Global chlorophyll a concentrations of phytoplankton functional types with detailed uncertainty assessment using multi-sensor ocean color and sea surface temperature products

Abstract

<p>With the extensive use of ocean color (OC) satellite products, diverse algorithms have been developed in the past decades to observe the phytoplankton community structure in terms of functional types, taxonomic groups and size classes. There is a need to combine satellite observations and biogeochemical modelling to enable comprehensive phytoplankton groups time series data and predictions under the changing climate. A prerequisite for this is continuous long-term satellite observations from past and current OC sensors with quantified uncertainties are essential to ensure their application. Previously we have configured an approach, namely OLCI-PFT (v1), to globally retrieve total chlorophyll a concentration (TChl-a), and chlorophyll a concentration (Chl-a) of multiple phytoplankton functional types (PFTs). This algorithm is developed based on empirical orthogonal functions (EOF) using satellite remote sensing reflectance (Rrs) products from the GlobColour archive (https://www.globcolour.info/). The algorithm can be applied to both, merged OC products and Sentinel 3A OLCI data. Global PFT Chl-a products of OLCI-PFT v1 are available on CMEMS under Ocean Products since July 2020. Lately we have updated the approach and established the OLCI-PFT v2 by including sea surface temperature (SST) as input data. The updated version delivers improved global products for the aforementioned PFT quantities. The per-pixel uncertainty of the retrieved TChl-a and PFT Chl-a products is estimated and validated by taking into account the uncertainties from both input data (satellite Rrs and SST) and model parameters through Monte Carlo simulations and analytical error propagation. The uncertainty of the OLCI-PFT products v2 was assessed on a global scale. For PFT Chl-a products this has been done for the first. The uncertainty of OLCI-PFT v2 TChl-a product is in general much lower than that of the TChl-a product generated in the frame of the ESA Ocean Colour Climate Change Initiative project (OC-CCI). The OLCI-PFT algorithm v1 and v2 have also been further adapted to use a merged MODIS-VIRRS input. Good consistency has been found between the OLCI-PFT products derived from using input data from the different OC sensors. This sets the ground to realize long-term continuous satellite global PFT products from OLCI-PFT. Satellite PFT uncertainty, as provided for our products, is essential to evaluate and improve coupled ecosystem-ocean models which simulate PFTs, and furthermore can be used to improve these models directly via data assimilation.</p>

Seasonal dynamics of major phytoplankton functional types in the coastal waters of the west coast of Canada derived from OLCI Sentinel 3A

“…The Chla concentration of PFTs was derived from OLCI satellite imagery using the EOF-based algorithm by assessing the variance or spectral shape in the OLCI R rs spectra (Bracher et al, 2015). This algorithm was recently tested globally by Xi et al (2020), and we further adapted it considering the local matchups between OLCI R rs and the CHEMTAX-derived PFTs Chla. A schematic workflow of the EOF-PFT algorithm is shown in Figure 2.…”

“…Step 1: OLCI R rs data at 10 bands (400, 412, 443, 490, 510, 560, 620, 665, 674, 681 nm) were standardized by subtracting the spectral average and normalized by the spectral standard deviation according to Xi et al (2020). The standardized R rs was then subjected to a singular value decomposition (SVD) to derive the EOF scores (U), singular values (L), and EOF loadings (V).…”

“…The EOF approach is a statistical technique that reduces the high dimensionality of spectral data into dominant modes, explaining the variance of structures in the data. Specifically, the EOF approach allows the extraction of the spectral signature associated with specific PFTs, thus enabling the development of empirical models for various PFTs (e.g., Xi et al, 2020). This approach has been successful in retrieving the concentration of different accessory pigments from Medium Resolution Imaging Spectrometer (MERIS) and hyperspectral data (Bracher et al, 2015), Chla concentration of multiple PFTs using GlobColor merged and OLCI data (Xi et al, 2020), and cell abundance of Prochlorococcus, Synechococcus, and autotrophic picoeukaryotes based on multispectral and hyperspectral data (Lange et al, 2020).…”

“…Specifically, the EOF approach allows the extraction of the spectral signature associated with specific PFTs, thus enabling the development of empirical models for various PFTs (e.g., Xi et al, 2020). This approach has been successful in retrieving the concentration of different accessory pigments from Medium Resolution Imaging Spectrometer (MERIS) and hyperspectral data (Bracher et al, 2015), Chla concentration of multiple PFTs using GlobColor merged and OLCI data (Xi et al, 2020), and cell abundance of Prochlorococcus, Synechococcus, and autotrophic picoeukaryotes based on multispectral and hyperspectral data (Lange et al, 2020). However, most of these studies were performed in open oceans or optically Case-1 type waters, with less application to coastal systems (e.g., Craig et al, 2012).…”

“…Here, we adapted an EOF-based algorithm developed by Xi et al (2020) to the Strait of Georgia, British Columbia (BC), Canada, to retrieve TChla concentration and the Chla concentration of multiple PFTs. The model was developed for each PFT using the matchup between CHEMTAX-derived PFTs and EOF scores derived from the OLCI remote sensing reflectance (R rs ; sr -1 (l)).…”