Optimal-Interpolation Data Fusion and Gap-Filled Satellite Background for Chlorophyll-a and Phytoplankton Functional Types along the RV Polarstern PS113 Atlantic Transect (2018)

This dataset contains daily gridded chlorophyll a products generated for the Atlantic Ocean transect of the RV Polarstern expedition PS113, conducted from 10 May to 9 June 2018. The spatial domain is restricted to a 1° buffer—approximately 100 km—surrounding the expedition track, and the grid has a spatial resolution of approximately 4 km. Two corresponding products are provided: Background field (\(x_b\)): a satellite-derived total chlorophyll-a and phytoplankton functional type product reconstructed using the Data Interpolating Convolutional Auto-Encoder (DINCAE) gap-filling method. The regional reconstructions were merged and cropped to the 1° expedition-track buffer. Analysis field (\(x_a\)): the output of an optimal interpolation data-fusion procedure that combines \(x_b\) with chlorophyll-a concentrations retrieved from underway hyperspectral absorption measurements collected with an ACS spectrophotometer during PS113. The products include total chlorophyll a and chlorophyll a concentrations associated with five major phytoplankton functional types: diatoms, dinoflagellates, haptophytes, green algae, and prokaryotic phytoplankton. Concentrations are expressed in mg m\(^{-3}\). The optimal interpolation analysis was calculated as \[ x_a = x_b + K(y-Hx_b), \] where \(y\) represents the ACS-derived in situ observations, \(H\) maps the background field to the observation locations, and \(K\) is the gain determined from the background covariance and ACS observation-error estimates. Spatial covariance localization was applied to limit unrealistic long-distance influences.