Effects of differential correction on accuracy of a GPS animal location system

The location error of uncorrected data from Global Positioning System (GPS) collars range from ca. 45.5 to 65.5 m (Rempel et al. 1995). Improvements that potentially could reduce locational error to <10 m include correcting systematic bias by reference to CPS data collected at a known position (differential correction), increase in the proportion of positions based on ranging 4 rather than 3 satellites (3-dimensional mode), and increase in the proportion of positions based on a well-spaced satellite configuration (low dilution of precision). Design changes meant to achieve these results were implemented in the second generation CPS collars we evaluated (Lotek Eng. Inc. 1996). We tested the performance of these collars under the controlled canopy conditions of the Thunder Bay experimental forest. Differential correction caused location error to decrease from 80 to 4 m (P < 0.0001), and the range of 25-75th percentile location error to decrease from 74.3 to 5.0 m. Location error among sample sites was greatest under tall red pines (Pinus resinosa; 15.7 m), possibly because the tall trunks interfered with signal reception, resulting in the acceptance of either 2-dimensional mode positions or positions based on poor satellite configuration (high dilution of precision), or because of multipathing effects caused by signal bounce off the tree trunks. Implementation of differential correction may involve substantial costs to maintain a CPS basestation and data handling, so effectiveness of this enhanced technology must be judged against study objectives and data requirements of the hypotheses being tested.