Evaluating the long short-term memory (LSTM) network for discharge prediction under changing climate conditions

Moura, Carolina Natel de; Seibert, Jan; Detzel, Daniel Henrique Marco

doi:10.2166/nh.2022.044

Cited by 11 publications

(3 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, if the current measured water quality is worse than the historical records, a ML model is likely to perform worse than a physical model. There is evidence that this claim might also hold true for the modern and more sophisticated data-driven models (de Moura et al 2022 ). However, large-sample hydrology (LSH), which is aimed at understanding hydrological processes at multiple spatiotemporal scales and under changing conditions (Addor et al 2020 ) is somewhat challenging this regime.…”

Section: Different Modeling Approachesmentioning

confidence: 91%

Uncertainty in Environmental Micropollutant Modeling

Ahkola,

Kotamäki,

Siivola

et al. 2024

Environmental Management

View full text Add to dashboard Cite

Water pollution policies have been enacted across the globe to minimize the environmental risks posed by micropollutants (MPs). For regulative institutions to be able to ensure the realization of environmental objectives, they need information on the environmental fate of MPs. Furthermore, there is an urgent need to further improve environmental decision-making, which heavily relies on scientific data. Use of mathematical and computational modeling in environmental permit processes for water construction activities has increased. Uncertainty of input data considers several steps from sampling and analysis to physico-chemical characteristics of MP. Machine learning (ML) methods are an emerging technique in this field. ML techniques might become more crucial for MP modeling as the amount of data is constantly increasing and the emerging new ML approaches and applications are developed. It seems that both modeling strategies, traditional and ML, use quite similar methods to obtain uncertainties. Process based models cannot consider all known and relevant processes, making the comprehensive estimation of uncertainty challenging. Problems in a comprehensive uncertainty analysis within ML approach are even greater. For both approaches generic and common method seems to be more useful in a practice than those emerging from ab initio. The implementation of the modeling results, including uncertainty and the precautionary principle, should be researched more deeply to achieve a reliable estimation of the effect of an action on the chemical and ecological status of an environment without underestimating or overestimating the risk. The prevailing uncertainties need to be identified and acknowledged and if possible, reduced. This paper provides an overview of different aspects that concern the topic of uncertainty in MP modeling.

show abstract

Section: Different Modeling Approachesmentioning

confidence: 91%

Uncertainty in Environmental Micropollutant Modeling

Ahkola,

Kotamäki,

Siivola

et al. 2024

Environmental Management

View full text Add to dashboard Cite

show abstract

“…The LSTM is a type of neural network that allows the autocorrelation often seen in hydrological variables to be modelled. It is commonly used in hydrology for simulation, forecasting, and hydroclimate predictions (e.g., Kratzert et al, 2018;Le et al, 2019;Natel de Moura et al, 2022) Perceptron Artificial Neural Network to forecast short-term irrigation water demand. They found that, although the forecasts had a skill comparable to previous studies, the lack of inclusion of physical understanding of the system limited the performance of the method.…”

Section: Data-driven and Hybrid Methodsmentioning

confidence: 99%

Connecting hydrological modelling and forecasting from global to local scales: Perspectives from an international joint virtual workshop

Dasgupta

Arnal

Emerton

et al. 2023

J Flood Risk Management

View full text Add to dashboard Cite

The unprecedented progress in ensemble hydro-meteorological modelling and forecasting on a range of temporal and spatial scales, raises a variety of new challenges which formed the theme of the Joint Virtual Workshop, "Connecting global to local hydrological modelling and forecasting: challenges and scientific advances". Held from 29 June to 1 July 2021, this workshop was coorganized by the European Centre for Medium-Range Weather Forecasts (ECMWF), the Copernicus Emergency Management (CEMS) and Climate Change (C3S) Services, the Hydrological Ensemble Prediction EXperiment (HEPEX), and the Global Flood Partnership (GFP). This paper aims to summarize the state-of-the-art presented at the workshop and provide an early career perspective. Recent advances in hydrological modelling and forecasting, reflections on the use of forecasts for decision-making across scales, and means to minimise new barriers to communication in the virtual format are also discussed. Thematic foci of the workshop included hydrological model development and skill assessment, uncertainty communication, forecasts for early action, co-production of services and incorporation of local knowledge, Earth Observation, and data assimilation. Connecting hydrological services to societal needs and local decision-making through effective communication, capacity-building and co-production was identified as critical. Multidisciplinary collaborations emerged as crucial to effectively bring newly developed tools to practice.

show abstract

“…The primary focus of the studies cited above was so far on model training and validation on a future part of the time series or on catchments not used for calibration. The question whether this success is transferable to the prediction of the consequences of modified driving forces in these catchments is less certain (Bai et al, 2021;Natel de Moura et al, 2022). For the prediction of the effects of climate change and water management measures on the hydrology of catchments, it is of particular interest to modify driving forces beyond the patterns observed in the past.…”

Section: Introductionmentioning

confidence: 99%

Metamorphic Testing of Machine Learning and Conceptual Hydrologic Models

Reichert

Höge

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. Predicting the response of hydrologic systems to modified driving forces, beyond patterns that have occurred in the past, is of high importance for estimating climate change impacts or the effect of management measures. This kind of predictions requires a model, but the impossibility of testing such predictions against observed data makes it still difficult to estimate their reliability. Metamorphic testing offers a methodology for assessing models beyond validation with real data. It consists of defining input changes for which the expected responses are assumed to be known at least qualitatively, and to test model behavior for consistency with these expectations. To increase the gain of information and reduce the subjectivity of this approach, we extend this methodology to a multi-model approach and include a sensitivity analysis of the predictions to training or calibration options. This allows us to quantitatively analyse differences in predictions between different model structures and calibration options in addition to the qualitative test to the expectations. In our case study, we apply this approach to selected conceptual and machine learning hydrological models calibrated to basins from the CAMELS data set. Our results confirm the superiority of the machine learning models over the conceptual hydrologic models regarding the quality of fit during calibration and validation periods. However, we also find that the response of machine learning models to modified inputs can deviate from the expectations and the magnitude and even the sign of the response can depend on the training data. In addition, even in cases in which all models passed the metamorphic test, there are cases in which the quantitative response is different for different model structures. This demonstrates the importance of this kind of testing beyond the usual calibration-validation analysis to identify potential problems and stimulate the development of improved models.

show abstract

Evaluating the long short-term memory (LSTM) network for discharge prediction under changing climate conditions

Cited by 11 publications

References 52 publications

Uncertainty in Environmental Micropollutant Modeling

Uncertainty in Environmental Micropollutant Modeling

Connecting hydrological modelling and forecasting from global to local scales: Perspectives from an international joint virtual workshop

Metamorphic Testing of Machine Learning and Conceptual Hydrologic Models

Contact Info

Product

Resources

About