This paper presents a surrogate modelling approach applied to a geochemical system of uranium sorption on clay, relevant for the safety assessment of nuclear waste repositories. The surrogate uses knowledge about the underlying process to recursively partition the dataset into regions of reduced dimensionality. It achieves high accuracy with a competitive prediction throughput, and is a promising method to speedup  computationally demanding coupled reactive transport models.

POET is a parallel reactive transport simulator which implements a mechanism to store and reuse previous results of geochemical simulations through distributed hash tables. POET parallelizes chemistry using a master/worker design with noncontiguous grid partitions to maximize its efficiency and load balance on shared-memory machines and compute clusters.

Barite formation is an important factor for many use cases of the geological subsurface because it may change the rock. In this modelling study, the replacement reaction of celestite to barite is investigated. The steps that were identified to play a role are celestite dissolution followed by two-step precipitation of barite: spontaneous formation of small crystals and their subsequent growth. Explicitly including the processes improve the usability of the models for quantified prediction.

RedModRphree is an R extension package to leverage the PHREEQC engine for geochemical models, providing convenience functions to efficiently setup computations and program algorithms involving geochemical models. Version 0.3.6 ships with a novel implementation of Pourbaix (potential/pH) diagram computation which considers the full speciation of the solution at each diagram point.

DecTree evaluates a hierarchical coupling method for reactive transport simulations in which pre-trained surrogate models are used to speed up the geochemical subprocess, and equation-based full-physics simulations are called only if the surrogate predictions are implausible. Furthermore, we devise and evaluate a decision tree surrogate approach designed to inject domain knowledge of the surrogate by defining engineered features based on law of mass action or stoichiometric reaction equations.