Reproduction Package for Machine Learning for One Day Ahead Prediction of High Fire Weather Index Conditions from Area Weighted ERA5 Data in Mersin

This repository provides the reproducibility package accompanying the revised manuscript “Machine Learning for One Day Ahead Prediction of High Fire Weather Index Conditions from Area Weighted ERA5 Data in Mersin”. The package contains the derived area-weighted Fire Weather Index series, spatial weights for the 41 ERA5 cells intersecting Mersin Province, season-aware modelling data, out-of-fold development predictions, frozen 2024–2025 temporal-holdout predictions, model specifications, calibration and uncertainty outputs, statistical comparisons, final tables and figures, and the analysis scripts and environment files required to reproduce or independently verify the reported results. Raw ERA5 files are not redistributed and can be reacquired from the Copernicus Climate Data Store using the supplied acquisition workflow and archived input manifest. The study outcome is next-day regional meteorological High+ FWI status; it does not represent wildfire ignition occurrence, burned area, suppression outcome, or an operational forecast product.