Seismic risk and loss assessment for Bucharest, Romania

Seismic risk and loss assessment for Bucharest, Romania

Sergio Molina, Dominik Lang, Conrad Lindholm & Stefan Balan Introduction

The city of Bucharest, capital of Romania, comprises around 2 million inhabitants and a considerable number of high-risk buildings and infrastructure facilities of nationwide importance. Economically, Bucharest is the most prosperous city in Romania and among the main industrial centers and transportation hubs of Eastern Europe. As the most developed cities in Romania, Bucharest hosts a wide range of educational facilities.

Administratively, the city proper is known as the ‘Municipality of Bucharest’ being divided into six sectors as illustrated by Figure 1. The geological setting of the city is characterized by the presence of deep sedimentary and partly soft deposits which are known to imply significant amplification effects of earthquake ground motion. Bucharest is located in the Vrancea region which is characterized by a multitude of different earthquake types. Figure 2 demonstrates some of the features of the seismic activity of the Vrancea zone where the most important is the frequency of earthquakes in the depth range 100–150 km.

To quantify the region’s earthquake hazard a peak ground acceleration of around 0.20 g for a rock site with a 10% exceedance probability in 50 years is predicted by the Global Seismic Hazard Assessment Program (GSHAP; Giardini et al., 1999).

The largest recent earthquake in the Vrancea region happened on March 4, 1977 with a moment magnitude Mw 7.3 at a hypocentral depth of 90 km. This event caused significant structural damage to the building stock in Bucharest and surroundings. Due to the political circumstances in Romania during that time (era?), a thorough documentation of occurred building damage and losses has not been conducted and is consequently not available for comparative studies. The earthquake threat of the city of Bucharest has been mapped and modeled by several investigators, as e.g. Radulian et al. (2000), Sokolov et al. (2004), Musson (2000), Mantyniemi et al., (2003), Kienzle et al. (2006), Enescu and Enescu (2007), among others.

The main goal of this article is the computation of seismic risk scenarios for the city of Bucharest using the SELENA software (Molina and Lindholm, 2007; Lang et al., 2008; Molina et al., submitted) considering a repeat of the 1977 Vrancea earthquake. The main differentiating factor from previous risk studies for Bucharest consists in the application of the capacity spectrum method (CSM) and logic tree computation scheme to account for uncertainties in the scenario parameterization. Thus, a sensitivity study of the results to the size of the earthquake and the related performance point identification on the capacity curve is facilitated.

1

Figure 1. Overview map of the city of Bucharest with its six main city sectors (areas filled with bright color) which are extended to the administrative city limits (areas filled with toned color).

Vrancia PDE earthquakes0 9 18 27 364.5

Kilometers

:

"Bucharest

28°0'0"E

28°0'0"E

27°0'0"E

27°0'0"E

26°0'0"E

26°0'0"E

25°0'0"E

25°0'0"E

46°0'0"N 46°0'0"N

45°0'0"N 45°0'0"N

Robinson ProjectionCentral Meridian: 30.00

Legend

" Bucharest

Depth range: Deeper than 200 kmDepth range: 150 - 200 kmDepth range: 100-150 kmDepth range: 50-100 kmDepth range: Less than 50 kmRiversLakes

Figure 2. The city of Bucharest and its distance to the earthquake-active Vrancea zone. Note the different colors reflecting on focal depth ranges of the PDE (what is PDE?) located earthquakes between 1973 and 2009.

legend cannot be read

2

margin of Bucharest needs to be sketched

Compilation of input and inventory data for seismic risk scenarios

Building inventory

A complete review of the building inventory and population database was conducted for Romania in the year 1999 (Marmureanu, 2007). For the current study this database was reviewed and rearranged in terms of city distribution and prevalent building typologies. The provided database subdivides the city of Bucharest into six sectors and comprises a total of 117,592 individual buildings (i.e. 813,075 living units) and a population of 1.9 million inhabitants.

A distribution of the building inventory and population numbers over the six sectors of Bucharest is given in Table 1. However, the building inventory only takes into consideration residential buildings neglecting residential, industrial or other occupancy types. Consequently, the seismic risk computation can only be conducted for buildings of residential occupancy along with casualty estimation. The damage and loss results will be computed on the level of city sector which will serve as the minimum geographical entity.

Table 1. Distribution of the population and building inventory (residential only) in each of the six sectors of Bucharest.

City sector

Population Number of buildings

Building floor area A [m2]

Average no. of people/building

Average no. of m2/person

1 230,724 30,596 6,650,008 7.5 28.8 2 370,436 25,845 9,765,130 14.3 26.4 3 404,804 13,604 10,185,345 29.8 25.2 4 304,285 11,503 7,648,074 26.5 25.1 5 277,138 27,201 7,549,997 10.2 27.2 6 372,008 8,843 9,178,462 42.0 24.6

Total 1,959,395 117,592 50,977,016 – –

The available building database contained several parameters that could be used for the subdivision into building classes: main construction material; building class; period of construction; height class; number of dwellings (i.e. number of apartments or living units) and number of occupants. The information given in the database was combined with the experience of local engineers and provided basis for a building typology classification according to the main structural material so that seven ‘Bucharest’-specific building typologies were initially defined (Table 2). These seven main residential typologies were further subdivided into 31 model building types for which damage computations are conducted and which establish the basis for the loss computation below.

3

Table 2. Classification of the building typologies prevalent in the city of Bucharest and assignment with ‘HAZUS’ model building types (mbt).

Level of code design ‘Bucharest’ mbt Description HAZUS

mbt 1) No. of stories Post-1978 Pre-1978

Index 2)

M1A-LM C2L 1–3 M1A-LL

M1A-MM C2M 4–7 M1A-ML M1A-HM

M1A RC shear walls

C2H ≥ 8

moderate-code low-code

M1A-HL C1L 1–3 M1B-LM

M1B-MM C1M 4–7 M1B-ML M1B-HM

M1B RC large panel

C1H ≥ 8

moderate-code low-code

M1B-HL M1C-LL C3L 1–3 M1C-LP M1C-ML C3M 4–7 M1C-MP M1C-HL

M1C RC frames

C3H ≥ 8

low-code pre-code

M1C-HP M2-LM RM2L 1–3 M2-LL M2-MM RM2M 4–7 M2-ML M2-HM

M2 brick masonry with rigid floors

RM2H ≥ 8

moderate-code low-code

M2-HL M3-LM RM1L 1–3 M3-LL M3-MM M3

brick masonry with flexible

floors RM1M 4–7

moderate-code low-code

M3-ML M4-LM M4 wood W1 1–3 moderate-

code low-code M4-LL M5-LL M5 adobe URML 1–3 low-code pre-code M5-LP

1) a detailed description of the ‘HAZUS’ model building types are given in FEMA (2003) 2) the final index for the 31 model building types is composed of a letter combination between the index

for the Bucharest model building type (M1 to M5), the height range of the structure (low-rise L, mid-rise M, high-rise H) and the level of anti-seismic code design (pre-code P, low-code L, moderate-code M)

The seven main building typologies were assigned according to the descriptive keywords indicated in Table 2 and after consultation of local structural engineers. The highest number of buildings belongs to model building types M2, M3 and M5 which were mainly erected before 1970 (Figure 3). As Figure 3 illustrates, these buildings at the same time have relatively small building floor areas (in [m2]) which is caused by the fact that the majority is characterized by smaller story numbers and smaller plan dimensions. Consequently, buildings of M2, M3 and M5 typology represent 87% of the total number of individual buildings but only some 34% of the total building floor area.

The current study did not allow for a thorough development of capacity curves and corresponding vulnerability functions customized for each of the 31 model building types. Therefore available building fragility information which was initially developed for other

4

building typologies, e.g. HAZUS model building types (FEMA, 1999; 2001; 2003), were assigned to the building typologies prevalent in the Bucharest study area according to descriptive features given in Table 2.

Figure 3. Distribution of the seven main model building types in Bucharest in terms of number of buildings and building floor area. Demographic information and casualty model

In addition to structural information, the available database (Stefan must find the proper reference) on the census of Bucharest also includes the number of inhabitants (occupants) residing in each building category and the total number of inhabitants per city sector. Based on this information criteria were established to estimate day- and night time occupancy rates, i.e. the percentage of people that are indoors and outdoors as function of the time of day.

To estimate the number of people who are indoors at different times of the day the following simplified model was used:

1. Daytime scenario (10 am): During daytime (working) hours 10% of the population stay indoors (DRES). 2. Nighttime scenario (2:00 am): During nighttime 90% of the population reside in their homes (NRES). 3. Commuting time scenario (5:00 pm): During commuting hours it is estimated that 50% of the population will be outdoors (COMM).

The absolute numbers of indoor occupancy (DRES, NRES, COMM) are given in Table 3. This will serve as the basis for estimating the number of casualties as a direct effect of building damage and collapse.

A small percentage of injuries will additionally be inflicted on people staying outside but being injured by e.g. falling objects (FEMA, 1999)). The number of injured persons is calculated according to four disaggregated severity levels ranging from slight (severity level

5

1), moderate injury (severity level 2), heavy (severity level 3) to fatal injuries (severity level 4).

Table 3. Indoor occupancy numbers in the six city sectors during different times of day.

City sector

Population Daytime indoor occupancy (DRES)

Nighttime indoor occupancy (NRES)

Indoor occupancy during commuting time (COMM)

1 230,724 23,072 207,652 115,362 2 370,436 37,044 333,392 185,218 3 404,804 40,480 364,324 202,402 4 304,285 30,428 273,856 152,142 5 277,138 27,714 249,424 138,569 6 372,008 37,201 334,807 186,004

Economic loss model There is at present no detailed economic model for Bucharest available that can be used

to precisely determine the repair and replacement costs for different building types.. For the year 2006, the estimated average construction price (AP) per square meter in the city of Bucharest was estimated to 975 Euro. A more precise estimate as a function of building typology could not be elaborated, and hence a uniform replacement value of 975 Euro/m2 was used as a best estimate average.

The repair or replacement costs naturally depend on the damage state of the structure. As provided by ATC-13 (ATC, 1985) and later adopted by FEMA (2003), the direct economic losses are expressed in terms of 2 % (for slight damage state), 10 % (for moderate damage state), and 50 % (for extensive damage state) of complete replacement costs for the respective building.

The economic loss model used is independent of model building type but varies slightly with the height range of the building considering that the relative repair cost of a certain damage at a larger building (including many apartments) is cheaper than the repair of a similar damage at a smaller building (with only one living unit). Having this in mind, a slight reduction factor was introduced as given in Table 4 as a function of building height range. These rates are similar to those applied in a recently published paper on seismic risk for Naples (Lang et al., 2008).

Table 4. Relation between height of the building and repair/replacement cost.

Height range Related repair costs low-rise 100 % mid-rise 95 % high-rise 90 %

Definition of seismic risk scenarios for Bucharest

6

In general, risk is defined as the destruction or expected loss obtained from the convolution of probability of occurrence of hazardous events and the vulnerability of the exposed elements to a certain hazard (United Nations Disaster Relief Organization). According to McGuire (2004), seismic risk entails a set of events (earthquakes likely to happen), the associated consequences (damage and loss in the broadest sense), and the associated probabilities of occurrence (or exceedance) over a defined time period. Thus, seismic risk can be expressed as the convolution of seismic hazard, exposure and vulnerability. For a deterministic analysis, seismic hazard refers to the shaking effects at a certain site caused by a scenario earthquake. . While the term exposure represents the availability and inventory of physical buildings and infrastructure facilities in the respective study area being subjected to a certain seismic event, structural (i.e. physical) vulnerability stands for the susceptibility of each individual element (building, infrastructure, etc.) to suffer damage in dependence on the level of earthquake shaking. This resulting in direct structural and non-structural damages which involve economic losses and human casualties.

The SELENA approach There are basically two approaches in modern earthquake risk estimation: a) The

empirical approach (sometimes called the “intensity” approach) which is based on expert judgment and reports from historical earthquakes and b) the analytical approach (sometimes called “spectral displacement” approach) which is based on analytical computations of the behavior of the structure exposed to lateral forces. The SELENA tool is capacity spectrum based and analytical, and since the core computation is similar to HAZUS-MH we could use the structural vulnerability functions developed under FEMA (1999).

The height of the buildings in Bucharest can generally reach up to 10 stories. The non-quantitative terms used for height are defined as follows: Low Rise (≤ 3 stories), Mid Rise (4 – 7 stories) and High Rise (≥ 8 stories). As an approximation for the Bucharest seismic risk scenario, we also classified the buildings in two time periods of construction related with the design level of the buildings. Buildings constructed after 1975 were assumed to be designed with moderate earthquake resistant design (moderate code) and building constructed before 1975 were assumed to have a low code design level. This distinction only applied to the M1C and M5 classes.

a) Deterministic earthquake scenarios for Bucharest

Statistics based on historical earthquake records (Oncescu et al., 1999) indicate that three to five destructive subcrustal earthquakes do occur in the Vrancea region per century, whereas Bucharest is the main city exposed to these strong events with intermediate hypocentral depth between 60 and 180 km.

7

Table 5 lists the main historical events of the Vrancea focal region which caused structural damage to the city of Bucharest. For those events which occurred in the last century more detailed information is available both on the focal parameters and the effects to the building stock in Bucharest (after Oncescu et al., 1998, Mândrescu and Radulian, 1999).

Table 5. Historical events in the Vrancea region which caused damage in the city of Bucharest (Oncescu et al., 1998; Mândrescu and Radulian, 1999).

Date Latitude (N)

Longitude (E) Magnitude Focal depth

h [km] 1681, August 19 n.a. n.a. MW 7.1 n.a.

1738, June 11 n.a. n.a. MW 7.7 n.a. 1802, October 26 n.a. n.a. MW 7.9 n.a.

1829, November 23 and 26 n.a. n.a. MW 7.3 n.a. 1838, January 11 n.a. n.a. MW 7.5 n.a.

1940, November 10 45.80 26.70 MW 7.7 150 1977, March 4 45.77 26.76 MW 7.4 94

1986, August 30 45.52 26.49 MW 7. l 131 1990, May 30 45.83 26.89 MW 6.9 91

For the present study a repeat of the 1977 Vrancea earthquake is assumed to derive

damage and loss estimated for Bucharest. The focal parameters (depth and magnitude) of this event delivered by different agencies showed significant variations. ISC Bulletin provided the highest magnitude of MW 7.4..

Since the seismic risk assessment tool SELENA is based on a Logic Tree computation scheme which accounts for uncertainties connected to any input parameter, the variations in magnitude and focal depth can be accounted for by the assignment of different weighing factors. Table 6 illustrates the definition of the Logic Tree with nine scenario hypocenters having different magnitude-depth combinations but the same epicentral coordinates.

Table 6. Scenario Earthquakes with corresponding weight used for the seismic risk computation.

Branch Weighting factor Latitude Longitude Magnitude Focal depth h [km]

1 0.09 60 2 0.12 90 3 0.09

7.20 180

4 0.09 60 5 0.12 90 6 0.09

7.30 180

7 0.12 60 8 0.16 90 9 0.12

45.77° 26.76°

7.40 180

8

b) Empirical ground motion prediction equations A variety of ground motion prediction equations for intermediate-depth subcrustal

earthquakes in the Vrancea region was proposed by Marmureanu et al. (2006). These relationships are based on ground motion records of the major Vrancea earthquakes of 1977 (MW 7.4), 1986 (MW 7.1) and 1990 (MW 6.9 and MW 6.4).

Soil amplification effects were removed from the records so that the obtained relations provide PGA predictions at rock outcrop. Marmureanu et al. (2006) proposed the following five attenuation relationships as suitable for intermediate-depth subcrustal earthquakes of the Vrancea region in which A is PGA in units of gravity [g], MW is moment magnitude, Rh is hypocentral distance, and Δ is epicentral distance:

ln A = –2.8042 + 1.1804 MW – 1.4553 ⋅ ln(Rh + 100) with σln A = 0.463 (1)

ln A = –6.4789 + 1.1896 MW – 0.8870 ⋅ ln(Rh) with σln A = 0.468 (2)

ln A = –8.0615 + 0.9756 MW – 0.3204 ⋅ ln(Δ) with σln A = 0.495 (3)

ln A = –9.9261 + 1.1664 MW – 0.005259 ⋅ Rh with σln A = 0.458 (4)

ln A = –9.0056 + 0.9669 MW – 0.003672 ⋅ Δ with σln A = 0.474 (5) In order to test different scenarios with varying depth we focus on those equations

depending on hypocentral distance Rh (rather than epicentral distance Δ) and with low sigma values. The computation of ground motion parameters is finally based on equation (1) both for mean value and for mean value plus one sigma.. Both ground motion prediction equations are assigned equal weighting factors (0.5) for the Logic Tree computation scheme. By the consideration of mean value plus one sigma value in addition to the mean value it is tried to avoid the calculation of under-conservative ground motion values.

The large scatter of strong ground motion especially from large earthquakes is well known. In probabilistic seismic hazard computations it is therefore customary to add 2 or 3 sigma in the integration procedure in order to avoid under-conservative estimates. A similar thinking lies behind the choice here.

c) Local site conditions for Bucharest

Bucharest is located in the central part of the Moesian Platform at an average epicentral distance of about 140–170 km from the Vrancea region. The relief of the city is generally plane with a slight dipping towards southeast (which is followed by the Danube river). The Dâmboviţa and Colentina rivers divide the city into several morphological units: Bucharest Plain (Dâmboviţa–Colentina interstream), Băneasa–Pantelimon Plain, Cotroceni–Văcăreşti Plain, and the meadows along the above mentioned rivers (in parts cited from Moldoveanu and Panza, 1998).

9

The soils of the Bucharest area are represented by Quaternary deposits (Middle and Upper Quaternary). Even though detailed knowledge of the soils of Bucharest exists, a simplified model has been used in the present study. The soils and sedimentary thicknesses vary from north to south. However, the upper 30 meters of soil stratigraphy are relatively homogeneous throughout the city. Based on information taken from Ciugurdean and Stefanescu (2006) ground type C according to the Eurocode 8 classification scheme is assumed for the whole city. Eurocode 8 ground type C is characterized by deep deposits of dense or medium-dense sand, gravel of stiff clay with shear wave velocities vs,30 ranging from 180 to 360 m/s.

Following the provisions of Eurocode 8, elastic design spectra for soil type C are provided to represent seismic ground motion in the different geographical units of Bucharest. Taking into consideration the aforementioned variations in earthquake source (magnitude M and focal depth h) and whether standard deviation sigma is considered in the chosen ground motion prediction equation (1), different values of peak ground acceleration PGA are derived which are used to up- or downscale the respective design spectrum. Figure 5 illustrates the design spectra for those two magnitude-focal depth combinations which lead to lowest and highest ground motion values.

The PGA value for the lower design spectrum represents the mean value derived by ground motion prediction equation (1), while PGA for the higher design spectrum reflects mean value + one sigma using the same equation.

Figure 5. Elastic design spectra following the provisions of Eurocode 8 – Type 1 (Ms > 5.5) for soil class C. Both spectra are scaled to the lowest and highest PGA values considered for the computation, respectively.

SEISMIC RISK SCENARIOS USING SELENA: RESULTS AND DISCUSSION

a) Damaged buildings and built area

Figure 6 shows the cumulative damaged area including all damage states for the four most important building types for each sector. It is seen that the damage distribution is quite homogeneous across city sectors due to the similar ground shaking (small hypocentral

10

distance differences). The model building type suffering most damage was M1C while the model building type suffering least damage was M2.

In Table 7 we show a comparison between the two computational methods (CSM and

MADRS) in terms of damage percentage of built area. The CSM method is described in ATC-40 and the MADRS method is described in detail in FEMA 440. From Table 7 it is seen that the MADRS method yield consistently higher damage results than the CSM methodology. The same results are concluded from Table 8.

11

Table 7. Damage distribution in terms of % of built area for all building types in all in sectors for the CSM and MADRS methodologies respectively. See text for description on the CSM and MADRS methods. The table is cumulative over all 31 building typologies. Damage (% of total built area)

At least slight At least moderate

At least extensive

Complete

CSM 66.4 41.5 14.0 3.4 MADRS 78.6 59.0 26.9 8.9

Table 8. Damage distribution in terms of damaged buildings for all building types in all sectors for the CSM and MADRS methodologies respectively. See text for description on the CSM and MADRS methods. The table is cumulative over all 31 building typologies. Damage (number of buildings)

At least slight At least moderate

At least extensive

Complete

CSM 82813 62934 31750 9612 MADRS 79052 60820 32030 10322

The differences in predicted damage exemplify the important and inherent uncertainty in risk computations even when using very similar methodologies. The MADRS methodology can be considered as a second generation CSM method, and as such should be initially considered as the best of the two methods.

Figure 6. Damage in % of built area across city sectors and four dominant building typologies.

12

Figure 7.

b) Comparing model results with historical damage

The reports from the 1977 Vrancea earthquake indicated that approximately 33 000 building were seriously damaged or destroyed (Balan et al., 1982).The results obtained using both methods (Table 8) indicated about 30 000 buildings suffering at least extensive damage, and this number is in perfect agreement with the 1977 historical evidence.

There is, however, one important caveat in such comparisons: Following a large earthquake many of the buildings are restored and remain in the building database, and on the other side many new buildings are constructed over the past 30 years, and these are largely using different and improved building techniques. With this caveat one should therefore not attempt to put too much confidence in the small comparison above.

c) Economic losses

We have again attempted to estimate the total economic losses using the models described above and also here comparing the MADRS and the CSM methods. Firstly in Figure 7 it is seen that the losses in each of the six sectors are not widely different. This points to some homogeneity in terms of building vulnerability. Beyond that it is seen that sectors 2 and 3 are the ones with the highest economic losses independently of the method, however, when using the MADRS method the economic losses are approximately 2 billions Euro higher compared to the CSM method.

13

Figure 7. Median monetary loss for each of the Bucharest sectors.

The economic losses obtained through the use of the MADRS method (10,001 ± 2,266 millions Euros) and the CSM method (8,197 ± 1,738 millions Euros) are close to the reported economic losses of 2 billions US dollars (currency value of 1977) taking into consideration the change in the European economy since 1977 and the increase of the prices after the introduction of the Euro. However, also here one should remember the caveats in the above comparison between modelling results and the historical losses in 1977.

d) Number of deceased persons

We have computed the number of deceased people with the earthquake occurrence at three different time-of-day scenarios: A midnight earthquake (02:00) when most of the people are supposed to be at home; A working hour (10 am) earthquake when most of the people are outside of the residential buildings and a commuting hour (5 pm) scenario when many people are travelling to the residential buildings after work.

As expected the maximum number of deceased (we are below only focusing on injury

class 4) are found with a midnight earthquake (Figure 8) adding to a total of 1396 ± 470 dead (0.07% of total population) when using the CSM method and 1958 ± 649 dead (0.10% of total population) when using the MADRS method. Since this seismic risk scenario only considers residential buildings (the only available) the number of dead people is at a minimum for a working hour earthquake (10 am) reaching a total of 95 ± 32 dead when using the CSM method and 134 ± 44 dead when using the MADRS method. Finally, an earthquake happening when people have started to arrive to the residential buildings after work (5 pm) is predicted to cause a total of 572 ± 190 deaths with the CSM method and 804 ± 266 deaths with the MADRS method.

14

Figure 8. Human deaths in each Bucharest sector as computed with the CSM and MADRS methods (only median value is represented). When we compare these results with the fatalities of the 1977 Vrancea earthquake we observe that the CSM method yields human losses that are only slightly lower than the number reported from the 4 March 1977 earthquake (when 1424 people were reported killed in Bucharest city). It should be noted that the 4 March 1977 earthquake occurred at 19:21 GMT, comparing to 21:21 local time, which should be comparable to a night time scenario. The reported human losses from the Vrancea earthquake are clearly within the confidence limits of the model results provided by both the CSM and the MADRS methods. However, since 1977 the population of Bucharest has increased dramatically, and as warned above the good correlation between historical and predicted deaths should not be over interpreted.

SUMMARY and CONCLUSIONS To the best of our knowledge the present study represents a first ever quantitative risk

study for Bucharest using the capacity spectrum method and applying vulnerability functions to 31 building classes. We have developed a model in terms of population and building inventory that can also be applied in a real time scenario context allowing the static model parameters and waiting for the dynamic shaking (which will depend on magnitude, focal depth and distance of the causative earthquake).

The experience throughout the work has indicated that the applied method is

convenient and straight forward to implement once the important building and population data are made available, and both the CSM and the MADRS methodologies are adequate, possibly with a small preference for the MADR methodology.

From the obtained results the following conclusions can be drawn:

15

Although both methods give results of similar order in all of the cases, the damages and losses using CSM are lower than the ones obtained with MADRS. Regarding the number of buildings damaged and the deceased people, the results of the scenarios are “acceptable” for today’s situation, taking into consideration that quite some changes in the population and building stock have taken place since 1997 when the basis building and population information used in this scenario was established.

Taking into consideration that ground motion from the different scenarios is relatively similar in all the Bucharest sectors, the damage distribution is also similar in all sectors. For the prevalent model building types in the city it seems that the model building type M1C shows the highest damage and the model building type M2 shows the lowest damage.

When using a simple economic loss model, both the CSM and the MADRS methods provide results close to the economic losses reported after the 1977 Vrancea earthquake, taking into consideration the increase of the damaged area and the change of the economy between 1977 and 1997 (year of the database establishment).

Finally, SELENA has been proved as a useful computation tool in order to estimate seismic risk scenarios after calibrating the results with the reported damages (buildings, economic and human losses) from the 1977 Vrancea earthquake and to analyze the sensitivity of the results to changes in the magnitude and depth of the earthquake.

ACKNOWLEDGEMENTS

The work was partly financed through the Seismic eArly warning For EuRope (SAFER) project under the EU contract N. 036935, and through the International Centre of Geohazards (ICG) financed through Norges Forskningsråd (NFR). REFERENCES (not yet alphabetical) Giardini D. (1999): The Global Seismic Hazard Assessment Program (GSHAP). Annali di

Geofisica, Vol. 42, No. 6 Radulian M., F. Vaccari, N. Mandrescu, G. F. Panza and C. L. Moldoveanu (2000): Seismic

Hazard of Romania: Deterministic Approach, J. of Pure and Applied Geophys. Issue Volume 157, Numbers 1-2

Sokolov V., K. Bonjer and F. Wenzel (2004): Accounting for site effect in probabilistic assessment of seismic hazard for Romania and Bucharest: a case of deep seismicity in Vrancea zone. Soil Dynamics and Earthquake Engineering. Vol. 24, No. 12, pp. 929-947

Musson R. (2000): Generalized Seismic Hazard Maps for the Pannonian Basin Using probabilistic Methods. Pure appl. geophys. 157, pp. 147–169

Mantyniemi P., V. I. Marza, A. Kijko and P. Retief (2003): A New Probabilistic Seismic Hazard Analysis for the Vrancea (Romania) Seismogenic Zone. Natural Hazards 29: pp. 371–385

Kienzle A., D. Hannich, W. Wirth, D. Ehret, J. Rohn, V. Ciugudean and K. Czurda (2006):

A GIS-based study of earthquake hazard as a tool for the microzonation of Bucharest. Engineering Geology. Vol. 87, pp. 13-32

16

17

Molina S. and C. Lindholm (2007): Estimating the confidence of earthquake damage scenarios; examples from a logic tree approach. Journal of Seismology, 11, pp. 299-310.

Lang D. H., S. Molina Palacios, C. Lindholm (2008): Towards near real-time damage estimation using a CSM-based tool for seismic risk assessment. Journal of Earthquake Engineering. Vol. 12, pages 199 – 210, DOI: 10.1080/13632460802014055

FEMA (1999): HAZUS®99: Earthquake Loss Estimation Methodology: User’s Manual, Federal Emergency Management Agency, Washington, DC, Unites States.

FEMA (2001): HAZUS®99 Estimated Annualized Earthquake Losses for the United States (FEMA 366), Federal Emergency Management Agency, Washington, DC, Unites States, February 2001, 33 pp.

FEMA (2003): HAZUS-MH: Multi-hazard Loss Estimation Methodology, Federal Emergency Management Agency.

Oncescu, M.-C., Mârza, V.I., Rizescu, M., Popa, M. (1999). The Romanian earthquake catalogue between 984 - 1996, Vrancea Earthquakes: Tectonics, Hazard, and Risk Mitigation, Editors: Wenzel, F., Lungu, D., Kluwer Academic Publishers, 43-48.

Marmureanu, G. (2007). Internal Report. National Institute for Earth Physics (NIEP), Bucharest, Romania.

Mândrescu, N., Radulian, M. (1999), Seismic microzoning of Bucharest (Romania): a

critical review, Vrancea Earthquakes. Tectonics, Hazard, and Risk Mitigation, Editors: Wenzel, F., Lungu, D., Kluwer Academic Publishers, 109-122.

Moldoveanu C.L. and Panza G.F. (1998). Vrancea source influence on local seismic response in Bucharest. In The Abdus Salam international centre for theoretical physics, Report IC/98/209, Miramare, Trieste, pp 1- 28.

Applied Technology Council ATC (1996). Seismic Evaluation and Retrofit of Concrete Buildings, Report ATC-40, Redwood City, California.

Applied Technology Council ATC (2005). Improvement of Nonlinear Static Seismic Analysis Procedures, FEMA-440, California, USA.

European Committee for Standardization CEN (2002): prEN 1998-1:200X, Eurocode 8: Design for structures for earthquake resistance, Part 1: General rules, seismic actions and rules for buildings. May 2002.

Enescu D. and B.D. Enescu (2007). A procedure for assessing seismic hazard generated by Vrancea earthquakes and its application. iii. A method for developing isoseismal and isoacceleration maps. Applications. Romanian Reports in Physics, Vol. 59, No. 1, pp. 121–145

Ciugurdean, V., and Stefanescu, I. (2006). Engineering geology of the Bucharest city area, Romania. Paper no. 235 subm. IAEG-2006, Engineering Geology for Tomorrow's Cities.

Balan, S., Cristescu, V., Cornea, I. (coordination) et al. (1982). The Earthquake of March 4, 1977 from Romania. Romanian Academy Printing House.