# A machine learning approach elucidates spatial patterns of environmental properties driving microbial composition over Santos Basin, South Atlantic

## TL;DR

This study uses machine learning to show how environmental factors shape microbial communities in Brazil's Santos Basin, offering a new way to monitor marine ecosystems.

## Contribution

A novel hybrid machine learning framework predicts depth-specific microbial associations with 86% accuracy in the Santos Basin.

## Key findings

- Five depth-specific microbial associations were identified, driven by temperature, salinity, density, and nutrients.
- Microbial diversity increases with depth, while cell abundance is higher in nutrient-rich shallow waters.
- Regional oceanographic processes like upwelling and plumes influence microbial community patterns.

## Abstract

Marine microbial communities are vital to biogeochemical cycling, yet their dynamics in regions of ecological and industrial significance, such as the Santos Basin (SB), Brazil’s largest offshore oil-producing basin, remain poorly resolved. To address this gap, we combined 16S rRNA amplicon sequencing, flow cytometry, and a hybrid machine learning framework (Self-Organizing Maps and Random Forest) to analyze microbial community stratification across pelagic depths in the SB. We identified five depth-specific microbial associations predicted with 86% accuracy, driven primarily by temperature, salinity, water density, and nutrient availability. Shallow epipelagic and mesopelagic zones were dominated by temperature-driven assemblages, while deeper bathypelagic communities responded to salinity and density gradients. Temporal and spatial patterns further highlighted the influence of regional oceanographic processes, including the Cabo Frio upwelling and Rio de la Plata plume. Microbial diversity increased with depth, contrasting with higher cell abundances in nutrient-rich shallow waters. We provided new insights into the relative importance of oceanographic processes, suggesting that vertical stratification and regional hydrography may play a more central role shaping microbial communities than previously recognized. We also established a predictive framework for microbial dynamics in marine ecosystems, with direct implications for assessing anthropogenic impacts in industrially active regions like the SB.

A data-driven study using machine learning reveals how environmental gradients structure depth-specific microbial communities in Brazil’s oil-rich Santos Basin, offering a predictive framework for ecosystem monitoring.