[Research Article] AI-driven ensemble forecasting of extreme wind gusts: Random Forest modeling and case studies from the western Mediterranean

A new study published in Big Earth Data develops a stochastic random forest (RF) classifier to predict extreme wind-gust (WG) occurrences along the western Mediterranean coast by integrating instantaneous and trailing 24-hour meteorological variables across multiple height levels, enabling the model to capture both short-term evolution and immediate atmospheric forcing. It demonstrates strong forecasting performance with high precision, low false-alarm rates, and reliable uncertainty estimation, while identifying barometric-pressure tendencies and humidity as key predictors, highlighting the framework’s potential for early-warning systems, climate resilience, and disaster-risk management. UERRA reanalysis data can be downloaded from the Copernicus Climate Data Store platform: https://cds.climate.copernicus.eu/. Supplementary materials supporting the findings of this study are openly accessible at the public GitHub repository: https://github.com/AI4OCEANS/.

Citation

Guerrero-Navarro, G. H., Martinez-Amaya, J., & Nieves, V. (2025). AI-driven ensemble forecasting of extreme wind gusts: Random Forest modeling and case studies from the western Mediterranean. Big Earth Data, 1–20. https://doi.org/10.1080/20964471.2025.2593745

Abstract

Forecasting extreme wind-gust (WG) is challenging because thedriving processes evolve rapidly and non-linearly. We tailora stochastic random forest (RF) classifier that predicts extremeWGs occurrences along the western Mediterranean coast andranks the meteorological factors that trigger WG intensification.Uniquely, each atmospheric variable—pressure, humidity, wind-direction and temperature—enters the model twice, both as aninstantaneous value and as its trailing 24-h mean, allowing the RFto learn momentary forcing and short-term evolution in a singlestep. These dual-time predictors are evaluated at 11 height levels(15–500 meters), yet the single (15 meters) version already deliversrobust skill. Local and regional models attain 85% mean precisionacross all lead-times, maintaining robust skill (~80% precision at the48-h horizon) with false-alarm ratios below 20% and well-calibratedprobabilistic reliability. The ensemble’s internal spread providesa first-order measure of forecast uncertainty, making the frameworkrisk-oriented and ready for early-warning forecasting workflows,with direct value for civil-protection agencies and regional adapta-tion planning. Variable-importance analysis highlights barometric-pressure tendencies and humidity as the dominant early-warningcues. The lightweight, interpretable, and data-efficient design iswell poised for broader use across other wind-prone regions—subject to local data availability and validation—thereby strength-ening climate-resilience and disaster-risk management.

#geoscience #remote sensing #earth observation #GIS #data analysis #Big Data #visualization

Big Earth Data is an interdisciplinary Open Access journal which aims to provide an efficient and high-quality platform for promoting the sharing, processing and analyses of Earth-related big data, thereby revolutionizing the cognition of the Earth’s systems. The journal publishes a wide range of content, including Research Articles, Review Articles, Data Notes, Technical Notes, and Perspectives. It is now included in ESCI (IF=3.8, Q1), Scopus (CiteScore=9.0, Q1), Ei Compendex, GEOBASE, and Inspec. Starting from 2023, Big Earth Data has announced a new award series for authors: Best and Outstanding Paper Awards.