Tracking small animals in complex landscapes: a comparison of localisation workflows for automated radio telemetry systems

Abstract

AbstractAutomated radio telemetry systems (ARTS) have the potential to revolutionise our understanding of animal movement by providing a near-continuous record of individual locations in the wild. However, localisation error in data generated by ARTS can be very high, especially in natural landscapes with complex vegetation structure and topography. This curtails the ecological questions that may be addressed with this technology. Here, we set up an ARTS grid in a valley with heterogeneous vegetation cover in the Colombian high Andes and applied an analytical pipeline to test the effectiveness of localisation methods. We performed calibration trials to simulate animal movement in high-or low-flight, or walking on the ground, and compared workflows with varying decisions related to signal cleaning, selection, smoothing, and interpretation, along with four multilateration approaches. We also quantified the influence of spatial features on the system’s accuracy. We tested the grid by deploying tags on two high-altitude hummingbirds, the Great Sapphirewing (Pterophanes cyanopterus) and Bronze-tailed Thornbill (Chalcostigma heteropogon). Results showed large variation in localisation error, ranging from only 0.4–43.4 m from known locations up to 474–1929 m, depending on the localisation method used. The lowest average median error across calibration tracks was 105 m. In particular, we found that the selection of higher radio signal strengths and data smoothing based on the temporal autocorrelation in movement data are useful tools to improve accuracy. Moreover, the variables that significantly influence localisation error include terrain ruggedness, height of movement, vegetation type, and the location of animals inside or outside the grid area. In the case of our study system, thousands of location points were successfully estimated for two hummingbird species that previously lacked movement ecology data. Our case study on hummingbirds suggests ARTS grids can be used to estimate small animals’ home ranges, associations with vegetation types, and seasonality in occurrence. We present a comparative localisation pipeline, highlighting the variety of possible decisions while processing radio signal data. Overall, this study provides guidance to improve the resolution of location estimates, broadening the application of this tracking technology in the study of the spatial ecology of wild populations.