Deep learning rainfall–runoff predictions of extreme events

Frame, Jonathan; Kratzert, Frederik; Klotz, Daniel; Gauch, Martin; Shelev, Guy; Gilon, Oren; Qualls, Logan M.; Gupta, Hoshin; Nearing, Grey

doi:10.5194/hess-26-3377-2022

Cited by 129 publications

(82 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning methods provide great versatility (Shen, 2018;Shen et al, 2018;Reichstein et al, 2019) and have demonstrated unprecedented accuracy in various modelling tasks like predictions in ungauged basins (PUB; e.g. Kratzert et al, 2019b;Prieto et al, 2019), transfer learning to data-scarce regions (Ma et al, 2021) or flood forecasting (Frame et al, 2022;Nevo et al, 2022). Nonetheless, deep learning remains a field of progress with gaps to fill.…”

Section: Machine Learning In Hydrologymentioning

confidence: 99%

“…Yet, internal investigation of machine-learning models relies on additional methods that come with their own assumptions and caveats, and the current straightforward interpretability of conceptual models serves as a benchmark in the hydrologic community. Much environmental research is dedicated toward extrapolation in space and in time and of boundary conditions, in order to investigate extreme events (Frame et al, 2022), climate change projections (Nearing et al, 2019) and so on. In all these fields, ease of interpretability is desirable.…”

Section: Machine Learning In Hydrologymentioning

confidence: 99%

See 1 more Smart Citation

Improving hydrologic models for predictions and process understanding using neural ODEs

Höge

Scheidegger

Baity‐Jesi

et al. 2022

Hydrol. Earth Syst. Sci.

View full text Add to dashboard Cite

Abstract. Deep learning methods have frequently outperformed conceptual hydrologic models in rainfall-runoff modelling. Attempts of investigating such deep learning models internally are being made, but the traceability of model states and processes and their interrelations to model input and output is not yet fully understood. Direct interpretability of mechanistic processes has always been considered an asset of conceptual models that helps to gain system understanding aside of predictability. We introduce hydrologic neural ordinary differential equation (ODE) models that perform as well as state-of-the-art deep learning methods in stream flow prediction while maintaining the ease of interpretability of conceptual hydrologic models. In neural ODEs, internal processes that are represented in differential equations, are substituted by neural networks. Therefore, neural ODE models enable the fusion of deep learning with mechanistic modelling. We demonstrate the basin-specific predictive performance for 569 catchments of the continental United States. For exemplary basins, we analyse the dynamics of states and processes learned by the model-internal neural networks. Finally, we discuss the potential of neural ODE models in hydrology.

show abstract

Section: Machine Learning In Hydrologymentioning

confidence: 99%

Section: Machine Learning In Hydrologymentioning

confidence: 99%

Improving hydrologic models for predictions and process understanding using neural ODEs

Höge

Scheidegger

Baity‐Jesi

et al. 2022

Hydrol. Earth Syst. Sci.

View full text Add to dashboard Cite

show abstract

“…ML methods are an integral and critical part of the river stage forecasting and flood inundation models. For stage forecasts, it has already been shown that LSTMs outperform standard hydrological models, not just for gauged basins but also for ungauged basins (Kratzert et al, 2019a, b) and even for extreme out-of-training events (Frame et al, 2022). For inundation modeling, a comparison of the two ML-based inundation models with the physics-based hydraulic model found the ML models exhibited higher accuracy (Table 1).…”

Section: Discussionmentioning

confidence: 92%

Flood forecasting with machine learning models in an operational framework

Nevo

Morin

Rosenthal

et al. 2022

Hydrol. Earth Syst. Sci.

Self Cite

View full text Add to dashboard Cite

Abstract. Google's operational flood forecasting system was developed to provide accurate real-time flood warnings to agencies and the public with a focus on riverine floods in large, gauged rivers. It became operational in 2018 and has since expanded geographically. This forecasting system consists of four subsystems: data validation, stage forecasting, inundation modeling, and alert distribution. Machine learning is used for two of the subsystems. Stage forecasting is modeled with the long short-term memory (LSTM) networks and the linear models. Flood inundation is computed with the thresholding and the manifold models, where the former computes inundation extent and the latter computes both inundation extent and depth. The manifold model, presented here for the first time, provides a machine-learning alternative to hydraulic modeling of flood inundation. When evaluated on historical data, all models achieve sufficiently high-performance metrics for operational use. The LSTM showed higher skills than the linear model, while the thresholding and manifold models achieved similar performance metrics for modeling inundation extent. During the 2021 monsoon season, the flood warning system was operational in India and Bangladesh, covering flood-prone regions around rivers with a total area close to 470 000 km2, home to more than 350 000 000 people. More than 100 000 000 flood alerts were sent to affected populations, to relevant authorities, and to emergency organizations. Current and future work on the system includes extending coverage to additional flood-prone locations and improving modeling capabilities and accuracy.

show abstract

“…This has yet to be applied to a large-sample LSTMs in regionalization, but it is possible that some research will elucidate this in the near future. Furthermore, some studies have started investigating the possibility of adding physical constraints within the LSTM structure (such as ensuring mass-balance) (Frame et al, 2022;Hoedt et al, 2021), which might pave the way to a better understanding of the underlying relationships built within the LSTM structure. Another limitation is the need for long observation data time series to adequately train the LSTM models.…”

Section: Comparison Of Hydrological Model-based and Lstm Regionalizationmentioning

confidence: 99%

“…On half their test catchments, the RNNs performed better than GR4H after training on a 14-year window, and for the other catchments the RNNs still performed acceptably well. With respect to the extrapolation to extreme events problem, Frame et al (2022) showed that LSTM models are relatively accurate at producing high flows when compared to SAC-SMA and a process-based model (US National Water Model), even when extreme events were excluded from the training. This suggests that LSTM models are able to not only extract relevant hydrological information from the training dataset but actually learn from it.…”

mentioning

confidence: 99%

Continuous streamflow prediction in ungauged basins: Long Short-Term Memory Neural Networks clearly outperform hydrological models

Arsenault¹,

Martel²,

Brunet³

et al. 2022

Preprint

View full text Add to dashboard Cite

Abstract. This study investigates the ability of Long Short-Term Memory (LSTM) neural networks to perform streamflow prediction at ungauged basins. A series of state-of-the-art, hydrological model-dependent regionalization methods is applied to 148 catchments in Northeast North America and compared to a LSTM model that uses the exact same available data as the hydrological models. While conceptual model-based methods attempt to derive parameterizations at ungauged sites from other similar or nearby catchments, the LSTM model uses all available data in the region to maximize the information content and increase its robustness. Furthermore, by design, the LSTM does not require explicit definition of hydrological processes and derives its own structure from the provided data. The LSTM networks were able to clearly outperform the hydrological models in a leave-one-out cross-validation regionalization setting on most catchments in the study area, with the LSTM model outperforming the hydrological models in 93 to 97 % of catchments depending on the hydrological model. Furthermore, for up to 78 % of the catchments, the LSTM model was able to predict streamflow more accurately on pseudo-ungauged catchments than hydrological models calibrated on the target data, showing that the LSTM model’s structure was better suited to convert the meteorological data and geophysical descriptors into streamflow than the hydrological models even calibrated to those sites in these cases. Furthermore, the LSTM model robustness was tested by varying its hyperparameters, and still outperformed hydrological models in regionalization in almost all cases. Overall, LSTM networks have the potential to change the regionalization research landscape by providing clear improvement pathways over traditional methods in the field of streamflow prediction in ungauged catchments.

show abstract

Deep learning rainfall–runoff predictions of extreme events

Cited by 129 publications

References 37 publications

Improving hydrologic models for predictions and process understanding using neural ODEs

Improving hydrologic models for predictions and process understanding using neural ODEs

Flood forecasting with machine learning models in an operational framework

Continuous streamflow prediction in ungauged basins: Long Short-Term Memory Neural Networks clearly outperform hydrological models

Contact Info

Product

Resources

About