Global patterns and predictions of seafloor biomass using random forests

Area of Science:

• Marine Ecology
• Oceanography
• Biogeography

Background:

• A comprehensive global seafloor biomass and abundance database was compiled from 24 international oceanographic institutions.
• Existing data gaps hindered a complete understanding of deep-sea benthic ecosystems and their role in global processes.

Purpose of the Study:

• To model and predict global seafloor standing stocks (biomass and abundance) for various faunal groups.
• To generate detailed maps illustrating the spatial distribution of benthic life.
• To identify key environmental drivers influencing seafloor ecosystems.

Main Methods:

• Utilized the Random Forests machine-learning algorithm to predict seafloor standing stocks.
• Input variables included surface primary production, particulate organic matter (POM) flux, seafloor relief, and bottom water properties.
• Generated predictive models explaining 63% to 88% of stock variance across major size groups.

Main Results:

• Created global maps of predicted biomass and abundance for bacteria, meiofauna, macrofauna, and megafauna.
• Benthic standing stocks positively correlated with surface primary production and particulate organic carbon (POC) flux to the seafloor.
• Highest biomass observed at the poles, continental margins (coastal upwelling), and equatorial divergence zones; lowest on abyssal plains.
• Depth-related shifts in biomass dominance are driven by decreasing average body size, not abundance, linked to food availability.

Conclusions:

• The developed biomass census and maps provide critical data for deep-sea food web models and global carbon cycling studies.
• Findings offer fundamental information for evidence-based marine management and conservation strategies.
• Understanding benthic biomass distribution is key to assessing ecosystem health and function.