We combine probabilistic and deterministic approaches in an attempt to tackle a challenging topic in engineering seismology, that of the prediction of strong ground motion in urban areas located close to active faults, with sparse modern seismicity. The case we study is the city of Xanthi in northern Greece and we generate bedrock synthetic time series to be used as input motions into site-specific site effect analysis. We re-assess the seismic hazard using the probabilistic method and then perform hazard deaggregation in order to define realistic earthquake scenarios, which we examine in a deterministic way. The scenarios examined are: (a) a reference earthquake of M6.7, return period 100 years and at *30 km distance from the city and (b) a reference earthquake of M5.8, return period 475 years and at *3 km distance from the city of Xanthi. To compute synthetic time series for the distant source, we apply the stochastic method for finite faults. For the near-fault earthquake scenario we adopt a hybrid deterministic-stochastic approach that includes the semi-analytical method implemented in the COMPSYN code for the low frequency (*\3 Hz) part of the spectrum and the stochastic method for the high-frequency (*[3 Hz). The results of the probabilistic and the deterministic approaches converge to a satisfactory degree.