Stochastic simulation of stratigraphic sequences from well log data using continuous wavelet transforms

@inproceedings{PERRIER_RM2021,
 abstract = { The analysis of well log data is essential for subsurface studies, as it forms the basis for petrophysical, facies and stratigraphic interpretation. However, significant uncertainty can be associated to the stratigraphic interpretation of well logs. The main goal of this project is to investigate about the continuous wavelet transform (CWT) to generate stratigraphic sequence interpretation scenarios from well logs at multiple scales. Multi-scale signal processing has already been proposed to help such interpretations. However, most existing methods often focus on petrophysical interpretation, and yield lithostratigraphic units rather than sequence stratigraphic units and time-equivalent boundaries. A second limitation of existing methods is that they often produce a best guess result, whereas we would rather like to sample the interpretation uncertainties before considering well correlation problems to possibly reduce these uncertainties. For this, our approach identifies the likelihood to have a maximum regression surface (MRS) and a maximum flooding surface (MFS) along the well given the recorded logs. This approach uses the continuous wavelet transform, as it has the ability to combine information about the local frequency content without losing the spatial content. Also, the CWT allows for choosing the spatial scales deemed to correspond to the cycle order of the targeted MFS and MRS. The methodology is implemented as a stand-alone computer program in C++ and Python. Applications and qualification of the method's results will be made on North Sea data sets which have already been interpreted by sedimentologists using cored wells. },
 author = { Perrier, Yassine AND Baville, Paul AND Caumon, Guillaume },
 booktitle = { 2021 RING Meeting },
 publisher = { ASGA },
 title = { Stochastic simulation of stratigraphic sequences from well log data using continuous wavelet transforms },
 year = { 2021 }
}