Introduction

Landslides always cause significant loss of life and property1. Their stability analysis and spatial distribution patterns have attracted much attention, which can provide support for predicting the probability of potential landslides in a region2,3. Additionally, the spatial distribution pattern of earthquake-induced landslides contains earthquake-related information4. The distribution pattern of earthquake-induced landslide including the information on landslide number density, landslide area percentage, type, and size, is crucial for understanding the causal mechanisms resulting from the complex multifactorial interactions5,6,7. These factors include topography (elevation, orientation, slope angle, slope curvature), seismic parameters, slope materials (soil cover and lithology), hydrology, geomorphology, and land use8,9. However, when it comes to earthquake-induced landslides, seismic endogenous factors, such as seismogenic fault, focal mechanism, peak ground acceleration (PGA), earthquake magnitude, depth, intensity, and coseismic slip, provide more direct insight into the underlying mechanisms10,11,12,13,14,15,16. Previous studies have found a positive correlation between earthquake magnitude and the distance from the outermost landslides to the seismogenic fault10,11,12,13,14,15,16,17, as well as a negative correlation between the distance from the epicenter and the number of landslides18. Recent research suggests that the distribution patterns of coseismic landslides are closely related to seismogenic structure and the seismic rupture process11,19,20,21,22,23. Examples of such distribution patterns can be seen in landslides triggered by 2005 Kashmir earthquake24, 2008 Wenchuan earthquake5,9, 2010 Yushu earthquake25 and 2020 Qiaojia earthquake26. Typically, the distribution pattern of coseismic landslides triggered by the 2022 Ms 6.8 Luding earthquake is strongly influenced by the earthquake’s endogenous factors. Numerous studies analyzing thousands of coseismic landslides have highlighted a clear association between their distribution and the seismogenic fault27,28,29. Landslide density is closely correlated with earthquake intensity and proximity to the macroscopic epicenter30,31. Areas with the highest landslide density coincide with regions experiencing the greatest surface rupture displacement32. Additionally, as the rupture extends deeper into the shallow portion of the fault plane, the frequency of landslides increases32.Thus, understanding the distribution patterns of coseismic landslides offers valuable insights into the characteristics of seismogenic faults and earthquake rupture processes, helping to identify the seismic endogenous factors that contribute to variations in damage.

On October 7, 2014, an Ms 6.6 earthquake struck Yunnan Province, China, following the Ms 6.5 Ludian earthquake on August 3, 2014 (China Earthquake Networks Center (CENC), http://www.cenc.ac.cn). The former, known as the Jinggu earthquake, occurred in Jinggu County, at the southwestern boundary of the Sichuan-Yunnan block, while the Ludian earthquake took place at the eastern boundary of the same block33,34,35 (Fig. 1). Although both earthquakes had similar magnitudes, the damage they caused was highly different. The Ludian earthquake caused significant geological hazards, severe casualties, and extensive damage to houses and other infrastructure36 After the event, Xu et al.4 identified 1024 landslides larger than 100 m2 using visual interpretation of satellite images with resolutions exceeding 2 m, alongside field photos and aerial images37. Tian et al.38 conducted a spatial distribution analysis of pre-earthquake and coseismic landslides triggered by the Ludian earthquake, examining seven factors: elevation, slope, aspect, curvature, lithology, seismic intensity, and river proximity. Chen et al.2,39 found that the relative landslide area ratio was strongly correlated with terrain factors and the seismic source process, based on an analysis of 114 coseismic landslides greater than 1000 m2 Chang et al.40 studied the largest landslide induced by the Ludian earthquake and identified contributing factors. Wu and Xu41 later compiled a more detailed and accurate inventory of the landslides triggered by the Ludian earthquake using pre- and post-earthquake high-resolution (~ 0.5 m) satellite images. Several studies also examined aftershocks42,43, surface ruptures, and the seismogenic fault associated with the Ludian earthquake44,45,46,47,48,49. In contrast, the Jinggu earthquake caused relatively minor damage. Some researchers have investigated its strong-motion records50,51, gravity fields52, seismogenic fault and rupture process53,54, earthquake sequence relocations, and focal mechanism solutions54,55. However, it wasn’t until 2021 that He et al.56 reported a more complete database of coseismic landslides triggered by the Jinggu earthquake and analyzed their relationship with topographic factors. Additionally, the some non-kinetic factors, such as human activity, topography and geomorphology, lithology, and vegetation50,57,58, have been discussed as possible reasons for contrasting damage caused by the two events,. From a seismic dynamics perspective, Xu et al.59 suggested that faster anelastic attenuation may have contributed to the more severe damage during the Ludian earthquake. Other researchers have inferred that the seismic rupture process likely concentrated landslides and aftershocks in the southeast of the hypocenter, leading to the severe damage during the Ludian earthquake37,39,47. To date, there has been no comprehensive discussion comparing the seismic rupture models of these two events and their effects on seismic damage, particularly in relation to the distribution patterns of coseismic landslide.

Fig. 1
figure 1

Tectonic maps of study area. (a) Tectonic setting of Qinghai-Tibetan plateau and adjacent regions. (b) Tectonic setting of the 2014 Jinggu earthquake and Ludian earthquake in the Sichuan-Yunnan rhomb block, southeast Qinghai-Tibetan plateau. The maps were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

This paper focuses on the effect of seismic parameters, such as PGA, earthquake magnitude and intensity, and distances to epicenter and seismogenic fault, on the distribution pattern of coseismic landslides. It also compares the differences in rupture processes between the Jinggu and the Ludan earthquakes. Our goal is to uncover the seismic-dynamic reasons behind the significant disparity in earthquake damage between these two events, despite their similar magnitudes.

Study area

According to the China Earthquake Networks Center (CENC http://www.cenc.ac.cn), an Ms 6.6 earthquake struck on October 7, 2014, at 23.4° N, 100.5° E, approximately 26 km southwest to Jinggu County, Yunnan Province, China (Fig. 1), with a focal depth of 5 km. The highest intensity, based on Chinese seismic intensity scale (CSIS), reached degree VIII, with a longitudinal axis oriented in a northwest direction (Fig. 2a). The Jinggu earthquake caused one fatality and 324 injuries.

Fig. 2
figure 2

Detailed tectonic setting, seismic intensity zones, and landslides distribution of the 2014 Jinggu earthquake (a) and Ludian earthquake (b). The two figures were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Earlier that year, on August 3, another earthquake, rated Ms 6.5, hit Ludian County, Yunnan Province, China (Fig. 1). Its epicenter was located at 27.11° N, 103.35° E, with a focal depth of 12 km44,45,46. The maximum intensity on the CSIS scale reached IX. This event resulted in extensive casualties, with 617 deaths, 112 people missing, and 3143 injuries59.

Both earthquakes occurred near the Sichuan-Yunnan block (SYB), one of the most seismically active regions in China60 (Fig. 1). Since the twentieth century, the area within 100 km of the Jinggu earthquake’s epicenter has experienced 38 earthquakes greater than M 5.0, 12 greater than M 6.0, and 2 above M 7.051,55. The Jinggu earthquake occurred in a seismic gap61 on the southwest of the SYB. In contrast, the Ludian region has seen numerous moderate-to-strong earthquakes in recent years (Fig. 1), including the 1974 M 7.1 North Daguan earthquake, the 2003 Ms 5.0 and 5.1 Ludian earthquakes, the 2004 Ms 5.6 Ludian earthquake, the 2006 Ms 5.1 Yanjin earthquake, the 2012 Ms 5.7 and 5.6 Yiliang earthquakes, and the 2014 Ms 6.5 Ludian earthquake47.

Tectonically, the Jinggu earthquake is located at the northwestern extension of the western branch of the Wuliangshan fault, specifically along the Puwen fault (expressed as a dotted line in Fig. 2a), east of the Lancangjiang fault54. Focal mechanism solutions indicate that the earthquake was triggered by a dextral strike-slip fault. Based on the distribution of intensity, aftershocks, and landslides, it is inferred that the seismogenic fault of the Jinggu earthquake is part of the northwestern segment of the Wulianshan fault zone, specifically, the extended Puwen fault, which is a blind fault37,53,54 (Fig. 2a). The Wulianshan fault zone in western Yunnan stretches over 220 km in a NNW direction53 and is composed of three secondary faults: the Mohei, Pu’er and Puwen faults. The Puwen fault starts at the southern end of the Jinggu Basin, arcs slightly westward, passing west of Simao and Puwen, before extending northward of Mengla in a nearly north–south direction after reaching Mengxing. It is considered an active Quaternary fault57. In contrast, The Ludian earthquake occurred east of the Xianshuihe–Xiaojiang fault system (XXFS), which lies between the SYB and the South China block61,62,63,64,65,66,67 (Fig. 1). The XXFS intersects the Zhaotong–Lianfeng fault zone, a NE-SW striking fault located near the Ludian earthquake’s epicenter. However, the observed primary surface ruptures suggests that the seismogenic fault was an unmapped active fault, the sinistral strike-slip Xiaohe-Baogunao fault42,45,46,47,68. This fault intersects the Zhaotong–Lianfeng fault zone in Longtoushan Town and displays NW-trending, discontinuous linear topographic features (Fig. 2b).

Topographically, both earthquake regions mainly are characterized by plateau mountains with valleys and basins. However, the terrain around the epicenter of the Jinggu earthquake is less rugged59 and Jinggu County enjoys better ecological environment, with approximately 70% of the land covered by forests.

Lithologically, Ludian seismic area is predominantly composed of basalt and carbonate rocks, which are more susceptible to erosion and weathering50. In contrast, the Jinggu earthquake area mainly features the sedimentary rocks such as sandstone, mudstone, and some tuffs.

Data and methods

Landslide dataset

The Jinggu earthquake triggered some small- to medium-sized landslides. This study builds on a coseismic landslide inventory obtained through visual interpretation of satellite images56. The visual interpretation method is available for identifying coseismic landslides in mountainous areas, as it distinguishes between the sliding area and surrounding terrain based on microtopography, tone, texture, and other markers. By comparing pre- and post-earthquake images, we can identify and exclude landslides that formed before the earthquake. The pre-quake images include Gaofen satellite images with a resolution of 2 m and RapidEye images with a resolution of 5 m, both taken on May 17 and 18, 2014. The post-earthquake images consist of 0.8 m resolution fused Gaofen images and 5 m resolution RapidEye images, captured on October 9 and 10, 2014. From the interpretations of these images, 441 coseismic landslides were identified, covering an area of 1.08 km2 within a 2160.86 km2 elliptical study area, which corresponds to the VII and VIII seismic intensity zones (Fig. 2a). The area of these landslides ranges from 86.47 to 56,143.90 m2, with about 58% being smaller than 1000 m2. The landslides are relatively concentrated in the southeast of the epicenter and are generally distributed along a NW–SE trending zone (Fig. 2a). The scale and frequency of these earthquake-triggered landslides follow a logarithmic relationship, indicating the completeness of the coseismic landslide inventory9,69. To date, this inventory represents the most comprehensive published record of landslides induced by the 2014 Jinggu earthquake.

For comparative analysis, we also cite the coseismic landslide inventory of the 2014 Ludian earthquake, compiled by Wu and Xu41. This inventory is based on the interpretation of ultra-high resolution (0.5 m) remote sensing images before and after the earthquake on Google Earth platform. Their results indicate that the Ludian earthquake triggered at least 10,559 landslides covering a total area of 14.975 km2 within an elliptical region of about 360 km2. The total volume of these landslides is estimated at 1.24 × 108 m3. The smallest landslide is 13 m2, while the largest is 345,292 m2. Of there, 9948 landslides have an area greater than 100 m2, a number much larger than 1024 landslides recorded in the inventory of Xu et al.37. The long axis of the overall landslide distribution is oriented NW–SE, consistent with the direction of the seismogenic Xiaohe-Baogunao fault (Fig. 2b). This inventory also follows a logarithmic relationship between the scale and frequency of landslides, which indicates that the coseismic landslides are relatively complete9,69.

Research methods

The occurrence of coseismic landslides is affected by various factors, such as seismological, geological, and topographical conditions. Strong ground shaking and significant surface deformation caused by earthquake are the direct triggers for coseismic landslides69. Areas near the earthquake epicenter and the seismogenic fault, where seismic energy is released, are particularly prone to landslides9,21,22,37,70. Therefore, in this study, the main seismological factors considered for the distribution of landslides triggered by the Jinggu and Ludian earthquakes include the seismogenic fault, PGA, epicenter, and seismic intensity. Fault-related factors include the vertical and parallel distances to the seismogenic fault. Vertical distance refers to the shortest distance to the fault line, while parallel distance measures the distance along the fault strike to the northwest point where the study area boundary touches the perpendicular to the fault line.

Landslides can be represented as either points indicating their location or polygons representing their scale71. In this work, landslides were initially delineated as polygons based on their shapes, and then converted to points located at the center of gravity of each polygon. The shapes of the landslides were determined by analyzing the boundaries of sliding areas and the surrounding terrain. Buffer zones were created for each factor, and the study area was classified into several grades intervals. For each interval, we calculated the number and area of landslides, examining correlations between these factors and coseismic landslides through two measures: landslide number density (LND) and landslide area percentage (LAP). The detail analysis method is outlined in previous studies26,72,73,74. Furthermore, we infer the seismogenic characteristic and rupture processes of the 2014 Jinggu earthquake and Ludian earthquake by analyzing the distribution models of coseismic landslides and aftershocks.

Correlation between coseismic landslides and seismology factors

Seismogenic fault

The Lancangjiang, Jinggu, and Puwen faults are the primary structural features around the Jinggu epicenter (Fig. 2a). Based on previous research53,54,55, we specifically examine the role of the Puwen fault as the seismogenic fault responsible for the occurrence of coseismic landslides. Since the northwest section of the Puwen fault is a blind fault and its exact location cannot be precisely determined, the extension of the identified southeastern section of the Puwen fault is used as the fault location for this analysis (Fig. 3a). The study area was divided into 23 bands parallel to the fault extension, each with a 2 km interval. Figure 3b shows that multiple peaks in LAP and LND appear in the region 4–14 km from the fault extension line. Although landslide distribution patterns of the strike-slip earthquakes are generally symmetrical4,22, it may also be altered with fault geometry.

Fig. 3
figure 3

The correlation between landslide distribution and the seismogenic fault of the Jinggu earthquakes. (a, b) Vertical distance to the extension line of the Puwen fault; (c) and (d) Parallel distance along the extension line of the Puwen fault. LAP: landslide area percentage; LND: landslide number density; CA: Classification area. The panels (a) and (c) were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Additionally, the study area is classified into 31 bands perpendicular to the Puwen fault, again with an interval of 2 km, extending from the northwest to the southeast (Fig. 3c). The highest LND presents in the band with the distance of 48 km, and both LAP and LND in southeast is obviously larger than that in northwest, which suggests that landslides are concentrated in the southeastern segment of the fault (Fig. 3d).

The Xiaohe-Baogunao fault (XBF) is considered the seismogenic fault of the Ludian earthquake, according to the field investigations, aftershock distribution, and focal mechanism solution42,45,46,47,68. Figure 4a,b show the landslide distribution in bands parallel to the XBF. The majority of landslides are concentrated within 14 km on both sides of the fault. The LND generally decrease with increasing distance from the fault. Its peak occurs within 2 km of the fault at a value of 12 km−2. The LAP peak of 2.5% is observed in the band 2–4 km from the fault. This pattern indicates that the landslides induced by the Ludian earthquake conform to the typical near-fault distribution.

Fig. 4
figure 4

The correlation between landslide distribution and the seismogenic fault of the Ludian earthquake. (a, b) Vertical distance to the Xiaohe-Baogunao fault; (c, d) Parallel distance along the Xiaohe-Baogunao fault. LAP: landslide area percentage; LND: landslide number density; CA: Classification area. The panels (a, c) were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Figure 4c,d display landslide distribution along the fault’s length. Most landslides are concentrated in bands 22–36 km from the starting point, with the epicenter located in the 24–26 km band. Peaks in LND and LAP are found in the bands 26–28 km and 30–32 km from the starting point, respectively. Taking the epicenter as the reference point, the landslide number and area densities are higher in the southeastern section of the fault compared to the northwestern section. Furthermore, Ludian earthquake-induced landslides extend about 10 km to the northwest and 14 km to the southeast from the epicenter along the seismogenic fault.

Peak ground acceleration

The PGA data are collected from United States Geological Survey (USGS, https://earthquake.usgs.gov/earthquakes/eventpage/usb000sjim/executive). The PGA of the Jinggu earthquake is between 0.06 and 0.30 g, which are divided into 13 classes (Fig. 5a). As shown in Fig. 5b, the LND begins to roughly increase with increasing PGA, but decreases significantly at 0.18 g and 0.30 g. The LAP also increases roughly with the increase of PGA at the beginning, but then fluctuates irregularly.

Fig. 5
figure 5

The correlation between landslide distribution and the peak ground acceleration (PGA) in the Jinggu earthquake (a, b) and the Ludian earthquake (c, d). LAP: landslide area percentage; LND: landslide number density; CA: Classification area. The panels (a, c) were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

In the Ludian event, the PGA is between 0.12 and 0.50 g, which are divided into 20 classes (Fig. 5c). As the PGA increases, the LND generally rises, whereas the LAP begins to decline from a PGA of 0.38 g onwards (Fig. 5d). This indicates that in areas with large PGA values, the number of landslides increased, but their sizes became smaller.

Theoretically, PGA is the maximum value of absolute ground acceleration during an earthquake, which is used to describe the maximum impact of an earthquake on ground motion and is proportional to the landslide density. However, the occurrence of landslide is also related to multiple factors such as topography, earthquake, geology and human activities. For example, relatively hard rock layers can better resist deformation during earthquakes, thereby reducing the occurrence of landslides. Therefore, it is normal that the PGA and landslide density do not have a perfect positive correlation.

Earthquake intensity

The area of the VIII-degree intensity zone of the Ms 6.6 Jinggu earthquake is smaller than that of the VII-degree zone. However, the number and area densities of landslides in the VIII-degree zone are obviously larger than those in the VII-degree zone (Fig. 6a,b). The LND and LAP in the VII-degree zone are 0.13 km−2 and 0.03%, respectively, while in the VIII-degree zone, they rise to 0.55 km−2 and 0.13% (Fig. 6b).

Fig. 6
figure 6

The correlation between landslide distribution and the seismic intensity in the Jinggu earthquake (a, b) and the Ludian earthquake (c, d). LAP: landslide area percentage; LND: landslide number density; CA: Classification area. The panels (a, c) were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Regarding the Ludian earthquake, the distribution density of coseismic landslides is also positively correlated with the intensity magnitude (Fig. 6c,d). In the IX-degree zone, the LND and LAP are as high as 39.34 km−2 and 17.55%, respectively (Fig. 6d). The LND and LAP in the VIII-degree zone are 21.62 km−2 and 3.56% (Fig. 6d), while in the VII-degree zone, they are much lower, similar in magnitude to those of the Jinggu earthquake (Fig. 6).

Earthquake epicenter

Figure 7a,b show the correlation between landslide distribution and distance from the epicenter of the Jinggu earthquake. A peak in LND and LAP appears in the zone 4–6 km from the epicenter, with another notable increase at around 16 km (Fig. 7b). Overall, the number and area densities of landslides do not strictly decrease with distance from the epicenter, as several secondary peaks are observed.

Fig. 7
figure 7

The correlation between landslide distribution and the distance to epicenters of the Jinggu event (a, b) and the Ludian earthquake (c, d). LAP: landslide area percentage; LND: landslide number density; CA: Classification area. The panels (a) and (c) were generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Figure 7c,d show the correlation between landslide distribution and distance from the Ludian earthquake epicenter. The highest LND and LAP values occur 2–4 km and 4–6 km from the epicenter, rather than in the immediate proximity (0–2 km) (Fig. 7d). While LND generally decreases with distance from the epicenter, the LAP shows a secondary peak at 8–10 km, indicating larger landslides in this zone (Fig. 7d).

Discussion

The seismogenic characteristics of the Jinggu and Ludian events

The seismogenic fault and rupture process of the Jinggu earthquake

Earthquake focal mechanisms, aftershock distribution, and coseismic landslide patterns can provide valuable insights into the seismogenic structure and rupture process of an earthquake4,54,74. The focal mechanism solutions indicate that the primary rupture plane of the Jinggu mainshock is a right-lateral strike-slip fault with a NW-trending nodal plane55,75. Relocated aftershocks of this event align within a narrow NW-trending zone with a steep dip, approximately parallel to the extension of the Puwen fault54 (Fig. 8a). These confirm the Puwen fault as the seismogenic fault of the Jinggu earthquakes. Since earthquake energy is released primarily around the epicenter and along the seismogenic structure, the coseismic landslides, as significant surface deformations, typically occur along linear seismogenic faults or around the epicenter. Therefore, the distribution of coseismic landslides often coincides with the seismogenic structure and the epicenter9,21. Reasonably, the major axis direction of the coseismic landslide distribution of the Jinggu event is almost consistent with the extension of the Puwen fault (Figs. 3a and 8a). However, the landslides induced by the Jinggu earthquake are dominantly distributed on the northeast side of the the Puwen fault’s extension (Fig. 3a,b and 8a), which contrasts with the typical distribution expected from a strike-slip earthquake. Normally, landslides induced by a steep strike-slip earthquake would be distributed symmetrically on both sides of the fault4,22. This symmetry can be affected by the fault’s dip, geometric characteristics, and movement direction. Generally, more landslides are located on the hanging wall of a dip-slip fault4. The Puwen fault, however, is not a pure strike-slip fault. The earthquake focal mechanisms of the Jinggu mainshock (strike [333°], dip [68°], rake [− 170°]) and the seismogenic fault parameters (strike [332°], dip [89°], rake [− 167°]) derived from aftershock inversion indicate that the fault rupture contains some normal fault components53,76. The northeast wall is thus inferred to be the hanging wall of the fault, explaining the greater concentration of landslides on that side. Additionally, fault zones are often composed of multiple secondary faults77. Wu et al. suggest that the seismogenic fault zone of the Jinggu earthquake comprises small-scale, intermittently connected faults with suble surface linear features in the middle section of the seismic region53. Based on this, we propose that the seismogenic fault of the Jinggu earthquake is the northwest section of the Puwen fault. Its location is in alignment with the extension of the southeast section of the Puwen fault, combining aftershock distribution patterns53 (Fig. 8a).

Fig. 8
figure 8

The distribution patterns of coseismic landslides and aftershocks and rupture processes in 2014 Jinggu event. (a) Superposition of landslides and aftershocks with reference to faults location; (b) The individual area of each landslide along the Puwen fault. The black solid line is the fitting lines for the variation of landslide area with distance from northwest to southeast; the black dashed lines show the epicenter location. (c) The hypocenter distribution of the Jinggu earthquake sequence after relocation in NW–SE direction (the data are from Wang et al.54). Red arrows show the variation trends of aftershocks depth; (d) Schematic diagram of the rupture process of 2014 Jinggu earthquake. The thick red line shows the fault trace on the surface. The panel (a) was generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

In seismic events where surface ruptures are minimal, surface change or damage intensity can sever as proxies for evaluating the earthquake source and mechanism21. Earthquake-induced landslides, as a form of surface damage, can provide valuable insights into the source mechanism of the earthquake. The distribution pattern of coseismic landslides is not only correlated with the spatial location of the epicenter and seismogenic fault (Figs. 3 and 7) but also reflects the rupture process of the earthquake9,11,21,37 Xu4 inferred variations in rupture depth based on spatial changes in the distribution and scale of coseismic landslides. He suggested that the large-scale, densely distributed landslides might indicate surface ruptures, whereas small-scale, scattered landslides imply subsurface ruptures. By considering the location of the epicenter, the rupture propagation process can also be inferred. In the Jinggu epicenter, the Puwen fault in the affected area can be divided into northwest and southeast sections. Landslides are more concentrated in the southeast section compared to the northeast (Fig. 3c,d). However, from northwest to southeast along the seismogenic fault, the area of individual landslides decreases gradually in the linear fit (Fig. 8b). As an index of surface damage, the scale of these landslides reflects the destructive power of the earthquakes in specific locations. Accordingly, the distribution pattern shows fewer but larger landslides in the northwest section, and more frequent but smaller landslides in the southeast section. This indicates a shallower rupture with greater destructive power in northwest section, and a dominated and deeper rupture in the southeast section.

Wang et al. applied the double-difference method to relocate 2076 aftershocks between October 7 and December 31 and noted that the Jinggu earthquake sequences follow a NW–SE linear distribution54. The sequence can be temporally divided into two stages: first, aftershocks extended bilaterally from the epicenter along the NW–SE direction, and then the rupture propagated south-southeast. The depths of the whole sequence decreased toward the northwest and increased toward the southeast (Fig. 8c). Xie and Han76 obtained the rupture features of the Jinggu seismogenic fault by aftershock sequence inversion. Their results showed the fault ruptured along a 332°strike under the main compressive stress in the NNE direction, with a dip that did not breach the surface. The rupture depth ranged from 19.085 km at the southeast end to 2.746 km in the shallower northwest end. Additionally, swarm activity and aftershock migration indicate that the rupture of the Ms6.6 earthquake briefly stalled as it expanded southeastward due to a rock bridge obstructing the fault’s oblique connecting position. This temporary block triggered subsequent Ms5.8 and Ms5.9 earthquakes after breaking the barrier53. This behavior is consistent with the segmented rupture pattern of the earthquake source54.

Combining the distribution pattern of coseismic landslides analyzed in this paper with previous analyses based on aftershock sequences (Fig. 8c), we can conclude that the Jinggu earthquake occurred along a blind section of the Puwen fault in the northwest direction. The rupture propagated from the earthquake source toward deeper areas in southeast and shallower regions in northwest, before further extending southeastward at greater depths (Fig. 8d). This is consistent with the phenomenon that there only are tensor gravity fractures and no surface rupture zone78. Such rupture process reasonably interprets the presence of more but smaller landslides at the southeast section and fewer but larger landslides in the northwest section of the seismogenic fault (Fig. 3 and 8).

The rupture process of the Ludian earthquakes

The correlation between landslide distribution and the seismogenic fault of the Ludian earthquake reveals that the landslides were concentrated near the Xiaohe-Baogunao fault (Figs. 4a,b and 9a). Along the fault, the number and area densities of landslides in the southeastern side of the epicenter are significantly greater than those in the northwestern, with LAP peaking farther from the epicenter in the southeast section than LND (Fig. 4c,d). Moreover, LND generally increases with the PGA, while LAP increases initially and then decreases with increasing PGA (Fig. 5d). This suggests that areas with higher PGA values, generally closer to the epicenter, are associated with smaller landslides. Additionally, the trends of LND and LAP with increasing distance from the epicenter are inconsistent (Fig. 7d). These results indicate the larger-scale landslides formed in the southeast section, rather than near the epicenter or in high-PGA regions, as evidenced by the linear trend of individual landslides area from northwest to southeast along the seismogenic fault (Fig. 9a,b). It can be deduced that seismic rupture affects the surface more strongly in the southeast section, where the rupture plane is closer to the surface.

Fig. 9
figure 9

The distribution patterns of coseismic landslides and aftershocks and rupture processes in 2014 Ludian event. (a) Superposition of landslides and aftershocks with reference to epicenter location; (b) The individual area of each landslide along the A-A’ profile. The black solid line is the fitting lines for the variation of landslide area with distance from northwest to southeast; the black dashed lines show the locations of Ludian epicenters; (c) The hypocenter distribution of the Ludian earthquake sequence after relocation in NW–SE direction (the data are from Zhang et al.43). Red arrows show the variation trends of aftershocks depth; (d) Schematic diagram of the rupture processes of 2014 Ludian earthquake. The thick red line shows the fault trace on the surface. The panel (a) was generated using the GeoScene 2.1 (https://www.geoscene.cn/).

Full size image

Zhang et al. relocated the Ludian Ms 6.5 mainshock and 647 aftershocks using the double-difference algorithm43. Their results showed an asymmetric, conjugate-shaped aftershock sequence stretching 17 km in the east–west direction and 22 km in the northwest direction. Along the NW–SE profile (B–B′), aftershocks extended toward shallower area in the southeast and deeper area in the northwest along the Xiaohe-Baogunao fault, originating from the mainshock (Fig. 9a,c). Zhang et al. also suggested the rupture propagated toward the surface along the southeast direction, based on a revised rupture model from the double couple’s two nodal planes45. Xu47 analyzed the rupture process of the Ludian earthquake using the spatial distribution and scale variations of 1024 coseismic landslides4. His findings also indicate that most landslides occurred southeast of the epicenter, characterized by smaller scales and scattered distribution in the northwest and larger scales with more concentrated distribution in the southeast. This implies that the rupture plane was relatively deep in the northwest and shallower, possibly reaching the surface, in the southeast section of the seismogenic fault4,37. These conclusions align with field investigation results. According to focal location of the Ludian earthquake68 and the distribution of the primary surface rupture zone36,44,46,47,48, it is inferred that the rupture initiated at the focal depth and propagated simultaneously upward and southeastward along the Baogunao-Xiaohe fault47.

In conclusion, the distribution pattern of coseismic landslides induced by the Ludian earthquake, the rupture model derived from aftershock sequences, and field investigation results consistently indicate that the seismic rupture began at focal depth and spread both upward toward the surface in the southeast and downward into deeper areas in the northwest (Fig. 9d). Such rupture process explains the occurrence of more and larger landslides in the southeast section, and fewer but smaller landslides in the northwest section of the Xiaohe-Baogunao fault.

Contrasting impacts of the Jinggu and Ludian earthquakes

The Ms 6.5 Ludian earthquake triggered at least 10,559 coseismic landslides, covering a total area of 14.975 km2 in an elliptical region of about 360 km2,41. The event resulted in 617 fatalities, 112 missing persons, 3143 injures, and economic losses approaching 20 billion yuan59,79. In contrast, the Jinggu earthquake of similar magnitude induced significantly fewer hazardous geologic effects, with only 441 coseismic landslides occupying an area of 1.08 km2 in a region of 2160.86 km2, leading to 1 death, 325 injuries, and economic losses of about 5.1 billion yuan59,79. Based on a number of coseismic landslides data worldwide, Keefer10 and Rodriguez et al.17 established the relationship between earthquake magnitude (Mw) and coseismic landslide affected area (Fig. 10a). Our findings indicate that the distribution area of landslides from the Ludian earthquake is smaller than that of other events of comparable magnitude, while the Jinggu earthquake aligns with the eatablished envelope (Fig. 10a). However, according to Keefer’s relationship between earthquake magnitude and the number of coseismic landslides80, the Ludian earthquake resulted in a relatively higher number of landslides compared to similar magnitude events, with the Jinggu event fitting this correlation (Fig. 10b). In summary, the Ludian earthquake produced concentrated and numerous landslides, whereas the Jinggu earthquake resulted in fewer, more widely distributed landslides.

What account for these differences in seismic impacts? Previous studies suggest that the major damage differences between the two earthquakes, including coseismic landslides, arise from variations in houses structure, population density, topography and geomorphology, lithology, and vegetation50,57,58. For the Ludian earthquake, the significant topographic relief, widespread basalt and carbonate rocks, and strong river erosion, along with freezing and chemical weathering, contribute to the frequent occurrence of coseismic landslides50. These factors are typical in seismic zones10,17,80. In this paper, we mainly focus on the influence of the seismic factors on the amount of landslides or damage induced by earthquakes of similar magnitude. In both events, relationships between seismic factors and landslides exhibit some similarities. For example, landslides are generally aligned with the fracture orientations, areas with higher seismic intensity show larger landslide densities, and the highest landslide density occurs approximately 4 km from the epicenter rather than in the proximity area. Such distribution pattern of coseismic landslides is also evident in the 2022 Ms6.8 Luding earthquake27,28,29,30,31. However, the varying seismic rupture processes in two earthquakes lead to different influence of these factors on landslide development. Xu et al. focused on the characteristics of the ground motions for the two events, and highlighted that faster anelastic attenuation probably explain more severe damage caused by the Ludian earthquake59, a notion supported by the distribution characteristics of landslides triggered by the Ludian earthquake37. In the Ludian earthquake, while the landslide distribution area is small, there are many large landslides, likely linked to its shallow focal depth and rapid seismic energy attenuation37. The Jinggu earthquake also has a shallow focal depth, indicating that focal depth alone is not the main factor affecting the distribution of landslides. Furthermore, the distribution of aftershocks shows that the rupture propagation angle of the Ludian earthquake is steeper (Fig. 9c), whereas the Jinggu earthquake exhibits a gentler angle (Fig. 8c). This difference may explain the concentrated distribution of landslides in the Ludian earthquake versus the scattered distribution in the Jinggu earthquake.

Based on these analyses, the difference in the rupture processes of the two events should be considered key factors influencing their geologic effects and the degree of damage. In the Jinggu event, a minor portion of the rupture propagated from the source to shallow depths in the northwest but did not reach the surface, resulting in a wider area of landslide influence with relatively larger individual landslides (Fig. 8). In the southeast, the rupture propagated deeper (Fig. 8). Consequently, while more landslides occurred in the southeast due to this dominant rupture, their individual scales are smaller due to deeper rupture depth. In contrast, the Ludian event’s rupture extended from the source to the surface in a southeastward direction with a relatively steep slope, while extending deeper to the northwest (Fig. 9). The landslides induced by the Ludian earthquake were scattered in the northwest with smaller scales but concentrated in the southeast with larger scales. Typically, landslides triggered by blind fault earthquakes exhibit wider distribution with smaller scales compared to the exposed fault earthquakes4,81,82. Both seismic events demonstrate bidirectional rupture propagation mainly toward the southeast; however, the distinction lies in that the Jinggu earthquake rupture propagates deeper, whereas the Ludian earthquake rupture reaches the surface during its southeastward propagation. The Luding earthquake case also proves that the closer the earthquake rupture is to the surface, the more landslides occur32. Such oblique upward extension of the rupture to reach the surface is likely a significant factor contributing to the severe impact of the Ludian earthquake.

Fig. 10
figure 10

Correlations between earthquake magnitude (Mw) and coseismic landslide distribution (affected) area (a) and the landslides number (b). The cases shown by black circles reference from Keefer10,80, Rodriguez et al. 17, Xu and Xu83 and He et al.26.

Full size image

Conclusions

According to the analysis on the distribution patterns of landslides triggered by the 2014 Jinggu and Ludian earthquakes in term of seismic factors, we conclude that the differences of two earthquakes.

  1. (1)

    The long axis of the coseismic landslide distribution zone in the Jinggu event aligns closely with the strike of the Puwen fault, while the distribution associated with the Ludian earthquake corresponds to the Xiaohe-Baogunao fault. Along the seismogenic faults, the landslides from the Jinggu event are characterized by fewer and larger instances in the northeast section and more numerous and smaller instances in the southeast section of the fault. In contrast, the Ludian event exhibits fewer and smaller landslides in the northeast section, with and a greater number of larger in the southeast section of the fault.

  2. (2)

    Based on the distribution patterns of previous aftershocks and field investigations, the pattern of the coseismic landslides suggests that the rupture from the Jinggu earthquake initially propagated upward into the northwest shallow layer but did not reach the surface, while it extended downward into the deep layer in the southeast direction. In contrast, the Ludian seismic rupture originated from a focal depth and spread upward to the shallow layer, breaching the surface in the southeast direction, while also extending downward into the deep layer in the northwest direction. The innovative study of the fault rupture process using the spatial distribution pattern of coseismic landslides may provide new useful information and exploration paths for similar earthquakes with complex source structures.

  3. (3)

    The steeper rupture propagation angle of the Ludian earthquake compared to the Jinggu earthquake may lead to more concentrated landslides in the Ludian earthquake. Additionally, the Jinggu earthquake rupture extending deeper while the Ludian’s reaching the surface during southeastward propagation contributes to significantly less geologic hazards and damage in Jinggu earthquake than Ludian’s. In the future, we will expand our research to encompass multiple factors influencing seismic damage, transcending the limitations of solely seismic factors.