earthquake_back_projection

Back-projection of high-frequency radiation from earthquake source using multiple arrays. Methodology is based on Ishii (2012). For details see Kumar et al., 2017 [doi: 10.1093/gji/ggw438]

dependency

• Matlab
• GMT4
• SEIZMO [http://epsc.wustl.edu/~ggeuler/codes/m/seizmo/] SEIZMO is a toolbox for seismic data porcessing in Matlab and it is mainly used for taup toolkit to calculate travel times of seismic phases.

---

Method

• Algorithm as follows:

1. The source region, surrounding the hypocentre is parameterized as a horizontal plane with grids. The plane is fixed in space at the hypocentral depth.

2. The teleseismic P-wave, from each station within the array, is windowed 30 s before and 150 s after the theoretical arrival time computed using the IASP91 velocity model (Kennett &Engdahl1991).

3. Within a given array, the windowed P-waves are cross-correlated with a reference station (chosen at the centre of the array) to estimate the time shift required to account for the 3-dimensional variation in the velocity structure.

4. Each grid point on the parameterized hypocentral plane is assumed as a potential source and the travel time of P-wave is calculated, using IASP91 velocity model, to all the stations within the array.

5. Waveforms from these stations, adjusted by the respective time shifts (computed from cross-correlation), are stacked together,starting from the calculated P-wave arrival time for a given source grid. During stacking, the waveforms are weighted (i) inversely by the density of stations within an array and (ii) by the cross-correlation coefficient (obtained above).

6. The weighted stacked energy is then back projected onto the source grid. This procedure is repeated for all potential source grids,for every array.

7. Information from all the arrays are combined by following a weighting procedure, where the stack from the hypocentral grid is cross-correlated for each array, with respect to a chosen reference array.

8. For every potential source grid, stacked waveforms from each array are adjusted by the estimated time shift and summed by nor-malizing the amplitude.

9. For removing the high-frequency artefacts in the stacked waveform at each potential source grid, a 10 s moving averagetime window is applied. In order to test the influence of the moving average time window on the final result, we performed our analysisfor 5, 10, 15 and 20 s time windows. It was observed that the 10 swindow is optimal in removing unwanted high-frequency artefacts,while preserving the necessary details of the source time function(STF).