Advances in land surface  forecasting: a comparison of LSTM, gradient boosting, and feed-forward neural networks as prognostic state emulators in a case study with ecLand

Wesselkamp, Marieke; Chantry, Matthew; Pinnington, Ewan; Choulga, Margarita; Boussetta, Souhail; Kalweit, Maria; Bödecker, Joschka; Dormann, Carsten F.; Pappenberger, Florian; Balsamo, Gianpaolo

doi:https://doi.org/10.5194/gmd-18-921-2025

Articles | Volume 18, issue 4

https://doi.org/10.5194/gmd-18-921-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-18-921-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 18, issue 4

Development and technical paper

|

19 Feb 2025

Development and technical paper |

| 19 Feb 2025

Advances in land surface forecasting: a comparison of LSTM, gradient boosting, and feed-forward neural networks as prognostic state emulators in a case study with ecLand

Marieke Wesselkamp, Matthew Chantry, Ewan Pinnington, Margarita Choulga, Souhail Boussetta, Maria Kalweit, Joschka Bödecker, Carsten F. Dormann, Florian Pappenberger, and Gianpaolo Balsamo

Supplement

https://doi.org/10.5194/gmd-18-921-2025-supplement

Data sets

Advances in Land Surface Model-based Forecasting: A Comparison of LSTM, Gradient Boosting, and Feedforward Neural Networks as Prognostic State Emulators in a Case Study with ECLand. European and Global training and test data Marieke Wesselkamp et al. https://doi.org/10.21957/n17n-6a68

Advances in Land Surface Model-based Forecasting: A Comparison of LSTM, Gradient Boosting, and Feedforward Neural Networks as Prognostic State Emulators in a Case Study with ECLand. European and Global training and test data Marieke Wesselkamp et al. https://doi.org/10.21957/pcf3-ah06

Model code and software

Advances in Land Surface Model-based Forecasting: A Comparison of LSTM, Gradient Boosting, and Feedforward Neural Networks as Prognostic State Emulators in a Case Study with ECLand. Model code. Marieke Wesselkamp et al. https://doi.org/10.17605/OSF.IO/8567D

Short summary

We compared spatiotemporal forecasts of three machine learning models that learned water and energy
states on the land surface from a physical model scheme. The forecasting models were developed with reanalysis data and simulations on a European scale and transferred to the globe. We found that all approaches deliver highly accurate approximations of the physical dynamic at long time horizons, implying their usefulness to advance land surface forecasting with synthetic data.