1 METHODS AND UNCERTAINTIES IN LIQUEFACTION HAZARD ASSESSMENT FOR NUCLEAR POWER PLANTS Erzsébet GYŐRI 1 , Tamás János KATONA 2 , Zoltán BÁN 3 and László TÓTH 4 ABSTRACT Experience shows that the nuclear power plant can be safely designed for vibratory effects of earthquakes. Contrary to this, the plants can be heavily damaged by effects of earthquake-induced phenomena like tsunami, soil liquefaction after surviving the ground shaking effects. In the paper, the problems in nuclear power plant’s safety analysis methodology are outlined for the case of soil liquefaction. Depending on the applied method, the liquefaction hazard assessment may provide very scattering results and might lead to controversial conclusions. In the paper different approaches and methods for liquefaction hazard assessment are considered. Practical problems of the selection of appropriate methods are demonstrated on the example of Paks Nuclear Power Plant, Hungary. For comparison, records of one SPT, one CPT and one shear wave velocity measurement, that were located very close to each other, were chosen from the site and the liquefaction potential was assessed by altogether nine of the newest and most commonly used empirical correlations. The results show big variance even by the same test-based methods. Epistemic uncertainties increase further when post liquefaction settlements are computed. To handle the uncertainties we propose a probabilistic multiple-scenario approach that is appropriate for safety analysis of the power plant on a given probability level. INTRODUCTION Proper understanding and assessment of safety of nuclear power plants (NPPs) with respect to external hazards became very important after 11th of March 2011, the accident of Fukushima NPP. Analysis of accident sequences initiated by extreme rare external events is required for demonstration of low annual probability of reactor core damage (≤10 -5 /a) or large early releases (≤10 -7 /a) and for the development of severe accident management methods and procedures. The analyses have to account events with extreme low annual probability of exceedance, up-to 10 -7 /a, while the probability of exceedance for the design base level is usually set for 10 -4 /a. There are well-developed deterministic and probabilistic methods for safety analyses of nuclear power plants (see ASME 2008; IAEA 2009, 2010). In case of external events the adequacy of safety analyses and conclusiveness of the results is mainly limited by the epistemic uncertainty of the hazard definition that is growing rapidly with decreasing probabilities considered. Safety analysis for external events becomes even more difficult if the secondary phenomena to the external events or combinations of external events have to be considered, e.g. earthquake and earthquake induced soil liquefaction or tsunami. The difficulties are 1 MTA CSFK GGI Seismological Observatory, Budapest, gyori@seismology.hu 2 MVM Nuclear Power Plant Paks LTd, Paks, bata01@t-online.hu 3 Budapest University of Technology and Economics, Budapest, ban.zoltan@epito.bme.hu 4 MTA CSFK GGI Seismological Observatory, Budapest, laszlo@seismology.hu