You are here:
Exploring coral reef communities in Puerto Rico using Bayesian networks
Citation:
Carriger, John F. AND William S. Fisher. Exploring coral reef communities in Puerto Rico using Bayesian networks. Ecological Informatics. Elsevier Science Ltd, New York, NY, 82:102665, (2024). https://doi.org/10.1016/j.ecoinf.2024.102665
Impact/Purpose:
This journal article presents a Bayesian network machine learning analysis approach to examining biological monitoring data for coral reefs. The first phase uses the Bayesian network for exploratory analysis of the strengths and types of connections in the database. Next the data are clustered based on community components (gorgonians, corals, sponges, fishes) of the reef. Finally, an overall cluster for all of the community components is identified. The techniques used for this are under explored with coral reef community assessments and may be useful tools for future coral reef surveys and other types of biological assessments, particularly for identifying community types and risk assessment endpoint selection for ecological communities.
Description:
Most coral reef studies focus on scleractinian (stony) corals to indicate reef condition, but there are other prominent assemblages that play a role in ecosystem structure and function. In Puerto Rico these include fish, gorgonians, and sponges. The U.S. Environmental Protection Agency conducted unique surveys of coral reef communities across the southern coast of Puerto Rico that included simultaneous measurement of all four assemblages. Evaluating the results from a community perspective demands endpoints for all four assemblages, so patterns of community structure were explored by probabilistic clustering of measured variables with Bayesian networks. Most variables were found to have stronger associations within than between taxa, but unsupervised structure learning identified three cross-taxa relationships with potential ecological significance. Clusters for each assemblage were constructed using an expectation-maximization algorithm that created a factor node jointly characterizing the density, size, and diversity of individuals in each taxon. The clusters were characterized by the measured variables, and relationships to variables for other taxa were examined, such as stony coral clusters with fish variables. Each of the factor nodes were then used to create a set of meta-factor clusters that further summarized the aggregate monitoring variables for the four taxa. Once identified, taxon-specific and meta-clusters represent patterns of community structure that can be examined on a regional or site-specific basis to better understand risk assessment, risk management and delivery of ecosystem services.
