In this dissertation, the analysis of the local site effects on the amplification of seismic ground motions in Croatia was performed using 1-D equivalent–linear (EQL) stochastic Random Vibration Theory (RVT) method. The main reason for choosing the relatively new 1-D EQL site response analysis with RVT-based method is the limited existing strong motion database in Croatia. In the RVT-based method, single theoretical point source Fourier Amplitude Spectrum (𝐹𝐴𝑆) defined by the local and regional seismological parameters is adequate to represent the input ground motion. First part of thesis covered calculation of the high-frequency attenuation parameter kappa (𝜅) and its local site-specific component (𝜅0) to describe high-frequency decay of 𝐹𝐴𝑆. The recordings from earthquakes with local magnitudes 3.0≤𝑀L≤ 5.7, focal depths less than 30 km and epicentral distances 𝑅e ≤ 150 km from ten seismological stations were used for the estimation of the 𝜅 using Anderson and Hough (1984) method. Local attenuation parameter 𝜅₀ was estimated using the linear 𝜅–𝑅e dependence by least-square regression for horizontal and vertical ground motion components (𝜅hor and 𝜅ver) for each station. The use of error-invariable regression could limit influence of uncertainty in 𝑅e on the final value of 𝜅0 and slope 𝜅R. Estimated 𝜅0 values for Croatian seismological stations are consistent with the global 𝜅0 for rock sites. Comparison between 𝜅hor and 𝜅ver models was performed to determine local (𝜅0) and regional attenuation (slope 𝜅R)contributions to the 𝜅. Observations from spatial distribution of the 𝜅 values indicate that beside isotropic local and regional geology and complex tectonic structure, other effects such as attenuation anisotropy from different causes (e.g., scattering due to heterogeneity, beamforming, fracturing, flow of fluids in rocks) possibly have effect on the 𝜅 distribution. Observed discrepancies between the frequency-dependent 𝑄(𝑓) and the frequency-independent 𝑄est(𝜅R) for the high-frequency range (10–25 Hz) are mostly within the respective confidence limits, and can be attributed mainly to different techniques to estimate 𝑄(𝑓) and 𝜅, and complexity and variability in the whole-path attenuation contributions to 𝜅. Second part of thesis is focused on 1-D EQL site response analysis using RVT-based method for different local site profiles around Croatia and for different input ground motion levels (peak ground acceleration at the bedrock—𝑃𝐺𝐴ROCK). Seismological parameters (magnitude, epicentral distance, focal depth, seismic attenuation, near-site attenuation) are varied to define the input rock motion based on the regional 𝐹𝐴𝑆. For lower levels (𝑃𝐺𝐴ROCK<0.1 g) , the input motion is significantly amplified at the top layers of the profile (𝐴𝐹@𝑃𝐺𝐴) and the amplification factor (𝐴𝐹) is most prominent at predominant peak period particularly for the softer soils with lower 𝑉S30 and thicker alluvium layers overlying bedrock. At higher levels (𝑃𝐺𝐴ROCK > 0.1 g), softer soils with lower values of 𝑉S10,𝑉S20, and 𝑉S30 (average shear wave velocity in top 10 m, 20 m and 30 m) shows non-linear behaviour, therefore, 𝐴𝐹@𝑃𝐺𝐴 decreases significantly below the 𝐴𝐹 = 1 line at shorter spectral periods, and predominant peak period is prolonged (increased) with decreasing 𝐴𝐹@𝑃𝑃. At different levels 𝑃𝐺𝐴ROCK, 𝐴𝐹 varies significantly with chosen spectral period for different site characteristics parameters (𝑉S10,𝑉S20,𝑉S30). Finally, the third part of thesis presents the empirical nonlinear site amplification model developed for Croatia for a range of local soil profiles (160 < 𝑉S30 < 1389 m/s) as a function of the local site parameter (𝑉S30) and intensity of input rock motion (𝑃𝐺𝐴ROCK). Proposed 𝐴𝐹 model for Croatia is in good agreement with Sandikkaya et al. (2013) and Kamai et al. (2014) empirical 𝐴𝐹 models. Observed differences between models may be related to different equivalent linear soil properties utilized in RVT site response methods, developed site amplifications based on empirical database or definition of the soil profiles (real/measured soil profiles in this study vs. generic randomized soil profiles in others). Proposed 𝐴𝐹 models are strongly nonlinear for soft sites and heavily dependent on the period compared to weakly period-dependent 𝐴𝐹 given in Eurocode 8. Significant amplifications were observed for stiffer 𝑉S30 sites compared to EC8–𝐴𝐹𝑠 for site classes B and A. Nonlinear site amplification based solely on single “questionable” site parameter 𝑉S30 can misled to wrong conclusions and needs to be further investigated.