Kazakhstan: Strengthening Catastrophe Risk Transfer ...siteresources.worldbank.org/FINANCIALSECTOR/Resources/World-Ban… · Strengthening Catastrophe Risk Transfer Supervision FIRST

SPA Risk LLC

Kazakhstan: Strengthening Catastrophe Risk Transfer Supervision

FIRST Initiative Project #8135

Prepared for

THE WORLD BANK Washington DC

Purchase Order 0007744725

Rev 1: 21 Dec 2010

Kazakhstan Seismic Risk Analysis Rev 1: 21 Dec 2010, The World Bank

2 SPA

EXECUTIVE SUMMARY

Kazakhstan is highly vulnerable to earthquakes. In order to examine potential losses that earthquakes might cause Kazakhstan, a seismotectonic model was developed and combined with information on Kazakhstan’s building characteristics, to estimate the probability of loss exceeding various values. These loss exceedance curves can be used for examination of earthquake insurance rates and solvency.

Earthquake loss begins with the initiation of an event at the seismic source, whose effects are attenuated to a site or sites of interest, where they are adjusted for site conditions, to determine the site hazard. The site hazard is then combined with the building or other asset characteristics via vulnerability functions, to determine the damage to the asset. The financial cost of this damage is quantified as the loss, and the insured loss as the financial cost inuring to the insurer per the terms of the insurance contract. In order to quantify this process for the study region, a seismotectonic model for Kazakhstan was developed, including a detailed characterization of site conditions.

Analyses were then conducted to generate a stochastic set of 10,000 seismic events affecting the study region. For each event, response spectral attenuation (5% damped, 1 second period) relations were employed to determine shaking intensities at various locations within the seismic portions of Kazakhstan. Uncertainty in ground motion and seismic vulnerability were considered in these analyses. Results based on the above methodology were provided in a 5mb zip file containing 33 files. Additionally, an Annual Probability of Exceedance Loss (“APEL”) table was provided for 2% deductible losses for 24 locations in Kazakhstan, for APELs ranging from 0.02 to 0.005 frequency per annum (50 to 200 year ‘return periods). A rule was also provided on combining these APELs to account for more than one location being affected by the same event. Lastly, a PowerPoint presentation was prepared for use in Kazakhstan by World Bank personnel, illustrating various aspects of the analysis.


3 SPA

Table of Contents

EXECUTIVE SUMMARY ............................................................................................................ 2 1. INTRODUCTION .................................................................................................................. 5

1.1. Background and Purpose of Project ................................................................................ 5 1.2. Scope of Work ................................................................................................................ 5 1.3. Study Region ................................................................................................................... 6 1.4. Organization of Report ................................................................................................... 6

2. EARTHQUAKE LOSS ESTIMATION ................................................................................. 7 2.1. Overview of Earthquake Loss and Its Estimation........................................................... 7 2.2. Exposure Database .......................................................................................................... 7 2.3. Hazard Characterization .................................................................................................. 8 2.4. Seismic Vulnerability Functions ..................................................................................... 9

3. ANALYSIS AND Results .................................................................................................... 11 3.1. Analysis......................................................................................................................... 11 3.2. Stochastic Loss Sets: COM, RES and IND losses by deductible ................................. 11 3.3. Annual Probabilities of Exceedance Losses (APELs) .................................................. 13 3.4. Combination of APELs for Country-wide reserve ....................................................... 14 3.5. PowerPoint presentation ............................................................................................... 14

REFERENCES ............................................................................................................................. 15 APPENDIX A – BACKGROUND ON SEISMIC VULNERABILITY FUNCTIONS .............. 18 TABLES ....................................................................................................................................... 19 FIGURES ...................................................................................................................................... 21

Revision Date Comment

1 21 Dec 2010 Results revised after review; additional results in term of Annual Probability of Exceedance (“APE”) tables, and a ‘correlation factor’. Figures 16 to 20 revised or added.


4 SPA

List of Tables Table 1 Units of Aggregation for the Project................................................................................ 19 Table 2 Example Expected Annual Loss (EAL) for various types of Construction, Almaty ....... 20

List of Figures Figure 1 Elements of Probabilistic Earthquake Loss Estimation .................................................. 21 Figure 2 Earthquake Loss Model Components ............................................................................. 22 Figure 3 Kazakhstan Historic Seismicity ...................................................................................... 23 Figure 4 Kazakhstan Seismic Zones Source: (Giardini, 1999) .................................................... 23 Figure 5 Kazakhstan overlaid on Munich Re Earthquake Zones ................................................. 24 Figure 6 Regional seismotectonics ............................................................................................... 24 Figure 7 Regional Historic Seismicity with major plate boundaries ............................................ 25 Figure 8 Regional CMT focal mechanisms and plate boundaries ................................................ 25 Figure 9 Vs30 Soil Conditions, Kazakhstan ................................................................................. 26 Figure 10 Vs30 Soil Conditions, Almaty ...................................................................................... 26 Figure 11 Buildings overlaid on Vs30 Soil Conditions, Almaty .................................................. 27 Figure 12 Structure type distributions for Alma Aty region ......................................................... 28 Figure 13 Sample seismic vulnerability function from Porter (2009b) ........................................ 29 Figure 14 Vulnerability Functions used in JICA Study ................................................................ 30 Figure 15 Vulnerability Functions Employed in this study .......................................................... 31 Figure 16 Example Hazard Curves, 24 locations.......................................................................... 32 Figure 17 Loss Exceedance Curves, various types of COM construction, various locations ...... 33 Figure 18 Fraction of each building type in each Unit of Aggregation ........................................ 34 Figure 19 Annual Probability of Exceedance Losses (APELs) for 2% deductibles, for COM, RES and IND occupancies for 24 locations for several ‘return periods’. ............................................. 35 Figure 20 COM “200 year” losses overlaid on historic seismicity, showing many parts of Kazakhstan with zero or near-zero losses, but South and East Kazakhstan have substantial seismic risk, particularly Almaty. ................................................................................................. 36 Figure 21 Radius of 200 km drawn about Almaty City. Red, blue and green circles are reverse, normal and transform earthquake mechanisms, resp. – the predominance of red events indicates primarily a reverse or thrust environment. A 200 km long reverse fault would correspond approximately to a Mw 8.2 earthquake, similar to the 1911 Kemin event. .................................. 37 Figure 22 Isoseismals (MSK) of 1911 Mw 8.2 Kemin earthquake – Almaty City is MSK 8+, while Almaty Oblast generally MSK 6-8 and Bishkek is MSK 6+. ............................................. 38 Figure 23 Isoseismals (MSK) of 1889 Mw 8.3 Chilik earthquake over historic seismicity – Almaty City is MSK 8, while Almaty Oblast generally MSK 7-8 and Bishkek is MSK 6+. ....... 39 Figure 24 Thumbnail view of powerpoint presentation prepared for the project, for use by World Bank personnel in Kazakhstan. ..................................................................................................... 40


5 SPA

1. INTRODUCTION

This section provides the background, scope of work and organization of this report.

1.1. Background and Purpose of Project Kazakhstan is located in and to the north of the Persia-Tibet-Burma (i.e., Himalayan) orogeny, and Tien Shan mountain chain. As such, portions of Kazakhstan are among the most seismically active places on earth, so that several of Kazakhstan’s cities, including its capital Almaty, are at great seismic risk. Property insurance in Kazakhstan often covers not only the normal peril of fire, but also earthquake. However, insurance company exposure to earthquake is not well monitored. In order to better monitor earthquake insurance exposure, the World Bank in cooperation with the Kazakhstan insurance industry, is developing an exposure tracking and monitoring tool. This report provides basic loss data in support of that tool.

Because insurance industry data collection in Kazakhstan is not highly detailed, the results presented here are aggregated into several sites and simple categories of construction, so as to be manageable for business purposes. Results are presented for 24 aggregate locations, and several materials of construction (masonry whether fired brick or earthen/stone, reinforced concrete, and wood), subcategorized by occupancy and in some cases era of construction. The overall stochastic results nevertheless were quite voluminous, consisting of 33 files (three classes of occupancy for 0 to 10% deductibles) each of which consisted of several thousand or more events for 13 classes of exposure, for the 24 aggregate locations, resulting in about 50 million data points (note that zero loss events were eliminated).

1.2. Scope of Work The scope of work for the project consisted of:

a) Perform probabilistic analyses to develop a stochastically generated dataset of earthquake events (based on the historic seismic catalogues); for each event, the consultant will provide peak ground acceleration (PGA) and / or intensities for each grid cell (at least a zip code level).

b) b) Calculate economic losses to all major classes of construction from different seismic events in defined locations, including residential, commercial, and industrial facilities. The final results of the calculations should be presented in the form of PML curves for each construction class. The consultant should supply vulnerability matrices for all chosen building vulnerability classes used in calculating the vulnerability curves for every class of construction, including housing, office space and industrial facilities. The work should result in an Excel modeling tool allowing to modify the replacement cost estimates for different classes of construction.

c) The event set will also account for correlations that may exist between economic losses to different insured property classes from seismic events in different key locations.


6 SPA

d) Develop insured loss estimates for given insurance products (exact deductibles and limits will be provided by the project team at the start of the assignment 1

e) The output of the project will be presented in an excel data file containing at least 30,000 loss events; estimates of PMLs for different classes of construction (e.g. residential, commercial and industrial) at above specified locations; a table of correlations between insured losses for different classes of construction in key locations.

) for different construction classes in key locations and for the portfolio as a whole. The consultant will provide the tabulated estimates of insured loss for any given level of insurance penetration by location and construction class, as well as the table of correlations between different losses in different locations.

1.3. Study Region The project study region was Kazakhstan and also the city of Bishkek, Kyrgyzstan. Figure 3 shows Kazakhstan’s historic seismicity and largest cities, from which it can be seen that only the southern portions of Kazakhstan are significantly affected by earthquakes. As a result, only the southern portions are zoned for earthquake, Figure 4 and Figure 5. However, it was agreed that the reporting scheme for the project region would consist of 23 major cities throughout Kazakhstan as well as the city of Bishkek, Kyrgyzstan, as shown in Table 1. For each of these units, Residential, Commercial and Industrial risks will be analyzed, for deductibles of 0 to 10%, in 1% steps.

1.4. Organization of Report The next section provides an overview of the data developed for this study. Section 3 describes our analysis, the deliverables, our results and validation. References, Tables and Figures complete the report.

1 Deductibles for the analysis were agreed to be zero to 5%, in 1% increments.


7 SPA

2. EARTHQUAKE LOSS ESTIMATION

2.1. Overview of Earthquake Loss and Its Estimation Figure 1 summarizes the elements of Earthquake Loss Estimation. In the figure, assets – that is property locations – are shown as black dots. These assets are affected by seismic hazards, especially shaking but also other hazards as discussed below. Damage is the physical degradation of the asset resulting from the occurrence of the hazard. Loss is the financial cost of correcting the degraded condition – the cost of repair of cracked walls, or the replacement value of a totally damaged building or other asset (e.g., the contents), or water-damaged contents.

Earthquake loss is therefore the product of a series of components, beginning at the seismic source, attenuated to a site, adjusted for site conditions, resulting in the site hazard. The site hazard then is combined with the building or other asset characteristics, via vulnerability functions, to determine the damage to the asset. The financial cost of this damage is quantified as the loss. These components, and the way in which information flows in an earthquake loss model, are shown in Figure 2.

2.2. Exposure Database Estimation of earthquake loss begins with determining the assets at risk, also termed the ‘exposure’ or ‘inventory’. Kazakhstan buildings are typically built of several materials: masonry (either fired brick or earthen/stone), reinforced concrete, or wood. The broad material categories may be further subdivided according to era of construction, number of stories, and occupancies. In order to develop a Kazakhstan building inventory, actual building data was not available for all of Kazakhstan so an exposure inventory was developed based on proxy information, which is a common approach in the insurance industry in this situation. Specifically, population and industrial data for selected portions of each country were accumulated from various sources such as census data. This data was then employed to estimate residential, commercial and industrial building area and value based on unit building area per capita, and urban and rural value per unit area for each country, values being adjusted by GDP pc. This approach was further justified due to the purpose of the analysis, which is overall loss monitoring.

In order to characterize the building stock corresponding to this exposure, researchers at the US Geological Survey, as part of the USGS’s Prompt Assessment for Global Earthquakes for Response (PAGER) project have estimated the global distribution of construction types by country (Jaiswal and Wald 2008). The work categorizes building stock among approximately 100 structure types, using a taxonomy referred to as PAGER-STR, developed by SPA Risk LLC and USGS personnel that merges three existing standards: that of the US Federal Emergency Management Agency (FEMA) and used for example in FEMA 154 (ATC 2002); the World Housing Encyclopedia (Earthquake Engineering Research Institute 2006); and the 1998 European Macroseismic Scale (European Seismic Commission Working Group—Macroseismic Scales 1998). The database provides an estimate of the fraction of the total area of buildings (urban and rural, residential and nonresidential) in each type, by interpreting a wide variety of available sources. Figure 11 for example shows detailed data for Almaty. The particular sources used for countries in the present study were: (Ashimbayev et al., 2001, AXCO, 2009a, AXCO,


8 SPA

2009b, Itskov et al., 2003, JICA and AKIMAT, 2008, UN-HABITAT, 2010a, UN-HABITAT, 2010b). The JICA study in particular was useful, Figure 12. The resulting exposure was then aggregated into the following categories:

• Fired Brick Masonry

• Unfired Masonry

• Pre 1998 Reinforced Concrete

• Post 1998 Reinforced Concrete

• Wood

• Average Building (weighted average of above) In addition to these basic structural categories, losses were also estimated for the contents for each category. That is, exposure data was collected (and losses reported) in the end for 13 categories of loss: two for each of the above six structural types (structure only, contents only), plus one additional category which was average building structure plus contents.

2.3. Hazard Characterization The next step in the estimation of earthquake loss is seismic hazard characterization. The seismotectonics of Kazakhstan is characterized by interactions between the Northern Eurasia, Arabia, India, Yangtze and Amur major lithospheric plates and continental blocks, resulting in the Persia-Tibet-Burma (i.e., Himalayan) orogeny, which includes the Tien Shan mountain chain along the Kazakhstan- Kyrgyzstan border, Figure 6. North of the Tien Shan, the Northern Eurasia plate is locally a platform with very low, diffuse seismicity, Figure 7. Focal mechanisms are shown in Figure 8, which indicates mostly reverse faulting in the Tien Shan, with some transform motion. Regarding historic seismicity for the region – the earliest record event in or within 200 km of Kazakhstan dates from 250 BC, with four events in the record greater than Mw8 and the largest event being a Mw 8.9 event that occurred in 1911 at 43.5N 77.5E. This data has been used by SPA to develop seismic source zones for the region, seismicity for each of which is characterized by a magnitude frequency relation. These magnitude frequency relations are randomly sampled for a simulated 10,000 year period, to develop a stochastic event set, which is a series of events specified by magnitude, epicentral location and associated annual frequency that simulate expected earthquakes for the region. For each event in the stochastic set, the shaking intensity for each population centroid was then calculated using the attenuation equation by (Campbell and Bozorgnia, 2008), one of the so-called “Next Generation Attenuation” equations (Abrahamson et al., 2008) which are the most recent and widely considered the most reliable for this purpose, having been developed using the largest strong ground motion dataset, appropriate for active crustal regions such as the portions of South East Europe considered in this study (Stafford et al., 2007). Uncertainty in strong ground motion is significant and was taken into account in this study. Local soil conditions were also taken into account, using the global Vs30 mapping by (Allen and Wald, 2007), Figure 9 and Figure 10. The specific shaking intensity measure employed was 1.0 hertz response spectral acceleration. Results are shown in Figure 16 for 24 locations in Kazakhstan.


9 SPA

2.4. Seismic Vulnerability Functions In order to estimate damage, building types for the study region were characterized based on a recent global building inventory characterization (Porter et al., 2008). In order to quantify damage, seismic vulnerability functions are required that determine damage given hazard, where damage is defined as the repair cost normalized by the replacement cost of the building. Estimation of seismic vulnerability functions is a complex and difficult task – see Appendix A for some background on this subject. In brief, determination of seismic vulnerability functions begins with calculation of the seismic response of a building (its displacement and acceleration, denoted here by Sd and Sa) to a given level of seismic excitation using the capacity spectrum method of structural analysis (ATC 1996). Given the displacement and acceleration, one calculates the probabilistic damage state of each of three general building components: structural, nonstructural drift-sensitive, and nonstructural acceleration-sensitive. The damage calculation takes each component as being in one of 5 damage states (none, slight, moderate, extensive, and complete), each of which has an associated fragility function that gives the probability of reaching or exceeding the damage state as a function of either acceleration or displacement. The fragility functions take the form of cumulative lognormal distributions, each with an associated median and logarithmic standard deviation value that must be specified. Thus the probabilistic damage state of each component can be calculated as shown in Equation (1). In the equation, P denotes probability; D2 denotes the uncertain damage state of the nonstructural drift-sensitive component, which can take on the values d = 0 (none), 1 (slight), etc.; Sd is displacement, Φ denotes the standard cumulative normal distribution, and θ and β are parameters of the distribution.

( )

( ) ( )

( )

12

1

1

1

4

4

ln1 0

ln ln1 3

ln4

d

d d

d d

xP D d S x d

x xd

xd

θβ

θ θβ β

θβ

+

+

= = = −Φ =

= Φ −Φ ≤ ≤

= Φ =

(1)

Each component and damage state is associated with a mean cost to restore the component from the given damage state. The total building repair cost is given by Equation (2), in which L denotes mean damage factor (repair cost as a fraction of replacement cost), D1 and D3 denote the uncertain damage to the structural component and nonstructural acceleration-sensitive components, respectively (calculated similarly to D2) ; L1d, L2d, and L3d denote the cost to repair the structural, nonstructural drift-sensitive, and nonstructural acceleration-sensitive components, respectively, as a fraction of the building’s replacement cost; and Sa denotes structural acceleration. The process of calculating L is repeated at many levels of seismic excitation to produce a seismic vulnerability function. Separate seismic vulnerability functions are created for each combination of structure type, design level, occupancy class, magnitude, distance, seismic domain, and NEHRP site soil class.


10 SPA

5 4 4

1 1 2 2 3 31 1 1

d d d d a dd d d

L P D d S x L P D d S x L P D d S y L= = =

= = = + = = + = = ∑ ∑ ∑ (2)

One important challenge related to employing HAZUS-MH is that the vulnerability relationships are derived in part using the capacity spectrum method of structural analysis, which tends to require iteration, followed by the calculations of probabilistic damage state and loss. The result is that loss calculations can be time consuming, and loss can be difficult to relate back to a structure-independent intensity measure.

Under research sponsored by the US Geological Survey and the Southern California Earthquake Center, SPA personnel (Porter 2009a, b) showed how a seismic vulnerability function can be created that honors all HAZUS-MH methodologies and data, yet that tabulates mean loss as a function of a structure-independent intensity measure, in particular, geometric-mean-component, site-soil-adjusted 5%-damped spectral acceleration response at 0.3-sec and 1.0-sec periods. Here, “honoring all HAZUS-MH methodologies” means that the methodology actually uses the capacity spectrum method of hazard and structural analysis to determine structural response to a scenario earthquake. It accounts for the effects of magnitude, distance, site amplification, seismic regime (plate boundary vs. continental interior), hysteretic energy dissipation, and using the HAZUS-MH structural model of an elastic-softening-perfectly plastic single-degree-of-freedom oscillator. The work produced tables of repair cost as a fraction of replacement cost, again as a function of site-soil-adjusted Sa(0.3 sec, 5%) and Sa(1.0 sec, 5%) for a combination of any of 5 NEHRP site soil classes, 4 magnitude ranges, 4 distance ranges, two seismic regions (western US or central and eastern US), 36 model building types, 4 code eras, and 33 occupancy classes. A sample is shown in Figure 13.

This approach was employed to develop preliminary vulnerability functions for various structural types (more detailed than as aggregated in the final reporting). These preliminary vulnerability functions were then combined with other vulnerability data, particularly that from the JICA study, Figure 14, to develop final vulnerability functions as used in this study, Figure 15.


11 SPA

3. ANALYSIS AND RESULTS

3.1. Analysis Using the above data, analyses were conducted to generate a stochastic set of 10,000 seismic events affecting the study region. For each event, response spectral attenuation (5% damped, 1 second period) relations were employed to determine shaking intensities at centroids in the study region. The resulting hazard table was then used to estimate seismic damage at each of the locations for all 10,000 events, considering uncertainty on the seismic vulnerability functions. The resulting damage array was then used to determine insured loss as a function of insurance deductible and penetration. All Event losses were weighted by the corresponding Event frequency, to determine Expected Annualized Loss (EAL), for the corresponding categories. Table 2 shows example EAL for Almaty, for various types of exposure. Losses are expressed in fraction of value for each event. An example Loss Exceedance curve is shown in Figure 17 for various locations (0% deductible). The next section describes how these results were formatted for delivery.

3.2. Stochastic Loss Sets: COM, RES and IND losses by deductible Results based on the above methodology were provided a 5mb zip file containing 33 files (shown in part below):


12 SPA

The file naming convention for the CSV files is clear: each file is a CSV file for one of three occupancies: Res/Comm/Indust and a deductible ranging from 0% to 10%. Each file has the appearance:

when opened in Excel. If pasted into the template spreadsheet, the appearance is:

Note that in the upper figure, the table is shown at the upper left corner, while in the lower figure, the bottom right corner is shown. For Residential occupancies, there are 5913 non-zero events (for 0% deductible, fewer for higher deductibles, but these zero loss events are included per World Bank direction), each event with 0.0001 frequency per annum. Note that the table, after the event info on the right, consists of 312 columns = 13 columns for each of 24 exposure


13 SPA

groups. The exposure groups (ie, cities and provinces) were all for Kazakhstan with the exception of Bishkek (to represent exposures in Kyrgyzstan). Some of these exposures have virtually no seismic hazard, while others (Almaty, Bishkek) are perhaps the highest seismic hazards in the populated world.

For any one of the exposures, there are 12 columns, as shown below for the first exposure group (Uralsk). There are five structural material categories: masonry, pre1998 Reinforced Concrete (RC), post-98 RC, Unfired Brick (ie, earthen buildings) and Wood. For each of these five types, there is an associated column for contents in that building type, for a total of 10 columns. Lastly, there is a column for an “Average Building”, “Average Building Contents” and lastly “Average Building AND Contents”.

Each cell in the tables is therefore the loss for that Exposure site (eg, Uralsk) for a particular building type (eg, masonry) or its contents, for a particular event. These losses are ‘pure losses’ – that is, there is no weighting for relative proportion of masonry vs. wood etc. Thus, if a company specifies its exposure by geographic exposure site and by material of construction, and separates out contents, these losses can be applied directly to their building and contents exposures to estimate their losses.

The Average Building is a weighted average of these losses, taking into account contents. That is, for each of the 24 exposure sites, census and economic data were reviewed to estimate the relative proportion of the five materials for residential, commercial and industrial type occupancies. Contents were uniformly considered to be 50% of the building value for Residential, 100% of the building value for Commercial, and 150% of the building value for Industrial occupancies. Based on this, the weights shown in the large table on the next page were arrived at. While the input data varied significantly, the results were surprisingly uniform – it can be seen for example that typically anywhere in Kazakhstan, masonry is about 17% of residential occupancies (some variation – low of 17% and a high of 20%). These weights were applied to the individual loss rates in the other 10 columns to develop the losses for an “Average Building” and an “Average Building Contents”. These Average values can be used if the insurer does not have a breakdown of exposure by materials. Per World Bank direction, the actual fractions are provided, in Figure 18.

3.3. Annual Probabilities of Exceedance Losses (APELs) The above results are provided in another format in Figure 19, which provides Annual Probabilities of Exceedance Losses (APELs) given a 2% deductible, for COM, RES and IND


14 SPA

occupancies, for average building and contents, for 24 locations in Kazakhstan. The annual probabilities of exceedance correspond to ‘return periods’ of 50, 75, 100, 125, 150 and 200 years. Figure 20 overlays the ‘200 year’ losses for the COM occupancies over historic seismicity, where it can be seen that the locations with significant losses are in or near regions of significant seismicity.

3.4. Combination of APELs for Country-wide reserve The APELs discussed in the previous section are estimates of the potential losses to individual locations in Kazakhstan. There is a possibility that a large earthquake can affect more than one location, so that the question arises “how should APELs for two or more locations be combined to cover the largest loss country-wide, for a given return period?” Review of the stochastic sets discussed above, as well as the largest events likely for various parts of Kazakhstan (see Figure 21) indicates that the largest aggregate loss would be due to a large earthquake in or near Almaty City – that is, one likely to cause major damage in Almaty City that would correspondingly also cause major damage in Almaty Oblast, and possibly also cause significant damage in Bishkek (for example, if the fault rupture began near Bishkek and extended to the vicinity of Almaty City). Figure 22 and Figure 23 for the 1911 Mw 8.2 Kemin and 1889 Mw 8.3 Chilik events, respectively, show that large earthquakes can affect all three locations.

Based on this review, it would appear that, at a minimum, a company’s reserves for a major earthquake should be the sum of reserves for Almaty City and Almaty Oblast, or the sum of reserves for Almaty City and some fraction of Almaty Oblast and Bishkek, whichever is larger – that is, for example:

Min Reserve = 𝑚𝑚𝑚𝑚𝑚𝑚 � 𝑅𝑅𝐴𝐴𝐴𝐴𝑚𝑚𝑚𝑚𝐴𝐴𝐴𝐴 𝐶𝐶𝐶𝐶𝐴𝐴𝐴𝐴 + 𝑅𝑅𝐴𝐴𝐴𝐴𝑚𝑚𝑚𝑚𝐴𝐴𝐴𝐴 𝑂𝑂𝑂𝑂𝐴𝐴𝑚𝑚𝑂𝑂𝐴𝐴

𝑜𝑜𝑜𝑜𝑅𝑅𝐴𝐴𝐴𝐴𝑚𝑚𝑚𝑚𝐴𝐴𝐴𝐴 𝐶𝐶𝐶𝐶𝐴𝐴𝐴𝐴 + 0.5𝑅𝑅𝐴𝐴𝐴𝐴𝑚𝑚𝑚𝑚𝐴𝐴𝐴𝐴 𝑂𝑂𝑂𝑂𝐴𝐴𝑚𝑚𝑂𝑂𝐴𝐴 + 0.5𝑅𝑅𝐵𝐵𝐶𝐶𝑂𝑂ℎ𝑘𝑘𝑘𝑘𝑘𝑘

� (3)

where 𝑅𝑅𝐴𝐴𝐴𝐴𝑚𝑚𝑚𝑚𝐴𝐴𝐴𝐴 𝐶𝐶𝐶𝐶𝐴𝐴𝐴𝐴 is the fractional Reserve for Almaty City for example, for COM, RES or IND as determined from Figure 19 for a given ‘return period’. Therefore, if the COM Total Sums Insured by a company for Almaty City, Oblast and Bishkek are USD 3 million, 1 million and 2 million respectively, and a ‘100 year return period’ is selected as the level of protection, then the Min Reserve is the maximum of

0.14 * $3 million + 0.14 * $1 million = $ 0.56 million

or

0.14 * $3 million + 0.14 * 0.5 * $1 million + 0.5 * 0.02 * $2 million = $ 0.51 million

that is, $0.56 million.

3.5. PowerPoint presentation Lastly, a PowerPoint presentation was prepared for use in Kazakhstan by World Bank personnel, illustrating various aspects of the analysis, Figure 24.


15 SPA

REFERENCES Abrahamson, N., Atkinson, G., Boore, D., Bozorgnia, Y., Campbell, K., Chiou, B., Idriss, I. M.,

Silva, W. & Youngs, R. (2008) Comparisons of the NGA Ground-Motion Relations. Earthquake Spectra, 24, 45–66pp.

Allen, T. I. & Wald, D. J. (2007) Topographic slope as a proxy for global seismic site conditions (VS30) and amplification around the globe. pp. 69 p. U.S. Geological Survey.

Ashimbayev, M. U., Itskov, I. E. & Lobodryga, T. D. (2001) Seismic Hazard and Earthquake Engineering In the Republic of Kazakhstan. Global Blueprints for Change, First Edition--Prepared in conjunction with the International Workshop on Disaster Reduction convened on August 19-22, 2001, Theme A: Living With Natural and Technological Hazards, Topic A.2: Reducing Vulnerabilities in Existing Building and Lifelines. Almaty.

AXCO (2009a) Insurance Market Report, Kazakhstan: Non-Life (P&C). pp. 130. AXCO (2009b) Insurance Market Report, Kyrgyzstan: Non-Life (P&C). pp. 147. Campbell, K. & Bozorgnia, Y. (2008) NGA Ground Motion Model for the Geometric Mean

Horizontal Component of PGA, PGV, PGD and 5% Damped Linear Elastic Response Spectra for Periods Ranging from 0.01 to 10 s. Earthquake Spectra, 24, 139-171pp.

Giardini, D. (1999) The Global Seismic Hazard Assessment Program (GSHAP) - 1992/1999 [A compendium of articles describing the GSHAP project. ]. Annall di Geofisica, 42, 957-1230pp.

Itskov, I. E., Umarbayevich, A. M. & Chernov, N. B. (2003) Kazakhstan, Prefabricated large panel concrete buildings with two interior longitudinal walls. World Housing Encyclopedia, pp. 36.

JICA & AKIMAT (2008) The Study on Earthquake Disaster Risk Management for Almaty City in the Republic of Kazakhstan, Interim Report. pp. 374. prepared by OYO International Corp., Nippon Koei Co., Ltd. and Aero Asahi Corporation for the Japan International Cooperation Agency (JICA) and AKIMAT of Almaty City of the Republic of Kazakhstan.

Porter, K. (2003) Seismic Vulnerability. Earthquake Engineering Handbook (eds W. F. Chen & C. Scawthorn). CRC Press, Boca Raton.

Porter, K. A., Jaiswal, K. S., Wald, D. J., Earle, P. S. & Hearne, M. (2008) Fatality Models for the U.S. Geological Survey’s Prompt Assessment of Global Earthquakes for Response (PAGER) System. The 14th World Conference on Earthquake Engineering. Beijing, China.

Stafford, P. J., Strasser, F. O. & Bommer, J. J. (2007) Preliminary Report on the Evaluation of Existing Loss Estimation Methodologies. pp. 108pp. Network of Research Infrastructures for European Seismology, Istanbul Meeting.

UN-HABITAT (2010a) Kazakhstan Settlement Indicators. United Nations Human Settlements Indicators, Global Urban Observatory Databases, United Nations Human Settlements Programme, http://www.unchs.org/programmes/guo/guo_databases.asp (accessed 25 Jan 2010).

UN-HABITAT (2010b) Kazakhstan Urban Indicators. United Nations Human Settlements Indicators, Global Urban Observatory Databases, United Nations Human Settlements

http://www.unchs.org/programmes/guo/guo_databases.asp�


16 SPA

Programme, http://www.unchs.org/programmes/guo/guo_databases.asp (accessed 25 Jan 2010).

(ATC) Applied Technology Council, 1985. ATC-13, Earthquake Damage Evaluation Data for

California, Redwood City, CA, 492 pp. Beck, J.L., A. Kiremidjian, S. Wilkie, A. Mason, T. Salmon, J. Goltz, R. Olson, J. Workman, A.

Irfanoglu, and K. Porter, 1999. Decision Support Tools for Earthquake Recovery of Businesses, Final Report, CUREe-Kajima Joint Research Program Phase III, Consortium of Universities for Earthquake Engineering Research, Richmond, CA.

(ATC) Applied Technology Council, 2002. FEMA 154: Rapid Visual Screening of Buildings for Potential Seismic Hazards: A Handbook, Second Edition, Federal Emergency Management Agency, Washington, DC.

(ATC) Applied Technology Council, 1996. ATC-40: Seismic Evaluation and Retrofit of Concrete Buildings, vols 1 and 2. Redwood City, CA.

Czarnecki, R.M., 1973, Earthquake Damage to Tall Buildings, Structures Publication 359, Massachusetts Institute of Technology, Cambridge, MA, 125 pp.

European Seismic Commission Working Group—Macroseismic Scales, 1998. European Macroseismic Scale 1998 EMS-98. Luxembourg. http://www.gfz-potsdam.de/pb5/pb53/projekt/ems/eng/index_eng.html [17 Jul 2006]

Goulet, C., C. Haselton, J. Mitrani-Reiser, J. Beck, G. Deierlein, K. Porter, and J. Stewart, 2007. Evaluation of the seismic performance of a code-conforming reinforced-concrete frame building - from seismic hazard to collapse safety and economic losses. Earthquake Engineering and Structural Dynamics. 36 (13), 1973-1997.

Jaiswal, K., and D. Wald, 2008. Creating a Global Residential Building Inventory for Earthquake Loss Assessment and Risk Management. Open File Report 2008-1160. US Geological Survey, Reston VA.

Kircher, C.A., A.A., Nassar, O. Kustu, and W.T. Holmes, 1997. Development of building damage functions for earthquake loss estimation. Earthquake Spectra, 13 (4) 663-682

Krawinkler, H., ed., 2005. Van Nuys Hotel Building Testbed Report: Exercising Seismic Performance Assessment, Report 2005-11, Pacific Earthquake Engineering Research Center, Richmond, CA

Kustu, O., D.D. Miller, and S.T. Brokken, 1982, Development of Damage Functions for Highrise Building Components, for the US Department of Energy, URS/John A Blume & Associates, San Francisco, CA

McGowan, S.M., 2009. Extracting Values of Some Key HAZUS-MH Seismic Vulnerability Parameters from Dynamic Test Results, with Application to Adobe Dwellings. Master’s thesis, Department of Civil, Environmental, and Architectural Engineering, University of Colorado, Boulder

(NIBS and FEMA) National Institute of Building Sciences and Federal Emergency Management Agency, 2003. Multi-hazard Loss Estimation Methodology, Earthquake Model,

http://www.unchs.org/programmes/guo/guo_databases.asp�

http://www.gfz-potsdam.de/pb5/pb53/projekt/ems/eng/index_eng.html�

http://www.gfz-potsdam.de/pb5/pb53/projekt/ems/eng/index_eng.html�


17 SPA

HAZUS®MH Technical Manual. Federal Emergency Management Agency, Washington, DC, 690 pp.

Porter, K.A., A.S. Kiremidjian, and J.S. LeGrue, 2001. Assembly-based vulnerability of buildings and its use in performance evaluation. Earthquake Spectra, 17 (2), pp. 291-312, www.sparisk.com/publications.htm

Porter, K.A., J.L. Beck, and R.V. Shaikhutdinov, 2002a. Sensitivity of building loss estimates to major uncertain variables. Earthquake Spectra, 18 (4), 719-743, www.sparisk.com/publications.htm

Porter, K.A., J.L. Beck, H.A. Seligson, C.R. Scawthorn, L.T. Tobin, and T. Boyd, 2002b. Improving Loss Estimation for Woodframe Buildings, Consortium of Universities for Research in Earthquake Engineering, Richmond, CA, 136 pp.

Porter, K.A., 2009b (expected). Cracking an open safe: more HAZUS vulnerability functions in terms of structure-independent spectral acceleration. Earthquake Spectra 25 (3)

Porter, K.A., 2009a. Cracking an open safe: HAZUS vulnerability functions in terms of structure-independent spectral acceleration. Earthquake Spectra 25 (2), 361-378

Robinson, D., T. Dhu, and J. Schneider, 2006. Practical probabilistic seismic risk analysis: a demonstration of capability. Seismological Research Letters, 77 (4): 452-458

Steinbrugge, K.V., 1982. Earthquakes, Volcanoes, and Tsunamis, An Anatomy of Hazards, Skandia America Group, New York, 392 pp.

Steinbrugge, K.V. and S.T. Algermissen, 1990, Earthquake Losses to Single-Family Dwellings: California Experience, Bulletin 1939, U.S. Geological Survey, Washington, DC.

Whitman, R.V., J.W. Reed, and S.T. Hong, 1973. Earthquake damage probability matrices. Proceedings of the Fifth World Conference on Earthquake Engineering, 25-29 June 1973, Rome, Italy, pp. 2531-2540

http://www.sparisk.com/publications.htm�


18 SPA

APPENDIX A – BACKGROUND ON SEISMIC VULNERABILITY FUNCTIONS

Seismic vulnerability functions express the relationship between damage or loss to a structure or facility and earthquake effects (i.e., intensity). The earthquake effect most commonly addressed by seismic vulnerability functions, and the only effect addressed in this study, is earthquake shaking.

Authoritative seismic hazard information is readily available for some areas of the world, such as through various online services of the US Geological Survey. Authoritative seismic vulnerability functions, however, are more problematic to acquire. They generally fall into three categories: empirical (derived from large quantities of historic loss data), expert opinion, and analytical (derived from mathematical models of structural response and construction contracting principles). Examples of empirical models include a study by Whitman et al. (1973) of building damage caused by the 1971 San Fernando earthquake, and various post-earthquake investigations presented in Steinbrugge (1982) and Steinbrugge and Algermissen (1990). When the Federal Emergency Management Agency (FEMA) wanted to develop an exhaustive set of vulnerability functions models for California, however, researchers working for the Applied Technology Council (1985) on ATC-13 found that inadequate earthquake experience data existed to create vulnerability functions for a wide variety of structure types, and resorted instead to the use of expert opinion. They applied a modified version of the Delphi Process to elicit and process that expert opinion in a rigorous, transparent way. Still, expert opinion can be seen to lack authoritativeness.

Analytical methods seem to hold the promise of developing seismic vulnerability functions for buildings that either have not yet experienced earthquakes or for which empirical loss data are not publically available, without relying on expert opinion. Important pioneering examples of analytical methods include work by Czarnecki (1973) and Kustu et al. (1982). More recently, researchers at Caltech and the Pacific Earthquake Engineering Research Center developed and applied second-generation performance-based earthquake engineering (PBEE-2) principles to derive vulnerability functions for several dozen particular woodframe, concrete, and steel buildings – see, e.g., Beck et al. (1999), Porter et al. (2001, 2002a,b), Krawinkler (2005), or Goulet et al. (2007). For sheer diversity of structure types and thorough coverage, however, nothing else compares with HAZUS-MH (Kircher et al. 1997; NIBS and FEMA 2003), which provides analytical seismic vulnerability information for a large number of engineered and non-engineered construction, including structure types present in the Balkans. Some work has been done to employ the HAZUS-MH methodology for additional structure types, e.g., Robinson et al. (2006) and McGowan (2009), who developed HAZUS-MH input parameter values for cavity-wall brick-masonry buildings and adobe buildings, respectively.

For more information, see (Porter, 2003).


19 SPA

TABLES

Table 1 Units of Aggregation for the Project

ExpNo.* City 5 Uralsk 6 Aksai 22 Temirtau 24 Karaganda 33 Kyzylorda 39 Almaty 40 Aktobe 60 Almaty oblast 69 Ust-kamenogorsk 85 Atyrau 88 Aktau 99 Taraz 111 Zhezkasgan 112 Balhash 125 Kokshetau 144 Kustanay 165 Pavlodar 167 Ekibastuz 180 Petropavlovsk 193 Semey 218 Astana 231 Akmola oblast 239 Shykment 999 Kyrgyzstan-Bishkek

*Number of location, for project purposes


20 SPA

Table 2 Example Expected Annual Loss (EAL) for various types of Construction, Almaty

Type of Construction EAL

Masonry 0.024 Masonry Contents 0.041 pre98RC 0.020 pre98RC Contents 0.038 post98RC 0.010 post98RC Contents 0.012 Unfired Brick 0.047 Unfired Brick Contents 0.046 Wood 0.011 Wood Contents 0.013 Avg Bldg 0.014 Avg Bldg Contents 0.021 Avg Bldg AND Contents 0.017


21 SPA

FIGURES

Figure 1 Elements of Probabilistic Earthquake Loss Estimation


22 SPA

Figure 2 Earthquake Loss Model Components

Risk Engine

Insured Loss


23 SPA

Figure 3 Kazakhstan Historic Seismicity

Figure 4 Kazakhstan Seismic Zones

Source: (Giardini, 1999)


24 SPA

Figure 5 Kazakhstan overlaid on Munich Re Earthquake Zones

Figure 6 Regional seismotectonics


25 SPA

Figure 7 Regional Historic Seismicity with major plate boundaries

Figure 8 Regional CMT focal mechanisms and plate boundaries


26 SPA

Figure 9 Vs30 Soil Conditions, Kazakhstan

Figure 10 Vs30 Soil Conditions, Almaty

See detail next figure


27 SPA

Figure 11 Buildings overlaid on Vs30 Soil Conditions, Almaty


28 SPA

Figure 12 Structure type distributions for Alma Aty region

Source: (JICA and AKIMAT, 2008)


29 SPA

Figure 13 Sample seismic vulnerability function from Porter (2009b)


30 SPA

Figure 14 Vulnerability Functions used in JICA Study

Source: (JICA and AKIMAT, 2008)


31 SPA

Figure 15 Vulnerability Functions Employed in this study


32 SPA

Figure 16 Example Hazard Curves, 24 locations

0.0001

0.001

0.01

0.1

1

-0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5

Annu

al P

roba

bilit

y of

Exc

eeda

nce

Response Spectral Acceleration (T = 1.0 sec)

5

6

22

24

33

39

40

60

69

85

88

99

111

112

125

144

165

167

180

193

218

231

239

999


33 SPA

Figure 17 Loss Exceedance Curves, various types of COM construction, various locations (0% deductible)

0.0001

0.001

0.01

0.1

0 0.2 0.4 0.6 0.8

Annu

al P

roba

bilit

y of

Exc

eeda

nce

Damage

5: Uralsk-Avg Bldg AND Contents6: Aksai-Avg Bldg AND Contents

22: Temirtau-Avg Bldg AND Contents24: Karaganda-Avg Bldg AND Contents33: Kyzylorda-Avg Bldg AND Contents39: Almaty-Avg Bldg AND Contents40: Aktobe-Avg Bldg AND Contents60: Almaty oblast-Avg Bldg AND Contents69: Ust-kamenogorsk-Avg Bldg AND Contents85: Atyrau-Avg Bldg AND Contents88: Aktau-Avg Bldg AND Contents99: Taraz-Avg Bldg AND Contents111: Zhezkasgan-Avg Bldg AND Contents112: Balhash-Avg Bldg AND Contents125: Kokshetau-Avg Bldg AND Contents144: Kustanay-Avg Bldg AND Contents165: Pavlodar-Avg Bldg AND Contents167: Ekibastuz-Avg Bldg AND Contents180: Petropavlovsk-Avg Bldg AND Contents193: Semey-Avg Bldg AND Contents218: Astana-Avg Bldg AND Contents231: Akmola oblast-Avg Bldg AND Contents239: Shykment-Avg Bldg AND Contents999: Kyrgyzstan-Bishkek-Avg Bldg AND Contents

Figure 18 Fraction of each building type in each Unit of Aggregation


35 SPA

COM 2% deductible loss RES 2% deductible loss IND 2% deductible loss Return Period (yrs) 50 75 100 125 150 200 50 75 100 125 150 200 50 75 100 125 150 200 Annual Prob. Of Exceedance 0.020 0.013 0.010 0.008 0.007 0.005 0.020 0.013 0.010 0.008 0.007 0.005 0.020 0.013 0.010 0.008 0.007 0.005

5: Uralsk 6: Aksai 22: Temirtau 24: Karaganda 33: Kyzylorda 39: Almaty 0.03 0.08 0.14 0.18 0.22 0.26 0.03 0.12 0.16 0.22 0.26 0.30 0.02 0.06 0.10 0.14 0.16 0.20 40: Aktobe 60: Almaty oblast 0.02 0.08 0.14 0.18 0.20 0.26 0.03 0.10 0.18 0.22 0.26 0.30 0.02 0.06 0.10 0.14 0.16 0.20 69: Ust-kamenogorsk 0.04 0.08 0.12 0.18 0.05 0.10 0.16 0.20 0.03 0.04 0.07 0.10 85: Atyrau 88: Aktau 99: Taraz 0.02 0.03 0.02 0.02 0.03 0.02 111: Zhezkasgan 112: Balhash 0.03 0.05 0.10 0.20 0.03 0.07 0.14 0.24 0.02 0.04 0.07 0.14 125: Kokshetau 144: Kustanay 165: Pavlodar 167: Ekibastuz 180: Petropavlovsk 193: Semey 218: Astana 231: Akmola oblast 239: Shykment 0.02 0.02 0.02 999: Kyrgyzstan-Bishkek 0.02 0.03 0.06 0.09 0.02 0.02 0.04 0.08 0.12 0.02 0.02 0.04 0.05

Figure 19 Annual Probability of Exceedance Losses (APELs) for 2% deductibles, for COM, RES and IND occupancies for 24 locations for several ‘return periods’.


36 SPA

Figure 20 COM “200 year” losses overlaid on historic seismicity, showing many parts of Kazakhstan with zero or near-zero losses, but South and East Kazakhstan have substantial seismic risk, particularly Almaty.


37 SPA

Figure 21 Radius of 200 km drawn about Almaty City. Red, blue and green circles are reverse, normal and transform earthquake mechanisms, resp. – the predominance of red events indicates primarily a reverse or thrust environment. A 200 km long reverse fault would correspond approximately to a Mw 8.2

earthquake, similar to the 1911 Kemin event.


38 SPA

Figure 22 Isoseismals (MSK) of 1911 Mw 8.2 Kemin earthquake – Almaty City is MSK 8+, while Almaty Oblast generally MSK 6-8 and Bishkek is MSK

6+.


39 SPA

Figure 23 Isoseismals (MSK) of 1889 Mw 8.3 Chilik earthquake over historic seismicity – Almaty City is MSK 8, while Almaty Oblast generally MSK 7-8

and Bishkek is MSK 6+.


40 SPA

Figure 24 Thumbnail view of powerpoint presentation prepared for the project, for use by World Bank personnel in Kazakhstan.

Kazakhstan: Strengthening Catastrophe Risk Transfer ...siteresources.worldbank.org/FINANCIALSECTOR/Resources/World-Ban… · Strengthening Catastrophe Risk Transfer Supervision FIRST

Documents