Incorporating Uncertainty in Automated Seismic Interpretation: Geobody Volumetrics

We apply a method for estimating deep learning model uncertainty to automated seismic interpretation of complex geobodies from the Paraihka 3D seismic dataset, offshore New Zealand. With the aim of estimating the Gross Rock Volume of a potential stratigraphic trap within part of a complex gully system, including quantification of the uncertainty in those estimates stemming from model parameter choices and variation. This work was presented at the DIGEX2019 conference as an illustration of the potential for automated seismic interpretation based on deep learning to provide static earth models with uncertainty estimates over them alongside rich and detailed output for visualisation.