README: Humpback Whale Caller ID

https://doi.org/10.5061/dryad.w9ghx3fw2

The data included here are processed measurements and code from analysis of humpback whale tag records collected as part of a large collaborative study in the Gulf of Maine. The data include caller ID classifications for calls detected on 16 tags across 7 different groups of 2-3 humpback whales, as well as calculated inter-call intervals within and between individuals, and whale depth at the time of production of each focal call.

Description of the data and file structure

1. Focal/non-focal call analysis
   The file "CallAnalysis_Focal_Nonfocal.csv" lists all of the selections which were made manually in the tag record, as described in the manuscript. The file lists call number, call start time, group ID, and then columns for each of the tags in the group (Tags A, B, and C). The Tag C column is listed as NA if there were only two tags in a given group. The Tag A, B, and C columns list the caller ID classifications for each call. When a call was detected on one tag record, other tags in the group were inspected for the same call, and then selections were classified based on relative amplitude comparisons to determine which animal likely produced the sound. These classifications include focal (fc), non-focal (nfc), miss, miss noise, non-focal but with noise present (nfcnoise), or indeterminate (ind). A focal call refers to the call on the tag of the animal that produced it, while a non-focal call is the instance on a tag of a conspecific that did not produce the call but was still received. Miss refers to when a call was not detected on a tag, even though it was detected on another tag at that time. Miss noise is the same as a miss but with noise present. Unique focal calls were used to calculate individual and group-level call rate.

   The file "FocalNonfocalRL_final.csv" contains received levels (dB re 1µPa) calculated with the rms function in Matlab for all focal and non-focal call detections where we were able to confidently measure the signal (and not the noise). The first column contains the selection start time (s), the second column contains the call duration (s), the third column contains the caller ID classification, as in the previous file, and the fourth column contains the filename, which indicates the tag ID. Tag IDs are a combination of a two letter species identifier (i.e., mn for Megaptera novaeangliae) followed by a two digit year, an underscore, the Julian day of the tag deployment, and a letter indicating the order of the tag deployments that day.

   The file "RLfocalnonfocalcorrespondence.csv" gives the calculated received levels (dB re 1µPa) for those calls that were detected on multiple tags. For each call, the first column lists the received level of the focal instance of that call and the second column lists the received level of the non-focal instance.

2. Call timing analysis
   Inter-call intervals (i.e., the time between two consecutive calls measured from the start of one call to the start of the next call) for consecutive calls within individuals are included in "InterCallIntervals_WithinIndividuals.csv". Inter-call intervals for consecutive calls between individuals (i.e., in potential vocal exchanges) are included in "InterCallIntervals_BetweenIndividuals.csv". All inter-call intervals are given in seconds.

3. Call depths
   This file lists the tag depth (meters) logged on the tag at the time when each call was produced.

Code/Software

We also include here two R scripts for making calculations, summarizing data, and creating figures. Each R script lists all of the packages needed to run the code.

The R script "FocalNonfocalAnalysis.R" walks through summarizing the occurrence of the different caller ID classifications in the dataset and summarizing the received level measurements for focal and non-focal calls, which are then plotted in figures that are included in the manuscript.

The other R script included is "CallTiming.R" which includes the code for calculating the Bout End Criterion (BEC) using the package "diveMove", as described in the manuscript, as well as calculating a probability density curve for the inter-call interval between individuals.