Program

Day 1:  September 4th, 2024
Venue: Jackson Center - 6001 Moquin Dr NW, Huntsville, AL 
7:30 - 8:30
Registration/ Continental Breakfast

8:30 - 8:45 

Workshop Opening and Logistics
Dr. Rahul Ramachandran, Dr. Budhu Bhaduri, Dr. Dalton Lunga

8:45-9:00

NASA MSFC Welcome - Mr. Larry Leopard Associate Director, Technical, National Aeronautics and Space Administration, Marshall Space Flight Center

Session 1
09:00 - 10:30

Progammatic Needs
Moderators
Dr. Tsengdar Lee, National Aeronautics and Space Administration
Dr. David Page, Oak Ridge National Laboratory
AI Foundation Models and Geospatial Digital Twins are a paradigm shift in geospatial applications that offers the potential for real-time insights for “what-if'' discoveries at a higher tempo than is possible in the physical world alone. Additionally, the automation through AI and the data-centric emphasis through digital twins has the potential for programmatic efficiences to achieve the scaling necessary for the eminent (or some might argue at-hand) Trillion Pixel Challenge. To achieve this potential, programmatic needs demand investments in software algorithms, computational infrastructure, workforce training, workflow modernization, policy changes, and perhaps even cultural mindsets among other needs. This panel of experts from government, academia, and industry will discuss these needs and the challenges along with possible solutions to address these needs.

Key Questions

  • From a programmatic level, do you agree that AI Foundation Models and Geospatial Digital Twins are a paradigm shift for geospatial applications? If so or if not, what benefits (if any) do these technologies bring to your geospatial organization and to the Trillion Pixel Challenge? What downfalls should we also be aware of?
  • Most paradigm shifts follow a hype cycle before adoption finds practical grounding. From a programmatic level, do you perceive a hype cycle for AI Foundation Models, Geospatial Digital Twins, and/or their integration together? If so, where are we in such a cycle, and what are the key programmatic challenges to move toward widespread practical adoption and deployment? Have you experienced similar paradigm shift successes or hype failures  in the geospatial industry in the past? If so, can you discuss in the context and applicability to the Trillion Pixel Challenge?
  • Beyond geospatial practitioners, who are the stakeholders that will resource (funders, suppliers, procurers, etc.) and utilize (users, decision-makers, etc.) the integration of AI Foundation Models and Geospatial Digital Twins? What challenges do you anticipate to gain trust and buy-in from the stakeholders and how do you recommend addressing such challenges?

Panelists

  • Mr. Todd Johanesen, National Geospatial-Intelligence Agency 
  • Dr. May Yuan, National Science Foundation
  • Dr. Mike Tischler, U.S. Geological Survey
  • Dr. Sid Ahmed Boukabara, National Aeronautics and Space Administration
  • Dr. Kunhikrishnan (Kunhi) Thengumthara, Office of Science and Technology Policy / Interagency Council for Advancing Meteorological Services

    10:30–11:00

    		Coffee Break

    Session 2
    11:00 - 12:30

    Artificial Intelligence
    Moderators
    Professor Wenwen Li, Arizona State University
    Dr. Lexie Yang, Oak Ridge National Laboratory
    Dr. Sujit Roy, National Aeronautics and Space Administration
    This session will delve into the innovative application of AI Foundation Models within Geospatial Digital Twins, highlighting the specific advancements and unique challenges involved. AI foundation models are large models trained on a massive amount of data to gain an unprecedented level of generalizability, making them easily adaptable to a wide variety of downstream tasks, from mapping landscapes and manmade features to detecting flooded regions and forecasting wildfire spread. The strong predictive capability of AI foundation models makes them an important tool for building the Digital Twins of natural and built environments. They can also advance automated processing of multimodal data, supporting high-stakes decision-making in a Digital Twin environment. The session aims to provide a comprehensive understanding of the role of AI in advancing Geospatial Digital Twins. By highlighting both the opportunities and the challenges, the discussion seeks to inform and inspire stakeholders to leverage AI Foundation Models for more robust and reliable Digital Twins.

    Key Questions

    This session will provide a technical examination of:

    • How are AI foundation models pushing the boundaries of what's possible in Geospatial Digital Twins? How do they compare to traditional physics-based models in terms of computational load, efficiency, accuracy, and explainability?
    • What are the barriers to effective AI integration within Geospatial Digital Twins?
    • What approaches can be taken to define uncertainty quantification (UQ) characteristics with respect to data or models?
    • What are the real-world applications and success stories of AI foundation models in understanding the changing natural and built environments?
    • What opportunities does AI present for contributing to the next wave of Earth scientific discovery?

    Panelists

    • Professor Shawn Newsam, University of California, Merced
    • Dr. Dan Lu, Oak Ridge National Laboratory
    • Mr. Zongyi Li , California Institute of Technology
    • Dr. Johannes Schmude, IBM's Thomas J. Watson Research Center
    • Dr. Raju Vatsavai, North Carolina State University
    • Dr. David Hall, NVIDIA

    12:30-14:00

    		Lunch Break + Talk,  Science at NASA Marshall Space Flight Center, Dr. Renee Weber, Chief Scientist

    Session 3
    14:00–15:30 

    Data and Infrastructure
    Moderators
    Ms. Katie Baynes, National Aeronautics and Space Administration
    Dr. Forrest Hoffman, Oak Ridge National Laboratory
    Dr. Aaron Kaulfus, National Aeronautics and Space Administration
    This session will seek an in-depth discussion featuring experts from governments, academia, and industry to explore the crucial role of data and compute infrastructure in advancing the integration of models and data, with a focus on the needs of foundation models and geospatial digital twins.

    Key Questions

    This session will provide a technical examination of:

    • Advanced strategies and methods for efficient data collection, storage, curation, management, and analysis tailored to supporting large-scale/high-performance processing
    • Approaches to improve data discovery and accessibility, enhance data system scalability and reliability, and deliver compute capacity at the data location
    • Techniques to enhance data assimilation and model performance, scalability, and predictive capabilities in Earth Science
    • Real-world case studies highlighting deployments and practical implications of large-scale Earth Science systems
    • Collaborative frameworks and partnerships essential for developing and sustaining robust data and compute infrastructure ecosystems

    Panelists

    • Dr. Kjiersten Fagnan, US Department of Energy Joint Genome Institute
    • Mr. Ian Schuler, Development Seed
    • Mr. Hook Hua, Jet Propulsion Laboratory - California Institute of Technology
    • Ms. Laura Carriere, High-Performance Computing Lead, NASA Center for Climate Simulation
    • Professor Ahmed Eldawy, University of California Riverside
    • Dr. Jay Hnilo, Department of Energy

    15:30–15:45

    		Coffee Break

    Session 4
    15:45–17:15 

    People and Partnerships
    Moderators
    Dr. Beth Plale , Indiana University
    Dr. Samantha Arundel, U.S. Geological Survey
    This session focuses on the pivotal roles of individuals and collaborative efforts in progressing AI-enhanced geospatial tools within the Trillion Pixel Challenge. Discussions will cover the trustworthiness and dependability of AI-augmented tools, optimal user interface design, and the appropriate autonomy levels for AI-driven Digital Twins. The session will also spotlight existing community consortiums and forums for ongoing dialogue, effective ways to engage stakeholders, and emerging commercial partners in the geospatial AI field. The goal is to identify strategies for cultivating strong, trust-based partnerships and improving human-AI interactions in geospatial applications.

    Key Questions

    • People:  How much trust can be placed in AI-augmented tools and results of AI-augmented tools? How will people interact with these systems? How much autonomy should an AI enabled Digital Twin have?
    • Partnerships:  Are there community consortiums/venues for continued (or related) conversations?  What are the vehicles for stakeholder involvement?   Who are the emerging partners from commercial companies?

    Panelists

    • Mr. Mark Korver, Taylor Geospatial Institute
    • Dr. Sambit Bhattacharya, Fayetteville State University
    • Mr. Mark Cygan, Environmental Systems Research Institute, Inc.
    • Dr. Lingbo Liu, Harvard University
    • Venkat Vishwanath, ANL and Trillion Parameter Consortium

    17:30 

    		Dinner own your own

    Day 2 : September 5th, 2024
    Venue: Jackson Center - 6001 Moquin Dr NW, Huntsville, AL 
    8:00 - 8:30Check in/ Continental Breakfast

    Session 5
    08:30–10:00

    Hardware and Software Architectures
    Moderators
    Ms. Shubha Ranjan, National Aeronautics and Space Administration
    Mr. Valentine Anantharaj, Oak Ridge National Laboratory
    The volume and velocity of earth observations require scalable high-performance computing (HPC) hardware and software infrastructure in order to transfer, store and analyze the vast amounts of data in GeoAI applications. Planetary-scale GeoAI digital twins, capable of learning and inferring from trillions of pixels streaming daily, will require tight integration of network, storage, and computing ecosystems that span CPUs, GPUs, accelerators, and potentially domain-specific architectures such as a GeoAI spatial processor. Also needed are scalable open-source software ecosystems to analyze these datasets by leveraging state-of-the-art computational resources. This session will explore the hardware and software architecture challenges posed by GeoAI and discuss potential and holistic architectural and infrastructure solutions to overcome these challenges.

    Key Questions

    • What are our present day GeoAI challenges & problems from the hardware and software architecture & infrastructure perspective?
    • How can the GeoAI community leverage existing hardware and platforms for GeoAI applications today, with respect to data, computing and workflows?
    • How can resource providers and sponsors help find solutions to advance GeoAI opportunities today and explore solutions for the future?
    • What are the motivations behind the current generation of digital twins for geospatial, climate and environmental applications? What are the challenges and lessons learned in developing and deploying the geospatial and climate digital twins?

    Panelists

    • Mr. Chris Zimmer, Oak Ridge National Laboratory
    • Dr. Neena Imam, Southern Methodist University
    • Mr. Kyle Lamb, VAST Data
    • Dr. Seetharami Seelam, IBM
    • Mr. Bijan Varjavand, Hewlett Packard Enterprise
    • Dr. Catherine Schuman, University of Tennessee

    10:00-10:30 

    		Coffee Break

    Session 6
    10:30–12:00

    Climate and Water Security
    Moderators
    Dr. Carter Christopher, Oak Ridge National Laboratory
    Dr. Assaf Anyamba, Oak Ridge National Laboratory
    The concept of a digital twin for water is getting increasing visibility among the Earth science and engineering communities, yet these communities define and model water systems uniquely. Additionally, the data used within these communities are themselves not well integrated, and across these communities even less so. With climate change expected to impact water security across the globe, an integrated view is essential to situational awareness and risk and resilience assessments. Harmonizing these datasets and modeling frameworks also can enable the development of a digital twin of water. If we define a digital twin as a temporally accurate and specific representation of Earth’s water state, availability, and quality, there are significant challenges that emerge to this harmonization. This session will explore the need, opportunity, and challenges with creating a digital twin of water, including the data disparities and gaps, and the role of GeoAI foundation models as an integrating framework.

    Key Questions

    • What are the key datasets characterizing Earth’s water system that need integration, and what are the impediments to harmonization that this community can address?
    • What are the data gaps, sensing gaps, and model gaps that must be addressed to enable a digital twin for water?
    • What is the role of GeoAI relative to traditional remote sensing techniques for harmonizing dataset spatial resolution, filling data gaps, and improving temporal cadence and consistency?

    Panelists

    • Dr. Sujay Kumar, National Aeronautics and Space Administration
    • Dr. Wenwen Li, Arizona State University
    • Ms. Laura Rogers, National Aeronautics and Space Administration
    • Dr. Jitendra Kumar, Oak Ridge National Laboratory

    12:00 - 12:30

    		Wrap Up/Workshop Close