Post provided by Hanna K Nuuttila

To celebrate the 10th Anniversary of the launch of Methods in Ecology and Evolution, we are highlighting an article from each volume to feature in the For Volume 9, we have selected ‘Estimating effective detection area of static passive acoustic data loggers from playback experiments with cetacean vocalisations’ by Nuuttila et al. (2018).  In this post, the authors discuss the background and key concepts of the article, and the application of the article for assessing abundance of cetaceans.

Background: How to Hear Harbour Porpoise

Fifteen years ago, we knew surprisingly little about absolute abundance of some of the most common cetaceans off our shores, and this was certainly the case for harbour porpoises. Myself, like many other marine acousticians were excited to use novel static acoustic monitoring devices to learn how harbour porpoises were using our shores, but as the technologies were new and untested to a degree, there were many unknowns about their acoustic performance and data collection abilities. We had learned to use these devices to identify cetaceans, some even to species level and were able to pinpoint when and where they were vocalising – effectively allowing us to ‘trap’ them acoustically. But we still needed to understand how to use static acoustic devices to estimate animal abundance, and this requires information on vocalisation rates, group sizes and crucially, the working range of acoustic dataloggers for different species and scenarios and the consequent Effective Detection Area (EDA). After all, abundance is specifically about understanding animal numbers in a defined space. And this method would be useful for any logger used for abundance estimates.

Mother and calf harbour porpoise. Picture credit: Sea Trust.
Team setting up the playback station. Photo: Hanna Nuuttila.

During my PhD I was lucky to be advised by a team of cetacean scientists who were working on these very questions for a major international project called SAMBAH – The Static Acoustic Monitoring of the Baltic Harbour porpoise. They had been attempting playback experiments to reveal the effective detection area in the study and were willing to repeat a similar exercise in my project area. So, there we were, planning a research expedition to a tiny village in west Wales which included major transportation of equipment, staff and knowhow from researchers in Sweden, Denmark, Germany and UK, with major contributions from the German Oceanographic Museum and Chelonia Ltd in Cornwall. For two weeks I ran a team of nine or ten staff and volunteers as we built hydrophone arrays, made up rope moorings, carried and lifted enormous amount of weights, deployed C-PODs, and built an acoustic lab on an wildlife watching boat and used a small RIB to drift around pretending to be an echolocating porpoise (no mean feat!). The resulting data required months of manual validation by a wonderfully dedicated team of volunteers without whom I would not have succeeded, until a suitable dataset was ready for further analysis. The paper however, would not see the light of day until nearly five years later.

Determining Effective Detection Area (EDA)

The published article, ‘Estimating effective detection area of static passive acoustic data loggers from playback experiments with cetacean vocalisations’, covers basic concept of acoustic monitoring, and details two relevant statistical approaches to density estimation, one based on individual vocalisations, and the other based on a snapshot of time where animal movement is negligible. A variant of the latter method can also deal with the situation where animals are in groups and multiple animals can be detected in a single snapshot. Both formulas rely on estimation of the detection function g(y), and hence the Effective Detection Area (EDA). Ideally this information would be attained from wild, free moving animals, but as this is not often possible, this paper presents an alternative approach, based on playback of animal vocalisations. Here, both artificially created click as well as real animal recordings were used to achieve two objectives; the performance of the datalogger’s hydrophone in detecting animal clicks (artificial and real) and the performance of the datalogger’s software in classification and identification of the recorded porpoise vocalisation sequence used.

Relatively small effective detection radii across all the C-PODs tested. Source: Nuuttila et al., 2018.

The playbacks were conducted with varying levels of click amplitude and from varying distances away from the datalogger. Additional variables were the known sensitivity levels of the dataloggers.  To estimate the detection function the detected clicks were analysed using a Generalised Additive Mixed Model (GAMM) for both types of signal.  Source level and distance from logger influenced the detection probability. Whilst differences between loggers were evident, detectability was influenced more by the deployment site than within-logger variability. Maximum distance for detecting real recorded porpoise clicks was 566 m. Mean EDR for artificial signals with source level 176 db re 1 μpa @ 1m was 187 m and for a recorded vocalisation with source level up to 182 db re 1 μpa was 188 m. For detections classified as harbour porpoise click sequences, the mean EDR was 72 m. This information, together with data on vocal rates and typical group sizes can help as assess absolute abundance of cetaceans based on static devices alone.

Applying Acoustics for Abundance Estimates

Harris et al. (2018) used acoustic bearings derived from sparse arrays to estimate fin whale density and distribution. Picture: Hanna Nuuttila.

Since this study was conducted, we still know strikingly little about cetacean abundance in our oceans and coastal seas, but many exciting studies are definitely changing that. The SAMBAH project estimated the population number of the critically endangered Baltic harbour porpoise and identified breeding area for the species. They also conducted an in-situ playback study to assist with the estimates, but the results of that study are yet to be published. The other pieces of the density estimation puzzle from static acoustic devices are the cue rate or the vocalisation rate and the group size estimation.  In the avian field, a study has been published, detailing  a vocal activity index to assess terrestrial bird abundance and others have succeeded in estimating group size of Blainville’s beaked whales from acoustic footprints. Simultaneously to our study, a group published a study demonstrating density estimation of vocalising terrestrial animals using just a single acoustic recorder; whilst others have demonstrated a density estimation method from sparse vocalisation. Barlow et al. (2018) perfected the use of DASPRs, which are custom designed floating sensors for three-dimensional acoustic localization and tracking, and we’re keenly waiting on the publication of their current work on detection ranges and abundance.

Another interesting project that will greatly add to the field of acoustic density estimation is the ACCURATE – “ACoustic CUe RATEs for passive acoustics density estimation” led by Tiago Marques, which aims to synthesise current state of knowledge on acoustic cue rates and cue stability of various different species of deep-diving and baleen whales;  develop methods for cue rate estimation; explore factors that determine cue rate variability and evaluate the impacts of cue rate variability on density estimates from cue-based methods.

To read the selected papers for the other volumes of Methods in Ecology and Evolution article in full, visit Method’s 10th Anniversary page on the MEE blog.