b'Education matters to displacement processes. The T 2 Full waveform inversion (FWI) is a high-relaxation time was measured to assessresolution velocity modelling technique the corresponding rock/\x1buid interplayused to image subsurface geological at the pore level, analyse the trappingstructures. The velocity pro\x1dle is one of behaviour, and demonstrate displacementvital physical information to identify the mechanisms, all of which are directlyproperties of medium. The conventional connected to \x1deld \x1bow functions. ways to conduct FWI can be gradient-In two-phase \x1bow systems, water-wetbased optimisations, Bayesian frameworks, samples consistently demonstratedor even evolutionary algorithms based greater residual CO 2trapping thanon stochastic theorem. The optimisation analogous oil-wet cores. Comparablemethods are usually driven by minimising results of CO 2trapping were observed inthe mis\x1dt between the observed and the three-phase \x1bow systems, althoughthe simulated waveform data on the certain aspects are di\x1eerent, andcurrent model to update the velocity displacements are signi\x1dcantly complex.model iteratively. The common issue of Furthermore, the oil recovery factor inthis method is easily being trapped into oil-wet cores was substantially lower thanlocal minima, highly depending on the Carbon Geo-sequestration (CGS) is a keythat of analogue water-wet cores. quality of an initial model. In order to component of the net-zero approach ofvalidate di\x1eerent methods, the easiest way achieving the global decarbonisationFinally, a thorough and self-consistentto obtain an initial model is a smoothed programme by 2050. When combineddataset that signi\x1dcantly impacts theversion of real velocity model (which with Enhanced Oil Recovery (CGS-EOR), itCO 2storage in saline aquifers and inis impossible in reality). Speaking of is an economically appealing techniquehydrocarbon reservoirs is the end result,inversion, it always represents the non-as it o\x1esets a fraction of the expensesdemonstrating how physicochemicalunique solutions due to its non-linear and associated with CO 2extraction. However,characteristics (wettability) adjustmentill-posed problems. Bayesian provides an despite recent cutting-edge technicalat the pore scale causes large-scalealternative option to bring uncertainty research, predicting quantitative CO 2 declines in trapping. These signi\x1dcantquanti\x1dcation of inverted models in trapping in geological formations viainnovative advancements emphasiseinstead of a best-\x1dt solution. But FWI is of capillary trapping remains enigmatic.the importance of CGS as an integralcomplexity and multi-parameters involved, Moreover, physical underpinnings ofpart of net-zero missions and CO 2 -EORthe intensive computation cost becomes several aspects of multiphase \x1bowproject designs relative to reservoir-scalean inevitable issue to deal with within the characteristics of CO 2 /brine/rock systems,implementations in terms of budgets,Bayesian framework.that greatly impact carbon capture anddelivery of additional energy resources containment security, require further(more oil recovery), and storage capacityMy thesis is aimed to apply a investigations. and containment integrity. geostatistical algorithm called Random Mixing (RM) for FWI. It takes Nuclear Magnetic Resonance (NMR) isAo Chang, University of Queensland: Fulladvantages of spatial correlation of the of particular interest when investigatingwaveform inversion with random mixing. unknown velocity field represented the petrophysical characteristics ofby variogram and univariate marginal multiphase \x1bow in porous media as it isdistribution as prior information to highly sensitive to pore scale rock-\x1buidguide the inversion, the univariate and \x1buid-\x1buid. From T 1 -T 22D images andmarginal distribution can be T 2relaxation time, conclusions regardingrepresented by any parametric or Porosity, Relative permeability, Pore-Sizenon-parametric distribution function. Distribution (PSD), Saturation (S w ) andThe prior information can be acquired thus residual CO 2saturation (S CO2,r ) caneasily according to literature, be inferred. experimental knowledge or direct observations. Meanwhile, RM allows This thesis investigates laboratory corefor uncertainty quantification based on \x1booding measurements with datasets of inverted realisations by means collected on water-wet (hydrophilic) andof variance. The realisations generated oil-wet (hydrophobic) sandstone andhere are possible velocity models of carbonate formation rocks under reservoirthe assumed variogram and marginal conditions. I was speci\x1dcally focused ondistribution, and all observations the pore-scale (micro-scale) \x1buid physicsare reproduced to some accuracy. in a multiphase \x1bow pore network dueThus, a multitude RM running can be to its signi\x1dcant impact on the Darcy- regarded as sampling from the random scale (macro-scale) \x1bow functions and,field which is modelling the solution ultimately, on the reservoir-scale CO 2 space of the inversion problems. residual trapping and oil recovery factor.The probability density function for I used the robust in-situ NMR T 1 -T 2 , 2Deach element can be characterised images to visualise \x1buid con\x1dgurationsby calculating and visualising in the pore-space and utilised T 1 /T 2ratioselement-by-element moments. The to assess the microscopic wettability ofspatial distribution of the mean and the rock to pore-space \x1buids subsequentstandard deviation across a group of 25 PREVIEW DECEMBER 2023'