ROSES 2025: A.9 User-Centered Applications with Large Earth Foundation Models

The National Aeronautics and Space Administration, through its Science Mission Directorate, has released the Research Opportunities in Space and Earth Sciences 2025 program element A.9, titled User-Centered Applications with Large Earth Foundation Models. This program is part of NASA’s long-standing effort to promote the integration of advanced Earth science into decision-making tools that benefit society and align with the Earth Science to Action strategy. The opportunity specifically leverages NASA and IBM’s Prithvi models, which are open-source, large Earth foundation models developed for Earth observation and climate applications. These foundation models are trained on vast datasets such as the Harmonized Landsat and Sentinel-2 and MERRA-2, enabling more efficient development of applications with reduced computational demands compared to traditional machine learning approaches. The purpose of this grant is to fund the creation of decision-support applications that directly incorporate end-user needs into the development process. Applicants are required to build applications that advance one of the Prithvi models while obtaining regular feedback from end users and demonstrating the ability to scale horizontally to more users or vertically to increase efficiency. End users must be included as part of the project team, and teams are encouraged to include data scientists, Earth scientists, and social scientists to ensure multidisciplinary expertise. Applications must begin at Application Readiness Level 3 or higher and are expected to advance at least one level each year. Projects are generally expected to last up to two years but may run for as long as three years if justified. The scope of funded work includes development of tools that support economic and societal sectors such as insurance, real estate, infrastructure, and transportation. Projects must demonstrate practical improvements in decision-making and quantifiable enhancements in time savings, computational efficiency, or scalability. Applicants must also provide a clear transition plan to ensure that the developed tools are sustainable and remain in use by the identified end users after the period of NASA funding. Proposals that do not include end users, lack a credible pathway to readiness milestones, or propose new foundation models instead of building on existing Prithvi models will be deemed nonresponsive. Eligibility for this program is broad, including U.S. academic institutions, nonprofits, private sector organizations, and partnerships with international institutions, although non-U.S. organizations are limited to participation under a no-exchange-of-funds basis. Federal agencies other than NASA and federally funded research and development centers, with the exception of JPL, are not eligible to receive funds. NASA-badged members of the IMPACT team currently funded on Prithvi-related work cannot participate as funded or unfunded collaborators, though the IMPACT team will be available as a resource. All proposals must comply with dual-anonymous peer review requirements, anonymizing the scientific and technical content while providing separate non-anonymized documentation of expertise and resources. The application process requires a two-step submission. Step-1 proposals, which serve as pre-proposals, are due on 2025-11-04. Step-2 full proposals are due on 2025-12-09, and only applicants who submitted Step-1 proposals are eligible to proceed. The estimated start date for funded projects is eight months following release of the solicitation. Proposals are submitted electronically via NSPIRES, with Grants.gov submission available only if requested at least 30 days in advance. Proposals must adhere to strict formatting rules, including a 15-page limit for the scientific and technical section, and must include an Open Science and Data Management Plan detailing how outputs will be openly shared. Applicants are encouraged to budget for participation in kickoff and cohort meetings, end-user engagement, scientific conferences, and computational costs, including $0.09 per service billing unit when using NASA’s high-end computing resources. The program anticipates funding up to eight new awards, with a total program budget of up to $2 million annually for new projects. Typical award durations are one to three years, and budgets are encouraged to remain below $250,000 per year per project. Reporting requirements include quarterly updates, annual progress reports, and a final performance report within 120 days of the project’s conclusion. Key points of contact include Argyro Kavvada (argyro.kavvada@nasa.gov, 202-213-7891), Tsengdar Lee (tsengdar.j.lee@nasa.gov, 202-251-9863), and Rahul Ramachandran (rahul.ramachandran@nasa.gov) for technical questions on the Prithvi models. Additional resources, FAQs, and details will be available on the NSPIRES program page, where a pre-proposal webinar will also be announced.