Stratiform rain situations are generally associated with the presence of a melting layer characterized by a strong signature in polarimetric radar variables. This layer is an important feature as it indicates the transition from solid to liquid precipitation. The melting layer remains poorly characterized, particularly from a polarimetric radar point of view. In this work a new algorithm to automatically detect the melting layer on polarimetric RHI radar scans using gradients of reflectivity and copolar correlation is first proposed. The algorithm was applied to high-resolution X-band polarimetric radar data and validated by comparing the height of the detected layer with freezing-level heights obtained from radiosoundings and was shown to give both small errors and bias. The algorithm was then used on a large selection of precipitation events (more than 4000 RHI scans) from different seasons and climatic regions (South of France, Swiss Alps and plateau, and Iowa, USA) to characterize the geometric and polarimetric signatures of the melting layer. The melting layer is shown to have a very similar geometry on average, independent of the topography and climatic conditions. Variations in the thickness of the melting layer during and between precipitation events was shown to be strongly related to the presence of rimed particles, to the vertical velocity of hydrometeors and to the intensity of the bright band.