You are here:
Assessing Fuel Treatment Performance with Probabilistic Inference
Citation:
Thompson, M., J. Carriger, M. Barron, K. Vogler, AND J. Scott. Assessing Fuel Treatment Performance with Probabilistic Inference. 9th International Fire Ecology and Management Congress, N/A, OH, November 30 - December 04, 2021.
Impact/Purpose:
Present extreme event attribution concepts for a session we are chairing at the 9th International Fire Ecology and Management Congress. Causal counterfactual theory will be the focus of the slides developed by the EPA authors. The content of the slides provides an overview of powerful tools available for assessing and managing risks from extreme wildfires.
Description:
Fire behavior modeling is widely used to comparatively assess risk under alternative fuel treatment scenarios, with increasing use of probabilistic ensembles. Although the modeling frameworks are naturally aligned for causal inference by isolating differences between ensembles, causal math and reasoning has not typically been applied. Here we borrow from the climate event attribution literature to address causality in fuel reduction and risk mitigation treatments using simulation modeling. To do so we use ensembles of model simulations representing (1) observed fire potential, which simulates the actual occurrence probability of exceeding a threshold for annual area burned given the current untreated landscape; and (2) counterfactual fire potential, representing an alternative world where the landscape is treated. In this presentation we will introduce key aspects of causal counterfactual theory, describe methods and examples of extreme event attribution in climate science, and illustrate translation of these ideas into prospective evaluation of fuel treatment performance. We will summarize simulated results from a landscape in northern New Mexico, USA, and offer ideas for future expansions and improvements with the methods.