| Abstract |
The Great Basin is the largest area of contiguous endorheic watersheds located in the western USA and covers Nevada, much of Oregon and Utah, and portions of California, Idaho, and Wyoming. It has multiple geothermal reservoirs ranging from low- to high-temperature resources and a vast area is yet to be explored to discover hidden geothermal resources. In this study, we aim to characterize the geochemical characteristics of low-, medium-, and hot-temperature geothermal resources of the Great Basin. Geochemistry is expected to provide critical information about groundwater temperature, groundwater type, flow path, recharge areas, possible in / outflow, spatial extent of reservoirs, and hydrogeochemical interaction between water and rock of a geothermal reservoir. The geochemical data are also expected to include hidden information that is a proxy for geothermal anomalies. As a result, analyses of this proxy data with ML methods can provide insights on locations of geothermal resources. To achieve this, we processed geochemical data in the Great Basin at 14,258 locations for 18 attributes, which are temperature, quartz and chalcedony geothermometers, pH, TDS (total dissolved solids), Al3+, B+, Ba2+, Be2+, Br–, Ca2+, Cl–, HCO3–, K+, Li+, Mg2+, Na+, and ∂18O. Three datasets are generated based on following criteria: low-temperature (30–60ºC), medium-temperature (61–90ºC), and high-temperature ( more than 91ºC) resources. Finally, an unsupervised ML method called non-negative matrix factorization with customized k-means clustering (NMFk) is applied on each dataset. NMFk identifies hidden signals in the dataset that are representative of hidden geothermal resources. In this study, ML analyses define key attributes that best characterize each type of geothermal resource. Also, spatial clusters define spatial boundaries with critical attributes of each cluster. |