Probabilistic methods for the estimation of potential seismic damage: Application to reinforced concrete buildings in Turkey A. Askan a, * , M.S. Yucemen b a Department of Civil Engineering and Earthquake Engineering Research Center, Middle East Technical University, Ankara 06531, Turkey b Department of Civil Engineering and Earthquake Studies Program, Middle East Technical University, Ankara 06531, Turkey article info Article history: Received 14 February 2009 Received in revised form 11 September 2009 Accepted 6 April 2010 Keywords: Earthquake damage estimation Damage probability matrix Fragility curve Reliability theory Discriminant analysis Recent major earthquakes in Turkey abstract Three probabilistic methods, for the prediction of potential seismic damage to low and mid-rise rein- forced concrete buildings in Turkey, are presented. As the first method, ‘‘best estimate” damage probabil- ity matrices for each seismic zone are developed by combining expert opinion and the damage statistics compiled from the recent earthquakes occurred in Turkey. Second method involves a reliability-based model, which treats the earthquake force and seismic resistance as random variables. This model expresses potential seismic damage in the form of a damage ratio distribution, which is a function of modified Mercalli intensity or peak ground acceleration. As the third methodology, discriminant analysis technique is utilized to carry out a statistical analysis on the damage data compiled during recent earth- quakes that occurred in Turkey. Based on the classification procedure involved in this technique, the damage state probabilities corresponding to different modified Mercalli intensity levels are obtained. These three methods are applied on the building damage databases compiled in the aftermath of several recent earthquakes in Turkey: 1992 Erzincan, 1995 Dinar, and 1999 Duzce earthquakes. The probabilistic damage profile obtained according to these three methods is expressed in terms of damage ratios and the results are compared with each other. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Earthquakes are natural hazards with high damage potential. However, by utilizing several principles, it is possible to identify and minimize the resulting social and economic losses, which mainly involve loss of human lives, damage to major industrial facilities, civil engineering structures and lifelines. In general, earthquake loss estimation methodologies involve the following steps: identification of seismic hazard, vulnerability assessment and risk analysis. Major earthquakes occurred all over the world during the past decades increased the number of studies on seismic vulnerability assessment techniques. Majority of these studies focus on empirical and theoretical techniques for estima- tion of seismic risk for buildings by incorporating the uncertainties related to structural capacity under the assumption of perfectly spatially correlated seismicity (e.g. [1–9]). However, for realistic loss estimations, it is important to take into account the uncer- tainty related to ground motion variability. Recently several stud- ies have proposed loss models that include spatial variations of seismicity parameters (e.g. [10–12]). Similarly, in Turkey, where 91.4% of the total area is located in seismically active regions and 95.1% of its population being under earthquake threat [13], many recent methodologies on seismic hazard analysis and damage prediction are developed (e.g. [13– 18]). At this stage, in order to estimate and reduce possible future losses, immediate assessment of the seismic resistance of existing structures is quite important. A large number of the studies on damage prediction are deterministic. However, due to aleatory and epistemic uncertainties involved both in seismic demand and seismic capacity, damage prediction procedures should be car- ried out based on statistical and probabilistic concepts [13]. The objective of this study is to present three different probabi- listic approaches in order to estimate the possible future damage to buildings due to earthquakes. These methods are damage probabil- ity matrices, reliability-based damage rate model and discriminant analysis. All of the three methods are developed by utilizing statis- tical and probabilistic techniques, which can be applied on any building group in a certain region, in order to obtain potential dam- age profile. Both the reliability-based model and the implementa- tion of the discriminant analysis technique are novel approaches for the vulnerability assessment of existing reinforced concrete buildings in Turkey. We note that in all three methods imple- mented in this study, hazard is assumed to be perfectly spatially correlated. 0167-4730/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.strusafe.2010.04.001 * Corresponding author. Tel.: +90 5065416154; fax: +90 3122105401. E-mail address: aaskan@metu.edu.tr (A. Askan). Structural Safety 32 (2010) 262–271 Contents lists available at ScienceDirect Structural Safety journal homepage: www.elsevier.com/locate/strusafe