75 / 2023-07-28 12:16:04

A DEEP-LEARNING-BASED MODEL FOR PHYTOPLANKTON PIGMENT ESTIMATION

Phytoplankton pigments; Remote sensing; Deep learning; Phytoplankton community; El Niño;

Understanding phytoplankton taxonomy and community structure is critical for advancing marine ecological research and facilitating accurate global climate prediction. Pigment-based approach for phytoplankton community analysis is widely used for calibration and validation of satellite phytoplankton functional types (PFTs) retrievals. A deep-learning-based model has been proposed to estimate concentrations of 17 different phytoplankton pigments globally using satellite data. The model takes into account ocean color parameters, satellite-derived environmental factors, and the slope of above-surface remote-sensing reflectance as inputs. Validation of the model was carried out against in-situ HPLC data, demonstrating its advantages in analyzing phytoplankton community dynamics on a large spatiotemporal scale.

The model can be used to analyze the phytoplankton community dynamics on a large spatiotemporal scale, which can be useful for understanding the impact of environmental factors on the distribution of phytoplankton groups. To analyze global pigment concentrations during 2003-2021, time series analysis was performed on MODIS retrieved pigment concentrations using the established DL-PPCE model. The findings revealed that during the 2015/2016 El Niño event, the prokaryotes-dominated area extended eastward from 180°E to 150°W. Over the period from 2003 to 2021, prokaryotic abundance exhibited a positive correlation with El Niño intensity but a negative correlation with the overall abundance of the entire phytoplankton community.