2021 ESIP Winter Meeting

Recent efforts to standardize data sharing and archiving guidelines within research institutions, professional societies, and academic publishers make clear that the scientific community does not have consistent recommendations regarding data produced as output from computational simulations. The massive size of some simulation outputs, as well as the large computational cost to produce these outputs, present problems related to data storage, preservation, duplication, and replication. Large-scale and smaller-scale simulations present additional challenges related to usability, understandability, documentation, and citation.

The goal of the session is to engage data facilities and other ESIP participants in a discussion of the following questions:
How to determine the value of computational model outputs? Who makes this determination?
How are requests to archive data (including model outputs) currently funded?
How do archives maximize the utility of the model-based data?
How should model codes be made accessible?
How should citation work for model-based research, including all of its aspects (input data, model codes, processing codes, model output, metadata related to all of these)?

This session will feature multiple speakers who will present on different aspects of the challenges related to preserving and replicating simulation-based research. The first set of speakers, Gretchen Mullendore, Matt Mayernik, and Doug Schuster, will talk about the “What about Model Data?” project (https://modeldatarcn.github.io/), funded by the NSF as an EarthCube Research Coordination Network, which is focused on determining best practices for preservation and replicability of simulation experiments. Peter Fox will discuss his experiences as Editor in Chief of the AGU Earth and Space Science journal in working with authors of computational model-based papers, especially those on intercomparisons and evaluations. Mark Parsons will discuss challenges that computational models present for citation and identification of data and software. Gary Strand will discuss how these questions are dealt with for data produced by the Community Earth System Model (CESM) community.

View Recording
View Notes

Description: The challenges confronting the Earth and Space Sciences (ESS) are increasingly complex, avoiding categorization or solution within neatly defined disciplinary boxes. Transdisciplinary, or "antidisciplinary," approaches are required to address threats in ESS like climate change and space weather.  However, existing approaches to integrating data and knowledge remain crippling to progress and collaboration. Thus, now is the critical time to bring together the antidisciplinary communities to create a new paradigm of knowledge integration using knowledge graphs/networks and semantic technologies. We will collect contributions that utilize these approaches to better structure knowledge across projects. This session will emerge the success stories and best practices for knowledge graphs across ESS.

Format: We propose to highlight the 'home run stories' in the Earth and Space Sciences to create knowledge graphs/networks that have improved information flows between individuals, projects, groups, and data systems. We will first feature these use cases, then hold a discussion period to emerge new articulation, methods, and research vistas for KGs across the Earth and Space Sciences. We hope that the discussion will emerge data models that can be generally used and processes for tailoring them to new applications.
Conveners:

• Ryan McGranaghan - Principal Data Scientist and Aerospace Engineering Scientist at ASTRA LLC
• Barbara Thompson - NASA Goddard Space Flight Center
• Emily Law - NASA Jet Propulsion Laboratory
• Lewis McGibbney - NASA Jet Propulsion Laboratory

Agenda

• (4 - 4:10 PM) Presentation - opening remarks    
  • Speaker: Ryan McGranaghan
• (4:10 - 4:20 PM) Introduction to panel speakers    
  • All; Ryan McGranaghan to moderate
• (4:20 - 4:45 PM) Panel Discussion    
  • General topics (panel member):
    • Different approaches to building a KG (Deborah L. McGuinness)
    • The socio-technical perspective of KGs (Juan Sequeda)
    • Role of KGs in AI/ML (Leilani Gilpin)
    • Using KGs to link domains (Krzysztof Janowicz)
• (4:45 - 5:10 PM) Breakout sensemaking groups    
• (5:10 - 5:25 PM) ‘What, so what, now what?' Exercise to get feedback from breakouts 
  • Moderator: Ryan McGranaghan and session conveners
• (5:25 - 5:30 PM) Concluding chat   

Panelists

• Juan Sequeda - Principal Scientist at data.world
• Krzysztof Janowicz - Professor of Geographic Information Science and Geoinformatics at the Geography Department of the University of California, Santa Barbara (UCSB)
• Leilani H. Gilpin - Research Scientist at Sony AI; PhD MIT CSAIL
• Deborah L. McGuinness - Tetherless World Senior Constellation Chair and Professor of Computer and Cognitive Science at Rensselaer Polytechnic Institute (RPI); Founding director of the Web Science Research Center at RPI

View Recording
View Notes

Session Description: Artificial intelligence (AI) is a leading technology that may transform the research and applications of Earth and space science and amplify the benefits to society. One of the key bottlenecks limiting exploring/applying AI in Earth and space sciences at scale is the lack of analysis-ready data for AI applications. It's often cited that researchers can spend 80% of their project time wrangling with the large heterogeneous volumes of data that are needed for the application. To accelerate the application of AI in Earth and space sciences, it is essential to minimize users’ burden of data wrangling by providing AI-ready data with machine-readable metadata that is not only compliant with FAIR principles but also tuned for AI applications (e.g., data biases). In this session, we will convene a panel of stakeholders from data providers, users, and data engineers to discuss the path of defining the requirements and creating inter-organization matrices to assess and describe data “AI-readiness”. The outcome of this session will be summarized in a short report, and the document will be used as a guiding document to pursue further collaboration across Earth and space science data communities on this topic.

Session Agenda:

1. Session Scope (5 min)
2. Establish A Baseline (5 min)
3. Invited Presentations (Gregory Dusek, Kirstine Dale, Sophie Hou) (25 min)
4. Breakout Discussion (35 min)
5. Setting Community-Driven Priority (15 min)
6. Close & Depart (5 min)

Slides & Relevant Documents:

• Data AI-readiness matrix (presented during 2nd NOAA AI Workshop).
• Session slide deck
• All slides from speakers can be found at the bottom of this page.

Relevant Sessions:

• Analysis Ready Data in Science and Industry (01/27 11:00 am -12:30 pm)
• Best Practices and Fundamental Challenges of AI in Earth and Space Sciences (01/28 1:30 pm - 3:00 pm)
• Research Showcase (01/27 6:00 pm - 7:30 pm)

View Recording
View Notes