Explainable Earth Surface Forecasting Under Extreme Events

Oscar J. Pellicer‐Valero; Miguel‐Ángel Fernández‐Torres; Chaonan Ji; Miguel D. Mahecha; Gustau Camps‐Valls

doi:10.1029/2024ef005446

journal article Open Access Sep 01, 2025

Explainable Earth Surface Forecasting Under Extreme Events

Oscar J. Pellicer‐Valero

Miguel‐Ángel Fernández‐Torres

Chaonan Ji Miguel D. Mahecha

Gustau Camps‐Valls

Earth's Future Vol. 13 No. 9 · American Geophysical Union (AGU)

View at Publisher Save 10.1029/2024ef005446

Abstract

AbstractWith climate change‐related extreme events on the rise, high‐dimensional Earth observation data present a unique opportunity for forecasting and understanding impacts on ecosystems. This is, however, impeded by the complexity of processing, visualizing, modeling, and explaining this data. We train a convolutional long short‐term memory‐based architecture on the novel DeepExtremeCubes data set to showcase how this challenge can be met. DeepExtremeCubes includes around 40,000 long‐term Sentinel‐2 minicubes (January 2016–October 2022) worldwide, along with labeled extreme events, meteorological data, vegetation land cover, and a topography map, sampled from locations affected by extreme climate events and surrounding areas. When predicting future reflectances and vegetation impacts through the kernel normalized difference vegetation index, the model achieved an score of 0.9055 in the test set. Explainable artificial intelligence was used to analyze the model's predictions during the October 2020 Central South America compound heatwave and drought event. We chose the same area exactly 1 year before the event as a counterfactual, finding that the average temperature and surface pressure are generally the most important predictors. In contrast, minimum evaporation anomalies play a leading role during the event. We also found the anomalies of the reflectances in the timestep before the extreme event to be critical predictors of its impact on vegetation. The code to replicate all experiments and figures in this paper is publicly available at https://github.com/DeepExtremes/txyXAI.

Topics

No keywords indexed for this article. Browse by subject →

References

56

[1]

Permutation importance: a corrected feature importance measure

André Altmann, Laura Toloşi, Oliver Sander et al.

Bioinformatics 10.1093/bioinformatics/btq134

[2]

10.1109/cvpr52733.2024.02625

[3]

Bommasani R. "On the opportunities and risks of foundation models" arXiv (2021)

[4]

A unified vegetation index for quantifying the terrestrial biosphere

Gustau Camps-Valls, Manuel Campos-Taberner, Álvaro Moreno-Martínez et al.

Science Advances 10.1126/sciadv.abc7447

[5]

Cong Y. (2022)

[6]

Copernicus DEM—global digital elevation model. (2024).Copernicus DEM—Global digital elevation model.https://doi.org/10.5270/ESA-c5d3d65 10.5270/esa-c5d3d65

[7]

Diaconu C. A. (2022)

[8]

Gao Z. "Earthformer: Exploring space‐time transformers for Earth system forecasting" arXiv (2022)

[9]

10.1016/j.agsy.2019.05.010

[10]

Goodfellow I. J. Pouget‐Abadie J. Mirza M. Xu B. Warde‐Farley D. Ozair S. et al. (2014).Generative adversarial nets. arXiv. Retrieved fromhttp://www.github.com/goodfeli/adversarial

[11]

10.1002/qj.3803

[12]

Long Short-Term Memory

Sepp Hochreiter, Jürgen Schmidhuber

Neural Computation 10.1162/neco.1997.9.8.1735

[13]

10.1109/mgrs.2024.3467001

[14]

Hong S. (2017)

[15]

10.3390/agriculture13050971

[16]

Ji C. Fincke T. Benson V. Camps‐Valls G. Fernández‐Torres M.‐Á. &Gans F.(2024).Deepextremecubes[Dataset].Universität Leipzig.https://doi.org/10.25532/OPARA-703 10.25532/opara-703

[17]

10.1038/s41597-025-04447-5

[18]

10.5194/hess-26-6339-2022

[19]

10.1016/j.jag.2021.102520

[20]

10.1016/j.ecoinf.2024.102474

[21]

Kokhlikyan N. Miglani V. Martin M. Wang E. Alsallakh B. Reynolds J. et al. (2020).Captum: A unified and generic model interpretability library for pytorch. arXiv. Retrieved fromhttps://arxiv.org/abs/2009.07896v1

[22]

10.1029/2022gl099368

[23]

10.1016/j.scitotenv.2024.170330

[24]

Lundberg S. M. "A unified approach to interpreting model predictions" Advances in Neural Information Processing Systems (2017)

[25]

10.1029/2023ef003963

[26]

10.1175/aies-d-22-0058.1

[27]

10.1175/jhm-d-23-0036.1

[28]

10.1002/joc.7365

[29]

10.5194/npg-31-535-2024

[30]

10.1109/lgrs.2023.3244064

[31]

Molnar C.(2023).Interpretable machine learning a guide for making black box models explainable. Independently published. Retrieved fromhttps://christophm.github.io/interpretable-ml-book/

[32]

10.5194/essd-13-4349-2021

[33]

Paszke A. (2019)

[34]

Pellicer‐Valero O. J. (2025)

[35]

Programme U. N. E. (2020)

[36]

10.1016/j.wace.2022.100525

[37]

10.1109/iccv51070.2023.00378

[38]

10.1038/nature12350

[39]

10.1109/cvprw53098.2021.00124

[40]

Ribeiro M. T. (2016)

[41]

Robin C. "Learning to forecast vegetation greenness at fine resolution over Africa with convlstms" arXiv (2022)

[42]

10.1109/access.2020.2976199

[43]

10.5194/isprs-annals-iv-2-w7-153-2019

[44]

10.1007/s11263-019-01228-7

[45]

Shi X. "Convolutional LSTM network: A machine learning approach for precipitation nowcasting" arXiv (2015)

[46]

Shrikumar A. "Not just a black box: Learning important features through propagating activation differences" arXiv, 1, 0‐5 (2016)

[47]

Smilkov D. "Smoothgrad: Removing noise by adding noise" arXiv (2017)

[48]

Smith M. J. "Earthpt: A time series foundation model for Earth observation" arXiv (2024)

[49]

10.1109/igarss.2019.8900532

[50]

10.1109/tgrs.2022.3194732

Showing 50 of 56 references

Metrics

2

Citations

56

References

Details

Published: Sep 01, 2025
Vol/Issue: 13(9)
License: View

Authors

O

Oscar J. Pellicer‐Valero

Image Processing Laboratory (IPL) Universitat de València València Spain

M

Miguel‐Ángel Fernández‐Torres

Image Processing Laboratory (IPL) Universitat de València València Spain

C

Chaonan Ji

Remote Sensing Centre for Earth System Research (RSC4Earth) Leipzig University Leipzig Germany; Institute for Earth System Science and Remote Sensing Leipzig University Leipzig Germany

M

Miguel D. Mahecha

Remote Sensing Centre for Earth System Research (RSC4Earth) Leipzig University Leipzig Germany; Institute for Earth System Science and Remote Sensing Leipzig University Leipzig Germany; Image and Signal Processing Group Leipzig University Leipzig Germany; Helmholtz Centre for Environmental Research—UFZ Leipzig Germany

G

Gustau Camps‐Valls

Image Processing Laboratory (IPL) Universitat de València València Spain

Funding

Horizon 2020 Framework Programme Award: 101003469

European Space Agency Award: AI4Science

Cite This Article

Oscar J. Pellicer‐Valero, Miguel‐Ángel Fernández‐Torres, Chaonan Ji, et al. (2025). Explainable Earth Surface Forecasting Under Extreme Events. Earth's Future, 13(9). https://doi.org/10.1029/2024ef005446

Explainable Earth Surface Forecasting Under Extreme Events

You May Also Like