Offshore Freshened Groundwater Prospecting using Machine Learning

Offshore freshened groundwater (OFG) refers to fluids stored in sediment pores and rock fractures below the seafloor, with a salinity lower than seawater.

This phenomenon has been identified globally in continental margins and proposed as a resource that can potentially alleviate water stress in coastal regions. However, the scarcity of data to constrain the distribution and volumes of the reservoirs remains a challenge.

OPTIMAL project aims to: (i) develop an interdisciplinary methodology to predict the occurrence and distribution of OFG resources built on Artificial Intelligence and (ii) apply the model globally to infer OFG occurrence and quantify the resource feasibility as a function of distribution characteristics such as offshore extent, depth below the seafloor and fresh to brackish water ratio.

The proposed methodology uses a surrogate model to create a dataset of input parameters, representing key geological and geomorphological components influencing OFG systems, such as aquitard thickness and seafloor bathymetry.

The output data will be generated via numerical simulation of variable-density groundwater transport on the suite of surrogate models using high-performance computing. These data will be used to train and test machine learning algorithms. The successful models will be validated using real-world data from the existing global OFG database.

The predictive model proposed in this fellowship contributes to achieving Sustainable Development Goals related to technologies for improving access to water resources. The primary beneficiary of this funding will be the University of Malta. Partner organizations will be Utrecht University and Deltares in The Netherlands.

The action presents a unique opportunity for the fellow to transfer his expertise in stochastic reservoir modelling and characterization of OFG systems to the host, while learning about marine geology, seafloor landforms and applied machine learning.