Editor's highlights are summary of recent papers by AGU journal editors.
sauce: Water Resources Research
The promise and appeal of Deep Learning (DL) is the algorithmic fusion of all available data to achieve model generalization and prediction of complex systems. Therefore, multivariate training and prediction tasks should be designed to identify all the relevant connections between variables of different spatial and time scales.
Ouyang et al. [2025] We propose a multitasking long-term memory (LSTM) neural network to predict the time series of multiple hydrological variables. The application of the approach improves physical consistency and accuracy by combining different variables from the prediction task and sharing information between them. The authors demonstrate this in various predictive exercises for river flow and evapotranspiration, including data shortage conditions.
This study is a great example of how innovation within DL can realize the promise of generalizable hydrological models and predictions of future complex systems. It also implicitly encourages hydrologists to expand their DL approach for multitasking. After all, there is plenty of data and computing resources available to achieve the promise of DL.
Quotes: Ouyang, W., Gu, X., Ye, L., Liu, X. , & Zhang, C. (2025). We investigate enhancing the interconnection of waterway variables and predictions of data-limited basins through multitasking learning. Water Resources Research61, E2023WR036593. https://doi.org/10.1029/2023WR036593
– Editor, Stefan Kollet, Water Resources Research
Text ©2025. author. CC by-nc-nd 3.0
Except in other cases, images are subject to copyright. Reuse without express permission from the copyright owner is prohibited.
Related
