NATO Science for Peace and Security Programme CLOSING CONFERENCE OF THE NATO SfP 983054 (BSHAP) PROJECT Harmonization of Seismic Hazard Maps for the Western Balkan Countries October 23, 2011 Ankara, Turkey PSHA results for the BSHAP region Neki Kuka Institute of Geosciences, Energy, Water and Environment Polytechnic University of Tirana
Mathematical Background Ø Mathematical background background:: total probability theorem (Cornell 1968 1968)). P(U > u ) = òò mmax m0 R P(U > u ) m, R) f M (m) f R ( R) dm dr P(U>u m,r): probability that ground motion level u will be exceeded, P(U>u m,r): given an earthq earthq.. of mag mag.. m on a source at distance R from the site site;; fm(m) (m):: probability distribution of magnitude, fr(r) (R):: probability distribution of site site--source distance distance;; m0, mmax: lower and upper bound of magnitude, m0<m< mmax. Ø Summation over all seismic sources gives the anual rate of ground motion exceedance exceedance:: n mu = å i= 1 n mi (u ) = å i= 1 l i (m0 ) ò ò R mmax m0 P(U > u ) m, R ) f M (m)i f R ( R)i dm dr λi(m0): the annual rate of occurrence of earthquakes with m m0. Assuming earthquake process is Poissonian Poissonian,, prob prob.. that level u will be exceeded at least once during the time period T: PT (U > u) = 1- e- m T u
Mathematical Background Ø Except probability distributions of the magnitude fm(m), and site site--to to-source distance fr(r) (R),, implementation of the PSHA also requires requires:: A spatial model to describe reliably the seismic activity rate λi(m0). A model to characterize the process that generates the ground motion P(U>u m,r) P(U>u m,r).. fm(m) is characterized by doubly doubly--truncated exponential model model:: b e- b ( m- m0 ) f M ( m) mo, mmax ) =, m0 m mmax - b ( mmax - m0 ) 1- e Ø λi(m0): different approaches proposed, are distinguished on the way they model seismicity within an area area:: assuming the earthquake rate of occurrence is uniform throughout (source zone approach), or considering it as spatial variable (gridded seismicity seismicity:: spatially smoothed, kernel methodology methodology)).
Mathematical Background P(U>u m,r): Assessment of the conditional exceedance probability of a specified level u of seismic intensity U on a certain site, requires an adequate model to predict the ground motion generated at a site when on a source zone has occurred an earthquake with magnitude m and distance R from this site: F U (u*): value of the cdf of the random variable U that corresponds to the fixed values of mag. m, and the distance to the eq. source r. Usually U is characterized by the lognormal law. In this case: µ U: conditioned mathematical expectation of ln(u) for a fixed (m,r); Φ: cdf of the standardized normal distribution N(0,1); σ lnu : logarithmic standard deviation of the regressive model (PGME). Generally: P U u m r F u * * ( >, ) = 1 - U ( ) * * ln u - mu P( U > u m, r) = 1 - F ( ) s ln( U ) = b + f ( M ) + f ( R) + f ( M, R) + f ( S) + f ( F) + e lnu 0 1 2 3 4 5
Seismicity Modeling Generally, four different classes of earthquake source models are used for seismic hazard assessment (NSHM, USGS, 2008): 1) Smoothed gridded seismicity, 2) Uniform background source zones, 3) Geodetically derived source zones, 4) Faults. The first two models are based on the earthquake catalog and characterize the hazard from earthquakes between M5 and M6.5-7.0. The geodetically derived source zones are used to assess the hazard between M6.5 and the largest potential earthquake in a region. Faults mostly contribute to the hazard for earthquakes stronger than M6.5.
Seismicity Modeling Fault sources an geodetically derived source zones Within the frame of the BSHAP project, it has been impossible to provide the relevant information, even for the faults that have generated strongest earthquakes with Mw> 6.5 in our region. Seismic hazard assessment is accomplished using the smoothed gridded seismicity methodology. Random seismicity-derived sources account for two types of earthquakes: 1) those that occur off known faults, and 2) moderate-size earthquakes that are not modeled on faults. The gridded-seismicity models are based on historical earthquakes and account for the observation that stronger earthquakes occur at near clusters of previous smaller earthquakes.
Application of PSHA requires: PSHA implementation (BSHAP) a homogenous catalog of historical earthquakes, a description of possible faults and earthquake sources, the parameters describing adequately seismicity for faults and earthquake sources, appropriate PGM models in region. The problems related with compiling of a homogenous earthquake catalog - unified in terms of M W mag. scale, completeness magnitude levels, declustering, etc., were just presented by Prof. Duni. Dr. J. Mihaljevic also presented in details the identification, delineation and characterization of the seismotectonic source zones in the BSHAP project region.
Estimation of recurrence statistics To calculate the hazard from a particular source, a doubly-truncated exponential model for G-R magnitude-frequency distribution is used: l m exp[ - b( m - m0)] - exp[ - b ( mmax - m0 )] = l m 0 1- exp[ - b( m - m )] max 0, m 0 < m< m max λ m : the mean annual number of earthquakes with M m. λ m0 : the mean annual number of earthquakes with M m 0. m 0 : minimum magnitude with engineering interest (m 0 =4.0 is used) m max : maximum magnitude that can be generated in a seismic source. The recurrence statistics (a- and b-values, λ m0 ) are obtained from analysis of the BSHAP catalog, using a MLE method that accounts for variable completeness (Bollinger et al. 1993; Weichert 1980, Berril and Davis 1980). The recurrence statistics are estimated for about 70 source zones.
Estimation of maximum magnitude M max is a difficult parameter to be assessed because the database to derive it is statistically very limited. M max, should be relatively large, because big earthquakes may have a very large interval occurrence, that sometimes exceeds 10000 years and probably are not evidenced in the historical or geological documents. Estimation of M max to some extent should reflect the uncertainties that associate this parameter. M max used in our hazard calculations, is source zone dependent. Its assessment is based considering the maximum observed magnitude in a seismotectonic zone and the respective geological settings. For every zone, we have accepted the estimations given by each partner country for the relevant zones.
The doubly-truncated G-R exponential model is characterized by three parameters: λ m0, b, and m max. Interactive fitting 0f the G-R model After their assessment, a final manual check and tuning is applied for every zone, inspecting carefully how the model obtained fits the respective observed data. This procedure enables an accurate calculation of the seismicity rates. Cumulative number / year
Predictive Ground Motion Models Due to the absence of sufficient strong motion data, an adequate attenuation model is not available so far for our region. Hence, we have to consider PGM models from regions surrounding the country, or models derived for similar seismotectonic characteristics. New PGME are derived last years for the European-Middle East region (Bindi et al. 2009, Akkar and Bommer 2010), and global models worldwide used (NGA project (USA 2008), Cauzzi & Faccioli 2008), using larger and improved ground motions databases. PGME-s are generally the component with the largest influence on the seismic hazard assessment. Therefore, the seismic hazard outputs obtained using different models are combined into a single map, according to a weighting scheme in the framework of a logic-tree approach.
Results of seismic hazard assessment Evaluation of the seismic hazard is performed using the smoothed- gridded methodology (Frankel 1995, Lapajne et al., 2003), which is based on seismic activity rate inferred from the earthquake catalogue. Hazard was calculated at grid cells (10x10 km) covering the BSHAP region. Hazard calculations are accomplished using the OHAZ 6.0 software, a joint development of Environmental Agency of the Republic of Slovenia and the Institute of Geosciences of Albania, which is greatly improved for the BSHAP project. At first, the seismicity rates are determined at every grid cell within [12.5-24.5 E, 38-47.5 N], by counting the earthquakes with magnitude greater or equal to the minimum magnitude (M W =4.0), and adjusting this value using a maximum likelihood method (Weichert, 1980) that accounts for variable completeness.
Results of seismic hazard assessment Then, the adjusted earthquake rates are spatially smoothed using a two-dimensional Gaussian smoothing operator with correlation distance 20 km, and an elliptical smoothing oriented according to the main tectonic faults within specific seismotectonic zones. Hazard curves that depict the annual frequency of exceedance at given ground-motion levels are calculated at the cells included within a smaller grid ([13.5-23.5 E, 39-47.0 N]) N]). To calculate hazard from a particular source, we apply a doubly-truncated exponential magnitude-frequency distribution, with b-value corresponding to this zone. M min is M W =4.0, while M max varies according to the respective zones from 5.6 up to 7.5.
Results of seismic hazard assessment The hazard is calculated for potential earthquakes at each grid cell. Earthquakes smaller than M6.0 are characterized as point sources at the center of each cell, whereas earthquakes larger than M6 assume hypothetical finite vertical or dipping faults centered on the source grid. Lengths of the finite faults are determined using the Wells and Coppersmith relations, accounting for faulting styles. Calculations are accomplished using four PGMEs: Bindi et al. 2009 (Bi09), Akkar and Bommer 2010 (AB10), Boore and Atkinson 2008 (BA08), and Cauzzi and Faccioli 2008 (CF08). Assessment is applied for rock conditions, with 800 m/sec shear-wave velocity in the upper 30 m of the soil section. The maximum source-sitesite distance and the magnitude range used, were choosen in accordance with their magnitude-distancedistance domain: D max =100 km and 5 M 7.5 for Bi09 and AB10; D max =200 km and 5.0 M S 7.5 for BA08; D max =150 km and 5 M W 7.5 for BA08.
Seismic Hazard Maps The mean hazard map is calculated using the relevant estimates from these models. The respective weights are selected based on a report on the evaluation of PGMEs within the context of SHARE project (M.Segou and S.Akkar, August 2010), as well as on the results of a investigation by Prof. S. Akkar on the validity and ranking of selected PGMEs, using the available strong motion records in the BSHAP region. Following the recomendations of Prof. S. Akkar, we have accepted the weights w=0.3 for AB10 and Bi09, and w=0.2 for BA08 and CF08. The seismic hazard maps we obtained by interpolation of the mean hazard curves at specified annual frequency of exceedance. Seismic hazard maps for PGA corresponding to 10% PE in 10 years (95- year return period), and 10% PE in 50 years (475-years return period) are calculated. Mapping software: GMT (version 4.5.5).
Fig. shows the seismic hazard map for PGA on uniform firm rock site conditions (800m/s shear-wave velocity in the upper 30 m of the crust) at 10-percent probability of exceedance in 50 years, corresponding to the 47-year return period. In compliance with EC8 standards, the map presenting the PGA at 10- percent PE in 10 years, (95- year return period) is also prepared. Seismic Hazard Maps
Conclusions The PSHA for the Western Balkan Countries builds upon extensive research and database compilation carried out over the last three years by the institutions participating in the BSHAP project. Hazard assessment is based on the smoothed-gridded seismicity approach. The seismic hazard maps derived in this project are a good basis to characterize the seismic hazard in our region. They will help the national authorities, public and private institutions, civil emergencies agencies, etc. for urban planning, disaster preparedness, etc. But they should not be considered as national documents for design bulding codes. Every country, based on the seismological database (BSHAP catalogue), and the present seismotectonic zones delineation and characterization, methodology and experience from BSHAP project, as well as the present maps, have to improve the seismic hazard assessment for the relevant territories.
Recommendations Improving of the seismological and seismotectonic databases. Completing the BSHAP catalogue with events M W 3.5 (better for M W 3.0); eleminating of possible inaccuracies; completing of the extended database in format we already have defined and agreed. Improving of the BSHAP seismotectonic databases (some zones are too small and difficult to estimate reliably the seismicity parameters, especially for the low seismicity areas). Identifying and characterization of the large faults in the BSHAP region, which have generated earthquakes with M W 6.5; combining the smoothed gridded seismicity with the fault generated seismic hazard. Creating of the strong motion database for the BSHAP area; deriving a PGME model - more adequate for our region.
Thank You!