Large fractions (15-40%) of the Arctic and Subarctic land surface are covered by lakes and rivers, e.g. in the Lena Delta in Northern Siberia more than 30%. The timing and duration of ice-onset is a relevant climate signal since it is strongly driven by the air temperature. Remote sensing provides a means for obtaining regional information in the high Arctic where ground-based data networks are sparse. Space-borne imaging radar is a suitable tool for the detection of freeze processes, as the backscattering of radar waves is highly dependent on the dielectric properties of the surface. As radar operates independently from sunlight and cloud cover, data retrieval is possible on a regular basis. Radar systems can monitor the earth surface at comparatively high spatial resolutions (on the order of 1 to 20 m), which allows to detect small-scale surface characteristics like ice-onset on small lakes and narrow river channels, which are often present in permafrost environments. For this study, a time series of high resolution TerraSAR-X images was analyzed. The images, recorded during fall and winter 2012, cover several lakes and river channels located in the central Lena Delta, North Siberia, Russia. Images are available every five to six days. After geocoding and multilooking, the pixel size is 13.2 m in azimuth and range direction. In 2012, ice-on is first visible in the SAR images at October 6th on the lakes and narrow river channels. The main channels of the Lena River were partly ice-free until October 12th and frozen from October, 17th on. These results were achieved via visual inspection by an experienced user. Automatic detection, however, is difficult, as the backscattering intensities are similar for calm water surfaces and thin ice without air inclusions or cracks. Thus the often used threshold-based method might miss the ice-onset. In addition, threshold-based methods can get confused with open water affected by wind, as the radar intensity scattered from a rough water surface at higher wind speeds is larger than the intensity from calm water and similar to the intensity of thicker ice. A lake or river channel with a thin ice cover might be characterized by low backscattering intensities on average, which is similar to ice-free water bodies, but can show surface features such as thin cracks, displayed by higher backscatter values. These features are clearly detectable via visible inspection. Techniques like edge detection or other gradient-based methods can help to visualize structures within the ice and ice-water boundaries. The presence of strong radar intensity gradients over lakes and rivers can be taken as a signal for the presence of ice. Later in winter, when the ice thickens and air bubbles, e.g. from methane emissions, are trapped in the ice, or the surface roughens due to break-up and refreeze, these features vanish and the backscatter rises to values close to 0 dB (HH-polarization). The main reflection of the radar waves now occurs at the ice-water boundary. As soon as the ice is frozen to the ground, the backscatter drops again. Thus it is possible to determine from SAR images whether the whole water column is frozen. In the Lena Delta, several lakes and river channels freeze to the ground, whereas others don't. Thus a map partitioning shallow water bodies from deep ones can be established on the base of space-borne SAR observations during winter.