Data and code from: Variable evidence of radio-tag backpacks affecting hummingbird time budgets in captivity
Data files
Jan 28, 2026 version files 77.44 KB
-
birdAndTagSpecs.csv
2.30 KB
-
finalAnalysisBiologgingEthics_2025_post-review.Rmd
52.48 KB
-
hummingbirdBehaviors_fullDataset.csv
11.30 KB
-
Inter-raterReliabilityDataset.csv
1.10 KB
-
overnightTrials.csv
386 B
-
README.md
9.89 KB
Abstract
Background: With wildlife-tracking devices miniaturizing rapidly to enable ever-more research on ever-smaller taxa, there is a newfound urgency for affordable, field-accessible biologging ethics studies. We designed a 3-hour time-budget experiment to investigate how radio-transmitter backpacks affect hummingbirds’ behavior in Colombia.
Methods: Using a large flight arena, we individually filmed 25 Black-throated Mangoes (Anthracothorax nigricollis) under two randomized treatments, tagged and untagged, to characterize and quantitatively compare behavior. We created time-budget breakdowns of our behaviors of interest—flying, hover-feeding, preening, and perching—then fit a series of linear mixed-effects models to determine the effects of tagging and additional experimental and environmental variables on behavior. We also designed an aviary-style “Entanglement Experiment” (n = 30) to determine if any individuals would snag on vegetation while equipped with the backpack harness, and tested 6 additional birds in this enclosure overnight for any longer-term negative effects.
Results: When tagged, individuals on average spent overall less time flying (with fewer and shorter bouts); and more total time feeding, preening (with more and longer bouts in both cases), and perching (with fewer but longer bouts)—however, this difference in total duration was only statistically significant in the case of preening. Our best-supported models also highlighted the importance of the following additional effects: whether or not the bird was undergoing its first or second 1.5-hour treatment (birds flew significantly more in their second treatment), bird mass (lighter birds fed significantly longer overall), and the time of day (birds preened significantly more in the afternoon than the morning, and more in the evening than the afternoon). No individuals in this captive study became entangled in vegetation or exhibited any adverse overnight effects from harness wear.
Conclusions: In our captive study, radio-transmitter backpacks significantly affected the amount of time that hummingbirds spent preening, and additional environmental variables helped explain behavioral differences in each bird’s treatments. While being in a confined space doubtlessly affected the behavior of individuals, our experimental model is relatively straightforward to fine-tune to other small avian taxa and is suitable for remote conditions, providing a useful basis for examining species-specific effects of biologging prior to starting field studies.
https://doi.org/10.5061/dryad.44j0zpcqw
Description of the data and file structure
We experimentally compared the individual behavior of 25 Black-throated Mangoes when they were and were not wearing radio-transmitter backpack harnesses. Each treatment (tagged and untagged) was randomized in order and we filmed the birds for 60 minutes. We broke down the birds' behaviors (flying, hover-feeding, preening, and perching) using the Behavioral Observation Research Interactive Software (BORIS). We also ran aviary-style experiments with 6 separate individuals overnight to determine if there would be any negative device effects that only became apparent after a longer period of time.
Files and variables
File: birdAndTagSpecs.csv
Description: This csv breaks down the weight specifications of each bird that we worked with and the various tags involved in this study (radio-transmitters and passive integrated transponders, aka "PIT tags").
Variables
- trialID: Identifier for specific individual and trial.
- birdWeightPreTrial: Weight of the bird, in grams, before the trial began.
- birdWeightPostTrial: Weight of the bird, in grams, after the trial began.
- birdWeightAvg: Average weight of the bird, in grams, across the course of the trial.
- CTTweight: Weight of the Cellular Tracking Technologies (CTT) radio-tag ("LifeTag"), in grams, that we used for that specific trial.
- CTTpercentBodyweight: The CTT tag's percent of the bird's bodyweight.
- CTTandHarnessWeight: Combined weight of the CTT tag and the harness used to equip it to the bird, in grams, that we used for that specific trial.
- CTTandHarnessPercentBodyweight: Combined percent bodyweight of the CTT tag and the harness used.
- CTTandHarnessAndPITweight: Combined weight of the CTT tag, harness used, and the PIT tag implanted in the individual prior to beginning the study.
- CTTandHarnessAndPITPercentBodyweight: Combined percent bodyweight of the CTT tag, harness used, and PIT tag.
File: hummingbirdBehaviors_fullDataset.csv
Description: This is the full csv of the time budget breakdowns and the additional covariates that we considered likely to affect hummingbird behavior.
Variables
- individual: Identifying number for the individual tested (1–25).
- trial: Unique trial identifier (a trial refers to the full experimental procedure for a single bird, containing two treatments).
- treatmentNumber: Refers to whether the individual was undergoing its first or second 1.5-hour treatment, regardless of whether it was tagged or not (2 levels: treatment 1 and treatment 2).
- treatmentType: Refers to whether the individual was tagged or not, regardless of whether it was undergoing its first or second 1.5-hour treatment (2 levels: tagged and untagged).
- treatmentCombo: Refers to the combination of treatment number and treatment type that the bird underwent (2 levels: A if the bird's first treatment was untagged and second treatment was tagged, and B if the bird's first treatment was tagged and second treatment was untagged).
- temperature: The tent’s average internal temperature for the treatment (continuous, in Celsius).
- sucroseConcentration: Percent sucrose of the provided nectar throughout the trial (continuous).
- birdWeight: Individual’s weight at the start of the trial (continuous, in grams).
- timeOfDay: Period of day wherein the majority of the treatment took place (3 levels: morning, 6:00–11:59; afternoon, 12:00–15:59; and early evening, 16:00–on).
- flightDuration: Calculated, from analysis in BORIS, the number of seconds individual spent flying (we considered “flying” to be whenever the bird was airborne but not hover-feeding) in the given treatment.
- flightPercent: Percent of time in the treatment that the individual spent flying.
- flightBoutNum: Calculated, from analysis in BORIS, the number of times ("bouts") an individual flew in the given treatment.
- flightBoutDur: Calculated, from analysis in BORIS, the average length in seconds of each flight bout in the given treatment.
- feedDuration: Calculated, from analysis in BORIS, the number of seconds individual spent hover-feeding (we considered “hover-feeding” to be whenever its bill was in contact with the syringe feeder) in the given treatment.
- feedPercent: Percent of time in the treatment that the individual spent hover-feeding.
- feedBoutNum: Calculated, from analysis in BORIS, the number of times ("bouts") an individual fed in the given treatment.
- feedBoutDur: Calculated, from analysis in BORIS, the average length in seconds of each feeding bout in the given treatment.
- preenDuration: Calculated, from analysis in BORIS, the number of seconds individual spent preening (we considered “preening” to be at any time that the bird's bill was in contact with its feathers) in the given treatment.
- preenPercent: Percent of time in the treatment that the individual spent preening.
- preenBoutNum: Calculated, from analysis in BORIS, the number of times ("bouts") an individual preened in the given treatment.
- preenBoutDur: Calculated, from analysis in BORIS, the average length in seconds of each preening bout in the given treatment.
- perchWithPreenDuration: Calculated, from analysis in BORIS, the total number of seconds individual spent perching (considered “perching” at any point when its feet were in contact with the perch) in the given treatment.
- perchNoPreenDuration: Calculated, from analysis in BORIS, the number of seconds individual spent perching (considered “perching” at any point when its feet were in contact with the perch) in the given treatment, having subtracted the amount of time the bird spent preening.
- perchNoPreenPercent: Percent of time in the treatment that the individual spent perching without preening.
- perchWithPreenBoutNum: Calculated, from analysis in BORIS, the number of times ("bouts") an individual perched in the given treatment.
- perchWithPreenBoutDur: Calculated, from analysis in BORIS, the average length in seconds of each perching bout in the given treatment.
- birdWeightChange: Difference in bird's weight over the course of the trial, in grams. Negative if the bird lost weight over the course of the trial, positive if the bird gained weight.
- feederWeightChange: Difference in feeder's weight over the course of the trial, in grams, indicating how much nectar the bird drank. The two NA values in this column correspond to a single trial wherein we unintentionally did not weigh the feeder at the end of the trial, and so could not calculate the difference.
- observer: The individual scorer whose BORIS analysis was used for the given trial.
- trialDate: The date of the trial.
- timeStart: The local (Bogotá, Colombia) time at the beginning of the treatment, not broken down categorically as morning, afternoon, or early evening, but instead in H:MM:SS format.
- timeEnd: The local (Bogotá, Colombia) time at the end of the treatment, not broken down categorically as morning, afternoon, or early evening, but instead in H:MM:SS format.
- ambientTemp: The average local ambient temperature in degrees Celsius over the course of the treatment, measured by the Ambient Weather Station on-site.
- tempDif: The difference between the average internal tent temperature and average external ambient temperature over the course of the treatment.
File: overnightTrials.csv
Description: This csv merely breaks down the amount of time that we tested individuals overnight in the aviary-style enclosure outside, while equipped with a radio-transmitter backpack harness
Variables
- TrialID: Unique trial identifier.
- TrialStart: Start time (local, Colombia) of trial, in HH:MM:SS format.
- TrialEnd: End time (local, Colombia) of trial, in HH:MM:SS format.
- TrialDuration: Calculated duration of trial in in HH:MM:SS format.
- DurationHours: Simplified duration of trial in hours (integer/decimal format).
File: Inter-raterReliabilityDataset.csv
Description: This csv breaks down the behavioral durations calculated by 6 of our "raters" (team members analyzing the videos in BORIS), for all behaviors, when scoring the same 4 videos.
Variables
- trial-treatment-behavior: Unique identifier indicating the trial, treatment (1 or 2), and behavior (flying, hover-feeding, preening, or perching) being scored.
- rater1-aec: Durations, in seconds, calculated by Rater 1.
- rater2-anc: Durations, in seconds, calculated by Rater 2.
- rater3-slm: Durations, in seconds, calculated by Rater 3.
- rater4-lrh: Durations, in seconds, calculated by Rater 4.
- rater5-ys: Durations, in seconds, calculated by Rater 5.
- rater6-mcf: Durations, in seconds, calculated by Rater 6.
Code/software: file finalAnalysisBiologgingEthics_2025_post-review.Rmd
To generate and run this R Markdown file, we used RStudio 2024.04.2 Build 764. All packages are listed within the code itself in the setup chunk at the top of the file. Packages, however, include: stats, ggplot2, lme4, AICcmodavg, DHARMa, irr, lmtest, glmmTMB, tweedie, tidyr, dplyr, coin, ggpubr, lmerTest, and MetBrewer. The rmd file includes a thorough breakdown of the full analysis that we conducted, which included calculating simple statistics on the weight of birds and tracking devices, as well as the length of overnight trials; paired-sample tests comparing behavioral durations, bout numbers, and bout lengths between tagged and untagged birds; model diagnostics and selection to determine the covariates that best predict the behavioral durations exhibited by the birds and plotting top-performing models and visually comparing behaviors of tagged and untagged birds.
As described above, we performed enclosure-style experiments (n = 25) consisting of two randomized treatments, tagged and untagged, to characterize and quantitatively compare behavior of Black-throated Mangoes when equipped (or not) with radio-transmitter backpack harnesses. Each treatment included a 30-minute acclimation period and a subsequent 60-minute observation period, in which we filmed the hummingbird's behavior. As such, we generated 50 one-hour videos throughout this study, which we analyzed using the Behavioral Observation Research Interactive Software (BORIS). With BORIS, we created time-budget breakdowns of our behaviors of interest: flying, hover-feeding, preening, and perching.
In this dataset you can find our time budget breakdowns for each individual treatment for each bird. Also included are additional covariates that we considered relevant to hummingbird behavior: treatment type (2 levels: tagged and untagged); individual’s weight at the start of the trial (continuous, in grams); treatment number (2 levels: treatment 1 and treatment 2); tent’s average internal temperature for the treatment (continuous, in Celsius); percent sucrose** of the provided nectar throughout the trial (continuous); and period of day wherein the majority of the treatment took place (3 levels: morning, 6:00–11:59; afternoon, 12:00–15:59; and early evening, 16:00–on). These data are included in one master dataframe.
Furthermore, you can find an inter-rater reliability dataframe, which we used to determine the reliability of our ethological video analyses and the protocols with which we differentiated behavior types. These data comprise 24 measurements (6 raters scoring the same 4 videos). We used these data to compare raters’ calculated durations (of flying, hover-feeding, preening, and perching) for the subset of videos in a single statistical test.
Lastly, you can find two smaller dataframes in this dataset: one which includes the full amount of time each bird was tested in our aviary-style "entanglement tests" overnight (n = 6), and another which breaks down the weight specifications of all the birds we tested, the radio-telemeters they wore in the experiment, the harness used to equip the radio-telemeters, and the passive integrated transponders that we tagged individuals with prior to beginning this study.
We also include all code used to conduct our analyses.