Bulletin of the Geological Society of Greece
Vol. 50, 2016
SIGNIFICANT EARTHQUAKES NEAR THE CITY OF THESSALONIKI (NORTHERN GREECE) AND PROBABILITY DISTRIBUTION ON FAULTS
Paradisopoulou P.M. Aristotle University of Thessaloniki, School of Geology, Geophysics Department Papadimitriou E.E. Aristotle University of Thessaloniki, School of Geology, Geophysics Department Mirek J. Institute of Geophysics Polish Academy of Sciences http://dx.doi.org/10.12681/bgsg.11852
Copyright © 2016 P.M. Paradisopoulou, E.E. Papadimitriou, J. Mirek
To cite this article:
Paradisopoulou, Papadimitriou, & Mirek (2016). SIGNIFICANT EARTHQUAKES NEAR THE CITY OF THESSALONIKI (NORTHERN GREECE) AND PROBABILITY DISTRIBUTION ON FAULTS. Bulletin of the Geological Society of Greece, 50, 1389-1398.
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | Δελτίο της Ελληνικής Γεωλογικής Εταιρίας, τόμος L, σελ. 1389-1398 Bulletin of the Geological Society of Greece, vol. L, p. 1389-1398 Πρακτικά 14ου Διεθνούς Συνεδρίου, Θεσσαλονίκη, Μάιος 2016 Proceedings of the 14th International Congress, Thessaloniki, May 2016
SIGNIFICANT EARTHQUAKES NEAR THE CITY OF THESSALONIKI (NORTHERN GREECE) AND PROBABILITY DISTRIBUTION ON FAULTS
Paradisopoulou P.M.1, Papadimitriou E.E.1 and Mirek J.2 1Aristotle University of Thessaloniki, School of Geology, Geophysics Department 54124, Thessaloniki, Greece, [email protected], [email protected] 2Institute of Geophysics Polish Academy of Sciences, Poland, [email protected]
Abstract Stress triggering must be incorporated into quantitative earthquake probability estimate, given that faults are interacted though their stress field. Using time dependent probability estimates this work aims at the evaluation of the occurrence probability of anticipated earthquakes near the city of Thessaloniki, an urban center of 1 million people located in northern Greece, conditional to the time elapsed since the last stronger event on each fault segment of the study area. A method that calculates the macroseismic epicenter and magnitude according to macroseismic intensities is used to improve the existing earthquake catalog (from AD 1600 - 2013 with M≥6.0) in order to compute new interevent and elapsed time values which form the basis for time-dependent probability estimates. To investigate the effects of stress transfer to seismic hazard, the probabilistic calculations presented here employ detailed models of coseismic stress associated with the 20 June 1978 M=6.5 Thessaloniki which is the latest destructive earthquake in the area in the instrumental era. The combined 2015-2045 regional Poisson probability of M≥6.0 earthquakes is ~35% the regional time-dependent probability varies from 0% to 15% and incorporation of stress transfer from 0% to 20% for each fault segment. Keywords: Intensity, interevent time, coseismic stress change.
Περίληψη Στόχος της παρούσας εργασίας είναι η εκτίμηση της χρονικά εξαρτώμενης πιθανότητας γένεσης ισχυρών σεισμών (M≥6.0) στα ενεργά ρήγματα γύρω από την πόλη της Θεσσαλονίκης, περιοχή που στο παρελθόν έχει πληγεί από αρκετούς κατατροφικούς σεισμούς με τελευταίο αυτόν του 1978. Με τη χρήση της μεθόδου υπολογισμού μακροσεισμικού επικέντρου και μεγέθους χρησιμοποιώντας μακροσεισμικές εντάσεις και από τις μακροσεισμικές περιγραφές παλαιότερων σεισμών με M≥6.0 που έγιναν στην περιοχή τα τελευταία 500 χρόνια καταβλήθηκε προσπάθεια επαναπροσδιορισμού των εστιακών παραμέτρων των ιστορικών σεισμών με σκοπό τον υπολογισμό μέσης περιόδου επανάληψης και του χρόνου από τον τελευταίο σεισμό σ’ ένα ρήγμα. Για να γίνει η εκτίμηση της πιθανότητας λήφθηκε υπόψη η μεταβολή της τάσης που προκύπτει μετά από κάθε ισχυρό σεισμό και η οποία έχει ως αποτέλεσμα να επιταχύνει ή να επιβραδύνει τη γένεση ενός επόμενου σεισμού. Ενσωματώθηκε έτσι στους υπολογισμούς της χρονικά εξαρτώμενης πιθανότητας, η μεταβολή της τάσης που προκλήθηκε από το σεισμό της 20 Ιουνίου 1978 με Μ=6.5. Η συνδυαστική πιθανότητα Poisson για τα επόμενα 30 χρόνια
1389
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | (2015-2045) βρέθηκε ίση με ~35%, η δεσμευμένη πιθανότητα εκτιμήθηκε από 0% έως 15% και από0% έως 20% με το συνδυασμό των τάσεων. Λέξεις κλειδιά: Μακροσεισμική ένταση, περίοδος επανάληψης, μεταβολές τάσης.
1. Introduction The city of Thessaloniki is the second largest city in the territory of Greece surrounded by several small towns and villages. In the 20th century five M≥6.0 earthquakes occurred and since 600 A.D. nine M≥6.0 earthquakes have shaken city and its satellite settlements, causing severe damage and casualties. The most severe episode took place in 1759 (M=6.5) when the majority of the inhabitants abandoned the city for about two years (Papazachos and Papazachou, 2003).The 1978 earthquake (M=6.5) was the latest destructive one, causing the collapse of buildings and loss of life in the city and nearby villages. Using a time dependent probability method, this work aims at the evaluation of the occurrence probability of anticipated earthquakes in the city of Thessaloniki, associated with known fault segments conditional to the time elapsed since the last strong event onto each segment. For the probability estimates an earthquake catalog as long in time is possible and the calculation of Coulomb stress changes were necessary. The historical earthquake catalog for the period between A.D. 1600 and 2014 was improved by reevaluation of the earthquake source parameters using the technique of Bakun and Wentworth (1997), for earthquakes with M≥6.0, and the regional attenuation relation of Papazachos and Papaioannou (1997) which enables better calculation of historic earthquake magnitudes from intensity values. Interevent and elapsed times where then calculated along with Coulomb stress changes due to the coseismic slip of the 1978 Stivos earthquake resulting to earthquake advances and delays onto each fault segment at a depth of 10km. The probability calculations for the next 30 and 50 years were performed and given in the form of tables and maps.
2. Earthquake catalogue In the period between A.D. 1600 and 2014 seven M≥ 6.0 earthquakes occurred near Thessaloniki metropolitan area (Figure 1, table 1). Strong earthquakes are documented before 1600, back to 1 A.D. (Papazachos and Papazachou, 2003; Ambraseys, 2009), but more damage descriptions begin with the 1759 Thessaloniki earthquake. Nevertheless set the 1677 Vasilika earthquake is also included in the data set due to its proximal location to Thessaloniki, although the damage reports were limited. These observations are required for the quantitative approach to historic earthquake locations and magnitudes used here. 2.1. Methodology - Data processing Damage descriptions for the study area were published in Papazachos and Papazachou (2003) and Ambraseys (2009) and interpreted on the Modified Mercalli intensity (MMI) scale for seven earthquakes from A.D. 1600 up to now. MMI values were assigned to about 1500 damage descriptions (608 were for 1978 earthquake). These values were used to infer M and epicentre location from MMI according to the methodology of Bakun and Wentworth (1997). The intensity attenuation relation used in this study is given by equation (1):
1390
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | Table 1 – Information on source parameters of earthquakes with M≥6.0, assigned to specific fault segments, the names of which are given in the first column. Bakun and Papazachos and Fault Date Ambraseys (2009) name Wentworth (1997) Papazachou (2003) φ(° Mi λ(°) M φ(°) λ(°) M φ(°) λ(°) ) Anth1 1677 6.0±0.2 40.48 23.24 6.2 40.50 23.00 Anth2 22/06/1759 6.5±0.3 40.55 23.10 6.5 40.60 22.8 Assiros 05/07/1902 6.6±0.3 40.89 22.97 6.5 40.82 23.04 6.3 40.79 23.05 Ierissos 26/09/1932 6.8±0.33 40.44 23.83 7 40.45 23.86 6.8 40.50 23.80 Sohos 29/09/1932 6.4±0.26 40.79 23.44 6.4 40.97 23.23 6.3 40.77 23.48 Volvi 11/05/1933 6.3±0.26 40.66 23.55 6.3 40.62 23.53 6.3 40.70 23.67 Stivos 20/06/1978 6.3±0.26 40.74 23.24 6.5 40.61 23.27 6.4 40.73 23.25 Equation 1- Intensity attenuation relation (Papazachos and Papaioannou, 1997)
M 1.260.62 ()0.003283.28log()/1.61MMIdd i i i i where di is distance in kilometres between MMI observation and epicentre. Felt reports (MMI< III) were excluded and MMI≥ III observations were used in the calculations. The method uses a grid search for an intensity center, thus distance in equation (1) is to a point source. The input file includes the point source (which is the center of grid search), the radius of search and the grid search spacing. The input MMI values yield an output grid of moment magnitudes and confidence intervals (RMS values for each magnitude). Outputs are highly dependent to the input parameters mainly in cases of limited MMI values (such as 1677, 1759 and 1902 earthquakes). Hence a multivariable method was developed to the existing code from Bakun and Wentworth (1997). A geographical area which is defined by the user and includes the epicenter of every earthquake in the dataset is used. Within this limited area we calculate the input parameters as follows: a) the grid center point per 0.15°, b) the radius of search for each 5km and c) the grid search spacing for every 2km. The average of the moment magnitudes yields from the inputs was chosen as the best solution.
Figure 1 - Main faults of the study area (after Tranos et al., 2003, So. F: Sohos Fault, As-An. F: Assiros-Analipsi Fault, L-AV. F: Lagina-Ag. Vasileio Fault, P-A. F: Pefka- svestochori Fault, A-Ch. F: Asvestochori-Chortiatis Fault, P-P. F: Pilaia-Panorama Fault, S-G. F: Stivos-Gerakarou Fault, TGFZ: Thessaloniki-Gerakarou Fault Zone, AEDF: Anthemoudas Fault). The intensity centers for each earthquake are depicted by the black triangles (Bakun and Wentworth, 1997). Red and blues circles denote the epicenters form Papazachos and Papazachou (2003) and from Ambraseys (2009) respectively.
1391
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | The historical earthquake rupture most consistent with the MMI data would be the one that minimizes M and falls within 95% confidence limits on minimized misfit to input MMI values. The segment nearest to the zone of minimum magnitude of sufficient length to accommodate the historical rupture is identified as the earthquake source, with the rupture being centered in the 95% confidence interval (Figure 2). 2.2. Interevent and elapsed times The A.D. 1600-2015 relocated earthquake catalogue comprises seven strong (M≥6.0) earthquakes (table 1, Figures 1, 2) that are used to build interevent and elapsed times for use in probability calculations. Assignment of historical earthquakes to faults indicates possible repeated rupture of some segments. The 1677, M ~6.0 and the 22 June 1759, M~6.5 earthquakes appear to have broken two parts called Anth1, Anth2 respectively, table 1, figure 2. The 5 July 1902, M~6.6 earthquake appears to have ruptured Assiros segment (table 1, figure 2), the 26 September 1932, M~6.8 and the 29 September 1932, M~6.4 earthquakes broke the Ierissos and part of Sohos faults, respectively, whereas the 11 May 1933, M~6.3 and 20 June 1978, M~6.3 shocks ruptured the Volvi and Stivos faults segments respectively (table 1, Figure 2). Based on the above assignment the mean earthquake interevent times (Tr) and elapsed times were found and are given in table 2. Specifically the mean interevent time was taken equal to 500 years due to limited event numbers reported for a particular fault segment, and adopting a mean return period suitable for continental faults. Table 2 - Estimated parameters that used for probability calculations on the 7 fault segments near the city of Thessaloniki.
Fault Tr Elapsed Time tα name (years) (years) (bars/year) Anth1 500 338 0.0076 50 Anth2 500 256 0.0073 50 Assiros 500 113 0.006 50 Ierissos 500 83 0.0076 50 Sohos 500 83 0.006 50 Volvi 500 82 0.0067 50 Stivos 500 37 0.0065 50
1392
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 |
Figure 2 – Relocated earthquakes with M≥6.0 during 1600 – 2015 A.D., using MMI values estimated from damage descriptions compiled by Papazachos and Papazachou (2003).Brown circles are sites with MMI assignments with symbol size increasing with the intensity. The intensity centre shown by a black solid triangle and contours of Mi by green lines. The rms(Mi) contours corresponding to the 95% confidence for location are depicted with thick blue lines.
1393
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | 3. Probability calculations Occurrence probabilities for an earthquake with M≥6.0 and in the next 30 years close to the city of Thessaloniki, are calculated, onto the seven fault segments mentioned above. 3.1. Regional Poisson probability We started with the stationary Poisson probability model which is one that treats earthquake as occurring at random in time (t) about an average interevent time Tr. The probability of at least one event in the time interval (t, t+Δt) is given by: Equation 2- Poisson probability model
( 1)ettTtP/ Ttr The estimated Poisson probabilities are given in Table 3. This model can be applied to the A.D. 16 00-2015 interevent catalog such that compined probabilities are calculated as: Equation 3 – Combined Probability
PPPPPPPP1(1)(1)(1)(1)(1)(1)(1)VasAnAsISVSt...... where: PVas. , PAn. , PAs. , PI. , PS., PV. , PSt. are probabilities for the Vassilika, Anthemounta, Assiros, Ierissos, Sohos, Volvi and Stivos faults, respectively. The combined 30 year probability of an eart hquake to occur near the city of Thessaloniki is 34.4%. 3.2. Time dependent Probability Earthquakes are likely not completely random in time and space as Poisson model assumes, thus time-dependent probabilities are also calculated and given in Table 3. A time-dependent probability (in any time interval (t, t+Δt)) is calculated by a probability density function f(t) as (Working Group of California Earthquake Probabilities, 1990): Equation 4 – Time dependent probability formulation
tt P( tTttf t dt )( ) t where f(t) is the lognormal probability distribution (e.g. Nishenko and Buland, 1987). 3.2.1. Coulomb stress change calculations An earthquake can be modeled as a slipping dislocation in an elastic half space (Okada, 1992) enabling estimation of stress transfer to other faults. Earthquakes occur when stress exceeds the strength of the fault. The closeness to the failure was quantified by using the change in Coulomb failure function (ΔCFF). In its simplest form, the Coulomb failure stress change is (modified from Scholz, 1990): Equation 5 - Coulomb failure stress change formulation
CFF (1) B where, Δτ is the change in shear stress on a fault, Δσ is the change in normal stress, μ is the friction coefficient and B is the Skempton’s coefficient (in this study we assume μ =0.75 and B=0.5 as in Robinson and McGinty, 2000 among others). The Coulomb stress changes are calculated for 1978 earthquake, the last strong earthquake in the study area and used for earthquake probabilities estimates.
1394
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | 3.2.2. Incorporating stress changes from the 1978 earthquake into time dependent probability calculations For time dependent probabilities calculations the methodology followed is that by Toda et al. (1998) and Parsons (2004) who support an earthquake renewal process. The probability of a future event grows as the time elapsed from the previous event increases and considering both permanent and transient effect of the stress changes. All parameters (interevent and elapsed times, tectonic stressing rate (τ), duration of transient effect (tα) that used for calculations are taken from previous investigations (Paradisopoulou et al., 2010, 2013) and given in Table 2. 3.2.3. Assumptions of parameters For Coulomb stress changes calculations it is assumed that are performed in a homogeneous elastic half space and require a coefficient of friction (μ) and the Skempton’s coefficient (Β). One more uncertainty concerns the dip and rake angles of the target fault which are known approximately, e.g. from surface projection, or they are defined from structural information or by moment tensors or focal mechanisms. In all cases there are uncertainties that lead to variation in stress change calculations.
For probability estimations the uncertainties concern: a) the calculation of Tr and elapsed time where historical and paleoseismic data used. We try to reduce these assumptions using the Bakun and Wentworth (1997) method and recalculate the location and magnitude of historical earthquakes based on intensity assigns. b) The estimation of stressing rate from geodetic data. The calculation of this parameter was improved using the distribution of across and onto every fault and taking the average value for probability estimations (Karakostas et al., 2013). For the dominant transient effect of the stress changes, rate-state constitutive relations were applied, which require parameters such as tα (aftershock duration) and Ασ (a state parameter). We assume, according to seismicity of each subarea of the study area and the mean recurrence time (Dieterich, 1994) that tα is equal to 10% of Tr. With given values of the parameters tα and , the Ασ was calculated using the equation: Aσ=tα∙ . 3.2.4. Results Detailed earthquake occurrence probabilities are provided by gridding the target fault areas and performing calculations on the nodes of the grid spacing by 1km (Figures 3). The probability calculations were carried out during the next 30 (Figure 3a) and 50 years (Figure 3b) and are given in Table 3. In these figures shades of blue denote faults where the probability values are low (0.00≤P≤0.03) due mostly to the effect of negative changes in Coulomb stress. Shades of green and yellow represent higher probability values (0.04≤P< 0.08 and 0.09≤P< 0.20, respectively) due mainly to positive ΔCFF values on these faults. The entirely blue lines represent faults that have already failed whereas orange and red lines correspond to faults that are candidate to host an incoming earthquake (higher probability values 0.20≤P< 0.35).
1395
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | a)
b)
Figure 3 - Probability distribution along each fault of the study area. Colors between blue and green correspond to low probability values (0.00≤Ρ<0.08) whereas between yellow and red indicate higher probability values (0.09 ≤ Ρ < 0.20 and P≥0.30). Table 2 - 30 and 50 years probability values (Poison and time dependent) for the occurrence of an earthquake with M≥6.0 onto each fault segment. Probability after Probability before stress step stress step Poisso Time Poiso Time Avg Average Average Fault n (30 Dependen n (50 Dependen ΔCFF value value name yrs) t (30 yrs) yrs) t (50 yrs) (bars) [P(30)] [P(50)] Anth1 0.06 0.09 0.095 0.15 -0.21 0.09 0.14 Anth2 0.06 0.06 0.095 0.11 -1.65 0.005 0.01 Assiros 0.06 0.01 0.095 0.02 -4.66 0.05 0.06 Ierissos 0.06 0.002 0.095 0.005 0.018 0.13 0.20 Sohos 0.06 0.002 0.095 0.005 0.15 0.02 0.03 Volvi 0.06 0.002 0.095 0.005 0.05 0.05 0.07 Stivos 0.06 0 0.095 0 -1.60 0 0
1396
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | The probability of an M≥6.0 earthquake rupturing near Thessaloniki city (if a time-dependent model that includes coseismic stress changes is applied) is estimating ~10% and 13% for Anthemoundas1 and Sohos faults, respectively, in the next 30 years although the estimated probability on the same faults for 50 years is ~14% and 20% respectively. Lower probability values are found for the other five faults because of the relatively long interevent time (500 years) and because they are located inside low (negative) ΔCFF values.
4. Conclusions Historical and instrumental records reveal that the city of Thessaloniki is frequently affected by disastrous earthquakes associated with proximal fault segments, which exhibit recurrence intervals of the order of 200 to 500 years. Intensity descriptions and attenuation relations were used in a grid search to find source locations and magnitude that best fit the intensity assignments (following Bakun and Wentworth, 1997). The same approach was applied for the recent 1978 main shock, which served as a calibration event. The historical earthquake rupture most consistent with the MMI data would be the one that minimizes M and falls within 95% confidence bounds on minimized misfit to input MMI values. Additionally, the 1677 earthquake according to 95% confidence contour is associated with a fault segment of ~17 km in length and the 1759 earthquake with a ~20km length fault segment. The results are in accordance with active mapped faults and with fault lengths calculated from empirical relationships. The applied method (Bakun and Wenworth, 1997) revealed that the intensity centre locations for the verification events are acceptably close to the epicentres from Papazachos and Papazachou (2003) and Ambraseys (2009). More specific for the 1978 earthquake the intensity centre is exactly at the same location with the instrumental epicentre, and for the 1759 and 1677 earthquakes the relocated epicentres are 10 and 15km, respectively, from the historical ones (table 1). The earthquake catalogue that resulted with revised magnitudes, Mi, and intensity centre locations, enables the estimation of interevent and elapsed times necessary for probability calculations that are performed for seven faults being proximal to the city of Thessaloniki. The Coulomb stress changes were calculated for the last strong earthquake that occurred 1978 earthquake in order to combine earthquake renewal and stress transfer into probability calculation. The ΔCFF values affect the estimated probabilities, by increasing or decreasing them when positive or negative ΔCFF values, respectively, were calculated for the particular fault segment. Thus, the probability values for an earthquake to occur for the next 30 years from 2015 on Sohos and Anthemountas1 faults were affected from positive stress changes from the 1978 earthquake and found equal to 10% to 13%, respectively (for 50 years is ~14% and 20%, respectively). The probability estimates derived from the renewal model, as shown in Table 3, reveal that for most of the faults are smaller than those based on the Poisson model (~0.06 and 0.1). Besides, Poissonian probabilities are larger than the conditional ones due to the fact that the time elapsed since the last event of M≥6.0 onto each of the considered faults is shorter than the estimated mean interevent time. The combined 30 year probability for the city of Thessaloniki is estimated to be equal to 34.4%. For certain faults the differences between Poissonian and conditional estimates before the stress step are significantly different than those incorporating the stress step, and worthy to mention for future seismic hazard assessment. For the Anthemountas and Sohos faults the Poissonian probabilities are found equal to 6% for 30 years and ~10% for 50 years, whereas the corresponding ones after the stress step are equal to 10% and 13% (for the next 30 years) and 14% and 20% for the next 50 years, respectively. The opposite consequence of ΔCFF effect is observed to the rest five faults, yielding a 30-year (50-year) Poisson of 6% (10%) and conditional probability ranging in 0.2-1%, respectively, whereas the time-dependent probabilities on these faults are from 0% to 7%.
1397
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 | 5. Acknowledgments All calculations concerning macroseismic data have been achieved using the Fortran code from Prof. William Bakun. The stress tensors were calculated using a program written by J. Deng (Deng and Sykes, 1997) based on the DIS3D code of S. Dunbar, which later improved (Erikson, 1986) and the expressions of G. Converse. Maps are from ArcMap tools. This work was partially supported within statutory activities No 3841/E-41/S/2015 of the Ministry of Science and Higher Education of Poland. Geophysics Department contribution 854.
6. References Ambraseys, N.N., 2009. Earthquakes in the Mediterranean and Middle East: a multidisciplinary study of seismicity up to 1900, ISBN 9780521872928, Cambridge University Press, 947 pp. Bakun, W.H. and Wentworth, C.M., 1997. Estimating earthquake location and magnitude from seismic intensity data, Bull. Seism. Soc. of America, 87, 1502-1521. Deng, J. and Sykes, L.R., 1997. Evolution of the stress field in southern California and triggering of moderate-size earthquakes: A 200-year perspective, Journ. of Geophys. Res., 102, 9859-9886. Dieterich, J.H., 1994. A constitutive law for rate of earthquake production and its application to earthquake clustering, Journal of Geophysical Research, 99, 2601-2618. Erikson, L., 1986. User’s manual for DIS3D: A three-dimensional dislocation program with applications to faulting in the Earth, Master’s Thesis, Stanford Univ., Stanford, Calif., 167 pp. Karakostas, V., Papadimitriou, E., Xueshen, J., Zhihui, L., Paradisopoulou, P. and Zhang, H., 2013. Potential of future seismogenesis in Hebei Province (NE China) due to stress interactions between strong earthquakes, Journal of Asian Earth Sciences, 75(1-12). Nishenko, S.P. and Buland, R., 1987. A generic recurrence interval distribution for earthquake forecasting, Bull. Seism. Soc. of America, 77, 1382-1399. Okada, Y., 1992. Internal deformation due to shear and tensile faults in a half space, Bull. of Seism. Soc. of America, 82, 1018-1040. Papazachos, C.B. and Papaioannou, Ch.A., 1997. Macroseismic field of the Balkan area, Journal of Seismology, 1, 181-201. Papazachos, B.C. and Papazachou, C., 2003. The earthquakes of Greece, Ziti publications, Thessaloniki, 289 pp. Parsons, T., 2004. Recalculated probability of M≥7 earthquakes beneath the Sea of Marmara, Turkey, Journal of Geophysical Research, 109, doi: 10.1029/2003JB002667. Paradisopoulou, P., Papadimitriou, E., Mirek, J. and Karakostas, V., 2013. Coseismic stress distribution along active structures and their influence on time dependent probability values, Bull. of the Geol. Soc. Greece, XLVII. Paradisopoulou, P.M., Papadimitriou, E.E., Karakostas, V.G., Lasocki, S., Mirek, J. and Kilias, A., 2010. Influence of stress transfer in probability estimates of M≥6.5 earthquakes in Greece and surrounding areas, Bull. Geol. Soc. Greece, XLIII, 2114-2124. Robinson, R. and McGinty, P.J., 2000. The enigma of the Arthur’s Pass, New Zealand, earthquake. 2. The aftershock distribution and its relation to regional and induced stress fields, Journal of Geophysical Research, 105, 16139-16150. Scholz, C., 1990. The mechanics of earthquakes and faulting, Cambridge University Press, Cambridge, 439 pp. Toda, S., Stein, R.S., Reasenberg, P.A. and Yoshida, A., 1998. Stress transferred by the 1995 Mw=6.9 Kobe, Japan shock: Effect on aftershocks and future earthquake probabilities, Journal of Geophysical Research, 124, 439-451. Tranos, M.D., Papadimitriou, E.E. and Kilias, A.A., 2003. Thessaloniki-Gerakarou Fault Zone (TGFZ): the western extension of the 1978 Thessloniki earthquake fault (Nothern Greece) and seismic hazard assessment, J. Struct. Geol., 25, 2109-2123. Working Group on California Earthquake Probabilities (WGCEP), 1990. Probabilities of large earthquakes in the San Francisco Bay region, California, U.S. Geol. Surv., Circ. 1053, 51.
1398
http://epublishing.ekt.gr | e-Publisher: EKT | Downloaded at 28/07/2017 10:12:06 |
Powered by TCPDF (www.tcpdf.org) See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/281837227
EARTHQUAKE-HAZARD ASSESSMENT IN THESSALONIKI, GREECE: LEVEL I: PROBABILISTIC & DETERMINISTIC APPROACH
Conference Paper · October 2002
CITATIONS READS 6 257
3 authors:
Maria Alexoudi Theodoros Chatzigogos Aristotle University of Thessaloniki Aristotle University of Thessaloniki
63 PUBLICATIONS 217 CITATIONS 18 PUBLICATIONS 80 CITATIONS
SEE PROFILE SEE PROFILE
Kyriazis Pitilakis Aristotle University of Thessaloniki
347 PUBLICATIONS 5,244 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
GRATIS - Ground Real-time Active Tremor Inhibition System View project
REAKT WP7 - Strategic Applications and Capacity Building View project
All content following this page was uploaded by Maria Alexoudi on 17 September 2015.
The user has requested enhancement of the downloaded file. EARTHQUAKE LOSS ESTIMATION AND RISK REDUCTION (2002_ELE & RR) INTERNATIONAL CONFERENCE, Bucharest, 24-26 October 2002
EARTHQUAKE-HAZARD ASSESSMENT IN THESSALONIKI, GREECE: LEVEL I: PROBABILISTIC & DETERMINISTIC APPROACH
M. Alexoudi1 (*), Th. Hatzigogos2, K. Pitilakis2
SUMMARY
Seismic hazard and risk analysis for urban areas consist an essential element of the strategy of protection against earthquakes. In the present paper we present a deterministic and probabilistic assessment of the seismic hazard for the city of Thessaloniki, in the framework of RISK-UE research project (ENVK4-CT-2000-00014). Four scenarios were developed: a deterministic one, which corresponds to the maximum probable/ historical earthquake and three probabilistic scenarios for a recurrence period of 475years: One for uniform distribution of the seismicity in two seismic source zones (SSZ) and two where the seismicity is concentrated in two seismic fault zones describing the complex seismotectonics of the area. Ground motion intensity is described in terms of spectral ordinates as SA(T) = 0.0, 0.3, 0.6, 1.0, 1.6 sec. All resulting maps are portrayed in Geographical Information System (GIS).
1. SEISMOLOGICAL AND GEOLOGICAL INFORMATION
The city of Thessaloniki is in the Northern Greece. Thessaloniki with its suburbs extends towards the E-NE direction in the Thermaikos gulf coastal zone starting from the sea level and reaching an altitude of 100 to 150m, with smooth ascents. Thessaloniki is located on the Axios-Vardar seismogenic zone, which is adjacent to the Servomacedonian massif, one of the most seismotectonically active regions in Europe (Papazachos et al., 1979). Servomakedonian zone is considered as a Tertiary structure but probably represents a reactivated tectonic line, related to an old (Jurassic) subduction of the Vardar Ocean beneath the Servomacedonian continental margin. There is clear evidence that this tectonic contact was reactivated by extension during Neogene-Quaternary times, with near-vertical normal faults along the boundary. Since the Miocene the area of Central Macedonia, where this neotectonic sheet is located, has
1 Civil Engineer MSc, Aristotle University of Thessaloniki, Department of Civil Engineering, Geotechnical Division, Laboratory of Soil Mechanics, Thessaloniki, Greece, P.O.B. 450, GR-54006, e-mail: [email protected] 2 Professor, Aristotle University of Thessaloniki, Department of Civil Engineering, Geotechnical Division, Laboratory of Soil Mechanics, Thessaloniki, Greece, P.O.B. 450, GR-54006 been intensely faulted, forming many tectonic grabens and depressions such as the Axios basin, the Anthemountas and Mygdonian grabens etc. These depressions are the result of a continuous extensional deformation, which was mostly associated with pure normal to oblique-normal faults trending mainly NW- SE (sinistral component), ENVE- WSW (dextral component) and E- W strikes (Pavlides S, Soulakellis N, 1990, Pavlides et al 1996, Voidomatis et al 1990). In addition, some long N-S trending faults complete the general fracture pattern. Most of the above mentioned faults have been active at least since the Miocene, while some of them (mainly the E-W trending faults) are associated with the present seismic activity or have a verified activity since the Quaternary. The active belt is characterized by an intense shallow seismic activity with earthquakes having magnitudes up to about 7.6. There were three seismic periods with great seismic activity in Servomakedonian zone during the present century. The first one started in the region of Volvi- Langada lakes with a main shock of Mo = 6.6 (Assiros 1902) and continued in Bulgaria (1903-1905) to the north with the main shock of Mo= 7.6 (Kresna 1904) and in the Athos peninsula to the southeast with the main shock of Mo= 7.4 (1905). The second period started in Yugoslavia with a main shock of Mo= 6.6 (Valandovo 1931) and continued southeastward (1932- 1933) with the main shock of Mo= 6.9 (Ierissos 1932) and with other shocks some of which had their epicentres in the region of Volvi- Langada lakes (Comninakis, Papazachos, 1986). In 1978 a strong earthquake (M=6.5) produced heavy damages in the city of Thessaloniki. The epicentre was located between the lakes Langadas and Volvis.
Fig. 1 Epicentres of large earthquakes in the area surrounding Thessaloniki (1901-2000)
From geological- geotechnical point of view, the city presents interesting features, as it comprises regions with various geological formations composed by a large variety of soil materials. Three main large-scale geology structures are identified, oriented in NW- SE direction. The first formation is the metamorphic substratum consisting of gneiss, epigneiss, and green schists, which are surficial near the city at the N-NE border of the urban area. These crystalline rocks constitute the bedrock basement beneath the city reaching a depth of 150-300m at the centre part of the city near the coastline in W-WS direction. The second formation is composed by alluvial deposits mainly of the Neogene period. In this geological structure the red stiff silty clay, series are dominant, covering the bedrock basement beneath the city. Finally, recent deposits of Holocene clays-sands-pebbles compose the third surficial formation.
Fig. 2 Seismotectonic map of Greece with Fig. 3 Geological map of Thessaloniki Seismologeological data Scale 1: 500.000 Ι.G.Μ.Ε 1989
2. GEOTECHNICAL ZONATION OF THESSALONIKI MUNICIPAL AREA
Anastasiadis et al (2001) give a detailed description of the geotechnical zones in the municipality of Thessaloniki. Field and laboratory determination of dynamic properties of natural soil deposits in Thessaloniki is given by Pitilakis et al, 1992; Pitilakis K., Anastasiadis A, 1996. In Thessaloniki we have analysed 450 geotechnical borings and many CH/DH/SWI data. Using a simplified classification of soil categories according to the Vs,30 (Table 1), Fig. 4 give a simplified geotechnical map of the city of Thessaloniki.
CITY HOTEL IPPO Table 1 Classification of soil conditions according to Ambraseys (1996) classification
Classification Vs,30 (m/ sec) Rock > 750 Stiff Soil 360- 750 Soft Soil < 360
Fig. 4 Simplified geotechnical map of the city of Thessaloniki
3. DETERMINISTIC HAZARD ASSESSMENT FOR THESSALONIKI URBAN AREA
The deterministic scenario for seismic hazard analysis corresponds to the maximum historical earthquake in the last century (Assiros, 1902, M=6.6, Intensity (IX.)), which (Fig.5) caused serious damages in Thessaloniki’s buildings and it’s suburbs (Papazachos B. Papazachou C; 1997, Papazachos B. Papazachou C; 1989). We selected Ambraseys et al (1996) attenuation relationship in both deterministic and probabilistic analysis.
Fig. 5 Assiros fault location in respect to the selected grid of the hazard analysis
4. PROBABILISTIC HAZARD ASSESSMENT FOR THE CITY OF THESSALONIKI
The Probabilistic Seismic Hazard Analysis (PSHA) for the city of Thessaloniki was conducted using CRISIS 99, a code developed by Professor M. Ordaz and co-workers of the “Institudo de Ingenieria de la U.N.A.M”, Mexico City. In CRISIS 99, the sources are modelled as area sources (Seismic Source Zones or SSZs) and fault sources. The present application is performed based on area geometry modelling for sources 35, 36 and on fault sources (fault geometry A, fault geometry B) that describe the complex geotectonic structure of the area surrounding Thessaloniki. Fault geometries A and B are representing parts of the faulting system were Thessaloniki’s 78 earthquake was produced. The seismotectonic zonation of Greece and consequently of the region of Thessaloniki is developed by Papazachos & Papazachou, 1989; Papazachos & Papazachou, 1997; Papaioannou & Papazachos, 2000. Six seismic zones affect Thessaloniki (zone 33, 34, 35, 36, 59, 64) - Fig. 5. SSZ 35& 36 give the 97% of the whole seismicity (Alexoudi M et al, 2001). The parameters of each zone were derived from data of previous earthquakes after completeness analysis (Papazachos, 1999b; Papaioannou & Papazachos, 2000) for all seismic zones of Greece.
Table 2 Seismicity parameters for seismogenic sources of the shallow earthquakes in Greece (Papazachos & Papaioannou, 1997)
No Source Name b a Area, A, km2 35 Volvi 0.84 4.04 5903 36 Kozani 0.87 3.84 19887
No Mmax Rate, r, (M>=5.0) ζ ε 35 7.1 0.723 89. 0.73 36 6.6 0.306 28. 0.79
The earthquakes in zones 35, 36 are characterized as shallow earthquakes having focal depth less than 20km.
36
35
Fig.6 Seismogenic sources of shallow earthquakes in Greece and surrounding area Fig. 7 Main seismic zones affecting Thessaloniki (Papazachos & Papaioannou, 1997)
Fig. 8 Position of Fault Geometry A Fig.9 Position of Fault Geometry B
5. RESULTS FROM SEISMIC HAZARD ANALYSIS BY APPLICATION OF G.I.S
The deterministic and the three probabilistic scenarios (T=475years recurrence period) are portrayed in terms of spectral ordinates (SA(T) = 0.0, 0.3, 0.6, 1.0, 1.6sec). In Fig.10-11 are presented the results for SSZ case, in GIS format. Also, for comparison reasons four maps are presented in Fig.12-15 (T=0sec) for the four developed scenarios.
0 1 2 K Fig.10 Probabilistic Acceleration Response spectral values for T=0sec, T=0.3sec, T=0.6sec (SSZ case)
Legend
Fig.11 Probabilistic Acceleration Response spectral values for T=1.0sec, T=1.6sec (SSZ case)
Fig. 12 Probabilistic approach Fig. 13 Probabilistic approach (SSZ case) –T=0sec (Fault geometry A)-T=0sec
Fig. 14 Probabilistic approach Fig. 15 Deterministic approach T=0sec (Fault geometry B)- T=0sec
6. COMPARISON BETWEEN THE DETERMINISTIC & PROBABILISTIC APPROACH
The deterministic scenario seems to be more severe but at the same time less reliable. Among the probabilistic scenarios the most reliable, according to our judgement and interpretation of the results together with the seismic history of the city, is probably the scenario using the seismic zones (SSZ). Besides the comparisons of PGA (T=0.0sec) values in Fig.12-15, we compared directly the computed PGA and spectral values in two specific locations in the city of Thessaloniki (See figure 4). One characterised as “soft soil” site, is located along the shoreline (CITY HOTEL) where we have a record of the 1978 earthquake, recorded at the basement of an 8 stories building. The second (IPPO) is a “stiff soil” site in the historical centre on the stiff red clay formation. In table 3, we gave the computed PGA (T=0.0sec) values and in Fig.16 -17 the acceleration response spectra.
Table 3 Comparison between Probabilistic & Deterministic scenario
Probabilistic Deterministic T=0sec SSZ Case Fault A Fault B ASSIROS Ippodromio 352.74gal 354.48 gal 264.45 gal 153.95 gal (Stiff Soil) City Hotel 358.20 gal 348.06 gal 262.34 gal 155.95 gal (Soft Soil)
Comparison between SSZcase, Fault geometry A & B Comparison between SSZ case, & Determninistic approach -STIFF SOIL (Ippodromio) Fault geometry A & B and Thessaloniki's 78 accelerogram SOFT SOIL (CITY HOTEL)
1200,00 1200,00 SSZ case SSZ case 1000,00 Fault A 1000,00 Fault A 800,00 Fault B 800,00 Fault B Deterministic scenario- Assiros 600,00 600,00 Deterministic scenario- Assiros
400,00 400,00 Thess'78_long Intensity (Amax inIntensity (Amax gal)
Intensity (Amax in gal) (Amax in Intensity Thess'78_trans 200,00 200,00 Deterministic scenario-Stivos 0,00 0,00 0,00 0,20 0,40 0,60 0,80 1,00 1,20 1,40 1,60 1,80 0,00 0,50 1,00 1,50 2,00 2,50 Structural period (T in sec) Structural period (T in sec)
Fig. 16 Comparison between probabilistic and Fig. 17 Comparison between probabilistic and deterministic (Ippodromio-Stiff Soil). deterministic and real record of Thessaloniki earthquake (City Hotel-Soft Soil).
SSZ and Fault A scenarios give in both site identical results while Fault B scenario base on a fault geometry somehow further from Thessaloniki gives in general lower spectral values. Comparing the deterministic scenario with the probabilistic ones we can notice that deterministic scenario gives lower PGA & PSA values in respect to probabilistic ones. 7. CONCLUSIONS
As a general conclusion the computed PGA values at T=0.0sec and for 475years recurrence period for the study area of Thessaloniki municipality are ranging from 0.24g- 0.40 although deterministic scenario gives even lower values ranging from 0.10g- 0.21g. The complex structure of the area doesn’t allow the precise estimation of the length of the faults, or the exact locations that are essential parameters for the seismic hazard. Deterministic approach based on an historical event with no reliable information of the epicentral distance or the magnitude is considered less reliable than the probabilistic approaches and especially this one using seismic zones. So it is believed that the probabilistic scenario SSZ is the most accurate approach in predicting the seismic hazard in Thessaloniki.
8. ACKNOWLEDGEMENTS
The warmest acknowledgements are due to Prof. M Ordaz for his permanent assistance and of course for providing the code (CRISIS 99). Many thanks are also addressed to Prof. Faccioli and to V. Pessina and to the team of Politechico di Milano for their help especially providing the attenuation relationships. European Committee has funded this research in the framework of RISK-UE project (Contract: EVK4-CT-2000-00014).
REFERENCES
ALEXOUDI M, PITILAKIS K, XATZIGOGOS (2001) “Preliminary study of the seismic hazard of the city of Thessaloniki using CRISIS 99” Internal report in RISK-UE project.
AMBRASEYS N.N. (1995) ”The prediction of earthquake peak ground acceleration in Europe”. Publ. Earthquake Engineering and Structural Dynamics, 24, pp.467- 490.
AMBRASEYS N.N, SIMPSON K.A, BOMMER J.J. (1996) “Prediction of horizontal response spectra in Europe”. Publ. Earthquake Engineering and Structural Dynamics, 25, pp. 371- 400.
ANASTASIADIS A.J., RAPTAKIS D.G., PITILAKIS K.D (2001) “Thessaloniki’s Detailed Microzoning: Subsurface Structure as basis for Site Response Analysis”, Pure and Applied Geophysics – PAGEOPH, Vol. 158, no. 12, pp. 2597-2633.
COMNINAKIS P.E & PAPAZACHOS B.C (1986) “A catalogue of earthquakes in Greece and the surrounding area for the period 1901- 1985” Univ. Thessaloniki, Geophys. Lab, 1, 167 pp.
PAVLIDES S, SOULAKELLIS N. (1990) “Multifractured Seismogenic Area of Thessaloniki 1978 Earthquake (Northern Greece)”. Proceedings of the International Earthquake Sciences Congress on Aegean regions, IZMIR –TURKEY, II, pp.64-74.
PAVLIDES S, MOUTRAKIS D, ZOUROS N, CHATZIPETROS A (1996) “Active Fault Geometry and Kinematics in Greece: The Thessaloniki (Ms= 6.5, 1978) and Kozani- Grevena (Ms= 6.6, 1995) Earthquake – Two Case Studies” Proceeding of the 30th International Geological Congress, 5, pp. 73- 86.
PAPAZACHOS B.C, MOUTRAKIS D, PSILOVIKIS A AND LEVENTAKIS G (1979) “Focal Properties of the 1978Earthquake in Thessaloniki area” Bulgarian Geophysical Journal, 6, pp. 72- 80.
PAPAZACHOS B.C, MOUTRAKIS D, PSILOVIKIS A AND LEVENTAKIS G (1979) “Surface fault traces and fault plane solutions of the May- June 1978. Major shocks in the Thessaloniki area, Greece” Technophysics, 53, pp. 171-183.
PAPAZACHOS B.C & PAPAZACHOU C (1997)“The earthquakes of Greece” Ziti Editions, Thessaloniki, Greece, 304. pp
PAPAZACHOS B, PAPAZACHOU C (1989) "Greek Earthquakes”. Ziti Publ, Thessaloniki, (in Greek)
PAPAZACHOS B. (1987) “Introduction to Seismology”. Ziti Publ, Thessaloniki, (in Greek)
PAPAZACHOS C.B & PAPAIOANNOU CH.A (1997) “The macroseismic field of the Balkan area”. Journal. of Seismology, 1, pp.181- 201.
PAPAZACHOS, B.C, PAPAIOANNOU, CH.A., PAPAZACHOS, C.B., SAVVAIDIS, A.S. (1997), “Atlas of Isoseismal Maps for Strong (M>5.5) shallow (h<60km) Earthquakes in Greece and Surrounding Area 426BC-1995”, University of Thessaloniki, Geophysical Laboratory, Publication No 4.
PAPAIOANNOU CH.A & PAPAZACHOS B.C (2000) “Time- Independnt and Time- Dependent Seismic Hazard in Greece based on Seismogenic Sources” Bulletin of the Seismological Society of America, 90, 1, pp. 22- 33.
PITILAKIS, K., ANASTASIADIS, A., RAPTAKIS D. (1992) “Field and Laboratory Determination of Dynamic Properties of Natural Soil Deposits” Proceeding of the 10th World Conference on Earthquake Engineering, Madrid, Vol.5, pp.1275-1280.
PITILAKIS, K., AND ANASTASIADIS, A. (1998) “Soil and site characterization for seismic response analysis” Proceeding of the XI ECEE, Paris 6-11 Sept. 1998, Inv.Lectures, pp.65-90
VOIDOMATIS PH.S, PAVLIDES S.B, PAPADOPOULOS G.A (1990) “Active deformation and seismic potential in the Servomacedonian Zone, northern Greece” Technophysics, 179, pp. 1- 9.
View publication stats ANNALS OF GEOPHYSICS, VOL. 49, N. 4/5, August/October 2006
Microearthquake study of the broader Thessaloniki area (Northern Greece)
Parthena M. Paradisopoulou (1), Vassilis G. Karakostas (1), Eleftheria E. Papadimitriou (1), Markos D. Tranos (2), Costantinos B. Papazachos (1) and Georgios F. Karakaisis (1) (1) Department of Geophysics, School of Geology, Aristotle University of Thessaloniki, Thessaloniki, Greece (2) Department of Geology, School of Geology, Aristotle University of Thessaloniki, Thessaloniki, Greece
Abstract A temporary network of twelve portable digital seismological stations was operated around the city of Thessa- loniki (Northern Greece) for a period of 19 months (from July 2001 to April 2002 and from October 2002 to Au- gust 2003), providing data that enabled the study of the interconnection between microseismicity and active tec- tonics in the area. During the operation period 277 microearthquakes that were recorded in more than four sta- tions were accurately located and 64 fault plane solutions were determined. Seismic activity is associated with ENE-WSW, E-W and ESE-WNW striking normal faults and is nearly confined to the first 15 km, thus defining the seismogenic layer in the study area. The mean orientation of the axis of maximum extension (T-axis) is N- S to NNE-SSW, determined from fault plane solutions, in agreement with the regional extensional stress pattern, which strikes perpendicular to the orientation of the main WNW-ESE active faults of the area.
Key words microseismicity – fault plane solutions – was the latest destructive one, causing the col- active tectonics – Northern Greece lapse of buildings and loss of life in the city and nearby villages. After this earthquake, seismo- logical (Papazachos et al., 1979 a,b; Carver and 1. Introduction Bollinger, 1981; Soufleris and Steward, 1981; Soufleris et al., 1982) and neotectonic (Moun- The city of Thessaloniki is the second trakis et al., 1983, 1996a,b; Mercier et al., 1983; largest city in the territory of Greece surround- Pavlides et al., 1990; Tranos, 1998; Tranos et al., ed by several small towns and villages. It has 2003) research was carried out in the broader suffered significant damage due to strong earth- Thessaloniki area, trying to shed more light on quakes several times in the past, the most severe the faulting pattern and the associated seismic ac- being in 1759 (M=6.5) when the majority of tivity of the area which belongs to the extension- the inhabitants abandoned the city for about al Northern Aegean region (fig. 1). two years (Papazachos and Papazachou, 2003). The continuous recordings from the perma- The 1978 Thessaloniki earthquake (M=6.5) nent seismological network installed in the area after the 1978 mainshock and fully operating since 1981 by the Geophysics Department of the Aristotle University of Thessaloniki has signifi- cantly enriched the seismological data base for Mailing address: Dr. Parthena M. Paradisopoulou, De- that area. In addition, temporarily installed dense partment of Geophysics, School of Geology, Aristotle Uni- versity of Thessaloniki, GRÐ54124, Thessaloniki, Greece; local networks (Hatzfeld et al., 1986/1987; Scor- e-mail: [email protected] dilis et al., 1989; Hatzidimitriou et al., 1991; Pa-
1081 Parthena M. Paradisopoulou et al.
Fig. 1. Morphology and main regional faults (after Tranos et al., 2003) of the study area (N-X. F: Nikopolis- Xilopolis Fault; As-An F.: Assiros-Analipsi Fault; L-AV F: Lagina-Ag. Vasilios Fault; NA. F: Nea Aghialos Fault; P-A F: Pefka-Asvestochori Fault; PR. F: Peristera Fault; A-Ch. F: Asvestochori-Chortiatis Fault; P-P F: Pilea-Panorama Fault; AEDF: Anthemountas Fault; S-G. F: Stivos-Gerakarou Fault). The grey arrows show the extensional velocities estimated for the Mygdonian Basin from Papazachos et al. (2001). The inset map shows main tectonic features of the Aegean and surrounding regions (CTF: Cephalonia Transform Fault; NAT: North Aegean Trough; NAF: North Anatolian Fault). The arrows show the direction of maximum compression axis from continental collision to the north and oceanic subduction to the south. The arrows at the north-east part show the strike-slip motion along the NAT which is the westward continuation of the NAF zone. The rectangle depicts the study area. pazachos et al., 2000; Paradisopoulou et al., plane solutions could be used for a more detailed 2004) also provided more accurate and detailed investigation aiming to identify the active struc- information on the seismotectonic properties of tures of the study area. the area. The rather weak seismic activity record- ed by the permanent network (with earthquake magnitudes up to 5.1) in the vicinity of the city of 2. Tectonic regime Thessaloniki, do not identify active structures and their geometrical and kinematic properties or The study region consists of elongated hilly define the association of seismicity with already to mountainous chains striking NW-SE and large known active faults. Although microseismicity basins and grabens that interrupt these chains. recorded by local networks is not as representa- The exposed rocks are pre-Alpine and Alpine tive as the strong earthquake occurrence for the that resemble formations of Axios, Circum study of major tectonic features, it is a powerful Rhodope belt and Serbomacedonian massif and tool for the identification of active structures and made up the Circum Rhodope Belt Thrust Zone. their properties. Its recordings permit more accu- This thrust zone is formed by successive anticli- rate earthquake location and the derived fault noria and synclinoria that have been interrupted
1082 Microearthquake study of the broader Thessaloniki area (Northern Greece) by repeated SW directed thrusts. Above this Historical strong earthquakes mainly oc- basement, extensive NW-SE and E-W striking curred along the Mygdonia Basin and Anthe- continental-type basins and grabens have been mountas fault zone (AEDF) and were ascribed filled with Neogene and Quaternary sediments to the reactivation of the NW-SE and E-W strik- i.e. the NW-SE Pre-Mygdonian Basin, and the ing boundary faults of the basins. Similarly, the E-W Anthemountas and Mygdonia basins (Tra- latest 1978 Thessaloniki earthquake was attrib- nos et al., 1999) (fig. 1). The formation of these uted to the Stivos-Gerakarou Fault bounding the basins is attributed to the well established brittle southern part of the Mygdonia Basin. The pres- extensional deformation, since the Late Mio- ent study correlated the possibly active faults cene, mainly related to high-angle normal faults being well defined from the neotectonic studies (Pavlides and Kilias, 1987; Tranos, 1998; Tranos with the seismological data in an attempt to et al., 2003). This extensional deformation has identify the active rupture zones of the area. the least principal stress axis (σ3) oriented NE- SW in Late Miocene-Pliocene and N-S in the Quaternary (Pavlides and Kilias, 1987; Mercier 3. Data processing and Carey-Gaihardis, 1989; Pavlides et al., 1990; Tranos, 1998; Tranos and Mountrakis, 1998). A dense temporary network of twelve However, the fault pattern of the broader area is portable digital seismological stations was in- more complicated than initially thought and in- stalled and operated during July 2001-April cludes NW-SE, NE-SW, E-W and NNE-SSW 2002 and October 2002-August 2003 in the area striking faults. Many of them are inherited struc- around the city of Thessaloniki. The seismic net- tures, active at least since the Miocene (Pavlides work consisted of Reftek 72A-07 seismographs, and Kilias, 1987; Tranos et al., 1999, 2003). The equipped with 3-component broadband seis- recent neotectonic mapping carried out in the mometers (Guralp CMG40T), all connected to broader Thessaloniki city (Mountrakis et al., GPS for time correction. The average spacing 1996a,b; Tranos et al., 2003, 2004) recognize the between the stations was ∼10 km in order to en- following (fig. 1): sure the accurate estimations of the focal depths a) E-W and ENE-WSW striking Pilea- and to allow a good coverage of first motion po- Panorama Fault (P-P. F), Peristera Fault (PR. F), larities. Figure 2 shows the position of the seis- Asvestochori-Chortiatis Fault (A-Ch. F) and Pe- mological stations along with their code names. fka-Asvestochori Fault (P-A. F). The faults are During the operation period, an adequate segments of the recently defined complicated number of events were recorded both by the lo- fault zone named Thessaloniki-Gerakarou fault cal and the permanent network of Geophysics zone that strikes E-W from the city of Thessa- Department of Thessaloniki University. Eight of loniki towards the Stivos-Gerakarou Fault (S-G. the seismological stations of this seismological F) activated in 1978. network are located at distances less than 100 km b) NW-SE striking Lagina-Ag. Vasilios around the city. Since earthquake locations are Fault (L-AV. F). important for the seismotectonic interpretation c) NW-SE striking Assiros-Analipsi Fault of an area, attention must be given to the accura- (As-An. F). cy of the locations to relate the seismicity with d) WNW-ESE striking Anthemountas Ex- geological features. P- and S-arrival times and tensional Detachment Fault (AEDF), and the waveform durations were picked from sta- e) Nikopolis-Xilopolis Fault system (N-X. F). tions of both networks. With this process, 398 These studies revealed that the E-W-striking earthquakes were identified, and eventually faults are the most active ones associated with the those with clear phase arrivals that have been current seismic activity i.e. Thessaloniki-Ger- recorded from at least 4 stations were used in the akarou Fault Zone (TGFZ) (Tranos et al., 2003) present analysis. or recognized Quaternary activity, i.e. Anthe- Our data set consists of 277 earthquakes mountas Fault (AEDF, fig. 2) (Mountrakis et al., with at least 4 P- and 1 S-arrivals, which were 1996a,b). located using the Hypo71 (revised) computer
1083 Parthena M. Paradisopoulou et al.
Fig. 2. Epicentral distribution of 277 earthquakes that were recorded during the operation period in more than four stations. Also the morphology and main regional faults are shown. Triangles denote the positions of the twelve seismological stations along with their code names, and circles the earthquake epicenters that occurred in the study area during July 2001-April 2002 and October 2002-August 2003. Symbol sizes are proportional to earthquakes magnitude. The symbols and the abbreviations of the faults are as in fig. 1. program (Lee and Lahr, 1975). To ensure accu- Table I. P-wave velocity model used for the hypo- rate earthquake location a local velocity model center determination (modified from Papazachos was sought. Two such models have been pro- et al., 2000). posed for the broader Thessaloniki area P-wave velocity (km/s) Depth (km) (Hatzfeld et al., 1986/1987; Papazachos et al., 2000) and were used to locate 18 earthquakes 4.00 0.00 that had clear P- and S-arrivals and were 5.29 1.00 recorded by at least seven stations. It was ob- 5.36 1.50 served that the root-mean square values of the 5.76 2.00 time error were almost the same for both mod- 5.79 3.00 els. In order to minimize the root-mean square 6.16 4.00 value each model was empirically modified 6.23 6.00 several times. Finally an overlying upper layer 6.27 8.00 with a velocity of 4 km/s and a thickness of 1 6.30 10.00 km was added to a nine layer model which had 6.42 20.00 been proposed by Papazachos et al. (2000). The
1084 Microearthquake study of the broader Thessaloniki area (Northern Greece) finally adopted crustal model is presented in recorded from the local network and not located table I. From the data sample a νp/νs value by the permanent one, amplitudes and durations equal to 1.76 was estimated. The focal parame- on the seismograms were measured from the ters of the full dataset comprising 277 earth- recordings of the Thessaloniki permanent sta- quakes were estimated with an average root- tion. Appropriate formula (Scordilis, 1985) was mean-square value (rms) of 0.10 s (fig. 3a). The used for the local magnitude determination, horizontal (ERH) and vertical (ERZ) uncertain- which was subsequently converted to equivalent * ties in the hypocenter locations were less than 2 moment magnitudes (Mw) according to Papaza- km (fig. 3b,c respectively). Most earthquakes chos et al. (1997). The obtained magnitudes * have focal depths between 2 and 12 km with an range for 1.0≤ Mw ≤ 4.5. Figure 4 shows the average equal to 7 km (fig. 3d). magnitude distribution exhibiting completeness * Local magnitudes of the earthquakes re- for earthquakes of Mw ≥2.0 (ML ≥1.5). corded from both the local and the permanent The epicentral distribution of the 277 earth- network were taken from the routinely compiled quakes (fig. 2) covers a broad area around Thes- catalogue of the Geophysics Department of saloniki forming several distinct spatial clus- Thessaloniki University. For the earthquakes ters, some of them being clustered also in time.
a b
c d
Fig. 3a-d. Histograms of a) the root-mean square of the time errors, RMS; b) the horizontal, ERH; c) vertical, ERZ, uncertainties of the hypocentral determination, and d) focal depths of the located events.
1085 Parthena M. Paradisopoulou et al.
In general, the epicentral distribution follows the NW-SE grabens mentioned above. In a more detailed view, the earthquake clusters are aligned either WNW-ESE or E-W, forming en échelon structures that follow the NW-SE trend. Reliable fault plane solutions for events recorded at least at ten stations from both the lo- cal and the permanent networks were determined by using the FPFIT program (Reasenberg and Oppenheimer, 1985). We have calculated 64 fault plane solutions, listed in table II and shown in fig. 5. The majority of them exhibit nodal planes that strike E-W to WNW-ESE (mean strike ∼104¡) and reveal slip vectors that align with the Fig. 4. Magnitude distribution of the located NNE-SSW orientation. They indicate that the events. The data sample seems to be complete for study area is characterised by a prevailing exten- * magnitudes Mw ≥2.0. sional faulting driven by an extensional strain
Table II. Fault plane solutions for 64 microearthquakes. The strike (ξ¡), dip (δ¡) and rake (λ¡) correspond to the one nodal plane of the focal mechanism. The dip and azimuth of the P- and T-axis for each earthquake are given in last four columns.
* a/a Year Date Origin φ(¡N) λ(¡E) Depth Mw ξ(¡) δ(¡) λ(¡) P-axis T-axis time (km) Azimuth Dip Azimuth Dip 1 2001 22-Jul 15:21:15 40.737 22.729 16.55 2.6 80 68 Ð98 350 56 170 34 2 2001 0-Aug 21:49:39 40.673 23.376 18.41 3.7 102 60 Ð74 48 70 180 14 3 2001 11-Aug 18:44:49 40.662 23.435 17.82 3.1 128 52 Ð30 38 64 218 26 4 2001 21-Aug 06:01:22 40.720 23.125 4.12 2.3 132 58 Ð30 98 42 4 4 5 2001 28-Aug 23:58:30 40.889 22.835 15.07 2.2 90 50 −98 360 65 180 25 6 2001 3-Sep 10:04:16 40.679 23.080 10.32 2.3 132 56 −30 42 62 222 28 7 2001 15-Sep 17:47:13 40.813 22.995 11.23 2.7 130 48 −35 142 52 357 33 8 2001 15-Sep 20:09:52 40.816 23.001 12.43 2.4 132 44 −30 111 49 3 15 9 2001 28-Sep 21:06:37 40.657 22.991 10.50 2.0 114 60 −44 78 51 173 4 10 2001 8-Oct 04:50:21 40.596 23.116 13.70 4.5 70 34 −124 20 65 200 25 11 2001 8-Oct 05:25:31 40.597 23.143 14.18 2.5 110 50 −98 258 68 6 7 12 2001 8-Oct 05:26:44 40.593 23.116 13.68 4.1 74 42 −122 323 72 181 14 13 2001 8-Oct 05:28:44 40.595 23.164 13.73 2.6 82 60 −102 360 72 180 18 14 2001 8-Oct 05:29:28 40.589 23.145 14.40 3.4 90 36 −100 302 70 195 6 15 2001 8-Oct 05:32:16 40.589 23.126 13.57 3.9 88 54 −114 247 69 355 7 16 2001 8-Oct 05:36:20 40.588 23.148 13.78 3.1 64 42 −120 260 69 166 2 17 2001 8-Oct 06:52:48 40.593 23.124 14.78 3.3 56 50 −118 360 67 180 23 18 2001 8-Oct 06:55:29 40.599 23.111 13.23 3.2 90 46 −104 20 67 200 23 19 2001 8-Oct 07:04:27 40.588 23.133 13.14 3.0 110 46 −58 335 81 200 7 20 2001 8-Oct 07:15:19 40.588 23.134 14.10 3.0 104 52 −98 346 60 166 30 21 2001 8-Oct 07:16:09 40.593 23.128 14.18 2.9 76 60 −104 6 71 186 19 22 2001 8-Oct 07:25:54 40.591 23.125 14.56 2.9 96 38 −92 308 67 206 5
1086 Microearthquake study of the broader Thessaloniki area (Northern Greece)
Table II (continued).
* a/a Year Date Origin φ(¡N) λ(¡E) Depth Mw ξ(¡) δ(¡) λ(¡) P-axis T-axis time (km) Azimuth Dip Azimuth Dip
23 2001 8-Oct 07:26:32 40.596 23.117 14.76 2.8 96 54 −118 352 60 172 30 24 2001 8-Oct 08:00:25 40.590 23.126 13.45 3.6 82 60 −112 307 81 187 5 25 2001 8-Oct 08:21:26 40.595 23.128 14.43 2.9 96 50 −100 20 72 200 18 26 2001 8-Oct 13:47:57 40.600 23.124 14.13 2.1 110 36 −90 322 79 199 6 27 2001 8-Oct 13:49:19 40.596 23.123 14.13 2.1 100 52 −102 360 70 180 20 28 2001 8-Oct 19:00:28 40.593 23.134 13.86 1.6 90 40 −106 360 70 180 20 29 2001 9-Oct 06:59:05 40.597 23.169 16.21 2.7 90 40 −108 20 71 200 19 30 2001 11-Oct 16:31:46 40.584 23.152 12.75 3.0 110 38 −80 301 53 211 1 31 2001 14-Oct 09:06:00 40.555 23.582 16.27 2.7 90 56 −136 20 69 200 21 32 2001 25-Nov 20:09:46 40.590 23.183 14.48 2.4 110 42 −90 126 51 7 21 33 2001 4-Dec 17:03:51 40.693 22.956 6.51 2.1 138 36 −30 36 66 147 10 34 2001 13-Dec 08:48:35 40.475 22.717 13.11 2.5 76 58 −64 115 50 2 18 35 2001 14-Dec 22:39:00 40.791 23.018 13.07 2.1 132 40 −30 156 52 11 33 36 2002 17-Jan 06:35:56 40.781 22.945 14.28 2.2 148 20 −30 240 78 12 8 37 2002 24-Feb 09:31:13 40.929 22.967 19.65 2.5 92 38 −104 64 63 176 11 38 2002 24-Feb 21:50:00 40.575 23.257 16.70 2.3 106 60 −62 271 57 170 7 39 2002 7-Mar 08:08:12 40.748 22.719 19.28 2.8 54 60 −128 10 78 190 12 40 2002 18-Mar 10:42:02 40.609 23.005 12.90 1.5 100 24 −98 69 57 170 7 41 2002 13-Apr 08:10:41 40.702 23.132 10.27 3.0 106 60 −52 343 73 195 14 42 2002 13-Apr 08:13:34 40.707 23.133 10.55 3.0 98 60 −100 71 60 176 9 43 2002 13-Apr 13:52:41 40.710 23.127 10.02 2.2 110 60 −56 16 60 196 30 44 2002 13-Apr 17:46:06 40.699 23.139 10.63 2.3 128 60 −30 76 48 169 2 45 2002 13-Apr 21:06:01 40.706 23.126 9.64 1.7 112 60 −40 89 47 180 1 46 2002 16-Apr 11:11:39 40.682 23.169 8.01 2.7 124 60 −38 237 66 4 15 47 2003 29-Oct 21:15:48 40.702 22.785 10.80 4.5 76 70 −140 346 55 166 35 48 2002 21-Oct 09:10:07 40.706 23.102 9.62 2.1 150 40 −42 60 70 240 20 49 2002 4-Nov 01:01:56 40.791 23.021 11.97 2.5 142 20 −30 52 80 232 10 50 2002 24-Nov 19:43:34 40.754 23.132 13.67 2.2 140 40 −44 50 70 230 20 51 2002 25-Nov 00:15:00 40.754 23.125 13.37 1.9 106 36 −30 16 72 196 18 52 2002 15-Dec 04:45:12 40.619 22.927 14.75 2.2 96 50 −96 6 65 186 25 53 2002 18-Dec 11:29:33 40.757 23.054 11.47 1.8 132 40 −48 42 70 222 20 54 2003 1-Jan 04:48:00 40.717 22.747 10.33 1.1 84 60 −150 354 60 174 30 55 2003 6-Jan 17:34:00 40.764 23.081 9.05 1.4 120 46 −48 30 67 210 23 56 2003 10-Jan 11:38:30 40.679 22.962 2.95 1.3 118 40 −110 28 70 208 20 57 2003 11-Jan 10:01:28 40.699 23.131 4.12 1.4 110 28 −46 20 76 200 14 58 2003 12-Jan 15:47:04 40.854 23.000 7.39 1.6 108 44 −108 18 68 198 22 59 2003 18-Jan 14:10:42 40.718 22.449 19.49 2.0 112 20 −64 22 80 202 10 60 2003 4-Feb 04:12:08 40.640 22.812 8.46 1.7 144 28 −48 54 76 234 14 61 2003 9-Feb 02:10:36 40.821 22.942 1.04 1.9 104 24 −110 14 78 194 12 62 2003 10-Feb 05:09:24 40.659 23.440 21.21 2.9 148 22 −148 58 79 238 11 63 2003 10-Feb 15:46:52 40.372 22.653 3.19 1.8 96 60 −150 6 60 186 30 64 2003 10-Feb 18:12:37 40.801 22.820 5.40 1.7 106 20 −126 16 80 196 10
1087 Parthena M. Paradisopoulou et al.
Fig. 5. Fault plane solutions shown as equal-area lower hemisphere projections. The numbers correspond to the time order of earthquake occurrence (table II). The balloon size is a function of the earthquake magnitude. Com- pressional quadrants are shaded. The inset shows the nodal planes considered as the fault planes, the correspon- ding slip vectors and the mean position of T- and P-axes. The symbols and the abbreviations of the faults are as in fig. 1. regime, in accordance with previously published to fit the trend of the cluster, the strike of the fault plane solutions (Papazachos et al., 1979b; nearest fault segment and the focal mechanisms Carver and Bollinger, 1981; Soufleris et al., 1982, of the earthquakes occurred inside the polygon, 1983; Hatzfeld et al., 1986/1987; Hatzidimitriou and particularly that of the largest earthquake of et al., 1991). the cluster. Cross sections trending perpendicular to the mean strike of the respective fault segment were made, trying to identify the possible rela- 4. Seismotectonic analysis tionship between the microearthquake clusters and the known active faults. It must be mentioned The earthquakes’ spatial distribution, fault that in this procedure only the earthquakes inside plane solutions and neotectonic information were each polygon were considered. combined to recognize the seismotectonic prop- The Nea Aghialos swarm (polygon A in fig. erties of the study area. Earthquake locations 6) consists of 13 earthquakes that occurred at a from previous local experiments (Hatzfeld et al., distance of 15 km NW of the city of Thessa- 1986/1987; Hatzidimitriou et al., 1991; Papaza- loniki, close to the Nea Aghialos village. The * chos et al., 2000) were taken into account. Earth- larger earthquake with Mw =4.5 occurred on 29 quakes are considered in clusters defined by October 2003 and the seismic sequence contin- polygons and assigned to specific fault segments ued for twelve days. At this period the local net- (fig. 6). The shape of each polygon was selected work was not operational for software upgrade
1088 Microearthquake study of the broader Thessaloniki area (Northern Greece) purposes. The earthquakes of the seismic se- reaching a depth of 12 km. The only geological quence were recorded from the permanent net- information (Tranos, 2002) available for this work, from at least 5 stations located near Thes- area refers to the existence of an E-W striking saloniki. Using HypoDD program (Waldhauser, fault that dips at a high angle to the S. 2001) from 34 recorded earthquakes only 13 The Efkarpia-Asvestochori cluster (polygon were selected with at least four P- and one S-ar- B in fig. 6) comprises several microearthquakes rivals and their locations were determined with that disperse along the mountainous Chortiatis the mean rms error. The aftershock spatial dis- region showing a WNW-ESE preferred orienta- tribution shows an ENE-WSW trend and focal tion and with focal depth ranging up to 15 km. depths range between 2 and 10 km. The fault The nodal planes of the determined focal mech- plane solutions of the earthquakes in this area anisms (No. 9, 33, 40, 56, 58 in table II and fig. (No. 1, 39, 47, 54 in table II and fig. 5) and the 5) strike WNW-ESE with a mean value of 130¡. alignment of the hypocenters in the cross-sec- Taking into account the corresponding cross- tion (fig. 7A) indicate that the rupture zone, section (fig. 7B), this rupture zone should dip at namely Nea Aghialos Fault (NA. F) strikes about 50¡ to the NE, in good agreement with the ENE-WSW (∼80¡) and dips at 70¡ to SSE, NW-SE (130¡) striking Pefka-Asvestochori
Fig. 6. Epicentral distribution of the earthquakes recorded by the local experiments (Hatzfeld et al., 1986/1987; Hatzidimitriou et al., 1991; Papazachos et al., 2000 and the experiment carried out in June 2001-April 2002 and October 2002-August 2003). White polygons enclose the clusters for which cross sections were performed. The hypocentres included in each polygon were projected onto the lines that are perpendicular to the strike of the corresponding fault segment.
1089 Parthena M. Paradisopoulou et al.
Fig. 7. Depth section of the earthquakes within the six boxes shown in fig. 6. Gray lines depict the fault dip. The mechanisms are shown as equal area projections of the front hemisphere.
Fault segment (P-A. F) of the Thessaloniki-Ger- curred in this area with a main shock of M=4.5 akarou Fault Zone (TGFZ) that dominates in the on 8 October 2001 and lasted for six days. The area and dips to NNE (Tranos et al., 2003). A focal mechanisms (No. 10-32 in table II and fig. few earthquakes located at larger depths (10-15 5) have nodal planes striking WNW-ESE with a km) are probably associated with a smaller fault mean value of 114¡ and dip angles of about 50¡. parallel to the Pilea-Panorama Fault (P.-P. F). The cross-section (fig. 7C), which is perpendi- The cluster in Chortiatis mountainous area cular to the strike of the Thessaloniki-Ger- includes about 50 earthquakes that form a akarou Fault Zone (TGFZ) indicates an east- WNW-ESE to E-W trending seismic band ward prolongation of the Pilea-Panorama Fault (polygon C in fig. 6). A seismic sequence oc- (dashed line in fig. 6, PR. F).
1090 Microearthquake study of the broader Thessaloniki area (Northern Greece)
In the cluster of polygon D the fault plane (fig. 7E) indicates that the rupture zone dips at solutions (No. 4, 41, 42, 43, 44, 45, 46, 48, 50, an angle of ∼45¡ towards SW, which agrees and and 51, in table II and fig. 5) clearly show that fits well with the northwestward extended to- the one nodal plane dips at about 55¡ towards wards NW trace of the Assiros-Analipsi Fault SW, and the other dips about 60¡ towards NW. (As-An. F, fig. 6), indicating the association of Focal mechanisms along with the cross-section the microseismicity with this fault. (fig. 7D) reveal an active structure that strikes The cluster of polygon F is located in an NW-SE and dips at 55¡ towards SW, probably area where ENE-WSW striking faults form a associated with the Assiros-Analipsi Fault (As- fault array, named Nikopolis-Xilopolis Fault An. F). Additionally, its kinematics, defined by system (N-X. F) that bound very small and focal mechanisms is an oblique extensional shallow Quaternary basins (Mountrakis et al., movement with a right lateral strike-slip com- 1996a,b). The cross-section (fig. 7F) indicates ponent that agrees fairly well with the kinemat- that this active structure dips at a high angle to- ics suggested by neotectonic studies (Mercier ward SE and is rather related to south dipping et al., 1983; Mountrakis et al., 1983; Koukou- faults of the aforementioned system. velas and Aydin, 2002). The maximum extension axes (T) derived Fault plane solutions of events inside the from the focal mechanisms determination (fig. 8) polygon E (fig. 6) (No. 8, 9, 35, 49, 53 and 55 are sub horizontal trending almost N-S, in agree- in table II and fig. 5) are similar to those of the ment with previous investigations (Hatzfeld previous clusters. The distribution of the hypo- et al., 1986/1987; Scordilis et al., 1989; Hatzidi- centers on the SW-NE striking cross section mitriou et al., 1991; Papazachos et al., 2000). In
Fig. 8. Spatial distribution of the axis of maximum extension (T-axis) provided by fault plane solutions. The symbols and the abbreviations of the faults are as in fig. 1.
1091 Parthena M. Paradisopoulou et al. particular, the mean trend of the T-axes has a Given the fact that this study concerns one NNW-SSE strike in the area of N. Anghialos of the most populated areas in Greece, the ac- Fault (NA-F) and along the Chortiatis mountain tive structures identified in this work could be chain (P-P. F and A-Ch. F). To the north of the taken into account in future seismic hazard as- Langada Lake (D area) the mean T-axis trends sessment. N-S, whereas similar N-S trending has been de- fined for the area south of Lake Langada. How- ever, in some cases a significant variation is ob- Acknowledgements served with the orientation of the T-axes ranging from NNW-SSE to NE-SW. The NE-SW orient- The authors express their appreciation to the ed T-axes are probably associated with the NW- reviewer for the careful inspection and con- SE trending inherited structures, which can be structive comments. This work was supported reactivated under the dominating N-S extension. by the Region of Central Macedonia and Mu- The results confirm a previous study (Pa- nicipality of Thessaloniki. The GMT system pazachos et al., 2001) using the combined stress (Wessel and Smith, 1998) was used to plot the pattern and the corresponding moment rate ten- figures. Geophysics Department Contribution sors. From the determination of strain and ve- 655. locity tensors it was found that the correspon- ding extensional velocity is 0.63 mm/yr at the northern part of the study area and 0.21 mm/yr REFERENCES at the southern part (Papazachos et al., 2001). CARVER, D. and G.A. BOLLINGER (1981): Aftershocks of the June 20, 1978, Greece earthquake: a multimode fault- ing sequence, Tectonophysics, 73, 343-363. 5. Conclusions HATZFELD, D., A.A. CHRISTODOULOU, E.M. SCORDILIS,D. PANAGIOTOPOULOS and P.M. HATZIDIMITRIOU (1986/ 1987): A microearthquake study of the Mygdonian A microearthquake study around the city of graben (Northern Greece), Earth Planet. Sci. Lett., 81, Thessaloniki provides additional information 379-396. on the seismotectonic properties of the area. Al- HATZIDIMITRIOU, P.M., D. HATZFELD, E.M. SCORDILIS, E.E. PAPADIMITRIOU and A.A. CHRISTODOULOU (1991): Seis- though the epicentral distribution exhibits a motectonic evidence of an active normal fault beneath NW-SE general trend, the observed clusters Thessaloniki (Greece), Terra Motae, 3, 648-654. correlate better with the known active structures KOUKOUVELAS, I.K. and A. AYDIN (2002): Fault structure of the area, which are orientated E-W. This also and related basins of the North Aegean Sea and its sur- advocates the fact that the majority of fault roundings, Tectonics, 21, doi: 10.1029/2001TC901037. LEE, W.H.K. and J.C. LAHR (1975): A computer program plane solutions exhibit extensional faulting for the determining hypocenter, magnitude and first along the E-W strike with the T-axis oriented motion pattern of local earthquakes, U.S. Geol. Surv. almost N-S, in accordance with previous inves- Open File Rep., 75-311. MERCIER, J.L. and E. CAREY-GAIHARDIS (1989): Regional tigations (Papazachos et al., 1979b; Carver and state of stress and characteristic fault kinematic insta- Bollinger, 1981; Soufleris et al., 1982, 1983; bilities shown by aftershock sequences: the aftershock Hatzfeld et al., 1986/1987; Hatzidimitriou et al., sequences of the 1978 Thessaloniki (Greece) and 1980 1991). However, some microearthquake fault Campania-Lucania (Italia) earthquakes as examples, Earth Planet. Sci. Lett., 92, 247-264. plane solutions define T-axes that deviate sig- MERCIER, J.L., E. CAREY-GAILHARDIS, N. MOUYARIS,K. nificantly from the mean T-axis value, possibly SIMEAKIS,T. ROUNDOYANNIS and C. ANGHELIDHIS indicating that the inherited fault pattern partly (1983): Structural analysis of recent and active faults influences the seismicity. and regional state of stress in the epicentral area of the 1978 Thessaloniki earthquakes (Northern Greece), Tec- In addition, the present approach seems to tonics, 2, 577-600. be more important since it permits the detection MOUNTRAKIS, D., A. PSILOVIKOS and B. PAPAZACHOS of blind active faults such as the case of the Nea (1983): The geotectonic regime of the Thessaloniki Anghialos Fault, where the geological informa- earthquakes, in The Thessaloniki, Northern Greece, Earthquake of June 20, 1978 and Its Seismic Sequence, tion is not adequate and a recent seismic se- edited by B.C. PAPAZACHOS and P.G. CARYDIS, Techni- quence occurred. cal Chamber of Greece, 11-27.
1092 Microearthquake study of the broader Thessaloniki area (Northern Greece)
MOUNTRAKIS, D., A. KILIAS, S. PAVLIDES, L. SOTIRIADIS,A. macedonian zone and the surrounding area, Ph.D. The- PSILOVIKOS,TH. ASTARAS, E. VAVLIAKIS, G. KOUFOS,G. sis (University of Thessaloniki, Greece), pp. 250. DIMOPOULOS, G. SOULIOS, V. CHRISTARAS, M. SKO- SCORDILIS, E.M., G. KARAKAISIS, E. PAPADIMITRIOU and B. RDILIS, M. TRANOS, N. SPYROPOULOS, D. PATRAS,G. MARGARIS (1989): Microseismicity study of the Servo- SYRIDES, N. LAMBRINOS and T. LAGGALIS (1996a): Neo- macedonian zone and the surrounding area, Geologica tectonic Map of Greece, Langadhas Sheet, Scale Rhodopica, 1, 79-83. 1:100000 (Earthq. Plann. Prot. Organ. and Europ. Cen- SOUFLERIS, C. and G.S. STEWARD (1981): A source study of tre Prevent. Forecasting Earthq.). the Thessaloniki (Northern Greece) 1978 earthquake MOUNTRAKIS, D., A. KILIAS, S. PAVLIDES, L. SOTIRIADIS,A. sequence, Geophys. J. R. Astron. Soc., 67, 343-358. PSILOVIKOS,TH. ASTARAS, E. VAVLIAKIS, G. KOUFOS,G. SOUFLERIS, C., J.A. JACKSON, G.C.P. KING, C.P. SPENCER DIMOPOULOS, G. SOULIOS, V. CHRISTARAS, M. SKO- and C.H. SCHOLZ (1982): The 1978 earthquake se- RDILIS, M. TRANOS, N. SPYROPOULOS, D. PATRAS,G. quence near Thessaloniki (Northern Greece), Geophys. SYRIDES, N. LAMBRINOS and T. LAGGALIS (1996b): J. R. Astron. Soc., 68, 429-458. Neotectonic Map of Greece, Langadhas Sheet, Scale SOUFLERIS, C., J.A. JACKSON, G.C.P. KING and C.P. SPENCER 1:100000 (Earthq. Plann. Prot. Organ. and Europ. Cen- (1983): Thessaloniki 1978 Earthquakes: Locally tre Prevent. Forecasting Earthq.). Recorded Aftershocks. The Thessaloniki, Northern PAPAZACHOS, B.C. and C. PAPAZACHOU (2003): The Earth- Greece Earthquake of June 20 1978 and its Seismic Se- quakes of Greece (Ziti Publications, Thessaloniki), pp. quence, edited by B.C. PAPAZACHOS and P.G. CARYDIS 317. (Thessaloniki), 159-185. PAPAZACHOS, B.C., D. MOUNTRAKIS, A. PSILOVIKOS and G. TRANOS, M.D. (1998): Contribution to the study of the neo- LEVENTAKIS (1979a): Surface fault traces and fault tectonic deformation in the Region of Central Macedo- plane solutions of the May-June 1978 major shocks in nia and North Aegean, Ph.D. Thesis, (University of the Thessaloniki area, Greece, Tectonophysics, 53, Thessaloniki, Greece), pp. 346 (in Greek with extend- 171-183. ed English summary). PAPAZACHOS, B.C., A. MOUNTRAKIS,A. PSILOVIKOS and G. TRANOS, M.D. (2002): Geological study of the Dialogi re- LEVENTAKIS (1979b): Focal properties of the 1978 earth- gion. Municipality of Eleftherio-Kordelio, Perfecture quakes in Thessaloniki area, Bulg. Geophys. J., 6, 72-80. of Thessaloniki Report, pp. 76. PAPAZACHOS, B.C., A.A. KIRATZI and B.G. KARAKOSTAS TRANOS, M.D. and D.M. MOUNTRAKIS (1998): Neotectonic (1997): Toward an homogeneous moment magnitude joints of Northern Greece: their significance on the un- determination in Greece and surrounding area, Bull. derstanding of the active deformation, Bull. Geol. Soc. Seism. Soc. Am., 87, 474-483. Greece, 32, 209-219. PAPAZACHOS, C.B., P. SOUPIOS,A. SAVVAIDIS and Z. ROUME- TRANOS, M.D., A.A. KILIAS and D.M. MOUNTRAKIS (1999): LIOTI (2000): Identification of small-scale active faults Geometry and kinematics of the Tertiary post-meta- near metropolitan areas: an example from the Asvesto- morphic Circum Rhodope Belt Thrust System chori fault near Thessaloniki, in Proc. XXXII ESC Gen. (CRBTS), Northern Greece, Bull. Geol. Soc. Greece, Ass., 15-20 September 2000, Lisbon, Portugal, 221-225. 33, 5-16. PAPAZACHOS, C.B., D.A. VAMVAKARIS, G.N. VARGEMEZIS TRANOS, M.D., E.E. PAPADIMITRIOU and A.A. KILIAS and E.V. AIDONA (2001): A study of the active tecton- (2003): Thessaliniki-Gerakarou Fault Zone (TGFZ): ics and deformation in the Mygdonia Basin (N. the western extension of the 1978 Thessloniki earth- Greece) using seismological and neotectonic data, quake fault (Nothern Greece) and seismic hazard as- Bull. Geol. Soc. Greece, XXXIV, 303-309. sessment, J. Struct. Geol., 25, 2109-2123. PARADISOPOULOU, P.M., V.G. KARAKOSTAS, E.E. PAPADIM- TRANOS, M.D., I.D. MELADIOTIS and E.P. TSOLAKOPOULOS ITRIOU, M.D. TRANOS, C.B. PAPAZACHOS and G.F. (2004): Geometrical characteristics, scaling properties KARAKAISIS (2004): Microearthquake study of the and seismic behavior of the faulting of the Chortiatis broader Thessaloniki area, in 5th Intern. Symp. East. region and Anthemountas Basin (Northern Greece), in Mediter. Geol., 14-20 April 2004, Thessaloniki, 5th Intern. Sympos. Eastern Mediterranean Geology, Greece, 2, 623-626. Thessaloniki, Greece, edited by A.A. CHATZIPETROS PAVLIDES, S.B. and A.A. KILIAS (1987): Neotectonic and ac- and S.B. PAVLIDES, 2, 889-891. tive faults along the Serbomacedonian zone (SE WALDHAUSER, F. (2001): A program to compute double-dif- Chalkidiki, Northern Greece), Ann. Tectonicae, 1, 9-104. ference hypocenter locations, U.S. Geol. Surv. Open PAVLIDES, S., D. MOUNTRAKIS, A. KILIAS and M. TRANOS File Rep., 1-113. (1990): The role of strike-slip movements in the exten- WESSEL, P. and W.H.F. SMITH (1998): New, improved ver- sional area of the Northern Aegean (Greece), in Active sion of the generic mapping tools released, Eos Trans. and Recent Strike-slip Tectonics, edited by M. BOC- Am. Geophys. Un., 79, 579. CALETTI and A. NUR, Ann. Tectonicae, 4, 196-211. REASENBERG, P. and D. OPENHEIMER (1985): Fpfit (U.S. Ge- ological Survey, Menlo Park, California). (received January 25, 2006 SCORDILIS, E.M. (1985): Microseismic study of the Serbo- accepted April 4, 2006)
1093 International Journal of Disaster Risk Reduction 17 (2016) 49–66
Contents lists available at ScienceDirect
International Journal of Disaster Risk Reduction
journal homepage: www.elsevier.com/locate/ijdrr
Accessibility modeling in earthquake case considering road closure probabilities: A case study of health and shelter service accessibility in Thessaloniki, Greece
Kivanc Ertugay a,n, Sotiris Argyroudis b,H.Şebnem Düzgün c a Selcuk University, Department of Urban and Regional Planning, Konya, Turkey b Department of Civil Engineering, Aristotle University, Thessaloniki, Greece c Department of Mining Engineering, Earthquake Engineering Research Center, Middle East Technical University, Ankara, Turkey article info abstract
Article history: One of the main focuses of disaster management is to decrease the impact of disasters by developing Received 3 December 2015 efficient emergency management plans and strategies. In this context, accessibility modeling of emer- Received in revised form gency related services such as health, fire, security services, or shelter sites in earthquake case can be a 14 March 2016 valuable tool for disaster managers, especially in assessing the vulnerability of urban districts as well as Accepted 17 March 2016 the sufficiency and appropriateness of the emergency service sites. Although there are several studies on Available online 25 April 2016 the accessibility modeling of emergency related services in the aftermath of a disastrous event, the lit- Keywords: erature that integrates road closure probabilities into the accessibility estimation is quite limited. Hence, Accessibility modeling this paper aims to develop a methodology for modeling of accessibility based on a probabilistic esti- Road network mation of road closures in earthquake case. The mentioned probabilities are predicted for given road Road closure probability network and hazard acting upon it and then incorporated into the accessibility analysis. The developed Location/allocation Service area methodology is demonstrated through an application on health services and shelter sites in the city of Health services Thessaloniki in Greece with and without considering the road closures, corresponding to road network Shelter sites under earthquake and normal conditions respectively. In this particular study, the road closures are Disaster/emergency management associated to ground failure and damage to bridges, as well as damage of overpasses and collapses of Earthquake buildings adjacent to road edges. The developed methodology could provide advanced and more realistic Vulnerability of city infrastructures and ur- support to the emergency managers and decision-makers who deal with accessibility, location/alloca- ban districts tion, and service/catchment area of critical facilities. & 2016 Elsevier Ltd. All rights reserved.
1. Introduction Several researchers studied the vulnerability of transportation networks in case of earthquake following different approaches. Past experiences show that transportation networks are vul- They are distinguished according to the time scale (emergency or nerable to earthquakes and failure of their components can cause recovery phase), the geographical scale (e.g. urban, regional), the economic losses, disruption of activities and can affect the rescue objectives (e.g. emergency or mitigation planning) and the type of fi fl and evacuation procedures in the area. Therefore, one of the main analysis (e.g. connectivity, traf c ow, estimation of direct and goals of disaster management is to reduce the impact of disasters indirect losses). Several studies focus on the emergency operations immediately after the earthquake [24,36,57]. In some cases the by developing efficient emergency plans and risk mitigation stra- impact of collapsed buildings in the emergency services is also tegies. In this respect, accessibility modeling of emergency related considered [29]. The purpose is to identify which portions of the services such as health, fire and security services or shelter sites network have the higher risk for loss of connectivity. In other could provide a vital decision support for disaster managers to studies the losses due to physical damage and travel delays are fi assess the vulnerability of urban districts and to test the suf - evaluated [35], while the exposure of the network to ground ciency and appropriateness of the emergency service sites for an shaking, liquefaction and landslides is considered separately in the earthquake disaster. analysis [52]. Fragility curves are essential tools in evaluating the vulnerability of each component of the network. They describe the
n probability that a structure will reach or exceed a certain damage Corresponding author. E-mail addresses: [email protected] (K. Ertugay), state for a given ground motion motion. An overview of existing [email protected] (S. Argyroudis), [email protected] (H.Ş. Düzgün). fragility models for bridges, is made by Tsionis and Fardis [56], http://dx.doi.org/10.1016/j.ijdrr.2016.03.005 2212-4209/& 2016 Elsevier Ltd. All rights reserved. 50 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 while Argyroudis and Kaynia [4] summarize existing fragility and service/catchment area related issues. The GIS based accessi- curves for other road elements. Some researchers studied different bility modeling techniques can generally be divided into three aspects in the vulnerability assessment of bridges which are the categories, which are briefly presented hereinafter and described most critical elements. These for example include the performance in detail in Makrí and Folkesson [41] [34,14,18,20]. of integral and jointed bridges [46], or the effect of aging and environmental deterioration [2,28,60] and the impact of these 2.1. Zone-based technique processes to the seismic resilience of bridges and highway net- works [3,7]. Accessibility is computed for predefined zones, which are de- Although there are increasing literature on vulnerability as- termined based on the available data, the objectives and required sessment of transportation networks in case of earthquake, the detail of the study. While a national or regional scale accessibility research that integrate the vulnerability of transportation com- study generally requires a coarse zone representation such as ponents into accessibility modeling process is quite limited, which state, county or district, a local/city scale study usually involves fi can be considered as a signi cant lack in terms of providing ac- smaller zones such as neighborhood, quarter or parcel. It must be cessibility related support for the emergency related decision noted that is more difficult to obtain the required data for the makers. Hence, this paper aims to propose a methodology for smaller zones than the coarse zones [31]. In this method, traveling accessibility modeling based on a probabilistic estimation of road cost calculation between supply and demand points are usually closures in case of earthquake. In particular, the closure prob- based on the centroid of the zones. An example in GIS environ- abilities are predicted for a given road network and then in- ment is given in Fig. 1 (see several examples for detail in corporated into the accessibility analysis. The proposed metho- [14,31,37,12,11,59]). The basic advantage of this technique is that it dology is demonstrated through an application on health services enables easier comparison of accessibility scores between the and shelter sites in the city of Thessaloniki in Greece. In this ap- zones such as administrative boundaries. plication, road closures due to ground failure, damage to bridges In zone-based technique the cumulative time/distance cost is and overpasses and collapses of buildings adjacent to road edges calculated from each administrative district centroid to all ser- are considered. The accessibility modeling is performed both un- vices. For the calculation of accessibility score for each of the ori- der earthquake and normal conditions. In the proposed metho- gins ( Ai), Eq. (1) is used: dology, three major steps are followed: (a) Data collection and analysis phase. (b) Traveling cost calculation phase. 1 Ai = ∑ time OR distance() i, j (c) Accessibility modeling and visualization phase. The proposed ()j ()1 approach has been developed in the framework of the EU FP7 SYNER-G project [49], as part of a general methodology for the According to Eq. (2),(i) is the origin, (j) is the destination, (Ai)is systemic vulnerability and risk analysis of networks and the accessibility score for each of the origins. Ai is inversely pro- infrastructures. portional with the time or distance cost from one origin (centroid of the zone) to all destinations (health service or shelter site), which means that the less time or distance cost between districts and services the better is the accessibility. 2. Overview of the accessibility modeling techniques in GIS When a gravity measure (G) such as size, capacity or scale of services is accounted in the accessibility measure, then Eq. (2) is Physical accessibility measures describe the spatial character- used: istics of the study area and need large amount of computation and organization between huge and complex spatial data sets. Hence, Gj() Ai = ∑ accessibility modeling is commonly based on GIS technologies in timeOR distance() i , j ()j ()2 terms of data collection, manipulation, programming, analysis and presentation of results. As GIS have unique capabilities to handle According to Eq. (2),(i) is the origin, (j) is the destination, (G)is and analyze geospatial data, it provides a powerful interface to the the gravity measure of the service, (Ai) is the accessibility score for decision makers who deal with accessibility, location/allocation each of the origins. In this case Ai is directly proportional with the
Fig. 1. Centroid of a zone, zone-based representation of accessibility (Source: [14,18]). K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 51
Fig. 2. The isochronal representation of accessibility (Source: [18]). gravity of the services, which means that the higher the gravity of within a threshold considering some gravity coefficients that the services, the better the accessibility. describe the importance of the services based on different cri- teria such as capacity, size, type and role of the service). 2.2. Isochrone-based technique The isochrone-based technique is generally preferred for better Accessibility is evaluated in terms of isochronal boundaries, representation of service/catchment areas for a given time/dis- which are also known as the catchment or service area bound- tance threshold. Although this technique is widely used in the aries. They are obtained by connecting equal travel time or dis- literature, a major shortcoming is that accessibility measures are fi tance points away from one or more reference points (e.g. supply highly sensitive to traveling time/distance costs and user de ned or demand). In particular, they are calculated from either constant thresholds. Consideration of several costs and thresholds, can average traveling costs such as 120 km/h for highways, 50 km/h provide a more realistic support to decision makers (see several for main streets, 30 km/h for local streets or unconstrained costs examples in [15,44,10,26,9,45,59,47]). based on Euclidean distance (straight-line/bird-flight based dis- tances) such as buffer and Voronoi (Thiessen) polygons without 2.3. Raster-based technique considering the transportation network (see several examples in [17,50,23,40,42,39,8,51,53,30,32,43,38,58]). When an origin is de- Accessibility measures are represented by raster-based “pixels”, fined as a reference point such as a demand or supply location, also called “cells”, the smallest unit in the raster environment. By isochronal boundaries can be drawn by connecting all points in all considering traveling costs in the transportation network, each directions for an equal threshold of time or distance (Fig. 2). The pixel in raster environment generally gets an accessibility score, most frequent usage of isochronal based technique is to determine which is based on its proximity to the nearest supply or demand the catchment/service area of the facilities for a defined threshold. point (Fig. 3). Raster-based technique is generally preferred in The buffer and Voronoi based isochronal boundaries have regional studies, which do not require high spatial accuracy. regular shape because of their unconstrained structure. However, However, this technique enables continuous representation of transportation networks based on isochronal boundaries can have accessibility scores and opens a wide range of new raster analysis irregular shape as the traveling costs can be lower in some di- capabilities (see several examples in [16,13,33,59]). rections than in the others [55]. Isochrone-based technique can be used in various accessibility measures, for example: 3. The proposed methodology “x” min/m catchment area of supply/demand points can be calculated as a “travel time/distance” based measures (i.e. the time or distance The accessibility modeling considering road closures due to based polygon boundaries for some services for a given threshold, seismic damages includes the following three phases which are e.g. the polygon boundaries which are 5 min/5 km to schools) or incorporated in a common GIS environment: total number of cumulative supply/demand points within “x” min/m catchment area boundary can be calculated as a “cu- Data collection phase; where the necessary data are obtained mulative opportunity” type of measure (i.e. the total number of and converted into a common format. services/opportunities within a threshold, e.g. the number of Traveling cost calculation phase; where road closure prob- population within 5 min to schools) or abilities are predicted for each segment of the given road net- total number of weighted supply/demand points within “x” min/ work and then integrated into the accessibility modeling. m catchment area boundary can be calculated as a “gravity” type Accessibility modeling and visualization phase; where health of measure (i.e. the total number of services/opportunities service and shelter site accessibility is modeled considering 52 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 3. The raster-based representation of accessibility (Source: [34,21]).
Fig. 4. The five seismogenic zones (black polygons) considered for the case study (red rectangle). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
road closure probabilities by using the techniques described in Supply locations, which include emergency related services (such Section 2. as hospitals, fire stations, security services, shelter sites etc). Demand locations which are administrative zones of the study area. 3.1. Data collection phase Road closure probabilities.
The primary data required for the accessibility analysis in case 3.2. Traveling cost calculation phase: computation of road closure of earthquake includes the following: probabilities
The digital transportation network and its hierarchy for the In case of strong earthquakes, roads can be closed due to study area. ground failure (e.g. soil liquefaction), heavy damage or collapse of K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 53
TRANSPORTATION NETWORK Freeways Main arterial Secondary arterial Primary collective Secondary collective
± 021Km
Fig. 5. The road network data of Thessaloniki. tunnels, bridges and overpasses, failure of pipelines beneath the Performance Indicators (PIs) for components and systems, and for road segments, collapse of buildings or electric poles adjacent to the infrastructure as a whole. A probabilistic evaluation of the road edges, debris flow due to landslides. In the present study, an performance of networks is carried out by means of Monte Carlo integrated approach for the probabilistic systemic risk analysis of simulations. In particular, the network is analysed for separate the road network is applied and then incorporated into the ac- sampled scenarios and the results are statistically combined, in cessibility analysis. The model considers spatially correlated seis- order to obtain the performance of the network in the form of a mic hazard, vulnerability of the network components, and the risk curve (i.e. exceedance probability of loss in performance as effect of interactions between roadway components and built measured by selected performance indicators) [25,49]. Fragility environment [5]. This approach is part of the general methodology functions are used to assess the physical damages for all the net- developed in SYNERG project [54] for the seismic vulnerability work components (e.g. bridges, road pavements, tunnels) in order assessment of lifelines and infrastructures accounting for inter- to define their physical state (i.e. closed or open to traffic) under and intra-dependencies among infrastructural components, as each sampled event. Functional consequences of physical da- well as for the uncertainties characterizing the problem. It is es- mages, as well as inter- and intra- dependencies among compo- sentially based on two steps: (i) the seismic hazard characteriza- nents and systems are considered such as the failure of bridge tion through realizations of ground motions and (ii) the prob- structures or blockages due to debris of collapsed buildings. abilistic evaluation of the target system's performance conditional The calculation of traveling costs just after the earthquake re- on such realizations. In particular, the development of the hazard quires calibration (reduction) of the average traveling speed on each model enables: (a) sampling events in terms of location (epi- of the road segments according to the calculated road closure centre), magnitude and faulting style according to the seismicity of probabilities and the type of the roads (i.e. major or secondary road). the study region and (b) maps of sampled correlated seismic in- tensities at the sites of the vulnerable components in the infra- For the major road segments like freeways and main arterials structure. These maps, conditional on the occurrence of one that remain open (i.e. the calculated road closure probability is earthquake with given magnitude and hypocenter describe the “0”), an average speed of 30 km/h is assigned. variability and spatial correlation of intensity measures at different For the secondary road segments, like secondary arterials, pri- sites, which is a fundamental request for the analysis of spatially mary and secondary collectives that remain open (i.e. the cal- extended networks. The development of the physical model for culated road closure probability is again “0”), an average speed the network starts from the SYNER-G taxonomy and requires: of 25 km/h is assigned. (a) for each system within the taxonomy, a description of the For the road segments that present a blockage risk, the above functioning of the system (intra-system dependencies) under both mentioned average speeds are reduced based on the value of undisturbed and disturbed conditions (i.e., in the damaged state the closure probability. In particular, the reduction of the tra- following an earthquake); (b) a model for the physical and func- veling costs is related to the closure probabilities as follows: tional (seismic) damageability of each component within each system (fragility functions); (c) identification of all dependencies All the road segments that present a blockage risk are char- between the systems (inter-dependencies); and (d) definition of acterized by a closure probability score. When a road segment is 54 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
! Hospitals (with ID labels) Possible Shelter Sites Transportation Network
!9 !1
!6 !7 !4 !3 !8 !0 !5
!2 ± 021Km
Hospital ID Hos pital Nam e Gravity Coefficient LATIPSOH"OITARKOPPI"0 6 IKINOLASSEHTFOLATIPSOHCIRTAIHCYHSP1 57.2 "SOLVPA.GA"IKINOLASSEHTFOLATIPSOHLARENEG2 5.3 IKINOLASSEHTFOLATIPSOHSESAESIDSUOITCEFNI3 5.1 A"IKINOLASSEHTFOLATIPSOHLARENEGLANOIGER4 5.5 LATIPSOHESAESIDNIKSDNACAISIDORHPA5 5.1 IRTIMID.GA"IKINOLASSEHTFOLATIPSOHLARENEG6 57.3 AMINNEG.G"IKINOLASSEHTFOLATIPSOHLARENEG7 4 LATIPSOHRECNACITNAOINEGAEHT8 52.4 "UOIGROEGAPAP"LATIPSOHLANOIGERLARENEG9 5.4
Fig. 6. The hospitals data with gravity coefficients and possible shelter site locations data of Thessaloniki. exposed to several threats (e.g. bridge damage, ground failure, After normalization process, the following rules are applied: building collapse etc), the highest closure probability score is ac- cepted as the final aggregated closure probability score of the road ○ if the normalized closure probability score is equal or less than segment. In other words, when there is an overlap between the 0.25, the traveling costs are reduced by 25%, probability scores for a road segment, the greater score is con- ○ if the normalized closure probability score is between 0.25 and sidered as final aggregated closure probability value for that 0.50, the traveling costs are reduced by 50%, segment. ○ if the normalized closure probability score is between 0.50 and
The final aggregated closure probability scores ( ei) of each 0.75, the traveling costs are reduced by 75% and segment are normalized (rescaled) in a 0–1 scale based on the ○ if the normalized closure probability score is equal or greater following formula: than 0.75, the traveling costs are reduced by 100% (which means the road segment is totally closed). eEimin− Normalized()= ei EE− max min The calibration of the traveling costs just after the earthquake where Emin is the minimum and Emax is the maximum probability has been selected based on expert judgment and could be score among all of the road closure probabilities. The reason of modified accordingly if real data become available. Although normalization process is to make understanding of the closure these judgments are subjective to experts they provide a rea- probability scores easier for the further accessibility modeling sonable representation of the post-earthquake network condi- operations. tions since they are based on closure probabilities. Hence, the K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 55
Districts District centroids (with District ID) Transportation Network
11 District ID Dis tr ict Nam e 12 0 Ladadika-Dikastirion squ. 6 1 St.George-Rotonda 7 2 University-Int.Exhib.Center-White T 0 3PAOK-Malakopi 1 4 Papafio-St.Constantine-Military Hos 5 Toumba-St.Fanourios-Ydragogio 2 6 Vlatadon-St.Dimitrios-Lachanagora 7 Municipal Hospital-Fylakes (Peniten 8Athinon 4 9 Analipsi-Dermatologiko 10 Nea Paralia-Municipal Library 5 19 11 Railw ay Station-Xirokrini 3 12 Pasha Hamam-Tyroloi 17 13 Charilaou-Exohes 15 16 14 Nomarchia-Vafopoulio-Depau 8 15 School for the blind-Sailing Club 9 18 10
13 14 ± 021Km *The transportation network data under district layers are hidden in figure in order to improve the quality of visualization.
Fig. 7. The administrative districts data of Thessaloniki.
Fig. 8. Probability of road closure due to ground failure and damage to bridges. 56 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 9. Probability of road closure due to damage to overpasses.
Fig. 10. Probability of road closure due to collapses of buildings adjacent to road edges. K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 57
Transportation Network Normalized Probability of Road Closure
0.005 - 0.250 0.251 - 0.500 0.501 - 0.750 0.751 - 1.000
± 021Km
Fig. 11. The final aggregated road closure probability scores (after normalization). accessibility analysis is performed based on realistic road net- estimated by means of Monte Carlo simulations and used in the work conditions. accessibility analysis. In particular, during the runs, the physical damages are sampled for all network edges (e.g. bridges, over- 3.3. Accessibility modeling and visualization phase passes, road pavements) based on the given intensity measure and the associated fragility curves. If for example an overpass edge is in This phase includes the modeling of emergency related service the collapse state, then the state of the underlying edge is changed accessibility by using the GIS based techniques (see Section 2)in to blocked condition. normal conditions and in earthquake conditions considering road closure probabilities. 4.1. Data collection phase
In summary, the basic data used in the application are the 4. Case study following;
The proposed methodology is demonstrated through an ap- Digital transportation network data of Thessaloniki. In this plication to the city of Thessaloniki in Greece to analyze the ac- study, the main network of the urban area which is critical in cessibility to health services and shelter sites in case of earth- case of emergency is included, along with the ring road and the quake. The accessibility is estimated with and without considering main exits of the city where the majority of bridges and over- the road closures due to earthquake in order to highlight the effect passes are located (Fig. 5). Bridges are classified based on the of damage distribution by using the zone-based and isochronal number of spans, deck continuity, deck to pier connection, pier based accessibility modeling techniques. type and year of construction. Sixty bridges are included in the The study area is characterized by intense seismic activity with case study; all of them are reinforced concrete structures and 44 strong historical earthquakes of magnitudes larger than 6.0. The have more than one span. Roads are classified as freeways, main most recent destructive earthquake occurred in the broader area and secondary arterials, primary and secondary collectives. of Thessaloniki on 20 June 1978 (Mw¼6.5). The main shock caused Local roads, which are expected to have higher blockage risk extensive damage and loss of life in the metropolitan area and the due to closer distance from buildings, are not considered in the surrounding villages. For the seismic hazard input of the present analysis. The studied network configuration includes 594 nodes application, five seismic zones (Fig. 4) with Mmin ¼5.5 and Mmax and 674 edges, with a total length of about 235 km; 495 edges ¼7.5 are selected based on the results of SHARE European research are two-ways roads, whereas 179 are one-way roads. project [27]. The ground motion prediction equation introduced by Supply locations, which represent the health services and the Akkar and Bommer [1] is applied for the estimation of the ground shelter sites (as in year 2012). The health services are classified motion parameters on rock. The spatial variability of ground mo- based on a gravity coefficient that describes the importance of tion intensity measures, the effect of soil conditions and the li- the services based on different criteria such as capacity, size, quefaction susceptibility are accounted in the analysis (for more type of clinics and role (e.g. region, city). Higher coefficients details see [6,5,48]). The road closures are attributed to ground represent higher ranking of the service. The potential shelter failure and damage to bridges, damage to overpasses and collapses sites are basically the open spaces in the city with area greater of buildings adjacent to road edges. The closure probabilities are than 1 hectare (Fig. 6). 58 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 12. Health service accessibility as a travel time measure in isochrone-based technique in earthquake (A) and normal (B) conditions. *The urban regions which have health service accessibility over 1200 s (1200þ) are shown in transparent white color in order to improve the quality of visualization. K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 59
Fig. 13. The influence of the earthquake conditions on the health service accessibility.
Demand locations which represent the administrative districts services, accessibility is calculated for 5, 10, 15 and 20 min time. In of Thessaloniki as defined by Eurostat through the European zone-based approach, time based and gravity based traveling costs Urban Audit (EUA) approach (Fig. 7). are calculated between supply (health and shelter sites) and de- mand (centroid of the administrative districts) locations. The re- 4.2. Traveling cost calculation phase: computation of road closure sults are useful for the decision makers to demonstrate which probabilities in case of earthquake parts of the city have accessibility to health and shelter services in earthquake case. The road closure probabilities due to ground failure, bridges’ and overpasses’ damage and collapses of buildings adjacent to 4.3.1. Health service accessibility road edges are estimated as the average over 10,000 earthquake In isochrone-based technique, accessibility is evaluated in simulations (Figs. 8–11) based on the use of fragility curves. In case terms of isochronal boundaries, by connecting equal travel time of road pavements, the fragility curves provided in HAZUS [22] are points away from each of the health services considering the cal- used to estimate closure due to permanent ground deformation culated traveling costs (Fig. 12). In this case, the scores are related induced by soil liquefaction. Damages to bridges and overpasses to travel time, therefore the higher scores indicate lower health are estimated based on fragility curves that have been developed service accessibility. by Tsionis and Fardis [56] as a function of peak ground acceleration According to the results, most of the urban districts have 5 min (PGA). It is noted that integral single span bridges are considered (300 s) and 10 min (600 s) health service accessibility scores under as not vulnerable in this application. The road blockages due to both normal and earthquake conditions. However, significant ac- collapsed buildings are estimated by the models described in [5] cessibility gaps are observed in earthquake conditions, which considering the vulnerability, typology and collapse mechanism of mean that possible road closures after the earthquake could sig- buildings. The calculation of traveling costs just after the earth- nificantly affect accessibility to hospitals, especially in the north- quake requires calibration of the average traveling speed on each ern parts and partly in the southern parts of the study area. A road segment according to road closure probabilities as it is de- comparative analysis is also performed in order to provide better scribed in Section 3.2. understanding of the influence of the earthquake conditions on the health service accessibility results (Fig. 13). 4.3. Accessibility modeling and visualization phase The isochrone-based accessibility modeling is quite robust in understanding the accessibility level of urban services. However, it Hospital and shelter site accessibility is modeled by using iso- is relatively weak for understanding accessibility differences chrone and zone-based techniques both in normal conditions and among different administrative regions when compared with the in earthquake conditions considering road closure probabilities. In zone-based technique. In the latter case the cumulative time cost isochrone-based approach, starting from the health and shelter from each administrative zone to each health service is calculated 60 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 14. Health service accessibility as a travel time measure in zone-based technique in earthquake (A) and normal (B) conditions. by using Eq. (1). The accessibility scores are given in Fig. 14 for road closures in case of earthquake as they present lower health normal and earthquake conditions. The higher scores present service accessibility scores compared to the other districts. This higher health service accessibility in the zone-based accessibility indicates that the health service seekers in these zones may not modeling. efficiently reach these services and rearrangement and/or addition It can be observed that the administrative districts with ID of 1, of mobile health services are required to mitigate the impact of 12, 13, 14, 15 and 19 are significantly affected from the possible earthquake. In order to account for the importance of the services K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 61
Fig. 15. Health service accessibility as a gravity measure in zone based technique in earthquake (A) and normal (B) conditions. the analysis is also performed by using the gravity coefficients of services by using Eq. (2). The accessibility modeling scores are health services. Accessibility scores can be considered as more given in Fig. 15 with and without considering the road closures, realistic for the decision makers as they integrate some additional corresponding to road network under earthquake and normal factors like capacity, size, importance of the service etc. In this conditions respectively. case, the cumulative time cost from each administrative zone to It is observed that the administrative districts with ID of 1, 11, each health service is multiplied by the gravity coefficient of the 12, 13, 14, 15 and 19 are significantly affected from possible road 62 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 16. Shelter site accessibility as a travel time measure under earthquake (A) and normal (B) conditions.
closures in case of earthquake. Therefore citizens in these zones 4.3.2. Shelter sites’ accessibility may not efficiently reach health care sites, while appropriate mi- Shelter sites accessibility is again calculated by using both tigation measures should be taken to reduce the social impact in isochrone-based and zone-based techniques with and without case of earthquake. The zones with ID of 3 and 8 on the other considering the road closures, corresponding to road network hand, have still higher health service accessibility in case of under earthquake and normal conditions respectively. Due to lack earthquake. of gravity coefficient data for the shelter sites, the zone based K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 63
Fig. 17. The influence of the earthquake conditions on the shelter site accessibility. technique is performed as a travel time measure only. The tra- affected from the possible road closures in case of earthquake. The veling costs are again calculated from final aggregated road closure shelter service seekers in these zones may not efficiently reach probability scores due to ground failures, damage to bridges and these sites and rearrangement and/or addition of shelter sites are overpasses and collapses of buildings (see Fig. 11). For isochrone- required closer to these zones to mitigate the impact in case of based accessibility analysis, catchment/service area of 5, 10, 15 and earthquake. Although the districts with ID of 2, 5 and 17 are highly 20 min accessibility are calculated starting from each of the origin affected from the possible road closures in case of earthquake, they of the shelter sites (Fig. 16). In this case the scores indicate time, present relatively higher shelter accessibility compared to the other therefore the higher the scores, the lower the shelter site acces- districts. On the other hand, the district with ID of 3, has the highest sibility; for zone-based results, as accessibility scores are inversely shelter accessibility both in normal and earthquake conditions. proportional with the time, higher scores indicate higher shelter site accessibility. The results of the isochrone-based technique indicate that most 5. Conclusions of the urban districts in Thessaloniki have 5 min (300 s) and 10 min (600 s) accessibility scores to shelter sites. However, road The main objective of the present research is to integrate road closures in case of earthquake can significantly affect the shelter closure probability scores into accessibility modeling in earth- site accessibility especially in the northern parts of the study area. quake case. It is demonstrated that this approach can provide A comparative analysis is also performed in order to provide better advanced and realistic support to decision makers who deal with understanding of the influence of the earthquake conditions on accessibility, location/allocation, and service/catchment area of the shelter site accessibility results (Fig. 17). critical facilities before, during and after the earthquake, allowing As the isochrone-based accessibility modeling is relatively a better understanding of the accessibility level of different ser- weak for understanding accessibility differences among different vices under various seismic hazard scenarios. The results provide administrative regions, the zone-based technique is also applied useful input for civil protection authorities in order to assess the by using Eq. (1) to estimate the cumulative time cost from each vulnerability of urban districts, to test the sufficiency and appro- administrative zone to each shelter site. The accessibility scores priateness of the emergency service sites in case of an earthquake are given in Fig. 18 for normal and earthquake conditions;higher disaster, and therefore to define priorities and develop efficient scores mean higher shelter site accessibility. risk management plans. The results of the zone-based technique indicate that the ad- GIS based accessibility modeling considering road closure ministrative district with ID of 12 has the lowest shelter accessi- probabilities can be an important tool for disaster managers to bility both in normal and earthquake conditions. The districts with decrease the amount of affected population after a disaster as well ID of 4, 6, 7, 9, 10, 16 and 11, 15, 19 present relatively lower shelter as to help the determination of location and capacity of shelter accessibility compared to the other districts and they are highly sites, mobile hospitals, search-and-rescue (SAR) teams, equipment 64 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
Fig. 18. Shelter site accessibility as a travel time measure in zone-based technique under earthquake (A) and normal (B) conditions.
storage depots, temporal living depots etc. In this framework, it is administrative regions when compared with the zone-based found that the isochrone-based technique has the advantage technique. The latter provides easier comparison of accessibility of better representation of service/catchment area for a given scores between administrative units. Consequently, the two tech- time/distance threshold. However, it is relatively weak for under- niques are complementary, enabling an efficient and comprehen- standing the variation of accessibility among different sive risk management planning by the decision makers. K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66 65
6. Limitations and future extensions [8] L.R. Bixby, Spatial access to health care in Costa Rica and its equity: a GIS- based study, Soc. Sci. Med. (2004) 1271–1284. [9] M. Black, S. Ebener, M. Vidaurre, P.N. Aguilar, Z. El Morjani, Using GIS to The detail and complexity of the obtained/collected data in the Measure Physical Accessibility to Health Care,in: Health GIS Conference Pro- model can be arranged according to the data availability, specific ceedings, International ESRI User Conference, San Diego, 2004. objectives and detail needs of the study as well as the available [10] M.N. Boulos, A.V. Roudsari, E.R. Carson, Health Geomatics: an enabling suite of technologies in health and healthcare, J. Biomed. Info. 34 (3) (2001) 195–219, resources. Although the proposed methodology is applied to Published: 22 September 2006. health and shelter service accessibility, it can be extended to other [11] L. Brabyn, C. Skelly, Modeling population access to New Zealand public hos- kinds of urban related services under earthquake or other natural pitals, Int. J. Health Geogr. 1 (2002) 1–3, http://dx.doi.org/10.1186/ and man made disasters. In this specific application, the reduction 1476-072X-1-3, Published online 2002 November 12. [12] L. Brabyn, Modelling Population Access to New Zealand’s General Practi- of traveling costs just after the earthquake has been selected based tioners, International ESRI User Conference, San Diego, 2002. on expert judgment and can be modified accordingly if real data [13] P. Chapelet, B. Lefebvre, Contextualizing the Urban Healthcare System, Meth- become available. Although these judgments are subjective to odology for developing a geodatabase of Delhi's healthcare system, CSH Oc- casional Paper No. 11, Publication of the French Research Institutes in India, experts they provide a reasonable representation of network Rajdhani Art Press, New Delhi, 2005, 135p. conditions after the earthquake since they are based on road clo- [14] Q. Chen, Measuring Accessibility in GIS. 〈http://mather.ar.utexas.edu/students/ 〉 sure probabilities. Hence the accessibility analysis of health ser- cadlab/chen/measuring_accessibility_in_gis.htm (visited on 7.01.02). [15] M. Dodge, S. White, GIS and public services in Wales: analyzing supply and vices and shelter sites is performed based on realistic road net- demand for healthcare and education, Mapp. Aware J. 9 (7) (1995). work conditions. [16] S. Ebener, Z. El Morjani, N. Ray, M. Black, Physical accessibility to health care: The accessibility in earthquake case can also be modeled by from isotropy to anisotropy, Glob. Geospat. Mag. (2005), GIS@development. [17] M.P. Emelinda, K.S. Shashi, GIS-Based Allocation of Emergency Response Units considering each of the threads separately such as ground failure, Along a Transportation Route, in: ESRI International User Conference, San damage to bridges and overpasses, collapses of buildings etc. Diego, 1995. However, as the performance of the network is affected at the [18] K. Ertugay, H.S.B. Düzgün, GIS based stochastic modeling of physical accessi- bility by using GPS Based floating car data and Monte Carlo Simulations, Int. J. same time from different types of damages, the aggregated Geogr. Info. Sci. 25 (9) (2011). (overlayed) road closure probability scores could provide a more [19] K. Ertugay, H.S.B. Düzgün, GIS Based Stochastic Modeling of Physical Acces- complete and accurate view of the accessibility in earthquake case. sibility by Using GPS Based Floating Car Data and Monte Carlo Simulations (PhD Thesis), Geodetic and Geographic Information Technologies Department, Hence, in aggregation process (or overlay process) when a road Turkey, 2011, Submitted to the Graduate School Of Natural And Applied Sci- segment is exposed to several threats, the “highest closure” prob- ences of Middle East Technical University, September 2011. ability score is accepted as the final aggregated closure probability [20] K. Ertugay, S.H. Duzgun, Understanding accessibility: accessibility modeling with geographical information systems (GIS), in: E. Ocalir-Akunal (Ed.), Using score of the road segment. This is assumed as an aggregation for Decision Support Systems for Transportation Planning Efficiency, Engineering overlapped scores rather than a simplification. The “average” or the Science Reference, Hershey, PA, 2016, “sum” of the probability scores could also be used, however the pp. 223–257, http://dx.doi.org/10.4018/978-1-4666-8648-9.ch009. “ ” [21] ESRI, ArcGIS User Manual, Environmental Systems Research Institute, USA, average option could decrease the probability scores and there- 2011. fore mislead the results of accessibility analysis, while the “sum” [22] FEMA, HAZUS-MH Technical Manual, Federal Emergency Management option could be a reasonable alternative in future research. Agency, Washington DC, USA, 2003. In future studies, the GPS based floating car data and probe [23] J. Fortney, K. Rost, J. Warren, Comparing alternative methods of measuring geographic access to health services, J. Health Serv. Outcomes Res. Methodol. 1 vehicle traveling speed data can also be used to decide the re- (2) (2000) 173–184, Springer, June. duction percentage for traveling costs just after the earthquake [24] P. Franchin, A. Lupoi, P.E. Pinto, On the role of road networks in reducing – (see [18,19] for more information). As it is beyond the scope of the human losses after earthquakes, J. Earthq. Eng. 10 (2) (2006) 195 206. [25] P. Franchin, F. Cavalieri, Seismic vulnerability analysis of a complex inter- present research, possible other coefficients such as population of connected civil infrastructure, in: S. Tesfamariam, K. Goda (Eds.), Handbook of the districts, population of the victims etc. are not taken into Seismic Risk Analysis and Management of Civil Infrastructure Systems, consideration. Such improvements can also be taken into account Woodhead Publishing Limited, Cambridge, UK, 2013. [26] C. Ghio, L. Mark, N. Abdulkadir, A. Ahmed, Health GIS Tools and applications depending on the aim and specific requirements of the study and informing decisions in Yemen, Geospatial Application Papers, 2007. 〈http:// the available resources. www.gisdevelopment.net/application/health/overview/me05_012a.htm〉 (last visited on 01.01.09). [27] D. Giardini, et al., Seismic Hazard Harmonization in Europe (SHARE), Online Data Resour. (2013), http://dx.doi.org/10.12686/SED-00000001-SHARE. References [28] J. Ghosh, J.E. Padgett, Aging considerations in the development of time- dependent seismic fragility curves, J. Struct. Eng. 136 (12) (2010) 1497–1511. [1] S. Akkar, J.J. Bommer, Empirical equations for the prediction of PGA, PGV and [29] A. Goretti, A probabilistic model able to handle interaction among road net- spectral accelerations in Europe, the Mediterranean region and the Middle work,buildings and emergency services in urban areas. in: Proceedings of the East, Seism. Res. Lett. 812 (2010) 195–206. IX International Conference On Structural Safety And Reliability, ICOSSAR’05, [2] A. Alipour, B. Shafei, M. Shinozuka, Performance evaluation of deteriorating Rome, 19–23 June, 2005. highway bridges in high seismic areas, J. Bridge Eng. (2011) 597–611, http://dx. [30] K.G. Goulias, An Optimal Resource Allocation Tool for Urban Development doi.org/10.1061/(ASCE)BE.1943-5592.0000197. Using GIS-based Accessibility Measures and Stochastic Frontier Analysis, [3] A. Alipour, B. Shafei, Seismic resilience of transportation networks with de- University of California, United States, 2007, Santa Barbara California PATH teriorating components, J. Struct. Eng. (2016) C4015015, http://dx.doi.org/ Research Report, UCB-ITS-PRR-2007-7. 10.1061/(ASCE)ST.1943-541X.0001399. [31] D. Halden, D. Mcguigan, A. Nisbet, A. Mckinnon, Guidance On Accessibility [4] S. Argyroudis, A.M. Kaynia, Fragility functions of highway and railway infra- Measuring Techniques and Their Application, Scottish Executive Central Re- structure, in: K. Pitilakis, H. Crowley, A. Kaynia (Eds.), SYNER-G: Typology search Unit,, Scotland, 2000, ISBN: 1842680013 (visited on 11.11.01) http:// Definition and Fragility Functions for Physical Elements at Seismic Risk, www.scotland.gov.uk/cru/kd01/blue/guidance.pdf. Springer, The Netherlands, 2014. [32] C. Hélène, C. Evelyne, Poor prenatal care in an urban area: a geographic [5] S. Argyroudis, J. Selva, P. Gehl, K. Pitilakis, Systemic seismic risk assessment of analysis, J. Health Place 15 (2008) 412–419. road networks considering interactions with the built environment, Comput.- [33] P.M. Joseph, M.S. Ashton, E.G. Richard, V. Pariwate, J.F. Mark, Evaluating Aided Civil Infrastruct. Eng. 30 (7) (2015) 524–540. Michigan’s community hospital access: spatial methods for decision [6] S. Argyroudis, J. Selva, K. Kakderi, K. Pitilakis, Application to the city of Thes- support, Int. J. Health Geogr. 5 (2006) 42, http://dx.doi.org/10.1186/ saloniki, in: K. Pitilakis, P. Franchin, B. Khazai, H. Wenzel (Eds.), SYNER-G: 1476-072X-5-42. Systemic Seismic Vulnerability and Risk Assessment of Complex Urban, Utility, [34] R.P. Juliao, Measuring Accessibility Using GIS, Geo-computation-99, 1999. Lifeline Systems and Critical Facilities. Methodology and Applications. Geo- 〈http://www.geovista.psu.edu/sites/geocomp99/Gc99/010/gc010.ppt〉 (visited technical, Geological and Earthquake Engineering, Springer, Netherlands, on 17.12.01). 2014, p. 31. [35] A. Kiremidjian, J. Moore, Y.Y. Fan, O. Yazlali, N. Basoz, M. Williams, Seismic risk [7] F. Biondini, E. Camnasio, A. Titi, Seismic resilience of concrete structures under assessment of transportation networks, J. Earthq. Eng. 11 (2007) 371–382. corrosion, Earthq. Eng. Struct. Dyn. 44 (2015) 2445–2466, http://dx.doi.org/ [36] Y. Kuwata, S. Takada, Effective emergency transportation for saving human 10.1002/eqe.2591. lives, Nat. Hazards 33 (2004) 23–46. 66 K. Ertugay et al. / International Journal of Disaster Risk Reduction 17 (2016) 49–66
[37] M.P. Kwan, Space-time and integral measures of individual accessibility: a [49] K. Pitilakis, P. Franchin, B. Khazai, H. Wenzel (eds.), SYNER-G: Systemic seismic comparative analysis using a point-based framework, Geogr. Anal. 30 (1998) vulnerability and risk assessment of complex urban, utility, lifeline systems 191–216. and critical facilities. Methodology and applications, Springer ScienceþBusi- [38] S. Lotfi, M.J. Koohsari, Measuring objective accessibility to neighborhood fa- ness Media, Dordrecht, 2014. (under production). cilities in the city (A case study: Zone 6 in Tehran, Iran), Cities 26 (2009) [50] J.R. Ritsema van Eck, T. de Jong, Accessibility analysis and spatial competition 133–140. efects in the context of GIS-supported service location planning, Comput. [39] W. Luo, Using a GIS-based floating catchment method to assess areas with Environ. Urban Syst. 23 (1999) 75–89. shortage of physicians, Health Place 10 (2004) 1–11. [51] J. Scott, A. Larson, F. Jefferies, B. Veenendaal, Small-area estimates of general [40] W. Luo, F. Wang, Measures of spatial accessibility to health care in a GIS en- practice workforce shortage in rural and remote Western Australia, Aust. J. vironment: synthesis and a case study in the Chicago region, Environ. Plan. B: Rural Health 14 (5) (2006) 209–213. Plan. Des. 30 (2003) 865–884. [52] E.C. Stergiou, A.S. Kiremidjian, Risk assessment of transportation systems with [41] M.C. Makri, C. Folkesson, Accessibility Measures for Analyses of Land-Use and network functionality losses, Struct. Infrastruct. Eng. 6 (1–2) (2010) 111–125. Travelling with Geographical Information Systems, Aalborg University pub- [53] D. Sylvie, Gis-Based Accessibility Analysis for Network Optimal Location lications, 1999. Model, Article 407, Cybergeo, Systems, Modelisation Geostatistiques, 2007. [42] M.B. Makrí, Accessibility indices and planning theory, in: Proceedings of the [54] SYNER-G, Systemic seismic vulnerability and risk analysis for buildings, life- Eighth International Conference on Urban Transport and the Environment for line networks and infrastructures safety gain, European Collaborative research the 21st Century, Urban Transport VIII, 2002, pp. 37–46. project (FP7/2007-2013), 2013). 〈http://www.syner-g.eu〉. [43] R. Matthew, J. McGrail, S. Humphreys, Measuring spatial accessibility to pri- [55] Transportation Statistics Annual Report, Mobility, Access and Transportation, mary care in rural areas: improving the effectiveness of the two-step floating 1997. 〈http://www.bts.gov/transtu/tsar/tsar97/chap06.pdf〉 (last visited on catchment area method, Appl. Geogr. 29 (2009) 533–541. 23.12.06). [44] A.A. Murad, Creating a GIS application for local health care planning in Saudi [56] G. Tsionis, M. Fardis, Fragility functions of road and railway bridges, in: Arabia, Int. J. Environ. Health Res. 14 (3) (2004) 185–199. K. Pitilakis, H. Crowley, A. Kaynia (Eds.), SYNER-G: Typology Definition and [45] S. Nadine, S.F. Robert, C.W.G. Stefan, G. Darrin, Defining rational hospital Fragility Functions for Physical Elements at Seismic Risk, Springer, The catchments for non-urban areas based on travel-time, Int. J. Health Geogr. 5 Netherlands, 2014. (2006) 43, http://dx.doi.org/10.1186/1476-072X-5-43. [57] R. Umitsu, M. Fushimi, Shortest Path Problems for Ambulances in Case of [46] M. Ní Choine, A. O’Connor, J. Padgett, Comparison between the Seismic Per- Severe Earthquakes, in: The Sixth International Symposium on Operations formance of Integral and Jointed Concrete Bridges, J. Earthq. Eng. (2015) Research and Its Applications (ISORA’06), Xinjiang, China, August 8–12, 2006, 172–191, http://dx.doi.org/10.1080/13632469.2014.946163. pp. 283–291. [47] D. O’Sullivan, M. Alastair, S. John, Using desktop GIS for the investigation of [58] M.H. Vahidnia, A.A. Alesheikh, A. Alimohammadi, Hospital site selection using accessibility by public transport: an isochrone approach, Int. J. Geogr. Info. Sci. fuzzy AHP and its derivatives, J. Environ. Manag. 90 (2009) 3048–3056. 14 (1) (2000) 85–104. [59] D.H. Yang, G. Robert, M. Ross, Comparing GIS-based methods of measuring [48] K. Pitilakis, H. Crowley, A.M. Kaynia (eds.), SYNER-G: Typology definition and spatial accessibility to health services, J. Med. Syst. 30 (1) (2006) 23–32. fragility functions for physical elements at seismic risk, Geotechnical, Geolo- [60] M.A. Zanini, C. Pellegrino, R. Morbin, C. Modena, Seismic vulnerability of gical and Earthquake Engineering, 27, Springer, Netherlands, 2014a. http://dx. bridges in transport networks subjected to environmental deterioration, Bull. doi.org/10.1016/10.1007/978-94-007-7872-6. Earthq. Eng. 11 (2) (2013) 561–579.