Potential geothermal areas are identified through investigation of spatial relations between geothermal occurrences and their surrounding geological phenomena in western Anatolia, Turkey. The identification is based on only publicly available data. It is expected that the study will guide further preliminary investigations performed for large areas having limited information. Magnetic anomaly, Bouger gravity anomaly, earthquake epicenter and lineament datasets are used for the analysis. The first is used without any modification whereas the rest are utilized to extract three evidence maps: distance to major grabens, Gutenberg-Richter b-value and distance to lineaments, respectively. Predictor maps are produced from these evidence maps as well as from the unprocessed magnetic anomaly map by applying two different binarization procedures. From each binarization procedure a favorability map is produced separately using Index Overlay (IO) and Weights of Evidence (WofE) methods. The findings reveal that weighting predictor maps according to spatial association between evidence maps and training points lead to more accurate prediction in both WofE and IO methods. The potential areas in the final maps are Aydin, Denizli, Manisa, Balikesir and Kutahya of which first two have been explored and exploited, and thus found to be favorable, while the rest are nearly unexplored. (C) 2009 Elsevier Ltd. All rights reserved.