**5. Discussion**

The Squirt model and the Pride model are not suitable for predicting the attenuation of rocks containing micro-pores at the seismic frequency band. The energy loss caused by fluid flow is neglectable, and gas dissolution and exsolution become the dominant attenuation mechanisms.

Although the conventional constant Q model is widely accepted, it will not predict accurate velocity dispersion when the corner frequency occurs in the seismic frequency band. For the same rock properties, the velocity range differs from that of the pore fluid dissolution model. The Backus averaging method almost results in a similar velocity range to that of real data. The size of the window chosen will not improve accuracy of velocity. Neither of these two methods are suitable for upscaling. The SLS model is suitable for upscaling the pore fluid dissolution model. The velocity dispersion and attenuation curve can be well modeled by the SLS model for different porosity and gas saturation cases.

Synthetic seismic data generated by using velocities calculated by these methods show obvious difference. Due to similar velocity range from the Backus averaging method, the events in the synthetic seismic data show little stretch. The constant Q model underestimates the velocity dispersion, and thus, the synthetic seismic data show little stretch until 0.4 s. The attenuation becomes observable because of the accumulation of the attenuation effect. The SLS model reasonably predicts the attenuation, and the synthetic seismic data show obvious difference from the traces generated by using logging velocity. When choosing a rock physical model for well-to-seismic calibration, one needs to consider the geology, geochemistry, and core data analysis report, and then select the appropriate rock physics model and update the parameters in the model.
