Competition for acoustic space in a temperate-forest bird community
Cite this dataset
Staniewicz, Agata; Sokołowska, Emilia; Muszyńska, Adrianna; Budka, Michał (2023). Competition for acoustic space in a temperate-forest bird community [Dataset]. Dryad. https://doi.org/10.5061/dryad.73n5tb339
Abstract
Animals that communicate by acoustic signaling share a common acoustic environment. Birds are particularly vocal examples, using a wide repertoire of songs and calls for mate attraction and territorial defense. However, interference caused by sounds that overlap in frequency and time can disrupt signal detection and reduce reproductive success. This may be particularly important in temperate regions where breeding is restricted to short seasonal windows. Here we investigated competition avoidance mechanisms used by the bird community inhabiting a primeval lowland temperate forest in Białowieża, Eastern Poland. We recorded morning soundscapes at 84 locations in early and late spring and calculated song dissimilarity indices to examine how species with greater song similarities use spatial and temporal partitioning to avoid competition for acoustic space throughout the breeding season. The bird community changed its use of acoustic space throughout the day and season. Birds did not use spatial acoustic niche partitioning when we looked at recording locations over the whole study period, but they did in a seasonal context, with species more acoustically different than expected by chance recorded at the same point on the same day. Our results also indicate that daily temporal niche partitioning may only occur at certain times before sunrise, with no evidence of large-scale temporal partitioning between species vocalizing during the same one-minute recordings in daytime. These results contribute toward our understanding of the evolution of bird communication, and highlight the strategies employed by different species to optimize their acoustic niche.
README: Competition for acoustic space in a temperate-forest bird community
https://doi.org/10.5061/dryad.73n5tb339
We calculated the Song Dissimilarity Index (SDI) for each pair of the species present in Białowieża Forest using good quality recordings of 55 reference species from the area (8±2 individuals per species, 414 individuals in total). The index is based on the Acoustic Dissimilarity Index developed by Sueur et al. (2008). To obtain the song samples, we recorded songs of 62 individuals of 10 species in Białowieża Forest between 19-28 April 2021. The recordings were made using a digital recorder (Marantz PMD661) at a sampling rate of 48kHz with16-bit accuracy, with a Sennheiser directional microphone. We supplemented these with recordings of 352 individuals downloaded from Xeno-canto (xeno-canto.org), recorded at a minimum 44.1 kHz sampling rate. As the files from Xeno-canto were saved in mp3 format, we first converted our recordings to mp3 format using Audacity 3.1.3 software and converted all mp3 files into wav format using ‘fix_wavs()’ function in warbleR package (44.1 kHz sampling rate, 16-bit, mono file).
From each recording, which corresponded to each individual bird, we manually selected between 2-12 (mean=7.17, SD=2.40) songs that had no background noise using the spectrographic representation with the software RavenPro 1.6.1 (window = Hanning, FFT = 1024). For species with very long songs, we selected multiple 10 s portions of the song instead of individual songs. We applied bandpass filter to remove all noise above and below the frequency range of the songs from the file to ensure that only the signal of the song was used to compute the index. We used the selection tables to extract individual songs in R 4.1.2 and calculated the spectral dissimilarity index for each pair of songs using the ‘diffspec()’ function in the Seewave package for R.
The dataset contains:
- the .csv file with mean SDI values calculated for each pair of reference species
- 55 folders with filtered audio recordings in .wav format and corresponding RavenPro selection tables in .txt format; each folder corresponds to a different bird species inhabiting Białowieża Forest; each recording contains song examples of one individual; recordings were collected by the authors in Białowieża Forest (files with names starting with AS or ES), or were downloaded from Xeno-canto (files with names starting with XC); the file names of recordings from Xeno-canto correspond to the recording names in the online database at xeno-canto.org
- R scripts used to calculate the Song Dissimilarity Index (SDI) for each pair of songs from the audio recordings and supporting files used for this process.
We randomly selected 84 recording points in Białowieża Forest District. Points were located both in protected and unprotected areas. The distance between neighbouring recording points ranged from 470 to 1,180m. At each recording point, we recorded the dawn chorus twice during the breeding season: once between April 20 and May 02, 2021 (early survey) and once between May 18 and May 26, 2021 (late survey). During each survey, we recorded the soundscape from two hours before sunrise to four hours after sunrise. We used 10 Song Meter SM3 acoustic recorders (Wildlife Acoustics) equipped with one omnidirectional microphone SMM-A1 (sensitivity -11 +/- 4 dB; signal-to-noise ratio > 68 dB). We recorded mono .wav files (1-hour duration, 48kHz/16 bit sampling rate, low and high frequency filters off, gain 24 dB).
From the soundscape recordings, we analysed 1-min every 10 minutes (6 minutes per hour, 36 minutes per survey per recording point). 1-min sound samples were analysed by manual scanning and listening to recordings, in Raven Pro 1.6.1 software (Cornell Lab of Ornithology) with the following settings: Window = Hamming, window size = 23.1 ms, Overlap = 75%. We compiled a list of bird species recorded in each 1-min sound sample.
The dataset contains:
- .csv file with the bird species list manually identified from each 1-minute sound sample from Białowieża Forest.
None of the species included in the dataset are of conservation concern.
Description of the data and file structure
Bialowieza_final.csv the file contains data on bird species identified on the 1-minute sound samples recorded from the 84 points in Białowieża forest. The data include: fileName (name of the 1-min sound sample); hourID (timestamp of the 1-hour recording created by the SongMeter recorder); recorder (ID of the recorder); gpsPoint (individual ID point number, assigned to each of the 84 points); decimalLatitude; decimalLongitude; altitudeInMeters; date; year; month; day; time; sample (sequence number of the 1-min sample from the hour); timeInSeason; recDay (recording day); part (1 = early season, 2 = late season); observer (initials of the person identifying the species); startTimeInSeconds (beginning of the song, automatically measured in RavenPro); endTimeInSeconds (end of the song, automatically measured in RavenPro); lowFreqInHz (automatically measured in RavenPro); highFreqInHz (automatically measured in RavenPro); freq25PercentInHz (automatically measured in RavenPro); freq75PercentInHz (automatically measured in RavenPro); peakFreqInHz (automatically measured in RavenPro); energy (automatically measured in RavenPro); inbandPowerInDecibels (automatically measured in RavenPro); speciesCode (numeric ID of the species); vernacularNameInPolish; vernacularNameInEnglish; scientificName; quality (subjective assessment of the song clarity, 1 = more clear, 2 = less clear); "null" values refer to sound samples where no species were identified.
pairs-index-values.csv the file contains the calculated mean Song Dissimilarity Index values (SDI) for each pair of the 55 reference species, which were used when analysing the occurrence of species pairs on sound sample recordings from the 84 points in Białowieża Forest. The data include: scientificName1, scientificName2, meanDiffspecValue (the mean SDI value calculated for the pair of species 1 and 2).
ref-songs-metadata.csv the file contains metadata on all the reference recordings used to calculate the SDI. The data include: datasetName (sonata = recorded by the authors, xeno-canto = from Xeno-canto); eventID (name of recording); vernacularNameInEnglish; vernacularNameInPolish; scientificName; country; verbatimLocality; decimalLatitude; decimalLongitude; verbatimElevation; eventDate; eventTime; habitatType; distanceToBird (in meters); playbackUsed; microphone; recorder; samplingRate; vocalisationType; birdSeen; "NA" values refer to data not collected for a particular reference recording.
songs-dirs folder contains 55 folders, each corresponding to one bird species and containing the song recordings (wav) and song selection tables (txt) used to compare each song to every other song and calculate the SDI. The folder names are used to identify each bird species. The file names are used to identify individuals.
calculate-indices.R the script calculates the alpha (NP - number of peaks, ACI - acoustic complexity index, BI - bioacoustics index; all not further used in the analysis) and beta (SDI - spectral dissimilarity index) acoustic indices for the songs and creates a new folder with results in the temp-local folder within the project
libraries-functions.R the script contains the libraries and custom-made functions used to process the recordings to calculate SDI; it is called in the script calculate-indices.R
extract-songs.R the script extracts all selected songs from the audio files in song-dirs, using the song selection tables; it is called in the script calculate-indices.R
plot-indices.R the script assigns the scientific names to species and plots the matrix of SDI between all species pairs (Fig. 1) and the variation between intraspecific and intraindividual SDI values (Fig. 2)
bird-taxonomy.csv contains taxonomy data for the bird species, used in the analysis to add scientific names to figures and arrange the species in family groups; the "nickname" variable refers to the name of the folder created individually for each reference species and "NA" values refer to species not included in the analysis, which did not have a folder.
Supporting files to run the R scripts: renv folder, renv.lock .R-version, .Rprofile
<br>
Code/Software
Folder structure needed:
The SDI calculation is ran in its own R project folder (eg. "song-dissimilarity-index") which needs to contain the following folders and files:
Assets: contains the file with bird taxonomy (bird-taxonomy.csv) used to arrange the results
Scripts: contains all the R scripts
temp: contains the songs-dirs folder with source recordings and selection tables
renv: contains files with information on R environment needed to run the calculation
Additional files in the main "song-dissimilarity-index" project folder:
renv.lock file containing information on all the packages and versions
.R-version file containing the information on R version (4.1.2)
.Rprofile file bootstraps renv
The calculation can be ran in the terminal in Ubuntu or OSX:
# enter the project folder and and create temporary directories
cd song-dissimilarity-index
mkdir -p temp temp-local
# install required R packages using renv (only need to run this once)
Rscript -e "renv::restore()"
# run R script to calculate acoustic indices for songs
# the '-d' option specifies the input directory
# the '-t' option specifies the number of processing threads
# 'cores' = number of physical processors, 'threads' = number of computational threads
Scripts/calculate-indices.R -d temp/songs-dirs -t 8
Funding
National Science Center, Award: 2019/35/D/NZ8/04416