Moving toward finer scales in oceanography: Predictive linear functional model of Chlorophyll a profile from light data

• Prediction of Chlorophyll-a profiles is achieved in Antarctic ocean.
• A functional linear model is constructed using light curves as covariate.
• The predictive capabilities of the model associated to confidence intervals show that it is possible to predict Chlorophyll-a at fine scale.
• Chlorophyll-a prediction at fine scale highlights sub-mesoscale variations.

The Southern Ocean plays a key role in ocean–atmosphere carbon dioxide fluxes. Estimation of carbon exchanges between ocean and atmosphere must rely on accurate estimations of primary productivity which require measurements of phytoplankton concentration within the water column. In this paper, we are interested in relationships between primary productivity and light in the Antarctic ocean. The originality of this work is twofold. Starting from physical hypothesis, a statistical model is constructed for the prediction of Chlorophyll a (Chl a) profiles where light profiles are used as a covariate. Taking into account of the functional nature of the data, solutions are proposed to estimate continuous vertical profiles from discrete data sampled by elephant seals equipped with a new generation of oceanographic tags. Bootstrapped prediction intervals show a good quality of prediction of Chl a profiles, giving access to the shape of the profiles along depth and to the submesoscale structure of phytoplankton within the euphotic layer of the Southern Ocean.