Natural 222Rn as a tracer of mixing and volatilization in a shallow aquifer during a CO2 injection experiment

This study aims to evaluate the application of 222Rn in groundwater as a tracer for monitoring CO2 plume migration in a shallow groundwater system, which is important to detect potential CO2 leakage in the carbon capture and storage (CCS) project. For this research, an artificial CO2‐infused water injection experiment was performed in a shallow aquifer by monitoring hydrogeochemical parameters, including 222Rn. Radon in groundwater can be a useful tracer because of its sensitivity to sudden changes in subsurface environment. To monitor the CO2 plume migration, the data were analysed based on (a) the influence of mixing processes on the distribution of 222Rn induced by the artificial injection experiment and (b) the influence of a carrier gas role by CO2 on the variation of 222Rn. The spatio‐temporal distributions of radon concentrations were successfully explained in association with horizontal and vertical mixing processes by the CO2‐infused water injection. Additionally, the mixing ratios of each monitoring well were calculated, quantitatively confirming the influence of these mixing processes on the distribution of radon concentrations. Moreover, one monitoring well showed a high positive relationship between 222Rn and Total dissolved inorganic carbon (TIC) by the carrier gas effect of CO2 through volatilization from the CO2 plume. It indicated the applicability of 222Rn as a sensitive tracer to directly monitor CO2 leakage. When with a little effect of carrier gas, natural 222Rn in groundwater can be used to compute mixing ratio of CO2‐infused water indicative of CO2 migration pathways. CO2 carrier gas effect can possibly increase 222Rn concentration in groundwater and, if fully verified with more field tests, will pose a great potential to be used as a natural tracer for CO2.


| INTRODUCTION
Carbon Capture and Storage (CCS) is one of the most feasible greenhouse gas emission-reducing techniques (Metz, Davidson, De Coninck, Loos, & Meyer, 2005;Gielen, 2008). Massive CO 2 gas emissions have triggered global climate changes, affecting the management of water resources and fossil energy. Carbon capture, storage, and sequestration techniques have been proposed to solve these environmental problems (Anwar et al., 2018;Shin, Ryu, Choi, Yun, & Lee, 2020;Üçtu g, A gralı, Arıkan, & Avcıo glu, 2014). However, despite various efforts to maintain stable and safe CO 2 storage, this technology has some potential problems, such as the leakage of stored CO 2 . The fugitive CO 2 can affect the groundwater system by leakage along the cracks before finally flowing to a near-surface ecosystem. This is especially important problem because shallow groundwater can be used for main potable water resources to people. When CO 2 enters the aquifer, the pH decreases and harmful heavy metals in host rocks could be dissolved making the groundwater unsuitable for drinking. Thus, previous studies conducted groundwater quality monitoring related to the CO 2 leakage (Bond et al., 2013;Humez, Lagneau, Lions, & Negrel, 2013;Jones et al., 2015; R. H. Patil, 2012;R. H. Patil, Colls, & Steven, 2010;Romanak et al., 2012;Smith et al., 2013;Spangler et al., 2010). However, few studies include experiments performed in shallow aquifers (Kharaka et al., 2010;Lions, Humez, Pauwels, Kloppmann, & Czernichowski-Lauriol, 2014;Spangler et al., 2010).
An artificial CO 2 injection experiment was conducted in a shallow groundwater system at the Korea CO 2 Storage Environmental Management (K-COSEM) field site in Eumseong, South Korea, which is designed to perform multidisciplinary research on environmental management induced by artificially injected CO 2 . The artificial injection experiment can determine the time, the location, and the amount of injection accurately. Thus, the range of CO 2 concentration, which affects the groundwater environment, can be quantified. That is, this injection experiment would be useful to investigate the hydrogeological characteristics related to the migration and distribution of diffusive CO 2 plumes induced by unexpected leakage.
Radon is a naturally existing isotope in the subsurface system and is sensitive to sudden environmental changes such as the mixing between different concentration sources or seasonal effects, especially in groundwater. 222 Rn belongs to the 238 U decay series and has a relatively short half-life (3.8 days). The radon concentrations depend on rock microstructure, pore space, emanation coefficient, environmental decay conditions such as 226 Ra activity, and dry density of the aquifer material. In addition, radon would be transported by advection with carrier gas, molecular diffusion, or dissolution from soil/bedrock. For the baseline, radon concentrations can be also obtained in any set of study site because the radon concentrations distributed by the bedrock characteristics. Consequently, many studies use the principal isotope of radon, 222 Rn, as a tracer to identify mixing characteristics in the groundwater system (Bertin & Bourg, 1994;Burnett & Dulaiova, 2003;Burnett, Peterson, Moore, & de Oliveira, 2008;Genereux, Hemond, & Mulholland, 1993;Hoehn & Von Gunten, 1989;J. Kim, Choi, Kim, Ryu, & Lee, 2020;McCoy & Corbett, 2009). Radon concentrations in groundwater can especially be used as a tracer to characterize groundwater flow affected by other matter-such as artificially injected water-into the steady state groundwater system. Moreover, radon activity in soil gases has been used as a tracer to monitor CO 2 leakage, because CO 2 is often regarded as the primary carrier gas for 222 Rn in soil and/or groundwater (Elío et al., 2015a(Elío et al., , 2015bG. Etiope, Guerra, & Raschi, 2005;G. Etiope & Martinelli, 2002;Giammanco, Sims, & Neri, 2007;Huxol, Brennwald, Hoehn, & Kipfer, 2012;Voltattorni et al., 2009). CO 2 gas leaks and/or release from deep sources (e.g., CO 2 reservoir) can show anomalies in radon activity with peak patterns (Elío et al., 2015a;Giammanco et al., 2007). Thus, radon activity must be considered to monitor CO 2 leakage at CCS sites. Moreover, radon measurements are easy, lowcost, and effective compared to other tracers, and have also less seasonal variations than CO 2 . The radon concentrations in groundwater can also provide the early warning signal as major water resources for people. However, previous studies related to the radon monitoring did not use radon concentrations in groundwater. Therefore, it is possible that there are potential relationships (e.g., positive correlations) between radon activity in groundwater and CO 2 plumes. This research describes and discusses the characterization of a groundwater flow system induced by CO 2 -infused water injection based on 222 Rn monitoring data detected in groundwater. The primary goal of this study is to evaluate the application of 222 Rn in groundwater as a natural tracer to monitor the CO 2 plume distribution and migration in shallow aquifer system. For this research, the results were analysed focused on (a) investigating the spatio-temporal changes of radon concentration distribution with the mixing processes induced by CO 2infused water injection into a shallow aquifer and (b) tracing the carrier gas role of CO 2 -saturated water through volatilization from the CO 2 plume on the variation of 222 Rn in groundwater with artificial injection.

| SITE DESCRIPTION
The study site is located at the K-COSEM field site in Eumseong, South Korea, which has an approximate area of 10,000 m 2 (Figure 1a, b). Seven wells (IW, MW1, MW2, PS, SW1, SW2, and SW3) were primarily used to monitor the radon concentration and other hydrological parameters (Figure 1c). MW1 and MW2 (multi-level groundwater monitoring wells) were installed as the bundle type with four specific depths at 2 m intervals. Information concerning the hydrostratigraphic layers and well depth is provided in Table 1. Downhole log data from monitoring wells (BH-1, BH-2, BH-3, and BH-4) indicate three distinct layers of weathered soils consisting of medium to coarse grained silty sand (0-30 m), weathered biotite granite (30-60 m), and unweathered biotite granite bedrock (>70 m). The depth of the water table is approximately 14-15 m below the surface. Groundwater generally flows from northwest to southeast based on equipotential groundwater level lines (Lee, Kim, Joun, & Lee, 2017). The results of push and pull tracer tests performed with multiple tracers (Chloride and SF 6 ) showed that hydraulic conductivity was from 4.0 ⨯ 10 −6 to 2.0 ⨯ 10 −5 m/s (H. H. Kim et al., 2015).

| Water sampling and analysis
Water samples were collected using 2-L polyethylene sampling bottles and an MP-1 pump (Grundfos, USA) from November 2016 to February 2017. The wells of IW, MW1, MW2, PS, and SW3 were used for water sampling in every sampling campaign conducted after the artificial CO 2 -infused water injection. Groundwater level was measured in-situ using the LTC Levellogger Junior (Solinist, Canada) at 10-min intervals.
Temperature, electrical conductivity (EC), dissolved oxygen (DO), and pH were detected using an YSI ProDSS digital sampling system (Xylem, USA). Stagnant water was removed by pumping for at least 15 min considering the well volume (IW well diameter: 100 mm, other monitoring wells diameter: 50 mm, and pumping rate: 4-6 L/min). Then, groundwater was collected when the parameters stabilized. The alkalinity was also determined in situ by titration with a 0.05-N HNO 3 solution to minimize degassing of CO 2 from the sampled waters.

| CO 2 -infused water injection
The CO 2 -infused water injection experiment was conducted from The CO 2 -saturated solution was continuously bubbling pure CO 2 gas into groundwater, which is pumped from the field site inside an unsealed tank. The injection system was at equilibrium with 1 atm of pure CO 2 .
Sampled water was put in a 500-ml air-bubbling flask, which was connected to the monitor in a closed air loop. The radon concentration was determined from repeated measurements of the gas circu-

| Groundwater level monitoring
The groundwater level monitoring data in five monitoring wells (IW, MW1-2, MW2-2, PS, and SW3) had variations before, during, and after the artificial CO 2 -infused water injection ( Figure 3). The data was taken at 10 min intervals. Figure 3 was drawn at 1 hr intervals.
Before the injection, the water level differences between the groundwater wells were not high-less than 0.1 m.

| Vertical profile
The vertical profiles were analysed in MW1 and MW2, which were bundle type wells with four specific depths at 2 m intervals (Figure 4).
The sampling depth was 17 m (below ground level) at the MW#-1, Here, vertical mixing was originally caused by the influence of the injected water with the migration of a CO 2 plume near the injection depth. Thus, the dominant factor in mixing processes is the horizontal migration of the injected water; vertical mixing also occurs by physical processes, such as diffusion and dispersion between the injected water and the groundwater located in the upper and lower parts of injection point. If the influence of the lower groundwater is high, the pattern shows a curve similar to that of the pre-injection data and/or small differences between the injection depth and the deepest point.

| Mixing ratio calculation using 222 Rn
For a quantitative interpretation of the study site, the mixing ratios were calculated for the four sampling periods using a radon tracer.
Binary mixing ratios can also be calculated using a simple mass balance equation, as follows: where C m is the concentration in the mixed groundwater well. (Bq/m 3 ),

| DISCUSSIONS
The change patterns of spatio-temporal distributions in radon concentration induced by the artificial CO 2 -infused water injection could be attributed to two primary reasons: (a) the influence of local groundwater flow characteristics (e.g., mixing processes) by the water injection on 222 Rn concentration distributions and (b) the influence of CO 2 as a carrier gas by volatilization from the CO 2 plume on the variation of 222 Rn. To discuss the application of 222 Rn in groundwater as natural tracer to monitor CO 2 plume migration, the spatio-temporal distributions of radon concentrations were analysed based on two reasons.

| Influence of artificially injected CO 2 -infused water on the distributions of 222 Rn concentration by mixing processes
The first reason is that the spatio-temporal distributions of radon concentration are formed by the local groundwater flow induced by the artificial water injection. Radon concentrations are generally distributed by mixing processes between two or more different water types, such as surface water, river water, shallow aquifers, or groundwater from a deep source (Bertin & Bourg, 1994;Burnett & Dulaiova, 2003;Burnett et al., 2008;Genereux et al., 1993;Hoehn & Von Gunten, 1989;. If the water mixing processes occurred by water injection, which has relatively low radon concentrations compared to the background level in the study site, the radon concentration distributions showed distinct mixing characteristics with the flow direction. For the IW data, TIC values and radon concentrations do not have the same peak between the two parameters ( Figure 6), showing that the mixing processes were dominant in the IW because of the injected CO 2 -infused water, which has low radon concentration despite the centre of CO 2 plume staying near the IW well until the third sampling period. The active mixing between the injected water, which has very low concentrations, and background groundwater around the IW well induced the decrease of the radon concentrations in the monitoring wells. These mixing influences are more important than the role of CO 2 as a carrier gas on 222 Rn in groundwater. In accordance with the IW data, the PS and MW2-2 wells showed similar patterns in early sampling periods, which is the opposite of the change patterns between TIC and radon concentration. That is, TIC values and radon concentrations had the peak pattern not in same sampling periods. In the PS well, the CO 2 plume arrived during the first sampling period, as shown in the highest TIC value, and radon concentrations were low Note: B.G refers the background radon concentration as the first end-member, and I.W refers the artificial injected water as the second end-member.

F I G U R E 5
The vertical profile data of 222 Rn, TIC, and water level in wells (a) MW1 and (b) MW2 F I G U R E 6 The pie diagram graph for mixing ratios of the background groundwater sample (green colour) and the artificial injected water (pink colour). The background contour map is based on the radon concentrations data collected at each sampling date (Bq/m 3 ) until the third sampling period. During the first sampling period, radon concentrations were low because of the active mixing with the injected water. During the third sampling periods after the second sampling periods, the boundary of the CO 2 plume, which had higher TIC and lower radon concentrations, affected the distribution of radon concentrations. During the fourth sampling period, radon concentrations recovered to the background range.
The groundwater level monitoring data showed abrupt change patterns in IW during the injection (Figure 3). All other monitoring wells showed similar drastic water level change patterns within 1 hr after the start of the injection. These variations mean that the mixing between the injected water and background groundwater could be actively occurring around IW. Thus, during the first sampling period, the average mixing ratios of injected water reached a very high 90.3% ( Figure 5). The low TIC value during the first sampling period in the IW well was attributed to the influence of chaser fluid. The abnormal pattern of MW2-2, which has a relatively high proportion of background groundwater, could be explained by the vertical mixing processes ( Figure 4b). Moreover, the high TIC values in MW2-2 and PS could be affected by the momentary CO 2 plume migration by the chaser fluid.
During the second and third sampling periods, the average mixing ratio values of the injected water were high (88.0 and 89.5%, respectively). During the second sampling period, the high proportion of injected water in MW1-2 and MW2-2 could be attributed to the migration of the CO 2 plume (which has low radon concentrations) boundary, as shown by the high TIC values at these points ( Figure 6).
Additionally, SW3 had a high background radon concentration because of the late arrival of the CO 2 plume, which signified little influence of mixing at the CO 2 plume boundary. During the third sampling period, the high portion of background radon concentration in MW1-2 could be explained by vertical mixing, as proved in Section 4.3.2 (Figure 4a) or by the regional groundwater flow direction from northwest to southeast. This was also attributed to the preferential plume direction toward to SW3 and not to MW1.
At the last sampling period, 1 month after injection, radon concentrations of all the groundwater wells increased by more than 9,000 Bq/m 3 . In this period, average values for the injected water and the background radon concentration were 49.9 and 50.1%, respectively. These values meant that the high background radon concentration was caused by end-member setting. Most wells showed recovery toward the range of background radon concentrations before the injection. Comparatively, SW3 had radon concentrations higher than the pre-injection value, which could be attributed to the preferential flow path of the CO 2 plume migration. Horizontal mixing with longitudinal transport along a preferential flow path could occur at the study site because of heterogeneous permeability distributions (Ju et al., 2019;H. H. Kim et al., 2015;Lee et al., 2017).

| Influence of CO 2 as a carrier gas on the variation of 222 Rn
The second reason is that CO 2 is the primary carrier gas of 222 Rn (Elío et al., 2015a(Elío et al., , 2015bGiammanco et al., 2007). Because there is no research on the application of CO 2 as a carrier gas for interpreting the data of 222 Rn concentrations detected in groundwater samples rather F I G U R E 7 Spatial correlation graph between radon concentrations and TIC in the groundwater monitoring well. The grey box is the range of radon concentrations from the August 1, 2016 data (Pre 1). (The sensitivity of radon concentrations was better than 3 cpm/(kBq/m 3 ) and the precision of TIC values was ±2.0 μmol/kg) than in the soil, the analysis and interpretation of radon concentration variations with an artificial CO 2 -infused water injection is necessary.
Thus, the plots of TIC and radon concentrations in five monitoring groundwater wells were analysed ( Figure 6). If CO 2 functions as a carrier gas by volatilization from the CO 2 plume on radon concentrations, the same peak can be observed in the same period, which means that the CO 2 leaked directly (Elío et al., 2015a;Giammanco et al., 2007).
For this possibility, main hypothesis is that, by volatilization from the CO 2 plume, CO 2 goes upwards to main aquifer with carry on radon, so that the radon concentrations in soil gas increase. In general, the radon in soil can be dissolved in groundwater. That is, the increasing radon in soil by carrier gas effect can be dissolved in groundwater.
Consequently, higher radon concentrations than background level can be observed at the CO 2 plume migration point.
Among the monitoring wells, only SW3 had a positive relationship between TIC and radon concentrations with the same peak in the TIC and radon concentration data during the fourth sampling period ( Figure 6). This phenomenon can be attributed to the influence of CO 2 as a carrier gas by volatilization from the CO 2 plume on 222 Rn.
The other data ( 222 Rn, pH, and HCO 3 ) supported this phenomenon.
The highest radon concentration during the fourth sampling period, indicating a stronger influence of CO 2 on 222 Rn in groundwater than of mixing processes by local groundwater flow (Figures 7 and 5). This specific variation during the fourth sampling period was also represented in pH and HCO 3 data ( Figure 7) and agreed with the results of other research conducted at this study site, which showed the preferential movement of the CO 2 plume between IW and SW3 conditioned by local permeability anisotropy for a long-term monitoring basis (more than one month from the injection) ( (Figure 4), indicating the arrival of the centre of the CO 2 plume at a depth of 25 m in the MW2-3 well. Additionally, MW 1-2 experienced the possible influence of the carrier gas CO 2 on 222 Rn in groundwater. This well showed increased TIC after the fourth sampling period (Figure 7) and radon concentrations in the background concentrations range. If the radon concentrations increased beyond these values, this well would also support the carrier gas effect.
Finally, the CO 2 plume migration can be conceptually illustrated as a schematic diagram (Figure 8). The CO 2 plume mainly migrated toward SW3 and downwards. During the second period, the migration of the B-B 0 section (from IW to MW2) was dominant because of the regional flow direction of the study site; after that period, the A-A 0 section migration (from IW to SW3) was dominant because of the preferential flow path of the CO 2 plume in this experiment. Therefore, although the mixing ratio portions of the artificially injected water were relatively higher than the background radon concentrations with the end-member setting, the mixing ratio distributions using radon concentrations clearly showed the influence of local groundwater flow characteristics, CO 2 plume migration, and variations in radon concentrations in accordance with horizontal and vertical mixing in most monitoring wells. Moreover, some monitoring points had positive relationships between TIC and 222 Rn, confirming the role of CO 2 as a carrier gas for 222 Rn. Therefore, results indicated that natural 222 Rn can be applied as the tracer to monitor and generally interpret the distribution of CO 2 plume migration.

| CONCLUSION
A CO 2 -infused water injection experiment was designed and conducted in a shallow aquifer to verify the application of 222 Rn in groundwater as a natural tracer for monitoring the leakage of stored CO 2 . For a multilateral comparison, the 222 Rn, TIC, pH, HCO 3 , EC, DO, and temperature were monitored in several groundwater wells.
The data sufficiently explained that the spatio-temporal distribution of radon concentrations was formed by local groundwater flow induced by artificial water injection. Radon concentrations were distributed in accordance with the mixing processes by local groundwater flow. The calculated mixing ratio results quantitatively supported these assumptions. Abnormal patterns in some monitoring points could be explained by vertical mixing, regional groundwater flow, or the active lateral transport immediately after the injection. Additionally, the role of CO 2 as the carrier gas for 222 Rn, was established in well SW3. Unlike other wells that showed opposite change patterns between 222 Rn and TIC, SW3 had the same peaks during the fourth sampling period in the 222 Rn and TIC plot.
Thus, this study evaluated the application of radon concentrations in groundwater as a natural tracer to monitor CO 2 plume migration with a CO 2 -infused artificial water injection into a shallow groundwater system. The spatio-temporal distributions of radon concentrations showed the deep influence of mixing processes and local groundwater flow characteristics induced by the injected water in the shallow aquifer system. Additionally, the high signal of the influence of CO 2 as a carrier gas of 222 Rn was detected in one monitoring well. Therefore, natural 222 Rn in groundwater could be used as the tracer to broadly investigate the flow characteristics related to CO 2 plume migration, by the interpretation of horizontal and vertical mixing processes with a little consideration of carrier gas effect. For directly leakage monitoring by carrier gas effect, it is necessary to conduct more field tests to apply 222 Rn in groundwater as the high signal tracer completely.