An imaging radar polarimeter measures the dependence of radar backscatter intensity and relative phase as a function of both the transmitted and received wave polarization states. We have built such an instrument by adding a dual-polarized antenna and a four-channel data system to a conventional airborne synthetic aperture radar system. We measure directly the amplitude and phase of all elements of the scattering matrix corresponding to each individual pixel in a radar image. Subsequent data processing combines these elements to synthesize any desired combination of transmit and receive antenna polarizations, and thus we can measure the variation of intensity or phase with polraization. Different scattering models predict different functional dependences of intensity on polarizaton: observation of this dependence for actual targets permits identification of the dominant scattering mechanisms contributing to the measured backscatter. For example, we find that Bragg scattering closely approximates both the observed scattering from the ocean and, suprisingly, from lava fields in Idaho. Urban areas can be modeled as two-bounce dielectric corner reflectors. Scatter from forests appears to have two main components: one is a two-bounce corner reflector term and the other is a diffuse component that we attribute to multiple scatter. Knowledge of the polarization signature of an object permits the selection of optimum radar polarizations to enhance certain classes of terrain. For example, we can choose a set of polarization states to optimize contrast in an image. We have applied this technique to mapping and differentiation to lava flows and to differentiation of forested and clear-cut areas. ¿ American Geophysical Union 1986