Machine learning sheds light on physical-chemical and biological parameters leading to Abrolhos coral reef microbialization

Abstract

Microbes play a central role in coral reef health. However, the relative importance of physical-chemical and biological processes in the control of microbial biomass are unknown. Here, we applied machine learning to analyze a large dataset of biological, physical, and chemical parameters (N = 665 coral reef seawater samples) to understand the factors that modulate microbial abundance in the water of Abrolhos reefs, the largest and richest coral reefs of the Southwest Atlantic. Random Forest (RF) and Boosted Regression Tree (BRT) models indicated that hydrodynamic forcing, Dissolved Organic Carbon (DOC), and Total Nitrogen (TN) were the most important predictors of microbial abundance. The possible cumulative effects of higher temperatures, longer seawater residence time, higher nutrient concentration, and lower coral and fish biomass observed in coastal reefs resulted in higher microbial abundance, potentially impacting coral resilience against stressors.

Keywords: Abrolhos Bank; Benthic cover; Fish biomass; Hydrodynamics; Microbial abundance; Temperature.