Two intensity-attenuation models for central Europe are presented in terms of mean relations and probability distributions of their variability. They are based on 14,714 intensity data points (IDP) of 31 earthquakes located in Germany, France, the Netherlands, and the Czech Republic. The attenuation relations are derived from crustal events with moment magnitudes Mw from 2.4 to 5.7, source depths h from 1 to 20 km, and epicentral distances R up to 400 km. Two different regression techniques are applied to derive the appropriate parameters a and b of the attenuation model . A combined weighted least-squares regression utilizing intensities and distances of the IDP in a direct way provides a=2.80 and b=0.0013. A chi-square regression method for mean intensity-class distances estimated by intensity-level binning gives a=2.95 and b=0.0025. The variabilities in the derived models can be described by normal distributions centered at the mean relation with a standard deviation of about 0.7 intensity units. This specification of attenuation variability is one of the advantages compared with former relations, making both attenuation models especially suitable to be applied in seismic hazard assessments. The two independent models can be regarded as an epistemic uncertainty in logic tree based hazard calculations.