loading page

Machine learning to predict final fire size at the time of ignition
  • Shane Coffield
Shane Coffield
University of California, Irvine

Corresponding Author:scoffiel@uci.edu

Author Profile


The boreal forests of Alaska have been experiencing a changing fire regime which threatens human lives and vulnerable ecosystems. Given expected increases in fire activity with climate warming, insight into the controls on fire size from the time of ignition could provide guidance for decision support. Such insight may be especially useful in cases where many ignitions occur in a short time period. Here we investigated the controls and predictability of final fire size at the time of ignition. Using decision trees, we show that ignitions can be classified as leading to small, medium, or large fires with 50.4 ± 5.2% accuracy in cross-validation. This was accomplished using two variables: vapor pressure deficit (VPD) and the fraction of spruce cover near the ignition point. The model predicted that 40% of ignitions would lead to large fires, which accounted for 75% of the total burned area. Other machine learning classification algorithms, including random forests and multi-layer perceptrons, were tested but did not outperform the simpler decision tree model. Applying the model to areas with intensive human management resulted in overprediction of large fires. The overprediction is explained by (1) suppression of those fires and (2) the fact that ignitions in more human-influenced areas occurred during periods of higher VPD on average. Overall, this type of simple classification system could offer insight into optimal resource allocation, helping to maintain a historical fire regime and protect Alaskan ecosystems.
2019Published in International Journal of Wildland Fire volume 28 issue 11 on pages 861. 10.1071/WF19023