Are existing size-fractionated primary production models appropriate for UK shelf Seas?

The shelf seas are some of Earth’s most productive oceanic environments. Though they comprise <10% of the oceans’ area, it is estimated that they contribute between 15 – 20% to annual global marine primary productivity and support the majority of commercial fisheries. This primary production is modified by changes in the taxonomic composition and size spectrum of phytoplankton, which in turn influences pelagic biogeochemistry, by altering elemental stoichiometry and carbon export efficiency. Deriving primary production from remote sensing data has undoubtedly aided our understanding of these processes. However, few models have been developed for estimating taxonomic, size class or functional type-specific primary production in the shelf seas. Current methods to derive size-fractionated primary production depend on modelling size class-specific parameters from the entire community by using measurements such as diagnostic pigment concentrations. There is a paucity of direct photo-physiological measurements for distinct size classes in situ, and therefore limited data for validation of bio-optical production models applied to these dynamic regions. In this paper we use bio-optical parameters for three phytoplankton size classes (>20 µm, 20-2.0 µm and 2.0-0.2 µm) measured at two time-series stations in the Western English Channel during Spring 2014 to late Summer 2015 and from two cruises in the Celtic Sea during August 2014 and April 2015 to validate an existing model for deriving size-fractioned primary production from satellite data. The results indicate that current models of size-fractionated primary production, which are calibrated and validated using open ocean samples, are not accurate for the UK shelf seas due to errors in the assignment of size-specific photosynthetic parameters. Based on the size-fractionated bio-optical measurements made in the Celtic Sea and Western English Channel, we suggest ways in which these current models can be improved.