Identification of fluvio-deltaic facies based on 3D-seismic attributes analysis and unsupervised machine learning techniques: strategy to reduce geothermal exploration risk in the North German Basin.

The application of machine learning (ML) for reservoir characterization and prospect identification in seismic data is becoming standard practice in the exploration industry. This technique has proven useful in identifying patterns in the data that might be overlooked by the interpreter. In addition, it improve reservoir predictions and characterization at a lower computational cost.

In this study, we analyze the fluvio-deltaic seismic facies of the Upper Triassic Exter Formation in the North German Basin. For this purpose, we applied seismic attributes and an unsupervised machine learning algorithm based on waveform segmentation of seismic amplitude data. Furthermore, to evaluate the evolution of deltaic complexes, we implemented the stratal slicing technique through the resulting waveform segmentation and the generated attributes volumes.

The study resulted in the delineation of a number of fluvial architectural elements in the study area, i.e. lateral shifting of individual channels contributing to the formation of channel belt reservoirs within the Rhaetian Deltaic System.

These results contribute significantly to reducing the risk of geothermal exploration in the North German Basin by proposing for the first time a way to improve the prediction of Rhaetian reservoirs on a local scale based on seismic methods.