You are here:
The ladder of causation for risk assessment and decision making with wildfire ecology
Citation:
Carriger, John F., M. Thompson, AND Mace G. Barron. The ladder of causation for risk assessment and decision making with wildfire ecology. SETAC Europe 31st Annual Meeting, N/A, N/A, May 03 - 06, 2021.
Impact/Purpose:
Provide a platform presentation on recent research into causal modelling for wildfire risk assessment and management. This presentation will be at the SETAC SciCon3 virtual conference.
Description:
A conceptual ladder of causation was introduced in Judea Pearl’s “Book of Why” for classifying causal queries by the amount and types of causality used. The first level is seeing, the second is doing and the third level of the ladder is imagining. Causal Bayesian network or influence diagram models can be built for addressing questions at all rungs of the ladder of causation but their importance is compounded for doing and imagining. To examine the implications of these concepts for risk assessment, causal demonstrative structures that serve each rung for both assessment and decision models were developed in hypothetical risk-based case study applications for the ecological effects of wildfires. Here we explore the possibilities for using Bayesian networks for assessing wildfire impacts to ecological systems through levels of causal representation and scenario examination. Understanding causal linkages between wildfires and drivers of the cascading effects on ecological systems remains a critical need in fire management and impact assessment. The key benefit that will be highlighted in using causal models for each rung of the ladder is the capability to incorporate knowledge and concerns from multi-disciplinary experts and stakeholders. Ultimately, Bayesian networks may facilitate understanding the factors contributing to fire susceptibility and resilience, and the prediction and assessment of wildfire risks to and impacts on ecosystems. The flexibility and quantitative capabilities of Bayesian networks can accommodate most models used by frequentist statistics and machine learning but go beyond other analytic structures through easy and powerful causal representation and calculations for wildfire risk assessment and management questions at each level of the ladder.