About me
Model approaches for estimating location error in miniaturized wildlife tracking devices
Bryan Hughes, Kyle Elliott, Debbie Kelly, Mélanie Guigueno
Rapid advances in satellite technology and the miniaturization of electronics have allowed the use of increasingly smaller (< 20 g) devices for wildlife tracking. Particularly, miniaturized GPS devices allow for collecting large, fine-scale datasets from small-bodied, remote species. However, limitations in battery longevity and sensor accuracy can limit GPS accuracy given environmental interference and the complexity of mapping fine-scale movement in free-living species. Ecologists often control for location error through the removal of outliers, but an estimate of error is necessary to determine what points are necessary to remove. However, approximate outlier removal has been shown to remove reasonable data. Likewise, inaccuracy estimates in bio-logging devices have focused on use in large terrestrial mammals, but seldom in small-bodied mammals and birds. Here, we use a series of controlled tests to measure the average performance (accuracy, precision, and successful fix-rate acquisition) in three brands of commercially available bio-loggers. We provide an example of controlled stationary testing for bio-loggers prior to deployment on wildlife. Further, we use a series of random forest regression analyses to compare manufacturer-specific parameters in reporting location errors. Such metrics account for just 20 – 40 % of variability in GPS datasets in controlled settings, and thus further support that the arbitrary removal of outliers can potentially remove usable data. We posit a set of guidelines for ecologists to consider prior to the deployment of wildlife bio-loggers and encourage the use of controlled testing to calibrate and improve GPS-based tracking data in free-living wildlife.
