Skip to content
BY-NC-ND 3.0 license Open Access Published by De Gruyter Open Access December 22, 2016

Post-Earthquake People Loss Evaluation Based on Seismic Multi-Level Hybrid Grid: A Case Study on Yushu Ms 7.1 Earthquake in China

  • Xiaohong Yang , Zhong Xie EMAIL logo , Feng Ling , Xiangang Luo and Ming Zhong
From the journal Open Geosciences

Abstract

People loss is one of the most important information that the government concerns after an earthquake, because it affects appropriate rescue levels. However, existing evaluation methods often consider an entire stricken region as a whole assessment area but disregard the spatial disparity of influencing factors. As a consequence, results are inaccurately evaluated. In order to address this problem, this paper proposes a post-earthquake evaluation approach of people loss based on the seismic multi-level hybrid grid (SMHG). In SMHG, the whole area is divided into grids at different levels with various sizes. In this manner, the efficiency of data management is improved. With SMHG, disaster statistics can be easily counted under both the administrative unit and per unit area. The proposed approach was then applied to investigate Yushu Ms7.1 earthquake in China. Results revealed that the number of deaths varied with different exposure grids. Among all the different grids, we found that using the 50×50 exposure grid can get the most satisfactory results, and the estimated number of deaths was 2,203, with an 18.3% deviation from the actual loss. People loss results obtained through the proposed approach were more accurate than those obtained through traditional GIS-based methods.

1 Introduction

Frequently occurring massive earthquakes have caused severe losses, especially human casualties [1]. For example, Wenchuan earthquake (Ms8.0) in China in 2008 caused 69,142 death tolls, Haiti earthquake (Ms7.3) in 2010 resulted in approximately 222,500 death tolls, east Japan earthquake (Mw9.0) in 2011 yielded 15,884 death tolls, Nepal earthquake (Ms8.1) in 2015 reached 8,219 death tolls. In order to reduce losses caused by an earthquake, a quick and accurate post disaster emergency rescue is becoming more and more important. It would be very helpful for this rescue if an estimation of the impending catastrophe could be made on the basis of one of the recent earthquake prediction tools [25]. The rescue decision-making should be more effective with more detailed information about the earthquake, however, it is almost impossible to obtain comprehensive disaster information only within the first 1 to 2 hours after an earthquake [6]. In practice, as an alternative, descriptive earthquake parameters, such as earthquake magnitude and peak ground acceleration, are often used as input to estimate possible losses and to provide emergency disaster information for rescue decisionmaking.

Earthquake disasters cause losses, including humans, buildings, and economics. Among all these losses, human casualties maybe the most important factor for the government to determine the corresponding post-disaster emergency rescue strategies. For example, different levels of people loss correspond to different responses and rescue levels based on Chinese Earthquake Emergency Response Plan [7]. Therefore, the number of human casualties must be estimated rapidly and accurately within 2 hours after an earthquake to develop and implement appropriate earthquake emergency responses and rescue strategies [6].

Various approaches have been proposed to evaluate post-earthquake people loss. These approaches can be generally categorized into two classes, namely, empirical model-based approach and the building vulnerability-based approach. In the former, an empirical formula is established to determine the number of deaths by subjecting historical seismic data to regression analysis [8]. This approach requires low amounts of input data, including magnitude and intensity, which are suitable for rapid evaluation after an earthquake. However, this model is usually established on a global or national scale because of limited historical data on strong earthquakes [9]. Inaccurate evaluation results are also obtained when a global or national empirical model is used in a certain region. By comparison, building vulnerability-based approach is another extensively used method to assess people losses because 75% of deaths come from building damages [10]. In this approach, buildings and building damages are generally classified into several types, and destructive probability matrices (DPM) are used to present the destructive probability of various buildings at different seismic intensities. Based on DPM, the number of deaths under different building damages can be calculated. This approach is often constructed by using the geographic information system (GIS) technology [1114].

The GIS-based evaluation methods usually use the administrative unit-based exposure data, which regard an entire administrative area as the stricken region. However, due to a large land territory of administrative units, such as in China, it is a common situation that only a component of administrative regions is affected in one earthquake event. Therefore, an erroneous evaluation result may be obtained and inappropriate rescue decisions may be implemented when exposure data of an entire administrative area is used [15]. For example, Qinchuan county was not provided with an appropriate rescue level after Wenchuan earthquake occurred in China. One of reason for this problem is that the disaster exposure data, such as population and buildings, were extracted on the basis of administrative units and detailed local information was inappropriately considered.

In order to overcome the above problem, the spatial information grid based (SIG-based) approach has been applied to assess seismic disasters. With a high-precision grid, the efficiency of seismic damage assessment can be improved [15]. In existing SIG-based methods, an entire disaster region is divided into several grid units, which exhibit same shapes and sizes, and seismic intensity and exposure information are assumed to be evenly distributed within a grid unit. However, one grid unit may contain several different countries, and the intensity values within a grid may vary. Moreover, existing SIG-based methods generate evaluation results in a grid unit, which is inconvenient for disaster management and post-earthquake rescue in administrative units.

In current evaluation methods, seismic intensity and exposure data are often assumed to be evenly distributed within an administrative unit or a grid unit. However, they do not consider the spatial disparity of influencing factors, which is critical to the evaluation. In order to consider the spatial heterogeneity of seismic intensity and exposure data, this paper proposes a seismic multi-level hybrid grid (SMHG) approach for people loss evaluation. In the proposed SMHG, a whole area is divided into grids at different levels with varying sizes. Different people loss related data are stored in different levels of the grids to improve data management efficiency. Specially, a hybrid grid, which is generated by overlaying the isoseismal map, county administrative unit-based map and a regular grid format map, is introduced to SMHG. With the proposed hybrid grid, both the uneven distribution of spatial data in an entire region and the heterogeneity of internal information in a grid unit can be considered. On the basis of SMHG, we aimed to establish a novel scientihc and reasonable approach for the acquisition, intelligent storage, inte-grated management, and statistical analysis of earthquake spatial data and for the timely and accurate evaluation of post-earthquake people loss.

2 Methodology

2.1 Seismic Isoseismal Map

Seismic intensity is one of the most important influencing factors for earthquake casualties. In general, more casualties are produced in a high intensity region than in a low intensity region. The values of regional intensity often decrease as the distance from the epicenter increases, and the relationship between distance and intensity can be expressed by using a seismic isoseismal map. A seismic isoseismal map is a continuous planar intensity map that illustrates the spatial distribution of a seismic intensity field [15]. Seismic intensity at any point in a disaster area can be obtained from the seismic isoseismal map, which is considered a hazard map used to depict the seismic field of influence of the earthquake.

Obtaining an actual seismic isoseismal map requires a long time and cannot meet the time acquirement of rapid post-disaster rescue. Therefore, a theoretical seismic isoseismal map is used as a substitute map. If the hypocenter is assumed as a point source model, an elliptic attenuation model can be used to simulate seismic intensity attenuation [16]. A typical elliptic attenuation model is represented as follows:

Ia = a1 + a2M - a3ln(Ra + R0a)Ib = b1 + b2M - b3 ln(Rb + R0b)(1)

where Ia and Ib represent the average impact intensity along the major and minor axis directions of the ellipse, respectively. M is the magnitude. al, a2, and a3 are parameters of the major axis direction of the ellipse. Ra is the epicentral distance along the major axis directions of the ellipse. Roa is the near-field saturation factor along the major axis directions of the ellipse. bl, b2, b3, Rob and Rbare parameters of the minor axis of the ellipse. The parameters in Eq. (1) are usually obtained through the statistical regression of historical earthquake data. The values of the parameters used in this study are obtained from the Provincial Earthquake Administration (PEA) in China.

The direction angle of the major axis is another parameter necessary to generate a seismic isoseismal map. In general, the rupture direction of a fault zone is the direction angle of the major axis [16]. The fault zone that induced an earthquake can be obtained by locating the earthquake epicenter position on a fault zone distribution map. In this study, the nationwide fault zone distribution map of China is stored in the SMHG geodatabase. Thus, we can quickly locate the position of an earthquakeߣs epicenter on this map and use the nearest fault zone direction as the fault rupture direction of the earthquake.

2.2 Post-Earthquake Evaluation of People Loss

In an earthquake, deaths and injuries are related not only to seismic intensity and magnitude but also to other influencing factors, such as population density and building vulnerability. Ideally, considering all influencing factors for every earthquake will get a more accurate evaluation result. However, only several of the most important influencing factors are commonly used in practical evaluation models of people loss because numerous influencing factors can greatly reduce the efficiency and universality of models, and it is impossible to gather enough information of all influencing factors for every earthquake.

In this paper, besides several extensively used influencing factors, time and population density are also used as additional fix factors in the evaluation model of people loss. The number of deaths (ND) and the number of injuries (SD) are calculated as follows [17]:

ND = ft * fp * RD * Mp(2)
SD = ml * ND(3)

where ft is the time fix factor; fp is the population density fix factor; Mp is the total number of the population in the calculation area. Mlis the proportion factor value, which is generally between 3 and 5 according to the relationship between SD and ND[18]. RD is the people death ratio and is calculated as follows [17]:

log10RD = 9.0 * RB0.1 - 10.07(4)

where RB is the ratio of the area of severely damaged and completely destroyed buildings to the total area of all buildings.

The building vulnerability-based evaluation method, which is extensively used to evaluate building damage, is used in this study to calculate RB. The building damage area S is determined as follows [19]:

S(m, j|I)=λ(m,j|I)T(m)(5)

where m is the type of buildings. j is the damage level of buildings. I is the seismic intensity. T(m) is the construction area of the m type building. λ(m,j/i) is the building damage ratio of the m type building at the j damage level andI intensity, which is a kind of destructive probability matrices (DPM). S(m,j/I) is the building damage area of the m type building at thejdamage level and I intensity. When j is the severe damage level, the area of severe damage can be calculated on the basis of Eq. (5). The area of complete damage can also be determined using Eq. (5).

The DPM [referring to A(m,j/I)] is a relevant parameter to estimate people loss caused by an earthquake. In general, a statistical method based on past earthquake disaster data is extensively used to obtain the building damage ratio. For example, Huang et al. [20] utilized the statistical earthquake loss data of Wenchuan earthquake in 2008 and the achievements of earthquake loss estimation results from Fujian Province during the 9th, 10th, and 11th 5-year-plan periods in China to obtain the building damage ratio. Yin etal. [21] analyzed the relationship among building reactance, building type, seismic intensity, and building damage ratio on basis of the historical seismic damage data of thousands of buildings. They then established DPM.

In this study, DPM was obtained from Yin [21], which spatially included the whole Chinese mainland area. In Table 1, an example showing parts of the DPM for the A-type building and the B-type building is used for illustration. In Yin's method, the type of buildings (m) is classified as reinforced concrete structure (A), brick structure (B), lime mortar structure (C), and adobe/stone structure (D). The damage level of buildings (j) is categorized as basic intact (Bi), slight damage (Sl), moderate damage (Md), severe damage (Sd), and complete damage (Cd). Seismic intensity (I) in DPM ranges from VI to X because the damage caused by an earthquake with a seismic intensity of less than VI is low and thus not considered in the evaluation, although the seismic intensity ranges from I to XII based on the Seismic Intensity Scale of China [15]. Moreover, an earthquake of a seismic intensity XII has not been recorded in China. Therefore, the column values of the DPM range from VI to X [21], as shown in Table 1. With DPM, the building damage ratio can be easily obtained for every damaging earthquake occurring in mainland China. For example, when an A-type building is struck by an earthquake of seismic intensity VI, the damage ratio for “basic intact” is 88%.

Table 1

An example of DPM in different building types and intensities (%).

(a) The DPM of A-type building(b) The DPM of B-type building
IBiSdMdSdCdIBiSdMdSdCd
VI8812000VI66.9726.085.421.360.17
VII7523200VII63.3723.118.973.600.96
VIII553310.31.50.2VIII48.2524.1216.278.253.10
IX3530.525.57.51.5IX28.6722.9723.3916.088.90
X1520.540.516.57.5X8.4214.0424.4328.0525.06

Note: I = seismic intensity; Bi = basic intact; Sl = slight damage; Md =moderate damage; Sd =severe damage; Cd = complete damage.

Earthquakes occurring at night often cause a more serious damage than those happening at daytime because most people are sound asleep at night; as such, their escape consciousness is weak. Therefore, considering time factor in an evaluation model can improve evaluation results. Previous research results showed that, the difference in death ratio between earthquakes occurring at night and at daytime likely decreases when intensity increases [17]. In this study, daytime is set at 6:00 to 18:00, and nighttime is set at 18:00 to the next 6:00. Supposing the value of ft in the daytime is 1, values of ft at night are displayed in the Table 2[17]. However, dividing one day into daytime and nighttime is too absolute. For example, some people are already awake and engaged in outdoor activities if an earthquake occurs at 7:00; however, other people are still in bed. As such, whether 7:00 is daytime or nighttime is difficult to determine. Therefore, 6:00–8:00 and 17:00–19:00 are defined as the critical time and the corresponding values of ft at critical time are displayed in the Table 2[22].

Table 2

Values of ft in the nighttime and critical point.

ftVIVIIVIIIIXX
Night time178421.5
Critical point8.54211

Earthquakes occurring in high population density areas likely cause more deaths than those happening in low population density areas. The population density fix factor fp is used to incorporate this kind of information into the evaluation model. The values of fp are obtained on the basis of the National Population Census in 2010 [22]. Different population density ranges and their corresponding fp are shown in Table 3.

Table 3

Values of the fix factor fp for different population densities (person/km2).

Population Density<5050~200200~500500~1000>1000
fp0.81.01.11.21.3

2.3 Seismic Multi-Level Hybrid Grid

In the evaluation models of people loss, different types of data, such as fundamental geographical earthquake thematic data and evaluation model parameters, should be used. A relevant information management platform should also be constructed because earthquakes occur dynamically in time and space, and relative data are spatially heterogeneous and complex. Spatial information grid (SIG) is an effective approach to organize and dispose spatial data in a grid unit in a distributed network environment. SIG has been applied to various disaster assessment models; with this approach, result accuracy can be improved by considering the spatial disparity of disaster exposure data [23]. Generally, SIG can be categorized into two types, namely, irregular grid and regular grid. An irregular grid comprises arbitrary polygons with different sizes and shapes, whereas a regular grid is a series of polygons with the same sizes and specifications.

An irregular grid is often based on an administrative unit [24]; as such, relevant administrative departments can conveniently manage disasters and establish rescue strategies. However, this method is limited by some drawbacks. For example, the entire administrative regions are assumed to be affected in an earthquake disaster region and people are considered averagely distributed inside administrative units. In fact, however, earthquake often only affects a component of an administrative region and humans are likely found around the main settlement points. Therefore, entire administrative regions should not be assumed as the affected area because such assumptions may result in erroneous evaluations and generate inappropriate rescue decisions.

Regular grid tries to use grids with the same sizes to manage the disaster exposure data. Regular grid-based methods can represent the spatial heterogeneity of earthquake disaster data to a certain extent; however, the internal information of a grid unit is assumed to be evenly distributed, and different intensity regions are often wrongly combined into one grid unit. For example, Fig. 1(a) is a grid unit of a regular grid format map, and the intensity of the whole grid is set as VII. However, a spatial heterogeneity is found in the grid [Fig. 1(b)] by overlaying the seismic isoseismal map, the county administrative unit-based map, and the regular grid format map. Although Fig. 1(a) and Fig. 1(b) cover the same area, in Fig. 1(b), this area is divided into three parts. The intensity of part C is VI. Although part A and B have the same intensity value of VII, they belong to different counties and possess different exposure data. Therefore, simply regarding the whole gird as an intensity VII region with the same exposure data is in-accurate in some cases.

The use of either regular grid or irregular grid cannot fully express the spatial characteristics of various disaster data. To address this problem, this paper proposes a novel hybrid grid-based method. The grid units in Fig. 1(b) are actually an example of a hybrid grid. In Fig. 1(b), the uneven distribution of spatial data in a whole region and the heterogeneity of internal information in a grid unit are considered in a hybrid grid. Using the hybrid grid, disaster statistics can be easily be counted in the administrative unit and the per unit area. To improve data management efficiency, the multi-level hybrid grid is used. In an SMHG, a whole area is divided into grids at different levels with various sizes.

Figure 1 Grid unit in different grid format maps (a) Grid unit in the regular grid format map; (b) Grid unit in the overlaid map. The numbers in the figures represent the intensity values of the regions. In Fig.1 (b), the yellow line is the seismic isoseismal line and the red line is the boundary of the county.
Figure 1

Grid unit in different grid format maps (a) Grid unit in the regular grid format map; (b) Grid unit in the overlaid map. The numbers in the figures represent the intensity values of the regions. In Fig.1 (b), the yellow line is the seismic isoseismal line and the red line is the boundary of the county.

In this research, an SMHG platform is established for the post-earthquake evaluation of people loss in mainland China. The level of SMHG is very significant because it can affect data storage efficiency and evaluation result accuracy [25]. To ensure that the level division of SMHG is in accordance with the distribution characteristics of earth-quake data and the requirements of people loss evaluation, the SMHG is categorized into three levels, namely, regional grid, provincial grid and exposure grid.

  1. Regional grid

    As the first level of SMHG, the regional grid aims to divide mainland China into several components. Each component is a regional grid unit covering several provinces, which have similar values of people loss related factors. For example, small earthquakes generally occur near the Jiulingshan fault zone in Hunan, Hubei, and Henan provinces in central China, and the values of DPM are the same. Therefore, separately storing the rupture direction of the fault zone and the DPM of the three provinces consumes a large storage space and time. In SMHG, these three provinces are integrated into one regional grid. Among various ways used for earthquake regional division in China, the division proposed by Deng [26] is selected in our study. In this approach, mainland China is divided into seven regional grids, namely, Dongbei, Xinjiang, Ganqingning, Yunchuanzang, Huabei, Huazhong, and Southeast Coast.

  2. Provincial grid

    Provincial grid is the second level of SMHG. Each provincial grid unit covers one province in China. The second level grid is divided because the similarity of the internal elements of the regional grid unit is relative and incomplete. For example, an elliptic attenuation model used in different provinces often provides different parameters. Therefore, it is unsuitable to store these parameters in a regional grid unit. In China, each province is equipped with a PEA, which is a specialized research institution for earthquakes. Each PEA has established a specialized elliptic attenuation model on the basis of its natural and social characteristics.

  3. Exposure grid

    Exposure grid, which is a kind of hybrid grid, is the third level of SMHG. This level aims to provide exposure data, such as population data and building data, necessary to evaluate people loss. The provincial grid is unsuitable for exposure data storage because a strong earthquake often spreads in several provinces. However, some counties of an affected province do not suffer from earthquakes. In order to generate an exposure grid format map, the disaster-affected area is firstly gridded with a regular grid. Then, the regular grid format map is overlaid with a county administrative unit-based map to generate the exposure grid format map. A regular grid format map comprises n×n grids, which indicate that the whole area is divided into n rows and n columns, where n is an integer. In this study, this map is referred to as n×n regular grid format map. Accordingly, the exposure grid format map generated from the n×n regular grid format map is called n×n exposure grid format map. Using different n, we can generate different kinds of regular and exposure grid maps. The exposure grid map can express the exposure data more accurately when the value of n is higher, however, the evaluation process needs more time.

2.4 Post-Earthquake Evaluation of People Loss Based on SMHG

In SMHG, people loss related data are hierarchically stored and managed in different grid levels. Evaluation is closely connected to this hierarchical grid system. The technical route of SMHG-based post-earthquake evaluation of people loss is shown in Fig. 2. The maps used in this evaluation approach can be divided into three categories: hazard map, exposure map, and impact map. A hazard map shows the distribution of disasters caused by an earthquake. An exposure map is used to store exposure data, such as population density and building data. An impact map is utilized to provide the evaluation results of people loss.

Figure 2 Technical route of the evaluation of people loss based on the SMHG platform.
Figure 2

Technical route of the evaluation of people loss based on the SMHG platform.

The entire evaluation process can be divided into four stages. In the first stage, an earthquake database is established and an exposure map is developed in a grid format. The post-earthquake evaluation of people loss covers different types of data, including fundamental geographical earthquake data, rupture direction of the fault zone, evaluation model parameters, population data and building data. These data are hierarchically stored in SMHG.

In the second stage, a hazard map is generated based on the intensity attenuation model. To generate a hazard map, magnitude, epicenter intensity, direction angle of the major axis and parameters of the elliptic attenuation model are needed. In general, the magnitude, epicenter intensity, and location information of the epicenter are published by China Earthquake Administration (CEA) within a few minutes after an earthquake occurs. According to the location information, the epicenter can be located in the regional and provincial grid format maps to obtain the direction angle of the major axis and the parameters of the elliptic attenuation model. Based on the preliminary data, the hazard map can be generated according to the intensity attenuation model.

In the third stage, the exposure data of populations and buildings in different seismic intensity zones are obtained. The exposure data of populations and buildings in different seismic intensity zonings can be obtained by overlaying the hazard and exposure maps. These data are the preliminary data for the evaluation under different intensity zonings.

In the fourth stage, people loss is calculated and an impact map is generated. RB is calculated on the basis of the DPM database and the building damage evaluation model. People loss in different regions and damage levels can then be calculated by using the evaluation model of people loss and the preliminary exposure data. The impact map is then obtained, and the evaluation result can be presented in various ways, such as tables and rendering maps.

3 Case Study

3.1 Study Area

On April 14, 2010, Ms 7.1 earthquake occurred in Yushu County, Qinghai Province in northwest China. Yushu earthquake caused frequent aftershocks and severe secondary disasters. This catastrophe is also considered a typical earthquake in China. Therefore, Yushu earthquake is chosen in this study to illustrate the proposed evaluation approach. The epicenter intensity of this earthquake was IX, according to the data released by the CEA. The epicenter is located at 33.1 °N and 96.6 °E.

3.2 Data Preparation

Two key stages are involved in the determination of people loss by using the proposed evaluation model. In the first stage, a theoretical isoseismal map is generated by using preliminary data, such as magnitude, epicenter intensity, elliptic attenuation model parameters, and direction angle of the major axis. In general, magnitude and epicenter intensity are published by CEA within a few minutes after an earthquake. Elliptic attenuation model parameters were stored in the provincial grid. The direction angle of the major axis can be obtained on the basis of the rupture direction of the fault zone. The detailed steps are presented in Section 3.3.

In the second stage, people loss is calculated. RB and exposure data are essential for the evaluation model (Section 2.2). The building damage ratio obtained from the DPM database can be used to calculate RB. The exposure data of an earthquake include construction area of different types of buildings and population data. The total construction area of buildings is calculated by the population data and the average per capita living space, where the population data obtained from China statistical yearbook for regional economy in 2010 [27] and the average per capita living space was 21.33 square meters per person according to the 6th census data. The proportion of the four types of buildings should be identified to obtain the construction area. In this study, the proportion of the four types of buildings is based on the mini-census data of 2005 (Table 4) [28].

Table 4

The proportion of the four types of buildings (%).

TypesABCD
Town10155025
Countryside1.21.852.944.1

3.3 Hazard Map Generation Based on SMHG

The elliptic attenuation model parameters and the direction angle of the major axis are two key preliminary data to generate a hazard map. The parameters of the elliptic attenuation model are initially obtained from the provincial grid. The elliptic attenuation model is obtained from the Qinghai provincial grid because the epicenter of Yushu earthquake is located in Qinghai:

Ia = 5.643 + 1.538M - 2.109 ln(Ra + 25)Ib = 2.941 + 1.363M - 1.494 ln(Rb + 7)(6)

Afterward, the direction angle of the major axis is deter-mined. Yushu earthquake occurred near the Ganzi-Yushu fault zone, where the rupture direction is NW-SE. Therefore, the direction angle of the major axis is 135°.

On the basis of the elliptic attenuation model of Yushu earthquake and the direction angle, the theoretical seismic isoseismal map (hazard map) can be produced, and the result is shown in Fig. 3(a). Fig. 3 (b) shows the seismic isoseismal map produced on the basis of field investi-gations and published by the government several months after Yushu earthquake. It is noticed that Fig. 3(a) and Fig. 3(b) not only exhibit similar shapes but also cover similar areas. In addition, both contain eight affected counties, namely, Yushu, Zhiduo, Chengduo, Qumalai, Shiqu, Zaduo, Nangqian, and Shengda. A highly accurate theo-retical seismic isoseismal map provides a good foundation for the evaluation of people loss.

Figure 3 Seismic isoseismal maps (a) Theoretical seismic isoseismal map based on SMHG; (b) Real seismic isoseismal map.
Figure 3

Seismic isoseismal maps (a) Theoretical seismic isoseismal map based on SMHG; (b) Real seismic isoseismal map.

3.4 Impact Map Generation and Result Analysis Based on SMHG

3.4.1 Impact Map and People Loss Estimation Result

An impact map illustrates evaluation results and shows the people loss in different regions and various damage levels. A hazard map and an exposure map are necessary to create an impact map. The hazard map can be obtained by using the elliptic attenuation model described in Section 3.3. For the exposure map, an n×n regular grid format map that covers the same area with the hazard map is initially created. Afterward, the n×n exposure grid format map is generated by overlaying the n×n regular grid format map and the county administrative unit-based map. By using the hazard map and the exposure map as input, an overlaying process map, which has the exposure data of populations and buildings in different seismic intensity zones, is generated through doing a spatial intersect analysis. ND and SD are then calculated on the basis of the exposure data of populations and buildings, the post-earthquake evaluation model of people loss, and DPM. Thus, the impact map is generated.

Figure 4 Impact maps rendering by RD and ND (a) Impact map rendering by RD; (b) Impact map rendering by ND.
Figure 4

Impact maps rendering by RD and ND (a) Impact map rendering by RD; (b) Impact map rendering by ND.

Using the impact map, people loss statistics can be counted in all of the grid units. Fig. 4 illustrates two impact maps produced on the basis of RD and ND by using the 50×50 exposure grid in Yushu earthquake. In general, RD and ND are higher in the higher-intensity region than in the lower-intensity region (Fig. 4). This finding indicates that more deaths are recorded in the higher-intensity region than in the lower-intensity region, because earthquake is more destructive in the former than in the latter.

However, in the highest intensity region of Fig. 4(a), the death ratio in the epicenter is not the highest because this parameter is also related to building type, building area and other influencing factors. In several lower-intensity regions, such as Qumalai, RD is almost equal to zero because the number of deaths is extremely low. To obtain the RD for each grid, the extremely low ND is divided by the total number of the population; as a result, a low RD is obtained. By contrast, the impact map produced on the basis of ND can clearly express the spatial distribution of death number in grid unit [Fig. 4(b)]. In Fig. 4(b), ND in a higher-intensity region is not always larger than that in a lower-intensity region because the exposure grid is not the regular grids with the same sizes. Grids near the boundary between seismic isoseismal zones and the boundary between counties are often divided into several sub-grids. The total number of the population of a sub-grid is low because the area coverage of a sub-grid is much less than that of a grid, and ND is depend on RD and the total number of the population. Thus, a low ND is obtained in the sub-grid.

Overall, the rendering map produced on the basis of RD and ND expresses the severity of human loss and the number of the human loss after an earthquake, respectively. An evaluation based on these two aspects can provide a more comprehensive and reasonable result. Moreover, the number of deaths and spatial distribution are displayed in the impact map in a grid unit, which can provide more specific and accurate data supporting the deployment of rescue materials and relief workers.

Deaths and injuries can also be easily counted in an administrative unit except the statistics in a grid unit. The attribute table of the impact map provides detailed people loss information. Thus, this information can be used to calculate the number of deaths and injuries in different counties (Table 5). It is noticed that the largest number of deaths and injuries is recorded in Yushu County. In particular, the number of deaths in Yushu county is 1,985, which is several hundred times larger than that in other counties, such as Zhiduo, Qumalai and Zaduo. The main reason for this is that Yushu County covers four seismic intensity regions, particularly containing intensity IX and VIII regions, which are not covered in other counties. The hazard map reveals that almost the entire territory of Yushu is located in the destructive seismic intensity region. As such, the largest number of deaths and injuries is obtained in Yushu.

Table 5

People loss in the Yushu earthquake in different counties.

YushuZhiduoChengduoQumalaiShiquZaduoNangqianShengda
IntensityIX/VIII/VII/VIVII/VIVII/VIVIVIVIVIVI
Deaths1985411053092931
Injuries7940164402012036116124

Note: The bold means the largest number of deaths and injuries.

By contrast, Zhiduo and Chengduo counties cover intensity VII and VI regions. The number of deaths in Chengduo is 110, which is the second largest number of deaths. However, deaths and injuries are relatively low in Zhiduo, because only a very small component of this county is found in the intensity VII region, and its total number of the population is small. The number of deaths in Qumalai, Shiqu, Zaduo, Nangqian, and Shengda was low because the components of the area are located in the intensity VI region in these counties. These areas are also located in regions below intensity VI, which have no damage to be considered. In the evaluation model of people loss, the number of injuries is a multiple of the number of deaths. Therefore, the characteristics of SD are the same as those of ND.

3.4.2 Results Comparison and Analysis

The deaths and injuries in hve different kinds of exposure grids and the time consumed by grid generating and evaluation are compared to analyze the accuracy and speed of the evaluation results of people loss. The processing environment is the Intel(R) Core(TM) i5-3470, CPU@3.20 GHz, 4.00 GB RAM, Windows 7 Pro32-Bit with SP1 on a PC. Table 6 shows the comparison of the evaluation results with different precisions of exposure grids.

Table 6

Comparison of the evaluation results.

Exposure GridReal Casualties
2×210×1020×2050×50100×100
ND207821462176220322032698
SD8312858487048812881212135
Tg6s7s7s8s11snull
Te3s4s11s43s149snull

Note: ND = the number of deaths; SD = the number of injuries; Tg = the grid generating time consumed; Te = the evaluation time consumed. The bold means the best evaluation results.

The evaluation results and time consumed are different in various exposure grids. When the number of grids increases, both the grid generating time consumed and the evaluation time consumed are increased. The difference in the grid generating time consumed is not too large. Generating a 2×2 grid format map takes 6 seconds, but generating a 50×50 grid format map takes 8 seconds. The evaluation time consumed with different grids is considerably different. The evaluation based on the 2×2 exposure grid only takes 3 seconds. However, the evaluation based on the 100×100 exposure grid takes 149 seconds.

For all grids, the deviation between the theoretical number of deaths and the actual number of deaths is less than 30%, which is within the error range of the earthquake disaster evaluation specified by CEA. However, the deviation between the theoretical number of deaths and the actual number of deaths decreases as the number of exposure grids increases. In the 50×50 exposure grid, the deviation between the theoretical number of deaths and the actual number of deaths is 18.3%. In the 2×2 exposure grid, the deviation between the theoretical number of deaths and the actual number of deaths is 22.9%. That is because as the number of grid increases, the grid can express the spatial heterogeneity of exposure data more clearly. However, the evaluation result based on the 100×100 exposure grid is the same as the 50×50 exposure grid because the precision of the exposure data and the evaluation model of people loss are limited. Although the number of grids increases to 100×100, the evaluation results remain unchanged. Overall, from the running efficiency and the results, the evaluation based on the 50×50 exposure grid yields the most satisfactory results, and the estimated number of deaths caused by Yushu earthquake is 2,203.

SD follows the same patterns of change as ND. It is noticed that, based on the 50×50 exposure grid, the deviation between the theoretical number of injuries and the actual number of injuries is 27.4%, which is higher than the deviation in the number of deaths. This difference is obtained possibly because the multiple of SD and ND ranges from 3 to 5. In this study, the multiple is set at 4. However, the majority of the building types of Yushu earthquake are C-type and D-type, which easily collapse in earthquakes. Therefore, a more appropriate multiple is 4.5.

The proposed approach is compared with previous approaches to evaluate the accuracy of people loss as-sessment. Xu et al. [28] used the GIS and the building vulnerability-based approach to evaluate the people loss in Yushu earthquake. They found that approximately 30,000 lives were threatened and thousands of people died. Wu et al. [29] obtained a quantitative evaluation result on the basis of administrative unit-based exposure data and found that deviation between the theoretical loss and the actual loss is 23.3%. By contrast, we obtain an 18.3% deviation between the theoretical loss and the actual loss by using the 50×50 exposure grid. What is more, the evaluation result can be calculated and displayed both in grid unit and in administrative unit. Overall, the experimental results show that the SMHG-based approach is more accurate than the GIS-based approach because the former can more accurately calculate the number of people losses. Our adaptable statistical mechanisms also provide a beneficial basis for rescue-related decision making.

3.5 Discussions

The SMHG-based approach proposed in this study can obtain a satisfactory result for the rapid post-earthquake evaluation of human loss. In general, the evaluation result is better as the number of grids increases. However, it is noticed that the evaluation results remain unchanged when the number of grids is high, and the final evaluation result is different from the actual value. This is mainly because the following two reasons: First, the exposure data used in this study are obtained from the China statistical yearbook for the regional economy. However, these data are based on a county level and thus are not highly precise. Second, the evaluation model of people loss is a national unified model, which is relatively simple and slightly precise, although it is extensively used in China.

High-precision data and fine-grained models are necessary to improve the accuracy of evaluation results. The exposure data used in this study consider the spatial disparity by increasing the number of the exposure grid. This approach can improve the accuracy of the result to a certain extent. However, the data of the exposure grid obtained from the county administrative unit-based map are at a county level. In this situation, increasing the number of the exposure grid is insufficient. Thus, detailed data from administrative departments below a county level are essential. Moreover, updating the data timely and using the latest survey data are also an effective way to improve the data precision.

In this study, fine-grained elliptic attenuation models are used to generate hazard maps, and different provinces employ different elliptic attenuation models. Benefiting from this, the theoretical seismic isoseismal map is highly similar to the actual hazard map. Given fitting the people loss evaluation model needs lots of historical seismic data, at present, there is no special evaluation model for people loss in every PEA. Therefore, a national unified model was used in this study. In future studies, special evaluation models for people loss should be established in data-rich areas, in which evaluation results are expected to be more accurate. The SMHG-based system contains an interface to facilitate the users to add appropriate models, and the interface is convenient for the addition of more appropriate models in the future.

The SMHG-based system is universal and expandable. This system can be easily applied to other regions with similar or different earthquakes. Exposure data, DPM, evaluation model, and evaluation processing are universal in mainland China. Therefore, for a new earthquake, it can rapidly get the value of magnitude, epicenter intensity and location information of the epicenter from CEA. Then, based on the method in Section 2, the hazard map can be quickly generated. Finally, based on the evaluation processing in Section 2.4, the number of deaths and injuries can be rapidly calculated.

Overall, the SMHG-based method is an effective and universal method. Although this method can be further improved in terms of data precision and used model, current assessment results have already met the requirements of CEA. Our evaluation results can provide strong support for post-earthquake rescue-related decisions.

4 Conclusions

A timely and accurate evaluation of people loss is an effective method for emergency rescue responses in the gold 72 hours after an earthquake. This study proposes a SMHG-based post-earthquake evaluation method of people loss in earthquake emergency responses. In SMHG, exposure data are stored as different maps in grid format. Grid format maps are divided into three levels on the basis of evaluation. The hierarchical storage of people loss-related data in grid format can improve data management efficiency. Meanwhile, a spatial information hybrid grid fully considers the uneven distribution of the spatial distribution of exposure data and thus greatly improves the accuracy of the assessment. With SMHG, the statistical results of people loss can be calculated in administrative and grid units. Thus, this study provides a useful basis for disaster emergency rescue responses and emergency evacuation.

Yushu earthquake was selected as a case study to validate the novel method. We used the hybrid grid to manage the people loss-related data. The results showed that the more precision of the grid, the closer the evaluation result near the actual result. Compared with other GIS-based methods, the proposed approach could obtain a more accurate result of people loss. The proposed method could effectively improve the accuracy of people loss estimation.

Acknowledgements

The authors would like to express their appreciation for the supports of the Fundamental Research Funds for the Central Universities, China University of Geosciences (Wuhan) (Grant No. CUGL150822) and China Postdoctoral Science Foundation (Grant No. 2015M582306).

References

[1] Marano K.D., Wald D.J., Allen T.I., Global earthquake casualties due to secondary effects: a quantitative analysis for improving rapid loss analyses. Nat. Hazards, 2010, 52, 319-328.10.1007/s11069-009-9372-5Search in Google Scholar

[2] Sarlis N. V., Skordas E. S., Varotsos P. A., Nagao T., Kamogawa M., Tanaka H., et al. Minimum of the order parameter fluctuations of seismicity before major earthquakes in Japan. P. Natl. Acad. Sci. USA, 2013,110,13734-13738.10.1073/pnas.1312740110Search in Google Scholar PubMed PubMed Central

[3] Sarlis N. V., Skordas E. S., Varotsos P. A., Nagao T., Kamogawa M., and Uyeda S., Spatiotemporal variations of seismicity before major earthquakes in the Japanese area and their relation with the epicentral locations. P. Natl. Acad. Sci. USA, 2015, 112, 986-989.10.1073/pnas.1422893112Search in Google Scholar PubMed PubMed Central

[4] Huang Q.H., Forecasting the epicenter of a future major earthquake. P. Natl. Acad. Sci. USA, 2015,112, 944-945.10.1073/pnas.1423684112Search in Google Scholar PubMed PubMed Central

[5] Varotsos P.A., Sarlis N. V., Skordas E. S., Christopoulos S.-R. G., and Lazaridou-Varotsos M.S., Identifying the occurrence time of an impending mainshock:a very recent case, Earthq. Sci, 2015, 28, 215-222.10.1007/s11589-015-0122-3Search in Google Scholar

[6] Nie G., An J. W., Deng Y., Advances in earthquake emergency disaster service. Seismol. Geol., 2012, 34, 782-791.Search in Google Scholar

[7] Wang S., Liu J., Modeling the risk assessment of landslide-dammed lakes based on the emergency response measures in Wenchuan earthquake, 2008, China. Nat. Hazards, 2013, 67, 523-547.10.1007/s11069-013-0584-3Search in Google Scholar

[8] Chen Q. F., Hongliang M. I., Huang, J. A simplified approach to earthquake risk in mainland China. Pure Appl. Geophys., 2005, 162, 1255-1269.10.1007/3-7643-7375-X_10Search in Google Scholar

[9] Samardjieva E., Badal J., Estimation of the expected number of casualties caused by strong earthquakes. B. Seismol. Soc. Am., 2002, 92, 2310-2322.10.1785/0120010112Search in Google Scholar

[10] Badawy A., Gaber H., Ibrahim H, Earthquake risk assessment of Alexandria, Egypt. J. Seismol., 2015,19,159-170.10.1007/s10950-014-9456-xSearch in Google Scholar

[11] Mebarki A., Boukri M., Laribi A., Farsi M., Belazougui M., Kharchi F., Seismic vulnerability: theory and application to Algerian buildings. J. Seismol., 2014, 18, 331-343.10.1007/s10950-013-9377-0Search in Google Scholar

[12] Panahi M., Rezaie F., Meshkani S. A., Seismic vulnerability assessment of school buildings in Tehran city based on AHP and GIS. Nat. Hazard Earth. Sys., 2014, 14, 969-979.10.5194/nhess-14-969-2014Search in Google Scholar

[13] Armenakis C., Nirupama N., Estimating spatial disaster risk in urban environments. Geomat. Nat. Haz. Risk, 2013, 4, 289-298.10.1080/19475705.2013.818066Search in Google Scholar

[14] Alam M. N., Tesfamariam S., Alam M.S., GIS-based seismic damage estimation: case study for the City of Kelowna, BC. Nat. Hazards Rev., 2013,14, 66-78.10.1061/(ASCE)NH.1527-6996.0000082Search in Google Scholar

[15] Xu J., An J., Nie G., A dasymetric data supported earthquake disaster loss quick assessment method for emergency response in China. Nat. Hazard Earth. Sys., 2015, 3, 1473-1570.10.5194/nhess-16-885-2016Search in Google Scholar

[16] Wan S., Yu Y., Gao A., Yan X., Development of attenuation relations for ground motion in China. Earthq. Res. in China, 2000, 16, 99-106.Search in Google Scholar

[17] Ma Y., Xie L., Methodologies for assessment of earthquake casualty. Earthq. Eng. Eng. Vib., 2000,20,140-147.Search in Google Scholar

[18] Ergonul S., A probabilistic approach for earthquake loss estimation. Struct. Saf, 2005, 27, 309-321.10.1016/j.strusafe.2005.01.001Search in Google Scholar

[19] Yin Z., Zhao Z., Yang S., Relation between vulnerability of buildings and earthquake acceleration. Earthq. Eng. Eng. Vib., 2003, 23,195-200.Search in Google Scholar

[20] Huang H.S., Chen L., Zhu G.Q., Wang L., Lin Y.Z., Wang H.S., Rapid estimation of earthquake loss based on instrumental seismic intensity: design and realization. Earthq. Sci., 2013,26, 67-73.10.1007/s11589-013-0006-3Search in Google Scholar

[21] Yin Z., Classification of structure vulnerability and evaluating earthquake damage from future earthquake. Earthq. Res. in China, 1996, 12, 49-55.Search in Google Scholar

[22] Yang X.H., Rapid loss assessment for earthquake disaster using seismic spatial information grid. PhD thesis, Huazhong University of Science and Technology, Wuhan, 2014.Search in Google Scholar

[23] Chen K. P., McAneneyJ., Blong R., Leigh R., Hunter L., Magill C., Defining area at risk and its effect in catastrophe loss estimation: a dasymetric mapping approach. Appl. Geogr., 2004, 24, 97-117.10.1016/j.apgeog.2004.03.005Search in Google Scholar

[24] Li D., Xiao Z., Zhu Y., Gong J., Research on Grid Division and Encoding of Spatial Information Multi-Grids. Acta Geod. et Cartog. Sin., 2006, 25, 52-57.Search in Google Scholar

[25] Fang Z., Song W., Wu H., A Multi-Grid Model for Evacuation Coupling with the Effects of Fire Products. Fire Technol., 2012, 48, 91-104.10.1007/978-1-4419-9725-8_32Search in Google Scholar

[26] Deng Y., A study on regional difference of earthquake emergency in China. Master thesis, Institute ofGeology, China Earthquake Administration, Beijing, 2003.Search in Google Scholar

[27] National bureau of statistics. China statistical yearbook for regional economy. China Statistics Press, 2010.Search in Google Scholar

[28] Xu G., Yuan Y., Fang W., Shi P., Fast loss assessment of M 7.1 Yushu earthquake. Earthq. Eng. Eng. Vib., 2011,31,114-122.Search in Google Scholar

[29] Wu L., Li Z., Wang Z., Li J., Liu C., Rapid assessment of earthquake disaster: with Yushu earthquake as an example. Sci. Technol. Rev., 2010, 28, 38-43.Search in Google Scholar

Received: 2016-5-5
Accepted: 2016-9-3
Published Online: 2016-12-22
Published in Print: 2016-1-1

© 2016 Xiaohong Yang et al.

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Downloaded on 2.3.2024 from https://www.degruyter.com/document/doi/10.1515/geo-2016-0051/html
Scroll to top button