Spectral decomposition is a common seismic processing technique that takes apart the original seismic trace into its individual time frequency components. Three frequencies ('triplets') put together in a Red-Green-Blue color display on a seismic horizon, can show spectacular apparent geology as geological features can have different tuning frequencies. While there are models to explain certain frequency responses (e.g. a wedge model), the translation of color to geology is not trivial and relies on integration with well data.
A big issue of this present-day workflow is that the seismic and well domains have a mismatch (e.g. time/frequency vs depth) and that a 'Trial and Tweak' approach is needed. To overcome this mismatch and make results comparable, our workflow aims at bringing the two datasets into the same frequency and time domain in an interactive, semi-automated fashion. In addition, we provide a more quantitative way of selecting frequency triplets.
Uniquely, we start from the well side, where we know a zone of interest (e.g. a gas-charged sandstone), and generate dozens of synthetics by utilizing sonic and density. While the user is given the freedom to decide on the frequency combination(s) based on a depth/frequency/amplitude plot, our App suggests the best combination through cost-function minimization. With the chosen frequency triplets, optimized for the zone of interest, a well RGB log is generated. With the same parameters, spectral decomposition of the seismic cube is performed. Both seismic and well log are in the same frequency domain, and are thus comparable. The RGB response can now directly be related to lithology and/or fluid content.
The promise of this workflow is that it provides a more quantitative approach of finding optimal frequency triplets, and more importantly, that we can translate the RGB color space of both seismic and well into 'real' geology. With better constraints on what the colors represent, we can better interpret features we observe on seismic horizons, such that we can go beyond just pretty pictures.
OLD*RGB blends of SpecDecomp data can show spectacular apparent geology. We're intent on exploring the visualisation technique, spectral decomp and well data to go beyond just pretty pictures.