Evaluation of empirical and dynamic primary production estimates for the California Current System

High biological and physical variability of coastal upwelling ecosystems limits the usefulness of empirical satellite algorithms for estimating marine primary production (MPP), with errors greater than 45% (MAPE) for common empirical algorithms even after improvements from regional tuning. We investigate the performance of empirical and dynamic models for representing MPP in the California Current System (CCS). Specifically, we evaluate the vertically generalized production model (VGPM), which is based solely on sea surface temperature (SST), chlorophyll a (Chla), and solar irradiance (PAR) estimates derived from satellite. We compare those results to a vertically resolved, coupled physical and biogeochemical, data assimilative state estimate of the CCS, which assimilates extensive physical surface and subsurface information as well as satellite Chla. We evaluate the performance of data assimilation estimates relevant to the VGPM against records from the California Cooperative Oceanic Fisheries Investigations (CalCOFI), and report root mean square error for SST (0.75, 1.00 °C) and Chla (4.93, 1.60 mg/m³) between the satellite and the data assimilation state estimates, respectively. We consider the accuracy of the VGPM if satellite inputs were augmented by continuous dynamic state estimates, and we evaluate potential improvements to MPP estimation from addition of vertically-resolved Chla and light fields. Finally, we evaluate the potential to merge empirical and dynamic methods to provide a continuous MPP record that satisfies uncertainty thresholds for operational ocean measurements, for example as proposed by the Global Ocean Observing System.