Abstract: Under the Energy Act of 2020, the U.S. Geological Survey is tasked with completing assessments for five geothermal resource types: conventional geothermal, enhanced geothermal systems (EGS), low-temperature heating and cooling resources, underground thermal energy storage, and the potential for co-production of conventional hydrothermal with critical minerals. The mathematical tools employed for these assessments range from data-driven methods (e.g., Machine Learning) to process-based physically motivated models, with methods selected to capitalize upon what is known from past study. The goal is to make a best estimate of the resource along with estimates of uncertainty, typically using Bayesian paradigms. The reality of data often conflicts with the idealized assumptions of the underlying mathematical strategies, and understanding the math helps the practitioner understand and address bias (e.g., the use of regularization). As methods become increasingly complex and apparently “black box” in nature, understanding of the underlying mathematics becomes increasingly important. This presentation seeks to highlight healthy synergies between naïve application of data-driven methods and best professional judgment.

Bio: Erick Burns is a Research Hydrologist at the USGS Geology, Minerals, Energy, and Geophysics Science Center. He specializes in the development of methods and tools for analysis and simulation of groundwater and heat flow in the subsurface, particularly in the volcanogenic terranes of California, Idaho, Oregon, and Washington. In 2004 Erick received PhD from Bioresource Engineering and MS in Mathematics from OSU.