Task 3 Assessing factors informing sue rate stability
Cue rate changes over time and space will be investigated to understand the main drivers of cue rate variability. Covariates that may be used to predict cue rate at times and locations where they are not measured will be considered, as well as covariates that may not always be available in a particular setting, but that enhance the understanding of cue production. Such covariates would enable the evaluations of potential biases of the existing tag datasets. Using the datasets identified previously, key drivers of cue rate variability across a range of taxa will be explored and identified. Factors affecting cue rates might be as diverse as time of day, sex, season, depth, species, population, sub-population, behavioral state, and worse of all, density itself. Models will be built predicting cue rate as a function of possible factors of interest, providing a framework to estimate cue rate for other times and places based on available covariates.
Task 4a Methods for cue rate estimation in baleen whales
For large baleen whales, identifying cues of the tagged whale from the tag can be challenging due to close amplitude of nearby animals. A variety of methods will be investigated, and the work will be carried out using data on blue, fin, right and humpback whales. Some of these data sets or approaches could be also used for Task 4b, in particular inferring presence of acoustic cues in tag data that do not have an acoustic data stream. The results will also be used to improve cue rate estimates for these species as applicable (Task 2) and will improve the breadth of the results obtained under tasks 5 and 6. While it is anticipated that preliminary cue rates might be available for these species already when ACCURATE starts under these projects, these have also revealed that additional work will be required to obtain completely reliable cue rates, namely working with non-acoustic data streams. Within ACCURATE the plan is to further develop methods for assigning cues to focal tagged animals.
Team members working on this task: Ana Širović, Susan Parks, Julia Dombroski
Species considered: Right whale (Eubalaena sp), fin whale (Balaenoptera physalus), blue whale (Balaenoptera musculus), humpback whale (Megaptera novaeangliae)
Data available: We are working with sensor and acoustic data collected using a range from biologging devices (D-Tags, B-probes, and Acousondes) deployed on baleen whales across 20 years of collaborative research at a variety of locations.
Approach taken: The assumption that only calls emitted by the tagged individual can be detected both in the accelerometer and in the acoustic data streams of multisensory biologging devices is widespread in the field of cetacean bioacoustics. We are systematically testing this assumption by investigating how social context, tag positioning, tag type, call characteristics (i.e., frequency and received levels), and sampling rate of accelerometer data affect the strength and quality of accelerometer signature of calls.
Progress: We have developed a graphical user interface to synchronously process acoustic and accelerometer data. We have gathered over 400 samples of calls + corresponding sensor data from blue and fin whales in multiple contexts and after an exploratory analysis to select parameters that contribute to variability of presence and quality of accelerometer signature of calls, we are starting a comprehensive analysis.
Outputs: Peer-reviewed paper (in prep).
Figure 1: Northeast Pacific blue whale (B. musculus) B call on the audio spectrogram (top panel) and on the spectrogram created using each axis of the accelerometer (next 3 panels) of an Acousonde tag. Audio spectrogram settings: FFT 256, Hann Window, 90% overlap. Accelerometer spectrogram settings: FFT 256, Hann Window, 90% overlap.
Task 4b Estimating cue rate from proxy data
In some cases, direct information about acoustic cue production rate might not be available. However, some variables available from non-acoustic tags (for example behavioural state) may correlate with cue rates, and hence be potentially useful for estimating acoustic cue rate. Once such model is developed, it can also be used to estimate behavioral states in tag data without acoustic sensors and make inferences on cue rate based on those estimated states. Datasets for 3 species (sperm whale, Cuvier’s beaked whale and narwhal) have been identified for this task. There might be other historical datasets, where non-acoustic information might be used to derive cue rate estimates, along the lines described above. If such datasets are identified under Task 1 we might extend task 4b to them.
Team members: Tiago Marques
Task 4c Investigating cue-rate variability of deep-divers due to geographic region, behavioral state and group size and composition
Previous research on deep-diving cetacean species has shown that geographic differences in acoustic behavior must be considered when using acoustic cues for abundance estimation in a new region (Douglas et al. 2005, Warren et al. 2017). This task will focus on sperm whales and will utilize archived data from combined visual and acoustic surveys to assess cue rate variability in two LMR priority geographic regions. The data available include recordings from towed passive acoustic arrays, with concurrent visual sightings during daytime hours, as well as recordings from drifting passive acoustic buoys.
Team members working on this task: Yvonne Barkley, Karlina Merkens, Erin Oleson
Species considered: Sperm whale (Physeter macrocephalus)
Data available: Passive acoustic towed array data were collected as part of 5 years of visual/acoustic surveys in the Main Hawaiian Islands in the Central North Pacific Ocean. Continuous acoustic recordings were collected using custom-built hydrophone arrays (Figure 1) towed at approximately 4-10 m deep, 300 m behind the ship while traveling at 18.5 km/h (10 kt). The components of the towed hydrophone arrays and data acquisition systems varied between surveys (Table 1). During daylight hours, trained acousticians monitored the sounds aurally with headphones and visually using spectrographic software (ISHMAEL, Mellinger 2001; PAMGuard, Gillespie et al. 2008). When cetacean vocalizations were detected, a phone-pair bearing algorithm in ISHMAEL or PAMGuard was used to calculate the direction of the sound source relative to the bow of the ship. These bearings were plotted using a mapping software with a GPS interface (PAMGuard; Whaltrak, by Jay Barlow). Target motion analysis was used to estimate the perpendicular distance to the animals based on the convergence of plotted bearings with left/right ambiguity. In some cases, the ambiguity in perpendicular distances was resolved if the ship turned while the animals were still vocalizing. Acoustic encounters were linked to concurrent sightings when acoustic bearing estimates were consistent with the sighted location of the animal group (Rankin et al. 2008). When the ambiguity of the acoustic location estimate could not be resolved, the average perpendicular distance from both sides of the ship was recorded. A total of 327 acoustic encounters of sperm whales were detected. A total of 72 sperm whale encounters were concurrently detected by both visual and acoustic methods and 242 sperm whale encounters were detected by acoustic methods only, and only 13 sperm whale encounters were only detected by visual methods (Table 1, Figure 2).
Figure 1. Diagram of the linear hydrophone array towed 300 m behind the NOAA research vessels at approximately 10 m deep during a cetacean abundance line-transect survey. The line array consisted of two depth sensors (denoted with ‘D’) and two array nodes spaced 20 m apart, each housing three hydrophones (black dots) spaced approximately 1 m apart.
Table 1. Summary table with survey information, acoustic equipment, and total sperm whale encounters detected during PIFSC cetacean surveys within the Hawaiian Archipelago, excluding DASBR recordings.