Abstract: Quantifying reactive transport of dissolved compounds in fluids flowing through natural porous media (e.g. aquifers and permeable sediments) is a challenging task, mainly because of heterogeneity and uncertainty spanning a variety of spatial and temporal scales. My seminar focuses on novel (process-based and possibly data-driven) approaches developed to tackle these issues. They are formulated in a probabilistic framework that allows one to combine physical constraints (physically based models of uncertainty propagation) and data, moving in space and time and across spatial and temporal scales. In particular, we explore rigorous homogenizability conditions for multicomponent reactive systems; the resulting continuum scale transport formulation retains uncertainty, which can be propagated in space and time leveraging mass conservation laws. Our general framework is used to analyze the dynamics of nitrogen compounds in the hyporheic zone at the bedform scale, in order to quantify uncertainty (e.g. in production of nitrous oxide) and to construct physically based prior distributions that lend themselves to data assimilation. Finally, we present a novel formulation for bidirectional scaling between multiresolution models based on probabilistic equivalence of a set of quantities of interest.

Bio: Francesca Boso received her doctorate in environmental engineering from the University of Trento, Italy, specializing in hydrologic sciences. During her work as a postdoc at UC San Diego and, currently, as a research scientist at Stanford University, she has been investigating uncertainty quantification for environmental applications.