Predictive modeling of coral disease distribution within a reef system

Area of Science:

• Marine biology
• Ecology
• Statistical modeling

Background:

• Coral diseases have complex environmental associations due to varied causes and host-pathogen dynamics.
• Statistical modeling is underused for understanding spatial patterns in coral disease ecology.
• Research is needed to explore how environmental factors and their interactions influence distinct coral diseases.

Purpose of the Study:

• To test if coral diseases associate distinctly with environmental factors.
• To determine the importance of environmental variable interactions in predicting coral disease spatial patterns.
• To compare the predictive accuracy of modeling individual diseases versus combined disease prevalence.

Main Methods:

• Modeled four coral diseases (Porites growth anomalies, Porites tissue loss, Porites trematodiasis, Montipora white syndrome) and their interactions with 17 environmental predictors.
• Utilized boosted regression trees (BRT) statistical modeling in a Hawaiian reef system.
• Assessed predictive performance using cross-validation, comparing individual disease models with a combined disease prevalence model.

Main Results:

• Each coral disease demonstrated unique associations with environmental predictors, including biotic and abiotic factors.
• Turbidity, butterflyfish abundance, juvenile parrotfish abundance, and Porites host cover were key predictors for specific diseases.
• Incorporating interactions among predictors enhanced model predictive power, especially for Porites trematodiasis.
• Modeling combined disease prevalence resulted in a six-fold increase in predictive error compared to individual disease models.

Conclusions:

• Coral diseases should be modeled individually to improve predictive accuracy, unless their etiologies are known to respond similarly to environmental conditions.
• Statistical modeling, particularly when accounting for variable interactions, is a valuable tool for understanding coral disease ecology globally.
• Distinct environmental drivers necessitate separate modeling approaches for different coral diseases to accurately predict spatial patterns and inform conservation efforts.