Over a floating glacier ice tongue or an ice shelf, the glacier motion measured by a single, repeat-pass, radar interferogram is difficult to analyze, because the long-term, steady motion of the ice is intermixed with its cyclic, downward motion induced by tidal forcing. Multiple interferograms and a quadruple-difference technique are necessary to separate the tidal signal from the long-term, steady motion of the ice. An example of application of this technique is given here using ERS-1 radar images of Petermann Gletscher, a major outlet glacier of northern Greenland. Tidal displacements are measured with <5 mm statistical noise. The long-term ice velocity is measured with a precision of 1 m a-1. The inferred tidal displacements agree well with model predictions from a fixed elastic beam with an elastic damping factor of 0.47 ± 0.01 km-1. The hinge line is mapped with a precision of 20-80 m. Combining the interferometric ice velocities with ice thickness data, the glacier ice discharge is calculated at and below the hinge line. At the hinge line, the ice flux is 12.1 ± 1 km3 a-1. At the ice front, calf-ice production is only 0.59 km3 a-1, meaning that 95% of the ice that crosses the grounding line melts before it reaches the calving front. Assuming steady-state conditions, the melt rate of the glacier tongue averages 12 ± 1 m a-1, with peak values exceeding 20 m a-1 near the hinge line. This high melt rate cannot be accommodated by surface ablation alone (only about 2-3 m a-1) and is attributed to pronounced basal melting of the ice tongue. Basal melting, often assumed to be negligible in Greenland, is the dominant process of mass release from the floating section of Petermann Gletscher.