Inter-Research > MEPS > v664 > feature  
Marine Ecology Progress Series

via Mailchimp
Linking spatiotemporal information to North Sea cod recruitment via machine learning.
Image: B. Kühn, M. Taylor; Gears from
https:// commons.
(CC-BY-SA license)

Kühn B, Taylor MH, Kempf A


Using machine learning to link spatiotemporal information to biological processes in the ocean: a case study for North Sea cod recruitment


Environmental processes on different temporal and spatial scales shape the life cycle of many marine organisms. Given the complexity of the interactions, identifying environmental variables that best explain biological processes can be like searching for a needle in the haystack. Kühn and co-authors propose a regression-type machine learning framework to extract information from spatiotemporal environmental data and link it to biological data via dimension reduction, multi-objective genetic algorithm and cross-validation procedures. When applied to the case study of North Sea cod recruitment, the algorithm identified spawning stock biomass, sea surface temperature and salinity as important factors in different seasons, likely relating to specific life-history stages during the recruitment year.


Abstract   Back to contents page   Link to full PDF