Detection and characterization of microalgae blooms based on multi-source information from in situ measurements and Earth Observation imagery: From data to end-user information

In the frame of the ESA Innovation project EO4HAB (Earth Observation for Harmful Algal Blooms detection and characterization, ref. ESA CfP/6-60008/23/I-DT-bgh), we propose to develop a pre-operational tool for the detection, characterization and prediction of microalgae blooms for aquaculture professional thanks to:

• In situ physico-chemical and spectroradiometry data,
• Multispectral and Hyperspectral satellite imagery.

This project involves two complementary French SME companies: Hytech-imaging and BiOceanOr. While the ambition of Hytech-imaging is to develop and generalize the uses of spectral imaging, BiOceanOr aims to offer water quality monitoring and forecasting services to its customers.

Several innovations are addressed in this project:

• Exploitation of in situ spectroradiometry data (hydraspectra – CSIRO).
• Exploitation of a very large database of samples in Chile (MOWI).
• Development of species-specific bio-optical models.
• Take into account adjacency effects.
• Explore several approaches for bloom prediction.
• Evaluate the interest of high revisit EO (Earth Observation) from nanosatellite constellations.

The methodology includes two main steps:

• Detect and characterize the blooms based on the in situ spectroradiometry (hydraspectra – CSIRO) and physico-chemical (MOWI) data.
• Use the in situ information to extend the detection and characterization using multispectral (S2/S3/VIIRS) and hyperspectral (PRISMA/EnMAP) satellite data.

For both steps, physics-based (radiative transfer modeling and inversion), empirical (spectral indices) and data-driven (machine and deep learning) approaches are considered.

Two sites are studied: Spencer Gulf (Australia) and Lake Region (Chile).

The main challenges addressed are:

• Use of in situ detection and characterization data to calibrate the estimation of the satellite BOA (Bottom Of Atmosphere) reflectance. In particular, species-specific bio-optical models must be implemented.
• Consideration of adjacency effects that can affect the satellite BOA reflectance, with use of a distance-to-the-coast metric. Evaluation of atmospheric correction imperfections as a source of uncertainties (aerosols, reflections, etc.).
• Consideration of sensor noise that affects the satellite BOA reflectance.
• Estimation and propagation of uncertainties that are essential to demonstrate the usability of the data.

The presentation will summarize the results of the EO4HAB project, including in situ detection, large scale bloom detection, characterization of blooms on a local scale and prediction.