Roughly coincident with last week’s announcement of a planned tripling of its compute capacity, the National Oceanic and Atmospheric Administration issued an ambitious AI Strategy to transform NOAA’s use of AI technology.

“The goals and supporting objectives in this strategy are intended to directly improve the understanding, coordination, awareness, and application of AI across all of NOAA. By directing organizational and process improvements to more rapidly transition research results to improve services, strengthen and expand partnerships, and deliberately develop AI proficiency across the NOAA workforce, the guidance below will result in transformational advances in every NOAA mission area,” according to the newly-released document.

NOAA says an implementation plan with “detailed action items, deadlines, and responsibilities” is forthcoming.

Bringing such strategies to fruition is always the challenge. The just announced plans for acquiring two 12 petaflops supercomputers from Cray (now part of HP Enterprise) will certainly help; they represent a refresh of existing resources and are expected to be completed in 2022. Together with existing research and development systems (located at NOAA sites in West Virginia, Tennessee, Mississippi, and Colorado), NOAA’s computing power will reach an aggregate 40 peak petaflops. (See HPCwire coverage for further details.)

There are many areas in which NOAA could leverage AI technologies. Brian Gross, director, Environmental Modelling Center, National Weather Service (NWS), told HPCwire, “There’s a couple of areas that we’re considering the use of AI and one of them is on data thinning. There’s an awful lot of observational data out there that we could ingest. We can rely on machine learning to winnow that down so that we’re really assimilating the key observations that will make a difference in the quality of the forecast. In a similar vein, we can select particular physics schemes [for modeling] based on the phenomena of interest and having that having automated is another area of exploration that we’re undertaking.”

The new AI strategy lays out five goals, excerpted here:

• Goal 1: Establish an efficient organizational structure and processes to advance AI across NOAA.When achieved, this goal will enhance cross-line office coordination that has been so far limited in AI development, awareness, and application. This, along with prioritization of AI in budget formulation, strategic communications, and cloud computing applications will exponentially increase the use and utility across all of NOAA.
• Goal 2: Advance AI research and innovation in support of NOAA’s mission.NOAA will advance AI-based environmental research and innovation across every mission area by adopting and institutionalizing AI throughout NOAA’s research processes. The NOAA Research and Development Database (NRDD) will be a critical asset to help track NOAA research involving AI, with the intent to see the application of AI research projects continuously increase.
• Goal 3: Accelerate the transition of AI research to applications.NOAA will rapidly accelerate the transition of AI-based environmental research to NOAA operations (R2O) and private-sector commercialization (R2C – together R2X). Because many requirements for NOAA’s AI applications are available in private-sector and academic research environments, the majority of our AI research involves applying these applications to our mission where applicable. Coordination of AI initiatives and activities mentioned in Goal 1, and examples from the Earth Prediction Innovation Center (EPIC), will play leading roles in transitioning R2O. For successful transitions, we will support the growth of a nascent commercial environmental AI sector that will increasingly serve as a source for solutions to our operational requirements. OAR’s TPO will lead the transition of NOAA tailored R2C.
• Goal 4: Strengthen and expand AI partnerships.Cooperative partnerships serve as force multipliers to optimize resources and effort, and the scientific and technological exchange keeps NOAA current in the rapidly evolving field of AI. Partnerships in AI-based environmental applications are already creating a community of practice that is sparking innovation and has the opportunity to accelerate tremendous advances in NOAA’s capabilities.
• Goal 5: Promote AI proficiency in the workforce.Where appropriate, NOAA will provide resources to equip our workforce to fully leverage the rapid evolving field of AI. This can only be achieved by providing continuous, current, creative, and tailored training and learning opportunities. NOAA’s existing development programs are well suited to be adapted for these, and we will look to partners for new options to develop skill, understanding, and expertise.

NOAA’s push to expand AI use is noteworthy on at least two fronts. It is in keeping with the U.S. Administration’s broader Executive Order supporting AI use and development. Yet the new strategy comes at a time when the Administration is proposing budget cuts (FY21) in science spending. NOAA, for example, faces 14% cut overall its Office of Oceanic and Atmospheric Research faces a roughly 40% cut in the proposed budget. As is typical, no one expects the proposed budget to pass as is and wrangling with Congress has already begun. (See HPCwire article, Trump Budget Proposal Again Slashes Science Spending)

AI use is certainly not new at NOAA which offered the following examples of ongoing and prior AI use: 1) aerial and underwater surveys from ships and autonomous platforms to assess the abundance of marine mammal and fish populations, (2) robotics for deep-sea exploration, (3) quality control of weather or satellite observations, (4) improving physical parameterization for weather, ocean, ice modeling, and improving the computational performance of numerical models, (5) aiding weather warning generation, (6) operation of unmanned systems for bathymetric mapping, habitat characterization, hydrologic, oceanographic, atmospheric, fishery, ecosystem, and geographic surveys, (7) supporting partners in wildfire detection and movement, and (8) using machine learning (ML) for reliable and efficient processing, interpretation, and utilization of earth observations.

“Despite this notable progress,” reported NOAA in the strategy document, “the true potential for AI to advance NOAA’s mission has not been realized because all NOAA AI activity heretofore has originated within individual offices with no institutional support.”

Stay tuned for the implementation roadmap.

Link to NOAA AI Strategy: https://nrc.noaa.gov/LinkClick.aspx?fileticket=0I2p2-Gu3rA%3d&tabid=91&portalid=0