You are here:
A Probabilistic Ecological Community Analysis for Coral Reef Systems in Puerto Rico
Citation:
Carriger, John F. AND William S. Fisher. A Probabilistic Ecological Community Analysis for Coral Reef Systems in Puerto Rico. SETAC North America 43rd Annual Meeting, Philadelphia, PA, November 13, 2022.
Impact/Purpose:
These are slides for a Society of Environmental Toxicology and Chemistry (SETAC) North America 2022 presentation on Bayesian networks and coral reef research. The presentation provides a unique perspective on data analysis for ecological interpretation of biological field measurements and provides methods that can be further developed for community assessment endpoints.
Description:
Effects analysis in ecological risk assessment requires choosing endpoints that are representative of ecological communities. Indicators of different components of the community and summary indicators can both be chosen. However, summary indicators applied to a complex system can sometimes result in metrics that are ambiguous due to mathematical difficulties and loss of interpretability. Coral reefs are complex ecosystems, so patterns of ecological interactions were explored by probabilistic clustering of reef monitoring variables with Bayesian networks. This could provide objective insights that would be undetected from informal and subjective indicator approaches. In 2010 and 2011, scientists from the U.S. Environmental Protection Agency sampled coral reef communities along the coast of Puerto Rico with probabilistic surveys. These data were used in a multivariate analysis with Bayesian networks. Relationship analysis approaches such as a maximum spanning tree Bayesian network structures were used to detect and characterize correlations. Most of the sub-community variables such as gorgonians, sponges, fish, and coral variables were found to have stronger associations within than between sub-communities. Complexity weights for the score-based learning algorithm were lowered to ensure a fully connected network that contained all variables. A second phase used a variable clustering analysis to identify clusters for sponge, gorgonian, stony coral and fish monitored variables. These clusters were constructed using an expectation-maximization algorithm that created a factor node for each taxon with clusters identified by the algorithm. The clusters were interpreted in terms of their relationship with the monitoring variables used in their construction and the relationship to the monitoring variables for other taxa, such as stony coral clusters with fish variables. Each of these factor nodes were then used to create a meta-factor variable that further summarized the entirety of the coral reef community monitoring variables. Further work can be used to examine the clusters identified on a regional or site-specific basis and interpret the cluster characteristics in terms of known sources and levels of stressors to support ecological risk assessment and coral reef protection decisions.