Abstract

Accurate parameter estimation for reactive transport models (RTMs) is crucial for simulating subsurface hydrogeochemical processes. This study addresses the inverse problem of multicomponent solute-water-rock interactions using advanced AI-based algorithms. We propose an enhanced Tandem Neural Network Architecture integrated with an Adaptive Updating Strategy, designed to simultaneously identify reaction parameters, high-dimensional permeability fields, and optimal monitoring networks in heterogeneous aquifers. The effectiveness of the proposed inverse algorithms is validated through synthetic case studies and further demonstrated by their application to a well-documented RTM problem in the Aquia aquifer, Maryland, USA. The findings offer valuable insights into the complex dynamics of solute-water-rock interactions in subsurface systems, advancing the understanding and predictive capabilities of hydrogeochemical processes in heterogeneous media.

Biography

Dr. Zhenxue Dai earned his Ph.D. in Civil Engineering from the University of La Coruña, Spain, in 2000. From 2001 to 2004, he served as a Postdoctoral Associate at Wright State University, followed by a tenure as a Staff/Senior Scientist at Los Alamos National Laboratory from 2005 to 2017. Currently, he holds the position of Chair Professor at Jilin University.

 Dr. Dai’s research focuses on uncertainty quantification and across-scale modeling of radionuclide transport in multi-scale subsurface systems, as well as coupled simulations of fluid flow and chemical reactions in heterogeneous media. He is the Principal Scientist for a National Key Research and Development Program of China, titled “An Integrated Framework for AI-Based Real-Time Monitoring, Modeling, and Early-Warning of Contaminant Sites.” In recognition of his groundbreaking work on “Upscaling reactive transport parameters in fractured rocks”, he was elected as a Fellow of the Geological Society of America in 2010.