From field notebook to machine learning: digitized measured sections for predicting deepwater reservoir architecture

Stratigraphic measured sections capture outcrop observations including bed thickness, grain size, and sedimentary structures. These data underpin interpretations of depositional processes and their translation into preserved stratigraphy, forming the basis for predictive lithofacies models in sedimentary basins. Such models are essential for subsurface applications including petroleum exploration, resource extraction, and carbon dioxide sequestration, where realistic sedimentologic architecture is critical for reliable forecasting.

Traditional workflows capture field data in a field notebook or digital field platform (e.g., the NSF-funded StraboSpot project), typically producing vertical sections and a correlation panel to show surfaces and spatial sedimentary relationships. However, these representations are poorly suited for quantitative analysis, including machine-learning workflows, and limit direct comparison with subsurface datasets. As a result, much of the information contained in stratigraphic measured sections remains underutilized.

Over a decade of work in the Cretaceous Dorotea and Tres Pasos Formations of the Magallanes Basin (Chilean Patagonia) has produced >10,000 meters of data from >300 measured sections spanning a >2-km-thick, >50-km-long shelf-to-basin succession. These sections have been digitized into a relational database that captures grain-size profiles and bed thickness within a hierarchical framework. The measured section database is linked to a document webserver to provide contextual information (i.e., manuscripts, maps, correlation panels, and images). The database produces pathways to translate field data, observations, and interpretations into tangible, testable, and shareable data products. A case study using the database to develop machine learning and architectural model demonstrates the power of this digital data for modeling sedimentologic architecture in the subsurface.