Presenter: Camila Serra-Pompei (MIT)

Description:
Several satellite-derived algorithms estimate phytoplankton carbon biomass (C_phyto) by using the particulate backscattering coefficient (b_bp). However, the paucity and spatial/temporal bias in the underlying observations limits our ability for validation. Yet, these algorithms are still used to estimate rates (such as primary production) that are widely used in ecological studies. Here, we investigate the drivers of the b_bp signal and explore the performance of b_bp-based algorithms that estimate C_phyto. We also investigate if Chl-b_bp relationships can be used to derive C_phyto, since the amount of available data is much larger. To do so, we combine observations from ocean floats (BGC-Argo) and a global ocean circulation model (MITgcm) that includes plankton groups and associated inherent optical properties. We show that there is a background b_bp signal associated to non-algal particles that overrides the phytoplankton signal in low-productivity regions. In terms of algorithm performance, b_bp-based algorithms tend to have a relative mean absolute error (rMAE) close to 30% at low latitudes and perform worse at high latitudes (rMAE between 30% and 50%). Sampling biases towards low latitudes tend to underestimate C_phyto, but still perform fairly similar to the non-biased algorithm. Deriving C_phyto from Chl-b_bp relationships also performs fairly well. Next, we derive a C_phyto algorithm from the BGC-Argo data. We find that the algorithm has a similar slope to other existing algorithms. Finally, we show that the main differences between algorithms lay in the low b_bp regions, where the differences can be of up to one order of magnitude. Overall, by using the mechanistic model we generate hypotheses that can be tested in field campaigns, and assess biases of algorithms commonly used in global primary production models.