Keywords

1 Introduction

The Haiyuan earthquake of magnitude 8.5 on 12 December 1920 was recorded as one of the largest earthquakes in China. Over 23,000 people lost their lives in the earthquake, in which about 100,000 people died from the loess landslides triggered by the quake (Zhang and Wang 1995; Derbyshire et al. 2000). Weng et al. conducted a field investigation in the earthquake affected area and provided a general description of landslides and barrier lakes in 1921 (Yuan 2005). Close and McCormick (1922) took a large number of photos for the landslides triggered by Haiyuan Earthquake and gave corresponding description. However, it took over 60 years for the deep research about Haiyuan earthquake to arise. Bai and Zhang (1990) studied the sliding displacement on gently inclined loess slope triggered by Haiyuan Earthquake in Shibeiyuan of Guyuan County and held that landslides resulted from loess liquefaction. Zhang and Wang (1995) pointed out that the landslides are located in three concentration zones, with the area of 4,000 km2 and summarized that they are characterized by low moisture content, gentle slope gradient and fast sliding speed, etc. (Fig. 118.1). Zhang (1999) classified the co-seismic landslides into three categories according to the failure mechanism, seismic subsidence, loess liquefaction and shear deformation. Wang and Zhang (1999) indicated that the main mechanisms of the long run-out co-seismic loess landslides were the loess mass disintegration, oblique projection-collision between loess blocks and pulverization under the joint action of vertical and horizontal seismic acceleration. Wang et al. (2000) adopted soil dynamical methods to explore the loess liquefaction mechanism. In 2003, the Geological Environment Monitoring Station of Ningxia of Geological and Mineral Resources carried out investigation of landslides in Xiji County and a landslide database was generated (Yuan 2005). With the landslide database, Yuan (2005, 2006) conducted a detailed analysis of landslides in Xiji County and proposed that a large number of landslides were distributed in this area due to the control of many factors such as its special geomorphology, geological structure, stress dependence, soil characteristics and pre-earthquake macro water anomaly, etc. Zhang and Wang (2007) undertook numerous ring shear tests and found that the dramatic decrease of shear strength of loess in the earthquake was due to the sudden increase of pore water pressure.

Fig. 118.1
figure 1

Location of the study area (adapted from Lin and Xu 2008)

Full size image

Although many research has been made on the landslides triggered by Haiyuan Earthquake in recent years, they are mainly aimed at several representative landslides or landslides in typical areas (such as Xiji County), with little attention to the integrated study of landslides in the whole earthquake region. Based on high-resolution remote-sensing images provided by Google Earth, we conducted detained interpretation of landslides in the intensity IX zone of Haiyuan Earthquake, established the landslide database and analyzed their distribution patterns.

2 Landslide Interpretation

Landslides in the loess area with sparse vegetation and bare surface can be easily identified through the keys as follows.

  1. (1)

    Armchair- shaped rear wall of landslide

The rear wall of landslide is the most direct interpretation key. As it has been over 90 years since the occurrence of Haiyuan Earthquake, the rear walls of landslides induced by it suffered long-term water and soil erosion as well as artificial modification, but due to their steepness they still showed notable armchair-shaped features, which presented themselves in arc shape and dark tone in the images that were easily identified in Google Earth 3D view.

  1. (2)

    Texture

It is an important key to identify landslides for the reason that the loess slope usually displays continuous strip texture parallel to contour line but dislocated or discontinued when the landslide position appears.

  1. (3)

    Barrier lake

A great many large landslides induced by Haiyuan Earthquake blocked rivers to form a large number of barrier lakes, 43 of which are reserved and concentrated in Xiji County. These barrier lakes can be adopted as the auxiliary interpretation key of landslides, because they show dark tone in the images, which are easily identified.

The above-mentioned interpretation keys are utilized for the interpretation of 805 landslides in the study area in the preliminary stage. Following that field investigation of 473 landslides was conducted from July to August in 2012 to check the interpretation results. On the basis of field investigation, we have carried out the interpretation again and ascertained that the total number of landslides is 1,000, as shown in Fig. 118.2. The areas of all landslides added up to 1.026 × 108 m2, in which the slip source area occupied 4.52 × 107 m2 and the accumulation area occupied 5.74 × 107 m2. The area of the minimum landslide was 755 m2, whereas the maximum landslide was 2.3 × 106 m2,and the average area was 1.025 × 105 m2.

Fig. 118.2
figure 2

Landslide distribution of the study area

Full size image

3 Landslide Distribution Patterns

3.1 Relationship Between Epicentral Distance and Distribution of Landslides

This paper deemed that the epicenter is located in Ganyanchi of Haiyuan County (36º39″N,105º17″E), reported by Lanzhou Institute of Seismology. The longest distance from the landslide triggered by earthquake to the epicenter is approximately 140 km. For most landslides induced by earthquake, the shorter the distance to the epicenter is, the more the landslides are. This is not true of landslides triggered by Haiyuan Earthquake, for most of them (67 %) are distributed in the range of 80–100 km from the epicenter, indicating that their distribution are not controlled by the epicentral distance.

3.2 Relationship Between Seismic Intensity and Distribution of Landslides

The seismic intensity in the epicenter of Haiyuan Earthquake reached to XII degree and the interpreted landslides in this paper are distributed in the range of intensity IX to XII. From the statistical analysis, it is discovered that the concentration of landslides decreases with the intensity decreasing. 47.4 % of landslides are distributed in the intensity IX zone; 35.0 % of landslides are distributed in the intensity X zone; 10.3 % of landslides are distributed in the intensity XI zone; the proportion of landslides in the intensity XII zone is the smallest, occupying only 7.3 %. Clearly, there are more landslides in the zones of intensity of IX and X than in the zones of intensity of XI and XII. Furthermore, the landslides in the zones of intensity of XI and XII are relatively small on the whole, which may be attributed to that XI and XII zones belong to Liupanshan ridge with thin loess or bedrock outcrop (Zhang and Wang 1995).

3.3 Relationship Between Distance to the Seismogenic Fault and Distribution of Landslides

The relationship between landslides and seismogenic fault has been studied through calculating the distribution of landslides at an interval of 5 km. The landslides are concentrated in two areas, range of 0–5 km (22.0 %) and range of 40–70 km (66.8 %), matching the concentration zone in Haiyan County and in Xiji County. Between the two concentration areas is Liupanshan mountain, so it may be concluded that the landslides triggered by earthquake are mainly affected by the lithology and geomorphology.

3.4 Relationship Between Elevation and Distribution of Landslides

The elevation in the study area varies from 1,245 to 2,992 m, whereas the slip source area of landslides is distributed in the height range from 1,407 to 2,423 m, especially from 1,800 to 2,200 m (90.3 %).

3.5 Relationship Between Slope Height and Distribution of Landslides

The slope height of study area varies from 0 to 496 m, while the slope height of slip source area ranges from 0 to 224 m, especially from 15 to 100 m (74.0 %).

3.6 Relationship Between Slope Gradient and Distribution of Landslides

The gradient in the study area is relatively low and it is lower than 20° in most of the area (91.6 %). The gradient in the slip source area of landslides varies from 0 to 41°, especially from 5 to 20° (87.9 %).

3.7 Relationship Between Slope Aspect and Distribution of Landslides

The relationship between the slope aspect and the landslide sliding direction indicates that the dominant sliding direction is 40 to 80° and 260 to 330°. In view of that a majority of landslides are distributed along the southeast of epicenter and southwest of seismogenic fault, it may be induced that the sliding directions mainly face the epicenter and seismogenic fault, contrary to the distribution pattern of landslides induced by Wenchuan Earthquake (Xu and Li 2010; Xu et al. 2011).

4 Conclusions

Google Earth remote-sensing images have been adopted for the interpretation of landslides in the zones of intensity being higher than IX of Haiyuan Earthquake, and 1,000 landslides were identified. These landslides are distributed in the two concentration zones in the southwest of Xiji County and the southeast concentration zone of Haiyuan County, in which there were more landslides in the concentration zone of Xiji County, which could be attributed to the difference of loess thicknesses. The field surveys found that most of the landslides occurring on concave slopes gentler than 20° with long run-out and high speed may be due to the loess liquefaction.