Data from: Using advanced tri-axial accelerometer data to improve behavioral time budgets and bioenergetic estimates of wintering Lesser Scaup
Data files
Jan 05, 2026 version files 33.26 MB
-
ACC_behaviors.zip
31.53 MB
-
Changepoint_censoring.zip
447.49 KB
-
Energetics.zip
1.27 MB
-
README.md
9.08 KB
Abstract
Wildlife behavior studies have provided vital information towards understanding the natural histories of wildlife species and identified crucial components regarding their habitat and metabolic needs. For many species, typical behavioral data is collected using diurnal scan sampling techniques that have limitations in both when and where the data can be collected, ultimately leading to biases in behavioral patterns. With technological and analytical advancements of radiotechnology, behavior data can be collected more often and over larger spatial scales than with traditional methods. This study compares the behavioral time budget estimates between two different observational methods: ground-truthed diurnal scanning observations and 24-hr tri-axial accelerometer (ACC) GPS/GSM transmitter data that were classified using machine learning. We used the time budgets produced from the two methodologies and calculated the daily energy expenditure (DEE) for wintering Lesser Scaup (Aythya affinis) to explore the implications of biased behavioral data. We found significantly more feeding and less flight behavior of birds in the ACC data than in the visual scanning data. Using the ACC behavior proportions of the two most energetically demanding behaviors (feeding and flying), we found that feeding occurred 42% more during the day and flying occurred 23% more during the night. Lastly, we identified that the DEE estimated using the diurnal scanning observations produced a significantly lower estimate than with the 24-hr ACC data. This advanced way of interpreting wildlife behavior patterns can increase our understanding of wildlife species’ natural history and make improved decisions regarding wildlife conservation and management. Incorporating this new technique of wildlife behavioral observations, we provided a new framework to expand our current knowledge of wintering waterfowl behaviors and energetic needs that can be adapted to research the vast intricacies of wildlife behavior.
Author List
Hannah L. Schley, Department of Entomology and Wildlife Ecology, University of Delaware, Newark, Delaware, USA
Christopher K. Williams, Entomology and Wildlife Ecology, University of Delaware, Newark, Delaware, USA
Josh Homyack, Maryland Department of Natural Resources, Cambridge, Maryland, USA
William F. Harvey (ret), Maryland Department of Natural Resources, Cambridge, Maryland, USA
Glenn H. Olsen, U.S. Geological Survey, Eastern Ecological Science Center, Laurel, Maryland, USA
Sharon Johnson, U.S. Geological Survey, Eastern Ecological Science Center, Laurel, Maryland, USA
Funding for this research was provided by the Maryland Department of Natural Resources and Ducks Unlimited. Additional funding was provided by the U.S. Department of Agriculture Hatch (DEL00774) and the University of Delaware Waterfowl and Upland Gamebird Center.
Lesser Scaup trapping occurred between December 2021 - March 2022 and December 2022 - March 2023 at Eastern Neck Wildlife Refuge, Rock Hall, Maryland, USA, and all transmitter implant surgeries occurred at the U.S. Geological Survey, Eastern Ecological Science Center, Laurel, Maryland, USA.
Dataset DOI: 10.5061/dryad.wpzgmsc1m
DATA & FILE OVERVIEW:
Description of the data:
These data were collected to calculate and compare daily energy expenditure of wintering Lesser Scaup in the Chesapeake Bay using two different methods for assigning behaviors during the wintering period of 2022-2023. The first method used to collect wintering behavior was performing diurnal instantaneous scan observations while the second analyzed 24-h tri-axial accelerometer (ACC) data collected using GPS/GSM transmitters. Data collected using transmitters were additionally used to analyze a separate tracking study.
Methods and Materials
We used OrniTrak-I30 3 G Ornitela 30-g GPS/GSM internal transmitters that were set to take a GPS location once every hour and an ACC x-y-z directional degree reading every 10 min for 5 sec bursts at 10 hz. For every 1-hr scan sample, we performed six separate 10-min scans using a tripod spotting scope and binoculars on lookouts, beaches and piers.
Files and variables
Folder 1: ACC_behaviors.zip
Files in ACC_behaviors.zip contain ACC data from captive (n = 2) and wild caught (n = 29) Lesser Scaup that were used to verify behaviors from captive scaup based on their x-y-z ACC reading and predict those behaviors on data that were unassigned from wild caught scaup.
File 1: lesc_unassigned_ACC.csv
Number of Variables: 95
Variable List:
- device_id: individual Lesser Scaup identification
- timestamp: date (YYYY-MM-DD) and time (HH:MM:SS) in Coordinated Universal Time
- Latitude: Latitude given from transmitter
- Longitude: Longitude given from transmitter
Column numbers 5-94:
- Pattern displays x1, y1, z1, x2, y2, z2 ... x30, y30, z30
- x(1-30): x coordinate from tri-axial accelerometer reading
- y(1-30): y coordinate from tri-axial accelerometer reading
- z(1-30): z coordinate from tri-axial accelerometer reading
- behaviors: default "Rest" behavior needed for model to assign predicted behaviors.
File 2: verified_trial_3s_4beh.csv
Number of Variables: 91
Variable List:
Column numbers 1-90:
- Pattern displays x1, y1, z1, x2, y2, z2 ... x30, y30, z30
- x(1-30): x coordinate from tri-axial accelerometer reading from two recorded lesser scaup individuals in dive tank simulation verified with camera footage
- y(1-30): y coordinate from tri-axial accelerometer reading two recorded lesser scaup individuals in dive tank simulation verified with camera footage
- z(1-30): z coordinate from tri-axial accelerometer reading two recorded lesser scaup individuals in dive tank simulation verified with camera footage
- behaviors: assigned behaviors of individual Lesser Scaup in dive tank simulation verified with camera footage
Folder 2: Changepoint_censoring.zip
Files in Changepoint_censoring.zip contain 1-hr GPS location data of Lesser Scaup (n = 29) that survived > 14 day post transmitter implant to analyze changes in daily movement and when best to censor GPS data
File 1: lesc_14days_postop.csv
Number of Variables: 6
Variable List:
- device_id: individual Lesser Scaup identification
- UTC_datetime: date (YYYY-MM-DD) and time (HH:MM:SS) in Coordinated Universal Time
- UTC_date: date (YYYY-MM-DD) in Coordinated Universal Time
- UTC_time: time (HH:MM:SS) in Coordinated Universal Time
- Latitude: Latitude given from transmitter
- Longitude: Longitude given from transmitter
File 2: Data_censor_changepoint.R
RStudio script calculating the daily distance traveled for 20 days of 29 Lesser Scaup and identifying when the greatest distance traveled per day were made in order to censor GPS data accordingly
Folder 3: Energetics.zip
Files in Energetics.zip contain behavior data from instantaneous scan sampling and predicted behaviors from the ACC data (described in Folder 1: ACC_behaviors.zip) that were used to calculate the estimated daily and hourly energy expenditure of Lesser Scaup in the Chesapeake Bay
File 1: full_predictedbeh_4day_10day.xgb.csv
Number of Variables: 5
Variable List:
- device_id: individual Lesser Scaup identification
- timestamp: date (YYYY-MM-DD) and time (HH:MM:SS) in Eastern Standard Time
- Latitude: Latitude given from transmitter
- Longitude: Longitude given from transmitter
- predictions: behavior as predicted based on model from ACC data (see ACC_behaviors.zip)
File 2: scanning_observations_2022to2023.csv
Number of Variables: 28
Variable List:
- date: date of scan (MM/DD/YYYY)
- Site: individual scanning site identification name
- lat: Latitude of observation point at scanning site
- lon: Longitude of observation point at scanning site
- Sunrise: time at which sunrise began at each site (H:MM) in Eastern Standard Time
- Sunset: time at which sunset ended at each site (H:MM) in Eastern Standard Time
- TempC: Temperature (C) at the beginning of each individual 10-min scan
- WindSpM/S: Wind speed (m/s) at the beginning of each scan
- Wind Direction: Direction where wind is coming from
- N = North
- E = East
- W = West
- S = South
- NE = North east
- NW = North west
- SE = South east
- SW = South west
- Tide Phase: Tidal phase at the beginning of each scan
- Rising = tide is rising
- Falling = tide is falling
- Low = tide is at lowest point
- High = tide is at highest point
- Moon Phase: Phase the moon is in for the day the scan occurred
- LQ = last quarter
- FQ = first quarter
- NM = new moon
- FM = full moon
- WNC = waning crescent
- WXC = waxing crescent
- WNG = waning gibbous
- WXG = waxing gibbous
- Ice Over: Percent of water covered in ice at the beginning of scan
- Cloud Cover: Amount of cloud cover at beginning of scan
- 1 = no cloud
- 2 = partly cloudy
- 3 = half cloud cover
- 4 = full cloud cover
- Precipitation: Precipitation type at the beginning of scan
- 0 = snow
- 1 = no precipitation
- 2 = rain
- Scan Number: Individual 10-min scan number within the full hour with values ranging from the first can being 1 and last scan being 6
- Scan Start: Start of individual 10-min scan (H:MM) in Eastern Standard Time
- Scan End: End of individual 10-min scan (H:MM) in Eastern Standard Time
- Scan Direction: Direction each 10-min scan occurred either being Right to Left (R-L) or Left to Right (L-R)
- Species: Waterfowl Species identified in scan
- Swimming: Number of individuals swimming during 10-min scan
- Sleeping: Number of individuals sleeping during 10-min scan
- Flying: Number of individuals flying during 10-min scan
- Feeding_n: Number of individuals diving and eating during 10-min scan
- Loafing: Number of individuals loafing during 10-min scan
- Preening: Number of individuals preening during 10-min scan
- Agnostic: Number of individuals displaying aggressive behavior during 10-min scan
- Mating: Number of individuals showing mating displays during 10-min scan
- Walking: Number of individuals walking during 10-min scan
CODE & SOFTWARE OVERVIEW
All code was produced and ran in RStudio version 4.3.1.
Folder 1: ACC_behaviors.zip
Featured libraries:
tidyverse version 2.0.0
rabc version 0.1.0
Predicting_ACC_behaviors.R
RStudio script outlining analysis on creating an algorithm from the verified ACC trial data to identify behaviors based on x-y-z tri-ACC readings and create a model used to predict behaviors on unassigned x-y-z ACC readings using XGBoost decision tree analysis
Folder 2: Changepoint_censoring.zip
Featured libraries:
tidyverse version 2.0.0
amt version 0.2.1.0
changepoint version 2.2.4
Data_censor_changepoint.R
RStudio script calculating the daily distance traveled for 20 days of 29 Lesser Scaup and identifying when the greatest distance traveled per day were made in order to censor GPS data accordingly