Table 1.
Description of data coverage and sources (see references).
Data coverage has varied over time. Since 2011, pollock vessels have been fully observed; previously, vessels < 125 feet long were only observed for 30% of pollock fishing days at sea while longer vessels were fully observed.
Table 2.
Candidate predictor variables for predicting whether a trip is a fishing or non-fishing trip.
Trip-level predictors are based on the characteristics of all VMS records per trip that meet the given descriptions.
Fig 1.
Distributions of observed and VMS-estimated trip durations.
(a) Overlain histograms illustrating the distribution of durations for observed and VMS-estimated trips. Dark grey areas show overlap between the two histograms. For illustration purposes, figures are scaled to a maximum of 10,000 min which omits < 1% of trips with longer durations. (b) Scatterplot of the observed versus VMS-estimated duration for each trip. Data are log-transformed to better illustrate the clusters of data greater than and less than ~ 700 min. The vertical line shows the log-transformation of 700 min (6.55), the cutoff for exploring different models to estimate bias in duration estimation. The grey line represents the 1:1 line. (c) Histogram of the difference between the observed and estimated duration for each trip. For illustration purposes, values less than -500 and values greater than 500 have been grouped into single bins, “< -500” and “>500,” respectively. (d) Histogram of percent error (positive errors indicate over-estimation) of observed versus estimated trip durations. For illustration purposes, values less than < -25% and greater than 25% have been grouped into single bins, “< -25” and “> 25.”
Table 3.
Distribution of fishing and non-fishing trips.
Total numbers of trips and vessels, plus the percent of the total trips for AFA fishing, non-AFA fishing, and non-fishing trips. Annual tallies are provided by season (Winter “A” season, Summer “B” season, and “N” non-AFA season).
Fig 2.
Percent errors in the estimated trip duration as a function of time gaps in VMS transmissions.
Gaps in the regular transmission frequency greater than the expected 30 min intervals were simulated by removing 1–4 VMS observations from a random location within a trip’s sequence of VMS records. Removals yielded gaps in the VMS sequence of 60, 90, 120, and 150 min. Points outside of the whiskers represent outliers (> 1.5 times the upper quartile), whiskers represent the range (excluding outliers) and the boxes represent upper and lower quartiles with the median depicted by the horizontal line within each box.
Fig 3.
Cumulative distribution of the maximum time gap between VMS records for each trip.
For illustration purposes, the 10% of trips with maximum gaps > 300 min are not shown here. Vertical grey lines are shown at each of the gap durations for which we simulated removals of VMS records.