Semi-automated correlation of Silurian well-logs across Gotland, Sweden, using Dynamic Time Warping with Barycenter Averaging

Michiel Arts , Daniel Sopher, Mikael Calner, Mikael Erlström, Rohit Samant, Bradley D Cramer,, Anne-Christine Da Silva

Studies of the Silurian sequence on Gotland have significantly advanced our understanding of Silurian climate dynamics, with much of the research focusing on small outcrops and short cores. Gotland has an extensive network of abandoned oil and gas wells. Most of these include gamma-ray well logs, which have not yet been fully utilized for their stratigraphic value. While the Ordovician succession has been successfully correlated using these well logs, the correlation for the Silurian succession has yet to be fully realised. The present study addresses some of the limitations of correlations performed using Dynamical Time Warping (DTW) by combining DTW with Barycenter Averaging (DBA). This enables a semi-automated correlation of the Silurian well-logs over a ~60 km-long transect and the subdivision of gamma-ray log logs into parts that can be linked with their lithostratigraphic surface counterparts. Additionally, we tracked changes in gamma-ray logs associated with these subsurface units, providing new insights into biogeochemical events (e.g., δ¹³C excursions) previously mapped only at the surface of Gotland. Our results demonstrate the application of DTW/DBA-based methods for stratigraphic correlation and highlight the potential for integrating subsurface well-log data to refine the Silurian stratigraphy of Gotland.