| GPS units used in animal telemetry suffer from non-random data loss and location error. GPS units use stored satellite information to estimate locations, including almanac and ephemeris data. Using the smallest GPS collars (45-51g) available for mammals, we evaluated how satellite information and environmental conditions affected GPS performance in 27 mobile trials and field reliability during 56 deployments on Pacific marten (Martes caurina). We programmed GPS units to retain or remove satellite data (i.e., continuous or cold start) before attempting a location (fix), thereby mimicking differing fix intervals. In continuous mode, fix success was 2.2 times higher, was not influenced by environmental obstructions, and improved after a location with satellites (3-D). In cold start mode, fix success was negatively correlated with vegetation cover. Location error was lower for 3-D fixes. Censoring cold start fixes with only 3 satellites (2-D) and 2-D locations prior to the first 3-D fix in continuous mode decreased location error by 91% and 55%, ensuring <50m accuracy. The significance of previous fix success and reduced battery expenditure underscores the benefits of ephemeris data and short fix intervals. Only 66% of 56 units functioned upon delivery for field deployment. Once tested and deployed, 28% malfunctioned. This study demonstrates that GPS tests should use the same schedules as field deployments, and GPS data quality in dense cover improves with short fix intervals. Lastly, with a short fix interval we reveal these devices can provide exceptional data for territorial carnivores. Miniature GPS units are a promising tool, but the study design should be carefully considered. |