TL;DR
This paper introduces GS³, an open-source, flexible framework built on Velo, designed to manage the complex data and workflows involved in CO₂ geologic sequestration (GS) projects. GS³ streamlines collaboration, data sharing, and model comparison across diverse projects, from international model intercomparison studies to regulatory permit applications.
What Problem Does This Paper Address?
Geologic sequestration of CO₂ is crucial for climate change mitigation, but it involves extensive site characterization, numerical simulations, and long-term monitoring. Managing the vast amounts of data generated throughout these projects, ensuring its accessibility, and maintaining a clear record of how models were developed (provenance) is a significant challenge. Existing computational tools are often single-purpose, platform-specific, and impose rigid workflows, hindering flexibility and collaboration. This paper addresses the need for a unified, adaptable system that can handle the diverse data types and workflows inherent in GS projects, facilitating efficient knowledge management and regulatory compliance.
Methods
The core of the solution is the Velo platform, an open-source knowledge management system that integrates a collaborative web-based environment with a scalable enterprise content management system. Velo provides a flexible, unstructured data storage system accessible via a customized MediaWiki interface.
GS³ (Geologic Sequestration Software Suite) is a specialized layer built on top of Velo, tailored for GS studies. Key features include:
- User-tailored interface: For capturing, organizing, and querying data for computational models.
- Versioning and provenance: Tracks model inputs and associates results with specific input datasets, maintaining a detailed project history.
- Integration of third-party tools: Supports various modeling activities, including model development and visualization, through tight and loose integration.
- Remote simulation launching: Enables running simulations on distributed computational platforms, including supercomputers.
- Uncertainty analysis and model comparison tools: Facilitates evaluating model performance and understanding uncertainties.
GS³'s architecture is designed to be extensible, allowing for the incorporation of domain-specific tools and data types through published programmatic interfaces.
Datasets
The paper does not explicitly list specific datasets used for training or evaluation in a traditional machine learning sense. Instead, it describes how GS³ is applied to manage and analyze data generated from real-world GS projects and studies. These include:
- Site characterization data: Geophysical logs (LAS files), seismic data (SEGY files).
- Simulation results: Output from numerical simulators like STOMP-CO2.
- Conceptual model data: Information describing model assumptions and development.
- Regulatory data: Information submitted for Class VI Well permit applications.
The Sim-SEQ international model comparison study is highlighted as a key application where GS³ facilitates the comparison of multiple subsurface models using a common dataset.
Key Insights and Ideas
- Flexibility is paramount: The success of GS³ lies in its adaptability. It's not a one-size-fits-all solution but a framework that can be customized for diverse project needs, from research collaborations to regulatory reviews.
- Provenance is critical for trust: In regulated fields like CO₂ sequestration, being able to meticulously track how data was generated, how models were built, and what assumptions were made is essential for accountability and regulatory approval. GS³'s versioning and metadata management directly address this.
- Bridging the gap between science and regulation: GS³ serves as a crucial tool for both researchers developing models and regulatory bodies reviewing permit applications, fostering better communication and understanding.
- Open-source fosters collaboration: The open-source nature of Velo and GS³ encourages wider adoption and community contributions, accelerating progress in the field.
Key Results
The paper showcases GS³'s utility through three distinct implementations:
- Sim-SEQ Model Comparison Study: GS³ provided a collaborative framework for international modeling teams, enabling easy sharing, querying, and updating of site characterization data and simulation results. Custom tools like model description worksheets facilitated consistent model comparisons. Uncertainty quantification (UQ) analysis using GS³ revealed that modeling schemes played the most significant role in predicting breakthrough times and pressures.
- Typical CO₂ Geologic Storage Project: GS³ supports the entire lifecycle of a storage project, from site characterization to numerical simulation for permit applications. It facilitates information sharing among diverse team members and provides a transparent record for project management and external reviewers. A PDF extractor tool aids in semantic tagging of keywords within reference documents.
- US EPA Class VI Well Permit Application Review: GS³ is used by the EPA to evaluate Area of Review (AoR) delineations. It supports independent simulations by EPA and provides resources like STOMP-CO2 user manuals and input snippets to assist applicants. A "pressure front calculator" tool helps delineate the AoR based on site-specific parameters.
Limitations and Future Work
- Proprietary vs. Open-Source Tools: While GS³ aims to integrate various tools, the paper notes that many conceptual model development tools are proprietary or platform-specific, posing integration challenges.
- Customization Effort: While flexible, tailoring GS³ for specific applications still requires development effort.
- Future Work: The authors mention ongoing development of new methods for model comparison, ranking, and tuning, including Bayesian methods for information accrual and ensemble data comparison. They are also enhancing capabilities for launching and monitoring simulation ensembles, with analysis and visualization performed "in-flight" on high-performance compute clusters. The development of a new web-based job launcher that doesn't require a user account is also underway.
Summary updated