Investigating biophysical control of marine phytoplankton dynamics via Bayesian mechanistic modeling

Marine phytoplankton possess a key position in multiple ecological, biological, and environmental processes; therefore, constructing a systematic and comprehensive mechanistic understanding of the leading mechanisms and key biophysical drivers of phytoplankton dynamics is crucial for evaluating biophysical forcings and ecological responses. Here, we develop a process-based model to improve the characterization of controlling mechanisms and assess the relative importance of nutrients, light, temperature, and oceanic mixing for explaining chlorophyll-a variability in the western North Pacific Ocean. The model is calibrated using 20-year observational data of chlorophyll-a concentrations from 1993 to 2012 in a Bayesian framework to provide data-driven inference and a rigorous uncertainty quantification of important biophysical rates. The model explains approximately 61% of the variability in chlorophyll-a. Our results, supported by cross-validation, show that the chlorophyll-a concentration is much more responsive to temperature variability than nutrients, light, or mixing. By synthesizing the effects of multiple nutrients (i.e., nitrogen, phosphorus, silica, and iron) on phytoplankton dynamics, we find that iron and nitrogen largely determine the chlorophyll-a variability. Sensitivity analysis shows that future warming conditions with +3°C change in temperature will impede phytoplankton production by 24% due to an intensified phytoplankton mortality impact. Our findings highlight the potential of “deductive-and-inductive” modeling approaches combining embedded biophysical mechanisms and probabilistic uncertainty quantification to effectively integrate and leverage sporadic ocean monitoring efforts and ultimately improve the marine water quality predictions.