Learning evaporative fraction with memory

Zhao, Wenli; Winkler, Alexander J.; Reichstein, Markus; Orth, Rene; Gentine, Pierre

doi:10.5194/hess-30-5373-2026

Articles | Volume 30, issue 16

https://doi.org/10.5194/hess-30-5373-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-30-5373-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 30, issue 16

Research article

| Highlight paper

|

25 Aug 2026

Research article | Highlight paper |

| 25 Aug 2026

Learning evaporative fraction with memory

Wenli Zhao, Alexander J. Winkler, Markus Reichstein, Rene Orth, and Pierre Gentine

Supplement

https://doi.org/10.5194/hess-30-5373-2026-supplement

Data sets

Warm Winter 2020 ecosystem eddy covariance flux product for 73 stations in FLUXNET-Archive format -- release 2022-1, version 1.0 Warm Winter 2020 Team and ICOS Ecosystem Thematic Centre https://doi.org/10.18160/2G60-ZHAK

Model code and software

Representative analysis and plotting scripts for Learning Evaporative Fraction with Memory W. Zhao https://doi.org/10.5281/zenodo.21960468

Video supplement

Learning Evaporative Fraction with Memory (Video Abstract) Wenli Zhao, Alexander J. Winkler, Markus Reichstein, Rene Orth, and Pierre Gentine https://doi.org/10.5446/73776

Editorial statement

This study proposes a novel framework for predicting evaporative fraction (EF) by explicitly incorporating meteorological memory into machine learning models. By demonstrating that long short-term memory (LSTM) networks substantially outperform conventional approaches across a range of ecosystems, the study reveals that vegetation responses to water availability are fundamentally governed by temporal dependencies that are often neglected in current modeling frameworks. This work provides new process-level understanding of land–atmosphere interactions.