Reconstructing the Distribution of Turbiditic Gas Sandstones in the SW Pannonian Basin Using Machine Learning Methods

ABSTRACT Seismic interpretation in geologically complex environments still remains a challenge due to limited resolution and ambiguous facies differentiation. In this article the 3D seismic data from the Gola field, south‐western Pannonian Basin, were interpreted using unsupervised self‐organizing maps (SOM) on unlabelled data. The complexity of this active gas field lies in the distribution of alluvial fan sediments. The sediments were spatially not connected due to the interplay of turbiditic processes within an alluvial fan system, resulting in a heterogeneous facies distribution and rapid lateral and vertical lithological changes. However, the sandstone reservoir bodies have been identified in the well data. Therefore, the well data were used to verify the extracted geobodies based on multi‐attribute seismic analysis. The optimal SOM architecture (5 × 5) produced geologically meaningful clusters, with the extracted geobodies confirming sandstone continuity for the first time in this area, demonstrating the robustness and scalability of the Python‐based workflow for interpreting complex sedimentary systems. Seismic resolution was further enhanced through SOM‐driven geobody extraction, enabling the identification of sandstone bodies below the tuning thickness.