We demonstrate the use of pop-off satellite archival tags (PSAT)-derived geolocations to determine the most probable tracks (MPTs) of olive ridley sea turtles Lepidochelys olivacea off Costa Rica. We use a Kalman filter state-space model (KFSST) that uses light-based longitude and latitude and sea surface temperatures (SST). PSATs placed on 14 turtles remained fixed for an average 53 d (range: 29 to 111 d). The average reduction in longitude and latitude standard deviations was ϕlon = 0.62 and ϕlat = 0.28 between the raw and KFSST-derived MPTs, respectively. Geolocations were linked in time to oceanographic features such as SST and chlorophyll a, as reported by satellite-based sensors. Turtles went in all directions from their respective release points, independent of year and capture type (longline-caught vs. hand-caught). Turtles remained within a SST range between 23.3 and 30.5°C (mean = 27.1°C), with over 75% of all recorded temperatures between 25.0 and 28.0°C. Turtle locations were associated with mean chlorophyll a = 0.37 mg m -3 . MPT data suggest that tur- tles spent a disproportionate amount of time in the general region of the Costa Rica Dome, a nutrient- rich quasi-permanent cyclonic eddy. Taken together, these findings support the increased utility of filtered light-based geolocation data in identifying environmental features characteristic of sea tur- tles' preferred habitat, information which can be useful in managing regional fisheries.