A systematic, science-driven approach for predicting subsurface properties

Abstract

As human exploration of the subsurface increases, there is a need for better data- and knowledge-driven methods to improve prediction of subsurface properties. Present subsurface predictions often rely upon disparate and limited a priori information. Even regions with concentrated subsurface exploration still face uncertainties that can obstruct safe and efficient exploration of the subsurface. Uncertainty may be reduced, even for areas with little or no subsurface measurements, using methodical, science-driven geologic knowledge and data. We have developed a hybrid spatiotemporal statistical-geologic approach, subsurface trend analysis (STA), that provides improved understanding of subsurface systems. The STA method assumes that the present-day subsurface is not random, but is a product of its history, which is a sum of its systematic processes. With even limited data and geologic knowledge, the STA method can be used to methodically improve prediction of subsurface properties. To demonstrate and validate the improved prediction potential of the STA method, it was applied in an analysis of the northern Gulf of Mexico. This evaluation was prepared using only existing, publicly available well data and geologic literature. Using the STA method, this information was used to predict subsurface trends for in situ pressure, in situ temperature, porosity, and permeability. The results of this STA-based analysis were validated against new reservoir data. STA-driven results were also contrasted with previous studies. Both indicated that STA predictions were an improvement over other methods. Overall, STA results can provide critical information to evaluate and reduce risks, identify and improve areas of scarce or discontinuous data, and provide inputs for multiscale modeling efforts, from reservoir scale to basin scale. Thereby, the STA method offers an ideal framework for guiding future science-based machine learning and natural language processing to optimize subsurface analyses and predictions.

A Geo-Data Science Method for Assessing Unconventional Rare-Earth Element Resources in Sedimentary Systems

“…Geo-data science leverages tools and techniques such as data statistics, machine learning (e.g., clustering analysis), and GIS to efficiently collect, integrate, analyze, and mine geoscience datasets, which in turn leads to knowledge discovery and insights about spatial decisions (Gibert et al, 2018b;Zuo & Xiong, 2020). One such approach, the Subsurface Trend Analysis (STA) method developed by Rose et al (2020), is a systematic, science-and data-driven method that utilizes geologic systems knowledge to inform statistical analyses of subsurface properties and resources. The STA workflow integrates geologic data used in traditional assessment approaches (e.g., geologic maps, in situ measurements, core) with appropriate statistical methods (e.g., geostatistics, dimensional analysis) to characterize the subsurface property of interest.…”

“…This involves characterizing the geologic history of the region of interest and identifying areas with common geologic attributes, such as basin sediments with shared provenance or regions with similar structural character, following the Subsurface Trend Analysis (STA) method as described in Rose et al (2020). The STA method involves using geologic knowledge and data relating to lithology, structure, and secondary alteration-three primary factors affecting subsurface properties-to partition an area into genetically related spatial domains (Fig.…”

A Python Tool for Predicting and Assessing Unconventional Rare-Earth and Critical Mineral Resources

“…There are several ways that the URC Resource Assessment Tool can be configured to run, but fundamentally the tool takes in a collection of spatial domains which fall under the Lithological, Structural, and Secondary Alteration categories defined by the Subsurface Trend Analysis (STA) method (Rose et al, 2020). These domains are combined, clipped to a researcher-defined boundary, and grided to cells of a research-specified dimension (see Figure 1 for overview of the process).…”

“…Figure 1: High level overview of the tool workflow when carrying out the Create Grid task. Domains created using the STA method (Rose et al, 2020) are rasterized using the provided width and height values for each pixel. If desired a geospatial projection can be specified to be assigned to the results by providing a projection or European Petroleum Survey Group (EPSG) code as a "projection override".…”

Enhancing knowledge discovery from unstructured data using a deep learning approach to support subsurface modeling predictions

“…In either instance, the tool extracts visualizations, categorizes the visualization via a convolutional neural net (CNN) built on the VGG16 architecture, and gives the user the opportunity to view them ( Figure 1 ). The image imbedding tool is part of a software suite that emerged from the National Energy Technology Lab's (NETL) Subsurface Trend Analysis (STA) workflow, which was developed to assist subsurface research by bringing greater contextual knowledge to measured data such as cores, well logs, and seismic surveys (Rose et al, 2020 ).…”