Z-99 RAY-BASED STOCHASTIC INVERSION

A new seismic inversion method, Ray-Based Stochastic Inversion (RBSI) is presented, which is founded on the high-frequency Asymptotic Ray Theory (ART) description of seismic waves and rmly links Kirchhoff-Helmholtz pre-stack depth migration (PreSDM) with Stochastic Inversion (SI) for reservoir properties. Results from synthetic data tests show that: (1) RBSI, contrary to SI, can correctly determine lateral layer-density variations, a requirement for preserved-amplitude PreSDM, and (2) RBSI can delineate the correct reservoir properties in cases where SI produces erroneous results because of dip-dependent migration stretch. Ray-Based Stochastic Inversion Principle For RBSI, the subsurface space with surface boundary is parameterised as an over-burden macro-model overlying a layered target reservoir sequence (Fig. 1) and is thereby assumed to satisfy the standard ray-theoretical validity conditions [1] (elements of can be written as, , ) with and as = 0). Instead of the 1D convolutional model of SI, in RBSI 3D elastic ray-tracing is used. In principle, any wave type and acquisition conguration can be handled by RBSI. In this paper, we limit ourselves to surface-recorded single P-reections. The key vehicle for RBSI is then formed by a single pair of P-rays leaving a reection point ff within the reservoir sequence at angles fiffifl to the normal-vector "! #% $ of the reection surfac