Enabling high-resolution wildlife tracking: antenna beam-based approach including novel per-position error estimations for automated radiotelemetry systems.

Abstract

Background: The increasing importance of wildlife movement data in ecology and conservation has fueled the development of Automated Radiotelemetry Systems (ARTS) using very-high-frequency (VHF) transmitters. To make optimal use of this data, highly precise analysis methods are needed to detect even small-scale movement changes and thus provide high data quality. While various approaches have successfully minimized position errors in ARTS, they mostly rely on a single mean error estimate.

Methods: We present two contributions. First, an antenna geometry-based position finding method (antenna beams) for small scale movements that reduces position errors (PE) and increases the number of position estimates. Second, a model for per-position error estimation, predicting error as a function of signal and position characteristics, applicable for data without ground-truth information and across various position finding methods. Using ground-truth data from VHF transmitters recorded simultaneously with the ARTS trackIT and GPS, we validated and compared yield, position errors and predictive performance of our approach with the common angulation and multilateration methods.

Results: Our antenna beam-based method provided a substantial alternative to angulation for directional set-ups, achieving comparable mean PEs (41 m vs. 44 m) and especially higher number of localizations (up to 99% vs. 30 to 66%). The per-position error estimation model demonstrated a strong predictive performance (mean absolute deviation from true error down to 21 m) utilizing parameters such as the number of participating stations and antennas, maximum signal strength, normalized summed up signal strengths and positioning within the study area.

Conclusions: Our results indicate that (i) the antenna beam-based position-finding method outperforms common methods in both accuracy and yield, (ii) the novel introduced per-position error estimation model reliably reflects measured PE from ground-truth data, and (iii) the resulting setup provides a robust foundation for high-resolution wildlife movement analyses.