Study of ENMAP imagery for the application of methods for Phytoplankton Functional Types determination in coastal waters

Monitoring and mapping of major phytoplankton groups (PG) or functional types (PFT) has been targeted as a relevant topic for the understanding and study of marine ecosystem, especially under the present climate change scenario. Developing of algorithms that determine the structure of the phytoplankton communities is a reality since the last 20 years, but not much advanced has been done in the field of image spectroscopy due to the lack of spaceborne sensors with a systematic high temporal and spatial scales. Some operational sensors are changing the game right now, like PRISMA, ENMAP and, in the future, SGB, CHIME and other developments in the hyperspectral dimension. Most of the approaches for determining PFT or PG are based on phytoplankton abundance, cell size or bio-optical properties that use chlorophyll-a or spectral features (absorption, backscatter, and/or reflectance) on water in the VIS-NIR range as inputs. These and other approaches based on machine learning and deep learning are being tested on ENMAP imagery over the Baltic Sea. We will use reflectance data provided by DLR. Comparison of atmospheric correction approaches seems to be a necessary step, and radiance data will be process with current available algorithms. Since ENMAP has 240 bands, and high spatial resolution (30 m), we will tackle the dimensionality reduction problem adapting well-known machine learning approaches to the sensor characteristics (https://isp.uv.es/soft_feature.html).