A nonstationary search parameter for internal multiple prediction

Kris Innanen


With the internal multiple prediction algorithm re-formulated in terms of the output time, the opportunity arises for the search limiting parameter to be varied according to some schedule ε(t) . Any prior knowledge suggesting that within a single trace this output time requires aggressive prediction, but that output time requires cautious prediction, can be used to guide an appropriate selection of ε(t) . Here we will consider two fairly obvious types of criterion for selection: data driven selection strategies and strategies driven by geological prior knowledge. 1D and zero offset synthetic and physical modelling lab data are used to find situations where the precision of predictions is increased beyond what would be available given any single fixed ε parameter. The promising results are suggestive that a much broader study of selection strategies for ε as functions of time, frequency, offset, lateral wavenumber, source and receiver horizontal slownesses, etc., is warranted.

View full article as PDF (2.56 Mb)