Multiples in F3 dataset offshore the Netherlands Our F3 Demo dataset contains many interesting geological and geophysical features that have been studied widely since we release...
Multiples in F3 dataset offshore the Netherlands
Our F3 Demo dataset contains many interesting geological and geophysical features that have been studied widely since we released the data set in OpendTect’s self-study course (see TerraNubis: www.terranubis.com).
Especially, the package deposited in the Miocene by the Eridanos Delta is widely used for teaching principles of seismic sequence stratigraphy. The package between the green and magenta horizons in the slider, features large-scale sigmoidal bedding and text-book quality downlap, toplap, onlap, and truncation structures. However, some interpreters argue that this interval is contaminated with multiples.
To investigate this hypothesis, we applied Lundin GeoLab (AkerBP)’s SimpleHmult model. SimpleHmult is a 3D U-Net in OpendTect’s Machine Learning library of trained models. The model has learned to suppress horizontal multiples on synthetic data examples. Application of SimpleHmult to the seismic data in F3 Demo takes just a few minutes on a modern laptop. The result shown in the slider is a composite of seismic data outside the interval of interest and the SimpleHmult result inside. The composite is made in the attribute engine using a bit of math and logic. SimpleHmult indeed suppressed horizontal events that are crosscutting dipping events. The result supports interpreters who warn for multiple contamination at this interval.
Author: Paul de Groot from dGB Earth Sciences
#OpendTect #seismic #multiples #geoscience #geophysics #geology #ML #machinelearning
