Submitted to: Rangeland Ecology and Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/1/2007
Publication Date: 7/1/2007
Citation: Ganskopp, D.C., Johnson, D.D. 2007. GPS error in studies addressing animal movements and activites. Rangeland Ecology and Management 60(4):350-358.
Interpretive Summary: Global positioning system collars are relatively new equipment being used to document distribution and activity in studies of wild and domestic animals. There is no information available that assesses the accuracy of these units when this equipment is used to determine distances animals traveled over the landscape. We found that when the GPS collars were moved over surveyed distances from 30 to 270 feet, distance measures derived from GPS data were incorrect by only about 6 and one half inches regardless of distance. Data from GPS collars that are sitting still, however, erroneously suggested the units had moved about 5 feet 7 inches each time they acquired a new position. We watched grazing cattle to determine when they were resting or traveling about, and found their daily travels were overestimated by about 15 percent if one did not remove the perceived travel distances from GPS data accumulated while they were indeed resting and stationary. These findings reveal that GPS data very accurately estimates distances between successive positions as long as an animal is moving. If there are prolonged periods, however, where animals are not moving moving about, errors in GPS data can seriously inflate estimates of their daily travels. These findings will mostly benefit, biologists and scientists studying behavior, distribution, and resource use by wild and domestic animals.
Technical Abstract: One concern in animal behavior studies employing global positioning system (GPS) collars centers on effects of GPS error on measures of animal movement. Errors may be additive over time and potentially intensify with high frequency sampling or fine scale movements. We addressed issues related to this question with 2 separate studies: 1) identified sources and degree of error contributed by individual GPS units, moved several distances over time; and 2) evaluated proposed techniques for filtering perceived but faulty distance measures from GPS data. The first study employed a 10 X 10 Latin Square to determine effects of integration times (n = 10, rows), GPS collars (n = 10, columns), and unit movements (n = 10 at 0, 10, 20, 30, 40, 50, 60, 70, 80, and 90-m treatments) on error of distance measures. “Error” was the difference between surveyed distances and distances derived from GPS coordinates. For the second study, we observed 12, free ranging, GPS collared cattle (8 hours each) and quantified time devoted to several activities. A classification assessment was conducted by comparing observation data with corresponding 5-min periods of GPS collar integrations after filtering by 4 methods. Except for immobile GPS collars in study 1, error was inconsequential for movements of 10 to 90 m (differentially corrected mean error = 0.17 m ). When collars were immobile, GPS error generated about 1.7 m of travel per integration with differentially corrected coordinates and 3.9 m with uncorrected data. Study 2 showed one may effectively filter data sets to remove GPS error that accrued when cattle were inactive from estimates of total distance traveled using any of several techniques (81%–92% correctly classified). Our most effective regression technique suggested daily travels of cattle were overestimated by approximately 1.15 km or 15.2% without filtering.