Stress drop and source scaling of the 2009 April L'Aquila earthquakes

The empirical Green's function (EGF) technique is applied in the frequency domain to 962 broad-band seismograms (3.3 <= M-W <= 6.1) to determine stress drop and source scaling of the 2009 April L'Aquila earthquakes. The station distance varies in the range 100-250 km from the source. Ground motions of several L'Aquila earthquakes are characterized by large azimuthal variations due to source directivity, even at low magnitudes. Thus, the individual-station stress-drop estimates are significantly biased when source directivity is not taken into account properly. To reduce the bias, we use single-station spectral ratios with pairs of earthquakes showing a similar degree of source directivity. The superiority of constant versus varying stress-drop models is assessed through minimization of misfit in a least-mean-square sense. For this analysis, seismograms of 26 earthquakes occurring within 10 km from the hypocentres of the three strongest shocks are used. We find that a source model where stress drop increases with the earthquake size has the minimum misfit: as compared to the best constant stress-drop model the improvement in the fit is of the order of 40 per cent. We also estimate the stress-drop scaling on a larger data set of 64 earthquakes, all of them having an independent estimate of seismic moment and consistent focal mechanism. An earthquake which shows no directivity is chosen as EGF event. This analysis confirms the former trend and yields individual-event stress drops very close to 10 MPa at magnitudes M-W > 4.5 that decrease to 1 MPa, on the average, at the smallest magnitudes. A varying stress-drop scaling of L'Aquila earthquakes is consistent with results from other studies using EGF techniques but contrasts with results of authors that used inversion techniques to separate source from site and propagation effects. We find that there is a systematic difference for small events between the results of the two methods, with lower and less scattered values of stress drop resulting from the EGF approach.