Pore Pressure Analysis for Distinguishing Earthquakes Induced by CO2 Injection from Natural Earthquakes

It is important to distinguish between natural earthquakes and those induced by CO2 injection at carbon capture and storage sites. For example, the 2004 Mw 6.8 Chuetsu earthquake occurred close to the Nagaoka CO2 storage site during gas injection, but we could not quantify whether the earthquake was due to CO2 injection or not. Here, changes in pore pressure during CO2 injection at the Nagaoka site were simulated and compared with estimated natural seasonal fluctuations in pore pressure due to rainfall and snowmelt, as well as estimated pore pressure increases related to remote earthquakes. Changes in pore pressure due to CO2 injection were clearly distinguished from those due to rainfall and snowmelt. The simulated local increase in pore pressure at the seismogenic fault area was much less than the seasonal fluctuations related to precipitation and increases caused by remote earthquakes, and the lateral extent of pore pressure increase was insufficient to influence seismogenic faults. We also demonstrated that pore pressure changes due to distant earthquakes are capable of triggering slip on seismogenic faults. The approach we developed could be used to distinguish natural from injection-induced earthquakes and will be useful for that purpose at other CO2 sequestration sites.

A pilot project for carbon capture and storage (CCS) was in operation at Nagaoka, Japan, from 2003 to 2004 [24][25][26][27][28][29][30]. During the injection of CO 2 at Nagaoka on 23 October 2004, the Mw 6.8 Chuetsu earthquake struck the area [31][32][33][34] with its epicenter about 20 km from the site. In 2018, the Mw 6.6 Hokkaido Iburi-Tobu earthquake struck with its epicenter about 31 km from a CCS demonstration project in Tomakomai, Hokkaido [35]. Because the distances of both of these facilities were not far from the earthquake epicenters, people worried about the relationship between the earthquakes and CCS activities. Apart from this issue, the location of the Japanese Islands on an active subduction zone means that a reliable method is needed to distinguish between natural subduction-related earthquakes and those possibly induced by CO 2 injection. There will not be a perfect method to classify these earthquakes, because many geological and hydraulic parameters are related to the earthquake generation. However, we should develop a science-based, quantitative approach for the difficult issue we need to solve in CO 2 geological storage. Here, we focus on the pore pressure change to evaluate the earthquake rupture initiation. Comparing natural (or seasonal) pore pressure change with the pressure change due to CO 2 storage could enable us to evaluate whether the earthquake is a natural or artificial one.
In this study, the seasonal changes in pore pressure were calculated from rainfall and snowfall data from the Nagaoka CO 2 storage site during 2003 and 2004. Then, to allow comparison of those seasonal changes with the effect of CO 2 injection on pore pressure, a three-dimensional model of the hydraulic system at the site was used to numerically simulate pressure changes during CO 2 injection, including their effect on seismogenic faults at various distances. The results of the simulation were used to determine the lateral extent and magnitude of changes in pore pressure due to CO 2 injection and were compared with remote earthquake occurrences [36,37]. Application of the approach presented here will contribute to distinguishing natural earthquakes from those induced by CO 2 injection during CCS projects.

Study Area
We focused on the region around the Nagaoka CO 2 storage project, in the Niigata back-arc basin of northeastern Japan, shown in Figure 1 [26]. This region lies above the boundary between the Amur and Okhotsk tectonic plates [33,34,38,39], a region of high seismicity due to convergence of the two plates at a relative velocity of~1.5 cm/year. The sediments of the Niigata back-arc basin were deposited during the early Miocene opening of the Japan Sea [33]. They comprise a sequence of up to 7 km of sedimentary and volcanic rocks of Miocene to Holocene age that contain oil and gas reservoirs [38,39]. Earthquakes in this region have been caused by thrust-related fold ruptures within the Muikamachi fault system [31][32][33]. The 2004 Chuetsu earthquake (Mw 6.8) caused coseismic surface rupture and subsequent landslides that changed the topography in this region [34].
The CO 2 injection layer for the Nagaoka CCS project is in the early Pleistocene Haizume Formation of the Niigata basin. The Pleistocene sediments in the basin are deltaic and shallow-marine sediments (interbedded sand, silt, and mudstone) that provide both reservoirs and caprocks [27,[38][39][40] within an anticlinal structure related to seismogenic thrust-related folding [33]. CO 2 was injected into a saline aquifer at~1100 m depth and is capped by overlying mudstone [27,40]. The aquifer reservoir is composed of alternating sand and silt beds dipping at 15 • and was selected for pilot-scale storage of 10,400 t of CO 2 [27,40]. As previously mentioned, the 2004 Chuetsu earthquake occurred during CO 2 injection at Nagaoka with its epicenter~20 km from the site, shown in Figure 1 [31,33,34].

Data and Methods
We estimated and compared various pore pressure changes due to CO2 injection and natural pore pressure changes at the hypocentral depth (13 km) of the 2004 Mw 6.8 Chuetsu earthquake by the following approaches. Firstly, numerical simulations of fluid flow in a reservoir were used to estimate pore pressure during and after CO2 fluid injection. Secondly, the pore pressure diffusion equation was applied for snowfall and rainfall data from Nagaoka in order to determine changes in seasonal pore pressure. Lastly, the pore pressure change due to remote earthquakes was estimated (i.e., the 2011 Mw 9.1 Tohoku-Oki earthquake).

Data and Methods
We estimated and compared various pore pressure changes due to CO 2 injection and natural pore pressure changes at the hypocentral depth (13 km) of the 2004 Mw 6.8 Chuetsu earthquake by the following approaches. Firstly, numerical simulations of fluid flow in a reservoir were used to estimate pore pressure during and after CO 2 fluid injection. Secondly, the pore pressure diffusion equation was applied for snowfall and rainfall data from Nagaoka in order to determine changes in seasonal pore pressure. Lastly, the pore pressure change due to remote earthquakes was estimated (i.e., the 2011 Mw 9.1 Tohoku-Oki earthquake).

Change of Pore Pressure Due to CO 2 Injection
The change in pore pressure due to CO 2 injection was estimated using a dynamic fluid flow numerical simulation. The key parameters of the geological model are provided in Table 1 [28][29][30]42]. We first simulated fluid flow without CO 2 injection to determine the steady state reservoir conditions. We then simulated injection of supercritical CO 2 into the aquifer at rates of 20 to 40 t/day (comparable to the injection rate used at the Nagaoka site) over 18 months from July 2003 to January 2005. In total, an injection of 10,400 t of CO 2 into an aquifer was simulated [26,27,40]. The injection was restricted to a 10 m thick layer representing the porous and permeable reservoir sequence encountered at 1095-1105 m depth in the injection well [29,30,40]. To simplify the analyses and discussions for pore pressure and its influence upon the earthquake, the model we used for the simulation was relatively simple.

Seasonal Changes in Pore Pressure
Changes in subsurface pore pressure in response to seasonal changes in rainfall and snowfall can be large in Japan [21,43,44]. These seasonal changes were estimated by applying the pore pressure diffusion equation [15,20] to groundwater recharge from precipitation data [43] for the Nagaoka area ( Figure 1a) that we obtained from the Japan Meteorological Agency. The equation includes two hydrological loading mechanisms. The first of these considers instantaneous hydrological loading at depth (the undrained response), whereas the second mechanism (the drained response) is based on increased pore pressure diffusion by virtue of a hydraulically connected fracture system extending from the surface to the seismogenic fractures and faults [15,[18][19][20]45]. Here, we assumed that pore pressure changes were caused only by drained pore pressure diffusion. Pore pressure increases due to undrained instantaneous hydrological loading are difficult to estimate because of the many geological and hydraulic parameters that should be considered. Indeed, for this reason, previous studies by [46] and [15] considered only drained pore pressure changes. In reality, pore pressure changes due to overburden do occur under undrained conditions [15,47]. Therefore, the natural changes in pore pressure we estimated here are only part of total diffused response values.
Pore pressure P f at depth r and at time t, expressed by pore pressure diffusion equation, is described as follows where δt is the time increment of the rainfall (i.e., fixed 1 day), n is the number of time increments for two-year calculation (i.e., number of days), δp i is the precipitation load change (the product of water density ρ, gravitational acceleration g, and precipitation data δh at day n), r is the hypocentral distance from the surface, c is the hydrological diffusivity rate, and erfc is the complementary error function. The value of c for fractures associated with seismicity is in the range 0.1-10 m 2 /s, decreasing with depth [15,18,22]. Here, the diffused pore pressure after n days (P f ) at the hypocentral distance r after the start of rainfall series is the sum of the diffused pore pressure generated by daily precipitation load changes. Here, the diffused pore pressure was estimated at r = 13 km, and the sensitivity changes of c as 0.1, 0.5, 1, 2, and 4 m 2 /s were considered in this diffused pore pressure calculation. Because snow cover is deep in the Nagaoka area in winter, it is important to consider the effect of the supply of water from snowmelt (snow water equivalent). The rain precipitation data were revised to include snow water equivalent by considering snow depth. The authors of [48] determined a "degree-day factor" k (mm/ • C·day) that they derived from observations in central Japan (Iwate, near this study area; [48,49]) as follows where M is daily snowmelt rate (mm/day), T x is mean daily air temperature ( • C) on day x, and T x−1 is mean temperature on the day preceding day x. Finally, after snow depth was obtained and added to the precipitation data, Equation (1) was used to estimate seasonal changes in pore pressure in the study area.

Pore Pressure Change Due to Remote Earthquakes
We estimated pore pressure change due to remote earthquakes (i.e., the 2011 M w 9.1 Tohoku-Oki earthquake). Although this earthquake did not occur during CO 2 injection at Nagaoka, seismic velocity data acquired during the event are available [51,52] and can be used to estimate the changes in pore pressure that would have occurred in the Niigata Chuetsu region.
Laboratory experiments have shown that the relationship between S-wave velocity and pore pressure in sediments from southern Japan [53] can be expressed as where ∆Vs is the change in S-wave velocity (m/s) and ∆P f is the change in pore pressure or effective stress (MPa). Although this relationship does not necessarily hold for the data from this study area, similar relationships can be obtained for other sedimentary rock types. We calculated the decrease in S-wave velocity that would have occurred in the study area during the 2011 Tohoku-Oki earthquake (M w 9.1) to be 0.04% on the basis of research by [51] and [52]. If we assume that the S-wave velocity at 13 km depth was 2.5 km/s, the change in S-wave velocity due to that earthquake would have been 1 m/s. Then, according to Equation (4), the pore pressure increase during the 2011 Tohoku-Oki earthquake would have been 100 kPa.

Pore Pressure Changes Due to CO 2 Injection
As a result of fluid flow simulation considering the geological condition in this study area, the pore pressure and CO 2 saturation distribution results can be consistent with the history of CO 2 injection at the Nagaoka site in Figure 2 [29,30,40]. During the period of injection of CO 2 (Figure 2c), the simulated bottom-hole pressure in the well increased from an initial value of 11.1 to~12.2 MPa. After injection Sustainability 2020, 12, 9723 6 of 12 ceased, the pressure decreased dramatically, returning to the initial pressure and remaining there for the rest of the simulation period (Figure 2c).
injection at the Nagaoka site in Figure 2 [29,30,40]. During the period of injection of CO2 (Figure 2c), the simulated bottom-hole pressure in the well increased from an initial value of 11.1 to ~12.2 MPa. After injection ceased, the pressure decreased dramatically, returning to the initial pressure and remaining there for the rest of the simulation period (Figure 2c).
In this study, we mainly focus on pore pressure variation far from the injection well (or close to seismogenic fault). On the day of the Mw 6.8 Chuetsu earthquake (23 October 2004), the simulated elevated pressures due to CO2 injection extended ~2 km laterally from the injection site (Figure 2a). Therefore, the pore pressure increase due to injection resolved by the numerical simulation would not have extended as far as the seismogenic fault that caused the Chuetsu earthquake ( Figure 1).

Natural Variations in Pore Pressure
The natural variations in pore pressure were estimated at a depth of 13 km using precipitation and snow data, because the hypocenter is at 13 km depth in the epicentral area of the Chuetsu earthquake in Figure 1 [32]. The diffused pore pressure changes from 1 January 2003 to 31 December 2004 were calculated. In the study period, the increase in hydrological loading after adding snowmelt to rainfall data can be observed (Figure 3). Within two years, the estimates of diffused pore pressure at 13 km depth were varying up to 23 KPa in this region (Figure 3c). Therefore, a period of heavy snowmelt and long-term series rainfall considerably influenced groundwater level, leading to increased pore pressure due to subsurface pore pressure diffusion. In this study, we mainly focus on pore pressure variation far from the injection well (or close to seismogenic fault). On the day of the Mw 6.8 Chuetsu earthquake (23 October 2004), the simulated elevated pressures due to CO 2 injection extended~2 km laterally from the injection site ( Figure 2a). Therefore, the pore pressure increase due to injection resolved by the numerical simulation would not have extended as far as the seismogenic fault that caused the Chuetsu earthquake ( Figure 1).

Natural Variations in Pore Pressure
The natural variations in pore pressure were estimated at a depth of 13 km using precipitation and snow data, because the hypocenter is at 13 km depth in the epicentral area of the Chuetsu earthquake in Figure 1 [32]. The diffused pore pressure changes from 1 January 2003 to 31 December 2004 were calculated. In the study period, the increase in hydrological loading after adding snowmelt to rainfall data can be observed (Figure 3). Within two years, the estimates of diffused pore pressure at 13 km depth were varying up to 23 KPa in this region (Figure 3c). Therefore, a period of heavy snowmelt and long-term series rainfall considerably influenced groundwater level, leading to increased pore pressure due to subsurface pore pressure diffusion. Sustainability 2020, 12, x FOR PEER REVIEW 7 of 12

Discussions-Comparison of Pore Pressure Increase due to CO2 Injection with Natural Variations
Based on the results, the simulated increases in pore pressure due to CO2 injection were much higher than the seasonal variations in pore pressure close to the injection well (Figure 4), but only within a radius of ~2 km around the well (Figure 4). The vertical extent of the increase due to CO2 injection was restricted by the caprock overlying the reservoir and the almost impermeable layer 10 m beneath it [29,30,40,42]. The 10-year simulation showed that the bottom-hole pressure decreased dramatically when injection ceased, returning relatively quickly to the initial steady state pore pressure (Figure 2c).
The hypocenter of the Chuetsu earthquake is >20 km from the site of the injection well at a depth of 13 km in Figure 1 [32,34]. Therefore, the simulation results clearly showed that in the hypocenter the pore pressure increase due to CO2 injection (Figure 2) was very much smaller than the natural fluctuations in pore pressure (Figures 3c and 4). Furthermore, the estimated natural pressure variation is the total diffused value because we ignored the effect of undrained instantaneous hydrological loading. Therefore, the natural pressure variation is more dominant, suggesting that the 2006 Chuetsu earthquake was a natural event. In more detail, an earthquake could be triggered by

Discussions-Comparison of Pore Pressure Increase Due to CO 2 Injection with Natural Variations
Based on the results, the simulated increases in pore pressure due to CO 2 injection were much higher than the seasonal variations in pore pressure close to the injection well (Figure 4), but only within a radius of~2 km around the well (Figure 4). The vertical extent of the increase due to CO 2 injection was restricted by the caprock overlying the reservoir and the almost impermeable layer 10 m beneath it [29,30,40,42]. The 10-year simulation showed that the bottom-hole pressure decreased dramatically when injection ceased, returning relatively quickly to the initial steady state pore pressure (Figure 2c).
The hypocenter of the Chuetsu earthquake is >20 km from the site of the injection well at a depth of 13 km in Figure 1 [32,34]. Therefore, the simulation results clearly showed that in the hypocenter the pore pressure increase due to CO 2 injection (Figure 2) was very much smaller than the natural fluctuations in pore pressure (Figures 3c and 4). Furthermore, the estimated natural pressure variation is the total diffused value because we ignored the effect of undrained instantaneous hydrological loading. Therefore, the natural pressure variation is more dominant, suggesting that the 2006 Chuetsu earthquake was a natural event. In more detail, an earthquake could be triggered by shear stress in Sustainability 2020, 12, 9723 8 of 12 addition to pore pressure variation. However, the shear stress perturbation due to CO 2 injection is also minor around the hypocenter (similar order to pore pressure).
Sustainability 2020, 12, x FOR PEER REVIEW 9 of 12 Figure 4. Comparison of the change with distance from the injection well (inferred from Figure 2) of the simulated increase of pore pressure with those related to precipitation (rainfall and snowmelt) and a remote earthquake (here, the 2011 Mw 9.1 Tohoku-Oki earthquake).

Conclusions
We have developed a method to distinguish natural earthquake and CO2 injection-induced earthquake, based on pore pressure. The changes in pore pressure caused by injection of CO2 at the Nagaoka CO2 storage site were examined and compared to natural variation. The main conclusions are as follows: -Apart from an area within ~2 km of the injection site, the natural seasonal fluctuations of pore pressure in the broad area around the Nagaoka CO2 injection site are much greater than those induced by CO2 injection. Therefore, CO2 injection at Nagaoka during 2003 and 2004 could not trigger the 2004 Chuetsu earthquake; -Although seasonal increases in pore pressure are relatively small in the Nagaoka area, they can be large enough to trigger slip on seismogenic faults that are already in a critical stress state due to regional subduction tectonics, especially during periods of snowmelt and rainfall; -Further investigation is needed of other phenomena that may cause increases in pore pressure large enough to trigger earthquakes (e.g., changes in sea surface height, and large-magnitude remote earthquakes) in order to evaluate the influence of pore pressure change more precisely. The remote earthquake could largely influence upon pore pressure variation. In the result, it shows that the pore pressure increase of 100 kPa in Nagaoka area due to 2011 Mw 9.1 Tohoku-Oki earthquake is more significant than those from seasonal fluctuations; - The methodology presented here provides a means to distinguish natural earthquakes from those induced by CO2 injection and can be useful at other CO2 sequestration sites and at geothermal field developments.
Here, the natural pore pressure variations at hypocenters exceed those from CO2 injection of Nagaoka project. However, there are also some limitations in this study: (1) the simplified fluid flow model was used for simulating the changes in pore pressure during CO2 injection at the Nagaoka site (Japan); (2) the undrained response (i.e., overburden pressure) for underground pore pressure changes is not considered due to the complexity of geological and hydrological heterogeneity; (3) the magnitude of the increase in natural pore pressure (i.e., seasonal effects and stress field) required to induce earthquakes in the Nagaoka region should be investigated. According to the earthquake catalog by [36], none of the hypocenters of the earthquakes that occurred in the Nagaoka area in 2003 and 2004 were near the injection depth of~1100 m or within a 2 km radius from the injection site. This is likely because the injected reservoir is within a thick and stable sequence of deltaic and shallow-marine sediments of the Niigata basin [30,38,39]. The reservoir is sealed by a mudstone caprock that effectively traps the fluid, and there are no seismogenic fractures or faults within the reservoir (Figure 1b). In addition, post-injection fluid sampling and reactive transport modeling of the supercritical injected CO 2 indicate that most of it is dissolved and will be progressively precipitated in the host reservoir by mineral carbonation over the years [28][29][30]42]. The dissolution and mineralization processes can reduce the elevated pore pressures caused by the CO 2 injection.
Other natural phenomena that can affect pore pressure include changes in sea surface height, and remote large-magnitude earthquakes. The Nagaoka injection well is about 15 km from the coast; thus, the effect of sea surface height in this study was not considered (Figure 1). The effects of large-magnitude remote earthquakes on pore pressure, however, are not negligible. The pore pressure variation due to remote earthquakes is difficult to estimate, though it can be roughly estimated from temporal variation of seismic velocity. The authors of [54] investigated subsurface stress changes associated with the 2011 Tohoku-Oki earthquake, and found that during the earthquake seismic velocity decreased greatly over a wide area. A decrease in seismic velocity during the large earthquake was reported by several studies [4,51]. Seismic velocity variation phenomena such as these are likely caused by mobilization of fluids and elevation of fluid pressure in the crust during shaking; it is possible that the resultant pressurized fluids inflate shallower formations and open cracks in them [4]. Thus, large-magnitude earthquakes can change pore pressures in areas that are distant from them.
Several large and remote earthquakes occurred during the period of CO 2 injection at Nagaoka (e.g., the Mw 6.4 Northern Miyagi and Mw 6.8 Tokachi-Oki earthquakes). Pore pressure changes in response to these earthquakes might have influenced those at seismogenic faults near Nagaoka CO 2 injection. Here, we consider this possibility on the basis of previous research on the influence of the 2011 Tohoku-Oki earthquake (Mw 9.1) on the Niigata Chuetsu region, some 400 km from the epicenter of the 2011 earthquake (Figure 1a). According to Equation (4), the pore pressure increase during the Sustainability 2020, 12, 9723 9 of 12 2011 Tohoku-Oki earthquake was estimated to be 100 kPa (see Section 3.3), clearly demonstrating that a large earthquake can have a marked influence on pore pressure in a remote region up to 400 km from its epicenter (black horizontal line in Figure 4). This estimation demonstrates that the large earthquake significantly affects the remote region. Indeed, because of this pore pressure variation, aftershocks frequently occur around the seismogenic faults. Moreover, these increases in pore pressure were larger than the seasonal variations in pore pressure we obtained from daily precipitation data (Figures 3c  and 4), thus lending further support to the presumption that remote earthquakes can influence pore pressure at considerable distances from their epicenters, which is more significant than those from seasonal change behavior on the seismogenic faults.

Conclusions
We have developed a method to distinguish natural earthquake and CO 2 injection-induced earthquake, based on pore pressure. The changes in pore pressure caused by injection of CO 2 at the Nagaoka CO 2 storage site were examined and compared to natural variation. The main conclusions are as follows: -Apart from an area within~2 km of the injection site, the natural seasonal fluctuations of pore pressure in the broad area around the Nagaoka CO 2 injection site are much greater than those induced by CO 2 injection. Therefore, CO 2 injection at Nagaoka during 2003 and 2004 could not trigger the 2004 Chuetsu earthquake; -Although seasonal increases in pore pressure are relatively small in the Nagaoka area, they can be large enough to trigger slip on seismogenic faults that are already in a critical stress state due to regional subduction tectonics, especially during periods of snowmelt and rainfall; -Further investigation is needed of other phenomena that may cause increases in pore pressure large enough to trigger earthquakes (e.g., changes in sea surface height, and large-magnitude remote earthquakes) in order to evaluate the influence of pore pressure change more precisely. The remote earthquake could largely influence upon pore pressure variation. In the result, it shows that the pore pressure increase of 100 kPa in Nagaoka area due to 2011 Mw 9.1 Tohoku-Oki earthquake is more significant than those from seasonal fluctuations; - The methodology presented here provides a means to distinguish natural earthquakes from those induced by CO 2 injection and can be useful at other CO 2 sequestration sites and at geothermal field developments.
Here, the natural pore pressure variations at hypocenters exceed those from CO 2 injection of Nagaoka project. However, there are also some limitations in this study: (1) the simplified fluid flow model was used for simulating the changes in pore pressure during CO 2 injection at the Nagaoka site (Japan); (2) the undrained response (i.e., overburden pressure) for underground pore pressure changes is not considered due to the complexity of geological and hydrological heterogeneity; (3) the magnitude of the increase in natural pore pressure (i.e., seasonal effects and stress field) required to induce earthquakes in the Nagaoka region should be investigated.
Funding: This study was supported by the Japan Society for the Promotion of Science KAKENHI grant number JP20H01997 and JP20K20948.