1 Introduction

The Ms 8.0 Wenchuan earthquake of May 12, 2008, which occurred in the Longmen Shan thrust fault zone, was the greatest intra-continental earthquake in China (Fu et al. 2008, 2011; Zhang et al. 2008; Xu et al. 2009a) (Fig. 1). The Wenchuan earthquake induced widespread landslides along valleys, as well as producing surface ruptures (Qi et al. 2010; Fu et al. 2011; Dai et al. 2011; Gorum et al. 2011). The Wenchuan Earthquake Fault Scientific Drilling Project (WFSD), initiated following the earthquake, reported a zone of fault gouge with a maximum thickness of 3.79 m along the Yingxiu–Beichuan fault zone. This thick section of fault gouge indicates long-term repeated seismic activity in the Longmen Shan thrust belt (Li et al. 2013). Earthquakes and associated landslides play a significant role in mountain building and landscape evolution in tectonically active mountain belts (Keefer 1994; Hovius et al. 1997, 2000, 2011; Dadson et al. 2004; Malamud et al. 2004; Avouac 2008; Ouimet 2011). In the Longmen Shan, previous studies focused on the volume balance between co-seismic rock uplift and eroded material generated by co-seismic landslides to determine whether large dip- and oblique-slip earthquakes have contributed to the high mountains (Parker et al. 2011; Li et al. 2014a, b; Ren et al. 2014a). However, few studies have focused on the combined role of repeated large earthquakes and fluvial erosion shaping the present topography in the Longmen Shan.

Fig. 1
figure 1

a The Longmen Shan is located at the boundary between the Sichuan Basin and the eastern margin of the Tibetan Plateau. b Topography and active faults (F 1–F 3) in the Longmen Shan. F 1, Wenchuan–Maoxian fault; F 2, Yingxiu–Beichuan fault; F 3, Guanxian–Anxian fault; Red lines indicate the surface rupture associated with the Wenchuan earthquake. c A cross-sectional profile of the topography along AA′ (dashed line) is shown

Full size image

In this article, we focus on the Minjiang River watershed, which is characterized by high relief and experienced extensive earthquake-induced landslides in 2008. Based on interpretations of the post-earthquake high-resolution remote sensing images and field surveys, an inventory of earthquake-induced landslides on both sides of the Minjiang River and its tributaries has been developed. We quantify the erosion rate due to earthquake-induced landslides and compare this with erosion rates on Myr-timescale to evaluate the role of the earthquake-induced landslides in surface processes. Furthermore, we discuss the implications of landslide erosion related to repeated larger earthquakes and fluvial erosion on the landscape evolution of the Longmen Shan.

2 Study area

The Longmen Shan resides in a region of steep topography in southeastern China and is the boundary between the Tibetan Plateau and Sichuan Basin (Fig. 1a). The Longmen Shan thrust fault zone is approximately 500 km long and 30–50 km wide and consists of three main sub-parallel faults (Deng et al. 1994; Burchfiel et al. 1995) that experienced thrust and dextral deformation along the major fault systems (Burchfiel et al. 1995; 2008; Kirby et al. 2002). From northwest to southeast, the three main faults are the Wenchuan–Maoxian fault (F 1), Yingxiu–Beichuan fault (F 2), the primary structure responsible for the 2008 Wenchuan earthquake (Zhang et al. 2008; Xu et al. 2009a), and Guanxian–Anxian fault (F 3) (Fig. 1b). The Wenchuan earthquake produced a surface rupture over 300 km long (Fu et al. 2008; Zhang et al. 2008; Xu et al. 2009a). The surface deformation was characterized by oblique thrust/dextral slip with a maximum vertical displacement of 9–10 m (Fu et al. 2008, 2011; Li et al. 2008; Ran et al. 2010a; Xu et al. 2009a; Zhang et al. 2010a). The Wenchuan earthquake induced massive landslides that are linearly distributed along the surface rupture zones and river valleys (Huang and Li 2009; Qi et al. 2010; Dai et al. 2011; Fu et al. 2011). In total, the earthquake triggered more than 56,000 landslides (Dai et al. 2011; Gorum et al. 2011; Xu et al. 2014). The main types of landslides were shallow landslides, rock falls, deep-seated landslides, and rock avalanches (Qi et al. 2010; Dai et al. 2011) that ranged in size from a few cubic meters to tens of millions of cubic meters (Dai et al. 2011). The estimated volume of the vast Daguangbao landslide located in Anxian county, the largest of the co-seismic landslides, is at least 7.4 km3 (Huang et al. 2008; Yin et al. 2011; Chen et al. 2014).

The surface in the Longmen Shan has been deeply dissected by several major rivers (e.g., the Minjiang River, Tuojiang River and Fujiang River) (Fig. 1b). The landforms are characterized by steep hillslopes and deep-incised valleys (Fig. 1c). Originating in the eastern margin of the Tibetan Plateau, the Minjiang River flows along the Wenchuan–Maoxian fault (F 1), passes through the Longmen Shan Range cutting gorges more than 1–3 km deep (Burchfiel et al. 1995), and drains into the Sichuan Basin at Dujiangyan. The study area presented here is confined to the Minjiang drainage basin within the Longmen Shan Range (Fig. 2a). The Baisha River is a tributary of the Minjiang River (Fig. 2a). The Wenchuan earthquake produced a surface rupture over 14 km long along this river (Fig. 2b) (He et al. 2008a). A co-seismic knickpoint with a height of about 4.5 m was formed in the Baisha River at Bajiaomiao village near the town of Hongkou (Liu-Zeng et al. 2010).

Fig. 2
figure 2

a Shaded relief map shows the morphology of the Minjiang drainage basin. F 1, Wenchuan–Maoxian fault; F 2, Yingxiu–Beichuan fault; F 3: Guanxian–Anxian fault. The landslides have a zonal distribution along the main channel of the Minjiang River and its tributaries. b A 14-km long surface rupture was produced by the Wenchuan earthquake along the Baisha River (He et al. 2008a)

Full size image

The eastern margin of the Tibetan Plateau is influenced by a wet tropical climate. Monsoon rainfall dominates with annual precipitation as high as 1,000 mm year−1 in the Longmen Shan (Godard et al. 2010). Most of the precipitation is concentrated during the heavy rains of the summer monsoon from June through September, and this constitutes 70–80 % of the total annual rainfall. Thus, water discharge shows a strong monsoonal influence (Liu-Zeng et al. 2011) and sediment transport away from the Longmen Shan, with 80–90 % of the annual suspended sediment transport occurring from June to September (Liu-Zeng et al. 2011).

3 Data and methodology

Owing to severe weather conditions, cloud- and haze-free images of the study region were difficult to obtain. Consequently, multi-source and multi-temporal remotely sensed images were collected after the Wenchuan earthquake with the aim of identifying earthquake-induced landslides. The landslide inventory was based on interpretations of aerial photographs (with 1-m spatial resolution) obtained on May 16–24, 2008, SPOT images (with 10-m spatial resolution) obtained on May 23, 2008, and Aster images (with 15-m spatial resolution) obtained on May 14–15, 2008. All images were processed and analyzed using image analysis and GIS software, such as ENVI 4.8 and ArcGIS 9.3. The images were geo-referenced using ground control points (GCPS) selected from 1:50,000 scale digital topographic maps. The root mean square error (RMSE) of the accuracy of the geo-referencing was set to not exceed the size of a pixel. All images were projected on a Universal Transverse Mercator (UTM) zone 48 North projection using the WGS84 datum.

3.1 Landslide detection

Landslides were identified on the basis of one or more diagnostic features including morphology, vegetation, and tones of the images (Yang and Chen 2010; Dai et al. 2011). Morphological characteristics included slope angle, visible landslide debris movement paths, and run-out lobe forms. The landslides were also recognized from newly denuded vegetation on the slope, linear scars in vegetation along run-out paths, and disturbed vegetation on body contrast with surroundings on the images. Color characteristics of individual landslides were enhanced using the presence of bright white or dark brown contrast as compared to the surroundings. Landslide perimeters were captured through heads-up digitizing in ArcGIS and included both deposit and source area. For a larger landslide, the source area and deposit should be separated. Based on the above criteria, all landslides visible on the images were delineated as single polygons in ArcGIS and the total number of landslides and the areas of individual landslide were calculated.

3.2 Earthquake-induced landslide erosion

Quantifying the erosion rate is a useful way of identifying the contribution of earthquake-induced landslides to surface erosion. In the Longmen Shan, long-term seismic activity occurs along the Yingxiu–Beichuan fault zone (Li et al. 2013). In addition, the recurrence interval of large earthquakes similar to the Wenchuan earthquake is in the range 2,300–3,300 year (Ran et al. 2010b, 2013; Chen et al. 2013). We assumed that a large earthquake similar to the Wenchuan earthquake is the characteristic earthquake within the Longmen Shan region.

For a characteristic earthquake, the earthquake-induced landslide erosion rate can be measured by the following equation (Malamud et al. 2004),

$$E_{\text{L}} = V_{\text{TL}} /A_{\text{L}} \times T_{\text{L}}$$
(1)

where E L is the earthquake-induced landslide erosion rate, V TL is the total landslide volume, A L refers to the area over which the landslides are concentrated, and T L is the time interval over which the landslides accumulated. For the earthquake-induced landslide event inventory, T L can also be defined as the recurrence interval of an earthquake (Malamud et al. 2004). The total volume (V TL) of earthquake-induced landslides is defined as follows,

$$V_{\text{TL}} = \mathop \sum \limits_{i = 1}^{N} V_{\text{LN}}$$
(2)

where N is the total number of landslides and V LN is volume of an individual landslide. Each individual volume can be estimated from the empirical relationship between landslide volume and area (Simonett 1967; Rice et al. 1969; Hovius et al. 1997; Korup 2005; Imaizumi et al. 2008; Guzzetti et al. 2009; Parker et al. 2011; Li et al. 2014b).

Parker et al. (2011) and Li et al. (2014b) provided the formulas to estimate the landslide volume associated with the Wenchuan earthquake. However, due to the various kinds of co-seismic landslides present (Qi et al. 2010; Dai et al. 2011) and the ruggedness and inaccessibility of the terrain, it was difficult to obtain enough representative field data to establish the empirical relationship between the co-seismic landslide volume and area within the Longmen Shan. In contrast, Guzzetti et al. (2009) proposed a robust global scaling law, in which the relationship between landslide volume (V) and area (A) is largely independent of the physiographical setting. The formula of Guzzetti et al. (2009) to estimate each individual landslide volume is as follows:

$$V = 0.074 \times A^{1.450}$$
(3)

To test whether Guzzetti et al.’s (2009) equation could be applied to the landslides associated with the Wenchuan earthquake, four co-seismic landslides reported by Xu et al. (2009b) were selected to validate the equation (Table 1). Figure 3 shows that the trend of Xu et al.’s (2009a, b) data is consistent with the power-law relationship of Guzzetti et al.’s (2009). In addition, the main Quaternary (Q) strata in the Longmen Shan are near horizontal or with a gentle slope and are therefore not generally prone to slide. Massive landslides mainly occurred in pre-Tertiary (T) strata, accounting for more than 92 % of the total landslides (Qi et al. 2010). The above-mentioned correlation between landslides and lithology shows that most landslides associated with the Wenchuan earthquake can be thought to be rock landslides. For bedrock landslides, Larsen et al. (2010) suggested that landslide erosion was controlled by hillslope material, and the relationship of individual landslide volume (V) and area (A) takes the form of a power law: V = αA γ (α, γ are constant) with a scaling exponent γ = 1.3–1.6. The exponent 1.45 of Guzzetti et al.’s (2009) falls in the range 1.3–1.6. Thus, the Eq. (3) can be used to estimate individual landslide volume in the Longmen Shan.

Table 1 The area and volume of landslides associated with the Wenchuan earthquake (data from Xu et al. 2009b)
Full size table
Fig. 3
figure 3

Area–volume relationship (Guzzetti et al. 2009) with Wenchuan earthquake-induced landslides superimposed

Full size image

Based on the method of Malamud et al. (2004), the area of the landslides concentration (A L) is a rough estimate according to its spatial distribution in which a high percentage of landslides are concentrated. Note that the area is not necessarily as large as the area over which the landslide inventory was originally compiled. Furthermore, we defined the area of the landslide concentration (A L) and assumed that erosion in this area is dominated by these landslides. The area A L has a zonal distribution in the Minjiang River watershed. According to the spatial distribution (Huang and Li 2009; Qi et al. 2010; Dai et al. 2011), the area of the landslides concentrated (A L) was calculated based on the extent of the Minjiang River and its tributaries within which an overwhelming majority of landslides were located.

3.3 Determining post-seismic river erosion

To elucidate the river erosion after the Wechuan earthquake, we used the Baisha River as a case study. The river erosion after the Wenchuan earthquake was estimated based on a comparison between 2008 and 2012 digital elevation models (DEM) compiled from the field measurement. Surveys were carried out using a Topokon total station (TS) and real-time-kinematic global positioning system (RTK-GPS) in 2008 and 2012, respectively. To minimize error and uncertainty, areas that were relatively undisturbed by human activity were selected for analysis. Kriging was used to interpolate via ArcGIS to obtain the final DEM with a spatial resolution of 0.5 m. Finally, all DEMs were geo-referenced to the coordinate system WGS84 UTM.

4 Results

4.1 Earthquake-induced landslide erosion

Based on aerial photographs and satellite images, 7,567 landslides were identified along the main channel of the Minjiang River within 8,000 m of the drainage line (Fig. 2a). Based on this distance, the area of landslide concentration (A L) was measured as 4,520 km2 with the aid of ArcGIS. Combining Eqs. (2) and (3), the total volume (V TL) of the 7,567 landslides was estimated to be about 5.7–5.9 km3.

Historical records indicate no earthquakes larger than magnitude 7 have been reported in the Longmen Shan region (Wen et al. 2009). Pre-earthquake trenches along the Yingxiu–Beichuan fault zone show an earthquake recurrence interval of at least 2,000–3,000 years (Li et al. 2006). Post-earthquake trench excavation indicates that the recurrence interval of large earthquakes similar to the Wenchuan earthquake is in the range 2,300–3,300 year (Ran et al. 2010b, 2013; Chen et al. 2013). From these studies, it may be concluded that large earthquakes have a recurrence interval of 2,000–3,000 years. For the characteristic earthquake, the repeated time of co-seismic landslides is the recurrence interval of the earthquake (Malamud et al. 2004). Thus, the time interval over which co-seismic landslides accumulated is 2,000–3,000 years. From Eq. (1), the average erosion rate (E L) due to earthquake-induced landslides is estimated to be 0.4–0.6 mm year−1. Our result falls in the range 0.1–0.6 mm year−1 reported by Parker et al. (2011).

4.2 Post-seismic river erosion

A comparison between the 2008 and 2012 DEMs from the Baisha River (Fig. 4) suggests that the river has removed at least volume of 1.9 × 104 m3 of material during the 4 years after the 2008 Wenchuan earthquake. The surface ruptures of the Wenchuan earthquake caused the formation of the fluvial knickpoints (Fu et al. 2009; He et al. 2008a; Liu-Zeng et al. 2009; Xu et al. 2008). The knickpoints were rapidly removed during the post-earthquake period. For example, at Pingtong town, Pingwu county, the surface rupture cut through the river bed and formed a knickpoint with a height of about 2.3 m on May 12, 2008 (Xu et al. 2008), but this had been removed by river erosion within 1 month (He et al. 2008b). Another knickpoint, with a height of nearly 4.5 m, was produced in the Baisha River by coseismic deformation (Liu-Zeng et al. 2010) (Fig. 5a). However, knickpoint was not present in 2012 (Fig. 5b).

Fig. 4
figure 4

Post-seismic fluvial erosion of the Baisha River in the 4 years following the Wenchuan earthquake. a DEM of the Bajiaomiao reach. b DEM produced from Total Station data collected in 2008. c DEM produced from Real-Time-Kinematic Global Positioning System data collected in 2012

Full size image
Fig. 5
figure 5

Field photographs showing evidence of rapid fluvial erosion. a Knickpoint with a height of 4.5 m produced by co-seismic deformation in Hongkou, Dujiangyan City (Liu-Zeng et al. 2010). b The co-seismic knickpoint dissipated in 2012

Full size image

5 Discussion

5.1 Assessing the uncertainties associated with estimating landslide volume

The present study of the 7,547 co-seismic landslides produced about 5.7–5.9 km3 of hillslope material on both sides of the Minjiang River and its tributaries over an area of 4,500 km2. In contrast, Parker et al. (2011) suggest that more than 50,000 landslides produced 5–15 km3 of erodible material within the Longmen Shan region. The estimated landslide volume range of 5–15 km3 was revised to 2.1–3.7 km3, which is considered to be consistent with the range of 1.5–3.6 km3 reported by Ren et al. (2014b). Compared with these results, we may have overestimated the landslide volume because we only determined the landslide volume of part of the area within the Longmen Shan.

Previous studies used various methods to reduce the uncertainties associated with the calculated landslide volumes related to the Wenchuan earthquake (Parker et al. 2011; Li et al. 2014b; Ren et al. 2014b). These studies used the robust scaling relationship between landslide volume and area (Parker et al. 2011; Li et al. 2014b), or topographic changes between pre- and post-earthquake DEMs (Ren et al. 2014b) to estimate the landslide volume. In contrast, we gave a rough estimate by an empirical formula, lacking a region calibration. However, an important problem must be addressed: How to quantify the effect of a gigantic landslide on the estimated total volume of all landslides? For example, the Daguangbao landslide is the largest co-seismic landslide within the Longmen Shan region (Huang et al. 2008; Yin et al. 2011). The volume of the Daguangbao landslide is 7.4–11.99 km3 based on the differential DEM method (Huang et al. 2008; Yin et al. 2011). However, the volume of all co-seismic landslides is 2.1–3.7 km3 (Li et al. 2014b). This discrepancy suggests that the estimated landslide volume may contain large uncertainties no matter which empirical relationship between landslide volume and area is used within the Longmen Shan. Nonetheless, the landslide erosion rate estimated here is of the same order of magnitude as that of Parker et al. (2011).

5.2 Earthquake-induced landslides and long-term surface erosion

Landslide erosion rates are needed for comparison with erosion rates derived from low-temperature thermochronology over a Myr-timescale to assess the importance of landslides in the landscape evolution of the Longmen Shan. Low-temperature thermochronology is an important tool for estimating rates of erosion in mountain landscapes (Safran 2003). Bedrock low-temperature thermochronology may provide erosion rates with a thermal model or by collecting multiple samples in a vertical transect via point measurements. Long-term erosion rates in the Longmen Shan obtained from low-temperature thermochronology are averaged over the past 10–15 Ma (Ouimet 2010). Low-temperature thermochronological data in the Minjiang River watershed are sparse and yield location-specific results. The average Myr-timescale erosion rates are 0.5–0.8 mm year−1 in the Longmen Shan (e.g., Godard et al. 2009b). We assumed that this limited low-temperature thermochronological data can reflect the areal average erosion rate. The magnitude of the erosion rate due to earthquake-induced landslides (0.4–0.6 mm year−1) is similar to the Myr-timescale erosion rates derived from low-temperature thermochronology.

Earthquakes are capable of triggering a large number of deep-seated bedrock landslides that are considered to be significant agents of rapid surface erosion (Hovius and Stark 2006). For example, landslide erosion rates increase significantly with increasing exhumation rate over the Myr-timescale in the eastern Himalaya syntaxis (Larsen and Montgomery 2012). In the Longmen Shan, the long-term erosion rates obtained from low-temperature thermochronology were averaged over the last 10 Ma (Kirby et al. 2002; Godard et al. 2009b). Therefore, during the Cenozoic topographic evolution of the Longmen Shan, the long-term (Myr-scale) erosion integrates many repeated earthquake events, which is confirmed by the result of the WSDF (Li et al. 2013). Large seismic events accelerate surface erosion by fracturing rock and mass wasting (Molnar et al. 2007). Fracturing can reduce rock strength, which leads to pervasive weathering of fractured rock along fracture planes and facilitates the extent of erosion. Mass wasting strips rock or soil from hillslope, thereby exposing fresh, unaltered rock to surface and making it easier to be eroded (Fig. 6). A recent study has shown that bare rock surfaces were eroded by 10–30 cm within 1 year of the Wenchuan earthquake (Wang et al. 2011). Thus, the massive landslides associated with repeated large earthquake and aftershocks may contribute to the long-term surface erosion by enhancing erosion efficiency.

Fig. 6
figure 6

Mass wasting exposes soil or fresh, unaltered rock to erosion. Photograph (a) and (b) are from Fu et al. (2009) and Yang and Chen (2010), respectively

Full size image

5.3 Implications for the long-term landscape evolution of the Longmen Shan

The rapid removal of earthquake-generated knickpoints may indicate a significant fluvial erosion capacity during the post-earthquake period. Previous studies have shown that long-term storage of co-seismic sediment along the major Longmen Shan River valley is unlikely (Kirby et al. 2003; Ouimet et al. 2009), as it can be evacuated efficiently over the entire earthquake cycle (Parker et al. 2011). Similarly, Liu et al. (2013) reported that most of the landslide-eroded sediment due to the Wenchuan earthquake is likely to be efficiently evacuated out of the Longmen Shan in the Minjiang River watershed over a seismic cycle. Seasonal floods could explain this efficient sediment delivery. In the field, we observed that heavy rainfall increases runoff and further sediment flux during the summer. In addition, a large volume of coarse-grained sediment deposited on the Chengdu Plain during the late Cenozoic has been interpreted to be closely associated with the region climate cycles and repeated large earthquakes similar to the Wenchuan earthquake (Wang and Meng 2009), indicating that the fluvial systems of the Longmen Shan have a high transport capacity.

The Wenchuan earthquake induced large-scale landslides along surface ruptures and river valleys (Huang and Li 2009; Fu et al. 2011). This spatial distribution mirrors the erosion, which is intense along rupture zones and river valleys. Similarly, the Myr-scale and modern denudation pattern show narrow zones of higher erosion rates along the active thrust fault (Godard et al. 2009a; Liu-Zeng et al. 2011). This relationship suggests that the earthquake-induced landslides may play a role in enhancing the differential erosion during the long-term landscape evolution. With the occurrence of repeated large earthquakes, long-term surface erosion produced by earthquake-triggered landslides, together with denudation related to other surface processes, may lead to unloading of the underlying lithosphere and consequent isostatic compensation, resulting in the uplift of summits of the Longmen Shan. This isostatic uplift is expected unless the erodible material is transported out of the Longmen Shan by rivers. In fact, the fluvial system has a significant capacity of transporting the erodible material out of the Longmen Shan. Thus, isostasy cannot be ignored in maintaining the high terrain of the Longmen Shan (Molnar 2012).

Late Cenozoic crustal shortening is a primary driver of uplift and the topographic development of the Longmen Shan (Hubbard and Shaw 2009; Tian et al. 2013). In this tectonic context, the fluvial system deeply incised the Longmen Shan, which produced the high relief. Without sustained endogenetic forces, the topography will be beveled by exogenic geological processes. The climate of the Longmen Shan and adjacent regions is characteristically warm and wet, with more precipitation in summer due to the influence of yearly monsoon. Several major rivers cross or originate from the Longmen Shan (Fig. 1). However, these rivers, and other surface erosion processes, do not degrade the topography. This leads to the question: How is the high relief of the Longmen Shan sustained? The long-term fluvial unloading related to the transport of eroded material due to earthquake-triggered landslides and other surface processes may induce isostatic rebound between the lower crust and upper mantle and result in deep mantle convection beneath the Longmen Shan (Fu et al. 2011; Liu et al. 2013). Consequently, deep tectonic processes may maintain the present-day steep topography of the Longmen Shan. Thus, rapid erosion associated with repeated large earthquakes and fluvial incision during the interseismic interval may be the important driver that is shaping the landscape of the Longmen Shan (Fig. 7).

Fig. 7
figure 7

Schematic model of landscape evolution in the Longmen Shan. a Mountain building with gentle slope produced by tectonic uplift. b Large earthquakes strike the Longmen Shan, exposing fresh rock and furthering the extent of surface erosion. Heavy rainfall results in rapid surface erosion. The material released by the mass wasting following large earthquake is transported out of the Longmen Shan by rivers, and this triggers isostatic rebound at the boundary between the lower crust and the upper mantle (black arrow) (Fu et al. 2011; Zhang et al. 2009, 2010b). c Erosional unloading results in preferential rock uplift focused along the major rivers. Tectonic uplift enhances river incision, which in turn creates the deep-incised valleys and high topographic relief

Full size image

6 Conclusions

Based on interpretations of high-resolution remote sensing images, 7,567 landslides were identified that were related to the Wenchuan earthquake within 8,000 m of the Minjiang River and its tributaries. The volume of erodible material was estimated to be 5.7–5.9 km3. Assuming that estimated 2,000–3,000-year time interval over which the landslides accumulated is reasonable, the landslides on both sides of the Minjiang River produced an erosion rate of 0.4–0.6 mm year−1. The earthquakes and associated erosion events may influence long-term (Myr-timescale) surface erosion by enhancing erosional efficiency. Furthermore, the rapid erosion associated with repeated large earthquakes similar to the Wenchuan earthquake, and river erosion during the interseismic interval, may contribute to the present-day landscape of the Longmen Shan area.