The identification of low-level thermal fronts is particularly challenging in high-resolution model fields over complex terrain. Firstly, direct model output often contains numerical noise which spuriously influences the highfrequency variability of thermal parameters. Secondly, the boundary layer interferes via convection and consequently leaves its thermal marks on low levels. Here, an automated objective method for the detection of frontal lines is introduced which is designed to be insusceptible to consequences of small grid spacings. To this end, existing algorithms are readopted and combined in a novel way. The overall technique subdivides into a basic detection of fronts and a supplemental division into local fronts and synoptic fronts. The fundamental parts of the detection are: (1) a smoothing of the initial fields, (2) a definition of the frontal strength, and, (3) a localisation with the thermal front parameter. The local fronts are identified by means of a classification of open and closed thermal contours. The resulting data comprise the spatial outline of the frontal structures in a binary field as well as their type and movement. The novel methodology is applied to a 3 year high-resolution reanalysis over central Europe computed with the COSMO model using a grid spacing of 7 km. Grid-point based climatologies are derived for the Alpine region. Frequencies of occurrence and characteristics of motion are analysed for different frontal types. The novel climatology also provides quantitative evidence of dynamical properties such as the retardation of cold fronts ahead of mountains and the dissolution of warm fronts over mountains.