Publication

Back to overview

COMPEST, a PEST-COMSOL interface for inverse multiphysics modelling: Development and application to isotopic fractionation of groundwater contaminants

Type of publication Peer-reviewed
Publikationsform Original article (peer-reviewed)
Author Halloran Landon J.S., Brunner Philip, Hunkeler Daniel,
Project Tracking reactive processes in low permeability sediments and their effect on contaminant longevity in aquifers using compound-specific isotope analysis
Show all

Original article (peer-reviewed)

Journal Computers & Geosciences
Volume (Issue) 126
Page(s) 107 - 119
Title of proceedings Computers & Geosciences
DOI 10.1016/j.cageo.2019.02.001

Open Access

URL https://libra.unine.ch/Publications/Daniel_Hunkeler/44702
Type of Open Access Green OA Embargo (Freely available via Repository after an embargo)

Abstract

In the geosciences, inverse problems, wherein observations corresponding to model outputs are known and model parameters are unknown, are commonplace. Many of these problems involve coupled physical, chemical, and other processes that can be modelled using forward finite-element models. Here, we present a novel interface, COMPEST, that connects the parameter estimation and uncertainty analysis package, PEST, with the finite-element modelling package, COMSOL Multiphysics. To demonstrate some of the capabilities of this approach, we also develop and present a novel model for the degradation and transport of chlorohydrocarbons in low-permeability units. This model integrates isotopic fractionation arising from degradation and diffusion. Three implementations of this model with increasing complexity are used to demonstrate the functionality of the developed interface. This linkage provides a means for parameter estimation, uncertainty analysis, and singular value decomposition to gain insight into the behaviour, identifiability, and interdependence of the various parameters in the model. COMPEST is written so as to be suited to a wide range of scientific and engineering applications and thus can be used to link any COMSOL model with PEST. This enables the use of advanced inverse modelling techniques previously unavailable to COMSOL users.
-