Fig 1.
Map of the Baffin Bay-Davis Strait study area, situated between Canada and Greenland.
The Baffin Bay narwhal population summering areas of Admiralty Inlet (blue) and Eclipse Sound (pink) are shown, along with the two field tagging sites at Kakiak Point (2009) and Tremblay Sound (2010-2012). Bathymetry raster (4000 m by 4000 m) from the International Bathymetric Chart of the Arctic Ocean, IBCAO, dataset version 4.3 [34].
Table 1.
Metadata for narwhals (N = 22) satellite tagged in Admiralty Inlet (AI) in 2009 and Eclipse Sound (ES) 2010-2012 that reached the overwintering area in southern Baffin Bay. The first and last date (yyyy-mm-dd) in the defined overwintering area were estimated using time-series plots, resulting in the number of winter days/locations for each ID used in the hidden Markov model (HMM) analysis.
Fig 2.
Argos locations and tracks of Baffin Bay narwhals (N = 22) tagged in 2009 in Admiralty Inlet and 2010–2012 in Eclipse Sound that reached the overwintering area in southern Baffin Bay.
Tracks are split into the defined summer (turquoise) and winter (purple) seasons. Bathymetry raster (4000 m by 4000 m) from the International Bathymetric Chart of the Arctic Ocean, IBCAO, dataset version 4.3 [34].
Fig 3.
Hidden Markov model (HMM) histograms and decoded states (Viterbi) for three behaviours of overwintering Baffin Bay narwhals.
States 1 (surface activity; green), 2 (pelagic diving; blue), and 3 (deep diving; red) are shown with respect to data streams: mean speed (Ms), tortuosity (Tr), surface time (Sf), mean depth (Md), relative dive depth (Rd), and time at depth (Dp). Covariate effects on the emission probabilities for shape (κ), scale (λ), mean (μ), and standard deviation (σ) included effect of available bathymetry on diving behaviour (Md, Rd, Sf, and Dd), effect of day of winter on surface time, and effect of tag programming on diving data streams. Covariate effects are numbered using Roman numerals and correspond to the histogram on the same row. Distributions for Md, Rd, and Dp, and Sf assume 2010 + tag programming, and a fixed mean bathymetry of 1379 m and winter day of 100. The first column of data for Dp highlighted in dark grey represents zeros that were modeled by the zero-mass parameter. Ribbons represent the estimated 95% confidence interval. Histograms exclude values of tags deployed in 2009.
Fig 4.
Stationary state probabilities of all behavioural states as functions of covariates.
State 1 (green), 2 (blue), and 3 (red) shown as functions of winter day (wday), bathymetry (bath), seafloor slope (slope), distance to shore (shore) and local hour of day (hour; note x-axis is shown from 0 (midnight) to 23 hours). Shaded area represents the 95% confidence interval.
Fig 5.
Map of decoded states for all 22 satellite tagged narwhals that reached southern Baffin Bay-Davis Strait.
Tracks are coloured by decoded behavioural states (1 = green, 2 = blue, 3 = red) and data gaps are indicated in grey.
Fig 6.
Winter distribution of decoded predicted states across all years.
(a) Total number of steps that occurred in each cell (50 km by 50 km), which represents the total data that went into predictions for each cell (higher number of steps used indicates greater confidence in state predictions), and (b) calculated as the state S with the highest within-cell proportion relative to the frequency of each state across all cells. Cells with < 4 steps are not plotted, and hash lines represent cells with < 16 total steps, which indicate low certainty in those state predictions.
Fig 7.
Predicted winter distribution of decoded behavioural states.
(a) State 1 (green), (b) state 2 (blue), and (c) state 3 (red), dark areas represent high state frequency within each cell (50 km by 50 km). Cells with < 4 steps are not plotted, and hash lines represent cells with <16 total steps, which indicate low certainty in those state predictions.
Fig 8.
Map of predicted stationary state probabilities across narwhal overwintering area.
All three states (panels a: state 1, b: state 2, c: state 3) mapped based on covariate (bathymetry, distance from shore, seafloor slope, winter day, and hour of day) values at each location.
Fig 9.
Maximum dive depth reached within each behavioural state.
Maximum dive depth calculated as a proportion of available maximum depth for a) state 1, b) state 2, and c) state 3. Bathymetric contour lines (light to dark) indicate 0 m, 500 m, 1400 m, and 1800 m.
Fig 10.
Occurrence of four key winter prey types for Baffin Bay narwhal.
The four panels represent Greenland halibut (Halibut), squid species (Squid spp.), various shrimp species (shrimp spp. excluding Pandalid shrimp), and Pandalid shrimp found in multi-species bottom trawl surveys only conducted in Canadian waters (Divisions 0A and 0B) between October and November 2009-2012. Locations of all trawl sets are shown in open black circles and species occurrence coloured in by starting depth (m) of the set. Bathymetric contour lines (light to dark) indicate 0 m, 500 m, 1400 m, and 1800 m. Data (only shown for this study area) was supplied by Fisheries and Oceans Canada Central and Arctic Multi-species Stock Assessment Survey Database [72].