Northern Prairie Wildlife Research Center
Charles J. Deutsch, Dean E. Easton, Howard I. Kochman, and James P. Reid, Sirenia Project, U.S. Geological Survey, Biological Resources Division, Florida Caribbean Science Center, 412 NE 16th Avenue, Rm. 250, Gainesville, FL 32601 USA
Satellite-based telemetry has been widely used for tracking the movements of large animals over long distances. The extent to which spatial data provided by Service Argos' Data Collection and Location System may be useful for finer-scaled analyses of movements or habitat use depends upon the accuracy and precision of the locations. We have used platform transmitter terminals (PTTs) deployed in buoyant, tethered housings to track the movements of West Indian manatees (Trichechus manatus) along the Atlantic coast of Florida and Georgia since 1986 and in Puerto Rico since 1992. We carried out a study to assess the bias and precision of the Argos spatial data generated by these satellite-monitored tags in the coastal environment that is typical for manatee tracking.
Service Argos initiated the classification of locations according to their quality in April 1987 and switched to a new set of classification algorithms in mid-June 1994. The advertised standard deviations of the latitudinal and longitudinal components of the locational error for both classification systems are as follows: LC3, σy = σx = 150 m; LC2, 350 m; and LC1, 1000 m. Service Argos states that 68% of locations should fall within one error standard deviation from the true latitude and longitude. Assuming a bivariate normal error distribution, these errors translate into total error radii of 226, 528, and 1510 m, respectively (Keating et al. 1991). Also in June 1994, location class zero (LC0, σ > 1000 m) was split into four classes: a new LC0 (σ > 1000 m), 'A' and 'B' (no precision estimates), and 'Z' (rejected as implausible).
We conducted two experiments in the Banana River, an estuary on the east coast of central Florida. The first test used six PTTs moored in a 6-m radius circle for 51 days from June to July 1992. The second test was conducted from August to November 1994 with three PTTs moored in a 3-m radius circle over a 104-day period. PTTs were also attached to three manatees enclosed in pens adjacent to the fixed PTTs for periods of 16-30 days, as part of a "soft-release" project for captive rehabilitated manatees. "True" locations of the fixed PTTs and of the pen centers were determined to within 5 m accuracy using differentially-corrected GPS.
In 1992 the locations were significantly biased (by about 250 m) to the West of the actual test site, whereas in 1994 the extent of bias was minor. Locational precision improved significantly from the old to the new location class system, as indicated by the respective 68-percentile error radii for the fixed PTTs (1992/1994): LC3 - 312/218, LC2 - 622/360, LC1 - 1824/834, LC0 - 8824/4043, LCA - --/3527, and LCB - --/13294 m (n = 1394/1653). This change was accompanied by a marked redistribution in the relative frequency of location classes. Errors in longitude were significantly greater than errors in latitude, and only the latter consistently met Argos' specifications of expected precision. There was considerable variation in performance (i.e., % high-quality locations) across PTTs, but not in locational precision. The frequency of good locations (LC1-3) from tagged manatees during the 1994 experiment (67% for penned, 71% for free-ranging) was high compared to that obtained from other marine mammals but was lower than that from fixed PTTs (83%). For location classes 1-3, precision of locations from penned manatees was significantly poorer than that from stationary tags.
Empirical information on the bias and precision of PTT locations, combined with a knowledge of the spatial scale of the scientific or management question, should be used to determine which Argos location classes, if any, are appropriate for analyses of animal movements or habitat use. Caution should be applied when pooling or comparing Argos spatial data from before and after June 1994 because of the marked change in the relative frequency and precision of location classes.