Study of the seismicity of the Lacq gas field and analyze of seismic waveform by chirplet decomposition

The Lacq gas field is exploited since 1957 and seismicity began in 1969. It is a moderate seismicity (M · 4.4) induced by fluid extraction. After the cleaning of the available seismological database (1989-1997), the use of an Automatic Picker with parameters optimised by Simulated Annealing has permitted the update of P wave picks. Additioned to picks of events recorded between 1974 and 1988, we localised more than 1600 events using a 3D-velocity model, most of them within or above the field. The comparison of hypocenter locations and focal mechanisms with well and 3D-seismic data has shown that earthquakes were associated to relatively important faults well-oriented for stress perturbation accommodation. Whereas the lateral extension of the field is oriented WNW-ESE, the distribution of events is oriented SSW-NNE, occuring on WNW-ESE faults. The WNW and ESE parts of the field are strained aseismicaly. By comparing the seismicity distribution to the field exploitation history and the field modelisation, three main results have been highlighted : (i) a water over-pressure triggered swarm, (ii) an important connexion between the central zone of the field and the gas reserves to the North of the field, and finally (iii) an update of the interpretation of the field physical model. <br />The second aim of this study was the analyse of waveforms with atomic chirplet decomposition, which is a generalisation of the wavelet transform and the Gabor transform. We first genalised the chirplet construction to 7D (7 construction attributes). Then, we developed an algorithm of wave separation. Finally, we can decompose a seismogramm in a few number of atoms. Each of these atoms is a 9-tuplet, characterised by morphological attributes : (i) a time arrival, (ii) a central frequency, (iii) an energy, (iv) a phase, (v) a duration, (vi) a frequency modulation rate, (vii) a frequency modulation type, (viii) an asymmetric shape and (ix) a flat shape. With this wave characterisation algorithm, we can classify events. The classification and the associated dendrogramm representation have allowed to differentiate events triggered by water injection from others, revealing a shorter and more impulsive source for triggered events. By using only morphological attributes of atoms and muting time and frequency arrival components, we also shown that we can compare events with different locations.