Search

Worldwide carbon dioxide (CO2) emissions targets are unlikely to be met without large scale geological CO2 storage, with much of the storage capacity found in saline aquifers. The key aim of carbon sequestration is the long-term or permanent storage of CO2 within the sub-surface, minimising the potential for CO2 leakage back to the surface. When CO2 is injected into a saline aquifer, it rises until it reaches an impermeable horizon, known as a caprock, where it subsequently spreads out due to buoyancy forces. While the CO2 is spreading, the chance of encountering potential leakage pathways increases and the potential rates of trapping also increase. In this dissertation, I combine reduced-order fluid models and geophysical data to understand the controls on the rate of trapping and CO2 leakage in saline aquifers. Chapter 1 outlines the broader context of CO2 sequestration, both in terms of physical processes, observations, and policy implications. In Chapter 2, I investigate the rate at which CO2 dissolves when injected into heterogeneous saline aquifers. As CO2 dissolves into brine, the density of the brine is increased, thereby acting as a mechanism for the stable trapping of CO2. CO2 injected into heterogeneous geological formations results in preferential migration along high permeability pathways, thus increasing the CO2-water interfacial area and enhancing dissolution rates. I analyse the rate at which free-phase CO2 propagates in layered reservoirs and the quantity of CO2 dissolved, showing that for reservoirs with finely bedded strata, over 10% of the injected CO2 can dissolve in a year. In Chapter 3, I investigate the potential for fault zones, which are localised planes of brittle deformation, to act as leakage pathways which transect low permeability structural seals. I develop an analytical model to describe the dynamics of leakage through a fault zone cross-cutting multiple aquifers and seals. This is tested against a set of porous media tank experiments and applied to a naturally occurring CO2-charged aquifer system at Green River, Utah. In Chapters 4 and 5, I combine seismic observations of a carbon sequestration project with numerical modelling to investigate the extent to which reduced-order models can accurately predict CO2 movement in heterogeneous aquifers at the Otway CO2 sequestration project in Victoria, Australia. I analyse seismic measurements of the motion of CO2 in the stage 2C trial, in which 15,000 tonnes of CO2-rich gas was injected into a saline aquifer at 1.5 km depth. The geometry of the reservoir and extent and thickness of the CO2 current is extracted from a series of time-lapse seismic surveys. I solve a gravity current model which accounts for topographic gradients in the caprock and incorporates residual trapping of CO2 numerically to model the observed spreading. This vertically-integrated model is inverted to find bulk reservoir properties that best match the modelled and measured CO2 distributions. The sensitivity of the model to changes in bulk properties and topographic gradients is investigated. In the concluding Chapter 6, I review the use of reduced-order physical modelling in describing the movement, trapping and leakage behaviour of CO2 in geological porous media, with application to field scale CO2 sequestration projects.

The flux and composition of river suspended sediments contain valuable information on processes including erosion rates, the composition of the eroded bed rocks, chemical weathering and sediment storage and sorting within river channels. In this thesis, samples from the Kosi River in Nepal (a major tributary of the Ganges), are used to improve current methods of investigating the processes controlling river suspended loads. Samples were collected as depth-profiles from all major tributaries and progressively along the mainstem to explore the mechanisms controlling the downstream changes in the Kosi River suspended sediments. Time series samples are used to investigate if suspended sediment compositions vary within the hydrological cycle and between successive years. Initial sampling commenced to quantify the impact of the 2015 Gorka earthquake. The April 2015, Mw 7.8 mainshock occured ~80 km west of the Kosi basin and the May Mw 7.3 aftershock within the basin. This thesis is one of the first assessments of the impact of the Nepal earthquakes on Himalayan river sediments. This thesis builds on work by Lupker et. al, 2012 and Bouchez et al., 2012 who sought to separate the effects of hydrodynamic sorting from the true downstream evolution of suspended sediment compositions in the Amazon and Ganges Rivers. From the different settling velocities of the mineral phases, hydrodynamic sorting results in depth variations in the composition and concentration of river sediments. These published works (ibid) focus on downstream variations in major element sediment chemistry; however, little work has been conducted into the effect of mineral phases on sediment chemistry. Kosi River sediments are, therefore, considered as mineral mixtures in this theis by modelling mineral modes from silicate major element concentrations and mineral compositions. Kosi River suspended sediments were found to be poorly sorted with depth, likely from the the highly turbulent secondary currents. Sediments have statistically higher Na/Al ratios and marginally lower K/Al ratios downstream at the Himalayan Front compared to those in the Himalayan Range. This downstream variation occurs from a greater abundance of plagioclase within Himalayan Front sediments, with the chemical composition of mineral phases being statistically indistinguishable at both sites. It is concluded that this downstream modal increase in plagioclase arises from the substantial flux of the plagioclase (and Na) rich sediments from the Arun Nadi River. The composition of Kosi River sediments were found to be in "steady state" across the annual hydrological cycle from late 2015 to 2018, but with a clear seasonal switching of the dominant mineral phases. Early monsoon sediments (May-July) are dominated by micaceous phases, switching to high abundances of quartz and plagioclase from August onwards. Variation in sediment provenance were ruled out, using Nd and Sr isotopes, as the cause of the seasonal or downstream sediment changes. It is concluded that the seasonality of mineral phases likely arises from the preferential sorting and easy transfer of micaceous sediments in the early monsoon, such that they are significantly depleted by the mid-monsoon From annual sediment fluxes, it is shown in this thesis that coseismic landsliding following the 2015 Gorka earthquakes resulted in an increase in the erosion and sediment fluxes above "baseline" levels in the Kosi River. Sediments collected within the Himalayan Range immediately following the earthquake (June to September 2015) were found to be more strongly dominated by contributions from the High Himalayan Crystalline sequence than within the "steady state" conditions. Whilst coseismic landsliding immediately following the shocks was recorded mainly in the west of the Kosi basin, results presented in this thesis suggest that the low peak ground accelerations across the entire basin seismically weakened slopes leading to enhanced erosion during the 2015 monsoon.

Geological Carbon Storage is necessary for reduction of anthropogenic carbon dioxide (CO2) emissions. CO2 is captured at 'point sources' (i.e. heavy industries, where large, concentrated volumes of CO2 are produced) and subsequently purified, compressed and transported. Some CO2 can be used for other processes, but excess CO2 can be injected into geological formations where it can be securely stored as either a gas, supercritical fluid or dissolved in formation fluids. Understanding the interactions of CO2 within the subsurface is important for risk characterisation. The long-term storage of CO2 at reservoir scales involves fluids close to equilibrium, which is better characterised using natural analogues than lab experiments. At the Green River CO2 seeps in Utah, CO2 saturated brines migrate up two fault systems: the Little Grand and Salt Wash faults. Modern springs, a man-made cold water geyser (Crystal Geyser), and fluid sampling during drilling at Little Grand in 2012 constrain current fluid compositions from 100m to 10km scales. Tufa/travertine (CaCO3) deposits form at modern springs, and historic deposits form along both faults. Fossil travertine veins can be dated by U-Th chronology and preserve chemical signatures of the fluids they form from: trace metals, δ234Ui and 87Sr/86Sr. Laser ablation methods were developed to analyse these samples, allowing for rapid, high spatial resolution analysis. Accurate and precise show external reproducibility of ±42 x10-6 (2σ) for 87Sr/86Sr, 7 x10-7 (2σ) for 230Th/238U and ±1.3 x10-6 (2σ) for 234U/238U. These uncertainties are propagated using an excess variance approach, giving typical age uncertainties of 20±1.8kyr and 120±4kyr, and ±40‰ for δ234Ui. Individual veins grow over < 5kyr, and vein systems may be active for upto 15kyr. 87Sr/86Sr & δ234Ui show limited variability within samples, and much larger variability between sample localities. At Salt Wash, these trends can be interpreted using an analytical reactive transport model for flow in the Navajo sandstone. This model is calibrated using spring fluid samples, and applied to interpret vein samples. Trends in δ234Ui are more easily interpreted spatially, due to increased residence times inferred from α-recoil inputs. This is counter to previous interpretations at this field area, where trends were interpreted as purely temporal. However, samples on the Little Grand fault do not fit either model well. Next to Crystal Geyser, one vein has clear annual layering. Trace metal trends show similarity to mixing patterns observed during geyser eruptions. There is a tentative link that the draining of the Navajo responds to seasonal changes in crustal stress, prior to man-made geyser activity.

Since 1996, super-critical CO$_2$ has been injected at a rate of $\sim$0.85~Mt~yr$^{-1}$ into a pristine, saline aquifer at the Sleipner carbon capture and storage project. A suite of time-lapse, three-dimensional seismic reflection surveys have been acquired over the injection site. This suite includes a pre-injection survey acquired in 1994 and seven post-injection surveys acquired between 1999 and 2010. Nine consistently bright reflections within the reservoir, mapped on all post-injection surveys, are interpreted to be thin layers of CO$_2$ trapped beneath mudstone horizons. The areal extents of these CO$_2$ layers are observed to either increase or remain constant with time. However, volume flux of CO$_2$ into these layers has proven difficult to measure accurately. In addition, the complex planform of the shallowest layer, Layer 9, has proven challenging to explain using reservoir simulations. In this dissertation, the spatial distribution of CO$_2$ in Layer~9 is measured in three dimensions using a combination of seismic reflection amplitudes and changes in two-way travel time between time-lapse seismic reflection surveys. The CO$_2$ volume in this layer is shown to be growing at an increasing rate through time. To investigate CO$_2$ flow within Layer~9, a numerical gravity current model that accounts for topographic gradients is developed. This vertically-integrated model is computationally efficient, allowing it to be inverted to find reservoir properties that minimise differences between measured and modelled CO$_2$ distributions. The best-fitting reservoir permeability agrees with measured values from nearby wells. Rapid northward migration of CO$_2$ in Layer~9 is explained by a high permeability channel, inferred from spectral decomposition of the seismic reflection surveys. This numerical model is found to be capable of forecasting CO$_2$ flow by comparing models calibrated on early seismic reflection surveys to observed CO$_2$ distributions from later surveys. Numerical and analytical models are then used to assess the effect of the proximity of an impermeable base on the flow of a buoyant fluid, motivated by the variable thickness of the uppermost reservoir. Spatial gradients in the confinement of the reservoir are found to direct the flow of CO$_2$ when the current is of comparable thickness to the reservoir. Finally, CO$_2$ volume in the second shallowest layer, Layer~8, is measured using structural analysis and numerical modelling. CO$_2$ in Layer~8 is estimated to have reached the spill point of its structural trap by 2010. CO$_2$ flux into the upper two layers is now $\sim$40\% of total CO$_2$ flux injected at the base of the reservoir, and is increasing with time. This estimate is supported by observations of decreasing areal growth rate of the lower layers. The uppermost layers are therefore expected to contribute significantly to the total reservoir storage capacity in the future. CO$_2$ flow within Layer~9 beyond 2010 is forecast to be predominantly directed towards a topographic dome located $\sim$3~km north of the injection point. This dissertation shows that advances in determining the spatial distribution and flow of CO$_2$ in the sub-surface can be made by a combination of careful seismic interpretation and numerical flow modelling.

Reactions between CO2-charged brines and reservoir minerals might either enhance the long-term storage of CO2 in geological reservoirs or facilitate leakage by corroding cap rocks and fault seals. Modelling the progress of such reactions is frustrated by uncertainties in the absolute mineral surface reaction rates and the significance of other rate limiting steps in natural systems. This study uses the chemical evolution of groundwater from the Jurassic Navajo Sandstone, part of a leaking natural accumulation of CO2 at Green River, Utah, in the Colorado Plateau, USA, to place constraints on the rates and potential controlling mechanisms of the mineral-fluid reactions,under elevated CO2 pressures, in a natural system. The progress of individual reactions, inferred from changes in groundwater chemistry is modelled using mass balance techniques. The mineral reactions are close to stoichiometric with plagioclase and K-feldspar dissolution largely balanced by precipitation of clay minerals and carbonate. Mineral modes, in conjunction with published surface area measurements and flow rates estimated from hydraulic head measurements, are then used to quantify the kinetics of feldspar dissolution. Maximum estimated dissolution rates for plagioclase and K-feldspar are 2x10-14 and 4x10-16 mol·m-2·s-1, respectively. Fluid ion-activity products are close to equilibrium (e.g. DGr for plagioclase between -2 and -10 kJ/mol) and lie in the region in which mineral surface reaction rates show a strong dependence on DGr. Local variation in DGr is attributed to the injection and disassociation of CO2 which initially depresses silicate mineral saturation in the fluid, promoting feldspar dissolution. With progressive flow through the aquifer, feldspar hydrolysis reactions consume H+ and liberate solutes to solution which increase mineral saturation in the fluid and rates slow as a consequence. The measured plagioclase dissolution rates at low DGr would be compatible with far-from-quilibrium rates of ~1x10-13 mol·m-2·s-1 as observed in some experimental studies. This suggests that the discrepancy between field and laboratory reaction rates may in part be explained by the differences in the thermodynamic state of natural and experimental fluids, with field-scale reactions occurring close to equilibrium whereas most laboratory experiments are run far-from-equilibrium. Surface carbonate deposits and cementation within the footwall of the local fault systems record multiple injections of CO2 into the Navajo Aquifer and leakage of CO2 from the site over ca. 400,000 years. The d18O, d13C and 87Sr/86Sr of these deposits record rapid rates of CO2 leakage (up to ~1000 tonnes/a) following injection of CO2, but rates differ by an order of magnitude between each fault, due to differences in the fault architecture. Elevated pCO2 enhances rates of feldspar dissolution in the host aquifer and carbonate precipitation in fracture conduits. Silicate mineral dissolution rates decline and carbonate precipitation rates increase as pH and the CO2 charge dissipate. The Sr/Ca of calcite cements record average precipitation rates of ~2x10-6 mol/m2/s, comparable to laboratory derived calcite precipitation rates in fluids with elevated Mn/Ca and Fe/Ca, at cc of ~1 to 3. This suggests that far-from-equilibrium carbonate precipitation, which blocks fracture conduits and causes the leaking system to self-seal, driven by CO2 degassing in the shallow subsurface, can be accurately modeled with laboratory derived rates. Sandstones altered in CO2 leakage conduits exhibit extensive dissolution of hematite grain coatings and are chemically bleached as a result. Measurements of Eh-pH conditions in the modern fluid, and modeling of paleo-Eh-pH conditions using calcite Fe and Mn concentrations, suggests that the CO2-charged groundwaters are reducing, due to their low dissolved O2 content and that pH suppression due to high pCO2 is capable of dissolving and transporting large concentrations of metals. Exhumed paleo-CO2 reservoirs along the crest of the Green River anticline have been identified using volatile hosting fluid inclusions. Paleo-CO2-charged fluids mobilized hydrocarbons and CH4 from deeper formations, enhancing the reductive dissolution of hematite, which produced spectacular km-scale bleached patterns in these sediment.