Tackling 21st Century Geoscience Problems with Machine Learning

Curtis, Andrew; O’Malley, Daniel; Beroza, Gregory C.; Johnson, Paul; Li, Elita; Haber, Eldad

doi:10.1029/2020eo150184

Cited by 2 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For well-log analysis and prediction, DNN class of models would need a specialized dataset or research question to take advantage of the inherent model complexity and infrastructure. While further research will continue to refine which models and class of models have the best performance, future needs for most geoscience professionals will include easier model deployment, interpretation, and dataset augmentation (Burkov, 2020;Curtis et al, 2020;Jahic et al, 2019;Molnar, 2022). Future researchers can build on the results this study to improve ML model parameterization and deployment practices.…”

Section: Machine Learning Model Selectionmentioning

confidence: 99%

Digitalization of Legacy Datasets and Machine Learning Regression Yields Insights for Reservoir Property Prediction and Submarine-Fan Evolution: A Subsurface Example From the Lewis Shale, Wyoming

Martín

Tadla

Jobe

2022

The Sedimentary Record

View full text Add to dashboard Cite

Machine-learning algorithms have long aided in geologic property prediction from well-log data, but are primarily used to classify lithology, facies, formation, and rock types. However, more detailed properties (e.g., porosity, grain size) that are important for evaluating hydrocarbon exploration and development activities, as well as subsurface geothermal, CO2 sequestration, and hydrological studies have not been a focus of machine-learning predictions. This study focuses on improving machine-learning regression-based workflows for quantitative geological property prediction (porosity, grain size, XRF geochemistry), using a robust dataset from the Dad Sandstone Member of the Lewis Shale in the Green River Basin, Wyoming. Twelve slabbed cores collected from wells targeting turbiditic sandstones and mudstones of the Dad Sandstone member provide 1212.2 ft. of well-log and core data to test the efficacy of five machine-learning models, ranging in complexity from multivariate linear regression to deep neural networks. Our results demonstrate that gradient-boosted decision-tree models (e.g., CatBoost, XGBoost) are flexible in terms of input data completeness, do not require scaled data, and are reliably accurate, with the lowest or second lowest root mean squared error (RMSE) for every test. Deep neural networks, while used commonly for these applications, never achieved lowest error for any of the testing. We also utilize newly collected XRF geochemistry and grain-size data to constrain spatiotemporal sediment routing, sand-mud partitioning, and paleo-oceanographic redox conditions in the Green River Basin. Test-train dataset splitting traditionally uses randomized inter-well data, but a blind well testing strategy is more applicable to most geoscience applications that aim to predict properties of new, unseen well locations. We find that using inter-well training datasets are more optimistic when applied to blind wells, with a median difference of 0.58 RMSE when predicting grain size in phi units. Using these data and results, we establish a baseline workflow for applying machine-learning regression algorithms to core-based reservoir properties from well-log and core-image data. We hope that our findings and open-source code and datasets released with this paper will serve as a baseline for further research to improve geological property prediction for sustainable earth-resource modeling.

show abstract

Section: Machine Learning Model Selectionmentioning

confidence: 99%

Digitalization of Legacy Datasets and Machine Learning Regression Yields Insights for Reservoir Property Prediction and Submarine-Fan Evolution: A Subsurface Example From the Lewis Shale, Wyoming

Martín

Tadla

Jobe

2022

The Sedimentary Record

View full text Add to dashboard Cite

show abstract

A Perspective on Concretions: Deciphering Diagenesis from Earth to Mars

Chan

2022

Jour. Geol. Soc. Japan

View full text Add to dashboard Cite

There is an amazing array of concretions throughout the sedimentary record of Earth, and now iron oxide examples ( blueberries ) recognized in several regions of Mars. Two questions address the current state of knowledge on nodular cemented mineral masses as well as coloration patterns. Collectively the authigenic cements and patterns chronicle past diagenetic conditions, particularly in clastic rocks. What do we know? Concretions are cemented mineral masses that occur in many sizes, spanning three orders of magnitude (mm, cm, and m scales). Spheroidal forms are most common, as an minimum free stage dominated by diffusive processes. There are multiple cement mineralogies, sometimes even within single concretions, reflecting different water compositions in open systems. Other concretionary geometries are affected by primary textures such as bedding, grain size, and porosity/permeability, or later textures such as fractures, joints, and faults. Iron cycling is readily apparent where visual coloration patterns indicate histories of early iron reddening, secondary bleaching (removal of iron), and iron replacement or reprecipitation. Interfingering colors may indicate a possible interface of immiscible fluids.What are the remaining challenges? There are many aspects of concretion diagenesis that are still yet to be deciphered. Non-unique pathways or processes may produce similar-looking end products. Thus, it can be difficult to determine exact histories, as well as the fluid compositions and environmental conditions that initiate concretion formation, particularly if an obvious nucleus is lacking. Microbial life may enhance nucleation and precipitation, and geochemical gradients are potential places to search for biosignatures. Timing and events are mostly relative relationships in these open systems, but newer developments in U-Th/He dating may provide age constraints for iron oxide cements. Continued explorations, field studies, modeling approaches, analytical advances, and instrument precision will enlighten our understanding on the diagenetic histories of both Earth and Mars.

show abstract

Tackling 21st Century Geoscience Problems with Machine Learning

Abstract: A new cross-journal special collection invites contributions on how machine learning can be used for solid Earth observation, modeling and understanding.

Cited by 2 publications

References 0 publications

Digitalization of Legacy Datasets and Machine Learning Regression Yields Insights for Reservoir Property Prediction and Submarine-Fan Evolution: A Subsurface Example From the Lewis Shale, Wyoming

Digitalization of Legacy Datasets and Machine Learning Regression Yields Insights for Reservoir Property Prediction and Submarine-Fan Evolution: A Subsurface Example From the Lewis Shale, Wyoming

A Perspective on Concretions: Deciphering Diagenesis from Earth to Mars

Contact Info

Product

Resources

About