pyCoreRelator: automated correlation of data from cores and geophysical logging

Stratigraphic correlation of cores and/or geophysical logging data is fundamental to basin analysis, reservoir characterization, and more derivation of paleoclimate and paleoearthquake studies. However, conventional correlation approaches remain subjective and are often not reproducible. To address this, we developed pyCoreRelator, a Python tool that offers a more objective method for stratigraphic correlation. This tool converts visual core imagery (e.g., CT scans, color photos) into digital logs that can be processed alongside geophysical logs. To correlate stratigraphic units across sites, pyCoreRelator searches for all feasible correlation solutions that follow the principle of superposition, while honoring chronological constraints and potential lateral stratal discontinuities. For every plausible correlation solution, the tool employs dynamic time warping to maximize similarity and minimize signal distortion that arises from varying sedimentation and erosion. To evaluate the robustness of correlations, the tool compares similarity and distortion metrics against those of randomly stacked synthetic stratigraphy. We confirmed the effectiveness of this new approach by applying pyCoreRelator to Cascadia offshore sediment cores of turbidites previously interpreted as earthquake-triggered deposits. We find that correlation quality improves markedly for core pairs collected in close proximity to one another and within the same depositional environment. Notably, the robustness of correlations decreases over long distances and across different depositional settings, and age constraints do not always improve correlation strength. These findings call into doubt the inferred synchroneity of deposition based on long-distance correlations of certain Cascadia turbidites and their associations with the same paleoearthquake. This pyCoreRelator approach has broad applicability in energy, hazards, and paleoclimate investigations by generating more objective, reproducible correlations that account for uncertainty, but further testing is worthwhile in other settings.