Identification of blind geothermal resources in Surprise Valley, CA, using publicly available groundwater well water quality data

Abstract

Geothermal resource exploration is generally limited to areas with surface expressions of thermal activity (fumaroles and hot springs), or relies on expensive geophysical exploration techniques. In this study, the hidden subsurface distribution of geothermal fluids has been identified using a free and publicly available water quality dataset from agricultural and domestic water wells in Surprise Valley, northeastern California. Thermally evolved waters in Surprise Valley have element ratios that vary in response to Ca carbonate and Mg silicate mineral precipitation, and have elevated total dissolved solids (TDS). The arid climate in Surprise Valley leads to surface water evaporation in a closed basin, producing high TDS Na-Cl-CO3-SO4 brines in three ephemeral alkali lakes and in shallow groundwater under elevated soil CO2 conditions. Evaporated fluids in Surprise Valley follow a chemical divide that leads to Ca carbonate and Mg silicate mineral precipitation. Plots of dissolved element ratios can be used to distinguish groundwater affected by evaporation from water affected by thermal water-rock interaction, however it is challenging to select components for plotting that best illustrate different fluid evolution mechanisms. Here, we use a principal component analysis of centered log-ratio transformed data, coupled with geochemical models of fluid evaporation and thermal mixing pathways, to identify components to plot that distinguish between groundwater samples influenced by evaporation from those influenced by thermal processes. We find that groundwater samples with a thermal signature come from wells that define a coherent, linear geographical distribution that closely matches the location of known and inferred faults. Modification of the general approach employed here provides promise for identifying blind geothermal resources in other locations, by applying low-cost geochemical modeling and statistical techniques to areas where large groundwater quality geochemical datasets are available. Figure 1: Map of Surprise Valley showing locations of creek sampling stations, groundwater wells, alkali lake sampling stations and thermal springs.

GIS model for geothermal advantageous target selection

“…Then, the eigenvalues of the principal components of each indicator are obtained, and the weight of each indicator is obtained. In order to ensure that the selected indicators can scientifically and concisely reflect the advantageous areas of geothermal resources 34,35 . In this paper, the statistical software SPSS is used to carry out principal component analysis to calculate the weights of various indicators, and to determine the evaluation model of geothermal resource advantageous areas.…”

Target selection for mid-high temperature geothermal exploration based on AHP and PCA: A case study of Jiangxi Province, China

“…The PCA can screen out the main independent comprehensive factors from many variables (Yasemin et al, 2019), which can scientifically and concisely reflect the areas with geothermal resources advantage (Otero et al, 2005;Fowler et al, 2017). The statistical software SPSS is used for PCA and calculation.…”

Cluster Analysis as a Tool for Evaluating the Exploration Potential of Known Geothermal Resource Areas

“…Aitchison and Greenacre (2002) specifically cite its use in analyzing whole-rock oxide compositions. In hydrology and hydrogeology it is commonly applied to aqueous-phase geochemical variables (e.g., Otero et al, 2005;Fowler et al, 2017), as in this evaluation.…”

“…A set of compositional data form a finite-dimensional vector space, that provides a consistent definition of distance, norm, and scalar product, and satisfies the geometric requirements of a real space (Otero et al, 2005). However, compositional data are not truly independent, because a reduction in the fraction of one component results in a rescaling of all the other components (Fowler et al, 2017). On the other hand, non-compositional data form an open set, where the values of the individual components can vary independently.…”