Maren Böse - Geophysical Institute, University of Karlsruhe, Hertzstr. ۱۶, ۷۶۱۸۷ Karlsruhe, Germany
Friedemann Wenzel - Geophysical Institute, University of Karlsruhe, Hertzstr. ۱۶, ۷۶۱۸۷ Karlsruhe, Germany
Mustafa Erdik - Department of Earthquake Engineering, Kandilli Observatory and Earthquake Research Institute, Bogazici University, ۸۱۲۲۰ Cengelkoy, Istanbul, Turkey
Gheorge Marmureanu - National Institute for Earth Physics, POB MG-۲, ۷۶۹۰۰ Bucharest-Magurele, Romania

Kandilli-Observatory of the Bogazici University in Istanbul operates within the Istanbul Earthquake Rapid Response and Early Warning System (IERREWS) ten strong motion sensors with real-time communication link to two datacenters [Erdik et al., 2003a]. Defined by the time window between Pwave detection at the first sensor of IERREWS and the arrival of high-amplitude S-waves average prewarning times for Istanbul are expected to range between 8 s and 15 s. Here we discuss an improved version that utilizes the Neural Net technology and does not rely on on-site early warning only. In contrast to this, on-site warning is used in Romania, where the accumulation of strong earthquakes with resembling source mechanisms in the Romanian Vrancea zone, SE Carpathians, allows for designing a simple, cheep and robust earthquake early warning system for Bucharest. Average epicentral distances of 130 km towards the Romanian capital provide warning times of about 25 s for all intermediate-depth events. A scaling relation has been developed by Wenzel et al. [1999] but recently improved and expanded to all relevant ground shaking parameters by using additional weak and strong motion data. Predictive laws suitable for early warning are determined for peak ground acceleration PGA, spectral response at different periods (PSA0.3s, PSA1.0s, PSA2.0s at 5% damping), and seismic intensity I. The observations made during the October 27, 2004 Vrancea earthquake (Mw = 6.0) are fully compatible with all scaling relations within reasonable limits. Both systems are discussed in terms of design philosophy, assumptions included and performance.