ESTIMATED ABUNDANCES OF CETACEAN SPECIES IN THE NORTHEAST ATLANTIC FROM TWO MULTIYEAR SURVEYS CONDUCTED BY NORWEGIAN VESSELS BETWEEN 2002–2013

Two shipboard line-transect surveys of the Northeast Atlantic were conducted between 2002–2007 and 2008–2013 to meet the ongoing requirements of the Revised Management Procedure (RMP) for common minke whales (Balaenoptera acutorostrata acutorostrata) developed by the International Whaling Commission’s Scientific Committee. Here we present estimated abundances for non-target species for which there were sufficient sightings, including fin whales (Balaenoptera physalus), humpback whales (Megaptera novaeangliae), sperm whales (Physeter macrocephalus), killer whales (Orcinus orca), harbour porpoises (Phocoena phocoena), and dolphins of genus Lagenorhynchus. The 2 surveys were conducted using a multiyear mosaic survey design with 2 independent observer platforms operating in passing mode, each with 2 observers. The abundances of Lagenorhynchus spp. from the 2002–2007 survey were estimated using single-platform standard distance sampling methods because of uncertainty in identifying duplicate sightings. All other estimates were derived using mark-recapture distance sampling techniques applied to a combined-platform dataset of observations, correcting for perception bias. Most notably, we find that the abundance of humpback whales, similar in both survey periods, has doubled since the 1990s with the most striking changes occurring in the Barents Sea. We also show that the pattern in distribution and abundance of fin whales and sperm whales is consistent with our earlier surveys, and that abundances of small odontocete species, which were not estimated in earlier surveys, show stable distributions with some variation in their estimates. Our estimates do not account for distributional shifts between years or correct for biases due to availability or responsive movement.


INTRODUCTION
Two multi-year surveys, targeting North Atlantic common minke whales (Balaenoptera acutorostrata acutorostrata), were conducted in the Northeast Atlantic between 2002-2007 and 2008-2013. The intent of the surveys was to achieve management targets under the Revised Management Procedure (RMP) for common minke whales, developed by the International Whaling Commission's Scientific Committee (IWC, 1994). Similar surveys have been conducted in Norwegian and adjacent waters to varying degrees since 1988 (Christensen, Haug, & Øien, 1992;Øien, 2009, 1990. All surveys preceding 1995 covered portions of the total study area (described under Materials and Methods), while a complete synoptic survey of the region was achieved in 1995. A cyclical mosaic survey design was implemented in 1996 to cover the Northeast Atlantic with a patchwork of smaller-scale surveys over a multi-year timeframe (Øien & Schweder, 1996). These are the second and third complete surveys under the mosaic survey design. The survey methodology has remained essentially the same, with slight improvements to ensure best possible estimates of minke whale abundance as the target species (Schweder, Skaug, Dimakos, Langaas, & Øien, 1997;Skaug, Øien, Schweder, & Bøthun, 2004).
Earlier surveys have resulted in published estimates for nontarget species including fin, humpback, and sperm whales from surveys conducted in , 1989, 1995(Christensen et al., 1992Øien, 2009, 1990, in which abundance estimates were made assuming that all animals on the transect line were detected (p(0)=1). This analysis differs in that it uses the double platform configuration to estimate p(0), accounting for perception bias to improve the abundance estimates.

Survey Design
The study area covers the Northeast Atlantic from the North Sea to the ice edge, and from the Greenland Sea in the west to the Barents Sea in the east. It consists of the 5 Small Management Areas (SMA) of the North Atlantic Minke Whale Implementation (IWC, 2004): CM, ES, EB, EW, and EN ( Figure 2). Within each SMA, a block structure was fitted to create areas of similar densities of minke whales, with survey effort distributed proportional to area. Within each block, transects were constructed as zig-zag tracks with a random starting point (Buckland et al., 2001). Block areas used to estimate species density were adjusted for ice-cover. In 2003, the SMA structure was modified by the IWC Scientific Committee, shifting the eastern boundary of the Barents Sea SMA westward to 28°E and extending the upper boundary of North Sea SMA southward to 62°N (IWC, 2004). This necessitated splitting the blocks BAW and FI each into 2 blocks, and because block FI was surveyed before the boundary change, it was further subdivided into Fl1 and Fl2, and re-stratified ( Figure 2a).
Due to the fragmentation of the strata through redefinitions of SMA boundaries that occurred in 2003 (IWC, 2004), it was necessary to redesign the block structure within the SMAs prior to the 2008-2013 survey (Skaug et al., 2004). The updated block design and names used in the 2008-2013 survey are illustrated in Figure 2b.

The surveys
In 2002 (NAMMCO, 2018), the duplicate effort in some blocks was retained and used to improve abundance estimation. The block BA2 was modified from the original BAW block mid survey cycle, in 2003. As a result, it was partially surveyed twice, and due to differing amounts of ice cover affecting the total area of the block, 2 separate estimates were obtained (BA2_a and BA2_b).
During 2008-2013, one or two vessels conducted the surveys each year, with a total of 7 vessels operating over the 6-year period. In 2008 the Svalbard area was surveyed; in 2009 the North Sea; in 2010 the Jan Mayen area; in 2011 the Norwegian Sea; and in 2013 the Barents Sea was surveyed.

Field methodology
Both surveys used a double-platform design with two platforms that were visually and acoustically separated from each other and thus independent. Platform 1 was positioned in a barrel on the mast above platform 2, which was located on the roof of the bridge. The two platforms varied in eye height depending on the vessel, with an average of 13.8 m for platform 1 and 9.7 m for platform 2.
Each platform operated continuously during daylight hours (between 05:00 and 23:00, depending on the latitude) with a team of 2 observers. Each team worked on 1-or 2-hour shifts with teams rotating between platforms. The searching speed was 10 knots with surveys conducted in passing mode. Searching was conducted by naked eye. The designated search area was the 90 o sector centred around the transect line, within 1500 m of the vessel. When searching, one observer in each team scanned the port 45 o sector from the transect line while the other scanned the starboard 45 o sector. All sightings were recorded regardless of whether they were sighted within the designated search area.
Observers recorded observations using a microphone connected to a central computer equipped with a GPS. Each observation documented the species, the angle from the transect line read from an angle board, the radial distance estimated by eye, and the group size. Tracking procedures were followed for minke whales, where the observer dedicated their effort to recording each repeat surfacing until it passed abeam of the ship. During tracking procedures, the other team member took over searching the entire 90 o search area. When both observers were occupied tracking minke whales, other minke whale sightings, along with all non-target species, were recorded as initial sightings only. After each completed recording of a minke whale or other large whale sighting, observers reported the sighting to the team leader by radio. The platforms operated on separate radio channels to maintain independence. During the surveys, regular training in distance estimation was conducted, including accuracy of angle-board readings and distance estimation using buoys as targets. Measures of covariates including glare, visibility, Beaufort Sea State (BSS) and weather conditions were recorded hourly and/or when conditions changed notably. Covariate classifications and definitions are detailed in Øien (1995). Acceptable survey conditions were defined as BSS of 4 or less and meteorological visibility greater than 1 km.

Data treatment
Sightings used in the abundance-estimate analyses were included based on the following criteria: the sighting was initially detected before abeam; the sighting was recorded from platform 1 or 2; and the species (or genus in the case of Lagenorhynchus spp.) was confirmed.
Observations from the two independent platforms were combined through a process of determining duplicate sightings. When possible, duplicates were identified in the field by the team leader operating from the bridge; otherwise, they were determined post-cruise.
The criteria used to determine duplicates, both in the field and in the post-cruise analysis, involved accounting for the timing and position of the sightings relative to the vessel (given a speed of 300 m per minute and allowing for some error in radialdistance estimates by different observers). Since only the initial sightings were recorded for non-target species, there was occasionally the need to match duplicates of disparate surfacings of the same whale. Given the relatively short designated search distance for the target species (1500 m), it was possible to have one platform make an initial sighting of a whale thousands of meters away, while the second platform observed it much later, once the ship moved closer to it. The team leader played an important role in identifying these duplicates in the field. When only one platform reported a sighting, the team leader could assist by tracking the whale so that if the other platform detected it closer to the ship, it could be identified as a duplicate.
In rare cases where one observer of a clear pair of duplicate sightings recorded the species as 'unidentified large whale' while the other confirmed the species, the positive ID was accepted for that sighting. In cases where there was uncertainty in species identification by one or both observers, the team leader, operating from the bridge, used binoculars to confirm uncertain identifications. Species identification was not always possible, so some sightings were left recorded as 'unidentified large whales.' For all duplicate sightings, the information recorded by the platform from which the whale was first sighted was used in the combined-platform analyses, as the analytical method used requires that these fields be identical (Laake & Borchers, 2004). Abundance estimates were calculated for the double platform for all species apart from Lagenorhynchus spp. in the 2002-2007 survey, where only a single platform (platform 1) was used due to uncertainty in judging duplicates. The certainty in judging duplicates improved between the 2002-2007 and 2008-2013 surveys due to a change in emphasis for the observers who were instructed to make a greater effort to discern and report smaller groups of dolphins rather than larger aggregations. In earlier surveys, some observers would classify nearby groups of dolphins as a single group, while others would classify them as separate groups. This caused greater uncertainty in judging duplicates, such that we did not feel they were reliable.

Analysis
These analyses were performed using the DISTANCE 7.2 software package (Thomas et al., 2010). Encounter rate and group size for each species and each survey were estimated by block. The effective search half-width (eshw) was estimated using pooled data over all survey blocks (globally) for each survey period as there were insufficient data to support stratified estimates.
To account for perception bias by estimating p(0), markrecapture distance sampling (MRDS) techniques were used (Laake & Borchers, 2004). The fully independent platform design allowed for the "independent observer configuration" to be used (Laake & Borchers, 2004). Both "full independence" (FI) and "point independence" (PI) were tested (Laake & Borchers, 2004). Models were chosen based on a comparison of the Akaike's information criterion (AIC) values. The "point independence" configuration requires the estimation of 2 detection functions: one for the probability of detection by one or more observers (Distance Sampling model: DS model), and a second conditional detection function (Mark Recapture model: MR model) for detection probabilities conditioned on detection by the other platform (Laake & Borchers, 2004). The "full independence" configuration requires only the conditional detection function. The conditional detection function is modelled logistically with the same covariates available for the primary detection function, selected based on AIC values.
The detection function models were selected based on AIC, goodness of fit test statistics, and visual inspection, particularly of data around the transect line. Hazard-rate and half-normal models were tested. The covariates considered were BSS, vessel identity, weather, group size, glare, and visibility. Some covariates were aggregated into categories for simplification and to improve model convergence, as detailed in Table 1. Data exploration also included truncation of the data by up to 5% if it improved the test statistics (Chi-square and Kolmogorov-Smirnov) and the shape of the q-q plot.
Encounter-rate variances were estimated using R2, the default in the mark recapture (MRDS) engine in DISTANCE 7.2, which is a design-based empirical estimator that assigns weights to transect lines based on length (Fewster et al., 2009). The confidence intervals of the abundance estimates were calculated assuming that estimated abundance is log-normally distributed (Buckland et al., 2001).

General
In 2002-2007 a total effort of 27,009 km of transects were searched over the survey period (Figure 2a), covering a total area of 2,962,269 sq. km. The distributions of search effort by BSS were 3% in BSS 0, 15% in BSS 1, 22% in BSS 2, 32% in BSS 3 and 28% in BSS 4. The surveys conducted between 2008-2013 covered a total area of 3,268,243 sq. km and 24,300 km of transects were searched ( Figure 2b). The distributions of search effort by BSS were 0.5% in BSS 0, 16% in BSS 1, 20% in BSS 2, 29% in BSS 3 and 33% in BSS 4.
In both survey cycles there were parts of the survey area that were not covered due to ice and unsuitable survey conditions. In 2002-2007, blocks VSI and SVI were not covered due to ice

Large whales
In 2002-2007 there were 893 unique records of large whale sightings (Table 2) and of these, 218 were identified as fin whales, 229 as sperm whales, 245 as humpback whales, 11 as blue whales, and 1 was identified as a sei whale. 189 sightings were categorized as 'unidentified large whales '. In 2008 were 611 records of large whale sightings (Table 3) and of these, 224 were identified as fin whales, 92 as sperm whales, 179 as humpback whales, 2 as blue whales, and 1 as a sei whale. 113 were categorised as 'unidentified large whale'.

Smaller odontocetes
There were 1042 unique records of smaller odontocete groups sighted during the 2002-2007 survey period (Table 2). Of these, 96 were identified as killer whales, 294 as harbour porpoises, 628 as Lagenorhynchus spp., and 12 as northern bottlenose whales. In 2008-2013, there were 487 records of small odontocete groups sighted (Table 3) and of these, 35 were identified as killer whales, 50 as harbour porpoises, 392 as Lagenorhynchus spp., and 10 of the sightings were identified as northern bottlenose whales.
The observations by platform, duplicates, and estimated p(0) for each species are shown in Table 4. In all cases, the PI models produced lower AICs than the FI models. Therefore, the PI method was used exclusively. Covariates included in the final model for each species, for both the Distance Sampling model (DS model) and the Mark Recapture model (MR model), are detailed in Table 5.

2002-2007
The sightings of fin whales are shown in Figure 3a. They were found throughout the survey area but were especially abundant west of Spitsbergen, in the Barents Sea, and in the western survey blocks near Iceland/Jan Mayen (NVN, NVS, JMC). The final detection function models used a half-normal key function, truncated to a perpendicular distance of 4000 m and included BSS as a covariate in the DS model ( Figure 4a). The resulting eshw was 1858 m. The abundance of fin whales was corrected with p(0)=0.72 (CV=0.10) to 10,004 (CV=0.18, 95% CI: 6,937-14,426). Detailed results by survey block are reported in Table  6a.

2008-2013
The highest encounter rate of fin whales occurred west of Spitsbergen (ES1, ES2) and in the western Iceland/Jan Mayen survey blocks (CM2, CM3) ( Figure 3b). The best-fitting models used a half-normal key function with truncation to 4000 m. The DS model was fit with BSS and weather as covariates and the MR model was fit with BSS as a covariate. Plots of the detection probabilities for each model are shown in Figure 5a. The resulting eshw was 1909 m. The abundance estimate of fin whales was corrected with p(0)=0.77 (CV=0.08) to be 10,861 (CV=0.26, 95% CI: 6,433-18,339) (Table 6b).

2002-2007
Humpback whales were found almost exclusively around Bear Island, in the northern Barents Sea, and in the western-most survey block north and east of Iceland (NVS), as depicted in Figure 6a. The best-fitting models used a half-normal key function truncated at a perpendicular distance of 4000 m and resulted in an eshw of 2240 m. The fitted detection function and conditional detection probability plots are shown in Figure 4b.

2008-2013
Humpback whales concentrated in 3 main areas: north and east of Iceland (CM2), around Bear Island (ES1), and in the northern Barents Sea (EB3) (Figure 6b). Detection function models were fit with a half-normal key function truncated to 4000 m, producing an eshw of 1760 m ( Figure 5b). The probability of sighting a humpback whale on the trackline was estimated to be p(0)=0.79 (CV=0.05). Visibility was included as a covariate in the DS model and weather was included in the MR model. The total estimate of humpback whales (corrected for perception bias) was 12,411 (CV=0.30, 95% CI: 6,847-22,497) (Table 7b).

2008-2013
Similar to the 2002-2007 survey, most of the sightings were made over the deep waters of the Norwegian Sea (EW1), south of Jan Mayen (CM1) (Figure 7b). A half-normal key function produced the best fit to the data truncated at 4000 m ( Figure  5c). The resulting eshw was 1964 m. Sperm whale abundance was corrected with p(0)=0.91 (CV=0.03) to a total corrected estimate of 3,962 (CV=0.29, 95% CI: 2,218-7,079). Detailed results by survey block are reported in Table 8b.

2002-2007
Sightings of killer whales occurred mainly in the Norwegian Sea south of the Mohn Ridge in block NOS ( Figure 8a). They were also abundant in the Icelandic/Jan Mayen survey blocks (NVN, NVS). The best fitting models used a half-normal key function. Data were truncated at 2000 m and resulted in an eshw of 996 m. BSS and weather covariates improved the fit of the DS model and group size improved the fit of the MR model ( Figure 9a).  The probability of sighting a killer whale on the trackline was p(0)=0.93 (CV=0.04) and the total corrected estimate was 18,821 (CV=0.24, 95% CI: 11,525-30,735). Detailed estimates by block are reported in Table 9a.

2008-2013
As in 2002-2007, most of the sightings were made in the Norwegian Sea (EW1, EW2) south of the Mohn Ridge. They were also abundant in the Icelandic/Jan Mayen survey blocks (CM1, CM3) ( Figure 8b). Models were fit with a half-normal key function ( Figure 10a). Distances were truncated at 2200 m, resulting in an eshw of 1377 m. BSS improved the fit of the MR model. Once corrected for perception bias (p(0)=0.92, CV=0.05) the total estimate for killer whales was 9,563 (CV=0.36, 95% CI: 4,713-19,403). Detailed estimates by block are provided in Table 9b.

2002-2007
Harbour porpoises were found in highest concentrations in the North Sea blocks NS and NC2 with additional concentrations in the Barents Sea (blocks KO and GA). They displayed a general shelf distribution within the study region and were absent from the western and northern-most survey blocks (Figure 11a). A half-normal key function with distances truncated to 600 m generated the best fitting models, resulting in an estimated eshw=279 m and p(0)=0.52 (CV=0.15) ( Figure 9b). The DS model    Table 10a.

2008-2013
Harbour porpoises were sighted most commonly in the Barents Sea (EB1, EB2) and the Norwegian Sea (EW1) and were completely absent from the western and northern-most survey blocks (Figure 11b). A half-normal key function with distances truncated to 500 m generated the best fitting models, with an eshw of 375 m. The proportion of harbour porpoises sighted on the trackline was estimated to be p(0) =0.36 (CV=0.49). Both the DS model and MR models included BSS as a covariate ( Figure  10b). The corrected harbour porpoise abundance was 38,351 (CV=0.58, 95% CI: 13,158-111,777). Detailed estimates by block are provided in Table 10b.

2002-2007
Lagenorhynchus spp. were found in almost all blocks within the study area, with the highest number of sightings around Bear Island (Figure 12a). A hazard-rate key function, without covariates, provided the best fit to the data from platform 1, which were truncated at a perpendicular distance of 1200 m. The detection function (Figure 9c) resulted in an eshw of 498 m and a total Platform-1 estimate of 213,070 (CV=0.18, 95% CI: 144,720-313,690). Block-wise estimates are detailed in Table  11a. As noted previously, the abundance was not corrected for perception bias.

2008-2013
Lagenorhynchus spp. were found throughout the survey area and were most commonly sighted around Bear Island (ES1) and the Barents Sea (EB4) (depicted in Figure 12b). A half-normal key function was used to fit the data (Figure 10c) Table 11b.

Other species
Other species recorded, for which abundance has not been estimated due to an insufficient number of observations, include blue whales, sei whales, and northern bottlenose whales. Their distributions are displayed in Figure 13. No sightings of pilot whales were made, but block EW4 near the Faroes, where they would be expected (Pike et al., 2019a(Pike et al., , 2019b, has not been covered in recent surveys.

Survey coverage
Ice coverage hampered effort in the northernmost regions of the study area. In 2002-2007, the entire SVI block was not surveyed due to ice. However, given that SVI accounted for only 2% of the total sightings (all species) in the previous survey period (Øien, 2009), the lack of effort in this area is not expected to have had a large effect on total abundance. In 2008-2013, ice also reduced the survey area coverage in the northern regions by 2.4%. Additionally, the EW4 block was not surveyed in 2008-2013 due to poor weather. However, the EW4 block was also not covered in the 2002-2007 survey, nor in the earlier 1996-2001 and 1995 surveys because it was not included as part of the SMAs under the minke whale RMP until 2003 (Øien, 2009;IWC, 2004).

Species identification
This study used survey methods specifically designed for minke whales (Skaug et al., 2004), which resulted in less optimal data collection for other species. The effective search half-width (eshw) for minke whales is in the range of half to one third of that for larger baleen whales. The designated search area for the observers was within 1500 m of the ship and observers were instructed to dedicate more of their effort to look for minke whales and also track them; thus, the detection of large whales was likely reduced by these patterns.
Some negative bias was likely introduced in the abundance estimates given that the surveys were conducted in passing mode and none of the sightings were closed upon. An examination of effective search half-widths for 'unidentified large whale' sightings, truncated at 4000m, resulted in estimates of 2107 m (CV=0.06) in 2002-2007 and 2509 m (CV=0.07) in 2008-2013, indicating that they are associated with greater sighting distances. It can therefore be assumed that the unidentified sightings do not bias the estimates proportional to their occurrence in the dataset. Additionally, an effort to improve identifications has reduced the proportion of 'unidentified large whales ' in 2002-2007 and 2008-2013 to 19%, down from 30% in 1996-2001 (Øien, 2009). We did not allocate unidentified whales to species based on their occurrence in the dataset. The effect of uncertainty in species identification could be measured in future surveys by including a confidence rating for each identification, which would allow for a sensitivity analysis of the magnitude of bias in species identification.

Pooling robustness
The detection functions and effective search widths were fitted over the complete survey region because most blocks did not yield enough sightings to allow separate detection functions to be fitted. This may lead to bias in the estimates for some blocks if the detection distances vary between blocks. The bias is hopefully low simply due to the consideration that the survey blocks with the highest estimates-and therefore the greatest vulnerability to bias-also had the greatest influence over the detection functions.

Availability bias
The corrected estimates account for perception bias by estimating for the values of p(0), but do not correct for availability, which may be a concern for sperm whales in this study. Given that sperm whales have long dive times (Drouot, Gannier, & Goold, 2004;Watkins, Moore, Tyack, 1985), they may remain submerged during vessel passage, and therefore undetectable. Availability bias is likely less of a concern for fin and humpback whales, which exhibit shorter dives (Dolphin, 1987;Panigada, Zanardelli, Canese, & Jahoda, 1999) and are therefore more likely to be detected within the window of time that they are in proximity to the ship. This should also not be a concern with small odontocete species because they tend to surface frequently and display conspicuous surface behaviour.

Duplicate judgement
Our methods for recording observations of non-target species-by recording only initial observations, without tracking-likely results in a higher level of uncertainty in judging duplicates compared to survey designs with tracking, such as the Buckland-Turnock (BT) method (Buckland & Turnock, 1992). The level of uncertainty is also likely higher in our surveys because the analyses rely heavily on post-cruise duplicate judgements and a largely subjective approach. Developing a more empirical and reproducible method, like that used for minke whales (Bøthun et al., 2009;Solvang et al., 2015), would reduce the potential error associated with judging duplicates. Additionally, including a confidence rating would allow for a sensitivity analysis of the effect of error in duplicate judgement.

Responsive movement
Responsive movement (i.e. when animals move toward or away from the ship before they are first detected), is a source of potential bias in any line transect survey studying cetaceans (Buckland et al., 2001). Movement toward the ship would result in a larger than expected number of sightings near the trackline (positive bias), whereas avoidance behaviour would have the opposite effect. Avoidance behaviour has been detected in harbour porpoises (Palka & Hammond, 2001), while whitebeaked dolphins have been shown to display both attraction and avoidance behaviour, depending on their distance from the observation platform (Hammond et al., 2002;Palka & Hammond, 2001). Given the designated search distance for minke whales in the survey (1500 m), it is possible that responsive movement could occur with small odontocetes before they are first detected.
Evidence for responsive movement in baleen whales is more mixed. A 2007 survey conducted in European waters found some evidence that fin whales were attracted to vessels (Macleod et al., 2009), whereas a similar survey in 2016 found no responsive movement (Hammond et al., 2017). Similarly, minke whale avoidance behaviour has been detected in some surveys (Palka & Hammond, 2001), but not in others (Paxton, Gunnlaugsson, & Mikkelsen, 2009). These findings suggest that responsive movement may be survey-specific and depend on region, vessel type, and possibly other factors. Our survey did not measure responsive movement; thus, there is likely some unaccounted-for bias, although the degree and direction are unknown.

Distance estimation
There is a large potential for bias in distance measurements in line transect surveys such as ours, which rely on naked-eye estimates of distance by trained observers (Leaper, Burt, Gillespie, & Macleod, 2010). Error of this type can bias abundance estimates by influencing the detection function models and affecting the identification of duplicate sightings (Buckland et al., 2001). Leaper et al. (2010) have demonstrated that both distance and angle errors make a substantial contribution to the variance of abundance estimates and may cause considerable bias. They also found that naked eye estimates were negatively biased, but non-linear in that observers tended to overestimate shorter distances and underestimate greater distances.
To mitigate error in distance estimation, observers received regular training using buoys as targets and newer observers were paired with more experienced observers. Observers also tested and trained their distance estimation skills opportunistically using floating objects (such as buoys and birds) by estimating their distance, then verifying distances with a stopwatch using the speed of the vessel (300m/min). Leaper et al. (2010) have shown that using measurements of distance to objects at the surface such as buoys, were not predictive of the actual biases found in measurements during the surveys. In future surveys, more could be done to reduce this type of error by incorporating a means of validating some proportion of the measurements, for example using cameras or reticle binoculars.

Distributional shifts
Given that the survey is conducted over a multi-year period any shifts in distribution between survey years and between survey blocks could have an effect on the abundance estimates. To reduce additional variance due to distributional shifts, the goal of the surveys is to cover each minke whale SMA within one survey year (Skaug et al., 2004). This was achieved in the 2008-2013 survey cycle. However, in the 2002-2007 cycle, some SMAs were surveyed over multiple years and within the SMAs, some blocks were surveyed twice (NOS, FI), increasing the potential for this type of variance. As a result, there may be additional variance in the minke whale estimates for the 2002-2007 survey due to the added potential for the duplication/omission of sightings between years. The block design is for minke whales; thus, constraining the area surveyed to a single SMA in a given year doesn't necessarily reduce additional variance for other species, although it may help for more regional species (such as small odontocetes) due to the fact that the minke whale SMAs are oceanographic regions with natural physical and biological distinctions.
Variance due to distributional shifts likely differs between species. Killer whales in the Norwegian Sea and Lagenorhynchus spp. in the Barents Sea, for example, are local populations with large home ranges and their distribution is likely to vary within and between seasons in relation to prey distribution (Christensen, 1982(Christensen, , 1988Øien, 1996). Other species like humpback whales, which are mostly migratory, show a generally consistent pattern of annual habitat use (Kennedy et al., 2013), but they can also display complex variation in distribution affected by larger climatological patterns as well as small-scale local effects (Keen et al., 2017;Visser, Hartman, Pierce, Valavanis, & Huisman, 2011).
Additional variance due to year-to-year shifts in distribution has been accounted for in minke whale estimates (Bøthun et al., 2009;Solvang et al., 2015). The estimates from prior synoptic and multi-year surveys and knowledge about population growth are used to model the random effects and estimate additional variance assuming a closed population based on genetic evidence and historic catch statistics. Corresponding information is not available for the non-target species that are locally abundant in smaller parts of the survey area.

Encounter rate variance
Variance in estimating encounter rate can be problematic for species other than minke whales, for which this survey was designed. Ideally, a transect design is stratified across a species' density in order to ensure precision in estimating the encounter rate variance (Buckland et al., 2001). The survey stratification was not considered for species other than minke whales, which may affect the precision of the estimates for other species. To aim for higher precision, a spatial modelling method could be applied to take spatial variation into account. This type of analysis has been shown to reveal habitat preferences of minke, fin and sperm whales and Lagenorhynchus dolphins (Skern-Mauritzen, ).

Harbour porpoise estimates and Beaufort Sea State
Typically for harbour porpoises, only survey effort at a BSS of 2 or less is used to estimate abundance, due to a rapid decline in detection at higher sea states (Barlow, 1988;Hammond et al., 2002). This approach was tested initially; however, our surveys exhibited a relatively high encounter rate at higher BSS compared to what has been observed in other multi-species surveys (e.g. Hammond et al., 2002) and lower variance when using total effort. As discussed at the NAMMCO Abundance Estimates Working Group meeting in October 2019, due to these factors it was agreed that total effort (BSS 4 or less) could be used for all of our survey cycles (NAMMCO, 2019). Given that the maximum sighting distance for harbour porpoises in these surveys was 600 m, and observers were asked to focus within a 1500 m range to detect minke whales, our survey method might generate reasonable abundance estimates for harbour porpoises.

Fin whales
The fin whale estimates for both surveys were very similar with a total abundance estimate of 10,004 (CV=0.  (Leonard & Øien, 2020).
The humpback whales in our study area are part of a much larger population with a continuous distribution across feeding areas around Iceland, Greenland, and Eastern Canada and US (Smith, 2010;Smith et al., 1999). Humpback whales increased in abundance in the feeding grounds around Iceland at a rate of 11% between 1986-2001 ). Since 2001, humpback whale distribution around Iceland has seen a shift to higher densities to the north of Iceland and a significant overall decline in density between (Pike, Gunnlaugsson, Sigurjónsson, & Víkingsson, 2020b. The 2015 NASS survey, covering a broader region around Iceland and the Faroe Islands, also found a lower abundance compared to a 2007 survey (Pike et al., 2019b). The increase we observe in our surveys between 2002-2013 (and a more recent survey (Leonard & Øien, 2020) may be a northeastward continuation of the trend initially documented around Iceland ). However, without further effort to track and identify the humpback whales observed in our study area, it is not possible to know whether the increase reflects population growth or immigration from other feeding areas.
The increase in abundance of humpback whales occurred largely in the Bear Island shelf area (BJ) and the Barents Sea.  (Øien, 2009). In past surveys, humpback whales were largely absent from the Barents Sea blocks (BAE, KO, GA) (Øien, 2009), but were abundant in our surveys, with a substantial increase between our 2 survey periods (summed block estimates: 1,358 in 2002-2007 and 3,220 in 2008-2013). These increases are likely related to ecosystem changes affecting the distributions of important prey species, as was concluded for the increase in abundance around Iceland (Pike et al., 2020b;Víkingsson et al., 2015). Large ecosystem changes have occurred in the Barents Sea during the past few decades, including collapses and subsequent recoveries of Atlantic herring and capelin and changes in herring over-wintering areas to areas of northern Norway (Gjøsaeter, Bogstad, & Tjelmeland, 2009). An analysis of ecosystem surveys coincident with the 2002-2007 whale sighting surveys found that fin, humpback, and minke whales were spatially associated with the northern polar front and northern prey species including krill, amphipods and polar cod (Skern-Mauritzen, Johannesen, Bjørge, & Øien, 2011).

Sperm whales
The distribution of sperm whales is generally consistent between survey periods, as they are reliably found in the central Norwegian Sea (NOS, NON, and NVS), associated with deep water of the Norwegian Sea Basin. The estimate for 2002CI: 5,695-11,617) is most comparable to the 1996-2001 uncorrected estimate of 6,375 (CV=0.22; CI: 4,163-9,762) (Øien, 2009). However, our estimate for 2008-2013 (3,962 CV=0.29; CI: 2,218-7,079) is lower and more in line with the synoptic survey conducted in 1995, which estimated 4,319 (CV=0.20; CI: 2,903-6,424) whales (Øien, 2009). Although the earlier surveys were not corrected for perception bias, we found the probability of sighting sperm whales on the transect line was quite high (81% (CV=0.06) in 2002-2007 and 91% (CV=0.03) in 2008-2013 (Table 4); thus, the comparisons to uncorrected estimates are reasonable.

Killer whales
The estimated abundance of killer whales for the 2002 was roughly double the estimate from the 2008-2013 survey of 9,563 (CV=0.36, 95% CI: 4,713-19,403). In general, the abundance of killer whales in the study region is not well understood; however, a photo-identification study conducted in the fjords of northern Norway estimated a total population size of 731 individuals in 1986-2003 (Kuningas, Similä, & Hammond, 2014). The Norwegian Orca Project has identified approximately 1000 unique individuals associated with the over-wintering herring in Tysfjord over a 20-year-period (Jourdain & Karoliussen, 2018). This is likely a fragment of the total population of the North Atlantic. The population has been previously approximated to be 7,000 animals in the northern North Sea and the Barents Sea up to Bear Island (NAMMCO, 1998), consistent with our 2008-2013 estimate, but much lower than that from the 2002-2007 survey.
The variation between survey-cycle estimates may be due to survey design factors, such as spatial variation in density or interannual shifts in distribution (described under Bias and estimation issues), which are not considered for non-target species. The 2002-2007 survey cycle covered blocks within the Norwegian Sea in different years, with some repeat effort. The variation in estimates may also be a natural phenomenon, possibly due to dynamic prey distributions, with which killer whales in the Norwegian Sea have been shown to be closely associated (Nøttestad, 2015). Earlier studies of killer whale abundance in the eastern North Atlantic between the Faroe Islands and East Greenland, conducted in 1987Greenland, conducted in , 1989Greenland, conducted in , 1995Greenland, conducted in and 2001, also found high variability among their estimates, which ranged from 4,413 to 26,774 (Foote et al., 2007). A recent report on the status of killer whales in the North Atlantic summarizes all of the estimates currently available in the North Atlantic (Jourdain et al., 2019). Our survey estimates will hopefully aid in clarifying the population status of North Atlantic killer whales.

Harbour Porpoises
The total corrected abundance of harbour porpoises in 2002, is reasonable given that a 1989 survey estimated a total uncorrected abundance of 93,612 (CV=0.22) (Bjørge and Øien, 1995) and our probability of sighting harbour porpoises on the track line was roughly 50% (p(0)=0.52 CV=0.15). The 1989 estimate was based on a partial single-platform survey.
The estimate of harbour porpoises in the 2008-2013 survey is much lower than our 2002-2007 estimate and older estimates (Bjørge and Øien, 1995 (Leonard & Øien, 2020). This aberrant estimate was discussed at the NAMMCO Abundance Estimates Working Group in October 2019 and it was agreed that the corrected estimate based on total effort could be accepted but that it should be treated as anomalously low and inconsistent with other estimates for the same area (NAMMCO, 2019).

Lagenorhynchus spp.
According to past surveys, approximately 90% of the Lagenorhynchus spp. are white-beaked dolphins (Øien, 1996). Past surveys have indicated that the population size of whitebeaked dolphins may be about 60,000-70,000 animals in the Barents Sea (Øien, 1996). Comparing similar areas in this study, the summed estimates in blocks BAE, GA, KO, BA2 and FI1 was 70,426 animals (single platform) in 2002-2007 and 65,234 animals (corrected) in 2008-2013 in the EB blocks. In the North Sea, however, our estimates show a large disparity from other survey estimates. We estimated 54,294 animals (single platform) for blocks NS andNC1 in 2002-2007 and23,773 animals (corrected) in 2008-2013 compared to the 2005 SCANS survey, which produced an estimate of ~10,000 animals (whitebeaked only) in roughly equivalent blocks (Hammond et al., 2013). These differences may be due to our estimates having high contributions to the CV from both encounter rate and group size, which have resulted in wide 95% confidence intervals for the North Sea blocks (Table 11).

ADHERENCE TO ANIMAL WELFARE PROTOCOLS
The research presented in this article has been done in accordance with the institutional and national laws and protocols for animal welfare that are applicable in the jurisdictions where the work was conducted.