The NASA scatterometer (NSCAT) is a spaceborne radar sensor designed to measure the normalized radar backscattering coefficient &sgr;0 of the Earth's surface. Over the ocean, backscatter measurements are used to infer surface wind vectors. Wind retrieval is based on a statistical relationship between short-ocean wave roughness (that causes the backscatter) and the surface wind speed and direction. For NSCAT geometry, multiple antennas are used to provide backscatter measurements at several azimuth directions to resolve wind direction ambiguities. To achieve the desired wind vector accuracy, these antenna beams must be calibrated within a few tenths of a decibel. A simple relative-calibration method is applied to the NASA scatterometer backscatter from homogenous, isotropic, large-area targets. These targets exhibit both azimuth and time invariant radar response. A simple polynomial model for incidence angle dependence of &sgr;0 is used, and the mean radar response from all antenna beams is taken as the reference. Corrections (&sgr;0 biases) are calculated as differences (in log space) between measurements from particular beam and the reference. This simple model is applied to data from the Amazon rain forest and the Siberian plain. These areas are tested for temporal stability within the calibration period (several weeks). High-resolution masks are applied to extract suitable calibration data sets. Calculated corrections for each antenna beam are added to NSCAT &sgr;0 measurements as a function of incidence angle. The magnitudes of corrections show the necessity of on-orbit calibration. ¿ 1999 American Geophysical Union