Reverse time migration with frequency-dependent Q compensation accelerated with GPU computing

It is well known that the earth strata are far from perfectly elastic but demonstrate the properties of viscoelasticity, in which seismic waves suffer wavelet shape distortion and energy loss during propagation. Seismic attenuation is commonly characterized by the quality factor Q. The linear model of wave attenuation with frequency independent Q is widely used in exploration seismology (Kjartansson, 1979). Based on this theory, different frequency components of seismic waves propagate in anelastic materials with a frequency-dependent phase velocity, while in elastic media all frequency components travel with the same phase velocity. In anelastic media the energy loss is approximately proportional to the frequency. In general, higher frequency components in seismic waves tend to travel faster than low frequency components, and their amplitudes decay more quickly. Seismic imaging without proper accounting of dispersion and amplitude loss produces images with distorted phases, dimmed amplitudes and reduced resolutions, especially for deeper horizons under low quality factor (low Q) strata, making it more difficult to do AVO interpretation analysis and to tie seismic horizons to well data.