Editors' Highlights are summaries of recent papers by AGU journal editors.
sauce: water resources research
Kalman filtering and its derivatives are common methods for coupling models and observations through state and parameter updates. This is done through linear mapping based on linear statics and most notably Gaussian assumptions. Because geoscience processes are generally nonlinear and often non-Gaussian, methodologies frequently reach their limits.
Zhang et al. [2024] We propose a methodology to replace linear maps with nonlinear maps based on deep learning (DL). Furthermore, this methodology is advanced by adding predictors, local update steps, and data augmentations that demonstrate the strong performance and versatility of DL. The proposed methodology is tested in numerical experiments related to groundwater flow and transport that reconstruct complex hydraulic conductivity fields. Future studies will demonstrate the utility of this method in application to real-world observations.
Citation: Zhang, J., Cao, C., Nan, T., Ju, L., Zhou, H., Zeng, L. (2024). A new deep learning approach for data assimilation of complex hydrological systems. water resources research60, e2023WR035389. https://doi.org/10.1029/2023WR035389
—Stefan Kollet, editor water resources research
Text © 2024. Author. CC BY-NC-ND 3.0
Images are subject to copyright unless otherwise noted. Reuse without the express permission of the copyright owner is prohibited.
related
