Identifying internal multiples using 1D prediction: physical modelling and land data examples

Melissa J. Hernandez and Kristopher A. Innanen

ABSTRACT

Internal multiples, if not properly identified, are a significant impediment to seismic reflection data analysis. As researchers engage with the full multidimensional internal multiple prediction and removal problem, it has been suggested that some of the obstacles this overall problem presents, especially on land, can be addressed by applying 1D prediction algorithms to near offset or post-stack data. We examine this possibility by carrying out 1D predictions on a zero-offset physical modelling data set and a post-stack land data set, both of which are likely to contain significant multiple energy. Our results confirm the kinematic accuracy of the predictions by comparing them against synthetic traces, and flesh out the problem of optimally choosing the integration limit parameter e in the algorithm. The results also bear out the idea of using predictions alone as a quick interpretation tool. The prediction output in any given case may potentially be too noisy to permit effective subtraction. However, it may yet constitute a sort of "multiple probability map", useful for identifying both multiples themselves, and primaries whose amplitudes are likely to have experienced interference from them.

View full article as PDF (13.68 Mb)