Movement and location data collected via satellite-linked telemetry tags are often used to inform spatial conservation measures for threatened marine populations. Most applied telemetry studies aim to reconstruct the continuous utilization distribution underlying reported locations to characterize the relative intensity of space use. However, commonly applied space use estimators do not directly estimate the underlying distribution of interest and, perhaps more importantly, ignore correlations in space and time that may bias estimates. Here we describe how geostatistical mixed effects models, which explicitly account for spatial and/or temporal correlation using Gaussian random fields, can be applied to estimate utilization distributions from satellite telemetry data. We use simulation testing to compare the performance of the proposed models with several conventional space use estimators. Our results suggest that geostatistical mixed effects models outperform conventional estimators when the number of tag transmissions changes over time, a common source of bias in satellite telemetry studies that is rarely addressed. We illustrate this approach via application to satellite telemetry location observations collected from 271 large juvenile and adult loggerhead sea turtles in the western North Atlantic from 2004 to 2016. We demonstrate how such models can be used to predict the overall spatial distribution of tagged individuals, as well as seasonal shifts in densities at smaller time scales. For tagged loggerheads, overall predicted densities were greatest in the shelf waters along the US Atlantic coast from Florida to North Carolina, but monthly predictions highlight the importance of summer foraging habitat in the Mid-Atlantic Bight.