[2603.25469] Not a fragment, but the whole: Map-based evaluation of data-driven Fire Danger Index models

Computer Science > Machine Learning arXiv:2603.25469 (cs) [Submitted on 26 Mar 2026] Title:Not a fragment, but the whole: Map-based evaluation of data-driven Fire Danger Index models Authors:Shahbaz Alvi, Italo Epicoco, Jose Maria Costa Saura View a PDF of the paper titled Not a fragment, but the whole: Map-based evaluation of data-driven Fire Danger Index models, by Shahbaz Alvi and 2 other authors View PDF HTML (experimental) Abstract:A growing body of literature has focused on predicting wildfire occurrence using machine learning methods, capitalizing on high-resolution data and fire predictors that canonical process-based frameworks largely ignore. Standard evaluation metrics for an ML classifier, while important, provide a potentially limited measure of the model's operational performance for the Fire Danger Index (FDI) forecast. Furthermore, model evaluation is frequently conducted without adequately accounting for false positive rates, despite their critical relevance in operational contexts. In this paper, we revisit the daily FDI model evaluation paradigm and propose a novel method for evaluating a forest fire forecasting model that is aligned with real-world decision-making. Furthermore, we systematically assess performance in accurately predicting fire activity and the false positives (false alarms). We further demonstrate that an ensemble of ML models improves both fire identification and reduces false positives. Comments: Subjects: Machine Learning (cs.LG) Cit...