Assessment of ripple effect and spatial heterogeneity of total losses in the capital of China after a great catastrophe shocks

The total losses caused by natural disaster have spatial heterogeneity due to different economic development level inside the disaster-hit areas. This paper set the scenarios of direct economic loss to introduce the sectors’ loss caused by 2008 15 Wenchuan earthquake into Beijing, utilized Adaptive Regional Input-Output (ARIO) model and Inter-regional ripple effect (IRRE) model. The purpose is to assess the ripple effects of indirect economic loss and spatial heterogeneity of both direct and indirect economic loss at the scale of smallest administrative divisions of China: streets/ (villages and towns). The results indicate that the district of Beijing with the most severe indirect economic loss is Chaoyang district; Finance & Insurance industry (#15) of Chaowai Street suffers the most in Chaoyang district, which is 1.46 times of its direct economic loss. During 2


Introduction
Economic losses caused by frequent natural disasters have been increased dramatically, and pose serious challenges to the world sustainable development and human safety (Munich, 2002). In April 2016, earthquakes with M_S>7 occurred within 8 days in succession in Japan(M_S7.3), Ecuador (M_S7.8), Burma(M_S7.2), and Afghanistan (M_S7.1), which caused large total economic losses to the above countries. The large economic losses caused by natural disasters should be assessed more 5 accurately for improving the awareness of disaster impacts and the higher ability of disaster prevention and mitigation.
The Total economic losses caused by disasters usually incorporate the direct economic loss (Algermissen S.T., 1984;Lijun S, 1998;Jiren L, 2003;Ming T, 2014) and the indirect economic loss. For the indirect economic loss, it is generated by business interruption, imbalance between supply and demand, disorder between forward output and backward supply of sectors due to physical damage (Cochrane, 1997;FEMA, 2001). Its property of "invisible loss" makes it difficult to 10 evaluate, but it cannot be ignored. Because indirect economic loss can increase, even outnumber direct economic loss along with the economic development, which underscore the significance of capturing the ripple effects accumulated along inter-regional and interindustrial linkages. The Input-Output (IO) model Wu et al., 2012;Hallegatte, 2014;Zhou et al., 2014;Xia et al., 2016) and Computable General Equilibrium (CGE) model (Rose and Liao, 2005;Guivarch et al., 2009;Xie et al., 2013;Rose, 2015) are two representative models which are commonly used to assess indirect economic 15 loss.
However, most of the loss assessment methods only consider the overall loss value of the region (Xie et al., 2012), the losses in different regions inside the disaster-hit area have not yet been quantified. In modern city, the functional zoning have its own economic characteristics because of the industrial distribution, so there may be only one or two developed economic sectors existing in each functional zoning. If a catastrophe occurs, the direct economic losses of sectors will not evenly 20 distributed inside the disaster-hit area due to the range of disaster impacts (Figure 1). In addition, as a result of the industrial linkage, the ripple effect of indirect economic loss can make a sector's loss spread to other related sectors in other region (e.g. the production reduction of an automobile manufacturing sector increases the unintended inventory (or decreases production ) of steel sector in other region, eventually making its indirect economic loss increasing)( Figure 1). Therefore, there are Nat. Hazards Earth Syst. Sci. Discuss., doi:10.5194/nhess-2016-354, 2016 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Published: 14 November 2016 c Author(s) 2016. CC-BY 3.0 License. obvious spatial heterogeneity in direct and indirect economic losses of sectors inside the disaster-hit area. The distribution of total losses are inconsistent in two disaster-hit areas with the similar economic development degree (or whole total losses).  hits, both suffers DEL of USD 10 billion, but the losses inside areas are different due to functional zoning and disaster location. For ripple effect of IEL, the indirect economic losses is not only impacted by the distance of disaster occurring, but is impacted by inter-regional and interindustrial linkages) Much attention has been paid to assessing the total loss of the whole disaster-hit area, but further study on refined assessment of direct economic loss and ripple effect of indirect economic loss inside the area is crucial to revealing the 10 comprehensive disaster impacts, though internal economic data is difficult to acquire and the assessment models are rare.
The results of the above research will play a significant role in different aspects. For governments, the policy maker can make more refined post-disaster recovery and reconstruction policies, and give the sector's priority to recovery with closer industrial linkage to make the economic system recover to pre-disaster level faster. For the insurance firms, they can add insurance categories to increase profits, also enhance the economic resilience of economic system. For the public, people can fully aware 15 of the long-term impact of disaster and thus may change their economic activities (e.g. real estate investment).
2008 Wenchuan earthquake (2008 WCE) is designated as the catastrophe in this paper because it is the most destructive natural disasters since the founding of China in 1949. Beijing (BJ), one of the most important and developed metropolises in China, is chosen to be the disaster-hit area due to its location which located at the North China seismic belt (Figure 2), and Nat. Hazards Earth Syst. Sci. Discuss., doi:10.5194/nhess-2016-354, 2016 Manuscript under review for journal Nat. This paper is the first to assess and analyse the ripple effects of indirect economic loss and spatial heterogeneity of both direct and indirect economic loss inside the disaster-hit area at the scale of the smallest administrative divisions in China: streets (villages and towns) by means of the Adaptive Regional Input-Output (ARIO) model (Hallegatte, 2008) and 5 Inter-regional ripple effect (IRRE) model. This paper aims to decrease the long-term economic impacts caused by disasters and provide more comprehensive information for government to make better post-disaster recovery, disaster prevention and mitigation, rescue measures and funds allocation policies to the regions which may suffer seriously damage in the disaster aftermath.

Study area
The 2008 Wenchuan earthquake (2008 WCE) is chosen in this paper as the catastrophe because it's an earthquake with the greatest magnitude and the highest damage degree since the founding of China in 1949, which affected 417 counties of 16 provinces and cities, with 69,226 people dead, and direct economic loss of USD 124 billion (the exchange rate of CNY to USD is 0.14 in 2008). According to all the affected provinces, Sichuan Province (SCP) suffered most serious disaster with 15 the direct economic loss of USD 104.9 billion, accounting for 89% of the total losses (NCDR & MOST, 2008). The Wu's study (Wu et al., 2012) result shows that, indirect economic loss of SCP caused by the earthquake was about USD 42.1 billion, which was 40% of its direct economic loss. The potential impact of the disaster on indirect economic loss has aroused wide concern from both government and scholars (Rose et al., 1997;Okuyama, 2007;Li et al., 2013).

Data
The data used in this study are mainly divided into three parts:

10
(1) Direct economic loss of BJ: The loss ratio of all sectors of SCP is taken as the scenario for direct economic loss of BJ.
The loss of SCP comes from National Disaster Reduction Committee-Ministry of Science and Technology, with a total of 17 sectors (NCDR and MOST, 2008). The specific calculation method is shown in the section 3.1.

10
The reason for setting this scenario is that there is no direct economic loss of 17 sectors of Beijing (BJ) due to no real disaster occurring. In order to solve this problem, The Scenario is set that: if an earthquake with the same intensity of 2008 Wenchuan earthquake (2008 WCE) occurs in BJ, and BJ has the same direct economic loss ratio of 17 sectors with that of Sichuan province (SCP). Before setting up Scenario, the economic conditions of two areas should be understood.
(1) Comparison of development between BJ and SCP

15
In terms of social and economic development degrees, SCP was categorized as a less developed region, its GDP was USD 175.1 billion (calculated based on 2008 constant price, the same below). BJ was categorized as a developed region, its GDP in 2008 was USD 146.8 billion, almost equivalent to the GDP of the whole province of SCP. Meanwhile, in 2008, per Capita GDP of BJ was 4.31times of SCP, the fixed asset stocks was 2.09 times. The proportion of the tertiary industry of BJ reached 73 percent. So, the developed secondary industry and tertiary industry as well as closer industrial linkage may lead to larger 20 indirect economic loss is worthy of attention.
The core thought of scenario of direct economic loss is that both areas have the same sector loss rates. The core formula of the scenario is calculated as followed: Where, DEL (SCP (or BJ), n, 2008) is the direct economic losses of n th sector of SCP (or BJ) after catastrophe. CAP (SCP (or BJ), n, 2008) is fixed asset stock of n th sector of SCP (or BJ) in 2008, The fixed assets stocks were calculated by the 5 perpetual inventory method (Hall and Jones, 1999) according to the total investment in fixed assets of the whole society.
This method can visually reflect the shock of 2008 WCE on BJ under the scenario of same sector's loss ratio, and can reduce the calculation uncertainty of direct economic loss caused by different economic development degrees between two regions. The results are illustrated in the Table 1. It can be found that, economic pattern of BJ was mainly dominated by the tertiary industry, while SCP was dominated by the secondary industry, and it was in an industrial transformation period. (3) Setting up damage intensity scenarios of BJ It can be seen from Figure 2, scope of the areas with seismic intensity higher than IX degree covers all districts and counties of BJ. Furthermore, due to lack of real loss data of BJ, if we add the seismic source location and the different seismic 5 intensity ranges, the simulation uncertainty will be greatly increased instead. Therefore, we assume that the intensity of disaster damage is evenly distributed in the space, and neither seismic source location nor intensity attenuation range is set. The upper limit of the losses caused by a violent earthquake in BJ is given through such extreme assumption.

Setting rescue scenes
After a catastrophe, Chinese government's rescue policy are "concentrating the whole country's efforts on rescue at all costs ", the rescue and reconstruction are led by the central government, which are not depend on the market (e.g. insurance). For recovery and reconstruction work of 2008 WCE, Chinese government provided 120% rescue efforts compare with pre-disaster production capacity (NCDR and MOST, 2008).With the consideration of importance of BJ and its economic development degree, if a devastating earthquake occurs, rescue efforts may reach 150%, and recovery period may be shorter. For this purpose, three rescue efforts scenes are set and they are illustrated in the Table 2 Rescue scene A: Natural recovery. Only satisfying the demand on its basic disaster relief, just depending on BJ's own economic structure and industrial linkage to make the economic system recovered to its pre-disaster level.
Rescue scene B: BJ is provided with 120 % of rescue efforts compare with pre-disaster production capacity according to 20 the real rescue policy for 2008 WCE. Scene B1 is the maximization of production capacity accomplished within 6 months, and scene B2 is to improve the production capacity to maximum 3 months ahead of schedule of Scene B1. Rescue scene C: BJ is provided with stronger rescue efforts (150%) and shorter production capacity improvement period (3months), considering the economic development degree, position, and importance of BJ.

The assessment model of indirect economic loss
The paper utilized the Adaptive Regional Input-Output (ARIO) Model to dynamically assess the indirect economic loss. after the disaster, such as the change of production capacity of various economic sectors, production bottleneck, and capacity restriction caused by the impact of production reduction, production halts, and industrial linkage between various sectors , to 10 make a dynamic simulation on the change of balance of supply and demand of the economic system in the time period from catastrophes occurrence to the completion of reconstruction, so as to describe the impact of the disaster to the regional economy. The modeling time-step width is one month. Key parameters of the model (Table 2) and the core formula are illustrated as follows: The parameter on the left side of the formula is the total output(Y(j)), and parameters on the right side of the formula are

The Spatial diffusion (SDN) model of direct economic loss
Direct economic loss is mainly caused by the damage of fixed assets, belonging to inventory change. Therefore, by utilizing the 5 stock of fixed assets data of the streets/ (villages and towns) of BJ, post-disaster direct economic loss is spatial extended into areas of 321 streets/(villages and towns). The core of formula of SDN model is as follows: Where DELstr(p) is direct economic loss of p th street of BJ, CAPstr (p) stands for the stock of fixed assets of p th street, DELBJ/CAPBJ stands for direct economic loss/stock of fixed asset of BJ, and n stands for 321 streets/ (villages and towns) of 10 BJ.

The inter-regional ripple effect (IRRE) model of indirect economic loss
Indirect economic loss is mainly caused by the production reduction and industrial linkage of various economic sectors, belonging to flow change. Therefore, according to business income of sectors of streets/ (villages and towns), the ripple effects among sectors can be assessed at 321 streets/(villages and towns). According to established IRRE model, the former studies that only obtain the direct and indirect economic loss of a whole region, without more specific losses information inside the region can be resolved. we can not only accurately spread the post-disaster losses from the whole city to districts-streets/(villages and towns), which not only helps to understand spatial heterogeneity of the losses, but also can propose "uneven" post-disaster recovery measures by analyzing the relationship between direct and indirect economic loss of sectors in various streets, to help the government find out optimal recovery and 5 reconstruction solutions for different areas by choosing either increasing rescue efforts to improve production capacity, or importing substitutes product from outside, so as to achieve a quicker recovery to the post-disaster level.

Assessment results of overall losses under rescue scenes
According to the results of simulation, under an extreme scenario, Beijing (BJ) is likely to suffer the USD 150.6 billion (the exchange rate of CNY to USD is 0.14 in 2008) of total loss, including USD 88.1 billion of direct economic loss and USD 62.5 billion of indirect economic loss (Table 3). The indirect economic loss accounts for 41.5% of total loss, and almost the same as the indirect economic loss. The total loss has exceeded the 2.6% of BJ's GDP of 2008, and accounted for 3.6% of the national GDP of 2008. During the 6 years after the catastrophe occurred (2008-2014), the average annual GDP growth rate of BJ could be decreased from 8.55% to 4.91%, the decreasing value is up to 3.63%, however, the goal for GDP growth rate of BJ was 8%, therefore the economic impact caused by the catastrophe can seriously impede the sustainable development in the future (Table 4). The Table 3 shows that, along with the intensifying rescue efforts, the indirect economic loss decreases gradually and thus the total losses decrease. When rescue efforts intensify from rescue scene A(natural recovery) to rescue scene C(150% of rescue efforts), the indirect economic loss is reduced by USD 45.1 billion, and the percentage of indirect economic loss to total losses is reduced from 42% to 17%. By comparing the value of "IEL/ DEL", it can be seen that, when BJ suffers 100 CNY of direct economic loss, its indirect economic loss is reduced from 71 CNY of rescue scene A to 20 CNY of rescue scene C.

5
Where does the reduced 51 CNY of indirect economic loss come from? We can figure it out by further analyzing the reasons: after disaster, damage of fixed assets leads to the inevitable outcome of reduction of production capacity, and the impact of industrial linkage on production capacity could be large or small. However, BJ is featured with strong industrial linkage, which brings difficult economic recovery, long recovery period, and slow capital accumulation. Under the same rescue efforts, when the time for improve production capacity to maximization is 3 months ahead of schedule (scene from B1 to B2), the indirect economic loss reduces by 5 CNY; under the same recovery period of 3 months, when rescue efforts increase by 30% (scene from B2 to C), the indirect economic loss will reduce by 22 CNY. In conclusion, rescue efforts play a more important role in the post-disaster recovery and reconstruction. It's necessary to 15 study how to properly use and scientifically allocate rescue and recovery funds according to the losses of different regions and different economic sectors. Therefore, the study will conduct an in-depth comparative analysis of the ripple effects of indirect economic loss and spatial heterogeneity of direct economic loss in the next section.

Spatial heterogeneity of direct economic loss and ripple effects of indirect economic loss
The losses are closely associated with the economic development of various regions of BJ, so they are not evenly 20 distributed. If rescue funds are evenly distributed according to the total losses without considering spatial differences of specific losses, it is likely to cause improper allocation of funds and reduce rescue efficiency, which is not conductive to post-disaster recovery and reconstruction.
(1) Differences of direct and indirect economic loss on the various districts Statistics losses data on the spatial scales of districts/counties based on those of streets/(villages and towns), which can be found in a macro perspective (Figure 3). The three administrative districts with the most serious direct economic loss are District accounts for 76% of its direct economic loss, and its total economic losses is 1.76 times of its direct economic loss.
Besides, although the indirect economic loss of Chaoyang District is less than that of Haidian District, its indirect economic loss is USD 1.2 billion higher than that of Haidian District due to its developed tertiary industry and strong industrial linkages, demonstrating that, indirect economic loss is relatively higher in economically developed regions, so it shall not be ignored in those more developed regions. (2) Spatial heterogeneity of direct and indirect economic loss It can be seen from Figure 4, firstly, direct and indirect economic loss are mainly concentrated in main urban areas 5 (Dongcheng District, Xicheng District, Haidian district, Chaoyang District, Shijingshan District, and Fengtai District). The longer the distance to the main urban areas is, the smaller the losses will be. Secondly, according to spatial distribution, even though the indirect economic loss is smaller than direct economic loss, the indirect economic loss of some streets/ (villages and towns) are higher than their corresponding direct economic loss. Therefore, the result shows that even though the indirect economic loss of a region is smaller than direct economic loss, some places inside the region may greater than direct economic 10 loss. The result can help the government confirm the important rescue places with more clearly. Thirdly, according to spatial agglomeration degree (Figure 4), the spatial aggregation of direct economic loss is 0.14 (Moran's index), and that of indirect economic loss is 0.27 (Moran's index), both are significant at the 99 percent significant level. Therefore, the spatial distribution of high value of indirect economic loss is more concentrated in the city center, while distribution with high direct economic loss value are relatively distributed dispersedly.  The reasons may include: 1) Direct economic loss is mainly generated by the damage of fixed assets and according to the spatial structure planning of BJ of "Two-axis, two-belt, and multiple centers" (Song, 2009), most of traditional manufacture 5 enterprises has been moved outside the city center, so the fixed assets values of peripheral districts are relatively high, and the post-disaster direct economic loss is higher in these districts; 2) BJ has a developed tertiary industry, accounting for 77.9% of the total GDP of the city, and affected by the population and commercial distribution, the tertiary industry is mainly distributed in the city center, so the spatial distribution of the tertiary industry is mainly concentrated in the city center. In terms of the optimal recovery measure for Commerce & Catering industry (#14), due to its small direct economic loss, its indirect economic loss is mainly caused by industrial linkage, so measure to increase market demands can be taken to pull Nat. Hazards Earth Syst. Sci. Discuss., doi:10.5194/nhess-2016-354, 2016 Manuscript under review for journal Nat. Hazards Earth Syst. Sci.  By analyzing ripple effects, the overall loss can be spatial extended into each street, and sectors' losses in each street can be further evaluated, which helps government policy-makers to intuitively get the information about the loss distribution, so as to properly allocate rescue efforts when taking recovery measures. Under the rescue scene A of natural recovery, no rescue support will be given to the BJ's economic system, and recovery will be accomplished by relying on its own economic structure and industrial linkage, so its reconstruction and recovery period will be relatively longer with larger economic losses. It can be seen that the areas with high losses (Figure 6a)  decrease, and only some streets/ (villages and towns) see relatively high losses. The areas with high losses mainly concentrated on city center areas, such as Xicheng District and Haidian District (Figure 6b). Under rescue effort scene C, the losses of each region are further reduced, but the spatial heterogeneity shows an obvious difference between high value loss and low value loss. (Figure 6c) The high value regions like Yizhuang region (including Economic Development Zone) still have losses without showing much decrease along with the rescue effort improving, while the losses of other regions reduce significantly.

5
Compared with rescue scene A, post-disaster affected areas* is reduced by 4,364 km 2 , accounting for 26.6% of the total area of BJ. It indicates that when the rescue effort is increased to 150% 3 months ahead of schedule, fixed assets of various economic sectors can be recovered with the fastest speed, and the industrial linkage is established rapidly, which will help the BJ's economic system faster recover and finish reconstruction.
*Notes: Affected area calculation is based on disaster magnitude (Freckleton et al., 2012). According to the Geometrical Interval 10 spatial-based segmentation algorithm in GIS, the loss data of four scenes is categorized into five grades of catastrophe (USD 3.75 billion), disaster (USD 0.65 billion), medium disaster (USD 0.11 billion), small disaster (USD 0.019 billion), micro disaster (USD 0.003 billion). The regions involved in corresponding indirect economic loss of disasters with the grade higher than micro disaster is defined as the post-disaster affected areas.

15
In order to reduce the long-term economic impacts caused by disasters and accelerate the economic recovery in the disaster aftermath, this paper sets two scenarios and three models to calculate and analyse the spatial heterogeneity of direct economic loss and ripple effects of indirect economic loss at streets/(villages and towns) of BJ. We set the scenarios of direct economic loss to introduce the sectors' loss caused by 2008 WCE into BJ, set rescue scenes to assess the necessity of improving rescue efforts, adopted ARIO model to evaluate the indirect economic loss of BJ, established SDN model and 20 IRRE model to assess the direct and indirect economic loss of 321 streets/(villages and towns) of BJ. The results indicate that high value of indirect economic loss is more concentrated in the centre of BJ compared with that of direct economic loss.
Finance & Insurance industry (#15) of Chaowai Street in Chaoyang district suffers the most serious indirect economic loss.
During the six years after catastrophe occurred (2008)(2009)(2010)(2011)(2012)(2013)(2014), the average annual GDP growth rate of BJ could be decreased from 8.55% to 4.91%, In terms of the 8% of BJ's GDP growth rate target, it's a significant and noticeable economic impact.
Before the disaster, attention shall be paid to the adjustment and optimization of industrial structure; In disaster emergency rescue stage, attention shall be paid to the transportation of relief material allocation; during post-disaster recovery and reconstruction stage, attention shall be paid to give the priority to the development of the industries with high industrial linkage coefficient, so as to speed up the economic recovery. Therefore, adjustment of industrial structure and strengthen of industrial linkage can mitigate the impact caused by natural disasters. Besides, optimization of industrial structure, and close 5 industrial linkage is of great significance in regional growth. In order to achieve sustainable development of a region, a balanced point between economic development and disaster mitigation must be found, and by taking it as a reference, a strategy for the optimization and adjustment of industrial structure shall be established. The results can provide the scientific and effective support to find the above "balanced point" according to increase rescue effort and to priority support the industries which are located in the seriously damaged regions.