01a. “M-Inference”: A software package for rapid reference class forecasting from complex mine water geochemistry datasets

The United States is heavily reliant on imports of mineral commodities vital to national security and economic prosperity. Long lead times needed for adequate environmental review of new mining projects, which hinge on water quality predictions using scale-up of laboratory tests, is inhibiting development of domestic mineral supply. A critical gap in the existing approach for environmental review is comparison to operating mines, because no such database exists. This project supports parameterization and ground- truthing of water quality predictions for proposed mining projects through comparison to existing water quality at operational mines. This project has developed a software for reference class forecasting based on a proprietary water-chemistry database of real-world operational mine-water geochemistry coupled with geochemical speciation and statistical tools. Reference class forecasting is a predictive technique for complex systems based on past performance in statistically similar reference classes. The approach relies on large data sets and carefully curated metadata. In Phase I, database development focused on mine sites around the world targeting a specific commodity: nickel sulfide deposits. Metadata describing mine setting and chemistry data were organized to facilitate reference class identification. Software was developed to (i) perform statistical and visual analysis of the data, (ii) feed the data into a geochemical forecasting package, and (iii) deliver data and analysis results to the end-user through a web-dashboard. The goal of Phase I to produce a minimum viable product has been achieved. In Phase II, the database will be expanded from 10 mines to include a total of 200 mines, targeting deposits with minerals critical for low carbon technologies. This scale is needed for effective reference class forecasting across a range of mineral deposit types. Software tools for data visualization, analysis, and user interaction will be tested and refined. Phase II will also include a demonstration project to help communicate the benefits of this approach to potential customers. The predictive water quality models that currently form the basis for mine engineering design and environmental review are costly, time-intensive and, thus, hinder permitting and development of new domestic mines. The overall value lost from permitting delays on proposed mining projects is multi-billion dollars in magnitude; this has direct impact on the financial prospects on proposed mining projects, go/no- go decisions on new mines, and therefore our domestic supply of metals. This SBIR project will fill a critical and currently missing information-gap by providing credible water quality information from operational mines to stakeholders of mine development, including mining companies, investors, regulators, and the public.