Global-scale topographic data are of fundamental importance to many Earth science studies, and obtaining these data is a priority for the Earth science community. Several groups have considered the requirements for such a data set, and a consensus assessment is that many critical studies would be enabled by the availability of a digital global topographic model with accuracies of 2 and 30 m in the vertical and horizontal directions, respectively. Radar interferometric techniques have been used to produce digital elevation models at these accuracies and are technologically feasible as the centerpiece of a spaceborne satellite mission designed to map the world's land masses, which we denote TOPSAT. A radar interferometer is formed by combining the radar echoes received at a pair of antennas displaced across-track, and specialized data processing results in the elevation data. Two alternative implementations, one using a 2 cm-/spl lambda/ radar, and one using a 24 cm-/spl lambda/ radar, are technologically feasible. The former requires an interferometer baseline length of about 15 m to achieve the required accuracy, and this could be built on a single spacecraft with a long extendible boom. The latter necessitates a kilometers long baseline, and would thus be best implemented using two spacecraft flying in formation. Measurement errors are dominated by phase noise, due largely to signal-to-noise ratio considerations, and attitude errors in determining the baseline orientation. For the 2-m accuracy required by TOPSAT, the orientation must be known to 1 arc-second. For the single-spacecraft approach, where attitude would be determined by star tracking systems, this performance is just beyond the several arc-second range of existing instruments. For the dual-spacecraft systems, though, differential global positioning satellite measurements possess sufficient accuracy. Studies indicate that similar performance can be realized with either system.