Areas of active tectonics host many active faults and frequently experience moderate to large earthquakes. The possibility of devastating earthquakes makes the development of major infrastructure projects in these areas risky. World-class large-scale step hydroelectric projects have been built along the Jinsha River, such as the Xiangjiaba, Xiluodu, Baihetan and Wudongde reservoirs in the Daliang Mountains of the southeastern Tibetan Plateau. Using the SKUA-GoCAD modelling platform, we created a thorough 3D model of the active faults. Regional geological information, historical strong earthquake catalogues, small earthquakes with fine displacement and 3D seismic tomography are all integrated in this model. The Mabian–Yanjin fault belt consists of a number of discontinuous faults that are either exposed on the surface or concealed, according to the 3D fault model. Some destructive earthquakes, including two enormous Ms 7 and many moderate earthquakes, have occurred along this fault belt. Some pre-existing thrust faults, together with numerous immature faults in specific areas, may have been reactivated and changed into strike-slip faults. The Jinsha River basin's seismic and geological concerns must be carefully considered given the existence of such intricate fault networks and seismic activity.

Supplementary material: A video of the 3D fault model is available at

The Daliang Mountain block, located at the southeastern border of the Tibetan Plateau, is an active tectonic zone characterized by multiple active faults and significant seismic activity (Deng et al. 2003; Zhang et al. 2004, 2005; Wen et al. 2013; Xu et al. 2014). The area is traversed by the Jinsha River, home to several large-scale hydroelectric projects (Guo et al. 2022; Guo and Zhao 2023), including the Xiangjiaba, Xiluodu, Baihetan and Wudongde reservoirs (Fig. 1). This region has experienced a high level of seismic activity, with 10 major earthquakes of magnitude (M) six or greater recorded in the past millennium (Table 1). However, the seismogenic structures of some historical large earthquakes remain unknown, such as the 1216 Mahu M 7.0 earthquake in Sichuan (Cao et al. 1993). Previous studies have proposed the Leibo–Mahu fault, which strikes near east–west (Han et al. 2009), or the Mabian–Yanjin fault, which strikes near north–south (Guo 2016), as the potential seismogenic fault. Furthermore, there is continuing debate regarding the epicentre position and seismogenic fault for the 1844 Northern Daguan earthquake (Hou et al. 1999) and the 1974 Daguan–Yongshan M 7.1 earthquake in Yunnan (Han 1993).

The focus of this study is the Xiangjiaba–Xiluodu reservoir region located in the downstream area of the Jinsha River (Zhao et al. 2022), primarily owing to its significant seismic activity (Figs 1 and 2). The construction and operation of various hydropower facilities in the downstream area of the Jinsha River have led to a notable increase in small earthquakes and microearthquakes, which typically cluster around the reservoir area (Diao et al. 2014; Li et al. 2018; Luo et al. 2020; Su et al. 2020; Zhang et al. 2021). These seismic events are believed to be induced by the presence of the reservoirs, as they often originate from shallow depths (Yang 2020). However, some moderate to strong earthquakes have also occurred in the reservoir region (Zhao et al. 2022), and clusters of small earthquakes are common in areas adjacent to the reservoirs. Consequently, distinguishing between the natural and the induced earthquakes in large reservoir areas with active tectonic processes proves to be a difficult task owing to the mixed distribution of earthquakes.

Research and assessments on the active faults, seismotectonics and strong earthquake risks are crucial for the mitigation of potential seismic hazards. To accurately estimate the potential for earthquake rupture, a 3D geometric model of the major faults must be developed (Yue et al. 2005; Plesch et al. 2007; Shaw et al. 2015; Li et al. 2022; Lu et al. 2022). In this study, to find the seismogenic faults we evaluated the key characteristics of the seismogenic structures in the Xiangjiaba–Xiluodu region using data from fine relocated seismicity (Guo et al. 2022; Lei et al. 2022; Zhao et al. 2022) from 2010 to 2018. By combining information on local geology and focal mechanism solutions, we created a 3D model of the major fault belts. This study aims to provide crucial data for the identification of active, seismotectonic and seismogenic faults in the area. In particular, it reveals the presence of concealed active faults in the Mabian–Yanjin fault belt and the Xiluodu reservoir area. Our findings may serve as a reference for future assessments of seismic risks associated with strong earthquakes in the SE Tibetan Plateau.

The study area has a significant historical record of strong earthquakes with M 6.0 and greater. Notable examples include the 1216 M 7.0 Leibo–Mahu earthquake and the 1974 M 7.1 Northern Daguan earthquake (Fig. 2; Table 1). Moderate to strong earthquakes, with magnitudes of five and greater, are principally found in the Mabian–Yanjin fault region in strips oriented in a NNW direction. Previous studies suggested a strong correlation between these earthquakes and the Mabian–Yanjin fault belt (Zhang et al. 2005; Yi et al. 2010; Wen et al. 2013).

The Mabian–Yanjin, Lianfeng and Huayinshan fault belts, amongst others in the area, form part of a complex fault system (Fig. 2). The Mabian–Yanjin fault belt consists of several faults of varying orientations and short lengths. Previous studies have reported this fault belt to have experienced activity during the late Quaternary and even the Holocene period (Tang and Han 1993). Based on previous research on tectonic deformation, Han et al. (2009) hypothesized that the Mabian region belongs to a young seismotectonic belt. Furthermore, an abnormally low b-value zone has been observed for the Mabian–Yanjin fault belt, indicating that it is locked and susceptible to earthquakes (Xiang et al. 2010; Yi et al. 2010; Zhao et al. 2014).

There are over 10 prominent fault belts located in the downstream area of the Jinsha River (Fig. 2). Many of these faults, active since the Late Pleistocene, are also aligned with the Mabian–Yanjin fault belt (Xu et al. 2016). The Xiluodu reservoir began to store water in 2013, which subsequently resulted in the occurrence of numerous small earthquakes and two moderately strong earthquakes with a magnitude of five in close proximity to the reservoir (Luo et al. 2020). Previous research has suggested that these earthquakes may have been induced by reservoir activity (Diao et al. 2014; Su et al. 2020; Zhang et al. 2021).

The construction of 3D models of active faults relies on multivariate constraints derived from various data and observations, including surface fault traces and attitude, deep geophysical sounding, seismological observations, and regional tectonic and geological data (Lu et al. 2022). The Southern California Earthquake Center (SCEC) has developed advanced techniques for the 3D modelling of seismic activity (Carena and Suppe 2002; Plesch et al. 2007), particularly with the SKUA-GoCAD platform to create fault and velocity models for seismic risk assessment (Shaw et al. 2015). In this study, SKUA-GoCAD (2017, Paradigm, was employed to establish the 3D work region and import the data (Fig. 3a).

We collected data from 161 seismic stations, including permanent stations, Three Gorges stations and reservoir stations, from 2010 to 2018 (Duan and Zhao 2019). An initial 1D velocity model (Xin et al. 2018) was selected for tomography imaging using a method of combining 3D velocity structure and localization (Guo et al. 2022; Zuo et al. 2023). To accurately relocate small earthquakes, the double-difference seismic tomography imaging algorithm TomoDDMC was employed. Consistent wave velocity ratio models were used to constrain TomoDDMC. Out of the 23 175 relocation data points collected for the four reservoir regions in the downstream area of the Jinsha River, 10 331 were distributed across the study area (Fig. 3b). In addition, we adopted a deep S-wave velocity model (Liu et al. 2014) for the 3D seismic tomography imaging of the shallow seismogenic faults in the study region, to gain insights into the relationship between deep structures and shallow faults.

We adopted an operational workflow (e.g. Plesch et al. 2007; Lu et al. 2019) to construct the 3D fault model in the 3D modelling platform SKUA-GoCAD. All available datasets were projected to the Universal Transverse Mercator (UTM) coordinate system and combined into one project (Fig. 3). The P-wave velocity data were converted to a 3D seismic cube (also known as a ‘voxel’). Based on surface fault traces, fault strike and dip, the earthquake clusters and the location of strong earthquakes associated with the fault, we took these types of control data and constructed a spatial 3D fault model using the discrete smooth interpolation in GoCAD (Mallet 1992).

The study region lacks high-resolution geophysical data, such as seismic reflection profiles, making it challenging to precisely define the deep fault structures. Thus, we initially constructed the 3D fault model based on the surface features of the prominent fault belts (Fig. 2; Table 2) and the spatial occurrence of earthquakes, focusing on strong earthquakes with a magnitude greater than or equal to six (Fig. 3).

Previous studies have acquired magnetotelluric and wide-angle reflection or refraction data to analyse and interpret the deep geometry of the faults in the study area (Wan et al. 2010; Yang et al. 2011; Wen et al. 2013). Most of the major faults are located within the upper crust at depths less than 20 km, and there is no evidence of typical faults penetrating the Moho surface (Cheng et al. 2017). In this study, we established 3D models and key parameters for 16 major faults, including the Meigu–Yiping fault (F1), the Sanhekou–Yanfeng fault (F2) and the Jinyang fault (F3) (Table 2; Fig. 4). The latest activity periods of these major faults were determined based on previous studies (Cao et al. 1993; Zhang et al. 2005; Han et al. 2009; Xu et al. 2014; Guo 2016).

The 3D fault model provides a clear representation of the spatial distribution of significant faults within the study area (Fig. 4). It offers valuable information regarding the fault characteristics, including the length, orientation, dip, segmentation, penetration depth and intersections (Supplementary Material). It also serves as a crucial basis for the evaluation of the regional crust stability, and the design and construction of large-scale hydropower plants and other infrastructure projects, as well as the assessment of seismic rupture behaviour and associated risks (Avouac et al. 2015; Cesca et al. 2017; Klinger et al. 2017; Li et al. 2022).

Analysis of the small-earthquake relocation data from 2010 to 2018 (Duan and Zhao 2019; Guo et al. 2022; Zhao et al. 2022) reveals two distinct clusters in the study area. The first cluster, characterized by minor earthquakes, is observed in the Mabian–Mahu region (Figs 5 and 6), whereas the second cluster is located in the Xiluodu reservoir area along the Jinsha River (Figs 5 and 7). In the subsequent analysis, we focus on these two earthquake clusters and their corresponding seismogenic structures.

Mabian–Yanjin fault belt

The Daliang Mountain active block is bounded on the east by the Xingjing–Mabian–Yanjin fault system (Fig. 1). This fault system forms a 270 km long fault belt extending from Tianquan in the north to Yanjin in the south and consists of the Ebian–Mabian fault, the Lidian fault, the Zhongdu fault, the Manao fault, the Zhaziba fault and the Zhongcun fault (Zhang et al. 2005). These faults have a general orientation of N20°–30°W and are characterized by short segments with intermittent activity. In this study, the Mabian–Yanjin fault belt specifically refers to the active faults between Mabian and Yanjin (Fig. 2). Previous research suggests that this fault belt is prone to experiencing severe earthquakes, as well as frequent minor and medium-sized earthquakes (Yi et al. 2010).

For an active fault, the size of the fault and strong seismic activity are indicative of potential seismic risk (Wells and Coppersmith 1994; Wesnousky and King 2007).

According to the empirical formula of Wells and Coppersmith (1994) and the earthquake rupture scaling relations for Mainland China (Cheng et al. 2019), the Mabian–Yanjin fault belt (Figs 2 and 4) in the study region still has the potential for large earthquakes (M ≥ 7.0).

The results of the small earthquake relocation analysis (Fig. 5) indicate the presence of a narrow strip of small earthquakes, approximately 40 km in length, striking in a NNW direction to the east of the Manao fault (F5) (Fig. 6). The majority of these small earthquakes have an estimated depth of around 5 km, with a large amount also observed within the range of 5–10 km. Based on these findings, the Xiaxi fault (F15) is considered to represent a concealed active rupture.

In 1970, an M 5.4 earthquake occurred in this group of small earthquakes, exhibiting a strike-slip focal mechanism solution (Xie et al. 2003) (Fig. 6). This suggests that the Xiaxi fault may have transformed into a left-slip fault. Furthermore, there are three sets of NNE-trending, minor earthquake clusters oriented approximately perpendicular to the Xiaxi fault. The fault zone separating these two sets of active faults may be classified as a conjugate fault and its advancement is consistent with the current maximum horizontal stress field (SHmax) in the region, with NW to SE motion (Xiang et al. 2010; Zhang 2020).

Seismogenic characteristics of the Xiluodu reservoir area

The Xiluodu Hydropower Station, located at the confluence of Leibo and Yongshan along the main channel of the Jinsha River, has a dam height of 285.5 m and a total reservoir capacity of 12.67 billion m3. At present, this facility is the fourth largest hydroelectric plant in the world and the third largest in China. The water storage process began in May 2013 and terminated in September 2014 (Luo et al. 2020). A significant number of minor seismic events have been recorded following the filling of the reservoir (Diao et al. 2014), including two moderate to strong earthquakes with magnitudes of five close to the reservoir in 2014 (Fig. 7).

The seismic activity in the Xiluodu reservoir area is primarily observed along the banks of the Jinsha River, with a predominant depth of approximately 3 km (Fig. 7). It should be noted that previous studies have not identified any active faults with a high density of minor earthquakes in this region. However, the introduction of reservoir water into caves and its subsequent infiltration into fissures and joints must have increased pore pressure along pre-existing small fault surfaces and discontinuous surfaces (Diao et al. 2014). This rise in reservoir water loading can enhance the elastic deformation while simultaneously reducing both frictional strength and rock rupture strength. The combined effects of these factors are likely to contribute to the occurrence of shallow seismic activity in this area (Diao et al. 2014).

Seismic events with magnitudes less than three are generally not induced by conventional fault activity, rather they can generally be attributed to various factors such as minor fractures or the collapsing of caves (Feng et al. 2015). However, the occurrence of moderately strong earthquakes with magnitudes exceeding five typically requires the presence of faults of a certain size (Cheng et al. 2019). Two consecutive earthquakes with magnitudes of five occurred in 2014 within the Xiluodu reservoir area, indicating the possible existence of pre-existing concealed faults in the region (Fig. 7). Previous investigations have attributed some of the earthquakes occurring in Xiluodu to a seismic zone located at 5–15 km depth in the upper crust (Duan and Zhao 2019). Analysis of small earthquake clusters and the focal mechanism solutions of the two M 5 earthquakes (Fig. 7) implies that the study area harbours NE- to NNW-striking, concealed faults (Su et al. 2020). However, it is difficult to map the 3D fault geometry, because these small earthquakes are clumped in multiple directions. We inferred some concealed faults according to the focal mechanisms of two strong earthquakes (Fig. 7).

Northern Daguan magnitude 7.1 earthquake structure

On 11 May 1974, a strong earthquake with a magnitude of 7.1 occurred at the intersection of Yongshan and Daguan in Yunnan province, China (Fig. 8). The source area of this earthquake was determined by analysing the damage extent and the distribution of aftershocks, as no surface rupture was immediately observed (Yi et al. 2010). The location of the epicentre has been a subject of debate. Some researchers have positioned it to the north of the Lianfeng fault band (F7), whereas others have suggested a southern location (Han et al. 2009; Wen et al. 2013). The Northern Daguan earthquake is considered to be associated with the Lianfeng fault belt activity owing to its close proximity to the epicentre.

Liu et al. (1977) proposed that the seismogenic fault principally responsible for the Northern Daguan earthquake is a NNW-trending strike-slip fault. The aftershocks that occurred between 11 and 18 May 1974 exhibited a similar NNW–SSE trend, as well as the focal mechanism solution depicted in Figure 8. Furthermore, the rupture characteristics observed during the Northern Daguan earthquake sequence indicate the presence of bidirectional conjugate faults (Zhang et al. 1994). The epicentre location (Fig. 2, Table 1) implies the presence of a concealed fault in the NNW direction, referred to as the Northern Daguan fault (F14), which ruptured during this earthquake along with the conjugate Xintian fault (F13) (Fig. 8).

Conjugate faults are a common fault type, and numerous large earthquakes occurring in China have been associated with conjugate seismogenic faults (Guo et al. 2015). In relation to these, conjoint surface ruptures are frequently observed in many seismic zones (Zhang 1991). Such surface ruptures in conjugate directions include the 2012 M 8.6 earthquake that occurred in the northern Sumatra Sea (Yue et al. 2012) and the 2014 M 6.1 earthquake in Ludian County, Yunnan, China (Zhang et al. 2014).

Conjugate ruptures characterized the seismogenic structures of the 1216 Sichuan Leibo–Mahu M 7.0 earthquake (Fig. 6), as well as other significant earthquakes that occurred along the Mabian–Yanjin fault belt (Fig. 8). The occurrence of these earthquakes and the development of conjugate faults are closely associated with the prevailing stress field resulting from the southeastward extrusion of the southeastern Tibetan Plateau (Wang and Shen 2020; Sheng et al. 2022). The strike-slip faults concealed within the study area may be relatively young or newly formed faults (Han et al. 2009).

Major faults are widely distributed throughout the upper crust, and most of the moderately strong earthquakes (M 5.0) in the Xiangjiaba–Xiluodu reservoir areas occur at shallow depths of less than 20 km (Fig. 9). A convergent intersection is observed at 20 km depth between the NE-trending Leibo–Mahu fault belt and the NW-trending Mabian–Yanjin fault belt. This indicates the potential for cascade ruptures in the region as a result of other major earthquakes.

The 3D seismic tomography imaging determined from the S-wave velocity model (Liu et al. 2014) reveals a steep slope zone (Moho ramp) in the deep region of the study area. This zone is associated with a concentration of strain, similar to the Longmen Shan thrust belt in the eastern Tibetan Plateau (Lu et al. 2019; Tan et al. 2019). The study area has undergone significant denudation (Deng et al. 2016) and has experienced two magnitude seven earthquakes. Considering that the regional geological settings have the potential to generate severe and large-scale earthquakes, assessing the seismic hazard risk in the study area is an important task.

This study investigated previous large earthquakes and significant faults in the Xiangjiaba–Xiluodu reservoir region in the downstream area of the Jinsha River. We employed the SKUA-GoCAD software platform to construct a comprehensive 3D model of the primary fault systems within the study area. Datasets on small earthquake fine relocation, focal mechanism solutions, regional geology and deep structures were integrated into the model. The study presents findings on the dynamic tectonic and deformation properties of the Daliang Mountain active block. Our analysis suggests that the southeastward movement of the tectonic plates in the SE Tibetan Plateau led to the formation of numerous conjugate faults, as well as the transformation of previous thrust faults into strike-slip faults. As a consequence, a substantial number of strong earthquakes in the study area are of strike-slip type. However, the presence of small earthquake clusters indicates the potential activity of concealed or recently formed faults. This highlights the possibility of a cascade rupture of active faults, which poses significant challenges for the accurate assessment of seismic risk in the region.

We acknowledge data support from the China Earthquake Network Center National Center for Earthquake Science Data (, a seismic mechanism solution plug-in from Harvard University, source mechanism solution data from Shuzhong Sheng of Donghua University of Science and Technology and helpful discussions from Xueze Wen. We are grateful to the Subject Editor, Yuntao Tian, and two anonymous reviewers for their significantly comments and suggestions that improved the paper.

RL: conceptualization (lead), data curation (equal), funding acquisition (equal), software (supporting), writing – original draft (lead); CZ: data curation (supporting), funding acquisition (supporting); JZ: formal analysis (equal), writing – review & editing (equal); QW: data curation (equal); XS: investigation (equal); FX: investigation (supporting); HS: investigation (supporting)

This study is supported by the National Key Research and Development Project of China (Grant No. 2021YFC3000600) and the project ‘The seismogenesis and discrimination methods of cascade reservoir in the Lower reaches of the Jinsha River’ (JG/20023B) from the China Three Gorges Construction Engineering Corporation.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All data generated or analysed during this study are included in this published article and its supplementary information file.

This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License (