Vincent Bouchot

Structural, lithological, and geodynamic controls on geothermal activity in the Menderes geothermal Province (Western Anatolia, Turkey)

Western Turkey belongs to the regions with the highest geothermal potential in the world, resulting in significant electricity production from geothermal resources located predominantly in the Menderes Massif. Although geothermal exploitation is increasingly ongoing, geological, and physical processes leading to the emplacement of geothermal reservoirs are hitherto poorly understood. Several studies on the Menderes Massif led to different interpretations of structural controls on the location of hot springs and of the heat source origin. This paper describes geological evidence showing how heat is transmitted from the abnormally hot mantle to the geothermal reservoirs. On the basis of field studies, we suggest that crustal-scale low-angle normal faults convey hot fluids to the surface and represent the first-order control on geothermal systems. At the basin scale, connected on low-angle normal faults, kilometric high-angle transfer faults are characterized by dilational jogs, where fluids may be strongly focused. In addition, favourable lithologies in the basement (e.g., karstic marble) could play a critical role in the localization of geothermal reservoirs. Finally, a compilation of geochemical data at the scale of the Menderes Massif suggests an important role of the large mantle thermal anomaly, which is related to the Hellenic subduction. Heat from shallow asthenospheric mantle is suggested to be conveyed toward the surface by fluid circulation through the low-angle faults. Hence, geothermal activity in the Menderes Massif is not of magmatic origin but rather associated with active extensional tectonics related to the Aegean slab dynamics (i.e., slab retreat and tearing).

Feature Selection Using Second Order Derivative Pruning Methods: A Case Study in GIS Andes

Abstract In this paper, we show how neural networks can be applied to Geographic Information System (GIS) for feature selection. Precisely, we want to select the most relevant variables in a classification task between gold deposits and deposits without gold by applying pruning neural methods to a GIS Andes dataset. Two families of pruning methods based on an analysis of the weight saliencies are presented, their strengths and weaknesses are discussed. This work could help in the long term to better understand the formation of gold deposits. Copyright