Award: OCE-1558710

The objectives of this collaborative project between MBL and MIT focused on improving our understanding and ability to model the growth and metabolism of marine plankton that produce nearly half of the oxygen we breath, provide food for almost all marine organisms, as well as maintain Earth?s life support system. As marine plankton grow, they modify the chemistry of the ocean by taking up nutrients, such as ammonium, and expelling waste products, such as oxygen for photosynthetic plankton (phytoplankton) or carbon dioxide for herbivores and carnivores. Because the microscopic organisms are intrinsically linked with chemistry, and vice versa, the field of research is known as biogeochemistry. Marine biogeochemical models attempt to predict how the presence of plankton alter the cycling and transformation of important elements, such as carbon, nitrogen, phosphorous, sulfur and many trace metals, such as iron. The project combined two types of marine biogeochemical models, trait-based models and thermodynamic-based metabolic models. Trait-based models simulate marine plankton groups, such as phytoplankton, bacteria, and zooplankton, using a Darwinian-like approach, where each group is represented by 10s to 100s of individuals with slightly different traits chosen at random, such as their maximum growth rate or optimum growth temperature. A three-dimensional (3D) ocean circulation model is then populated with a large number of plankton with randomized trait values and allowed to compete ?in silico? (i.e., in a computer simulation). Because the ocean environment varies from place to place and from time to time, plankton grow only where and when their traits match the local environment. Some plankton flourish in certain areas, while others go extinct. The overall result is a better understanding and prediction of plankton biogeography and associated biogeochemistry. Thermodynamic-based metabolic models use a different approach to simulate marine biogeochemistry based on the conjecture that systems, both living and chemical, will likely organize to maximize the dissipation rate of available energy, such as food or sunlight, or in thermodynamic jargon, they organize towards maximum entropy production (MEP) states. The organization of hurricanes to dissipate the thermal gradient between the ocean and atmosphere built up over summer is an example of MEP at work. The main difference between nonliving systems, such as hurricanes, and life is that life maximizes entropy production over time, while nonliving systems maximize entropy production instantaneously, such as a rock rolling down a hill. For marine biogeochemistry, MEP can be used to predict the activity of metabolic reactions plankton can conduct, such as fixing carbon dioxide or dinitrogen from the atmosphere. However, to solve the MEP problem over time requires solving a computation problem that is difficult even in just one dimension and is currently not practical to solve in 3D ocean circulation models. The project combined the two approaches by using a trait-based representation of plankton; however, instead of using in-silico Darwinian-like selection to set trait values, they were determined by maximizing entropy production over time (Vallino & Tsakalakis, 2020). This combined approach has a few advantages. First, MEP theory is general, so biogeochemical models based on MEP should be better at prediction than models that rely on mechanistic interactions between plankton, which we still do not understand well. Second, it solves a problem known as mortality closure. In most food-web-based marine biogeochemistry models, predator prey interactions, including mortality factors, must be specified; however, there are many ways to do this, but only one can be chosen at run time. MEP-based optimization can be used to dynamically select the appropriate predator-prey configuration from a large set of possibilities, which is closer to how natural systems organize. In Vallino & Tsakalakis (2020) we demonstrate this in zero-dimensions, but we have a way to extend this approach to 3D ocean circulation models. Our development of MEP theory for a coastal pond highlights the importance of temporal strategies in living systems, such as circadian rhythms (Vallino & Huber, 2018). To explore this in the MIT 3D model, we added day-night (diel) light cycles to the model, which previously only considered seasonal light cycles (Tsakalakis et al. 2021). We found diel light cycles induced oscillations in limiting nutrients in the lower latitudes of oceans. These daily fluctuations allow diatoms to compete with small phytoplankton, which did not occur when only seasonal light forcing was used. Consequently, we showed that the addition of diel cycles changes the biogeography of plankton and illustrates the importance of including all timescales in biogeochemical models. Project results have been presented at several professional conferences (Association for the Sciences of Limnology and Oceanography, American Geophysical Union, and Society for Mathematical Biology) and at department seminars. The project supported two postdoctoral students and involved several undergraduate students at both MIT and MBL. All model code is open source and available via GitHub (https://github.com/darwinproject, https://github.com/maxEntropyProd). Last Modified: 07/30/2021 Submitted by: Joseph J Vallino