PREDICTING MARINE PRIMARY PRODUCTION BY MACHINE LEARNING

Abstract

 Recently, there has been concern about the impact of declining primary production on material cycles and ecosystems. The primary production depends on phytoplankton photosynthesis, and predictive models have been constructed based on various types of monitoring of the dynamics of phytoplankton. However, it is not easy to predict phytoplankton dynamics in the ocean, which are affected by various environmental factors. In this study, we constructed a prediction model for basic production rate using machine learning, which is good at predicting the future, and evaluated its performance. The results showed that the model could be estimated with a coefficient of determination of 0.75, using dissolved iron, light intensity, PO₄-P, and NOx-N as important variables.