Evaluating and optimizing the use of logistic regression for tree mortality models in the First Order Fire Effects Model (FOFEM)

Wildland fires burn millions of forested hectares annually around the world, affecting biodiversity, carbon storage, hydrologic processes, and ecosystem services largely through fire-induced tree mortality (Bond-Lamberty et al. 2007; Dantas et al. 2016). In spite of this widespread importance, the underlying mechanisms of fire-caused tree mortality remain poorly understood, (Hood et al. 2018). Post-fire tree mortality has been traditionally modeled as an empirical function of tree defenses (bark thickness) and fire injury (crown scorch, stem char) (Ryan and Amman 1996; Woolley et al. 2012). Empirical models are commonly used in fire management to predict fire effects (Reinhardt et al. 1997), from the finescale software tools for fire management planning, to process-based succession models (Keane et al. 2011), and global models of the terrestrial carbon cycle (Hantson et al. 2016). Nevertheless, many fire-caused tree mortality models have undergone little evaluation.