Ruido (version 1.0.2)

soundSat: Soundscape Saturation Index

Description

Calculate Soundscape Saturation for a combination of recordings using the methodology proposed in Burivalova 2018

Usage

soundSat(
  soundpath,
  channel = "stereo",
  timeBin = 60,
  dbThreshold = -90,
  targetSampRate = NULL,
  wl = 512,
  window = signal::hamming(wl),
  overlap = ceiling(length(window)/2),
  histbreaks = "FD",
  powthr = c(5, 20, 1),
  bgnthr = c(0.5, 0.9, 0.05),
  normality = "ad.test",
  beta = TRUE,
  backup = NULL
)

Value

A list containing five objects. The first and second objects (powthresh and bgnthresh) are the threshold values that yielded the most normal distribution of saturation values using the normality test set by the user. The third (normality) contains the statitics values of the normality test that yielded the most normal distribution. The fourth object (values) contains a data.frame with the the values of saturation for each bin of each recording and the size of the bin in seconds. The fifth contains a data.frame with errors that occurred with specific files during the function.

Arguments

soundpath

single or multiple directories to your audio files

channel

channel where the saturation values will be extract from. Available channels are: "stereo", "mono", "left" or "right". Defaults to "stereo".

timeBin

size (in seconds) of the time bin. Set to NULL to use the entire audio as a single bin. Defaults to 60

dbThreshold

minimum allowed value of dB for the spectrograms. Defaults to -90, as set by Towsey 2017

targetSampRate

desired sample rate of the audios. This argument is only used to down sample the audio. If NULL, then audio's sample rate remains the same. Defaults to NULL

wl

window length of the spectrogram. Defaults to 512

window

window used to smooth the spectrogram. Switch to signal::hanning(wl) to use hanning instead. Defaults to signal::hammning(wl)

overlap

overlap between the spectrogram windows. Defaults to wl/2 (half the window length)

histbreaks

breaks used to calculate Background Noise. Available breaks are: "FD", "Sturges", "scott" or any numeric value (foe example = 100). Defaults to "FD"

powthr

numeric vector of length three containing the the range of thresholds used to evaluate the Soundscape Power of the Activity Matrix (in dB). The values correspond to the minimum threshold, maximum threshold and step size respectively.
Defaults to c(5, 20, 1), which evaluates thresholds from 5 dB to 20 dB in increments of 1 dB

bgnthr

numeric vector of length three containing the the range of thresholds used to evaluate the Background Noise of the Activity Matrix (in %). The values correspond to the minimum threshold, maximum threshold and step size respectively.
Defaults to c(0.5, 0.9, 0.05), which evaluates thresholds from 50% to 90% in increments of 5%

normality

character string containing the normality test used to determine which threshold combination has the most normal distribution of values. We recommend to pick any test from the nortest package. Defaults to "ad.test".
"ks.test" is not available. "shapiro.test" can be used, however we recommend using only when analyzing very few recordings

beta

how BGN thresholds are calculated. If TRUE, BGN thresholds are calculated using all recordings combined. If FALSE, BGN thresholds are calculated separately for each recording. Defaults to TRUE

backup

path to save the backup. Defaults to NULL

Details

Soundscape Saturation (SAT) is a measure of the proportion of frequency bins that are acoustically active in a determined window of time. It was developed by Burivalova et al. 2018 as an index to test the acoustic niche hypothesis. To calculate this function, first we need to generate an activity matrix for each time bin of your recording with the following formula:

$$a_{mf} = 1\ if (BGN_{mf} > \theta_{1})\ or\ (POW_{mf} > \theta_{2});\ otherwise,\ a_{mf} = 0,$$

Where \(\theta_{1}\) is the threshold of BGN values and \(\theta_{2}\) is a threshold of dB values. Since we define a interval for both the threshold, this means that an activity matrix will be generated for each bin of each recording. For each combination of threshold a SAT measure will be taken with the following formula:

$$S_{m} = \frac{\sum_{f = 1}^N a_{mf}}{N}$$

After these equations are done, we check every threshold combination for normality and pick the combination that yields the most normal distribution of saturation values.

If backup is set to a valid directory, a file named "SATBACKUP.RData" is saved after every batch of five processed files. If the function execution is interrupted (e.g., manual termination, an R session crash, or a system shutdown), this backup file can be passed to satBackup() (e.g., ~path/SATBACKUP.RData) to resume the original process. Once a backup is created, all arguments and file paths must remain unchanged, unless they are manually modified within the .RData object.

References

Burivalova, Z., Towsey, M., Boucher, T., Truskinger, A., Apelis, C., Roe, P., & Game, E. T. (2018). Using soundscapes to detect variable degrees of human influence on tropical forests in Papua New Guinea. Conservation Biology, 32(1), 205-215. https://doi.org/10.1111/cobi.12968

Examples

# \donttest{
### Downloading audiofiles from public Zenodo library
dir <- paste(tempdir(), "forExample", sep = "/")
dir.create(dir)
recName <- paste0("GAL24576_20250401_", sprintf("%06d", seq(0, 200000, by = 50000)),".wav")
recDir <- paste(dir, recName, sep = "/")

for(rec in recDir) {
 print(rec)
 url <- paste0("https://zenodo.org/records/17575795/files/", basename(rec), "?download=1")
 download.file(url, destfile = rec, mode = "wb")
}

### Running the function
sat <- soundSat(dir)

### Preparing the plot
timeSplit <- strsplit(sat$values$AUDIO, "_")
sides <- sat$values$CHANNEL
date <- sapply(timeSplit, function(x)
  x[2])
time <- sapply(timeSplit, function(x)
  substr(x[3],1,6))
datePos <- paste(substr(date,1,4), substr(date,5,6), substr(date,7,8), sep = "-")
timePos <- paste(substr(time,1,2), substr(time,3,4), substr(time,5,6), sep = ":")
dateTime <- as.POSIXct(paste(datePos, timePos), format = "%Y-%m-%d %H:%M:%OS")
leftEar <- data.frame(SAT = sat$values$SAT[sides == "left"], HOUR = dateTime[sides == "left"])
rightEar <- data.frame(SAT = sat$values$SAT[sides == "right"], HOUR = dateTime[sides == "right"])

### Plotting results

plot(SAT~HOUR, data = leftEar, ylim = c(range(sat$values$SAT)),
col = "darkgreen", pch = 16,
     ylab = "Soundscape Saturation (%)", xlab = "Time of Day", type = "b")
points(SAT~HOUR, data = rightEar, ylim = c(range(sat$values$SAT)),
col = "red", pch = 16, type = "b")
legend("bottomright", legend = c("Left Ear", "Right Ear"),
       col = c("darkgreen", "red"), lty = 1)

unlink(dir, recursive = TRUE)
# }