State of Wildfires 2024-25: ConFLAME NRT Attribution Results

ConFLAME attribution outputs are split by region, using the same regions as in previous years, with each directory name ending in -2425 to denote the 2024/25 report version: Amazon-2425 – Northeast Amazonia Pantanal-2425 – Pantanal–Chiquitano LA-2425 – Southern California Congo-2425 – Congo Basin Each of these directories contains the files necessary for rerunning the model without redoing the optimisation.While you may not need to view these directly, they are useful for setting up your own ConFLAME model experiments (see the ConFLAME configuration settings). Core files: model_graph.png – Graphical representation of the trained model scalers-*.csv – Scaling parameters used for model input features trace-*.nc – Posterior traces from the Bayesian optimisation none_trace-params-*.txt – Parameter values without the trace variables_info-*.txt – List of variables used in the model and their definitions namelist_*.txt – Model configuration files Core directories: figs – Automatically generated evaluation figures and selected output plots time_series – Burned area (BA) time series for attribution experiments.Structure: <<model_id>>/<<experiment>>/<<range>>/<<members or percentiles>>/<<metric>>/<<variable>>.csv Model ID: currently only _19-frac_points_0.5 Experiments: factual – Simulations using observed climate conditions counterfactual – Simulations with climate change signal removed counterfactual-metmean – Counterfactuals using mean meteorological differences (HadGEM3 ALL–NAT) as in Barbosa et al. (submitted) Range: mean – Regional mean burned area pc-0.95 – Burned area for grid cells in the top 5% of BA values Members or Percentiles: members – Time series for individual ensemble members from the sampled posterior percentiles – Precomputed percentile ranges (e.g., 5th–95th) over all members Metric: absolute – Burned area in km² or equivalent units climatology – Long-term monthly mean burned area anomaly – Deviation from climatology for that month ratio – Burned area divided by climatology Variables: Evaluate.csv – BA simulated with stochasticity (used in the State of Wildfires report) Control.csv – BA simulated without stochasticity samples – Contains full spatial samples from the posterior for each experiment: <<model_id>>/<<experiment>>/<<variable>>/sample-predXXX.nc Variables align with those in the time_series directory. Pairing: Samples are paired across experiments within the same optimisation, meaning sample-predX.nc in the factual experiment matches sample-predX.nc in the counterfactual experiment.