EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Real‐Time Prediction of Salt Intrusion in Tidal Estuaries Using Long Short‐Term Memory Networks

Karoline Rummel, Tobias Strauß, Franziska Lauer, Ulf Gräwe

Journal of Geophysical Research: Machine Learning and Computation · 2025-12 · 2 (4)

DOI: 10.1029/2025jh000768 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Geophysical Research: Machine Learning and Computation, online erschienen am 01.11.2025. Verfasst von Karoline Rummel und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Universität Rostock.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Estuaries worldwide provide unique ecosystems and critical freshwater resources for drinking water supply or agriculture. They are vulnerable to salt intrusion, which can have serious environmental and socioeconomic consequences. Obtaining a detailed picture of the state of the estuary through continuous monitoring is therefore essential. However, this is challenging due to the limited number of measurement stations. Numerical models provide such detailed insights but are computationally expensive and have limited suitability for real‐time predictions. Data‐driven models, such as Long Short‐Term Memory (LSTM) networks, offer a promising alternative. In this study, LSTM networks are used to predict the location of the 2 g/kg bottom isohaline as an indicator for salt intrusion. Based on salinity, water level, wind, and river discharge data, the LSTM networks were trained on the salt intrusion length derived from a numerical model to estimate future salt intrusion accurately. Thus, detailed horizontal information from point measurements can be provided. This method is applied to the tidal Weser and Elbe estuaries, demonstrating that the network remains functional even when input stations are missing. Short‐term environmental variability hampers longer‐term forecasts, and limited data reduce prediction accuracy. Despite these constraints, our results show that LSTM networks replicate numerical model outputs from observational data and deliver real‐time estimates of estuarine salt intrusion. This approach provides an efficient, adaptable tool for coastal monitoring and management.

Förderung

  • Universität Rostock

Zitieren

APA

Rummel, K., Strauß, T., Lauer, F., & Gräwe, U. (2025). Real‐Time Prediction of Salt Intrusion in Tidal Estuaries Using Long Short‐Term Memory Networks. Journal of Geophysical Research: Machine Learning and Computation, 2(4). https://doi.org/10.1029/2025jh000768

Vancouver

Rummel K, Strauß T, Lauer F, Gräwe U. Real‐Time Prediction of Salt Intrusion in Tidal Estuaries Using Long Short‐Term Memory Networks. Journal of Geophysical Research: Machine Learning and Computation. 2025;2(4). doi:10.1029/2025jh000768

BibTeX
@article{Rummel2025,
  author = {Rummel, Karoline and Strauß, Tobias and Lauer, Franziska and Gräwe, Ulf},
  title = {Real‐Time Prediction of Salt Intrusion in Tidal Estuaries Using Long Short‐Term Memory Networks},
  journal = {Journal of Geophysical Research: Machine Learning and Computation},
  year = {2025},
  volume = {2},
  number = {4},
  doi = {10.1029/2025jh000768}
}
Kostenlos starten →