Developing GIS-based earthquake loss model: a case study of Baqiao District, China

Abstract

Large earthquakes frequently occurred and caused substantial damage and losses in Chinese densely populated urban areas. It is worth clarifying that the existing earthquake loss assessment tools in China do not consider its special characteristics (e.g., tectonics, geology and building inventory). In this paper, an integrated framework for earthquake loss assessment in China is proposed and a geographic information system (GIS)-based system, referred to as China Earthquake Disaster Loss Assessment System, is developed. The individual building is considered as the basic geographical unit and described in three levels (i.e., basic attributes, structural performance parameters, and functional parameters. A three-step data collection method is proposed, and the mobile GIS-based field collection tool is developed. A broad, reasonable, and upgradeable building typology is then presented for the seismic fragility assessment of building blocks. A new economic loss model and casualty model are also presented by considering the characteristics of economics and population. The proposed methodology is implemented in Baqiao District, and the estimation of economic losses and casualties under a scenario earthquake is discussed, which can provide information for urban disaster risk assessment and mitigation.

1 Introduction

Earthquake disasters have been identified as a serious threat to human development, and a challenge to earthquake prevention (Ahmad et al. 2014; Haldar et al. 2013). China, located in the joint of the Circum-Pacific seismic belt and Eurasian seismic zone, is one of the most seismically active regions in the world (Hu 1990). Due to its dynamic tectonic characteristics, China is exposed to the highest seismicity risk. Recently, major earthquakes have occurred and caused severe damage and losses in some densely population urban areas, especially in highly developed zones with structures vulnerable to earthquakes (Coburn and Spence 2003; Haldar et al. 2013; Lin et al. 2012; Xu et al. 2014). For instance, the 2008 Wenchuan earthquake caused 845.1 billion RMB in direct economic losses and more than 87,000 deaths, and the 2010 Yushu earthquake killed more than 2698 people and caused economic losses over 12.5 billion RMB. Considering the previous seismic disasters, the main reason underlying these losses is lack of well-planned preparedness activities in the country, which failed to consider the regional characteristics of structural performance and socio-economic level (Ahmad et al. 2014; Gong et al. 2015). As a vast area with diversity of economy and culture, potential threats in China also increased its challenges for disaster prevention and mitigation. The deficiencies and challenges include the following. (1) The structure type, seismic performance, and quantity of buildings vary from region to region, with high density in southeast coastal cities and low density in northwest cities of China. Masonry and adobe structures with poor seismic resistance performance (e.g., lower fortification level and poor construction without sufficient special details for ductile response), are found primarily in the Midwest rural areas, which have experienced a lower level of economic development (Gong et al. 2015; Wang 2008). (2) The new normal in China is urbanization. Statistics show that at the end of 2014, the mainland urban population was 749.16 million, 54.77% of the total population (NBSPRC 2015). With the high density and uneven distribution of people, the majority of low-income people live in older buildings or self-constructed buildings, mostly without code-compliant seismic design (Gong et al. 2015). The potential consequences including casualties and monetary losses would be immeasurable if earthquakes were to occur.

Earthquake loss assessment in China first began after the 1976 Tangshan earthquake. Loss assessment methodologies were achieved, especially in the early 1990s (Gao et al. 1989; Yin 1995). Earthquake loss assessment exercises mainly contain earthquake damage prediction, earthquake emergency response, and earthquake site disaster loss assessment (Sun et al. 2012). Earthquake damage prediction is mainly undertaken in the pre-earthquake period, providing estimation consequence for policy makers to guide disaster preparedness. As a vital work during earthquakes, earthquake emergency response can provide emergency relief for decision-makers and reduce casualties and economic losses. Earthquake site disaster loss assessment involves field investigation after earthquakes, which can assess the detailed consequent losses and provide guides for reconstruction. All exercises have made a great contribution to the work of earthquake loss assessment, especially in 2008 Wenchuan earthquake. China has experienced many destructive earthquakes. However, the existing earthquake loss assessment in China does not account for its special characteristics, such as the vulnerability of structures and the distribution characteristics of population and economy, which affects the accuracy of earthquake loss assessment.

The main objective of the present research work is to develop an integrated framework by combining the existing methodologies of China. At the same time, some research work is attempted to improve the overall methodology, which includes the following: a new data collection method is devised using a mobile Geographic Information System (GIS) -based field collection tool that can effectively collect detailed building inventory; a hierarchical, detailed, and extensible building typology is proposed, which can consider the differences in structural performance; and a new economic loss model and casualty model are developed for use on individual buildings. Then, the integrated framework is embedded into the China Earthquake Disaster Loss Assessment System (CEDLAS) for use in different assessment phases, i.e., pre-, intra-, and post-earthquake periods. Components of the CEDLAS are discussed here including the seismic hazard analysis, inventory, fragility functions, and loss assessment. Baqiao district is adopted as a case study and the loss assessment is carried out to provide useful information for urban disaster risk assessment and mitigation.

2 Earthquake loss assessment framework

2.1 General framework

The vision of the earthquake loss assessment methodology is to provide emergency relief and rehabilitation guidance for the decision-makers after earthquakes, and also to guide the planning and mitigation of earthquake disasters and earthquake insurance in seismic risk assessment (Federal Emergency Management Agency, FEMA 1997). Various studies have been conducted with earthquake loss assessment methodologies (Choun and Elnashai 2010; Duzgun et al. 2011; FEMA 1997), and also some unique tools for decision-makers to assess the impact have been developed, e.g., HAZUS-MH (FEMA 2012a, b), SELENA (Molina et al. 2010), OpenQuake (Pagani et al. 2014; Silva et al. 2014), MAEViz (2018), and SYNER-G (http://www.vce.at/SYNER-G/). Unfortunately, due to the differences in tectonics, geology, and building inventory, the current methodologies or tools may not apply for China. In order to consider the differences, three key points are summarized as below.

• 1. New and efficient techniques should be developed for acquiring useable building inventories in a large urban areas, an advanced data collection method and procedure are presented.

• 2. In general, building-specific information cannot be easily obtained from the raw inventory data (SVAPG 2013). Therefore, it is necessary to adopt a broad, reasonable, and upgradeable building typology for a large number of buildings.

• 3. The distribution of wealth, population, and external environment can vary in different urban areas. The loss model is a complex, dynamic, and cross process, which should take the different economic status among urban areas into consideration.

By addressing the above issues, this paper presented an integrated, flexible, extensible loss assessment framework which includes seismic hazard analysis, inventory, fragility, physical damage, and loss assessment (direct economic and casualty) as shown in Fig. 1. As it can be seen that modules are interdependent with output of some modules acting as input to others. In seismic hazard analysis module, both China probabilistic seismic hazard assessment (CPSHA) and deterministic seismic hazard analysis (DSHA) methods are used to assess seismic hazard of regional areas. The output of mapping peak ground acceleration (PGA) and spectral acceleration (Sa) can be used as inputs in physical damage. Inventory is a major and essential phase in loss estimation for a group of buildings and is also a collection set of information (e.g. building characteristics, construction practice, occupancy, and other social and economic aspects). In the module, the data are collected by multiple approaches and processed as inputs to others. Fragility quantifies the properties of structures to be damaged by specified ground motions. The fragility functions have been obtained as the basic input to physical damage. The loss assessment module is the combination of the above modules. The direct economic loss and number of casualties can be estimated for policy-makers to guide the planning. The relationship between modules is illustrated in Fig. 2.

Fig. 1

Architecture of the proposed integrated framework

Fig. 2

Methodology of the proposed framework

All components are refined and tailored to proposed GIS-based system CEDLAS to achieve a seamless integration and to arrive at an operational system. The system is established based on C/S architecture with.Net as the development platform and ArcGIS Engine components as the key technology. The source code is written in C# language with the integrated development environment. The technical sides of developing the system can be referred to a companion paper (Long et al. 2020), including plugin management, view definition, and interface design. Core source codes are also provided.

2.2 Uncertainties of earthquake loss assessment framework

Each component in the proposed framework carries associated uncertainties that can be propagated through the analysis (Ahmad et al. 2014). In general, uncertainties in the earthquake loss assessment can be categorized into two groups, i.e., aleatoric and epistemic. Aleatoric uncertainty is related to the inherent randomness of limited data. Epistemic uncertainty is associated with a lack of knowledge and assumptions in modeling techniques, which can generally be reduced with additional information and more comprehensive analysis (Ellingwood and Wen 2005; Jeon 2013). In this paper, the uncertainties of ground motion attenuation models and structural fragility functions are considered. The logic tree method is used to consider the uncertainty of ground motion attenuation models by attaching a weight value for each model. In order to access the earthquake loss, the classical structures are defined. The uncertainties in the earthquake and structure are combined by variety of incremental dynamic analysis (IDA).

3 Components of earthquake loss assessment framework

3.1 Seismic hazard analysis

The main goal of seismic hazard analysis is to estimate ground motions and return period at the site or in a region (Liu et al. 2013). Recent research shows that DSHA and probabilistic seismic hazard analysis (PSHA) are the most frequently used methods (Wang and Taheri 2014). DSHA estimates ground motion from a worst-case earthquake magnitude and location by considering a small number of scenarios (Liu et al. 2013; Wang and Taheri 2014). Conventional DSHA has the advantage of having a clear physical and statistical meaning; however, the temporal characteristics of an earthquake are not considered. PSHA aims to quantify the uncertainties related to earthquake occurrences with respect to time, location, space, and magnitude, and combine them to produce an explicit description of the distribution of future shaking that may occur at a site (Baker 2008; Cornell 1968; McGuire 2008). By considering spatially and temporally inhomogeneous seismic activities in China, a modified PSHA method (CPSHA) was proposed and implemented into the seismic hazard map of China (National Standard of People’s Republic of China, NSPRC 2015). The tri-class seismic source model in CPSHA consists of the seismic province, the background seismicity zone and the tectonic features zone (Pan et al. 2013). In the framework, the CPSHA method is used to determine seismic hazard assessment in regional areas. Mapping of PGA and Sa (e.g., 2% in 50 years and 10% in 50 years) can be described, and the magnitude and location of a potential earthquake (the worst-case earthquake) can be estimated. Furthermore, the DSHA method is used to estimate the seismic hazard analysis of local regions by considering the worst-case earthquakes and site conditions. In addition, it is worth mentioning here that ground motion attenuation model has a significant influence on seismic hazard analysis. In order to consider the geotectonic characteristics of China, the models are established in accordance with the models used in the western United States due to the limitation of ground motion recorders in China. In this paper, the ground motion attenuation models for the New Seismic Hazard Map of China by Yu et al. (2013) and other research on special regions of China (Fan et al. 2011; Yu and Wang 2003) are all integrated into the module.

3.2 Inventory

The accuracy of the estimation results depends on the data used for the analyses, including structure type, occupancy, building height, construction year, population and content value (Bal et al. 2008; Jaiswal et al. 2010; Mansouri et al. 2010; Mansouri and Amini-Hosseini 2014; Sahar et al. 2010). Nevertheless, due to the wide coverage and large number, it is difficult to obtain useable building inventory. Two significant technical challenges are faced here: (1) how to describe the inventory that can be suitable for earthquake loss assessment and (2) how to develop an advanced data collection method that can be used for data acquisition and inventory updates.

In practice, for the sake of efficiency, a building block or city district is mainly used as the basic geographical unit in some risk estimation software tools (FEMA 2012a, b; Molina et al. 2010). However, in reality, buildings consist of collective sets of information, varying widely in characteristics and numbers. In order to depict the inventory in detail, a unique hierarchical description method is proposed here. The description of individual buildings is divided into three levels, as shown in Table 1.

Table 1 Description of individual building levels

As mentioned previously, individual buildings vary widely in characteristics. However, this information is scattered in different sectors (e.g., the civil affairs department, public security department and statistics department) and presents huge data collection challenges. The methods in data collection can be categorized into three groups: census data, ground survey, and field data and remote sensing techniques. As basic datasets, census data represent fundamental inputs for vulnerability assessment (JRC 2013; Ketner 2008). The procedure of a field survey may be divided into two groups: paper forms and hand-held equipment-based forms (Annunziato et al. 2010; FEMA 2002; IDCT 2011). The techniques of GIS, remote sensing, and multi-source imaging are applied for rapid data collection (Wieland et al. 2012; Spence 2011). Based on existing methods, a data collection procedure is proposed with neighborhood community or administrative village as the basic geographical unit to collect and summarize the data. The data collection has three steps (see Fig. 3).

Fig. 3

Description of data collection procedures. From left to right: preparation step, sidewalk collection step, inside data processing step

As shown in Fig. 3, step one is the preparation before data collection, including the acquisition and geo-referencing of high-resolution remote sensing images, communicating with local government leaders, consulting the statistical reports and formulating the collection scheme, and training the collectors. Step two is sidewalk collection using two approaches: paper form-based and mobile GIS-based. The first approach is time-consuming but suitable for non-professional collectors. In order to improve the efficiency and accuracy, a mobile GIS-based field collection system is also applied. The approach can quickly locate the individual building position and efficiently interact with the basic data platform (see Fig. 4). Step three is the inside data processing, including the vector and matching treatment for the first approach results, checking data and uploading to a basic data server. To build an extensible basic data platform, the model combines ArcGIS with Geodatabase, which can integrate spatial data (e.g., administrative divisions, buildings, and roads) and attribute data into the geodatabase model (ESRI 2010).

Fig. 4

Screen shot of the mobile GIS-based field collection system

Some problems arise in data collection. Due to the diversity of information, some parameters cannot be obtained directly, such as floor area, occupants and content value. Multiple approaches are combined and conducted to cope with the issues in our research work. Census data and statistical reports are consulted to obtain the raw data such as functional parameters; in addition, some parameters can be obtained from the existing building design drawings. Under the assistance of local government, various building design drawings are gathered and consulted, which may obtain reliable information. However, a field survey must be conducted due to the limitation of resources. Thus, some equipment is needed in field surveys, such as measuring tape and a laser telemeter. For instance, a data collection method for the floor area and others is shown in Fig. 5.

Fig. 5

Data collection method for the floor area and others

3.3 Fragility functions for structures

Seismic fragility functions describe the probability associated to a given component or building of exceeding predefined damage limit states at the given level of ground shaking. Fragility of buildings implies in macro-seismic intensity scales is commonly used for the earthquake disaster prediction in nation scale. However, when using this method, it is difficult to show the seismic performance of buildings with different qualitative characteristics (such as building height, seismic design code, and construction year). In general, obtaining fragility functions for buildings in Chinese urban areas currently faces two significant technical challenges: (1) proposing a reasonable and upgradeable building typology for a large number of buildings and (2) carrying out fragility analysis for different building typologies.

3.3.1 Building typology

With the requirement of seismic loss estimation, a number of studies in building typology have been conducted in earthquake-prone areas (ATC-13 1985; Brzev et al. 2013; FEMA 1997; Pitilakis and Argyroudis 2014). Several parameters were taken into account, such as the material of construction, building height, structural frame system, and other qualitative characteristics. For instance, HAZUS developed by FEMA considered construction material, load resisting system, and building height (FEMA 2012a, b). GEM (2013) developed a refined taxonomy which considers most general building characteristics (e.g., roof, hazards, and occupancy), and main detailed characteristics (e.g., construction material, structural element, lateral load resisting system, and height). Furthermore, some new building typologies were also employed related to local building stocks (Mansour et al. 2013; Mansouri and Amini-Hosseini 2014). However, the typology cannot fully used for Chinese buildings due to the differences in the seismic design code of buildings, construction standard as well as local characteristics.

Considering the difficulty in obtaining detailed building-specific information from the raw inventory data, it is necessary to propose a broad building typology for China. Based on the Chinese Code for Seismic Design of Buildings (CCSDB) (NSPRC 2010), a hierarchical, detailed, and extensible building typology is proposed as shown in Table 2. Building categorization parameters are described below.

Table 2 Proposed building typology

• 1. Material Type

Based on the construction material, the buildings are generally classified into four types which are wooden structure/masonry/reinforced concrete/steel, respectively.

• 2. Structural framing system

The choices of structural framing system are determined by construction material, building height and seismic design code. The types of structural framing system are divided into five classes, including wall system (stone masonry/adobe/burnt clay brick/block masonry), moment resisting frame, shear wall structure, frames with walls (cast-in-place concrete shear walls/masonry infill walls/un-reinforced masonry infill walls), and braced frame.

• 3. Height level

Building height is a vital factor in determining the natural period of a building and its vulnerability. Various classifications for height level have been proposed in assessment methodology and codes (NSPRC 2010), hence, the height level of buildings are classified into four categories according to the number of stories. They include Low-rise (1–3), Mid-rise (4–9), High-rise (10–33 and < 100 m), Super high-rise (33+ or > 100 m).

• 4. Irregularities

The horizontal and vertical irregularities in buildings have a great influence on the seismic performance of buildings. In general, if a building’s mass center is not consistent with its stiffness center, the torsional action may intensify damage. Horizontal and vertical irregularities used here are defined in chapter 3.4 and 3.5 of CCSDB (NSPRC 2010).

• 5. Code level

The level of seismic design greatly determines the seismic performance of existing buildings (Ahmad et al. 2018). Based on the evolution of CCSDB (NSPRC 1979, 1989, 2001, 2010), building classification is divided into four stages: Pre-code (buildings constructed before 1978 or constructed without seismic code); Low-code (buildings constructed from 1978 to 1989); Moderate-code (buildings constructed from 1989 to 2001); High-code (buildings constructed after than 2001).

• 6. Seismic precautionary intensity (SPI)

The seismic precautionary intensity in buildings is divided into six types as defined in GB 18306-2015 (NSPRC 2015) and GB 50011-2010 (NSPRC 2010). It is evident from past fatal earthquakes around the world that the existence of vulnerable buildings in high intensity areas has caused the highest number of fatalities. This is VI (0.05 g), VII (0.10 g), VII (0.15 g), VIII (0.20 g), VIII (0.30 g), IX (0.40 g).

• 7. Current state of the building

Research shows that the current state of the building would influence the structural damage when earthquake occurred, especially for the case of the aged buildings under low seismic intensity (e.g., VI, VII). Therefore the current state of the building is divided into five types: Buildings without structural damage; Buildings with slight structural damage; Buildings with moderate structural damage and in need of repair; Buildings with severe structural damage and in need of reinforcement; Dangerous buildings.

3.3.2 Fragility analysis

The methods to derive fragility functions include empirical, expert opinion-based, analytical, and hybrid method (Ahmad et al. 2010; Pitilakis and Argyroudis 2014). Due to the aforementioned local characteristic of structures in China, different methods were adopted to create the corresponding fragility functions. Sun and Zhang (2012) used the observed data of buildings damaged by 2008 Wenchuan earthquake to generate the fragility matrix. Empirical methods take advantage of real earthquake disaster and experiments while cannot fully consider important factors including different building attributes and randomness in ground motion due to the limited data gathered so far. Considering these deficiencies, most fragility functions are built following analytical method (Wu et al. 2012; Jiang et al. 2015).

In the present study, the fragility functions of adobe, stone, and unreinforced brick masonry are built based on statistics of observed damage from past earthquakes and experiments (Sun and Zhang 2012; Zhang 2010). For other kind of building typology, numerical simulations are performed to derive the fragility functions. Firstly, multiple index buildings that can represent the class of proposed building typology are designed and numerical modeled. A set of 22 far field ground motion records are selected from PEER NGA strong ground motions database as recommended by a series reports (FEMA P-695 2009; D’Ayala et al. 2014) and a set of 10 ground motion records are selected from China Earthquake Networks Center (CENC, http://data.earthquake.cn). The details of Chinese ground motion set are summarized in Table 3. Incremental dynamic analyses (IDA) are carried out, where each record is scaled up until the collapse of the building occurs. According to the results of IDA, the correlation coefficients between the ground motion intensity measures (IM) and the structure damage measure (DM) were calculated. Then the fragility function can be obtained as follow.

Table 3 Ground motion records selected from China Earthquake Networks Center

In the above procedures, two main issues related to the fragility curves are selections of an appropriate seismic IM as well as a structure DM. An optimal IM can characterize the strong ground motion and even correlate with the structural response (NGI 2013). Due to the majority and diversity of buildings in urban areas, it is difficult to select a unified structure-specific IM. The structure-independent IMs should be considered such as peak ground acceleration (PGA) and peak ground velocity (PGV). In addition, PGA is also consistent with the ground motion attenuation models of China. Therefore, PGA is adopted to develop the fragility curves. As suggested by CCSDB (NSPRC 2010), the maximum peak inter-story drift over the height of building, denoted θ_max, is chosen as the damage measure to reflect the damage degree of the components as well as the structure. This parameter is suitable for the urban-scale seismic fragility estimation since it can be widely retrieved from both the experimental observations and numerical simulations. The damage states are divided as none damage (ND), slight damage (SD), moderate damage (MD), extensive damage (ED), and complete damage (CD) following the definitions provided in code of Classification of earthquake damage to buildings and special structures (NSPRC 2009).

3.4 Loss assessment

Seismic risk is represented by the probabilistic physical damage of an individual building while considering local hazard, vulnerability, and inventory. Based on the physical damage results, both direct economic losses and number of casualties are calculated.

With the rapid urbanization of China, wealth is mostly centered in urban regions. The affiliated value of a building is more apparent, which represents a large proportion of its replacement costs. To assess economic loss more precisely, a direct economic loss model is proposed based on the model of Sun and Chen (2009), which considers the structure loss, decoration loss, and content loss. In addition, the differences of economic level in urban areas are considered. The proposed direct economic loss model is developed as follows:

$$L_{{{\text{total}}}} = \alpha (L_{S} + L_{D} + L_{C} )$$

(1)

$$L_{S} = \sum\limits_{i = 1}^{n} {\sum\limits_{j = 1}^{5} {M_{i} \cdot P(DS = ds_{j} ) \cdot \mu_{{ds_{j} }}^{s} } }$$

(2)

$$L_{D} = \gamma_{1} \cdot \gamma_{2} \cdot \gamma_{3} \sum\limits_{i = 1}^{n} {\sum\limits_{j = 1}^{5} {\eta_{1} \cdot M_{i} \cdot P(DS = ds_{j} ) \cdot \mu_{{ds_{j} }}^{d} } }$$

(3)

$$L_{C} = \sum\limits_{i = 1}^{n} {\sum\limits_{j = 1}^{5} {\eta_{2} \cdot M_{i} \cdot P(DS = ds_{j} ) \cdot \mu_{{ds_{j} }}^{c} } }$$

(4)

where L_total is the expected total direct economic loss for any geographical unit; α is the loss multiplier accounting for the ignored losses (e.g., natural environment loss), suggested as 1.0 to 1.3; L_S, L_D, and L_C are the structure loss, decoration loss, and content loss, respectively; $P(DS = ds_{j} )$ is the probability for damage state ds_j to be incurred on an individual building in an earthquake; M_i is the replacement cost for structure i, which is equal to the product of replacement cost per floor area and total floor area; $\mu_{{ds_{j} }}^{s}$, $\mu_{{ds_{j} }}^{d}$ and $\mu_{{ds_{j} }}^{c}$ are the loss ratios for damage state ds_j on the structure, decoration and content, respectively; γ₁, γ₂ and γ₃ are the decoration loss correction factors considering the economic developmental variety, building occupancy, and decoration level, respectively; η_i is the ratio of decoration cost to the structure replacement cost; andη₂ is the ratio of the content cost to the structure replacement cost.

As the result of observations of past major historical earthquakes in China, several casualty models have been developed considering seismic intensity, earthquake time, and population density parameters. However, the uneven distribution of population caused by the occupancy levels and economic status is not considered in these models. By considering damage probability, occupancy, population density, structure type and earthquake time, a new casualty model is proposed that can estimate the number of deaths, severe injuries and minor injuries. As described in Eq. 5.

$$N_{s,i} = \sum\limits_{j = 1}^{n} {N_{j} = \alpha \cdot \gamma } \cdot \eta \cdot N_{0} \sum\limits_{j = 1}^{n} {A_{j} } P_{j}$$

(5)

where N_s,i is the total number of casualties in severity level i, i = 1 to 3, where 1—minor injury, 2—severe injury, and 3—death; N_j is the number of individual building j in severity level i; α is the regional multiplier accounting for the population density variation between the different regions; γ is population density correction factor in different occupancy levels; η is the people indoor rate in different occupancy levels at different earthquake times; N₀ is the number of people per floor area; A is the total floor area; and P_j is the mean damage ratio (e.g., death ratio is given as Eq. 6), which relates to the damage state and the casualty ratio in different damage states.

$$P_{j} = P_{l = 1} P_{d = 1} + P_{l = 2} P_{d = 2} + P_{l = 3} P_{d = 3} + P_{l = 4} P_{d = 4} + P_{l = 5} P_{d = 5}$$

(6)

where P_l is the probabilities in damage state l; l = 1 to 5, which represents none, slight, moderate, extensive, and complete, respectively. P_d is casualty rate in damage state d; d = 1 to 5, with the same values for 1 to 5 as previously outlined for l.

4 Application and discussion

4.1 General description

Xi’an is an earthquake-prone region due to its tectonic setting. According to the latest Seismic ground motion parameters zonation map of China (NSPRC 2015), Xi’an is located in seismic fortification intensity zone VIII (0.20 g). Baqiao, with a population of 500,000, is one of the oldest districts in Xi’an, Shaanxi province, which is located in the middle of Guanzhong basin. With an area of 332 km², the district is divided into nine sub-district regions (including 33 communities and 226 administrative villages) as shown in Fig. 6. Baqiao is an old industrial district known for its textile industry. Its buildings are typically composed of adobe and brick masonry. However, with the development of urbanization and the recession of the textile industry, some older buildings have been demolished and replaced by new reinforced concrete shear wall buildings, especially industrial buildings.

Fig. 6

Overview of Baqiao District. The district is divided into nine sub-district regions: Xinhe, Xinzhu, Baqiao, Shilipu, Fangzhicheng, Xiwang, Hongqing, Hongqi, and Dizhai

4.2 Structure classification

Considering the efficiency of field survey, a detailed survey and spot check of buildings were separately carried out in Baqiao. The detailed survey, considered the differences in individual buildings, was conducted in the urban and urban–rural areas. Information about structure type, condition, construction year, occupancy, floor area, and other parameters can be obtained from this survey. Spot checks, however, were used for the buildings with only minor differences (e.g., for rural areas). The parameters of construction year, height, occupancy, structure type, and regional difference were considered in the spot check. 61,625 buildings were checked in Baqiao with 28,969 building data collected in total. The entire data field work lasted for more than 10 months. According to structure type, construction year, occupancy, and number of stories, the statistical results of different building distribution are shown in Fig. 7.

Fig. 7

Distribution of buildings in detailed survey area. From left to right and top to bottom: a structure type b construction year c occupancy, and d number of stories

As shown in Fig. 7a, buildings in the case study area are mainly composed of reinforced brick masonry (69.0%) and unreinforced brick masonry (25.1%). Bottom frame structures are usually used for commercial due to large space at the bottom. Considering the relatively high cost of construction, RC frame structures and RC shear wall structures are rarely constructed compared to reinforced brick masonry buildings. With the development of urbanization, RC shear wall structures and RC frame with cast-in-place concrete shear walls structures are commonly used for high-rise buildings. According to the year of construction, 82% of surveyed buildings were constructed after 1990 following the introduction of earthquake-resistant design principles in 1989 CCSDB (NSPRC 1989). Since 1990, reinforced brick masonry buildings have rapidly expanded in China; at the same time, unreinforced brick masonry stocks have been restricted in cities, which largely limits such construction to rural and undeveloped areas (Fig. 7b). Statistics from the survey show that the buildings are predominately built for residential use, with only a few for commercial (0.38%) and education (0.37%) (Fig. 7c). As seen from the Fig. 7d, the buildings are almost low-rise (85.5%) and mid-rise (14.1%). The number of RC frame structures are mainly from 4 to 9, and the number distribution of RC shear wall are all above 12 in the range of high-rise.

According to the field survey data, the buildings are classified into six types based on construction material and structural system. Reinforced brick masonry (RBM), unreinforced brick masonry (UBM), masonry with bottom frame (MBF), concrete moment frame (C1), and concrete shear wall structures (C3) are built in the detailed survey area, and adobe (A) and unreinforced brick masonry (UBM) are in the spot check area. A qualitative description of building features is provided in the following and the photos of typical building types are shown in Table 4.

Table 4 Characteristics of building typology in Baqiao district

Adobe masonry (A) Adobe masonry is built with sun-dried earth brick in mud mortar or constructed mud walls with horizontal wood elements. Considering the advantages of easily availability material, simplified construction, and low cost, adobe masonry structures are widely constructed in rural areas of China. Unfortunately, those structures are highly vulnerable in earthquakes. Statistics show that adobe structures are severely damaged in earthquake intensity VII, and completely damaged and collapsed in earthquake intensity above VIII. Adobe structures in Baqiao were typically built in the undeveloped administrative villages or neighborhood communities before 1980 and are typically self-constructed with wooden roofs (Table 4). The characteristics of adobe are as follows: (1) the adobe is constructed with low strength and viscosity of mud; (2) many cracks can be seen in adobe walls because of the uneven foundation settlement; (3) the configurations of adobe are mostly arbitrary and unreasonable especially in the self-constructed buildings, which could lead to out-of-plane failure or torsional damage.

Unreinforced brick masonry (UBM) Unreinforced brick masonry buildings (UBM) are constructed without appropriate seismic measures (i.e., ring beams and structural columns) as shown in Table 4. The UBM buildings in Baqiao are mainly located in administrative villages and are almost exclusively low rise (1–3 stories). The survey indicates that the overall seismic performance of the UBM structure is poor due to the lack of effective seismic measures. Also, the floors are precast concrete slabs, and the roofs are mostly wooden structures. The phenomenon of foundation settlement and wall-cracks increase the potential risk to seismic effects as well. These buildings were mainly constructed in the 1960s without seismic retrofitting. Even new self-constructed buildings ignore the current seismic design codes.

Reinforced brick masonry (RBM) Different from adobe and UBM buildings, the ring beams and structural concrete columns can be observed in the RBM buildings. In Baqiao, buildings constructed after 1990 are typically with three to six stories and are structurally sound and follow good construction practice (Table 4). In addition, after the Tangshan earthquake in 1976, many unreinforced brick masonry buildings constructed between 1950 and 1970 with no seismic design code were retrofitted with the ring beam and structural concrete columns. After retrofitting, these buildings are considered as RBMs.

Masonry with bottom frame (MBF) Masonry with bottom frame is a special structure type with a reinforced concrete frame for the first floor and a masonry or concrete block structure at upper floors (Gong et al. 2015). The framing system of MBF consists of moment resisting frame and brick wall system. Because the material and mechanical properties are different, the seismic performance of the MBF is more complex. Many MBF structures can be founded in Baqiao including multi-storey residential, hotel, and office buildings with first floor restaurants (Table 4).

RC frame structures (C1) Most RC frame structures in Baqiao are designed and constructed following national seismic design codes, and mainly found in the developed, high population density areas due to its relatively high costs in construction. RC frame structures are designed to have the failure mechanism of strong-column and weak beam, and they are expected to have good seismic performance. These buildings are used for commercial and administration. The buildings are typically four to nine stories high and were constructed after 1990 (Table 4).

Concrete shear wall structures (C3) A concrete shear wall structure is a special structure that adopts the reinforced concrete walls to resist vertical and horizontal seismic force. Due to the perfect seismic performance with large lateral stiffness, these structures are commonly used for high-rise buildings. In recent years, with the development of urbanization in Baqiao, some older buildings have been demolished and replaced by new reinforced concrete shear wall buildings. Located along the Canhe River, these high-rise/super-high-rise buildings are mainly used for residential and commercial (Table 4).

4.3 Earthquake loss assessment

4.3.1 Seismic hazard analysis

The classical deterministic seismic hazard analysis is carried out based on the historical earthquake. 1556 Huaxian Ms 8.0 earthquake (located at 109.7E Longitude and 34.5 N Latitude in Shaanxi province) are used as scenario earthquake. The earthquake caused huge destruction, and killed more than 830,000 people (Lv et al. 2014). The main seismogenic structure includes Huashan piedmont fault, Weihe fault, and Weinan front fault. The seismic intensity of meizoseismal areas was XI as shown in Fig. 8.

Fig. 8

Earthquake intensity map and seismogenic faults for the Huaxian earthquake from Lv et al. (2014)

Baqiao is located in the southwest of Huaxian city, approximately 70 km from the epicenter. Historical statistics show that the Huaxian earthquake caused a seismic intensity of IX in Baqiao. Seismic hazard analysis is conducted based on the ground motion attenuation model and the characteristics of site amplification. The former has obvious regional characteristics depending on the properties of source, propagation medium, and site condition, etc. (Hu 1990). The present study used the ground motion attenuation model developed by Yu and Wang (2003), and the equations for major axis and minor axis are shown respectively as follow:

$${ }\lg Y = 2.206 + 0.532M - 1.954{\text{lg}}\left( {R + 2.018e^{0.406M} } \right)$$

(7)

$${ }\lg Y = 1.010 + 0.501M - 1{\text{.441lg}}\left( {R + 0.340e^{0.521M} } \right)$$

(8)

where Y represents the horizontal component of PGA in g; M represents the surface wave magnitude; and R represents the hypocentral distance (unit: km).

The impact of site conditions on the propagation of seismic waves is a complicated problem, which manifests as a strong amplification or reduction effect on different frequency components of ground motions as well as the PGA (Konakli and Der Kiureghian 2012). This impact is considered by adjusting PGA with the site amplification factor, k_sp. The site amplification model proposed by Chen and Duan (2013) is utilized because the model is built based on the borehole data in Xi’an area. The model is shown as follow:

$${ }k_{sp} = \exp (1.660 -0.412\lg H - 0.368\lg V_{s} )$$

(9)

where H represents the soil thickness (unit: m) and V_s represents the mean value of shear wave velocity (unit: m/s). Both the parameters can be obtained from the geological survey report from local government (Xi’an Geophysical Exploration Center 2011).

The seismic hazard analysis is carried out using the above equations and the theory in Sect. 3.1, which are integrated in the tool—CEDLAS. Considering the differences in the site conditions of the Baqiao district, the whole area has been resampled on a uniform grid of 0.5 km × 0.5 km, which results in a database of 92,597 geocells. The results of PGA map of Baqiao are shown in Fig. 9. The value of PGA from 0.3 to 0.5 g as shown in Fig. 9 is in accordance with the intensity distribution given by Seismic ground motion parameters zonation map of China (NSPRC 2015).

Fig. 9

The peak acceleration map of Baqiao calculated using Huaxian earthquake

4.3.2 Fragility functions for case study structures

As mentioned in Sect. 4.2, the buildings in Baqiao are classified by structure type, construction year, occupancy, and number of stories. Since seismic precautionary intensity of Baqiao is VIII (0.20 g), the fragility curves are determined considering only the structure type, number of stories, and seismic design code. The fragility functions for the case study structures can be derived using the methodology proposed in Sect. 3.3.2 and the results are shown in Fig. 10.

Fig. 10

Fragility functions for building typologies in Baqiao. From left to right and top to bottom: a A, Low-rise, Pre-code, b UBM, Low-rise, Pre-code, c RBM, Low-rise, All codes, d RBM, Mid-rise, All codes, e C1, Low-rise, All codes, f C1, Mid-rise, All codes, g C1, High-rise, High-code, h C3, High-rise, High-code. SD-slight damage, MD-moderate damage, ED-extensive damage, CD-complete damage

4.3.3 Damage analysis

The damage analysis for the case study area under the 1556 Huaxian earthquake is investigated. The seismic peak ground acceleration distribution, building inventory, and fragility curves are integrated as described in Fig. 11. The probability for different damage states to be incurred on individual buildings in Huaxian earthquake can be seen in the bar chart below (Fig. 11). The detailed percentages of five failure states for different building typologies are calculated as shown in Table 4. Meanwhile, according to the structural engineering reconnaissance of past earthquakes (Sun and Zhang 2012; Gong et al. 2015), typical seismic damage of corresponding buildings are summarized in Table 4. It appears that adobe structures have the most severe damage as reflected by their highest percentage of complete damage (26.92%) and extensive damage (39.57%). Comparing with unreinforced brick masonry buildings, the seismic performance of reinforced brick masonry buildings raised obviously due to the mandated use of special details. The seismic performance of MBF is similar to RBM. It’s noted that the ratios of complete damage of RC frame structures and shear wall structures are both low, especially the latter ones because they all built following the latest two versions of seismic codes. This manifests that current seismic codes are effective in preventing structural collapse in strong earthquakes.

Fig. 11

Probability of structure damage in bar chart

4.3.4 Loss assessment

According to the construction industry statistical yearbook in China from 2004 to 2013, a reasonable building replacement cost using the construction completion value and floor area in different year throughout all the provinces is obtained. Due to the complex tectonic with collapsible loess covered in Baqiao, the loss multiplier α is 1.1 accounting for the natural environment loss. The direct economic loss (e.g., structure loss, decoration loss, content loss, and total loss) can be seen in Fig. 12. It is noted that Fig. 12 is based on the results of average predictions. The corresponding upper and lower bounds of economic losses are shown in Table 5 to illustrate the possible range of losses in Baqiao district.

Fig. 12

Direct economic loss assessment in Baqiao at neighborhood community and administrative village level. From left to right and top to bottom: a structure loss, b decoration loss, c content loss, and d direct economic loss (in million RMB)

Table 5 The range of economic losses in Baqiao (in billion RMB)

As shown in Fig. 12a, most structure losses occurred in neighborhood communities and administrative villages in Fangzhicheng and Xiwang. The buildings, with high occupancy are mostly older brick masonry. After the 1976 Tangshan earthquake, most of UBM buildings were retrofitted by setting ring beams and structural columns. However, there are still some unreinforced brick masonry and self-constructed buildings found in this area, especially in Xiwang. The decoration loss can be seen from Fig. 12b, with the higher losses concentrated in the Fangzhicheng sub-district region. The economic characteristic is uneven due to the variation in decoration factor. High and moderate decoration factors are found in the densely populated Fangzhicheng sub-district regions, while low decoration factors are found in undeveloped areas. In general, the content losses are directly related to the economics of the region. However, the field survey shows that in some areas, the buildings with poor seismic resistance performance have high content values, which may increase in the direct economic losses (Fig. 12c). The total direct economic losses are shown in Fig. 12d, which are the sum of structure losses, decoration losses, and content losses as shown in Fig. 12d. The higher losses are mostly concentrated in Fangzhicheng, Hongqi, Xiwang, and Baqiao sub-district regions, where the proportions of losses caused by decoration and content damage are larger than these proportions in other regions.

The population distribution in Baqiao is highly uneven, with the greatest population concentrated in the Fangzhicheng and Shilipu sub-district regions. The regional multiplier α is set to 0.8 in view of the population density variation among different regions. The number of people per floor area N₀ and population density correction factor γ in different occupancy classes are carried out based on the census data and questionnaire survey as shown in Table 6.

Table 6 The number of people per floor area N₀ and population density correction factor γ in different occupancy classes

In order to obtain the occupancy rate in different types of buildings at different times of a day, a questionnaire survey was conducted among Baqiao residents. The analysis from 1200 questionnaires is shown in Table 7.

Table 7 The occupancy rate of Baqiao in different occupancy classes at different times

Based on the proposed casualty model (Eq. 5) and above parameters, the average casualties are calculated as shown in Fig. 13. Detailed numbers of casualties are summarized in Table 8 including the upper and lower bounds of predictions. The majority of casualties occurred in large, populated areas, such as Fangzhicheng, Shilipu, and Xiwang sub-district regions. The scenario earthquake happens at night (22:00–07:00) most of the residents would be sleeping, and the number of casualties would be higher than if the earthquake occurred during the day. The ground motion intensity decreased from northeast to southwest (Fig. 9). In the northeast, most areas are sparsely populated farmland and mountains, thus resulting in relatively fewer casualties. However, the southwest area is more densely populated with many unreinforced brick masonry buildings, thereby resulting in large majority of casualties. More casualties are found in the Fangzhicheng, Xiwang, and Hongqing sub-district regions which are similar to higher economic loss concentrated regions.

Fig. 13

The distribution of casualty estimation in Baqiao at neighborhood community and administrative village level. From left to right and top to bottom: number of minor injury, number of severe injury, and number of death

Table 8 The range of casualties in Baqiao

According to the estimation results of earthquake loss assessment in Baqiao, the losses are mainly distributed in Fangzhicheng and Xiwang sub-district regions. These regions have a higher economic level and a high building and population density. However, in some areas, buildings are constructed in close proximity and do not meet the seismic design code. Therefore, when an earthquake occurs, these buildings can also be damaged and destroyed by fire and debris. These secondary disasters greatly affect rescue activities. In other words, the method may underestimate the resulting earthquake disaster losses without considering secondary disasters. It is noted that the predicted range of casualties varies significantly, especially for the number of death. This is because there are many factors affecting casualties, and the uncertainty of each factor is relatively large due to incomplete statistical data.

4.4 Discussions

4.4.1 Discussion on the results of loss assessment

The overall loss assessment results of Baqiao district show that the socio-economic impacts are high due to the seismic prone characteristic and amount of pre-code buildings. It is noted that the government is pushing the program of retrofitting UBM buildings. To demonstrate if retrofitting may help the region in risk mitigation, all UBM buildings are replaced with RBM buildings in the CEDLAS, and another simulation is carried out. Two simulation results of loss assessment are archived in Appendix Table A for comparing the effect of retrofitting.

As presented in Table A, a slight decrease (1.8%) of direct economic loss is observed in Baqiao district after retrofitting. It is noted that the economic impact is not as obvious as expected. This mainly because the structure, decoration, and content costs of UBM are higher than those of RBM in the case of the same floor area. Meanwhile, the structural damage is certainly less after retrofitting, resulting in a mild decrease in economic loss. However, the estimated casualties are much lower after retrofitting. In Hongqing sub-district region, the number of death decreases from 142 before retrofitting to 71 after retrofitting. Despite the accuracy of the presented method to show the impacts of retrofitting UBM can be improved, the results are enough to illustrate the urgency and effectiveness of implementing retrofitting programs.

4.4.2 Discussion on the limits of proposed earthquake loss model

The proposed methodology is used for the assessment of urban-scale earthquake losses in China. However, to avoid the expensive computational cost, some effects that may have impacts on the final results of losses are simplified considered and discussed in the following. (1) In the procedure of seismic hazard analysis, site effects are included using an equation to amplify the parameter of ground motion. Recently, more ground shaking modelling techniques have been developed to directly reflect the spatial variability in ground motions. For example, Xiong et al. (2020) considered a coupled site-city effect for seismic damage simulation on a region-scale. (2) In the procedure of seismic fragility analysis, different building typologies are included while the corresponding numerical modeling techniques can still be improved for some typical buildings. For example, in-plane and out-of-plane behavior would be coupled under multi-directional seismic action (Ahmad et al. 2012; Al Hanoun et al. 2019) while only the in-plane mechanism is considered in the modeling of masonry walls.

5 Conclusions and future developments

This paper proposes an integrated framework for earthquake loss assessment in China. This new assessment methodology can account for Chinese special characteristics. Several key components are included: a new data collection method is devised using a mobile GIS-based field collection tool that can effectively collect detailed building inventory; a hierarchical, detailed, and extensible building typology was proposed to assess the fragility of buildings based on 2010 CCSDB; and a new economic loss model and casualties model are developed for assessindividual buildings to estimate the loss and number of casualties. Baqiao district is used as a case study and the economic losses and number of casualties are estimated. The results of retrofitting of UBM buildings in Baqiao district are presented. The case study can provide useful information for urban disaster risk assessment and mitigation. Using the proposed methodology, it is possible to accomplish the following: (1) scientifically evaluate the degree of damage to buildings in future earthquakes and reduce the socio-economic losses, (2) suggest mitigation measures for decision-makers, and (3) assist insurance departments in estimating insurance rates.

This earthquake loss assessment framework is a complex project that integrates the components of seismic hazard, inventory, fragility, and loss assessment. Based on the methodologies, the stand-alone system CEDLAS is developed in ArcGIS Engine and the “.NET” framework environment. Because some functions of the system are still under testing, the system is currently not publicly available. With successful development, it not only provides a tool for earthquake loss assessment, but also integrates the latest research findings, accurate data and new methodologies for risk management and loss assessment. Further development will primarily concentrate on the following aspects: (1) implementing recovery decision making by considering reinforcement methods and cost–benefit analysis, (2) improving the accuracy and efficiency of data collection, (3) improving the function module and testing the stability of overall system, and (4) integrating the latest research findings and new methodologies into the tool CEDLAS to improve the applicability to other countries.

Appendix

See Table

Table 9 Comparisons of loss assessment in sub-districts of Baqiao

9.