You are here

Application of Artificial Intelligence for Science Modeling and Instrumentation

Description:

Scope Title:

Accelerating NASA Science and Engineering Through the Application of Artificial Intelligence to Data Assimilation

Scope Description:

NASA, the National Oceanic and Atmospheric Administration (NOAA), and other Federal agencies maintain extensive Earth and space observation networks and are continuously developing the next-generation remote sensing platforms. The data from these observations are used in a wide variety of ways, including as input to scientific data analysis and physics-based computer models to make a wide range of forecasts for both Earth and space weather. The current state-of-the-art simulations remain coarse enough that the physics of certain phenomena cannot be modeled at the current grid scale. Therefore, subscale or subgrid parameterizations are used as approximations to a full physics model, such as for convection, microphysics, turbulence, radiation, cloud cover, and more. These parameterizations rely on heuristic approaches that are highly time consuming, are computationally intensive, and can result in large uncertainties.

NASA is seeking proposals that apply artificial intelligence (AI), machine learning (ML), and/or deep learning (DL) in support of model parameterization to improve efficiency of the model runs and the accuracy of model forecasts. Further, NASA is highly interested in approaches to subgrid parameterizations that are physically constrained, explainable, and have well defined uncertainties. As an alternative to traditional methods of model parameterization, the application of AI/ML/DL methods have the potential to result in a more complete state of the natural system while supporting faster and better forecasts for the following:

  • Short term, to understand the risk for localized extreme events.
  • Longer term, for seasonal to subseasonal and potentially decadal predictions.
  • Retrospective reanalysis, to provide a more accurate historical record.

Proposals MUST specify and be in alignment with existing and/or future NASA/NOAA. Research proposed to this subtopic should demonstrate technical feasibility during Phase I, and in partnership with scientists and/or engineers, show a path toward a Phase II prototype demonstration, with significant communication with missions and programs to later plan a potential Phase III infusion. It is highly desirable that the proposed projects lead to solutions that will be infused into government programs and projects.

Expected TRL or TRL Range at completion of the Project: 4 to 6

Primary Technology Taxonomy:

  • Level 1 11 Software, Modeling, Simulation, and Information Processing
  • Level 2 11.2 Modeling

Desired Deliverables of Phase I and Phase II:

  • Prototype
  • Software
  • Research

Desired Deliverables Description:

Data products developed under this subtopic may be developed for broad public dissemination or used within a narrow scientific community. It is expected that the labeled training data sets, models, and any resulting data products will be publicly accessible.

 

In general, the desired outcomes for this subtopic include, but are not limited to, the following:

  • New methods, approaches, and/or applications for model parameterization that can be used and infused into NASA/NOAA simulations.
  • Labeled training data sets and trained models specifically for a given problem but that can also be used as a basis for furthering other science and engineering research and development.

More specifically,

  • Phase I should be used to establish a proof of concept with deliverables that include a final report, any software developed, training sets, etc.
  • Phase II will expand on this proof of concept to a full prototype with a very similar set of deliverables, including a final report, software, training sets, etc.

State of the Art and Critical Gaps:

NASA, along with other Federal Agencies and commercial and foreign research organizations that perform science and engineering are making large strides in the use of AI technologies (which include both ML and DL). This subtopic is looking to improve this by providing trained models that have the possibility of creating a better understanding of the state of the physical system (i.e., Earth, solar wind, etc.) to improve predictability,

 

In addition, emerging computational platforms now provide significant improvements in computing capabilities to enable AI to be applied to a wide variety of applications in science and engineering. These emerging computational capabilities have the potential to dramatically speed up AI calculations, and these systems are even being used as the reference architecture for exascale high-performance computing systems.

Relevance / Science Traceability:

This subtopic has broad applicability across the decadal surveys and satellite development requirements to improve the quality and granularity of system forecasts:

  • Improved measurements could provide better gap analysis for future mission requirements.
  • Global Modeling and Assimilation Office (GMAO): Improved model parameterizations for increased computational performance and more accurate short-term, seasonal-to-subseasonal, and retrospective forecasts.
  • Goddard Institute for Space Studies (GISS): Improved model parameterizations for increased computational performance and more accurate decadal and retrospective forecasts.
  • Carbon Cycle Ecosystems Office (CCEO): Wide variety of applications, given the diversity of data sets from sparse in situ to global satellite measurements.
  • The Community Coordinated Modeling Center (CCMC): A multi-agency partnership performing research on space science and space weather models; improve predictability of short-term forecasts.
  • Earth Science Technology Office (ESTO/AIST): New technology and services to exploit NASA and non-NASA data leading to digital twins of physical systems.
  • NOAA Joint Center for Satellite Data Assimilation (JCSDA) - Joint Effort for Data assimilation Integration (JEDI)
  • ​​​​​​​NOAA Global Forecast System (GFS
  • Computational and Information Sciences and Technology Office (CISTO - Code 606): Computational, analytic and visualization technologies used for new data science.
  • NASA Center for Climate Simulation (NCCS - Code 606.2): Building applications toward exascale computing.

References:

In addition, proposers are encouraged to search the NASA Technical Report Server (NTRS) for additional information to help guide potential solutions:

US Flag An Official Website of the United States Government