A Seismic Landslide Hazard Assessment in Small Areas Based on Multilevel Physical and Mechanical Parameters: A Case Study of the Upper Yangzi River

Abstract

Many landslides triggered by earthquakes have caused a countless loss of life and property, therefore, it is very important to predict landslide hazards accurately. In this work, regional seismic landslide data were obtained via a field survey, remote sensing interpretation, and data collection, and a multilevel physical and mechanical parameter system for seismic landslide hazard assessment was established; this system included a landslide inventory, loose accumulation layers, and geological units, enabling higher accuracy in the data. The Newmark displacement model with a modified correlation coefficient was used to assess the regional seismic landslide hazard in four scenarios (a = 0.1, 0.2, 0.3, 0.4) to study the influence of the landslide hazard at different peak ground accelerations. Moreover, the information value model was used to modify the calculated results to improve their accuracy in the assessment. By assessing the potential seismic landslide hazard in Shimian County in the upper reaches of the Yangtze River, the regional landslide distribution and pattern at different peak ground accelerations were obtained. The results show that with decreasing parameter accuracy in the system, the importance of the landslide inventory increases. When the peak ground acceleration is a = 0.3, which can be defined as a high hazard grade, in which the landslide area demonstrates a large-scale sharp increase, a devastating hazard threshold is reached. As the peak ground acceleration increases, the factor controlling landslides transforms from the landslide inventory to the slope, which reflects the reasonableness of the parameters in the system. The input parameters were regarded as important factors for efficiently increasing the accuracy of the results of the Newmark displacement model in the discussion.

1. Introduction

Landslide hazards are among the priority issues in risk assessment systems. Many landslides triggered by earthquakes have caused countless loss of life and property. Thus, landslides have caused great concern among geoscientists.

At present, many scholars have researched seismic landslides from a statistical perspective [1,2]. Multiple historical seismic landslide datasets have focused on revealing the factors and mechanisms that trigger seismic landslides. Common seismic landslide models include information models [3], random forest models [4], and artificial neural network models [5], which are statistical analysis models that are data-driven and use large amounts of historical data to obtain results when assessing regional seismic landslide hazards. Therefore, the results are objective and credible in the spatial and temporal dimensions. However, the models are restricted by inconsistent data types or limited data coverage because different earthquakes have occurred across different regions.

Some scholars have researched seismic landslides from the perspective of engineering mechanics [6,7]. Mechanical models have been mostly utilized to simulate and predict regional seismic landslide hazards by evaluating the stability of slopes with seismic structural data or relatively complete records from a single earthquake. Such models require detailed variable spatial parameters, which are effective and require less work for a single landslide.

Commonly, the landslide occurrence mechanism, slope instability mechanism, and slip results are considered for grading the landslide hazard quantitatively [8]. The Newmark displacement model, a type of mechanical model, is commonly used for assessing seismic landslide hazards quantitatively on the basis of limit equilibrium theory. After the model was first proposed in 1965 [8], Ambraseys and Menu [9] established an analytical model for the critical acceleration ratio on the basis of a full collection of seismic network-monitoring records over many years. Milesa and Ho [10] incorporated GIS into the Newmark displacement model, combined with the ground motion parameters to establish a simplified Newmark model for calculating the permanent displacement of landslides. Peng et al. [11] combined computer technology with the Newmark model and proposed a procedure that considers topographic effects and runout behavior for analyzing seismic landslide hazards. Dreyfus et al. [12] discussed the accuracy of the Newmark evaluation model by comparing landslides caused by the 1994 Northridge earthquake in California with landslides on the basis of landslide displacement prediction. In comparison to statistical models, the Newmark displacement model can quantitatively assess the hazard of landslides despite in the lack of actual landslide samples. Currently, it is widely used in seismic hazard assessments worldwide.

Many scholars have used the Newmark displacement model to analyze landslide hazards, and have proposed improving the model by modifying the coefficients according to the characteristics of the research area [13,14]. The application scope of the Newmark model has been extended. However, some deficiencies are reflected in the limitations of the model. Jibson et al. [14] established a regression equation between the probability of failure and displacement based on the landslides inventory triggered by the Northridge earthquake. But the study area selected only the epicenter area of the Northridge earthquake where landslides are densely distributed, which does not include the whole affected area, thus likely causing discrepancy with the assessment of the whole affected area. In addition, the commonly available regional geological map data are suitable for rapid emergency assessment but insufficient for the model because of the poor data accuracy or the landslide inventory not being fully utilized [15,16]. Therefore, it is necessary to establish a parameter system to improve data accuracy for Newmark displacement model applications.

In this work, a multilevel parameter system of the Newmark displacement model was established by utilizing different accurate data that were obtained in multiple ways, and Shimian County and the surrounding area of the Yangtze River were taken as examples to verify the scientific reliability of the system in a small regional seismic landslide hazard assessment. In addition, taking the whole study area as an example, the model correlation coefficient is modified to adapt to the characteristics of the study area. The results of seismic landslide hazards in different peak ground acceleration scenarios are studied and will provide a reference for seismic landslide prevention in the study area.

2. Study Area and Geological Setting

2.1. Geographical Location

Shimian County is located around the middle reaches of the Dadu River at the upper reaches of the Yangtze River, and the geographical coordinates are 101°55′–102°34′ E and 28°51′–29°32′ N (Figure 1). The study area contains Shimian County town and its surroundings are 5.5 km long, 4.4 km wide, and 24.2 km² in total area. The area is regarded as a risk prevention and control region for geological disasters [17].

2.2. Geological Conditions

In the study area, 803.1 mm of annual rainfall is concentrated in the months from May to September, accounting for 86.4% of the total rainfall, and the annual rainfall is characterized by an uneven spatial distribution, with more rainfall in mountainous areas than in river valleys, with heavy rain or showers. The river network consists of a tertiary river—the Dadu River and its tributary—the Nanya River, in which the former runs through the area from west to east and the latter runs from south to north and finally merges into the Dadu River at the county seat.

The lithology in the district is relatively simple. The lithology can be divided into two main categories: magmatic rocks and loose accumulations (Figure 2a). The magmatic rocks are composed mainly of ordinary granite, rhyolite porphyry, diabase dikes, gabbro, granite porphyry, and ultrabasic rock, and ordinary granite γ22 is the main lithology of the stratum accounting for approximately 50% of the study area; the loose accumulation layer is divided into glacial fluvial deposits and alluvial accumulations and is distributed along both sides of the rivers.

The geological structure is significantly affected by the geostructure and presents a deep-cut topography, with rivers on three sides at the intersection of the Xianshuihe fault, Longmenshan fault and Anninghe fault. In particular, the Shimian fault is a branch of the Anninghe fault that passes through the study area directly in a nearly north–south direction (Figure 2b), breaking the rock mass and leading to the development of adverse geological engineering phenomena [18,19].

The geomorphology is a canyon landform with high terrain in the northwest, east and southwest regions, and low terrain in the central part of the area and the river valley. The study area has an elevation of 850 m, and the highest point is located in Dapinggang, with an elevation of 2030 m. The lowest point is located most downstream of the Dadu River, with a relative elevation difference of 1180 m (Figure 2c). The amount of topographic relief is 487.6 m/km². The gradients are distributed from the river surface to steep slopes, with degrees of 0° to 82°, due to the steep terrain and strong cutting of the river (Figure 2d). The average slope is 30.8°, which reaches the grading standard of steep slopes and triggers frequent geological disasters in the area.

In this work, lithology and fault data were from a 1:200,000 geological map. The slope angles of the study area were derived from a DEM with resolution 20 m. The study area is located at the intersection of three fault zones with strong structural movement and frequent earthquakes, and mass geological hazards have occurred because of strong geotectonic movement, the fault passing through the county, and rainstorms. The mountains and rocks are highly susceptible to landslides. Human engineering activities have intensified geological disasters, resulting in frequent seismic activity and high susceptibility to geological disasters [20].

3. Materials and Methods

3.1. Newmark Displacement Model

The Newmark displacement model is based on the limit equilibrium theory and is used to calculate the cumulative displacement of landslides. The principle (Figure 3) is that the landslide along the sliding surface produces transient displacement and continuous accumulation under the effect of earthquake acceleration. When the earthquake acceleration along the slope is greater than the critical load acceleration, a landslide is produced, and according to the theory of calculation, the second-order integral of the difference between the two accelerations is calculated, and the permanent displacement value is finally obtained [21].

The Newmark displacement parameters include the static factor of safety (Fs), critical acceleration (a_c), and cumulative displacement (D). The formula is used to calculate the static factor of safety (Fs) of landslides and calculate the critical acceleration (a_c) by gravity, the slope, and the static factor of safety (Fs). Then, the second-order integral of the difference between the peak ground acceleration (a) and the critical acceleration (a_c) is calculated so that the cumulative displacement (D) can be obtained. The workflow of the calculation is shown below (Figure 4).

(1)

Calculate the simplified critical acceleration (a_c)

The commonly static factor of safety (Fs) for an infinite-slope model is computed as follows:

$$Fs={\displaystyle \frac{c^{\prime }}{\gamma z\sin \beta }}+{\displaystyle \frac{\tan {\phi }^{\prime }}{\tan \beta }}+{\displaystyle \frac{m{\gamma }_w\tan {\phi }^{\prime }}{\gamma \tan \beta }},$$

(1)

where c′ represents effective cohesion; γ represents the weight of rock–soil mass; z represents the depth of the failure surface; β represents the degree of the slope; φ′ represents the internal friction angle; m represents the degree of slope saturation; γ_w represents the weight of underground water; and Fs represents the static factor of safety.

Considering the environmental conditions in which the study area is located, a region with little or no rainfall, the effect of groundwater is not considered [10]. The equations are simplified as follows:

$$Fs={\displaystyle \frac{c^{\prime }+\gamma z{\cos }^2\beta \tan {\phi }^{\prime }}{\gamma z\sin \beta \cos \beta }},$$

(2)

where z is the depth of the failure surface, in m; γ is the weight of the rock–soil mass, in N/m³; c′ is the effective cohesion, in Pa = N/m²; β is the degree of slope, in °; φ′ is the internal friction angle, in °; and Fs is the static factor of safety, which is nondimensional.

The critical acceleration (a_c) is calculated by assuming an infinite slope condition, and the equation is as follows:

$$a_c=(Fs-1)g\sin \beta ,$$

(3)

where a_c is the critical acceleration, in m/s²; g is the acceleration of gravity, which is 9.8 m/s²; and Fs is the static factor of safety, which is nondimensional.

(2)

Calculate the displacement (D)

In this study, the Jibson [21] empirical regression equation is utilized to calculate the Newmark displacement (D), which involves two regional seismic parameters, namely, the peak ground acceleration (a) and moment magnitude (M_w). The earthquake magnitude will influence peak ground acceleration (a) which is crucial for seismic landslide hazard assessments, and the M_w is calculated from the surface wave magnitude (M_s) via the empirical formula. The earthquake magnitude scale is one of the most fundamental earthquake source parameters used for measuring the strength of an earthquake [22,23]. The equation is as follows:

$${\log }_{10}D=-2.710+0.424M_w\pm 0.454+{\log }_{10}\left[{\left(1-{\displaystyle \frac{a_c}{a}}\right)}^{2.335}{\left({\displaystyle \frac{a_c}{a}}\right)}^{-1.478}\right],$$

(4)

$$M_w=0.844M_s+0.951,$$

(5)

where D is the cumulative displacement, in cm; M_w is the seismic moment magnitude; M_s is the surface wave magnitude; and a is the peak ground acceleration (approximately replacing the ground acceleration), in m/s².

The Newmark displacement empirical formula based on regression estimation can be used to assign values flexibly for M and a at different scales, which can effectively simplify the calculation process.

3.2. Modified Model

Statistical analysis models can deduce the distribution regularity, correlation, etc. of landslide data via statistical analysis of related datasets [24]. The information value model, which is a type of statistical analysis model, can comprehensively utilize environmental data from potential landslides to calculate the susceptibility of potential regional landslides. First, a parameter system was established, and the relevant data were collected, in which the factors leading to landslide hazard occurrence were identified. Then, the information value of each factor was calculated, and the seismic landslide hazard zone was calculated comprehensively according to the formula. The equation is as follows:

$$I={\displaystyle \sum _{i=1}^{n}I(Y,X_i)}={\displaystyle \sum _{i=1}^n\ln \left(\frac{S_i/S}{A_i/A}\right)}$$

(6)

where I represents the information value provided by the environmental factors; n represents the total number of factors; X_i represents the environmental factors; Y represents the seismic landslide events; S represents the gross area of landslides in the research region; S_i represents the area of landslides of different grades for X_i; A represents the area of the research region; and A_i represents the area of different grades for X_i. When I < 0, X_i represents information that is beneficial to the occurrence of landslides; and when I > 0, X_i represents information that is not beneficial to the occurrence of landslides.

In this work, Formula (2) is obtained by simplifying the empirical Formula (1) according to the rainfall conditions in the study area, which makes the results more realistic. In addition, Equations (4) and (5) were cited to calculate the Newmark displacement (D) under different peak ground acceleration (a) conditions. The information value model was introduced to supplement the Newmark displacement model, for which only the internal physical parameters were considered, by calculating the landslide hazard with external environmental factors (relative height difference, slope, lithological group, distance to the fault). In order to not excessively affect the results of the Newmark displacement model, the weight of 0.1 is assigned to the information value model result to modify the results of the Newmark displacement model by adding together.

3.3. Physical and Mechanical Parameter System

The data of large-scale rock groups that were defined via field investigations were divided into three levels. The first level was data from investigated landslide inventories, collected with sufficient accuracy to quantitatively analyze the spatial densities and seismic parameters of landslides [25]. The second level was made up of interpreted remote sensing data of the loose accumulation layer, and the third was data on geological units originating from regional geological maps.

The research region was divided into 45 areas on the basis of physical and mechanical parameters on the basis of the three levels (Figure 5), in which the number of landslide inventory units was 17 and the regional point density was 0.7 points/km². The loose accumulation layer differed greatly from the geological units; therefore, it was not based on the classification of geological units but rather divided into different categories, with 22 groups numbered SS01–SS22, including the alluvial proluvial accumulation layer, ice water accumulation layer, eluvial accumulation layer, and collapse accumulation slope. The geological units were divided into six groups on the basis of regional geological maps numbered DZ01–DZ06; rivers were excluded from the division.

In this work, the parameters and details for each group are adopted from geotechnical bibliography: Code for Geotechnical Engineering Investigation (GB 50021-2001) [26], Standard for Engineering Classification of Rock Masses (GB/T 50218-2014) [27], and Code for seismic design of building (GB 50011-2010) [28] and previous studies [18,19]. The physical and mechanical parameter values (

Table 1. Values of the physical and mechanical parameters.

Groups	Number	Z/m	γ/(t·m−3)	c′/kPa	φ′/(°)
Landslide inventory (first level)	SM448	2.5	19.0	27	30
	SM506	2.5	19.0	25	30
	SM505	2.5	20.0	25	31
	SM107	2.5	22.0	26	31
	SM311	2.5	21.5	25	31
	SM125	2.5	21.5	26	30
	SM121	2.5	21.5	25	30
	SM309	2.5	21.5	24	30
	SM310	2.5	21.5	25	30
	SM312	2.5	21.5	24	32
	SM313	2.5	21.5	25	30
	SM452	2.5	22.0	24	32
	SM205	2.5	21.5	25	31
	SM206	2.5	21.5	26	30
	SM203	2.5	22.0	24	30
	SM204	2.5	21.5	25	30
	SM209	2.5	21.5	26	30
Loose accumulation (second level)	SS01	3.0	21.5	28	35
	SS02	3.0	21.5	31	34
	SS03	3.0	21.5	30	34
	SS04	3.0	21.5	31	34
	SS05	3.0	21.5	32	34
	SS06	3.0	21.5	32	35
	SS07	3.0	21.5	32	35
	SS08	3.0	20.0	31	34
	SS09	3.0	22.0	31	34
	SS10	3.0	22.0	30	34
	SS11	3.0	21.5	31	34
	SS12	3.0	21.5	31	35
	SS13	3.0	21.0	30	35
	SS14	3.0	21.5	31	34
	SS15	3.0	20.0	32	35
	SS16	3.0	21.5	31	34
	SS17	3.0	21.5	30	34
	SS18	3.0	20.0	32	34
	SS19	3.0	20.0	31	34
	SS20	3.0	21.5	31	34
	SS21	3.0	21.0	30	35
	SS22	3.0	21.0	30	35
Geological units (third level)	DZ01	4.0	22.0	32	38
	DZ02	4.0	22.0	36	42
	DZ03	4.0	22.0	36	42
	DZ04	4.0	22.0	34	39
	DZ05	4.0	22.0	36	38
	DZ06	4.0	22.0	36	42

) of the 45 groups were assigned, and a map (Figure 6) was produced via ArcGIS 10.8 software.

4. Results

The static safety factor (Fs) and critical acceleration (a_c) were calculated via the raster calculator function of ArcGIS software on the basis of slope data and parameter values, according to Equations (2) and (3). The results are shown in Figure 7a,b.

The basic parameters of peak ground acceleration (a) and surface wave magnitude (M_s) were assigned as follows, by querying the relevant file: 1. a = 0.1, Ms = 7; 2. a = 0.2, Ms = 7; 3. a = 0.3, Ms = 8; and 4. a = 0.4, Ms = 9 [28]. The displacement value (D) is the basic parameter of seismic landslide hazard assessment under different working conditions and was calculated via Equations (4) and (5).

For the information value model, the hazard-causing parameter system of seismic landslides was established, and these factors included the relative height difference, the slope, the lithological group, and the distance to fault. In this work, topography and geological condition data were from detailed survey data [18] and each factor of the information value model (

Table 2. Hazard-causing parameter system and information values of the information value model.

Primary Index	Sub Index	Rank	Information Value/I
Topography	Relative height difference/m	<100	0.03
		[100, 200)	0.94
		[200, 300)	0.54
		[300, 400)	−0.28
		[400, 550)	−4.27
		[550, 700)	−2.75
		[700, 850)	−2.06
		[850, 1000)	−1.71
		>1000	0
	Slope/°	<10	0.4
		[10, 20)	0.08
		[20, 30)	−0.29
		[30, 40)	−0.18
		≥40	−1.81
Geological condition	Lithological group	Alluvial accumulation	0.19
		Ordinary granite	−0.54
		Rhyolite porphyry	−0.54
		Granite porphyry	0.74
		Ultrabasic rock	0.07
		River	−0.77
	Distance to the fault/m	[0, 100)	−0.43
		[100, 300)	0.18
		[300, 600)	0.36
		[600, 1500)	−0.22
		≥1500	0.04

) was calculated according to Equation (6). The study area was comprehensively divided into four grades for seismic landslide hazard zones. The results are shown in Figure 8.

The Newmark displacement model calculations were modified by the information value model and the seismic landslide hazard zones were calculated at different peak ground acceleration (a) scales via the raster calculator function of ArcGIS.

The calculated results could be classified into four categories by using the natural breakpoints tool of ArcGIS, which represents high, medium, low, and very low hazard levels (Figure 9). Obviously, with increasing peak ground acceleration, the area of high hazard increases, with a positive correlation.

The distributions of high- and medium-hazard areas under different seismic conditions were calculated via the raster reclassification function of ArcGIS (

Table 3. Areas of high- and medium-hazard zones under different working conditions.

Condition	Distribution Area/km2		Control Factors
	High Hazard	Medium Hazard
a = 0.1	2.78	4.01	Landslide inventory
a = 0.2	3.42	6.29	The rock properties and landslide inventory
a = 0.3	6.78	10.02	Slope and rock properties
a = 0.4	11.62	13.28	Slope

).

The landslide quantity and scale are positively related to the magnitude (peak ground acceleration (a)) of the earthquake and represent an inflection point where the area of the hazard zone increases sharply, which can be deduced via statistical analysis of the correlation of regional historical seismic landslides, such as the 2008 M_w 7.9 Wenchuan earthquake and the 2013 M_s 7.0 Lushan earthquake. The point is the critical value for a large regional disaster in which the landslide hazard is ruinous for the region once the magnitude (peak ground acceleration (a)) approaches or reaches it.

A power function linear curve was obtained by analyzing the correlation of the distributions of the high- and medium-hazard zones with different peak ground accelerations (a), as shown in Table 3. In this case, the correlation coefficient R² = 0.995 of the high-hazard zones and the correlation coefficient R² = 0.995 of the medium-hazard zones, with the peak ground acceleration (a), are well correlated. The area of the hazard zone sharply increases at a peak ground acceleration a = 0.3, which can be regarded as the critical point for a regionally destructive disaster.

• When a = 0.1

the area of the high-hazard zone is 2.78 km², accounting for 11.49% of the research region, and the area of the medium-hazard zone is 4.01 km², accounting for 16.57% of the research region. The hazard zone is distributed mainly from river valleys to steep slopes in the mountainous region and is significantly affected by a single landslide. The distribution of hazard zones is highly correlated with the landslide inventory.

• When a = 0.2

the area of the high-hazard zone is 3.42 km², accounting for 17.02% of the research region, and the area of the medium-hazard zone is 6.29 km², accounting for 25.99% of the research region. The area of the hazard zone increased significantly, with a growth rate of 48.02% compared with a = 0.1 g. The distribution characteristics of the high-hazard zone rapidly increases along the topographic features of the river valley, with a zonal distribution in the steep slope area in many regions, and the distribution is significantly controlled by the rock properties and landslide inventory in the high-altitude area at a grade of a = 0.2.

• When a = 0.3

the area of the high-hazard zone is 6.78 km², accounting for 28.05%, and the area of the medium-hazard zone is 10.02 km², accounting for 41.40% of the research region. The area of the hazard zone increases significantly, with a growth rate of 64.56% in comparison with a grade of a = 0.2 g. The distribution characteristics clearly reveal that the hazard zone spread to surrounding areas along the slope in multiple regions on the bases of the controlling factors, i.e., the slope and rock properties.

• When a = 0.4

the area of the high-hazard zone is 11.62 km², accounting for 48.03%, and the area of the medium-hazard zone is 13.28 km², accounting for 59.01% of the research region. Compared with a = 0.3 g, the area of the hazard zone increases significantly, with a growth rate of 71.38%. The area of the hazard zone increases geometrically at a seismic condition of a = 0.4 g, and the growth rate far exceeds the condition of low peak ground acceleration. The whole region is surrounded mostly by high- and medium-hazard zones. The distribution characteristics of the hazard zone show that the area increases sharply, the carrying capacity of the environment reaches the greatest limit, the hazard zone in the mountain region is divided into smaller areas, and the slope becomes the main factor controlling hazards. High- and medium-hazard zones are widespread in Shimian County and surrounding areas.

In summary, the hazard zone distributions at peak ground accelerations (a) of 0.1, 0.2, 0.3, and 0.4 were obtained by assessing the seismic landslide hazard in Shimian County and the surrounding areas. When the peak ground acceleration a = 0.1, 0.2, the seismic landslide hazard mainly occurred in the areas of landslide inventory. With the gradual increase of a, the larger slope region became the main hazard area. The controlling factors in the hazard zone gradually transitioned from the landslide inventory to the slope, demonstrating the rationality of the multilevel physical and mechanical parameter system. The linear fitting curves of the high- and medium-hazard zone areas with peak ground acceleration (a) indicate that when a = 0.3, the area of the hazard zone increases dramatically on a large scale, which can be regarded as the critical point for a regionally destructive disaster.

5. Discussion

The accuracy of the hazard assessment was determined by the precision of the parameters. To date, regional geological units have been used as the study units for small-scale seismic landslides in some studies [29,30]. Accurate rock group data are necessary for studying small-region geological hazards, but the rock groups cannot be divided via small-scale geological maps, such as the commonly used 1:50,000 geological maps. A multilevel physical and mechanical parameter system was established on the basis of field investigation data. Considering the data accuracy of the system, data of the rock–soil mass was divided into three levels: the landslide inventory data, data of the loose accumulation layer originating from remote sensing interpretation, and data of the geological units derived from the regional geological maps. With verification of examples, the precision of the parameters gradually decreases from the first level to the third level and reflects the importance of the landslide inventory in hazard assessment, which is closer to the actual conditions. However, whether the division of multilevel parameter groups is a more accurate method for hazard assessment in small areas can be further analyzed in future work.

Owing to the limitations of environmental conditions, the effect of groundwater was not considered for the Newmark displacement model (the study area during the season of less or no rainfall). Miles and Mo [10] considered the effect of groundwater to be simplified in dry conditions, Huang et al. [31] found the predicted displacement of landslides to be far less than the actual displacement in the condition of high groundwater. Dreyfus et al. [12] demonstrated that when different simplified Newmark models are compared, accurate input parameters may have a greater influence on the assessment results. Therefore, the prediction of sliding displacement by considering various groundwater levels should be emphasized in future work.

The calculation of Newmark displacement involves a large amount of regional data and seismic parameter data that are not easily accessible, making effective performance difficult for predicting earthquake landslide zones. Thus, many scholars have established empirical formulas to simplify the calculation process. As Jibson [21] calculated an empirical regression equation by analyzing a large amount of data in comparison with actual landslides, the equation was verified to be highly accurate and to meet the requirements for predicting the cumulative displacement of seismic landslides. Xu et al. [32] corrected the regression parameters and established a Newmark model for the Wenchuan earthquake on the basis of 221 ground motion records of the earthquake. Yuan et al. [33] developed new fitting coefficients for the Newmark model on the basis of the dataset of the 2013 Lushan Ms 7.0 earthquake, and in this study, on the basis of prior experience, the empirical formula of Jibson is used to assess seismic landslide hazards. Moreover, Ma and Xu [34] reported that the different empirical formulas for the Newmark model may have little effect on the accuracy of landslide hazard assessment. Therefore, the use of empirical formulas might not affect the accuracy of the assessment, even if not in similar or analogous regions.

Huang et al. [35] presented a new method of integrating statistical analysis models (linear regression (LR) and support vector machine (SVM)) with critical acceleration (a_c) for earthquake landslide hazard assessment, and demonstrated that the new assessment method is more effective than the simplified Newmark model for seismic landslide hazards. The information value model is a statistical analysis model that statistically analyzes relevant data to calculate the landslide hazard zone on the basis of historical seismic landslide events. Du et al. [36] showed that a combined model is a significant improvement over the Newmark displacement model alone. Related studies have shown that combining multiple methods provides better assessment results than a single model does. Thus, in this study, the information value model is used to assign different weights to correct the seismic landslide hazard results calculated by the simplified Newmark displacement model to improve the accuracy of the assessment results.

The research results show that when a = 0.3, the area of hazard regions will increase significantly in Shimian County and its surrounding areas. The distribution map of seismic landslide hazards, which can clarify the distribution region of hazard and the key prevention areas, has an important role for the prevention and control of landslides in Shimian County.

6. Conclusions

In this work, an assessment of the potential danger of landslide hazards was carried out in a small area. The following conclusions can be drawn:

(1)

A multilevel physical and mechanical parameter system was established on the basis of the Newmark displacement model, which includes a landslide inventory, a loose accumulation layer, and a geologic unit and the accuracy gradually decreases from the first level to the third level. The accuracy of input parameters is effectively improved, demonstrating the rationality of the parameter system;

(2)

The Newmark displacement model with a modified correlation coefficient is used to assess the hazards of regional seismic landslides in four scenarios (a = 0.1, 0.2, 0.3, 0.4). The calculated result was modified by the information value model and the seismic landslide hazard area was delimited scientifically on the basis of the results;

(3)

Improving the accuracy of input parameters is important for improving the assessment results of the Newmark displacement model, such as the division of the study unit, the rationality of parameter inputs or the quantity of seismic events. In future work, improving relevant parameters for the Newmark displacement model can be further studied.