Surface-Data-Driven Multiple Elimination in the Common-Midpoint Domain and Its Application to Land Seismic Data

Abstract

Surface-data-driven multiple elimination methods directly use surface data to predict and eliminate multiples without requiring a detailed subsurface velocity model, but can be computationally expensive and not always straightforward on land datasets. We present a unified framework in the common-midpoint (CMP) domain for suppressing both surface-related and internal multiples. The CMP-domain implementation reduces the dimensionality of the prediction operations and therefore has the potential to lower computational cost relative to a full shot-domain implementation, and it can be applied directly to land data. Under the local low-relief structural approximation, the method operates within individual CMP gathers and reduces the spatial convolutions of multiple predictions to scalar multiplications in the f-k domain. A practical workflow is proposed, with two implementation details highlighted: a first-Fresnel-zone time-variant taper applied before prediction, and the minimum zero-padding required to avoid wraparound aliasing. Application to a field dataset from the Tarim Basin demonstrates that the workflow effectively suppresses both surface-related and internal multiples. The low-velocity multiple energy is attenuated, the primary energy cluster becomes relatively more concentrated, and the continuity and quality of the deep stacked reflections are improved. The method is well suited for land seismic data with low signal-to-noise ratio and irregular offset distribution.