Reproducible eRFO-Assisted Wildfire Escalation Analysis

Le résumé fourni par la source

This reproducibility archive supports Science Progress manuscript SCI-26-1321, entitled “A Novel Optimized Ensemble Learning Framework for Early Wildfire Escalation Prediction in Mediterranean Environments.” It contains anonymized incident-level modelling data and coordinates authorized for public release by the Mersin Regional Directorate of Forestry, the original and audited eRFO implementations, analysis scripts, locked temporal protocols, dependency records, optimizer comparison outputs, bootstrap confidence intervals, locked 2021–2025 prediction probabilities, and the three figures used in the manuscript. The primary analysis includes 1,898 wildfire incidents with known burned-area outcomes recorded during 2004–2025 and predicts escalation to at least 10 ha using calendar, district, event-day meteorology, and Canadian Fire Weather Index variables. Model development uses three expanding-window temporal folds within 2004–2020, while 2021–2025 is retained as an independent recent-year evaluation period. The software is released under the MIT License, and the anonymized data are released under the Creative Commons Attribution 4.0 International License.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.