Light-based archival tags are increasingly being used on free-ranging marine vertebrates to study their movements using geolocation estimates. These methods use algorithms that incorporate threshold light techniques to determine longitude and latitude. More recently, researchers have begun using sea surface temperature (SST) to determine latitude in temperate regions. The accuracy and application of these algorithms have not been validated on free-ranging birds. Errors in both geolocation methods were quantified by double-tagging Laysan ( Phoebastria immutabilis Rothschild) and black-footed ( P. nigripes Audubon) albatrosses with both leg-mounted archival tags that measured SST and ambient light, and satellite transmitters. Laysan albatrosses were captured and released from breeding colonies on Tern Island, northwestern Hawaiian Islands (23°52′N, 166°17′W) and Guadalupe Island, Mexico (28°31′N, 118°10′W) and black-footed albatrosses from Tern Island. Studies were carried out between December 2002 and March 2003. For all birds combined, the mean ± SD great circle (GC) distance between light-based locations and satellite-derived locations was 400±298 km ( n=131). Errors in geolocation positions were reduced to 202±171 km ( n=154) when light-based longitude and SST-based latitude (i.e. SST/light) were used to establish locations. The SST/light method produced comparable results for two Laysan albatross populations that traveled within distinctly different oceanic regions (open ocean vs more coastal) whereas light-based methods produced greater errors in the coastal population. Archival tags deployed on black-footed albatrosses returned a significantly higher proportion of lower-quality locations, which was attributed to interference of the light sensor on the tag. Overall, the results demonstrate that combining measures of light-based longitude and SST-based latitude significantly reduces the error in location estimates for albatrosses and can provide valid latitude estimates during the equinoxes, when light-based latitude measurements are indeterminate.