Measurements to report when describing new animal vocalization

Question

This may be slightly similar to this question and this one.

I am in the process of describing a new marine animal vocalization (unknown species). Are there recommended call measurements to report when describing a new vocalization, especially when the species is unknown?

There are a ton of metrics one can extract/report on, including min, max, peak, and percentiles related to time, frequency, power, amplitude, RMS, etc. But I haven’t seen any consistency in the reported measurements among journal articles describing newly discovered animal vocalizations.

I would think min and max frequency, call duration, and inter call interval (at a minimum) would be useful from a biological perspective. Would mean and SD or median and 10/90th percentiles (frequency) be more useful? Any others?

Answer 1

I suspect the large amount of variety in the literature is because the answer depends a lot on what kind of sound you are describing. Also, I think there are features that are important to describe perhaps biologically to support that sound is animal produced and maybe start to discus function, but then there are other features that are required to ensure anyone else can easily identify that call in their own recordings in the future. I’m not sure these are always the same or not.

In addition min and max frequency, call duration, and inter call interval:

For an impulsive sound like an echolocation click peak frequency and 3 dB or 10 dB bandwidth would be essential for capturing how broadband or narrowband the call is.

For a tonal sound, slope or number of oscillations or some way to to capture any change in frequency over time would be really important.

In either case, median inter call interval as well as the range (90th and 10th percentile) would be informative.

Answer 2

Here I'd only echo what @selene said. The measurements to be performed will depend fully on the time of sound being quantified. However, about the last part of the question 'Would mean and SD or median and 10/90th percentiles (frequency) be more useful?' - here I'd argue the X-Yth percentiles for all measurements are of interest.

It seems like this is a fieldwork scenario, which means the measurement is always an attempt to guess where the animal is, in which direction it is calling, and what context it is vocalising in. For instance if distance to animal is known, the 90th percentile can be used to define the 'apparent source level' (e.g. as in [1]) and observe trends over time/context. In principle I could imagine the 'apparent bandwidth' and other 'apparent' measurement reports when the field/setup conditions aren't straightforward.

References

[1] Lewanzik, D., & Goerlitz, H. R. (2018). Continued source level reduction during attack in the low‐amplitude bat Barbastella barbastellus prevents moth evasive flight. Functional Ecology, 32(5), 1251-1261.

Answer 3

Understanding the variability is very valuable (some calls are more stereotyped/regular than other calls). In addition to SD & percentiles, but also understanding the total range (so the minimum & maximum for each measurement).