Amplitude-encoding FWI using different bases
He Liu, Daniel O. Trad, Kristopher A. Innanen
A super-shot or blended data strategy has been used in marine and land seismic surveys to reduce acquisition costs by reducing the time spent on the field. Full waveform inversion (FWI) has been used to estimate high-resolution subsurface velocity models. However, it suffers from expensive computational costs for matching the synthetic and the observed data. To reduce the costs of both data acquisition and processing, FWI using blended data has been recognized as very promising in future oil exploration. In this work, we use an amplitude-encoding strategy with different bases to accelerate the FWI process and compare their performance. The synthetic examples show that amplitude-encoding FWI using different bases as encoding functions can mitigate the crosstalk noise very well, providing good estimations of velocity models and convergence rate for both acoustic and elastic media. To further improve the calculation efficiency, we also adopt the dynamic encoding concept and reduce the number of super-shots every a few iterations. Since the encoding functions are not changed during the iterations, we can directly simulate the super-shots without the blending stage. From the updated velocity model comparison, we can see that the inversion results by dynamic encoding are almost identical to those by static encoding with further reduced calculation effort.