Developing remote sensing-based early phase geothermal exploration model via machine learning techniques

One of the biggest factors causing the global climate crisis is the widespread use of fossil fuels worldwide. The most effective measure taken globally to reduce fossil fuel consumption is renewable energy usage. Geothermal energy is considered a reliable alternative energy source for its sustainability and applicability in diverse areas. Despite its major advantages, geothermal usage is limited due to high exploration costs. Conventional exploration techniques are expensive and limited, particularly in geologically difficult terrains. To handle these problems, remote sensing and machine learning techniques can be utilized to increase exploration accuracy and reduce the costs. Accordingly, this study aims to create a geothermal exploration model that is compatible to work with remotely sensed data and machine learning algorithms. Land Surface Temperature, Lineament Density, and Hydrothermal Alterations were the input parameters of the model. K-means and Random Forest were selected as the machine learning algorithms due to their success to handle big and work on complicated data series. Buharkent and Germencik geothermal production fields in Türkiye were selected as the study areas. The findings revealed that the model had promising results particularly in Buharkent field where the model accuracy is 79% and recall value is 93%. Although the Germencik field produced relatively low results, it provided valuable insights for the upcoming studies with 59% model accuracy and 83% recall value. The results of the study indicate that integrating satellite data with machine learning algorithms presents a promising approach, particularly in the early phase of geothermal exploration.