(Color online) Diagram of study design. A porpoise swim track is made by combining consecutive surfacings by linear interpolation. Each swim track is then divided into snapshots of equal duration and a distance to the T-POD cluster is assigned to the midpoint of each snapshot. Each snapshot then constitutes a binary trial with a visual detection and a possible acoustic detection along with a given distance. If there is an acoustic detection it is a success and otherwise a failure. The binary trials and distances are the foundation of the detection function.
(Color online) All tracks obtained in 2003 and 2007. Stars represent the TPOD positions and the thin blue lines are interpolated swimming tracks between each surfacing. The black dot on land marks the theodolite position. The coastline is shown with a fat red line.
(Color online) Representative track from 2003 of a porpoise around a version 3 T-POD (indicated by ×). Each circle represents the estimated animal position at the midpoint of a single 15 s interval, with closed circles indicating that clicks were recorded simultaneously on the T-POD. Arrows at the top indicate direction of swimming.
Examples of click trains recorded by the two T-PODs during tracking of a porpoise. (A) Individual clicks on the version 3 T-POD, (B) clicks from the version 1 T-POD, and (C) interclick intervals for both recordings superimposed. As the drift of the internal clock of the T-POD becomes significant on the scale of milliseconds, the two recordings were aligned in time to obtain best possible fit.
Estimated probability of detection (smooth solid lines) and 95% bootstrap confidence intervals (smooth dashed lines) against distance for snapshots of 15 s and data in the category All Trains. Circles show the proportion of positive detections in ten equally spaced distance bands, and vertical lines show 95% binomial confidence intervals around these proportions. Closeness of the fitted detection line to these proportions indicates goodness-of-fit. The top three panels show estimates for T-PODs belonging to each of the three threshold levels used in 2007; the bottom left and center panels show estimates for the version 1 and 3 T-PODs used in 2003.
Summary table of used data.
Effective detection radius (ρ), probability of detection (P), density (standard deviation), D (confidence intervals), in porpoises · km−2 for two of the tested snapshots (15 and 60 s), T-POD detection thresholds and data categories. 95% confidence intervals on D (in parentheses) were derived from a nonparametric bootstrap (see the text). The two topmost T-PODs are from the 2003 dataset, where T-POD detection thresholds were not measured (denoted “na”). Note that the TPOD 745 is an outlier and that the apparent increased sensitivity is a result of internal abnormalities rather than more detected porpoises.
Article metrics loading...
Full text loading...