MOMENT TENSOR INVERSION OF POSSIBLY MULTIPLE EVENTS AT REGIONAL DISTANCES Petra Adamová 1, Jiří Zahradník 1, George Stavrakakis 2 1 Charles University in Prague 2 National Observatory of Athens, Institute of Geodynamics Abstract Moment tensor inversion of possibly multiple events is studied. All experiments are performed with new code ISOLA, based on time-domain iterative deconvolution of multiple point-source subevents. It is extension of Kikuchi and Kanamori (1991) for complete waveform inversion of regional data (Zahradník et al., 2005). The inversion is tested and applied to two selected M5 events from western Greece: near the town Amfilochia (December 31, 2002) and near the town Vartholomio (December 2, 2002). High non-shear CLVD component was reported by major agencies for them. Both these earthquakes have been interpreted as double events. Methodical lessons emerging from the applications are formulated. Acknowledgements This work was supported by projects (GOCE)-3HAZ-CORINTH and GAUK 279/2006/B-GEO/MFF. We thank E. Sokos for ISOLA Matlab GUI. References Kikuchi, M. and H. Kanamori (1991). Inversion of complex body waves-III, Bull. Seis. Soc. Am. 81, 6, Zahradník et al. (2005). Iterative Deconvolution of Regional Waveforms and a Double-Event Interpretation of the 2003 Lefkada Earthquake, Greece, Bull. Seis. Soc. Am. 95, 1, Adamová, P. (2006). Earthquake source process and its complexity, Master thesis. Charles University in Prague. Steps of inversion 1) Point source: first approximation of the centroid depth (fig. (a)) 2) Point source: optimization of the horizontal position of the centroid C in the depth found in step 1 (fig. (b)) 3) Multiple point source: plane passing through the centroid C of the step 2, using strike and dip found in step 2 (fig. (c)) Method – ISOLA Iterative deconvolution of multiple point source based on Kikuchi and Kanamori method (1991) – our modification for regional and local data, ISOLA Fortran code (Zahradnik et al., 2005) Least-square inversion of waveforms to retrieve moment tensor Optimization of waveform fit by grid search over trial source positions in space and time Application – Amfilochia earthquake, December 31, 2002; M w = 4.5 step 1: point solution (f = Hz) Optimum depth: 17km step 2: centroid in horizontal plane (f = Hz): 3 km to the West and 3 km to the South from the SED epicenter step 3: multiple point source (f = 0.04 – 0.3 Hz): 2 subevents Application – Vartholomio earthquake, December 2, 2002; M w =5.5 step 1: point solution (f = 0.04 – 0.08 Hz) Optimum depth: 17 km step 2: centroid in horizontal plane: 3 km to the East and 3 km to the North from the NOA epicenter step 3: split into subevents not successful Discussion Delicate instrument. Needs a lot of care in physical justification of subevents. Source multiplicity may explain apparently high non-DC component. Joint interpretation of the centroid moment tensor and hypocenter position may help to identify the fault plane. Data stations from the network of NOA (National Observatory of Athens, IG): Lennartz Le-3D/20 sec 1 station (SER) jointly operated by the Charles University Prague and University of Patras: Guralp CMG-3T/100 sec red stars: epicenters of Amfilochia and Vartholomio earthquakes black triangles: stations used for Amfilochia and/or Vartholomio earthquake Correlation and focal mechanism as a function of the depth. displacement blue: observed red: synthetic without centroid shift black: synthetic with centroid shift blue: observed displacement black: synthetic displacement Fixed mechanism Free mechanism Multiple point solution, red points: surface projection of the trial source positions. (Second nodal plane appears less likely candidate for the fault plane.) displacement blue: observed red: synthetic with centroid shift black: synthetic without centroid shift 17 km