Incorporating the multiscale Fourier transform, principal components, and significance measures in seismic prediction of rock properties

We present a workflow for predicting reservoir properties based on incorporating the Multiscale Fourier Transform(MSFT) into multi-attribute analysis. To use it efficiently, we make use of Principal Component Analysis(PCA) to extract the most important parts of the MSFT data. To evaluate the workflow’s effectiveness and guide training selection of attributes, we utilize various significance measures in addition to goodness of fit. We find an increase in the predictive power of regression models produced using our workflow compared to using only conventional attribute volumes. This improvement stems from the holistic and efficient frequency response representation of the data enabled by the properties of the MSFT.