How to quantify animal activity from radio-frequency identification (RFID) recordings.

Arne Iserbyt,Benny Borremans,Wendt Müller,Marcel Eens,Maaike Griffioen

doi:10.1002/ece3.4491

Abstract

Automated animal monitoring via radio‐frequency identification (RFID) technology allows efficient and extensive data sampling of individual activity levels and is therefore commonly used for ecological research. However, processing RFID data is still a largely unresolved problem, which potentially leads to inaccurate estimates for behavioral activity. One of the major challenges during data processing is to isolate independent behavioral actions from a set of superfluous, nonindependent detections. As a case study, individual blue tits (Cyanistes caeruleus) were simultaneously monitored during reproduction with both video recordings and RFID technology. We demonstrated how RFID data can be processed based on the time spent in‐ and outside a nest box. We then validated the number and timing of nest visits obtained from the processed RFID dataset by calibration against video recordings. The video observations revealed a limited overlap between the time spent in‐ and outside the nest box, with the least overlap at 23 s for both sexes. We then isolated exact arrival times from redundant RFID registrations by erasing all successive registrations within 23 s after the preceding registration. After aligning the processed RFID data with the corresponding video recordings, we observed a high accuracy in three behavioral estimates of parental care (individual nest visit rates, within‐pair alternation and synchronization of nest visits). We provide a clear guideline for future studies that aim to implement RFID technology in their research. We argue that our suggested RFID data processing procedure improves the precision of behavioral estimates, despite some inevitable drawbacks inherent to the technology. Our method is useful, not only for other cavity breeding birds, but for a wide range of (in)vertebrate species that are large enough to be fitted with a tag and that regularly pass near or through a fixed antenna.

Highlights

Automated monitoring of moving targets via radio-frequency identification (RFID) technology is inextricably embedded in our human daily lives, with examples like theft prevention, stock management, pet identification, and access badges (Ngai, Moon, Riggins, & Yi, 2008)
Erasing successive registrations of a unique visit is based on an arbitrary threshold that only can be determined by visual observations in a subset of individuals
Visual observations for a subset of individuals are still needed, as it allows to quantify how much time is spent on average on the in-and outside of the nest box, and on either side of the antenna. These visual observations should confirm a limited overlap between the data distribution of time spent within and outside the nest box

Summary

| INTRODUCTION

Automated monitoring of moving targets via radio-frequency identification (RFID) technology is inextricably embedded in our human daily lives, with examples like theft prevention, stock management, pet identification, and access badges (Ngai, Moon, Riggins, & Yi, 2008). Applying a longer than optimal cutoff time would lead to an exponential increase of missed “true” visits, because fast individuals with foraging trips that are shorter than the cutoff time are discarded This example illustrates that accurate identification of independent nest visits using RFID data will depend on optimizing the trade-off between accepting nonindependent registrations and rejecting independent ones. This third method determines exact arrival times, which further facilitates the quantification of additional behavioral parameters such as the coordination within social networks or coordinated feeding activities between pair members caring for their offspring. Our approach of data validation and data processing can be useful for a wide range of (in) vertebrate species that are large enough to carry a tag and that regularly pass near or through a fixed antenna (Whitham & Miller, 2016)

| METHODS

Findings

| DISCUSSION