Data to reproduce results from: "Forecasted increases in fire occurrence in Victoria under climate change"

<b>Scripts and data for analysis of the manuscript</b>: <i>"</i><i>Forecasted increases in fire occurrence in Victoria under climate change"</i><br><b>Authors:</b> Charlie Hart, Andrew Dowdy, Sarah C. McColl-Gausden, Hamish Clarke, Luke Collins, Amelia French, Trent D. Penman, Nevil Amos, Angie Haslem, Cindy E. Hauser, Jim Thomson, Josephine MacHunter, Matt White &amp; William L. Geary<b>Corresponding author:</b> William L. Geary ([e-mail])Description of the Data and File StructureThis repository contains the code and key data required to replicate the modelling and figures presented in the manuscript.<b>Note on Replication:</b> While the full workflow is documented for transparency, Scripts 1–4 and 9 rely on external raw data (DEECA spatial layers, AGCD/AWO climate grids, etc.) that are not included in this repository due to size and licensing restrictions.<b>To replicate the core results:</b><b>Run Step 5</b> to fit the Boosted Regression Trees (BRTs).<b>Run Step 6</b> to prepare cross-validation data.<b>Run Step 8</b> to consolidate and summarise the cross-validation performance.<b>Run Step 10</b> to generate the spatial fire probability predictions.<b>Run Step 11</b> to recreate the final publication figures.Folder: scriptsThis folder contains the sequential workflow for the analysis.<b>fun_fit_forced_brt.R</b> - Helper functions for fitting BRT models.<b>Step 1 Make analysis mask.R</b> - Generates the 75m study region mask and stratified random sampling points.<b>Step 2_Make burned area rasters.R</b> - Processes Victorian Fire History into annual binary burned area layers.<b>Step 3a_SPEI from AWO grids.R</b> - Calculates 12 and 24-month SPEI from climate data.<b>Step 3b_Make covariates.R</b> - Processes fire weather, ignitions, topography, and Time Since Fire (TSF) predictors.<b>Step 3c_Make covariate stacks.R</b> - Harmonises all spatial layers into unified annual stacks.<b>Step 4a_Extract data.R</b> - Extracts covariate values to sampling points for model training.<b>Step 5_Fit the model_lr_0.005.R</b> - Fits primary BRT models (learning rate 0.005).<b>Step 6_Prep Cross validation Data.R</b> - Prepares data folds for model validation.<b>Step 8_Summarise CVs.R</b> - Consolidates performance metrics from the cross-validation runs.<b>Step 10_Current and Future Fire Predictions.R</b> - Generates spatial probability rasters for baseline and future scenarios.<b>Step 11_Plots for publication.R</b> - Produces final binned maps, delta maps, and regional boxplots.Folder: dataThis folder contains essential input files for the analysis.<b>Modelling Data:</b> Extracted covariate values used to train the BRT models.<b>Districts:</b> Spatial boundaries for Victorian fire management districts.<b>Random Points:</b> The stratified random points used for data extraction and model training.Folder: outputsThis folder contains final datasets and model objects created at key stages.<b>Prediction_Stacks:</b> Multi-layer annual covariate stacks used to generate predictions.<b>Predictions:</b> Baseline and future fire probability GeoTIFFs (.tif).<b>Final_Models:</b> * <i>brt_base_lr0.005_ffdi_95.RDS</i> - Base model object.<i>brt_full_rac_lr0.005_ffdi_95.rds</i> - RAC model object.<b>Regional_Summaries:</b> * <i>regional_fire_risk_summary_RAC.csv</i> - Mean fire probabilities by district.Code and SoftwareAll analyses were performed in <b>R v4.3.0</b> or later. The complete, version-controlled codebase is hosted on GitHub:<b>GitHub Repository:</b>https://github.com/charliehart4/future-fire-victoriaKey R Libraries:<b>Spatial Data:</b> terra, sf, tidyterra<b>Modelling:</b> dismo, gbm (Boosted Regression Trees)<b>Visualisation &amp; Utility:</b> ggplot2, dplyr, tidyrUsage:To replicate the study, clone the GitHub repository and ensure the 21 GB data archive from Figshare is extracted into the <code>/data</code> and <code>/outputs</code> directories as specified in the scripts.