The goal with my project is to combine machine learning, ocean color products, and Lagrangian particle tracking to bridge the different spatial and temporal scales that affect primary production in the ocean. In earlier work we have developed a framework where we seed virtual particles in existing velocity fields, attach satellite data to the particles, and calculate rates of changes along the trajectories. I will expand this approach by analyzing each trajectory as sparse time series using modern ML-based time series models to create robust estimates and forecasts of fluxes and rates. The resulting rates will be compared with estimates from exiting CBIOMES model simulations. I will also use In-Situ drifter data from the Global Drifter Program to minimize the inherent problem of short de-correlation timescales between simulated drifters and the physical water mass they are assumed to follow. Part of the work will be to explore primary production over hourly timescales. Earlier work has used data from Geostationary satellites and found large differences in production rates on diurnal time scales. The results did, however, generate more questions than answers… I will combine machine learning models with hourly Ocean Color products from the Korean GOCI and GOCI2 geostationary satellites to explore how productivity rate changes over the day. By comparing biomass at the end of the dat and the beginning of the next day, I hope to also estimate community night respiration.