Fig. 4: Summary of ROC AUC scores from the ablation experiments.
From: Fire weather indices tailored to regional patterns outperform global models
The DT model using the full feature set achieves the highest performance with a score of 0.86. When using only one feature group—vegetation, meteorological variables, or fire weather indices (FWIs)—performance drops to 0.80. Combining any two groups yields intermediate results, with ROC AUC scores ranging from 0.83 to 0.85.
