Drilling through over presured zones is most critical for any exploration campaign. There are large number of accidents that have happened in exploration history which caused huge damage to human life and environment in addition to capital and reputation loss to E&P companies. Famous BP Macando Blow Out is the recent one that we all know.
Therefore, to drill safely it is of paramount importance to know about pore pressure profile of the well including geomechanical model and what is more important is to know these properties ahead of the bit while drilling.
The only tool available to E&P community to look ahead of the bit for these properties is VSP only. And that is the importance of VSP data which at many a times is ignored by E&P team in order to cut costs.
In my own experience I have seen many a critical wells with significant over pressured zoned planned to be drilled without any requirements of VSP data acquisition plans. However, when the bit reaches to critical depth - the entire E&P team starts panicking and in a hush calls for intermediate VSP acquisitions resulting in infereior project execution plus paying more too. And many a times they just keep drilling leading to such accidents. Unfortunately, that is how it happens.
The objective of VSP inversion is to try to estimate full band acoustic impedance also called absolute acoustic impedance. However, the lowest frequency available in VSP data acquired not less than 5Hz due to limitations on source side. Therefore, to estimate absolute acoustic impedance a low frequency model of 0-5Hz bandwidth is required. The missing low frequency model is provided through acoustic impedance constraints.
Absolute acoustic impedance upto TD of the well is estimated from measured VSP/Sonic and density data and same is used as constraints during VSP inversion but below TD no such reliable data is available in most cases. Therefore, the inversion has reliable constraints only upto TD and below TD the inversion is unconstrained. Therefore, in absence of constraints below TD the recovered reflectivity below TD are not full band and so the estimated acoustic impedance/interval velocity.
The uncertainty in estimated acoustic impedance increases as depth (time) from TD increases. This is the reason the uncertainty in predicted acoustic impedance ahead of the VSP TD, which acts as the reference point, increases with depth. Also the corridor stack just ahead of the TD is expected to be free of multiples while same may not be true further deep. Due to these reasons, and based on experience, it is recommended that look ahead VSP inversion could be relied for next 200m with an uncertainty of approximate 10 percent.
The VSP data in a Well is inverted to derive the interval velocity profile ahead of the bit. The VSP data is inverted to derive acoustic impedance using e.g. Bayesian inversion. The algorithm derives reflectivity and then computes a synthetic with the wavelet derived from the data. This synthetic is then compared with the actual VSP corridor stack and the error is minimized in the least square sense. Once the residual between the two is minimum, inversion is supposed to be doing a good job. The velocity profile is then derived from the inverted acoustic impedance by using a density approximation that would be valid for the well.
A comparison between LWD Sonic Velocity, acquired post VSP inversion, and Inverted Velocity can be done below the total depth of VSP and any deviation can be used to correct inverted velocity profile further down. And if deviations increase more than 10% post corrections then it is better to acquire another VSP data and do the same again.