There is increasing emphasis put on the correlation of marine predator behaviour and foraging performance with the bio-physical properties of the ocean environment. However, spatial error in the estimated position of animals and the accuracy of interpolated, physical oceanographic data require the assessment of the appropriate spatial resolution at which to assess relationships. We recorded surface temperature data from 17 archival tags attached to female southern elephant seals at Macquarie Island during the post-lactation foraging trip of 1999–2000. Archival-tag temperature data were associated with twice-daily, at-sea positions derived from light levels (i.e., “geo-location”). We compared these surface temperatures and their associated spatial error to satellite-derived Multi-Channel Sea Surface Temperature (MCSST) data to assess at what spatial scale the agreement between the two data sources was highest. We considered scales from 50×50 km through to 500×500 km grid cells, at 50×50 km increments. Averaged over all individuals and assessed in fortnightly time periods, we found a peak in agreement between the mean surface temperature recorded by the archival tags and the MCSST data at 350×350 km grid cells (122 500 km 2 ). We used logistic regression model selection to examine the effects of spatial scale, time period, latitude, longitude, and MCSST variance per grid cell on the agreement between the data sources. The most parsimonious model included all main effects, and a significant interaction between time period and latitude. There was also a trend for reduced agreement at higher latitudes, and there was a peak in temperature agreement in the general region around 180° longitude. There were two peaks in agreement during the early and latter parts of the post-lactation foraging trip, resulting probably from the more northerly latitudes travelled during these times and the improved MCSST coverage in the summer months. The 350×350 km scale recommended by this study should be appropriate for the study of coarse-scale associations between marine predators located using geo-location and physical oceanographic data during summer. However, the approach remains valid for predators located at sea using other positioning methods.