Introduction

The geometry of an active fault controls its seismic potential and segmentation1,2,3 and dictates behaviors such as earthquake nucleation, dynamic rupture, stress triggering, and seismic wave propagation4,5. Therefore, for a nonplanar fault, it is necessary to establish a three-dimensional (3D) fault model to effectively characterize its subsurface complex structures6. Accurate geometric shapes of fault systems can better define seismic parameters, such as the fault area and fault length, and have significant impacts on estimating slip rates7,8 based on geodetic data and rupture dynamics9. The modeling of fault system behavior and dynamic rupture simulations shows that the geometric shape of faults plays a crucial role in fault behavior throughout the entire seismic cycle3,10,11. High-precision data of small earthquakes can intuitively reflect the geometric shapes of active faults through seismic distribution relationships12.

On 5 September 2022, an Ms 6.8 (Mw 6.6) earthquake occurred in Luding County, Sichuan Province (Fig. 1). It was the strongest earthquake in eastern Tibetan Plateau since the 2017 Jiuzhaigou M 7.0 earthquake. The Luding earthquake occurred at the intersection of the Baryan Har Block, the Sichuan-Yunnan Block, and the South China Block (Fig. 1). It is also near the Y-shaped junction area formed by three large-scale boundary fault zones—the Xianshuihe Fault Zone (XSHF), Longmen Shan Fault Zone (LMSF), and Anninghe Fault Zone (ANHF) (Fig. 1). Surface ruptures from field investigations, surface deformation by InSAR observations, focal mechanism, and relocated aftershocks all show that the seismogenic fault of this earthquake is the Moxi segment of the Xianshuihe Fault Zone, along with the Daduhe fault in the east and some previously unknown faults in the west13,14 (Fig. 2).

Fig. 1
Fig. 1
Full size image

Topographic and tectonic map of the study area in eastern Tibetan Plateau. (a) Tectonic map of the Tibetan Plateau region. The blue rectangle indicates the study area. (b) Map of main active faults, historical and modern strong earthquakes at the Y-shaped junction area between the Xianshuihe, Longmen Shan, and Anninghe Fault Zones in the Sichuan-Yunnan area. The yellow rectangle indicates the location of Fig. 2. DEM map based on the publicly available Shuttle Radar Topography Mission data with 90 m resolution. Focal mechanism solutions are collected from references15,16,17,18,19,20,21,22,23. The rupture zone are collected from references22. The focal mechanism solution is the lower hemisphere projection. Abbreviations: ANHF: Anninghe Fault Zone; BHB: Baryan Har Block; DLSF: Daliang Shan Fault Zone; LMSF: Longmen Shan Fault Zone; MZG: Mozigou fault; QTB: Qiangtang Block; SCB: Sichuan Basin; XSHF: Xianshuihe Fault Zone. Figures and maps were generated using QGIS 3.32.1 (https://download.qgis.org/downloads/) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

Due to the lack of data in the early stage after the earthquake, previous studies suggested that the Mozigou fault, which is nearly perpendicular to the Moxi segment, was the seismogenic fault of the aftershocks on the west side of the mainshock24,25. This result however is inconsistent with the focal mechanism solution of the subsequent inversion. After the mainshock of the Luding earthquake, many aftershocks occurred. In particular, two strong aftershocks with a magnitude of Mw 5.1 on 22 October 2022 and Mw 5.4 on 25 January 2023, respectively, occurred on the west side14,23. In contrast to the strike-slip mainshock, both of them were characterized by normal faulting mechanisms. The relationship between mainshocks and aftershocks with different faulting mechanisms remains unclear. To address this issue, therefore, it is necessary to depict the geometric shape of the seismogenic faults in detail and analyze Coulomb stress change in the Luding earthquake area based on the fine geometric model.

Since 1700, the XSHF has experienced at least sixteen (16) M ≥ 6.5 earthquakes, including eight M ≥ 7.0 earthquakes, and coseismic surface ruptures have occurred along almost the entire XSHF (Fig. 1). However, the Moxi segment has not undergone any M ≥ 6.5 earthquake since the 1786 M 7.3 Kangding earthquake22. The potential seismic hazard after the Luding earthquake on this segment remains unclear. Moreover, there is a seismic gap of 50–60 km long in the southern LMSF between the 2008 Wenchuan Ms 8.0 earthquake and the 2013 Lushan Ms 7.0 earthquake26. The 1 st June 2022 Lushan Ms 6.1 earthquake indicates that this seismic gap still has high seismogenic potential18. Furthermore, the northern ANHF can also be identified as a seismic gap, since it has not experienced an M ≥ 6.5 earthquake since the 1480 M 7.5 and the 1536 M 7.5 earthquakes. How the occurrence of the 2022 Luding earthquake affects the seismic hazard of the adjacent seismic gaps requires pressingly further research. Based upon two-dimensional fault models, previous researchers computed the Coulomb stress changes caused by this earthquake23,24,27. However, Coulomb stress changes are highly sensitive to geometric variations in faults, and results based on simplified two-dimensional fault models may neglect stress changes at fault bends related to rupture termination or nucleation4.

Therefore, this study constructs a detailed, 3D geometric model for the seismogenic fault of the 2022 Luding earthquake by employing 3D modeling techniques on the SKUA-GOCAD software platform and incorporating published results of relocated aftershock catalog and focal mechanism solutions (Fig. 2). We further conduct a 3D Coulomb stress change analysis for the area where the two M > 5 aftershocks occurred, as well as the faults in the Y-shaped junction area. Furthermore, in conjunction with long term Coulomb stress and fault coupling characteristics28this study analyzes the potential seismic hazard in the surrounding two seismic gaps.

Tectonic setting

Since the late Cenozoic, the ongoing convergence between the Indian Plate and the Eurasian Plate has led to extensive extrusion of active blocks in the southeastern Tibetan Plateau29. The Bayan Har Block extrudes eastward but encounters strong resistance from the Yangtze Craton of the South China Block30leading to the thrust-dominant LMSF31and the Sichuan-Yunnan Block rotates clockwise around the Eastern Himalayan syntaxis32 (Fig. 1a). As a significant boundary fault between these two active blocks, the Xianshuihe-Anninghe-Zemuhe-Xiaojiang Fault Zone exhibits an overall eastward convex arc-shaped structure, coordinating the left-lateral strike-slip motion between the two blocks33.

The NE-striking LMSF consists of a series of thrust faults and nappe structures (Fig. 1b). Although the overall shortening rate of the LMSF is only approximately 1–2 mm/a32,33, strong earthquakes have occurred frequently in recent decades, including the 2008 Wenchuan Ms 8.0, the 2013 Lushan Ms 7.0 and the 2022 Lushan Ms 6.1 earthquakes.

The NW-striking, 400-km-long XSHF is the boundary fault between the Bayan Har Block and the Sichuan-Yunnan Block in the Tibetan Plateau (Fig. 1b). There is a clockwise deflection of ~ 10° near Kangding. According to the geometry, fault discontinuity, and historical earthquake distribution, the XSHF can be divided into seven segments, including, from NW to SE, the Luhuo, Daofu, Qianning, Yalahe, Salaha, Zheduotang, and Moxi segments. Since the Holocene, the XSHF has been dominated by left-lateral strike-slip with thrust components and is one of the most active tectonic zones on the Chinese continent34. Geodesy shows that the average slip rate of the XSHF has been ~ 9–15 mm/a32,33, and geological investigations show that the average slip rate is approximately 10–17 mm/a35,36.

South of the XSHF, the ANHF exhibits an almost N-S trend37 (Fig. 1b). Geodesy shows a slip rate of ~ 5–7 mm/a38. No strong earthquake has been recorded in the fault zone since the 1480 M 7.5 earthquake and the 1536 M 7.5 earthquake, showing an obvious “seismic gap”.

Fig. 2
Fig. 2
Full size image

Distribution of the Luding Mw 6.6 earthquake sequence and focal mechanism solutions. DEM map based on the publicly available Shuttle Radar Topography Mission data with 90 m resolution. The aftershock sequence of the Luding earthquake is from reference23; focal mechanism solutions in red and black represent the mainshock of the Luding earthquake and the two M > 5 aftershocks, respectively23; the focal mechanism solutions in purple represent seven (7) Ms ≥ 3.0 aftershocks of the Luding earthquake14 (same below). The focal mechanism solution is the lower hemisphere projection. The blue line indicates the position of eight representative depth profiles of earthquake clusters in Fig. 4. DDHF: Daduhe fault; MXS: Moxi segment of the Xianshuihe Fault Zone. Figures and maps were generated using QGIS 3.32.1 (https://download.qgis.org/downloads/) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

Data and method

An accurately relocated aftershock catalog can be used to constrain the geometry of the fault plane and is the first and foremost data source for establishing a 3D model of a seismogenic fault39. This study collects 5269 aftershocks that occurred within 34 days after the mainshock of the Luding earthquake, as well as the focal mechanism solutions of the two M > 5 aftershocks relocated by Zhang et al. (2023) (Figs. 2 and 3; Table 1). This aftershock catalog, as well as the focal mechanism solutions, obtained by relocation using the HypoDD method40is based on the complete waveforms of 34 stations within 120 km of the epicenter, 13 of which are located within 50 km of the epicenter. It has lateral and vertical uncertainties of 190 m and 300 m respectively, with an average root mean square (RMS) residual of 22 ms. This study analyzes the cluster characteristics of the aftershock catalog in 3D space using the SKUA-GOCAD platform. We build a series of vertical profiles perpendicular to the strike of the aftershock cluster (Fig. 2) and map the fault traces on each profile based on the linear distribution characteristics of the aftershock cluster41. We then construct a 3D fault plane of the Luding earthquake. The focal mechanism solutions of seven Ms > 3 earthquakes calculated by Yi et al. (2023), using fixed station waveform data within 350 km of the epicenter and the Cut and Paste (CAP) waveform inversion, are also collected to constrain the fault plane occurrence (Figs. 2 and 3; Table 1).

Table 1 Focal mechanism solutions of the Luding earthquake sequence obtained from different sources.
Full size table
Fig. 3
Fig. 3
Full size image

Three-dimensional fault model, velocity model, and distribution of earthquakes in the Luding earthquake catalog in the Y-shaped junction area. The P-wave velocity model is from reference42and the Y-shaped junction fault model is from the China Seismic Experimental Site. For visualization purposes, the vertical exaggeration is 3x. Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/en/products/sse/aspen-skua) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

The summation of the shear and normal stress changes provides the Coulomb stress change (ΔCFS)43failure is encouraged if ΔCFS is positive and discouraged if negative44. We conducted our study using the expression.

where τ is the shear stress, σN is the normal stress and µ’ is the apparent coefficient of friction45.

To further quantitatively understand the seismogenic environment and seismic hazard in the area adjacent to the Luding earthquake, we calculate the ΔCFS on nearby faults induced by the Luding mainshock. We also incorporate the coupling data of the Y-shaped junction fault inverted by Li et al. (2023) using geodetic data, to investigate (1) why the strike-slip mainshock has dip-slip aftershocks and (2) the influence of the Luding earthquake on the seismogenic potential of the two seismic gaps in the southern LMSF and the northern ANHF. Numerous coseismic slip models have been developed for the Luding earthquake24,47. We adopt the model proposed by Li et al. (2022), as it is derived from a joint inversion of multiple geophysical datasets—including GPS measurements, seismic station records, and InSAR-derived static displacements—and incorporates complex fault system. We project it onto the 3D fault model of Moxi fault established in this study, which is interpreted as the seismogenic fault. And then, on the Relax platform48we use the fault models of the southern LMSF, ANHF, and Daliang Shan Fault Zone published by the China Seismic Experimental Site (Fig. 3) as the receiver faults to obtain the ΔCFS result. The Poisson’s ratio, friction coefficient, and Young’s modulus are set to 0.25, 0.4, and 80 GPa respectively.

Results and discussion

The 3D model of the seismogenic fault

Fault trace mapping

The aftershock sequence of the Luding earthquake can be divided into four distinct clusters in 3D space: (1) a cluster distributed along the mainshock seismogenic fault, i.e. the Moxi segment of the XSHF; (2) a cluster along the Daduhe fault; and (3) two sub-parallel aftershock clusters to the west of the mainshock (Fig. 2, Supplementary Movie S1). To characterize the 3D geometry of the fault planes associated with each aftershock cluster, we construct a series of vertical depth profiles perpendicular to the trend of these clusters, and the aftershocks are projected onto the profiles based on their distribution density within a given range: 2 km for the first two clusters and 1.5 km for the latter two clusters. In total, we establish 56 profiles, of which 8 representative profiles are presented here to illustrate the characteristics of the aftershock clusters on different seismogenic fault segments of the Luding earthquake (Figs. 2 and 4).

Lines 1–3 are located in the northern, central, and southern parts of the Moxi segment (f1) (Figs. 4a-c). The linear relationship in Line 3 is notably better than those in Line 1 and Line 2. It is evident that the fault dips to the northeast, consistent with the nodal plane of the focal mechanism. The aftershock distribution along Line 2 appears relatively scattered, and we interpret the fault plane as southwest-dipping when considering the focal mechanism solution of the mainshock and the distribution of aftershocks to the north and south of Line 2. Similarly, the northern fault plane is found to be northeast-dipping.

On the eastern side of the mainshock, the aftershock cluster generated by the Daduhe fault (f2) exhibits a lower number of aftershocks, and this region lacks constraints from focal mechanism solutions (Fig. 4d). Combining previous geological surface investigations on this fault, we determine that the fault dips to west-southwest49.

Fig. 4
Fig. 4
Full size image

Eight representative depth profiles of vertically-oriented earthquake clusters, lateral aftershock projection, and interpreted fault traces. (a-c) Lines 1–3 are perpendicular to the Moxi segment of the XSHF (f1), and (d) Line 4 is perpendicular to the Daduhe fault (f2). Aftershocks within a 2 km range on both sides of the profiles are projected on the diagrams. (e-h) Lines 5–6 and 7–8 are perpendicular to two newly identified faults, f3 and f4, to the west of the mainshock. Aftershocks within a 1.5 km range on both sides of the profiles are projected on the diagrams. The gray areas represent the data range, and the red lines represent the fault traces on the profiles. Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/en/products/sse/aspen-skua) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

The aftershocks on the western side of the Moxi segment can be divided into two sub-parallel clusters and are interpreted as faults f3 and f4. Focal mechanism solutions show that these faults are normal faults, with little strike-slip components. Considering the consistency of the focal mechanism solutions within the clusters, we interpret both fault planes as northwest-striking, southwest-dipping, with top-down-to-the-southwest sense of motion (Figs. 4e-h). This interpretation also aligns with the northwest-trending valleys within the Gongga Mountains in this region.

3D fault model

We build the 3D model for the four seismogenic faults based on the interpretation of 56 profiles (Fig. 5a, Supplementary Movie S2). The fault models (Figs. 5b-e) reveal that the depth constrained by the seismic clusters is up to approximately 15 km for the Moxi segment of the XSHF and the Daduhe fault. For the secondary faults f3 and f4, the aftershocks are distributed at shallower depths, so the deepest depth constraint for these faults is approximately 12 km.

The Moxi segment (f1) (Figs. 5a, b) has an approximate length of 55 km and an area of approximately 860 km2. It exhibits an overall NNW-striking, sub-vertical (77°−89°) attitude. Along the strike of the fault plane, the dip direction changes from NE in the north to SW in the central, and then back to NE in the south. The aftershocks in the northern and central parts of the fault are mainly concentrated within a depth range of 2–7 km, with a higher number of shallow aftershocks. In the southern region, aftershocks are primarily distributed at depths of 3–15 km, and the distribution is relatively uniform.

The Daduhe fault (f2) (Figs. 5a, c) is approximately 14 km long, and the fault plane has an area of approximately 213 km2. It strikes nearly due N and dips to the SWW overall, with a steep average dip angle of 88°. However, it is crucial to exercise caution when using the dip angle range and variations in the 3D model of this fault due to its relatively weak activity during the Luding earthquake and the limited number of associated aftershocks.

The aftershock clusters on the western side of the Moxi segment have provided a more substantial number of aftershocks and have effectively constrained the geometry of f3 and f4 (Figs. 5a, d, e). The 3D model indicates that these two fault segments are approximately 10 km and 8 km long, respectively, with an area of approximately 137 km2 and 119 km2respectively. Both f3 and f4 exhibit sub-parallel, NW-strike and dip to the SW, with dip angles of 55°−64° for f3 and 52°−66° for f4. The geometric characteristics of these two fault segments are highly similar. However, given that faults f3 and f4 were not identified in previous field investigations, their near-surface geometries should be treated with caution.

In the regions where the aftershocks are sparse, the exact intersections or connections among the four faults are not entirely clear. Additionally, because the aftershock catalog provides relative positions, the fault models of the Luding earthquake may deviate from the surface fault traces determined through active fault detection. However, this deviation does not hinder the characterization of the fault geometry.

Fig. 5
Fig. 5
Full size image

Geometric representations of the 3D fault model of the 2022 Luding Mw 6.6 Earthquake based on the aftershock catalog. (a) Overall 3D fault model. Solid triangles represent thrust faults, and hollow triangles represent normal faults. (b-e) Detailed representation of the four seismogenic faults, on which the contours represent the dips. The color bar for the dip angles of all four figures is shown in the upper left corner of (b). Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/en/products/sse/aspen-skua) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

The triggering effect of the luding earthquake on aftershocks

Seismic triggering refers to the process in which the stress changes associated with an earthquake can accelerate or delay seismic activity in surrounding areas or trigger other earthquakes at greater distances43. Coulomb stress loading on the surrounding faults caused by a strong earthquake event is one of the important factors triggering a aftershock50. For example, the regional stress changes caused by the 1992 Landers Mw 7.3 earthquake in the United States triggered the 1992 Big Bear Mw 6.3 earthquake only 3.5 h afterward51 and also the 1999 Hector Mine Mw 7.1 earthquake52. The 2008 Wenchuan Ms 8.0 earthquake had an obvious triggering effect on most of its strong aftershocks, and may trigger the 2013 Lushan Ms 7.0 earthquake53,54.

After the mainshock of the Luding earthquake, two M > 5 aftershocks occurred in the seismic area, namely, the Mw 5.1 aftershock on 22 October 2022 and the Mw 5.4 aftershock on 25 January 202323. Their seismogenic faults are the branch faults f3 and f4 west of the Moxi segment. In contrast to the strike-slip mechanism of the mainshock, these two aftershocks exhibited normal faulting characteristics. Before the Luding earthquake, the 7-year earthquake catalog shows these two branch faults had obvious microseismic activities55but none of the earthquakes were greater than M 5. Hence, this section aims to explore whether there is a stress-triggering effect between the Luding earthquake and its strong aftershocks with different rupture mechanisms.

Fig. 6
Fig. 6
Full size image

The Coulomb stress change (ΔCFS) on the branch faults f2, f3, and f4 caused by the mainshock of the 2022 Luding earthquake. The black and gray lines on the fault surfaces represent contours of Coulomb stress changes. The red and black focal mechanism solutions denote the mainshock of the Luding earthquake and the two M > 5.0 aftershocks, respectively23; the purple ones represent Ms ≥ 3.0 aftershocks in the Luding earthquake14. Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/products/engineering/skua-gocad), Relax 1.0.7 (https://github.com/geodynamics/relax/releases/tag/1.0.7) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

We use f1 as the seismogenic fault and f2-f4 as the receiver faults to calculate the Coulomb stress changes (ΔCFS). The result (Fig. 6) shows that the ΔCFS on f3 and f4 is generally positive. The range of the ΔCFS on f4 is 0.27 ~ 1.59 MPa, and for f3, the range is −0.07 ~ 1.37 MPa. The ΔCFS on ~ 90% of the fault surface is ≥ 0.01 MPa. Therefore, for f3 and f4, the ΔCFS on most fault planes far exceeds the threshold (0.01 MPa) of triggering effects56,57. Meanwhile, the epicenters of the two strong aftershocks coincide with regions of high positive ΔCFS on faults f₃ and f₄ (Fig. 6). Although the Coulomb stress distribution patterns are generally similar if we set the two branch faults (f3 and f4) as normal faults and strike-slip faults (with rake angles of −90° and 0°, respectively, and consistent 3D fault geometries), respectively, the normal fault exhibits slightly higher positive Coulomb stress values than the strike-slip fault. We infer that this is one of the main reason why the Luding strike-slip earthquake can trigger dip-slip aftershocks. And the higher ΔCFS on these two fault planes is consistent with the larger number of aftershocks in this area (Fig. 2).

Unlike f3 and f4, positive and negative values of ΔCFS appear alternately on the Daduhe fault (f2), ranging from − 0.29 to 0.85 MPa (Fig. 6). The ΔCFS values are generally lower than those on the two branch faults, which is also consistent with the fact that there are fewer aftershocks around this fault (Fig. 2). The consistency of the distribution of ΔCFS values on the fault plane and the number of aftershocks suggests that our calculated results are relatively reliable.

Seismic hazard analysis in surrounding fault zones

The seismic gap in the southern LMSF exhibits the features of low b-values and relatively high seismic apparent stress and has the stress conditions to trigger a moderate-strong earthquake58which is proven by the occurrence of the Ms 6.1 Lushan earthquake in 2022. From the perspective of the seismic moment deficit, a previous study states that this area has the seismic potential to generate a Mw 7.2–7.3 earthquake26and the seismic moment released by the 2022 Ms 6.1 Lushan earthquake is not enough to fill the above deficit (Fig. 1b). The northern ANHF is a “seismic gap” area22 (Fig. 1b), and the elapsed times of the latest strong earthquakes, the 1480 M 7.5 earthquake and the 1536 M 7.5 earthquake, have reached 543 a and 460 a, respectively. Paleoseismic research shows that the recurrence interval of strong earthquakes in the northern ANHF is 520–660 a37. Hence, this seismic gap is very close to or has entered the recurrence stage. This area has the potential for a Mw 7.2 earthquake59. These two fault zones have a high degree of coupling28 (Fig. 7), so the strains received by the faults within a depth of approximately 10 km in the southern LMSF and a depth of approximately 20 km in the ANHF have all accumulated and are basically not released through creep28,60. We envision that the energy accumulated in this area have not been released because this area lacks strong earthquakes.

Fig. 7
Fig. 7
Full size image

Interseismic coupling distribution of the main fault zone in the Y-shaped junction area of the three main fault zones. The coupling data are from reference28. ANHF: Anninghe Fault zone; DLSF: Daliang Shan Fault Zone; S-LMSF: southern Longmen Shan Fault Zone; XSHF: Xianshuihe Fault Zone; ZMHZ: Zemuhe Fault Zone. Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/en/products/sse/aspen-skua) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

Therefore, the southern LMSF and the northern ANHF have high seismogenic potential. What impact did the Luding Mw 6.6 earthquake have on these two fault zones? Coulomb stress changes along the North Anatolian Fault Zone in Türkiye indicate that the Izmit region was located in a positive ΔCFS area with elevated seismic hazard, where the 1999 İzmit Mw 7.4 earthquake subsequently occurred61. Hence, calculating the ΔCFS induced by strong seismic events in the surrounding fault zones can aid in a more precise evaluation of seismic hazard49,62,63. We use the Moxi segment of the XSHF as the seismogenic faults and the southern LMSF, the ANHF, and the Daliang Shan Fault Zone as the receiver faults to calculate the ΔCFS (Fig. 8).

Fig. 8
Fig. 8
Full size image

Coulomb stress change (ΔCFS) on faults in the Y-shaped junction area of the main fault zones induced by the mainshock of the 2022 Luding earthquake. (b) Enlarged view of the dashed box in (a). (c) Cumulative Coulomb stress in the northern ANHF, incorporating results from reference17,64. The black and gray lines on fault surfaces represent contours of Coulomb stress changes. Figure were generated using SKUA-GOCAD 2017 (https://www.aspentech.com/products/engineering/skua-gocad), Relax 1.0.7 (https://github.com/geodynamics/relax/releases/tag/1.0.7) and CorelDRAW Standard 2024 (https://www.coreldraw.com/en/product/coreldraw/standard/).

The results (Fig. 8) indicate that most faults are primarily within the positive ΔCFS region, except for a small area in the northern ANHF, which exhibits a noticeable negative ΔCFS, corresponding to the fault rupture unfavorable zone. However, only the southernmost LMSF and the northernmost ANHF surpass the ΔCFS threshold for triggering an earthquake (0.01 MPa). In comparison to another area, the northern terminus of the ANHF has a significantly higher ΔCFS value, reaching 0.43 MPa. The long-term accumulation ΔCFS along the Anninghe fault zone — incorporating co-seismic, post-seismic, and interseismic tectonic loading — indicates that the northern segment has accumulated over 1 MPa of positive ΔCFS over the past centuries17,64. Based on our calculations, even in the co-seismic stress-shadow (negative ΔCFS) area following the Luding earthquake, the accumulated ΔCFS remains above 0.8 MPa, while in regions with co-seismic positive ΔCFS, the accumulated ΔCFS reaches nearly 1.5 MPa. Given the high level of coupling and its proximity to the recurrence interval of strong earthquakes, we suggest that the ΔCFS induced by the Luding earthquake has significantly elevated the seismic hazard of the northern ANHF.

Conclusion

We have established a three-dimensional model of the seismogenic fault associated with the 2022 Luding Mw 6.6 earthquake using the aftershock catalog. The three-dimensional fault model contains the Moxi segment of the Xianshuihe Fault (f1), the Daduhe Fault Zone (f2), and two previously unknown faults, named f3 and f4 in this article.

Based on the new 3D fault model, we have calculated the three-dimensional Coulomb stress changes (ΔCFS) in this region. Our results suggest that the ΔCFS on f3 and f4 can reach a maximum of 1.59 MPa, evidently triggering the two strong M > 5 aftershocks nearby. It also shows that only the northern Anninghe Fault Zone has reached the earthquake-triggering threshold (0.01 MPa), with the maximum value reaching 0.43 MPa, resulting in a significant increase in seismic hazard. Our study demonstrates that incorporating a three-dimensional model in the calculation of Coulomb stress changes leads to notably improved accuracy in assessing the seismic triggering effects following great earthquakes, as well as their influence on seismic gaps.