Rationale for Work

Introduction – current survey methods

Determining the density and distribution of cetacean species is fundamental to understanding their basic biology, and also to monitoring and mitigating the effect of man-made impacts on their populations.  Currently, the main method of achieving this is using visual line transect surveys.  Here, a set of randomly placed survey lines are traversed by an observation platform (a ship, airplane or helicopter) and all sighted animals of the target species are recorded, together with their perpendicular distance from the line.  In the standard method, it is assumed that all animals on the transect line (i.e., at zero distance) are seen with certainty, but that probability of detection declines with increasing distance from the line.  The distribution of observed detection distances is then used to estimate the average probability of detection, and this in turn allows estimation of total population size (abundance) or population size per unit area (density).  Line transects are a special case of distance sampling methods, which are described in detail in the two standard texts by Buckland et al. (2001, 2004).  A summary of distance sampling concepts relevant to this proposal is given by Thomas and Martin (2006).

Limitations of current methods

One problem with visual line transect methods is that many cetacean species spend the majority of their time too far below the surface of the water to be available for visual detection.  This means that even animals right on the line may not be detected with certainty, and has led to the development of methods to estimate probability of detection at zero distance using additional information such as observations from two or more independent observation platforms (see, e.g., Buckland et al. 2004: Chapter 6).  Another fruitful avenue of recent research has been the use of towed passive acoustic arrays either in place of a single visual platform or as a secondary observation platform.  Many cetacean species emit frequent characteristic vocalizations, making acoustic methods an attractive supplement or alternative to visual methods (Mellinger and Barlow 2003). There has been considerable research developing methods to detect and classify these sounds for various species (Desharnais and Hay 2004, Adam 2006)..  With an array of two or more hydrophones towed in a known configuration it is possible to estimate the bearing to a sound received at multiple hydrophones and, if the ship is moving fast relative to the animal, a sequence of these bearings can be used to estimate the perpendicular distance of the animal from the transect line.  Such approaches are the focus of ongoing research funded by the Office of Naval Research (ONR), with Thomas as PI.  Examples of combined visual-acoustic surveys include Barlow and Taylor (2005) and the SCANS II project.

Fixed passive acoustics

An alternative method of collecting undersea acoustic information is through fixed passive acoustic devices.  By “fixed”, we here include both devices that are anchored to the sea floor and those that are buoyed (and so potentially moving slowly), but whose position is known.  Such devices can be configured in dense arrays, such as those at the instrumented US Navy testing ranges, sparser arrays, such as those in the Integrated Undersea Surveillance System (IUSS) or as single sensors, such as typical deployments of High-frequency Acoustic Recording Packages (HARPs) and, in some cases, sonobuoys.  Many of these devices can pick up cetacean vocalizations, despite in some cases not being designed with this in mind, and so can potentially be used to monitor cetacean populations.  The species that may be detected depend on the frequency characteristics of the devices, as well as their location; for example the bottom-mounted sensors in the AUTEC range operate at low to high frequencies and detect multiple species such as beaked whales, sperm whales, and pilot whales (Moretti et al. 2002, Moretti et al. 2006), while the IUSS sensors are tuned to low frequencies and so detect low frequency baleen whales (Clark and Gagnon 2004).

Fixed passive acoustic devices have enormous potential for cetacean monitoring, because they enable large amounts of data to be collected over long time periods, or potentially processed in real time for select outputs, at low to moderate cost.  In areas where fixed arrays already exist, there would be very little additional expenditure required to extract and analyze the data, were suitable methods available.  Where fixed arrays do not exist at an appropriate scale, but where short to medium-term monitoring is required, temporary deployments of devices such as sonobuoys or various pop-up (i.e., anchored, retrievable) buoys could provide the required data.

Importance of estimating absolute density

This research focuses on using fixed passive acoustic devices to estimate density (or abundance) of cetacean species.  We believe it is an important goal to estimate absolute density, as opposed to some index of density such as received call rates, because indices often have only a weak relationship with population size (see, e.g., Anderson 2001, Pollock et al. 2002, Anderson 2003).  For example, if we find that the number of calls of a species of interest received per unit time are different between two areas or two seasons then this could be because of variation in calling frequency or in probability of detecting calls caused by differences in sound propagation in the water, average depth of the animals, call strength, etc.  Here, we attempt to deal with these potential sources of bias, focusing mainly on the application of distance sampling methods to the fixed passive acoustic scenario.

Extending previous research

The research builds upon capabilities for near real time monitoring of cetaceans using large arrays of acoustic sensors, recently developed under funding by ONR (Morrissey et al., 2006).  While there has been some initial work to give preliminary estimates of density for some species (e.g., Moretti 2006) this has used ad hoc methods.  Here, we propose to extend this work by developing a solid, statistically based set of tools that can be consistently applied to archive data as well as potentially be incorporated into near real time algorithms.  In the case of single hydrophone data, our proposed research builds on previous work in this area by, e.g., McDonald and Fox (1999) and Mellinger et al. (2003).