Author(s): Liam Duncan; Abbas El-Zein
Linked Author(s):
Keywords: No Keywords
Abstract: The recent emergence of several colloidal contaminants in the subsurface has prompted renewed interest in characterising the transport of colloids in porous media (Wong et al., 2020). This transport is modelled using the colloidal version of the reaction diffusion advection equation (CRDAE), which takes into account colloidal attachment and detachment kinetics. Several equation parameters can only be determined by solving the inverse CRDAE problem (Bradford et al., 2009). The most commonly used inversion algorithms generate best-fit estimates of unknown parameters but without any measure of uncertainty (Barati Moghaddam et al., 2021). An assessment of uncertainty is important as it would allow a probabilistic approach to be taken in predictive models and measures of risk to be generated (Barati Moghaddam et al., 2021). Stochastic inversion algorithms such as the iterative ensemble Kalman filter (iEnKF) are capable of estimating both the value and associated uncertainty of unknown parameters but, to the best of the authors’ knowledge they have not yet been applied to colloidal transport problems (Evensen, 2003). This study reports the development of an iEnKF solver for the inverse RDAE using a synthetic experimental dataset. The iEnKF is found to be capable of accurately and efficiently estimating parameter values and their respective uncertainty. The uncertainty measurements were used to assess the sensitivity of the CRDAE to each parameter. The iEnKF was also shown to be robust to noise in data measurements and to changes in the quality of the initial guess.
Year: 2024