DARPA, USGS Accelerate Critical Mineral Mapping with AI-Powered Workflow
RESTON, Va., Jan. 16, 2025 -- Geologists and innovators from the U.S. Geological Survey, the Defense Advanced Research Projects Agency (DARPA), the Advanced Research Projects Agency-Energy (ARPA-E), and other partners came together Jan. 13-17 to collaborate, train, and transition artificial intelligence (AI) tools to streamline mineral resource assessment workflows.
CriticalMAAS stakeholders gather at USGS in Reston for the effort’s final hackathon. Credit: DARPA.
The Critical Mineral Assessment with AI Support (CriticalMAAS) project set out to develop machine-learning tools to accelerate time-consuming parts of USGS data interpretation and critical mineral assessments. The collaborative five-day “hackathon” workshop served as a launch pad of progress for researchers and USGS users.
Graham Lederer, a USGS research geologist and lead for the collaboration with DARPA, explained that the current process for mineral resource assessments relies heavily on staff intensive data compilation and analysis.
“Typical mineral resource assessment will take us two years, start to finish, and that’s just for one deposit type, which may contain one mineral commodity in one area of the country,” said Lederer. “To assess 50 commodities across 100 deposit types throughout the entire United States would take many years to complete.”
The challenge now becomes to augment and accelerate the assessment timeline from years to days.
Since February 2024, CriticalMAAS has conducted a dozen pilot critical mineral assessments, and the results have been promising, said Erica Briscoe, DARPA Information Innovation Office program manager. Hackathons have demonstrated the AI tools’ ability to reduce the critical mineral assessment workflow to two and half days, start to finish.
A hackathon held earlier in 2024 reproduced assessments covering national-scale assessments of zinc, copper, and nickel. Another hackathon that took place at the end of 2024 pivoted to conduct regional assessments of regional Mississippi Valley Type zinc, magmatic cobalt and nickel in the upper midwest, lacustrine lithium, tungsten skarn in Alaska, national peralkaline and carbonatite rare earth elements, and regional and national porphyry copper.
The CriticalMAAS effort has four technical areas: extracting geospatial data from maps and documents; model extraction from knowledge; mineral potential mapping exploiting multi-modal fusion; and human-in-the-loop learning and mixed-initiative learning.
“So far, we’ve automated the process for pulling data from maps and documents and then compiling the data together and making it analysis-ready,” said Briscoe. “The process is already moving much faster than expected.”
AI-generated assessments would still need to pass the USGS’s scientific standards and peer review, where the fourth technical area comes into play.
“Over time, we’ve built more precise and accurate extraction tools because of having human-in-the-loop,” said Briscoe. “This human element also helps create better training data and customizable tools for further specific applications.”
At the conclusion of the effort, Lederer says the goal is for USGS scientists to be able to take a usable product of the program and begin implementing it operationally.
The timing of these tools could not be better, as the USGS Earth Mapping Resources Initiative (Earth MRI) is delivering an unprecedented volume of foundational data crucial to identifying resources of critical and industrial minerals, as well as other earth science applications.
Earth MRI is a partnership between the USGS and state geological surveys that is revolutionizing our understanding of the nation’s geology and critical mineral resources, which are vital to the U.S. economy and national security.
“These investments, both to Earth MRI and to CriticalMAAS, are setting the groundwork for future generations of earth scientists,” said Lederer.
CriticalMAAS was inspired in part by the Bipartisan Infrastructure Law. Since 2021, BIL has advanced scientific innovation through a $320 million investment in the USGS-Earth MRI to better map the nation’s mineral resources. Through the end of fiscal year 2024, more than $198 million has been obligated for Earth MRI initiatives, propelling efforts to make once-in-a-generation advancements in the nation’s geologic and geophysical data collections and mapping.
Source: DARPA
