A data‐driven approach for 3D VSP multiple prediction

Abstract

Migration images obtained with 3D VSP (vertical seismic profile) data can provide valuable high‐resolution information about the subsurface structures. One of the challenges in 3D VSP imaging is the blurring due to the interference from multiple reflections, similar to that in imaging with surface seismic data. For imaging with primaries acquired with VSP surveys the upgoing multiples will interfere with and generate artifacts on the migration images. For VSP imaging with multiples other multiples will be the interference. That is, when the imaging condition is tuned to a certain type of multiple, other types of multiples will be regarded as noise and will generate artifacts on the multiple migration image. This problem may be solved using a prediction and then subtraction strategy. To predict the interfering multiples, we use a data‐driven approach that combines the SSP (surface seismic profile) and VSP data. The method of stationary phase shows that by convolving the VSP data and SSP data the VSP interfering multiples could be predicted. The advantage of this approach is that no subsurface information is needed in the prediction, which is different from our previous model‐based method that requires an estimated (velocity) model. The limitation is that a separate SSP data will be needed. This method is applicable to both 2D and 3D VSP data; In marine environment it can also be applied to multi‐component VSP data recorded with geophones. The theory for the data‐driven approach is presented and the proof by method of stationary phase is provided. This abstract also shows another application of the data‐driven approach to predict the VSP down‐going components, which might be useful in VSP upgoing/downgoing component separation.