Information content of in situ and remotely sensed chlorophyll-a: Learning from size-structured phytoplankton model

Chlorophyll-a measurements in the form of in situ observations and satellite ocean colour products are commonly used in data assimilation to calibrate marine biogeochemical models. Here, a two size-class phytoplankton biogeochemical model, with a OD configuration, was used to simulate the surface chlorophyll-a dynamics (simulated surface Chl-a) for cyclonic and anticyclonic eddies off East Australia. An optical model was then used to calculate the inherent optical properties from the simulation and convert them into remote-sensing reflectance (R-rs). Subsequently, R-rs was used to produce a satellite-like estimate of the simulated surface Chl-a concentrations through the MODIS OC3M algorithm (simulated OC3M Chl-a). Identical parameter optimisation experiments were performed through the assimilation of the two separate datasets (simulated surface Chl-a and simulated OC3M Chl-a), with the purpose of investigating the contrasting information content of simulated surface Chl-a and remotely-sensed data sources. The results we present are based on the analysis of the distribution of a cost function, varying four parameters of the biogeochemical model. In our idealized experiments the simulated OC3M Chl-a product is a poor proxy for the total simulated surface Chl-a concentration. Furthermore, our result show the OC3M algorithm can underestimate the simulated chlorophyll-a concentration in offshore eddies off East Australia (Case I waters), because of the weak relationship between large-sized phytoplankton and remote-sensing reflectance. Although Case I waters are usually characteristic of oligotrophic environments, with a photosynthetic community typically represented by relatively small-sized phytoplankton, mesoscale features such as eddies can generate seasonally favourable conditions for a photosynthetic community with a greater proportion of large phytoplankton cells. Furthermore, our results show that in mesoscale features such as eddies, in situ chlorophyll-a observations and the ocean colour products can carry different information related to phytoplankton sizes. Assimilating both remote-sensing reflectance and measurements of in situ chlorophyll-a concentration reduces the uncertainty of the parameter values more than either data set alone, thus reducing the spread of acceptable solutions, giving an improved simulation of the natural environment.