You are here:
Bayesian Networks for Assessing and Managing Risks to Ecosystems in a Changing Climate
Citation:
Carriger, J. Bayesian Networks for Assessing and Managing Risks to Ecosystems in a Changing Climate. Society of Toxicology Annual Meeting 2022, San Diego, CA, March 27 - 31, 2022.
Impact/Purpose:
This presentation provides an overview of Bayesian networks and a case study assessing spatial risks to coral reefs from climate change and local stressors. Extensions are described to the analysis and the ORD research and several EPA initiatives on climate change are briefly discussed. This is an invited presentation at a multi-disciplinary session and panel discussion at Society of Toxicology 2022.
Description:
Estimating ecological risks from climate change requires knowledge of the exposure to multiple stressors along with the susceptibility of the ecological resources to these stressors. Uncertainty and variability are essential components for calculating risks in future conditions. Bayesian networks are probabilistic tools that have been under-utilized for inferring risks. A Bayesian network approach for considering risks from multiple global- and local-source stressors will be explored. The approach will be demonstrated with examples from a case study on coral reef risks to stressors from local sources such as watershed- and marine-based pollution and global sources such as acidification and thermal stress. The networks were developed using machine learning techniques and based on augmented naïve structures to accommodate multiple predictors and collinearities in the spatial data. However, causal knowledge was also used with a direct effects assessment to examine the potential changes in the reef health indicators from changes in threats. Although developed for coral reefs, the described methods can be adapted for screening¿level risk assessments with broad and complex spatial data sets for other contexts and ecological variables. Evaluating the temporal dimensions of risks for considerations of phase shifts and ecological resilience to climate change will also be discussed. Assessing and managing the changes to ecological condition from a changing climate will benefit from the future application, use and development of Bayesian networks. The views expressed in this presentation are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.