Abundance of Whales in West and East Greenland in Summer 2015

An aerial line transect survey of whales in West and East Greenland was conducted in August-September 2015. The survey covered the area between the coast of West Greenland and offshore (up to 100 km) to the shelf break. In East Greenland, the survey lines covered the area from the coast up to 50 km offshore crossing the shelf break. A total of 423 sightings of 12 cetacean species were obtained and abundance estimates were developed for common minke whale, from now on called minke whale, (Balaenoptera acutorostrata) (32 sightings), fin whale (Balaenoptera physalus) (129 sightings), humpback whale (Megaptera novaeangliae) (84 sightings), harbour porpoise (Phocoena phocoena) (55 sightings), long-finned pilot whale, from now on called pilot whale, (Globicephala melas) (42 sightings) and white-beaked dolphins (Lagenorhynchus albirostri) (50 sightings). The developed at-surface abundance estimates were corrected for both perception bias and availability bias if possible. Data on surface corrections for minke whales and harbour porpoises were collected from whales instrumented with satellite-linked time-depth-recorders. Options for estimation methods are presented and the preferred estimates are: Minke whales: 5,095 (95% CI: 2,171-11,961) in West Greenland and 2,762 (95% CI: 1,160-6,574) in East Greenland, fin whales: 2,215 (95% CI: 1,017-4,823) in West Greenland and 6,440 (95% CI: 3,901-10,632) in East Greenland, humpback whales: 993 (95% CI: 434-2,272) in West Greenland and 4,223 (95%CI: 1,845-9,666) in East Greenland, harbour porpoise: 83,321 (95% CI: 43,377-160,047) in West Greenland and 1,642 (95% CI: 319-8,464) in East Greenland, pilot whales: 9,190 (95% CI: 3,635-23,234) in West Greenland and 258 (95% CI: 50-1,354) in East Greenland, white-beaked dolphins 15,261 (95% CI: 7,048-33,046) in West Greenland and 11,889 (95% CI: 4,710-30,008) in East Greenland. The abundance of cetaceans in coastal areas of East Greenland has not been estimated before, but the limited historical information from the area indicate that the achieved abundance estimates were remarkably high. When comparing the abundance estimates from 2015 in West Greenland with a similar survey conducted in 2007 there is a clear trend towards lower densities in 2015 for the three baleen whale species and white-beaked dolphins. Harbour porpoises and pilot whales however, did not show a similar decline. The decline in baleen whale and white-beaked dolphin abundance is likely due to emigration to the East Greenland shelf areas where recent climate driven changes in pelagic productivity may have accelerated favourable conditions for these species.

Greenland has not been estimated before, but the limited historical information from the area indicate that the achieved abundance estimates were remarkably high. When comparing the abundance estimates from 2015 in West Greenland with a similar survey conducted in 2007 there is a clear trend towards lower densities in 2015 for the three baleen whale species and white-beaked dolphins. Harbour porpoises and pilot whales however, did not show a similar decline. The decline in baleen whale and white-beaked dolphin abundance is likely due to emigration to the East Greenland shelf areas where recent climate driven changes in pelagic productivity may have accelerated favourable conditions for these species.

INTRODUCTION
Most cetacean species that occur in West Greenland are subject to various levels of subsistence hunting where frequent abundance estimates are required for assessing the effects of the exploitation. Furthermore, the waters around Greenland are located in a climate sensitive area and large-scale changes in the marine environment may eventually influence the presence and abundance of whales in coastal areas of Greenland.
Aerial surveys for large whales have been conducted at regular intervals in West Greenland since 1984. The first two surveys in 1984 and 1985 were conducted with the intention of obtaining uncorrected line transect estimates of the abundance of common minke whales, from now on called minke whales, (Balaenoptera acutorostrata) and fin whales (Balaenoptera physalus), however, too few sightings were obtained to generate estimates.
After 1985 surveys were conducted as combined cue counting and line transect surveys. The eight aerial surveys conducted between 1984 and 2007 each provided between 9 and 44 primary sightings of minke whales. Most sightings were of single individuals and sightings were widely dispersed on the banks of West Greenland. Given the difficulties in visually detecting minke whales it is unlikely that future surveys will obtain significantly more detections.

Surveys for fin whales have been conducted regularly between 1982 and 2007 in West
Greenland but only three surveys were useful for abundance estimation (Heide-Jørgensen et al. 2008, 2010b. In 1987/88 fin whale abundance was estimated at 1,100 (cv=0.35) from an aerial cue counting survey (IWC 1992). In 2005, the abundance was estimated at 3,234 fin whales (cv=0.44) from an aerial line transect survey with independent observers that allowed for correction of perception bias but not availability bias (Heide-Jørgensen et al. 2008). A ship-based survey also conducted in 2005 gave a smaller abundance estimate (1,980 fin whales, cv=0.38) than the aerial survey ). The estimated abundance of fin whales from the aerial survey in 2007 (4,468 fin whales, cv=0.68) was corrected for perception bias but not for whales that were submerged during the survey (availability bias; Heide-Jørgensen et al. 2010b). This study presents results from an aerial line transect survey of small and large cetaceans in East and West Greenland conducted in August and September 2015. Options for converting the at-surface density of whales to fully corrected total estimates of abundance are explored and applied to earlier partially corrected surveys. This requires the application of correction factors that adjust for whales missed by the observers and for whales that are not available to be detected at the surface.

Aerial survey technique
An aerial line transect survey of whales was conducted from 18-25 August in East Greenland and between 27 August and 30 September 2015 in West Greenland. The survey platform was a Twin Otter, with long-range fuel tanks and two pairs of independent observers all with access to bubble windows. Sightings, survey conditions and a log of the cruise track (recorded from an external GPS) were recorded on a Geospatial Digital Video Recording System (Remote GeoSystems, Inc., Colorado, USA) that also allowed for continuous video recording of the trackline. Declination angle (θ) to sightings was measured when animals were abeam with Suunto inclinometers and converted to perpendicular distance (x) using the equation from Buckland et al. (2001): x= v * tan(90-θ) where v is the altitude of the airplane.
Time-in-view was calculated as the difference in time between the first detection and the time the sighting passed abeam. Target altitude and speed was 213 m and 167 km hr -1 , respectively. Survey conditions were recorded by the primary observers at the start of the transect lines and whenever a change in sea state, horizontal visibility or glare occurred.
The survey was designed to systematically cover the area between the coast of West Greenland and offshore (up to 100 km) to the shelf break (i.e. the 200 m depth contour) by placing transects evenly across strata. Transect lines (n=124) in West Greenland were evenly placed in an east-west direction except for south Greenland where they were placed in a north-south direction (Fig. 1). In East Greenland 90 transects were designed to systematically cover the area from the coast over the shelf break up to 50 km offshore. The surveyed area was divided into 11 strata in West Greenland and 10 strata in East Greenland. Additional smaller strata covered selected fjord areas in West Greenland, where lines were placed following a zig-zag design when possible..

Instrumentation of minke and fin whales and harbour porpoises with satellite tags
Five minke whales and two fin whales were tagged with satellite linked time-depth-recorders in July and August 2013-2017 (Table 1). The whales were pursued from an open skiff with outboard engine operating in the waters off the town of Maniitsoq (65°25'N and 52°54'W) in central West Greenland.
Instrumentation was conducted by using the Air Rocket Transmitter System (ARTS) initially developed by Heide-Jørgensen et al. (2001a, b) and now widely used internationally for tagging baleen whales (Mate et al. 2007, Silva et al. 2013, Kennedy et al. 2014. The ARTS consists of an air gun with adjustable air pressure delivered by a scuba tank. The barrel of the ARTS is large enough to carry a plastic tube that acts as both a carrying rocket for the tag as well as a floatation device if the whale is missed. The 'rocket' is a cylinder with a finned tailpiece that provides stabilization during flight. The pressure and distance to the minke and fin whales when the rocket was launched was 12 bars and 5-10 m. A cylindrical stainless steel implantable tag (22x113 mm, 1 AA cell, Mk10A Wildlife Computers, Redmond, WA) was used. It was equipped with triangular pointed steel arrow and had a transmission repetition period of 45s and a conductivity switch that prevented transmissions when the tag was underwater. The tags were not duty cycled, but were restricted to a maximum of 350 or 500 daily transmissions.
Nine live captured porpoises were instrumented with satellite-linked radio transmitters (SPLASH tags, Wildlife Computers) in 2014 with the same timing and locations as the minke whales (Nielsen et al. in press). The tag was attached to the dorsal fin using three 5 mmdiameter delrin nylon pins, that were pushed through holes drilled in the fin with a sterilized cork borer mounted on a cordless electric drill (Heide-Jørgensen et al. 2017, Nielsen et al. in press).
In order to collect data on the time spent at the surface, the satellite-linked dive recorders were equipped with a pressure transducer (WC ver. 10.2) and software (WC ver. 1.25p) that allowed for sampling of pressure every second. Information on the percentage of time spent at seven depth bins (TAD: 0, 0-1, 1-2, 2-3, 3-4, 4-5, >5m) were collected in 1hr intervals for the minke and fin whales. To correct for drift the pressure transducer were calibrated when breaking the surface by the salt-water switch that also controlled transmissions to the satellites. Information on the percentage of time spent between 0 and 2 m depth were collected in 1hr intervals for the harbour porpoises.

Mark-Recapture distance sampling correction for perception bias
The search method deployed used an independent observer configuration where the primary and secondary observers acted independently of each other. Detections of animals by the primary observer serve as a set of binary trials in which a success corresponds to a detection of the same group by the secondary observer on the same side. The converse is also true because the observers are acting independently; detections by secondary observers serve as trials for the primary observers. Analysis of the detection histories using logistic regression allows the probability that an animal on the trackline is detected by an observer to be estimated, and thus, abundance can be estimated without assuming g(0) is one, i.e. no perception bias. These methods combine aspects of both mark-recapture (MR) techniques and distance sampling (DS) techniques and so they are known as mark-recapture distance sampling (MRDS) methods (Laake and Borchers, 2004). Probability of detection is modelled as a function of the perpendicular distance from the track line including covariates like sea state, observer, side of plane and group size.
Although observers were acting independently, dependence of detection probabilities on unmodelled variables can induce correlation in the detection probabilities. Laake and Borchers (2004) and Borchers et al. (2006) developed estimators that assumed that detections were independent at zero perpendicular distance only -called point independence modelsthat are well suited for aerial surveys where no responsive movements are expected.
Sightings that were detected by both platforms (i.e. duplicates) were identified based on coincidence in timing (<3s), distance to sightings (+200m), group size (+20%) and species identification. In the few cases where different species were identified the more experienced front observer identifications were used. For duplicates identified with certainty a mean of group size and distance from both platforms were used for the density modelling.
Heterogeneity in the detection probabilities can be reduced by including explanatory variables in the MRDS model. The explanatory variables available in this survey were perpendicular distance to sighting, group size, sea state and observers and they were included in the MRDS models with model selection criteria to select the best model. Detection probability was estimated using the independent observer configuration implemented in Distance 6.2 (Thomas et al. 2010). Model selection and selection of either a uniform or halfnormal detection function model was based on lowest AIC (Akaike Information Criteria).
Group abundance was estimated in each stratum using: where A is the stratum area (in km 2 ), L is the effort (in km) and w is the truncation distance (in m), is a vector of explanatory variables for group i (possibly including the group size, ) and is the estimated probability of detecting group i obtained from the fitted MRDS model. Individual animal abundance is estimated by: .
The estimated group size in the stratum is given by: Effective search width was estimated at survey level (East and West Greenland combined) except for minke whales where strip width was estimated at global level (East and West Greenland separately). Encounter rates were estimated by stratum and detection probabilies were pooled across all strata. For each species it is estimated if group sizes were estimated at a global level (across all strata) in either East or West Greenland or at stratum level.

Perception bias in strip census estimates
The sample size for minke whales in West Greenland was too low to allow for distance sampling estimation and instead strip census estimation of density, with a constant probability of detection within a species-specific strip width, was used. The strip census estimate was developed with an average group size across all strata. A Chapman estimate was used to correct for perception bias ( ) by the observers: where n is the total number of sightings, S 1 and S 2 are the sightings by observer platform 1 and 2 only and B is the sightings by both platforms (Magnusson et al. 1978). Variance of ( ) was estimated with Jackknife methods.
Individual abundance in stratum A was developed from: where G is the average group size and w is the strip width.

Correction for non-instantaneous availability
Whales are available for detection for a short period of time during aerial surveys (i.e. some whales may be seen ahead of the plane). Therefore the probability that an animal was available to be seen was greater than the proportion of time it spends at depths at which it is visible. Laake et al. (1997)  It is assumed that the whales were only available for detection when they were close to the surface (0-2m) and that the proportion of time spent ( ) close to the surface was known from satellite linked-data recorders. In order to account for this availability bias, corrected abundance (denoted by the subscript 'c') was estimated by: with estimated cv:

Estimation of time at surface and time in view for minke whale, fin whale and harbour porpoise
The average time spent at all 7 sampled depth bins for the minke whales show that the largest proportion (percentage) of each hour was spent at depths >5 m and relatively small fractions of time were spent at the surface (Table 1).
The weighted average time spent between the surface and 2 m depth was around 16% for the five minke whales that provided dive data during day light hours (9:00-18:00) for July through 27 September in West Greenland.

1
The weighted average time spent between the surface and 2 m depth was around 19.4% for the two fin whales that provided dive data during day light hours (9:00-18:00) for the period 23 July through 8 September in West Greenland.
Nine harbour porpoises provided data on surfacing time (0 m) from August-September 2014 (Table 1). Data collected during daytime (07:00-19:00) were used and they were collected in 6 hr intervals providing around 720 hrs of measurements from each porpoise (2x6x60=720).
The mean surfacing time of the nine whales was 19.0% (se=1.14).

Estimation of abundance
A total of 423 sightings, covering 12 species of cetaceans and a few polar bear sightings, where obtained. Six species were detected in low numbers (n<12) leaving 6 species as candidates for abundance estimation.

Minke whale abundance
Sightings of minke whales were widely distributed in both East and West Greenland but there was a decline in numbers towards north with few or no sightings in the northern strata ( Greenland (Tables 2 and 3).
The distribution of perpendicular distances of sightings (Figs 3 and 4) showed a large proportion of sightings close to the trackline indicating that there was not a blind spot for observers beneath the plane.
All minke whale detections were of single individuals and the explanatory variables available to be included in the MRDS models were a) perpendicular distance to sightings, b) sea state and c) observer. Effort and sightings made in sea state above 2 was excluded from the model (Fig. 3). Estimates of correction factors for perception bias were developed for both MRDS methods and by the Chapman estimator for the strip census estimates (Table 4).
Time-in-view for minke whales was on average 2.9 s for detection distances <450 m from the track line and 2.4 s for detections <300 m from the track line but the distribution were not significantly different (K-S, p=1), thus a common correction for availability bias (<450m) was applied (Table 5).
In order to maintain consistency with the survey conducted in 2007 in West Greenland using identical methods (Heide-Jørgensen 2010a), separate analyses were chosen for East and West Greenland. The low number of sightings (n=12) precluded an MRDS analysis of minke whales in West Greenland alone but a strip census estimate (truncated at 300m), similar to that used in 2007, provided a fully corrected estimate for West Greenland of 5,095 (95% CI: 2,171-11,961) minke whales. About half the abundance of minke whales in West Greenland was found in the southernmost stratum next to the East Greenland survey area (Table 6).
An MRDS analysis based on sightings combined from both East and West Greenland, truncated at 450 m and at sea state<3, excluded 4 observations, and provided a fully corrected (including availability bias) MRDS estimate of 2,762 (95% CI: 1,160-6,574) minke whales for East Greenland. An alternative analysis using only sightings from East Greenland and applying strip census estimation with an assumed strip width of 450m gives an estimate corrected for availability bias of 1,784 whales (cv= 0.43, 95% CI: 796-4,000). This estimate could not be reliably corrected for perception bias because all sightings were seen by observer 1 (hence there was no missing re-sightings for the front observer which seems unlikely based on previous double observer minke whale surveys).  Table 7).

Fin whale abundance
There were a few scattered observations of fin whales in West Greenland whereas large numbers were detected on the southern strata in East Greenland (Fig. 5).
The number of sightings of fin whales in sea states <5 was about 4 times higher in East Greenland compared to West Greenland. The number of sightings in West Greenland was similar to the survey in 2007 (Tables 2 and 3).
The MRDS analysis was right truncated at 700 m leaving 75 observations for analysis with an overall expected group size of 1.2 (cv=0.09) fin whales in East and West Greenland combined. The explanatory variables available to be included in the MRDS models were a) perpendicular distance to sightings; b) group size, c) sea state and d) observer. Effort and sightings made in sea state above 4 were excluded from the model (Fig. 6). The best model, based on the lowest AIC, did not include any variables. There was little perception bias in the combined East and West Greenland data (
The observed surface time for the fin whales tracked in West Greenland was 19.45% (Table   1) and the mean time-in-view of fin whale sightings was 11s and 4s in East and West Greenland, respectively (Table 5). Heide-Jørgensen and Simon (2007) (Table 5).

Humpback whale abundance
In West Greenland most humpback whales were detected in the southern strata, but densities were low compared to East Greenland where humpback whales were detected in large numbers all along the coast (Fig. 7). Still, the number of sightings in West Greenland was similar to that detected in the survey in 2007 (Table 3).
The number of sightings of humpback whales in sea states <5 was almost three times higher in East Greenland compared to West Greenland ( Table 2). The northernmost stratum in East Greenland (stratum E1, Figs 1 and 7) had 7 sightings of humpbacks but the coverage was restricted to one line where only half the line (23 km) was covered in good survey conditions.
Because of the biased coverage of stratum E1 abundance was not estimated for this stratum.
The overall expected group size of humpback whales was 1.35 (cv=0.09) in East and 1.53 (cv=0.16) in West Greenland.
The average time-in-view for both observers for humpback whales was 8 s (n=124).
Adjusting the average surface time (

Pilot whale abundance
Except for four sightings in East Greenland pilot whales were only detected in West Greenland (Fig. 11). More pilot whales groups were detected in 2015 than in 2007 (Table 3).
A right truncation at 700 m left 32 observations in sea states<5 for abundance estimations.
The expected group size was 8.5 whales (cv=0.10). A half-normal key with no covariates was chosen for the MRDS model ( Fig. 12) (Table 7). Pilot whales are however detectable ahead of the plane and applying an instantaneous availability correction factor leads to a positive bias. The average time-in-view for pilot whales was 6.4 s (cv=0.76) but data on the number of surfacings/dives are missing and adjustment of the availability bias for time-in-view is not

White-beaked dolphin abundance
White-beaked dolphins were widespread in both East and Southwest Greenland (Fig. 13) Table 4).

Hansen and Heide-Jørgensen (2013) used data from a single white-beaked dolphin from
Iceland to develop an availability correction factor (Table 1) and applying this to the atsurface abundance gave a fully corrected estimate of 15,261 white-beaked dolphins (95% CI: 7,048-33,046) in West Greenland and 11,889 (4,710-30,008 in East Greenland (Table 8). As for pilot whales the availability correction factor is not adjusted for time-in-view but for a comparison with the survey in 2007 it is useful to apply the same availability correction factor.

DISCUSSION
MRDS estimates that correct for perception bias are considered the most accurate estimates for all species as they take into account the perception bias and the heterogeneity that affects  (Table 1).
We therefore estimated abundance of minke whales using a strip census with an assumed The development of an adjustment of the availability correction for the time-in-view was based on cue-rates for the three baleen whales with fully corrected abundance estimates. For these species a cue is defined as any part of the body of the whale appearing at the water surface. There will however often be several cue's involved in a surfacing event (the period between ascent above 2 m and before next descend below 2 m). It is therefore likely that the impact of the time-in-view correction is overestimated and the abundance estimates negatively biased. The magnitude of the bias is difficult to assess but it could be as large as 1-2 percentage points of the availability correction factor. A more accurate correction factor could be developed if data on the average duration of surfacings and dives were available.

The fully corrected abundance estimates of fin whales between 2005 and 2015 from West
Greenland fluctuate more widely than the other species and this is mainly due to a major difference in the group size estimates that reached its maximum in 2007 with a mean observed group size of 2.5 (range 1-25) compared to 1.7 (range 1-13) in 2005 and 1.5 (range 1-7) in 2015. The large groups observed in previous years in West Greenland were not detected in West or East Greenland (mean=1.6, range 1-6) in 2015, and the changes are most likely due to ecological changes that has influenced the distribution of fin whales in the North Atlantic.
All of the three large whale species (minke, fin and humpback whale) show a remarkable decline in West Greenland since the last abundance estimates were obtained in 2007 (Fig.   16). The minke whale has decline from 10,000 to 5,000, the fin whale from 4,400 to 1500 and the humpback whale from 2,704 to 1,000. The same survey design and some of the same observers were used in the two surveys, and survey conditions were in both surveys kept to sea states below 3 for minke whales and below 5 for fin and humpback whales. Identical survey techniques were deployed in the two surveys and the decline cannot be attributed to a lower detection probability due to new observers in the 2015-survey because detections of large whales in East Greenland (that was covered first and could be considered as training) were still unexpectedly high. In addition, detection of harbour porpoises, which is the shelf areas is related to an increased summer SST with temperatures above 6 o C. Even though mackerel is not known to be a primary prey item for baleen whales it seems likely that the 1 dramatic increase in a pelagic prey resource may be driven by some of the same factors that are also driving baleen whale and dolphin distribution (e.g. krill, copepods. Myctophidae and herring, Clupea harengus). Large-scale ecological changes on both sides of Greenland are most likely driving the observed shifts in abundance of whales in the two areas.   Table 4. Number of sightings seen by each observer and the number of duplicates (seen by both observers). The 'Total' column shows the number of sightings seen by observer 1 and observer 2 with the sightings seen by both removed. CV's are given in parenthesis.  Table 5. Development of availability correction factors for minke, fin and humpback whales for different distance truncations for detections from 0 to 2 m depth. The cue rates were obtained from Heide-Jørgensen and Simon (2007). The beta distribution of the surface availability (see Table 1) was skewed to the right and restricted to avoid unrealistically low (i.e.<12%) surfacing times. A Kolomogorov-Smirnov test detected a significant difference for the time-in-view distributions of fin whales from East and West Greenland in 2015 (p=0.013) but no difference was found for humpback whales (p=0.872) or minke whales (p=0.604) for the same year.