Coseismic and Early Postseismic Deformation of the 2024 Mw7.45 Noto Peninsula Earthquake

An unexpected Mw7.45 earthquake struck the Noto Peninsula on 1 January 2024, preceded by several long‐living earthquake swarms, providing a valuable opportunity to study seismic and aseismic slips, as well as their interactions. We derived coseismic and 19‐day postseismic slip distributions by inverting co‐ and post‐seismic displacements from Global Navigation Satellite System (GNSS) data. The inverted coseismic slip distribution shows two slip patches, with a maximum slip of ∼4 m. The early postseismic afterslip is 0.1–0.25 m within coseismic slip asperity and 0.1–0.6 m northward of the rupture area. The afterslip within the rupture area is accompanied by numerous aftershocks and coincides with a ∼6 MPa stress drop, suggesting that aftershocks are likely driven by the afterslip. The pattern of poroelastic rebound implies a potential effect of fluid flow on aftershock triggering. This study sheds lights on the intricate interplay between seismic and aseismic processes following large earthquakes.


Introduction
The Japanese islands are formed due to the subduction of the ancient Pacific Ocean floor beneath the Eurasian Plate (Taira, 2001).As the back-arc basin of the subduction zone, an extensional tectonic stress field was dominant in the Japan Sea from the Early Miocene to the Early Middle Miocene, and numerous normal faults formed due to basin rifting (Jolivet et al., 1992;Sato, 1994).From the Pliocene to the Quaternary, the eastward motion of the Amur plate resulted in strong compression that reactivated the normal faults to become thrusting faults (Taira, 2001).These reactivated reverse faults have the potential to host large earthquakes that may cause tsunamis and severe destruction.
Several large earthquakes have occurred in this fault system, including the Mw ∼6.6 earthquake in 1729 (Hamada et al., 2016) and the Mw 6.8 Noto Hanto earthquake in 2007 (Hiramatsu et al., 2008).Intense earthquake swarms have been observed in the Noto Peninsula since November 2020.Such earthquake swarms are usually observed in geothermal or volcanic regions (Cesca et al., 2022).However, there has been no volcanic activity in this region since the late Miocene (Kato, 2024;Nakajima, 2022).It has been suggested that these seismic activities may be attributed to fluid movement at depth (Nakajima, 2022;Yoshida et al., 2023).
On 1 January 2024 (UTC Time 07:10), an unexpected earthquake with a moment magnitude (Mw) of 7.5 occurred on the shallow reverse fault in the offshore region of the Noto Peninsula, making it the strongest earthquake in this region in the last century.The hypocenter of this event was determined to be at 137.242°E, 37.498°N, and a depth of 10 km by the U.S. Geological Survey (USGS).The focal mechanism solutions from the USGS indicate that the ruptured fault is a reverse fault.The fault has a moderately dipping angle of 35.0°and stretches at an angle of 51.0°from southwest to northeast.The seismic moment release was 2.3 × 10 20 Nm, equivalent to Mw7.5.The earthquake and induced tsunami caused severe damage to the Noto Peninsula.According to the earthquake field survey, the ground surface near the epicenter was uplifted by 4 m, which resulted in a seaward migration of the coastline up to 200 m.
The relationship between seismic and aseismic processes has gained significant attention in earthquake studies.Scholz (1998) suggested that seismic and aseismic slip are two complementary types of slip based on the rate-and state-friction law.Such a relationship between the seismic and aseismic slip has been observed in large earthquakes.For instance, short-term (e.g., over a few days; Kato et al., 2012) and long-term changes (Yokota & Koketsu, 2015) in GNSS time series and seismic activity preceding the 2011 Mw 9.0 Tohoku earthquake have been interpreted within this context.The aseismic segment may often act as a barrier and terminate seismic ruptures (Nishikawa et al., 2019;Perfettini et al., 2010).In nature, the occurrence of early aftershocks (first few hours) is often controlled by the early afterslip (Itoh et al., 2023;Jiang et al., 2021).These studies suggest that the relationship between the seismic and aseismic slip is complex.Meanwhile, studying aseismic processes at a short time scale (first few hours) is challenging because it requires high-rate GNSS time series rather than the daily GNSS time series.The availability of high-rate GNSS time series of this earthquake from the Nevada Geodetic Laboratory (NGL, http://geodesy.unr.edu/)provides a unique opportunity to explore the interaction between seismic and aseismic processes in nature.
In this work, we first derive the coseismic displacement and the 19-day postseismic deformation from the highrate GNSS observations of this earthquake.We then determine the coseismic slip distribution and subsequent postseismic afterslip distribution on the seismogenic fault by geodetic inversion of the well-documented coseismic and early postseismic displacements.Finally, we calculate the coseismic static Coulomb stress changes and poroelastic rebound following the earthquake and discuss their potential influences on aftershocks.

GNSS Data Processing
We used 5-min continuous GNSS data from NGL to derive the coseismic and 19-day postseismic displacements of the 2024 Noto Peninsula earthquake (Figure 1).These data, processed by NGL in the ITRF 2014 (Altamimi et al., 2016), encompass various signals, including coseismic and postseismic deformation triggered by the earthquake, interseismic secular trends, semi-annual and annual variations, as well as common mode noise.Given the relatively short duration of the time series (∼20 days), the contributions from secular trends, semi-annual, and annual variations were negligible, and thus ignored in the analysis.We used the difference of averaged positions 30 min before and after the mainshock from the GNSS data to estimate coseismic displacements.As there are four large aftershocks (shown as different colored squares in Figure 1) within 3 hr after the mainshock, we excluded the coseismic effects of these four earthquakes by discarding the time series of the first 3 hr.To obtain the postseismic deformation of the Noto Peninsula earthquake from 3 hr after the mainshock to the 19th-day, we first estimated the common mode noise using a method described in Huang et al. (2012) and removed it from the time series.Subsequently, we fitted the remaining postseismic GPS time series using a logarithmic function and an exponential function.Finally, we calculated postseismic displacements at specified time intervals from the fitted curve (Figure S1 in Supporting Information S1).

GNSS Data Analysis
We extracted coseismic and postseismic displacements at 581 and 55 GPS stations, respectively.The horizontal coseismic displacement reaches ∼100 cm in the mainshock region, then decreases rapidly with increasing distance from the rupture area (Figure 2a).The vertical coseismic displacement is up to ∼100 cm uplift within the rupture area and up to ∼30 cm subsidence on the southwestern edge of the rupture area (Figure 2a).The direction of the early postseismic deformation is similar to that of the coseismic deformation (Figures 1 and 2).Furthermore, the horizontal postseismic deformation is ∼3 cm and smaller than 1 cm within and outside the rupture area, respectively (Figure 2c).The pattern of vertical motion after the mainshock is more complex, transitioning from uplift to subsidence, and back to nearly zero as the distance from the rupture area increases, with a range of 2-2 cm.

Aftershocks
We focused on aftershocks with M JMA ≥ 1 and M JMA < 5, which occurred within 19 days after the earthquake in the study region.Because the uncertainties of locations of the aftershocks are quite large, we simply assumed that those aftershocks located within 10 km of the fault were aftershocks that occurred on the fault, which are shown as green dots in Figure S2a in Supporting Information S1.Hereafter the aftershocks mentioned in this work are these on-fault aftershocks.As shown in Figure S1 in Supporting Information S1, the number of aftershocks with M JMA of <2 (Figure S2b in Supporting Information S1), 2-3 (Figure S2c in Supporting Information S1), and ≥3 (Figure S2d in Supporting Information S1), have been decreasing over time, but the pattern of their distribution has remained relatively stable.The aftershocks in the target region are distributed at depths ranging from 0 to 20 km.

Fault Geometry
We delineated the fault geometry of the 2024 earthquake using the data from the active fault catalog along the Sea of Japan from the Geological Survey of Japan (Mulia et al., 2020).The strike directions of the two fault segments, F1 and F2, are 64°and 55°, respectively.The dip angles, widths, and lengths of F1 and F2 are the same, with values of 45°, 50 km, and 70 km, respectively.We will use this fault information to analyze the slip model of the 2024 earthquake.In addition, we will further discuss the impact of different dip angles of the fault geometry on the coseismic and postseismic slip distributions in Section 4.2.

Coseismic Slip Model
We inverted the coseismic GNSS displacement to determine the coseismic slip on the fault of the earthquake with a uniform elastic half-space model (Y.Okada, 1985).We use the code from Xu et al. (2016) to solve for the strikeslip and dip-slip components of coseismic slip on all patches of the two fault segments.We choose the smoothing factor based on the data fitting of the inverted coseismic slip model.
The optimal smoothing factor is the largest smoothing factor that does not induce significant degradation of the data fitting.The best smoothing factor of the coseismic slip model (model CM) is 0.07 (Figure S3a in Supporting Information S1).The modeled coseismic displacement reproduces the first-order pattern of both the horizontal and vertical GNSS observations (Figure 2a and Figure S4 in Supporting Information S1).The coseismic rupture of the Noto Peninsula earthquake, whose slip error is generally less than 15% (Text S1 and Figure S5 in Supporting Information S1), shows two main rupture areas along the fault segments, with a maximum value of ∼4 m at a depth of ∼6 km (Figure 2b).The total released moment is about 1.91 × 10 20 Nm, corresponding to a moment magnitude of 7.45 (assuming a shear modulus of 30 GPa), which is close to the USGS of Mw7.5.There is a ∼45 cm misfit in the horizontal direction at station J972, which may be due to the coseismic offset of the Mw6.2 aftershock that occurred 8 min after the mainshock.

Early Afterslip Model
It has been reported that the postseismic deformation in the near field within the first few months is controlled mainly by afterslip (Freed et al., 2012).Therefore, we only consider the 19-day postseismic afterslip in this work.
In the inversion method, fault parameters and selection strategy of optimal smoothing factor are the same as those used in the coseismic slip inversion.
The optimum smoothing factor was determined to be 0.15 (Figure S3b in Supporting Information S1).The inverted postseismic deformation closely matches both the horizontal and vertical GNSS observations (Figure 2c and Figure S6 in Supporting Information S1).The postseismic afterslip model (model AM) shows a slip of 0.1-0.25 m in the rupture area, with a peak slip error of approximately 0.1 m at the edge of the fault geometry (Text S1 and Figure S7 in Supporting Information S1).There is also an afterslip patch of approximately 0.1-0.6 m to the northeast of the coseismic slip area, where sparse GNSS stations are located, but this afterslip patch may still be reasonable (more details in Section 4.1) (Figure 2d).The postseismic afterslip model indicates a total moment released of about 1.995 × 10 18 Nm, which is equivalent to a moment magnitude of 6.13, accounting for ∼1% of the coseismic moment.

Synthetic Tests
In order to assess the influence of the uneven distribution of GNSS stations on the inversion results, we perform three synthetic tests to demonstrate the ability of the GNSS data to recover the details of the slip distribution on two fault segments, F1 and F2, of the source fault.We generate synthetic displacements at GNSS stations with the same fault parameters as those in the aforementioned coseismic and postseismic slip inversions in a uniform elastic half-space model (Y.Okada, 1985) based on slip models shown in Figures S8d-S8f in Supporting Information S1.The synthetic displacements at GNSS stations were then inverted with the same inversion method to recover the assumed slip models on the two fault segments.
For the first test model, the inversion was successful in recovering the input model for the F1, where most GNSS stations are located in the vicinity of the slip area (Figure S8g in Supporting Information S1).In addition, the inverted deformation at the GNSS stations closely matched the surface deformation (Figure S8a in Supporting Information S1).For the second model, the inverted surface deformation also well fits the overall pattern of the synthetic deformation at each GNSS station.Although the inverted model was able to recover the overall pattern of the input model, it overestimated the slip area due to a lack of GNSS data in that region (Figures S8b, S8e, and S8h in Supporting Information S1).For the third model, the inversion captures the bilateral rupture successfully recovered slip but overestimates the slip area in F2 (Figures S8f and S8i in Supporting Information S1).The inverted surface deformation is still consistent with the overall pattern of synthetic deformation at each GNSS station (Figure S8c in Supporting Information S1).These tests indicate that our inversion method can effectively recover the input model for F1.However, the sparse distribution of GNSS stations limits the accuracy of the slip pattern recovery for F2, although the overall pattern is still discernible.Therefore, the distribution of the postseismic afterslip model is considered to be reasonable, but the real slip distribution on F2 may be more concentrated than that shown in the inversion results.

Influence of Different Dip Angles on Fault Geometry
To further investigate the influence of different fault geometries on the distribution of coseismic and postseismic slip, we have constructed test models using different dip angles, while keeping other parameters consistent with the final models (CM and AM).Model results indicate that the inverted coseismic slip models are similar across different fault geometries, displaying two main rupture areas along the fault segments, but with different peak values of slip (Figures S9a-S9f in Supporting Information S1).Furthermore, the modeled surface deformation derived from these fault geometries effectively reproduces the GPS observations (Figures S10a-S10f in Supporting Information S1).Although there are slight variations in the distribution of slip in the F1, the overall features of postseismic afterslip models remain consistent (Figures S11a-S11f in Supporting Information S1).Nevertheless, the modeled deformation at each GNSS station remains consistent in different fault geometries (Figures S12a-S12f in Supporting Information S1).These results indicate that both the coseismic and postseismic slip distributions described here are independent of the dip angles of the fault geometry.

Interplay Between Aftershocks, Mainshock and Afterslip
It has been found that the distribution of on-fault aftershocks following large earthquakes is a result of stress changes caused by the mainshock (Harris, 1998;Harris & Simpson, 1998;King et al., 1994), and/or the subsequent afterslip (Hsu et al., 2006;Marone et al., 1991;Perfettini & Avouac, 2007).Several studies have used the Coulomb failure criterion (ΔCFS = Δτ + μ′Δσ n , where ΔCFS is the change of the Coulomb failure stress, Δτ is the change in shear stress, Δσ n is the change in normal stress, and μ′ is the effective coefficient of friction) to investigate slip episodes that are triggered by the static coseismic stress changes caused by nearby earthquakes (Harris, 1998;Harris & Simpson, 1998;King et al., 1994).A positive ΔCFS indicates that the receiver easily fails, and vice versa.We calculated the static Coulomb stress changes from the Mw7.45 event (the effective frictional coefficient is assumed to be 0.4) at the fault segments (F1 and F2).This analysis aims to investigate the correlation between the stress changes of the mainshock, the afterslip, and the aftershocks.
We first conducted a test to analyze the relationship between the aftershocks on F1, the mainshock, and the afterslip on F1.The calculation results indicate that there are significant negative Coulomb stress changes larger than 6 MPa within the rupture area on the F1 of this earthquake.This negative Coulomb stress change would prevent slip on fault (Harris & Simpson, 1998;Stein, 1999), and thus inhibit aftershocks on F1.However, there were still a large amount of aftershocks on F1.Recent studies proposed that afterslip and aftershock areas usually overlap with the rupture area of the mainshock (Hsu et al., 2006;Itoh et al., 2023;Milliner et al., 2020;Woessner et al., 2006;Wetzler et al., 2018).Milliner et al. (2020) analyzed the connection between the rupture area and the resultant aftershocks within 160 hr in the rupture area of the 2016 Kumamoto earthquake.They suggested that early afterslip loads in the rupture zone, lead to ruptures during aftershocks.This study also proposed that early afterslip could trigger aftershocks, despite the negative Coulomb stress changes in the rupture area (Dieterich, 1994).Itoh et al. (2023) also discovered that the distribution of aftershocks during different stages (27 hr, 2 days) overlapped with the afterslip distribution within the rupture area.They hypothesized that the early aftershocks were a result of the early afterslip following the 2014 Iquique earthquake.The inverted postseismic afterslip distribution in this study overlaps with the rupture area, which matches the location of the aftershocks on F1 shown in Figure 3a.This result is consistent with that of previous studies.Therefore, it may be reasonable to conclude that the occurrence of the aftershocks on F1 may be influenced by the early afterslip, albeit the negative Coulomb stress in the region.
Aftershocks on F2 occurred both inside and outside of the rupture area, whereas the aftershocks on F1 only occurred within the rupture area.The aftershocks that occurred along the rupture area on F2 overlapped with the inverted postseismic afterslip in regions where the ΔCFS was greater than 5 MPa.This indicates that, like on F1, the early aftershocks in the rupture area on F2 may be driven by the early afterslip, even though there are negative static Coulomb stress changes in this region.Furthermore, the aftershocks that occurred outside the rupture area on F2 also closely matched with the early afterslip, with a slightly less positive static Coulomb stress change of about 0.4 MPa.This suggests that the aftershocks outside the rupture area on F2 may also be attributed to the early afterslip.Nevertheless, these results suggest that the early afterslip may have a significant influence on the aftershock activity both within and outside the rupture area.
However, except static Coulomb stress changes (Harris, 1998;Harris & Simpson, 1992;King et al., 1994), dynamic stress changes (e.g., Gomberg, 2003;Hill et al., 1993) resulting from seismic wave propagation are also considered to impact seismic activity.For example, it has been proposed that dynamic stress changes resulting from seismic waves can explain the remotely triggered seismicity that occurred after the 1992 Landers earthquake (Hill et al., 1993).Moreover, Kilb et al. (2002) demonstrated that compared to static perturbations, dynamic stress changes better explain the distribution of aftershocks during the 10 months after the 1992 Landers earthquake.This implies that aftershock distribution may be influenced by dynamic stress triggering as well.It provides a possible explanation that the dynamic perturbations may influence the afterslip, thereby triggering aftershocks.
These results highlight the complex interplay between fault slip, stress changes, and seismic activity following large earthquakes.

Poroelastic Rebound
The poroelastic rebound (PE) in the oceanic crust predicts surface deformation mainly in a narrow zone near the trench (Hu et al., 2014).Therefore, this section focuses on a test model of poroelastic rebound in the continental crust where the effect of PF in the oceanic plate can be considered negligible due to its distance from the study area.The total effects of the time-dependent PE are defined as the difference of surface displacement calculated from a three-dimensional elastic coseismic model under undrained and drained conditions.Previous studies suggested that the top layer of 6 km of the continental crust is assumed to be poroelastic (Hu et al., 2014;Hughes et al., 2010).Following the previous study (Hu et al., 2014), the Poisson's ratio in the continental poroelastic layer is assumed to be 0.34 under undrained conditions and 0.25 under drained conditions.The shear moduli in the continental poroelastic layer, continental crust, and mantle wedge are 15, 48, and 64 GPa, respectively, for both the undrained and drained states.The horizontal displacement from PE is estimated to be up to 2 cm (Figure S13 in Supporting Information S1).The model also produces broader subsidence along the rupture area with a peak value of 7 cm and an uplift of approximately 1 cm at the southeastern boundary of the rupture area.
We take into account the poroelastic rebound effects by subtracting a portion of the poroelastic effects from the observed postseismic deformation.The remanent postseismic deformation was then used to invert the early afterslip.The trade-off between afterslip and poroelastic rebound in explaining the early postseismic deformation is thus assessed by varying the portion of the poroelastic effects from 0.001 to 1.The preferred model is the test model with the lowest fitting misfit and the portion of poroelastic rebound effects is determined to be 7.5% (Figure S14 in Supporting Information S1).Our results demonstrate that the observed subsidence in the near field is better explained by the test model including both afterslip and poroelastic rebound effects (Figure S15 in Supporting Information S1), while the pattern of the postseismic afterslip model is similar to that of AM.This indicates that afterslip remains the primary contribution to the early postseismic deformation, despite the potential influence of poroelastic rebound.
Studies have revealed that fluid movement seems to play a significant role in natural earthquake swarms that occur in the overriding plates of subduction zones (Bianco et al., 2004;Iio et al., 2002;Okada et al., 2015;Yoshida et al., 2016;Yukutake et al., 2011).Recent research (Kato, 2024;Nakajima, 2022;Nishimura et al., 2023;Yoshida et al., 2023) has also suggested that the upward fluid flow in the crust could drive large earthquakes and long-lasting earthquake swarms in the northeastern tip of the Noto Peninsula.Ishikawa and Bai (2024) have further demonstrated that deep fluid flow triggered the 2024 earthquake.Specifically, a thrust fault is typically characterized by a strong horizontal compression and lesser vertical stress, therefore the cracks around a thrust fault are prone to develop horizontally, which prevents vertical fluid migration.Consequently, this horizontal spread of fluids and strain accumulation across the broad expanse of the crust can result in prolonged earthquake swarms and large earthquakes.Our study verifies the hypothesis that poroelastic rebound may influence early postseismic deformation.The estimated poroelastic rebound predicts subsidence within the rupture area and uplift outside of the rupture area, indicating a potential outward flow of fluid from the epicenter.This result leads to a hypothesis that fluid may influence the occurrence and distribution of aftershocks, given the proximity of aftershock locations to regions affected by fluid dynamics.

Conclusion
In this study, we used GNSS data to determine the distribution of coseismic slip for the 2024 Noto Peninsula earthquake.Model results indicate that the earthquake ruptured up to 100 km along the southwestern side of the hypocenter, with a maximum slip of about 4 m at a depth of approximately 6 km.The northeastern side of the hypocenter, on the other hand, experienced little rupture.We have also estimated that the early afterslip occurred within the first 19 days following the earthquake.The inverted postseismic afterslip distribution shows 0.1-0.25 m of slip in the coseismic slip patch and 0.1-0.6 m of slip toward the north of the rupture area.We further calculate the static Coulomb stress changes of this earthquake.The significant negative values of earthquake-induced stress changes are in the rupture area.The afterslip and aftershocks also occurred in this region with negative stress changes.This implies that the aftershocks on the rupture area may be driven by the afterslip within the rupture area.Finally, the observed subsidence in the near-field is mainly due to the poroelastic rebound, indicating possible fluid flow in this region, which may influence the distribution of the aftershocks.This study indicates the potential impact of early afterslip and/or fluid flow on these aftershocks.

Figure 1 .
Figure 1.Tectonics of the Noto Peninsula.The black star represents the location of the 2024 Mw7.45 Noto Peninsula earthquake.Solid squares of different colors represent the aftershocks at different times since the main shock.Open triangles represent the locations of the Global Navigation Satellite System stations used in this study.Blue lines represent the location of the plate boundary.

Figure 2 .
Figure 2. Co-and post-seismic slip distributions.(a) Coseismic displacements.(b) Coseismic slip model.(c) 19-day postseismic deformation.(d) Early postseismic afterslip model.Green and black arrows in panels (a, c) represent the model and observed surface deformation, respectively.The solid squares in panels (a, c) represent the vertical observations, while the color contours in panels (a, c) represent the predicted vertical displacements.The color contours in panels (b, d) represent the modeled coseismic and postseismic afterslip on the fault, respectively.

Figure 3 .
Figure 3. Coseismic static Coulomb stress changes.Predicted static Coulomb stress changes on the F1 and F2 are shown in panels (a, b), respectively.The cyan dots, black crosses, and red crosses in panels (a, b) represent the 19-day aftershocks along the fault with the magnitude of 1-2, 2-3, and 4-5, respectively.The dashed red contour represents the coseismic slip distribution, while the dashed green contour represents the postseismic afterslip distribution.