Satellite, airborne, and marine geophysical surveys usually sample features better along-track than across-track. In maps, this aliasing together with errors in along-track data processing produce a systematic washborad effect across the survey lines known as track-line noise. In regions covered by two line surveys at different azimuths, the combined data generate maps with this noise orthogonal to both survey azimuths. Statistical efforts to minimize random noise will only be partially effective in attenuating systematic track-line noise. We present a method that exploits the spectral differences resulting from the different survey azimuths to reduce this systematic noise. In each survey, only two of the spectral quadrants are normally contaminated. Combining the four least contaminated quadrants from both surveys generates a map of minimal track-line noise. This method is used to improve the recovery of geoid undulations from satellite altimetry of the Barents Sea in the Arctic. ¿ 1998 American Geophysical Union