This document is an expanded abstract.
Summary
We introduce a generalized concept of blending and deblending, establish its models, and accordingly establish a method of deblended-data reconstruction using these models. The generalized models can handle real-life situations by including random encoding into the generalized operators both in the space and time domain, and both at the source and receiver side. We consider an iterative optimization scheme using a closed-loop approach with the generalized-blending and -deblending models, in which the former works for the forward modelling and the latter for the inverse modelling in the closed loop. We established and applied this method to existing real datasets offshore Abu Dhabi. The results show that our method succeeded to fully reconstruct deblended data even from the fully generalized, thus quite complicated blended data.
Introduction
In traditional acquisition, spatial and temporal interference between shots is avoided, often resulting in poor sampling in the source dimension. However, in blended acquisition, the interference is allowed, leading to dense and wide sampling in an economical way. We can achieve this by blending and deblending, i.e. blended acquisition followed by deblended-data-reconstruction processing.
As for blending at the source side, first, blended acquisition stands for continuous recording of seismic responses from incoherent shooting, the properties of which are characterized and encoded by random spatial distribution and time shifts among the involved source units of blended-source array (Berkhout, 2008). Second, blended acquisition uses inhomogeneous shooting, in which the blended-source array consists of different source units rather than traditional equal ones, e.g. narrow-frequency-banded versions instead of a certain broad-frequency-band one: dispersed source array, or DSA (Berkhout, 2012). Third, this acquisition also uses signature stamping, in which each source unit is encoded with its own signature, e.g. various sweeping for land (Bagaini, 2006); popcorn shooting for marine (Abma and Ross, 2013). Furthermore, another concept, particularly at the receiver side, is spatial sampling based on compressive sensing, which adopts non-uniform- and under-sampling acquisition followed by regularization and interpolation processing; where a signal can be recovered from far fewer samples than required by the Shannon-Nyquist theorem (Mosher et al., 2014).
