Short-Term River Flood Forecasting Using Composite Models and Automated Machine Learning: The Case Study of Lena River

Sarafanov, Mikhail; Borisova, Yulia; Maslyaev, Mikhail; Revin, Ilia; Maximov, Gleb; Nikitin, Nikolay O.

doi:10.3390/w13243482

Cited by 8 publications

(3 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A possible next step in the research is to use artificial intelligence and machine learning techniques. In fact, combining machine learning and physical-based models is becoming popular in the design of predictive systems [29,30]. Including the human component is especially relevant in southern Europe where water retention is a more common practice.…”

Section: Conclusion and Future Approachmentioning

confidence: 99%

Framework for Improving Land Boundary Conditions in Ocean Regional Products

Campuzano

Santos

Simionesei

et al. 2022

JMSE

View full text Add to dashboard Cite

The coupling of coastal or regional ocean models to hydrological models or observed data is currently an uncommon practice in operational oceanography. Though hydrological models are regarded as a powerful and useful tool for estimating the quantity and quality of freshwater running in a watershed, they fail to provide accurate results for river flow reaching the coastal area due to water-management activities occurring within the river catchment, activities such as human consumption, irrigation, storage, etc. For this reason, many coastal and regional ocean models continue to impose surface zero-salinity discharges as land boundary conditions for representing such a dynamic boundary. Moreover, river flows are based in climatologies, thus neglecting seasonal and interannual variability. To achieve those objectives, this study proposes an integrated methodology ranging from watershed models to validation in the coastal area and passing through methods and proxies for integrating the freshwater flows into regional ocean models. The main objective of this study is to explore the results obtained by using more sophisticated land boundary conditions based on the capacities of state-of-the-art hydrologic models combined with observation networks. In addition to the evaluation of the source of river-flow data, this work also explores the use of estuarine proxies based on simple modelling grids. The estuarine proxies enable the incorporation of the mixing processes that take place in estuaries into the land fluxes and obtain the plume momentum. The watershed, estuarine proxies, and ocean were modelled using the MOHID Water modelling system and evaluated in western Iberia waters. The modelling results served to illustrate the sea surface salinity extension of the Western Iberia Buoyant Plume (WIBP) during an extreme event in March 2018.

show abstract

Section: Conclusion and Future Approachmentioning

confidence: 99%

Framework for Improving Land Boundary Conditions in Ocean Regional Products

Campuzano

Santos

Simionesei

et al. 2022

JMSE

View full text Add to dashboard Cite

show abstract

“…Similar problems have been encountered in the prediction of midwinter breakups (the early breakup of ice cover brought about by midwinter thaws) using data‐based thresholds on single location (Carr & Vuyovich, 2014; Prowse et al., 2002) or regional scales (Newton et al., 2017) and machine learning analysis (De Coste et al., 2022a, 2022b), or in the prediction of breakup ice jams using artificial neural networks (Massie et al., 2002), neuro‐fuzzy systems (Mahabir et al., 2006), and stacking ensembles (De Coste et al., 2021). These techniques have also been extended to generalized spring flooding using boosting and random forest (RF; Kulin et al., 2021), ensembles of regression and snowmelt models (Sarafanaov et al., 2021), and recurrent neural networks (Cai & Yu, 2022). The management of data in these hydrological forecasting studies, especially when attempting modeling on regional or national scales, is often a challenge, as is delivering a user‐friendly means of deployment.…”

Section: Introductionmentioning

confidence: 99%

A hybrid ontology‐based semantic and machine learning model for the prediction of spring breakup

Coste

Khédri

2023

Computer aided Civil Eng

View full text Add to dashboard Cite

River ice breakups carry the potential for high flows and flooding and are of great interest to accurately predict. A challenge in forecasting these events is the management of the massive amounts of data associated with an ice season. This study couples ontological and machine learning models in a new hybrid modeling framework to predict spring breakup on a national scale. The Ice Season Ontology sorts the data and allows for a user‐friendly means of analyzing any ice season, providing insight on which variables are most and least central. With this, a refined variable selection is able to be made for machine learning models. The most successful developed model, a random forest, produced highly accurate forecasts when applied to a national scale case study, with a mean absolute error of 10.85 days and an R2 of .884. This new modeling framework provides a means for decision‐making support for river bound communities and a new methodology for modeling applications in other fields.

show abstract

“…Castelletti, 2012;Antonio Francipane, 2012;Singh, 2009;Silberstein, 2006). Machine learning approaches can automate data ingestion, (Mikhail Sarafanov, 2021; Randal S. Olson, 2016) but their opacity may limit insight and stakeholder trust (Jajarmizadeh, 2012). Both approaches require large amounts of computational power and data (R. Kumar, 2013;Moges B. Wagena, 2020).…”

Section: Introductionmentioning

confidence: 99%

Generating interpretable rainfall-runoff models automatically from data

Dantzer,

Kerkez

2023

Preprint

View full text Add to dashboard Cite

A sudden surge of data has created new challenges in water management, spanning quality control, assimilation, and analysis. Few approaches are available to integrate growing volumes of data into interpretable results. Process-based hydrologic models have not been designed to consume large amounts of data. Alternatively, new machine learning tools can automate data analysis and forecasting, but their lack of interpretability and reliance on very large data sets limits the discovery of insights and may impact trust. To that end, we present a new approach, which seeks to strike a middle ground between process-, and data-based modeling. The contribution of this work is an automated and scalable methodology that discovers differential equations and latent state estimations within hydrologic systems using only rainfall and runoff measurements. We show how this enables automated tools to learn interpretable models of 6 to 18 parameters solely from measurements. We apply this approach to nearly 400 stream gaging sites across the US, showing how complex catchment dynamics can be reconstructed solely from rainfall and runoff measurements. We also show how the approach discovers surrogate models that can replicate the dynamics of a much more complex process-based model, but at a fraction of the computational complexity. We discuss how the resulting representation of watershed dynamics provides insight and computational efficiency to enable automated predictions across large sensor networks.

show abstract

Short-Term River Flood Forecasting Using Composite Models and Automated Machine Learning: The Case Study of Lena River

Cited by 8 publications

References 53 publications

Framework for Improving Land Boundary Conditions in Ocean Regional Products

Framework for Improving Land Boundary Conditions in Ocean Regional Products

A hybrid ontology‐based semantic and machine learning model for the prediction of spring breakup

Generating interpretable rainfall-runoff models automatically from data

Contact Info

Product

Resources

About