Design site characterization via stochastic transfer functions functions M. Beneldjouzi Civil Engineering Faculty, Houari Boumediene University of Sciences and technology, BP 32 El Alia Bab Ezzouar, Algiers, Alegria. N. Laouami & A. Slimani CGS, National Center of Earthquake Applied Research, Rue Kaddour Rahim, BP 252, Hussein Dey, Algiers, Algeria SUMMARY: This paper provides an alternative and efficient means for characterisation of design sites classified in the Algerian Resistant Seismic Rules (RPA 99, 2003 version) by average transfer functions, through a stochastic approach, combined with a statistical study. For each soil type, the deterministic calculation of average transfer function is performed over a wide sample of 1-D soil profiles, for which, the average shear wave velocity, Vs, in soil layers is simulated using the random field theory. Thus, for any layer of each design site, Vs is defined between two bounds, so, to describe its variability neither the normal or lognormal distributions are appropriate. For that purpose, a probabilistic model that allows transforming the normal unbounded distribution to a bounded one has been used considering equivalent linear analysis method to approximate the nonlinear soil deposits response. These information help to decide on the aspect of structure type to build. Keywords: Random fields; transfer functions; soil classification; RPA99; statistics 1. INTRODUCTION It is well known that local site conditions have a strong influence on shaking behaviour of the site and, therefore, plays a major role in the damage potential of earthquakes. Hence, the local site analysis is a fundamental component of earthquake engineering when related to the geotechnical aspect of the problem. It makes possible the prediction of surface ground motion in terms of amplitude, frequency content and duration. Seismologists and earthquake engineers have put substantial efforts in the last several decades to understand and estimate, more accurately, the site effects on ground motion characterization, and take in account the associated effects in structure design. Overall, estimating site effects, namely potential amplification, pass mainly by classifying them with respect to a previously defined soil site classes, adopted in many earthquake resisting rules, which have tendency to categorize sites into different classes wherein the commonly used parameters for site classification are the average shear wave velocity over 30 meters of the subsurface geological materials, V s30 , and the dominant period, habitually associated again with normalized elastic response spectra. Furthermore, some others information such as surface geological and geotechnical description are also introduced. On the other hand, for including influences of site conditions on the earthquake resistant design of structures in the seismic codes and design guidelines, number of site classification techniques was developed by several researchers. Among those, the most elementary is the borehole data assessing and geologic maps and/or geomorphology data interpretation. However, with the increase in the number of strong ground motion stations as well as extension of the use of more simple and less onerous techniques around the world, many efficient other methods were successfully used. Almost of these methods are based on site predominant period determination’s, in order to assigning site class by reference to site class provisions, without providing reliable amplification levels, needed in structural dynamics analysis. In this study, we propose an alternative tool for characterizing design sites classified in the RPA99 by average transfer functions, through stochastic approach, combined with a statistical study. For many reasons, the transfer function is a useful means, unlike to response spectra, since it allows characterizing directly and fully the soil profile with previously known local