WaterRF Project: Task 2.0

Modeling Design of Field Push-Pull Tests

Preliminary modeling will be conducted to design various parameters for conducting push-pull tests (PPTs). This modeling analysis will examine carbonate and silicate minerals to determine optimal parameters for the field tests. Modeling will be conducted using codes such as CORE2D (Yang et al., 2007a, b; Yang et al., 2008; Samper et al., 2009; Yang and Samper, 2009) and TOUGHREACT (Xu et al., 2004). CORE2D is a model tool for single phase flow and may be used for the push-pull tests with CO2-equlibrated water, and TOUGHREACT is a model tool for multiphase flow and will be used for the push-pull tests with direct CO₂ gas injection. Parameters required for modeling include aquifer permeability, injection amount and rate, mineralogy, groundwater chemical composition, and mineral dissolution rate constants. Groundwater levels have been monitored at the Cranfield site quarterly. Preliminary estimates of aquifer permeability are available for the Cranfield site from pumping tests. Sediments have been examined using XRD to determine solid-phase chemistry and SEM to determine mineralogy at a reconnaissance level. Baseline groundwater chemistry (major and trace elements) has been monitored in nine wells on a quarterly basis for 1 yr. Mineral dissolution rates will be obtained from the batch experiments and the literature (Plummer et al., 1978; Busenberg and Plummer, 1982; Helgeson et al., 1984; Palandri and Kharaka, 2004). Sensitivity analyses will be conducted to evaluate effects of varying injected CO₂ concentration, injection rate, injection time, duration of the rest period, regional groundwater-flow velocity, and duration of pull phase of the tests on the basis of length of time for chemical breakthroughs to occur. Given these results, we will design a sequence of field experiments to evaluate variations in CO₂ concentrations that will result in different pH values and resultant groundwater chemistry changes. Modeling analyses will also be used to evaluate different approaches to CO₂ injection, i.e., direct injection of CO₂ gas using a sparging system versus CO₂ equilibration with groundwater.

A numerical model was set up for simulating push-pull tests which will be conducted at the Cranfield shallow confined aquifer. The model domain is 3-D axisymmetric which was represented by a vertical 2-D cross section (Figure 4). The test well for push-pull tests is located at the center of the domain with 7 m of the confined aquifer. The outer boundary to the vertical axis of the well is 35 m. Because the simulated domain is confined at the top and bottom boundaries, no flux boundary conditions were imposed at these boundaries. Because a small amount of water is going to be injected into the well, we assume the head at the outer boundary, 35 m from the well, will not to be disturbed. Then a prescribed water head boundary condition, equal to the initial head is imposed on the outer boundary. A flux boundary condition is imposed on the inner boundary. The injected water is uniformly distributed along the well screen.

Three zones are considered in the simulated domain: a zone for the well bore within the casing, a zone between the casing and the well bore that was infilled with sand and a zone for the aquifer. The aquifer in the simulated domain is assumed to be homogeneous. Mineralogic data for the aquifer sediments considered in the numerical model are obtained from XRD analyses of the sediment samples. The injected water is assumed to have the same chemical composition as groundwater from the target aquifer and is equilibrated with CO₂ gas at a pressure of 1.10325 bar (1 atm). Potential water-rock-CO₂ interactions are considered to be the same as those determined from the batch experiments. The amount of water volume to be injected and the injection rate during the push phase, the time duration for the resting phase and the pumping rate during the pull phase are the key model parameters.

A 2-D grid with gradually increasing size from the inner boundary to the outer boundary is used in the numerical model (Figure 14).

​