Reproducible Research, Seismic Attributes, and Shaping Regularization
Reproducible research, in application to computational sciences, implies a system for linking scientific publications with computational recipes. Following such system allows a community of researchers to verify and extend the results of each other's numerical experiments. A particular application area where reproducibility is vitally important is seismic data analysis.
Local seismic attributes measure seismic signal characteristics not instantaneously at each signal point and not globally across a data window but locally in the neighborhood of each point. I define local attributes with the help of regularized least-squares inversion and demonstrate their applicability for measuring a number of useful characteristics, such as local frequency or local similarity. I use shaping regularization, a general method for imposing constraints on inverted models, for controlling the locality and smoothness of local attributes. Reproducible computational experiments from time-lapse seismic monitoring of subsurface fluid flow illustrate practical applications of the proposed technique.
Sergey Fomel, University of Texas at Austin, USA
