Mesa et al. (2026) Quantifying atmospheric and land drivers of hot temperature extremes through explainable Artificial Intelligence

Identification

• Journal: Weather and Climate Dynamics
• Year: 2026
• Date: 2026-09-07
• Authors: Alejandro Mesa, Lluís Palma, Markus G. Donat, Stefano Materia, Bruna Gràvalos Talló, Raül Marcos-Matamoros
• DOI: 10.5194/wcd-7-1709-2026

Research Groups

• Earth Sciences Department, Barcelona Supercomputing Center (BSC), Barcelona, Spain
• Facultat de Física, Universitat de Barcelona, Diagonal 645, 08028 Barcelona, Spain
• Institutució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain
• CMCC Foundation–Euro–Mediterranean Center on Climate Change, Bologna, Italy

Short Summary

This study uses an explainable machine-learning framework to disentangle the respective influences of large-scale atmospheric circulation, soil-moisture anomalies, and rising CO2 concentrations on boreal-summer temperature extremes at six locations across Europe and North Africa. The results show that atmospheric circulation consistently dominates model explainability across all locations.

Objective

• To quantify the relative contributions of atmospheric circulation, soil moisture, and anthropogenic climate change to extreme temperature events during the boreal summer (JJA).

Study Configuration

• Spatial Scale: Regional scale, covering Europe and North Africa.
• Temporal Scale: Daily data from 1940 to present.

Methodology and Data

• Models used: Deep learning model with a CombinedModel architecture, including a Multi Layer Perceptron (MLP) for land and CO2 data, and a ConvNeXt convolutional neural network (CNN) for spatial features from circulation.
• Data sources: ERA5 reanalysis dataset, E-OBS gridded dataset for temperature and precipitation, and Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI) as proxies for soil moisture anomalies.

Main Results

• Atmospheric circulation consistently dominates model explainability across all locations.
• Soil-moisture influence exhibits a northward gradient: negligible at Marrakech, moderate at Córdoba, and substantial at Lyon.
• The dominant shallow soil-moisture signal shows an inverse relationship between moisture anomalies and SHAP values.

Contributions

• This study provides a novel framework for quantifying the relative contributions of atmospheric circulation, soil moisture, and anthropogenic climate change to extreme temperature events.
• The results highlight the importance of considering multiple drivers in understanding hot extremes.

Funding

• Not specified.

Citation

@article{Mesa2026Quantifying,
  author = {Mesa, Alejandro and Palma, Lluís and Donat, Markus G. and Materia, Stefano and Talló, Bruna Gràvalos and Marcos-Matamoros, Raül},
  title = {Quantifying atmospheric and land drivers of hot temperature extremes through explainable Artificial Intelligence},
  journal = {Weather and Climate Dynamics},
  year = {2026},
  doi = {10.5194/wcd-7-1709-2026},
  url = {https://doi.org/10.5194/wcd-7-1709-2026}
}