Passive Acoustic Monitoring (PAM) can be a great tool to monitor marine mammal presence over wide areas and long time periods. However, there can be large costs associated with retrieving and replacing the recorders – for example, those deployed offshore or in dynamic environments may require significant time, specialist vessels and personnel in order to retrieve and replace the unit which can be costly to do regularly. Often, in order to get around this issue, there is a trade-off between the deployment duration and the sampling regime. For example, if the battery life of your PAM device limits your deployment duration, then you can use a subsampling scheme to record data at specific times. I.e. you could choose to record data continuously, or for ten minutes of every hour, or for ten seconds of every minute, or for one hour every 10 hours, etc. The subsampling cycle is up to the researcher to determine and is dependent on the questions they are trying to answer and the species of interest (e.g. some species only make sounds in certain situations or at certain times of day, season etc.)
This kind of subsampling of PAM data is frequently used in data collection; however, this new publication highlights the issues with subsampling data and provides recommendations on subsampling cycles that could help improve the data collected. A key issue is that subsampling can reduce the accuracy of the PAM surveys and this bias varies with the species and questions of interest.
Karolin Thomish (Ocean Acoustics Lab, Alfred-Wegener-Institute) and co-authors have just published a new paper called ‘Effects of subsampling of passive acoustic recordings on acoustic metrics’.
The authors ask the question:
“If the recording period is limited, for example, to 1 h during a day, one wonders whether sampling once daily for an hour, twice daily for half an hour, or four times a day for 15 min represents the vocal behaviour of a given species best.”
This should depend on the vocalization characteristics of the target species. For example, the subsampling regime should be different if you are targeting a species that produces regular calls compared to a species that rarely vocalizes. This paper investigates the impacts of different subsampling regimes on the results of acoustic presence and call rates for North Atlantic right whales and Antarctic blue whales.
They tested the data using a range of subsampling regimes ranging from duty cycles of only 2% up to 50%. They found that the probability to correctly assess daily acoustic presence decreased with increasing cycle periods and that the variability of call rate estimates increased with decreasing duty cycle.
They conclude that subsampling acoustic data can seriously bias any estimates of animal presence and call rates. They suggest that for species that vocalize infrequently, the best subsampling regime is for many short samples (for example: 1 minute on, 1 minute off) to be recorded rather than fewer longer samples.
You can find out more about the PAM work we do here.