Catriona Harris

This paper describes the MOCHA project which aims to develop novel approaches for the analysis of data collected during Behavioral Response Studies (BRSs). BRSs are experiments aimed at directly quantifying the effects of controlled dosages of natural or anthropogenic stimuli (typically sound) on marine mammal behavior. These experiments typically result in low sample size, relative to variability, and so we are looking at a number of studies in combination to maximize the gain from each one. We describe a suite of analytical tools applied to BRS data on beaked whales, including a simulation study aimed at informing future experimental design.

ABSTRACT Cetacean sound-production rates are highly variable and patchy in time, depending upon individual behavior, social context, and environmental context. Better quantification of the drivers of this variability should allow more realistic estimates of expected call rates, improving our ability to convert between call counts and animal density, and also facilitating detection of sound-production changes due to acoustic disturbance. Here, we analyze digital acoustic tag (DTAG) records and visual observations collected during behavioral response studies (BRSs), which aim to assess normal cetacean behavior and measure changes in response to acoustic disturbance; data sources include SOCAL BRS, the 3S project, and Bahamas BRS, with statistical contributions from the MOCHA project (http://www.creem.st-and.ac.uk/mocha/links). We illustrate use of generalized linear models (and their extensions) as a flexible framework for sound-production-rate analysis. In the context of acoustic disturbance, we also detail use of two-dimensional spatially adaptive surfaces to jointly model effects of sound-source proximity and sound intensity. Specifically, we quantify variability in pilot whale group sound production rates in relation to behavior and environment, and individual fin whale call rates in relation to social and environmental context and dive behavior; with and without acoustic disturbance.

ABSTRACT Statistical analysis of data from multi-sensor acoustic tags presents several characteristic challenges. Datasets generally include time-series of many measurements on a small number of individuals; different data streams often have distinct temporal resolutions and precisions. The MOCHA project (Multi-study Ocean acoustics Human effects Analysis) is a three-year effort focused on developing innovative statistical methods for such data. Here, we present several approaches for appropriate, effective statistical analysis of such datasets, with an emphasis on quantitative assessment of changes in marine mammal behavior in response to acoustic disturbance. Issues to be addressed will include: combining data streams from multi-sensor tags (and also concurrent visual observation data) for statistical analysis; statistical methods to characterize or summarize normal behavior and detect departures from normal; methods for analysis of call-production-rate data from acoustic tags; and methods for combining analysis of data from multiple tags, individuals, and species. Specific statistical methods to be presented will include Mahalanobis distance as a summary of multivariate data, state-switching models, random effects, and other extensions of generalized linear models appropriate to tag data.