Probabilistic assessment of seawater intrusion under multiple sources of uncertainty

Coastal aquifers are affected by seawater intrusion (SWI) on a worldwide scale. The Henry's problem has been often used as a benchmark to analyze this phenomenon. Here, we investigate the way an incomplete knowledge of the system properties impacts the assessment of global quantities (GQs) describing key characteristics of the saltwater wedge in the dispersive Henry's problem. We recast the problem in dimensionless form and consider four dimensionless quantities characterizing the SWI process, i.e., the gravity number, the permeability anisotropy ratio, and the transverse and longitudinal Peclet numbers. These quantities are affected by uncertainty due to the lack of exhaustive characterization of the subsurface. We rely on the Sobol indices to quantify the relative contribution of each of these uncertain terms to the total variance of each of the global descriptors considered. Such indices are evaluated upon representing the target GQs through a generalized Polynomial Chaos Expansion (gPCE) approximation. The latter also serves as a surrogate model of the global system behavior. It allows (a) computing and analyzing the joint and marginal probability density function (pdf) of each GQ in a Monte Carlo framework at an affordable computational cost, and (b) exploring the way the uncertainty associated with the prediction of these global descriptors can be reduced by conditioning of the joint pdf on available information. Corresponding analytical expressions of the marginal pdfs of the variables of interest are derived and analyzed. (C) 2014 Elsevier Ltd. All rights reserved.