of this paper is that the local hydrological sensitivity in mountainous basins is determined by the relationship between the glacier area ratio and the mean annual precipitation.
Michael J. Vlah, Matthew R. V. Ross, Spencer Rhea, and Emily S. Bernhardt
Hydrol. Earth Syst. Sci., 28, 545–573, https://doi.org/10.5194/hess-28-545-2024, https://doi.org/10.5194/hess-28-545-2024, 2024
Short summary
Short summary
Virtual stream gauging enables continuous streamflow estimation where a gauge might be difficult or impractical to install. We reconstructed flow at 27 gauges of the National Ecological Observatory Network (NEON), informing ~199 site-months of missing data in the official record and improving that accuracy of official estimates at 11 sites. This study shows that machine learning, but also routine regression methods, can be used to supplement existing gauge networks and reduce monitoring costs.
Sungwook Wi and Scott Steinschneider
Hydrol. Earth Syst. Sci., 28, 479–503, https://doi.org/10.5194/hess-28-479-2024, https://doi.org/10.5194/hess-28-479-2024, 2024
Short summary
Short summary
We investigate whether deep learning (DL) models can produce physically plausible streamflow projections under climate change. We address this question by focusing on modeled responses to increases in temperature and potential evapotranspiration and by employing three DL and three process-based hydrological models. The results suggest that physical constraints regarding model architecture and input are necessary to promote the physical realism of DL hydrological projections under climate change.
Guillaume Evin, Matthieu Le Lay, Catherine Fouchier, David Penot, Francois Colleoni, Alexandre Mas, Pierre-André Garambois, and Olivier Laurantin
Hydrol. Earth Syst. Sci., 28, 261–281, https://doi.org/10.5194/hess-28-261-2024, https://doi.org/10.5194/hess-28-261-2024, 2024
Short summary
Short summary
Hydrological modelling of mountainous catchments is challenging for many reasons, the main one being the temporal and spatial representation of precipitation forcings. This study presents an evaluation of the hydrological modelling of 55 small mountainous catchments of the northern French Alps, focusing on the influence of the type of precipitation reanalyses used as inputs. These evaluations emphasize the added value of radar measurements, in particular for the reproduction of flood events.
Lena Katharina Schmidt, Till Francke, Peter Martin Grosse, and Axel Bronstert
Hydrol. Earth Syst. Sci., 28, 139–161, https://doi.org/10.5194/hess-28-139-2024, https://doi.org/10.5194/hess-28-139-2024, 2024
Short summary
Short summary
How suspended sediment export from glacierized high-alpine areas responds to future climate change is hardly assessable as many interacting processes are involved, and appropriate physical models are lacking. We present the first study, to our knowledge, exploring machine learning to project sediment export until 2100 in two high-alpine catchments. We find that uncertainties due to methodological limitations are small until 2070. Negative trends imply that peak sediment may have already passed.
Cited articles
Abramowitz, G.: Towards a benchmark for land surface models, Geophys. Res. Lett., 32, L22702, https://doi.org/10.1029/2005GL024419, 2005. a
Addor, N., Newman, A. J., Mizukami, N., and Clark, M. P.: The CAMELS data set: catchment attributes and meteorology for large-sample studies, Hydrol. Earth Syst. Sci., 21, 5293–5313, https://doi.org/10.5194/hess-21-5293-2017, 2017. a
Andréassian, V., Perrin, C., Berthet, L., Le Moine, N., Lerat, J., Loumagne, C., Oudin, L., Mathevet, T., Ramos, M.-H., and Valéry, A.: HESS Opinions “Crash tests for a standardized evaluation of hydrological models”, Hydrol. Earth Syst. Sci., 13, 1757–1764, https://doi.org/10.5194/hess-13-1757-2009, 2009. a
Berthet, L., Bourgin, F., Perrin, C., Viatgé, J., Marty, R., and Piotte, O.: A crash-testing framework for predictive uncertainty assessment when forecasting high flows in an extrapolation context, Hydrol. Earth Syst. Sci., 24, 2017–2041, https://doi.org/10.5194/hess-24-2017-2020, 2020. a, b
Best, M. J., Abramowitz, G., Johnson, H., Pitman, A., Balsamo, G., Boone, A.,
Cuntz, M., Decharme, B., Dirmeyer, P., Dong, J., Ek, M., Guo, Z., van den Hurk, B. J. J., Nearing, G. S., Pak, B., Peters-Lidard, C., Santanello Jr., J. A., Stevens, L., and Vuichard, N.: The plumbing of land surface models: benchmarking model performance, J. Hydrometeorol., 16, 1425–1442, 2015. a, b, c
Beven, K.: Facets of uncertainty: epistemic uncertainty, non-stationarity,
likelihood, hypothesis testing, and communication, Hydrolog. Sci. J., 61, 1652–1665, https://doi.org/10.1080/02626667.2015.1031761, 2016. a
Beven, K. and Binley, A.: GLUE: 20 years on, Hydrol. Process., 28, 5897–5918, 2014. a
Beven, K. and Young, P.: A guide to good practice in modeling semantics for
authors and referees, Water Resour. Res., 49, 5092–5098, 2013. a
Beven, K. J., Smith, P. J., and Freer, J. E.: So just why would a modeller
choose to be incoherent?, J. Hydrol., 354, 15–32, 2008. a
Blundell, C., Cornebise, J., Kavukcuoglu, K., and Wierstra, D.: Weight
uncertainty in neural networks, arXiv: preprint, arXiv:1505.05424, 2015. a
Clark, M. P., Wilby, R. L., Gutmann, E. D., Vano, J. A., Gangopadhyay, S.,
Wood, A. W., Fowler, H. J., Prudhomme, C., Arnold, J. R., and Brekke, L. D.:
Characterizing uncertainty of the hydrologic impacts of climate change, Curr. Clim. Change Rep., 2, 55–64, 2016. a
Cole, T.: Too many digits: the presentation of numerical data, Arch. Disease Childhood, 100, 608–609, 2015. a
Dehghani, M., Tay, Y., Gritsenko, A. A., Zhao, Z., Houlsby, N., Diaz, F.,
Metzler, D., and Vinyals, O.: The Benchmark Lottery, https://doi.org/10.48550/arXiv.2107.07002, 2021. a
Demargne, J., Wu, L., Regonda, S. K., Brown, J. D., Lee, H., He, M., Seo,
D.-J., Hartman, R., Herr, H. D., Fresch, M., Schaake, J., and Zhu, Y.: The science of NOAA's operational hydrologic ensemble forecast service, B. Am. Meteorol. Soc., 95, 79–98, 2014. a
Donoho, D.: 50 years of data science, J. Comput. Graph. Stat., 26, 745–766, 2017. a
Ellefsen, K. O., Martin, C. P., and Torresen, J.: How do mixture density rnns
predict the future?, arXiv: preprint, arXiv:1901.07859, 2019. a
Feng, D., Fang, K., and Shen, C.: Enhancing streamflow forecast and extracting insights using long-short term memory networks with data integration at continental scales, Water Resour. Res., 56, e2019WR026793, https://doi.org/10.1029/2019WR026793, 2020. a
Gers, F. A., Schmidhuber, J., and Cummins, F.: Learning to forget: continual
prediction with LSTM, IET Conference Proceedings, 850–855,
https://digital-library.theiet.org/content/conferences/10.1049/cp_19991218 (last access: 31 March 2021), 1999. a
Gneiting, T. and Raftery, A. E.: Strictly proper scoring rules, prediction, and estimation, J. Am. Stat. Assoc., 102, 359–378, 2007. a
Govindaraju, R. S.: Artificial neural networks in hydrology. II: hydrologic applications, J. Hydrol. Eng., 5, 124–137, 2000. a
Graves, A.: Generating sequences with recurrent neural networks, arXiv: preprint, arXiv:1308.0850, 2013. a
Ha, D. and Eck, D.: A neural representation of sketch drawings, arXiv: preprint, arXiv:1704.03477, 2017. a
Ha, D. and Schmidhuber, J.: Recurrent world models facilitate policy evolution, in: Advances in Neural Information Processing Systems, 2450–2462, https://papers.nips.cc/paper/2018/hash/2de5d16682c3c35007e4e92982f1a2ba-Abstract.html (last access: 28 March 2022), 2018. a
Hochreiter, S.: Untersuchungen zu dynamischen neuronalen Netzen, Diploma
thesis, Institut für Informatik, Lehrstuhl Prof. Brauer, Tech. Univ., München, 1991. a
Hochreiter, S. and Schmidhuber, J.: Long Short-Term Memory, Neural Comput., 9, 1735–1780, 1997. a
Kavetski, D., Kuczera, G., and Franks, S. W.: Bayesian analysis of input
uncertainty in hydrological modeling: 2. Application, Water Resour. Res., 42, W03408, https://doi.org/10.1029/2005WR004376, 2006. a
Kendall, A. and Gal, Y.: What uncertainties do we need in bayesian deep
learning for computer vision?, in: Advances in neural information processing
systems, 5574–5584, https://proceedings.neurips.cc/paper/2017/hash/2650d6089a6d640c5e85b2b88265dc2b-Abstract.html (last access: 28 March 2022), 2017. a, b
Klemeš, V.: Operational testing of hydrological simulation models, Hydrolog. Sci. J., 31, 13–24, https://doi.org/10.1080/02626668609491024, 1986. a, b
Klotz, D., Kratzert, F., Herrnegger, M., Hochreiter, S., and Klambauer, G.:
Towards the quantification of uncertainty for deep learning based rainfallrunoff models, Geophys. Res. Abstr., 21, EGU2019-10708-2, 2019. a
Kratzert, F.: neuralhydrology, GitHub [code],
https://github.com/neuralhydrology/neuralhydrology, last access: 21 March 2022. a
Kratzert, F., Klotz, D., Brenner, C., Schulz, K., and Herrnegger, M.: Rainfall–runoff modelling using Long Short-Term Memory (LSTM) networks, Hydrol. Earth Syst. Sci., 22, 6005–6022, https://doi.org/10.5194/hess-22-6005-2018, 2018. a
Kratzert, F., Klotz, D., Herrnegger, M., Sampson, A. K., Hochreiter, S., and
Nearing, G. S.: Toward Improved Predictions in Ungauged Basins: Exploiting
the Power of Machine Learning, Water Resour. Res., 55, 11344–11354, https://doi.org/10.1029/2019WR026065, 2019a. a
Kratzert, F., Klotz, D., Shalev, G., Klambauer, G., Hochreiter, S., and
Nearing, G.: Towards learning universal, regional, and local hydrological
behaviors via machine learning applied to large-sample datasets, Hydrol.
Earth Syst. Sci., 23, 5089–5110, https://doi.org/10.5194/hess-23-5089-2019, 2019b. a, b, c, d, e, f, g, h, i, j, k, l
Laio, F. and Tamea, S.: Verification tools for probabilistic forecasts of continuous hydrological variables, Hydrol. Earth Syst. Sci., 11, 1267–1277, https://doi.org/10.5194/hess-11-1267-2007, 2007. a, b
Lane, R. A., Coxon, G., Freer, J. E., Wagener, T., Johnes, P. J., Bloomfield, J. P., Greene, S., Macleod, C. J. A., and Reaney, S. M.: Benchmarking the predictive capability of hydrological models for river flow and flood peak predictions across over 1000 catchments in Great Britain, Hydrol. Earth Syst. Sci., 23, 4011–4032, https://doi.org/10.5194/hess-23-4011-2019, 2019. a
Li, W., Duan, Q., Miao, C., Ye, A., Gong, W., and Di, Z.: A review on statistical postprocessing methods for hydrometeorological ensemble forecasting, Wiley Interdisciplin. Rev.: Water, 4, e1246, https://doi.org/10.1002/wat2.1246, 2017. a
Liu, M., Huang, Y., Li, Z., Tong, B., Liu, Z., Sun, M., Jiang, F., and Zhang,
H.: The Applicability of LSTM-KNN Model for Real-Time Flood Forecasting in
Different Climate Zones in China, Water, 12, 440, https://doi.org/10.3390/w12020440, 2020. a
Makridakis, S., Spiliotis, E., and Assimakopoulos, V.: The M5 accuracy
competition: Results, findings and conclusions, Int. J. Forecast., https://doi.org/10.1016/j.ijforecast.2021.11.013, 2020. a, b
Montanari, A. and Koutsoyiannis, D.: A blueprint for process-based modeling of uncertain hydrological systems, Water Resour. Res., 48, W09555, https://doi.org/10.1029/2011WR011412, 2012. a
Muller, J. Z.: The tyranny of metrics, Princeton University Press, https://doi.org/10.1515/9780691191263, 2018. a
Naeini, M. P., Cooper, G. F., and Hauskrecht, M.: Obtaining well calibrated
probabilities using bayesian binning, in: Proceedings of the AAAI
Conference on Artificial Intelligence, vol. 2015, NIH Public Access, p. 2901, https://ojs.aaai.org/index.php/AAAI/article/view/9602 (last access: 28 March 2022), 2015. a
Nash, J. E. and Sutcliffe, J. V.: River flow forecasting through conceptual
models part I – A discussion of principles, J. Hydrol., 10, 282–290, 1970. a
NCAR: CAMELS: Catchment Attributes and Meteorology for Large-sample Studies – Dataset Downloads, https://ral.ucar.edu/solutions/products/camels,
last access: 21 March 2022. a
Nearing, G. S., Mocko, D. M., Peters-Lidard, C. D., Kumar, S. V., and Xia, Y.: Benchmarking NLDAS-2 soil moisture and evapotranspiration to separate
uncertainty contributions, J. Hydrometeorol., 17, 745–759, 2016. a
Nearing, G. S., Kratzert, F., Sampson, A. K., Pelissier, C. S., Klotz, D.,
Frame, J. M., Prieto, C., and Gupta, H. V.: What role does hydrological science play in the age of machine learning?, Water Resour. Res., 57, e2020WR028091, https://doi.org/10.1029/2020WR028091, 2020a. a
Nearing, G. S., Ruddell, B. L., Bennett, A. R., Prieto, C., and Gupta, H. V.:
Does Information Theory Provide a New Paradigm for Earth Science? Hypothesis
Testing, Water Resour. Res., 56, e2019WR024918, https://doi.org/10.1029/2019WR024918, 2020b. a, b
Newman, A. J., Clark, M. P., Sampson, K., Wood, A., Hay, L. E., Bock, A., Viger, R. J., Blodgett, D., Brekke, L., Arnold, J. R., Hopson, T., and Duan, Q.: Development of a large-sample watershed-scale hydrometeorological data set for the contiguous USA: data set characteristics and assessment of regional variability in hydrologic model performance, Hydrol. Earth Syst. Sci., 19, 209–223, https://doi.org/10.5194/hess-19-209-2015, 2015. a
Ovadia, Y., Fertig, E., Ren, J., Nado, Z., Sculley, D., Nowozin, S., Dillon,
J., Lakshminarayanan, B., and Snoek, J.: Can you trust your model's
uncertainty? Evaluating predictive uncertainty under dataset shift, in:
Advances in Neural Information Processing Systems, 13991–14002, https://papers.nips.cc/paper/2019/hash/8558cb408c1d76621371888657d2eb1d-Abstract.html (last access: 28 March 2022), 2019. a
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S.: Pytorch: An imperative style, high-performance deep learning library, in: Advances in neural information processing systems, 8026–8037, https://papers.nips.cc/paper/2019/hash/bdbca288fee7f92f2bfa9f7012727740-Abstract.html
(last access: 28 March 2022), 2019. a
Reichle, R. H., McLaughlin, D. B., and Entekhabi, D.: Hydrologic data
assimilation with the ensemble Kalman filter, Mon. Weather Rev., 130, 103–114, 2002. a
Renard, B., Kavetski, D., Kuczera, G., Thyer, M., and Franks, S. W.:
Understanding predictive uncertainty in hydrologic modeling: The challenge of
identifying input and structural errors, Water Resour. Res., 46, W05521, https://doi.org/10.1029/2009WR008328, 2010. a, b
Richmond, K., King, S., and Taylor, P.: Modelling the uncertainty in recovering articulation from acoustics, Comput. Speech Language, 17, 153–172, 2003. a
Rothfuss, J., Ferreira, F., Walther, S., and Ulrich, M.: Conditional density
estimation with neural networks: Best practices and benchmarks, arXiv:
preprint, arXiv:1903.00954, 2019. a
Shrestha, D. L. and Solomatine, D. P.: Data‐driven approaches for estimating
uncertainty in rainfall‐runoff modelling, Int. J. River Basin Manage., 6, 109–122, https://doi.org/10.1080/15715124.2008.9635341, 2008. a
Smith, L. and Gal, Y.: Understanding measures of uncertainty for adversarial
example detection, arXiv: preprint, arXiv:1803.08533, 2018. a
Weijs, S. V., Schoups, G., and van de Giesen, N.: Why hydrological predictions should be evaluated using information theory, Hydrol. Earth Syst. Sci., 14, 2545–2558, https://doi.org/10.5194/hess-14-2545-2010, 2010. a
Yilmaz, K. K., Gupta, H. V., and Wagener, T.: A process-based diagnostic approach to model evaluation: Application to the NWS distributed hydrologic
model, Water Resour. Res., 44, W09417, https://doi.org/10.1029/2007WR006716, 2008. a, b, c
Yu, K. and Moyeed, R. A.: Bayesian quantile regression, Stat. Probabil. Lett., 54, 437–447, 2001. a
Zhu, L. and Laptev, N.: Deep and confident prediction for time series at uber, in: 2017 IEEE International Conference on Data Mining Workshops (ICDMW), 103–110, https://ieeexplore.ieee.org/abstract/document/8215650 (last access: 28 March 2022), 2017. a
Zhu, S., Xu, Z., Luo, X., Liu, X., Wang, R., Zhang, M., and Huo, Z.: Internal
and external coupling of Gaussian mixture model and deep recurrent network
for probabilistic drought forecasting, Int. J. Environ. Sci. Technol., 18, 1221–1236, https://doi.org/10.1007/s13762-020-02862-2, 2020. a
Short summary
This contribution evaluates distributional runoff predictions from deep-learning-based approaches. We propose a benchmarking setup and establish four strong baselines. The results show that accurate, precise, and reliable uncertainty estimation can be achieved with deep learning.
This contribution evaluates distributional runoff predictions from deep-learning-based...
