Seismic ground motion is strongly affected by many phenomena depending on complex characteristics, some of them related to the source. Those phenomena must be taken into account in risk mitigation, for example through the evaluation of several simulations for a future fault scenario. However, each analysis has high computational cost in the case of physical-based numerical models, even with efficient numerical methods such as Spectral Elements Method (SEM). It makes difficult to evaluate more than dozens of alternatives with high refined mesh with SEM. In this work an alternative method is proposed, it takes advantage of the results from few SEM simulations, and by combining it with the Empirical Green Function method (EGF), we are able to evaluate thousands of alternatives that could affect a region. This combination of methods allows us also to recreate simulations at low and high frequencies as well with a reduce cost. This work focuses on the variability of some fault parameters, including stress drop and rupture geometry. Joining SEM and EGF methods, we analyzed the seismic hazard to the nuclear site of Cadarache, France. We use SEM method to simulate the ground motions for several point sources around the fault of Middle Durance (Southeastern zone). These results are combined and scaled by the EGF method, allowing to build with few SEM simulations the ground motions for thousands of simulation with different fault parameters. Those simulations are usually valid for low frequencies where the mesh is refined enough. At high frequencies, we employ the well-known method of EGF using previous records in the zone coming from the same fault. EGF method recreates several alternatives taking into account the variability of the source with empirical methods. The combination of both methods results in broad band accelerograms and it allows to consider several alternatives of faults in the seismic hazard analysis. Here, we are presenting our procedure and the precautions to be taken into account for combining SEM and EGF methods. The obtained results are showing the usefulness of considering the fault variability to better predict the seismic hazard in region.