Aquatic Carbon Footprint

Project title: Exploring the Potential of Hyperspectral EMIT Data to Assess Phytoplankton Diversity and Carbon Dynamics in the Wetland-Estuarine Systems of the Northern Gulf of Mexico

The hyperspectral observation by the Earth Surface Mineral Dust Source Investigation (EMIT) instrument offers an untapped potential for monitoring water quality and biodiversity across diverse habitats, from estuaries and wetlands to coastal pelagic. We propose to explore the EMIT hyperspectral capabilities in studying biogeochemical processes in the wetland-estuarine system of the northern Gulf of Mexico (nGoM), which has experienced rapid changes due to multiple, pressing environmental stressors. In particular, we will test and implement advanced machine learning algorithms, such mixture of expert (MoE) and Variational Autoencoder (VAE) to fully utilize the hyperspectral information from EMIT Level 2 surface reflectance data to infer phytoplankton-related metrics including, phytoplankton absorption coefficient (aphy), chlorophyll a (Chl-a) and phytoplankton community composition (PCC). The EMIT algorithms will be developed and validated using a comprehensive optical and biogeochemical dataset to be acquired through HydroLight simulation, uncrewed autonomous observation and lab analysis.

Our overarching goal is to apply novel machine learning retrieval algorithms to enhance the estimation of biogeochemical variables using remotely sensed hyperspectral measurements. To achieve the goal, we propose to accomplish the following specific objectives:

Objective 1: Collect hyperspectral measurements for algorithm development and validation by deploying an autonomous optical system (AOS) and ground-based field experiment and lab analysis.

Objective 2: Develop and validate AI-based algorithms for retrieving aphy, Chl-a, and PCC, along with other optically active biogeochemical parameters, including colored dissolved organic matter (CDOM) and total suspended solids (TSS).

Objective 3: Integrate AI-based algorithms for EMIT into an open-source python package (HyperCoast) and characterize carbon footprints across various wetland-estuarine systems in the northern Gulf of Mexico (nGoM).

Project Team

	Bingqing Liu
	University of Louisiana Lafayette	bingqing.liu@louisiana.edu https://bingqingliu.com https://hypercoast.org
	Eurico D’Sa
	Louisiana State University	ejdsa@lsu.edu
	Xiaodong Zhang
	University of Southern Mississippi	Xiaodong.Zhang@usm.edu
	Melissa M. Baustian
	U.S. Geological Survey	mbaustian@usgs.gov
	Xu Yuan
	University of Delaware	xyuan@udel.edu