[M-GI34] Data-driven geosciences

[MGI34-02] Application of a Bayesian algorithm to characterize deep temperatures based on shallow temperature profiles of the Berlin geothermal field, El Salvador.

*Osmany Rene Aparicio¹, Kazuya Ishitsuka², Noriyoshi Tsuchiya³ (1.M1 Tohoku University, 2.Kyoto University, 3.Tohoku University)

Machine learning approach has been used to estimate temperatures in deep unexplored regions in geothermal systems, like the Kakkonda geothermal study in 2021.This data driven methodology can become an important to assess potential resources tool. The estimation of deep temperatures in El Salvador’s Berlin geothermal field it has been implemented through a Bayesian inference algorithm approach. This algorithm correlates observed resistivity geophysical exploration data, measured stabilized temperature borehole data, and different reservoir parameters, like hydrostatic pressure, rock porosity and salinity, as shown in figure 1. The Berlin geothermal reservoir system is a basaltic-andesitic fragmented rocks, approximately at more than 1.5 km depth or equivalent -800 meters below sea level. It is a saline liquid dominated system with salinity values up to 12,000 ppm of Chloride (Cl) and temperature ranges from 260 °C up to 300 °C. It is characterized by resistivities greater than 30 Ω m as well as mean rock porosities averaged at 1 %. The adjustment between the inferred temperatures results with temperature measured in borehole data is the tune up task at an approximately 5 km² area, and therefore establish possible deep high temperatures areas for future exploration purposes.