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

Hydrogeophysical model calibration and uncertainty analysis via full integration of PEST/PEST++ and COMSOL

“…1). Previously, Halloran et al (2019) presented COMPEST, an interface built in Java to link PEST and COMSOL, that they satisfactorily applied to a case of isotopic fractionation of groundwater contaminants (Halloran et al, 2021). In this work we follow a different strategy, using MATLAB for connection instead of Java, and with some important additions, including the use of pilot points for spatial parametrization, Tikhonov regularization, singular value decomposition, parallelization to reduce computational burden and the integration with PEST++ that provides global sensitivity and uncertainty analysis.…”

“…3). Although the GLM algorithm was used in COMPEST (Halloran et al, 2019), the use of pilot points for spatial parametrization in COMSOL is a novelty of this work. A computational limitation, especially for highly parameterized models and long forward running times, is that the GLM algorithm requires to fill the Jacobian matrix at every iteration by computing a finitedifference approximation of the partial first derivative, which is directly related with the number of parameters, in our case the same number of pilot point locations, 636.…”

“…The strategy is achieved by integrating a fully coupled hydrogeophysical forward model developed with the commercial finite-element software COMSOL Multiphysics® and the model-independent calibration software PEST (Doherty, 2020) and PEST++ (White et al, 2020), widely used by the groundwater community. The coupling of PEST and COMSOL was performed previously by Halloran et al (2019), who developed a Java interface for communication. Here we follow a different strategy for communication using MATLAB instead of Java, and we include some novelties that were not implemented in COMPEST, such as the use of pilot points for spatial parameterization, Tikhonov regularization, application of the novel iterative Ensemble Smoother from PEST++, parallelization and a first application of the strategy to a hydrogeophysical problem.…”

Benchmarking multiphysics software for mantle convection

“…Accordingly, COMSOL® is also a viable option for introducing students to geodynamic modeling. COMSOL® has been utilized in the geosciences (Azad et al, 2016;Bird et al, 2014;Butler and Sinha, 2012;Butler and Zhang, 2016;Halloran et al, 2019;Li and Smith, 2015;Nardi et al, 2014;Pirnia et al, 2019), including mantle convection studies (Herein and Galsa, 2011;Lee, 2014;Lee and Wada, 2017;Shahraki and Schmeling, 2012;Sinha and Butler, 2009;Shiels and Butler, 2015). However, it appears the software has not been benchmarked in detail for mantle convection problems.…”

A COMSOL‐PHREEQC Coupled Python Framework for Reactive Transport Modeling in Soil and Groundwater

“…In addition, a COMSOL‐PHREEQC interface for modeling the multi‐components transport of saturated cement‐based materials was carried out to model the influence of the variation of the porosity on the ionic transport (Guo et al 2018). Besides, a novel interface, COMPEST, that couples the parameter estimation and uncertainty analysis package, PEST, with COMSOL were presented (Halloran et al 2019), and a methodological approach to perform coupled inversion by integrating the calibration software PEST/PEST++ with COMSOL was proposed using MATLAB (González‐Quirós and Comte 2021).…”