Mean annual soil temperature has important implications for crops as well as for soil classification and formation. Landsat 7 Enhanced Thematic Mapper Plus (ETM+) band-6 was analysed to determine its relationship with mean annual soil temperature (MAST) at 50 cm in the Transylvanian Plain, Romania. Band-6 is available in both high and low gain formats from the United States Geological Survey; for our study only high gain was evaluated because of the increased resolution that it provides. Both of the gain levels of band-6 are measured at 10.4-12.5 μm (thermal infrared), at 60-m spatial resolution. Four different months of Landsat 7 ETM+ data were used to predict MAST and compared with 50-cm soil temperature data measured on-site with in situ sensors and data logging stations. Despite no correction for land cover differences across the plain, strong relationships were found between the Landsat-predicted and field measured MAST with a coefficient of determination (R 2) for July, August, December and February of 0.63. Multiple regression analysis (MAST Regression) provided a weaker relationship, when compared with MAST in situ, with a coefficient of determination (R 2) of 0.42. Significant differences existed between urban and agricultural land covers, as identified by Coordination of Information on the Environment (CORINE) data. The use of Landsat 7 ETM+ could reduce the time and expense of large field studies for determining MAST. These data could then be used for temperature models of entire regions, for a range of land management options.