Machine Learning Estimates of Geothermal and Critical Mineral Prospectivity of the Great Basin

Abstract

Prospectivity mapping for geothermal and critical minerals is a complex undertaking reliant on integrating diverse datasets ranging from geological and geophysical surveys to geochemical analyses. Prospectivity mapping aims at the identification of areas likely to host commercially viable resources based on multiple criteria and risk factors. Subject-matter-driven methods such as traditional Play Fairway Analysis (PFA) borrowed from the oil & gas industry are commonly used. However, often they can be subjective and time-consuming. Also, the PFA’s may not account for all the intricate details hidden in the analyzed datasets. This study explores the application and comparative performance of alternative artificial intelligence (AI) and machine learning (ML) techniques for automated and objective prospectivity assessment. In our work, we have investigated the efficacy of SmartTensors’ Nonnegative Matrix Factorization (NMFk), Support Vector Machines (SVM), Graph Neural Networks (GNN), Gradient Boosting (XGBoost), Diffusion Models (DM), and Kriging Convolution Networks (KCN) in predicting geothermal potential. The analyses are based on Great Basin datasets collected under various past projects. These include the USGS/DOE’s GeoDAWN and INGENIOUS projects. We merged surface, structural geology, gravity, magnetic, heat flow, and geochemical data attributes. The performance of each model is evaluated using a series of metrics, as well as through visual inspection of the resulting prospectivity maps. We also analyzed feature importance for each model to gain insights into the key factors influencing prospectivity. The tested ML methods and tools are deployed and demonstrated on EnviCloud (https://envitrace.com/#envicloud; https://envitrace.com/saas). EnviCloud is a proprietary, comprehensive, cloud-based platform designed to optimize the entire reservoir lifecycle, from initial exploration and site assessment to real-time well monitoring and production optimization. It is developed to support Software-as-a-Service (SaaS) licensing as well as project consulting work. By leveraging cloud/high-performance computing, cloud data management, AI/ML, data analytics, and GIS, EnviCloud streamlines resource utilization and enhances decision making. The platform offers key features such as AI-powered geologic mapping, reservoir simulations, near-real-time data analytics, and risk/decision analysis tools for project feasibility, all at a centralized cloud dashboard for multi-user collaboration. It also includes tools for tracking sustainability metrics and ensuring regulatory compliance, especially related to groundwater contamination and induced seismicity. EnviCloud reduces exploration time and maximizes the potential of geothermal and critical mineral resources. It targets a broad audience, including exploration companies, energy providers, investors, regulators, and research institutions, aiming to promote sustainable geoengineering. Here, we demonstrate EnviCloud's application in processing the Great Basin datasets. Our ML analyses successfully extracted key features relevant to evaluating geothermal and critical mineral prospectivity. This research contributes to a more robust and efficient methodology for exploration, potentially reducing exploration costs and accelerating the discovery of new geothermal and critical mineral resources.