A short review of the distribution of short-beaked common dolphins ( Delphinus delphis ) in the central and eastern North Atlantic with an abundance estimate for part of this area

This paper uses data from 3 programmes: (1) the North Atlantic Sightings Surveys (NASS) surveys undertaken throughout much of the central and eastern North Atlantic north of about 40° N in 1987, 1989, 1995 and 2001; (2) the MICA-93 programme; and (3) the north eastern Atlantic segment of the Small Cetacean Abundance in the North Sea (SCANS) survey in 1994. The data from all surveys were used to examine the distribution of common dolphins in the NE Atlantic. No sightings were made north of 57° N. An initial attempt to examine distribution against 4 potential non biological explanatory variables was made. A simple interpretation of the preliminary analyses presented here is that the primary areas for groups of common dolphins were in waters over 15° C and depths of 400-1,000 m (there does appear a link with shelf features), between around 49°-55° N especially between 20°-30°W. An illustrative example of spatial modelling is presented. Only for 1 year (and part of the total survey area) were there sufficient data to attempt to estimate abundance: 1995. The estimated abundance in the W Block of the NASS-95 Faroese survey was 273,159 (cv=0.26; 95% CI=153,392-435,104) short-beaked common dolphins. This estimate is corrected for animals missed on the trackline (g(0)) and for responsive movement. Cañadas, A., Donovan, G.P., Desportes, G. and Borchers, D.L. 2009. A short review of the distribution of short-beaked common dolphins (Delphinus delphis) in the central and eastern North Atlantic with an abundance estimate for part of this area. NAMMCO Sci. Publ. 7:201-220.

A short review of the distribution of short-beaked common dolphins (Delphinus delphis) in the central and eastern North Atlantic with an abundance estimate for part of this area INTRODUCTION Although short-beaked common dolphins (Delphinus delphis) are known to occur in many areas of the North Atlantic (e.g.Perrin 2002), there have been few large scale systematic dedicated cetacean surveys to allow ocean basin estimates of their abundance to be obtained or even to allow their distribution to be described fully.In the north eastern Atlantic, the available information suggests that they are frequently encountered, most typically appearing in oceanic and shelf edge waters (Forcada et al. 1990, Goujon et al. (MS) 1993, Hammond et al. 2002, Harwood and Wilson 2001, Lopez 2003, Silva and Sequeira 2003).However, the overall distribution of the species and the biological and non biological factors affecting this, have been little studied.Abundance has been reported for only a few discontinuous areas (e.g.Goujon et al. (MS) 1993;Hammond et al. 2002, O'Cadhla et al. 2004), representing a patchy and sparse coverage of the distribution range.Stock structure is poorly understood.Since the beginning of the 1990s, concerns over the conservation status of the species in the area have been raised as a result of documented by catches, mainly in trawl and driftnet fisheries (Goujon et al. (MS) 1993, Lopez et al. 2003, Morizur et al. 1999, Silva and Sequeira 2003, Tregenza and Collet 1998).Without better information on abundance, stock structure and total removals it is difficult to assess the impact of by catches at the population level (e.g.see Hall and Donovan 2002).Better information on the distribution of short-beaked common dolphins and the distribution of fishing effort may serve to identify potential 'hot spots' for by catches and enable a more focussed examination of the issue.
This paper examines the available data on short-beaked common dolphins from the NASS (North Atlantic Sightings Surveys) multi year, multi national survey programme that covered a large part of the north-eastern and central North Atlantic over several years (Víkingsson et al. 2009), supplemented with data from 2 other survey programmes of lesser geographical scope, the MICA-93 (Goujon et al. (MS) 1993) and the SCANS-94 (Hammond et al. 2002) surveys.Most of the effort of the SCANS-94 programme was in the North Sea.It uses these to (1) review and expand upon what is known about the distribution of common dolphins in the central and north eastern Atlantic, (2) present a preliminary examination of the non biological factors that may influence that distribution and (3) provide an abundance estimate of common dolphins from the Faroese vessel operating as part of the NASS-95 programme, following on from the work of Cañadas et al. (2004).

Survey design and data collection
As noted above, this paper uses data from 3 programmes: (1) the NASS surveys undertaken throughout much of the central and eastern North Atlantic north of about 40° N in 198740° N in , 198940° N in , 199540° N in and 200140° N in (Víkingsson et al. 2009); (2) the MICA-93 programme (Goujon et al. (MS) 1993); and (3) the SCANS survey in 1994 (Hammond et al. 2002).The areas covered are shown in Fig. 1.Details of the survey procedures can be found in the relevant papers.The important points to notice are that all of the surveys were dedicated cetacean line transect surveys following a random cruise track design and that they all collected standard line transect sightings, weather and effort data.A summary of the vessels used and other pertinent information is provided in Table 1.
The completed 'on effort' cruise tracks are shown in Fig. 2, stratified by Beaufort Sea State (BSS) (0-2, 3+).However, it should be noted that the primary target species of the cruises varied (see Table 1) and this influenced inter alia choice of survey blocks, 'acceptable' search conditions and observer strategy (e.g.single versus double platform, naked eye versus binoculars, etc.).The implications of these differences for the results obtained in this paper are considered in the Discussion.

Distribution and relative abundance
Only data for sightings of confirmed species identity were considered.In examining distribution and relative density, the following data, where available, were used: position of sighting, group size, BSS at the time of sighting.In addition, account was taken of the overall levels of effort by sea state where this was available.Sea state is known to be an important factor that affects the ability of observers to detect cetaceans, including common dolphins (e.g.see Cañadas et al. 2004).In the simplest instance, plots of sightings against completed survey track were made to examine distribution (see Fig. 5).
As an initial attempt to begin to try to explain the observed distribution patterns, 2 crude indices of abundance were calculated: (1) encounter rate expressed as number of groups per nm; and (2) encounter rate expressed as number of animals per nm.These indices of abundance were examined against 4 of the many potential non-biological explanatory variables that may influence distribution: position (latitude and longitude); depth; and sea surface temperature.These were chosen because (a) there were some data available and (b) they have been implicated as being important factors in the distribution of other cetacean species (e.g.Cañadas et al. 2005, IWC 2006).The total area bounded by 42°-57° N (there were no sightings north of 57° N) and 1°-43° W was divided into 0.5°×0.5°squares and the average depth, latitude and longitude of the midpoint, and (for those cells for which it was available i.e.only north of 52° N) average sea surface temperature for July (1995 and 2001) were calculated (see Figs 3 and 4).
The variables listed above were stratified into suitable bins and encounter rates were then calculated as the average encounter rate for the grid cells included in each category (see Table 2).
Choice of bin was chosen by inspection of the data and available information on common dolphin distribution from other sources; sensitivity to the choice of bins was examined but the results are not presented here as they were not found to be significant.Interested readers can contact one of us (Cañadas) if they wish to receive more information.Average group sizes were calculated also for each category of the 4 variables; the sample sizes are slightly smaller than shown in Table 2 as 16 sightings had no group sizes recorded.Only grid cells with more than 10 nm sailed on effort were used in the analysis this value represented a compromise between having sufficient effort to consider it representative and to maintain adequate sample size.The use of this criterion resulted in  a reduction of accepted sightings from 298 to 273.The results were examined for effort in all sea states and separately for BSS 0-2 and 3+.In this paper, the results are presented for all sea states combined only.However, if there is a changing pattern by sea state, this is mentioned in the Results section.Again, interested readers can contact one of us (Cañadas) if they wish to receive more information.Given the preliminary nature of the analysis and the limi- tations of the data, we did not believe a complex statistical analysis was appropriate and thus restricted ourselves to a largely qualitative examination of graphical data.However, in order to provide an example as to how a more sophisticated analysis could be undertaken given more data, a simple spatial analysis (e.g.see discussion in IWC, 2006) is included as Appendix 1. (1) The cv of the abundance estimate was obtained using a non parametric bootstrap procedure, in which segments were the sampling units, with 1,000 resamples (Cañadas et al. 2004).

RESULTS
A total of 27,160 nm (50,300 km) were surveyed on effort (of which 33.6% were carried out in BSS) 0 to 2, 57.8% in BSS 3 to 5, and 2.8% in BSS 6 to 7) during the 6 years of survey considered in this paper (1987, 1989, 1993, 1994, 1995 and 2001).A total of 14,607 (27,052 km) were carried out within the grid cells analysed here.
A total of 298 sightings of confirmed common dolphin groups were recorded with an average group size of 15 animals per group (±2.2, range 1-239).There were 77 sightings classified as 'maybe common dolphin' or 'unidentified dolphin'; these were excluded from the analysis.
All confirmed sightings are plotted in Fig. 5.

Abundance estimate
The only data suitable for obtaining an absolute abundance estimate that have not been previously analysed were those from the Faroese vessel Miđvingur that took part in the 1995 NASS survey and used a double platform methodology; details of that survey including the important methodological details of the approach used to estimate density taking into account animals missed on the trackline and the movement of the dolphins towards the ship before detection are given in Cañadas et al. (2004).It is not appropriate to repeat those details here.Several transects were defined within each block: 2 for block W and 10 for block E, and effort was calculated for each transect.The extent of each block is shown in Fig. 1.Each transect was divided into segments of approximately 20 nautical miles (nm) each for bootstrapping purposes (non parametric bootstrap with replacement).
The estimated total number of animals in the survey area, N, is obtained by taking the estimated density D from Cañadas et al. (2004) and A, the surface area of the survey area:

Position
Despite considerable effort to the north (see Fig. 5), there were no sightings of short-beaked common dolphins north of around 57° N. The most northerly sighting in the surveys covered in this paper was at 56°45' N.
Fig. 9a shows the encounter rate of groups stratified into 6 latitudinal bins.South of around 49° N, encounter rates were lowest (around 0.01±0.001groups nm -1 ).However, they increased significantly (at the 5% level) after this before maintaining a relatively stable value of around 0.03 (±0.002) groups nm -1 between 49° N and 55° N and then declining to 0.023 (±0.003) north of 55° N. By contrast, encounter rates of animals (Fig. 9c) shows a pronounced and significant peak (of around 0.8±0.27animals nm 1) between 51-53° N.This reflects the second significant peak in average group size (26.4±5.72 animals) between 51-53° N an earli- er peak, (31.7±9.86 animals) but with larger SE occurred in the most southerly bin (42-45.5°N).In other latitudes, the average group size (Fig. 9b) was relatively consistent at around 8-10 animals with small SEs (0.68-1.46).Fig. 10a shows the encounter rate of groups stratified into 4 longitudinal bins.Encounter rates of groups were highest (0.03±0.003 groups nm -1 ) between 20-30° W and lowest (0.005±0.0007 groups nm -1 ) in the most easterly bin (at the 5% level).East of 20° W, the rates were reasonably constant at around 0.02±0.002groups nm -1 .The pattern is somewhat different for encounter rates of animals (Fig. 10c) where the variation with longitude is much less pronounced, although there is a slight but insignificant peak at 20 -30° W. This reflects, in particular, the much higher average group size (49.5±27.21animals).East of 30° W, the average group size (Fig. 10b) was consistent at about 12 -15 animals (SEs=1.24-2.79).

Depth
Fig. 11a shows the encounter rate of groups stratified into 5 depth range bins.The lowest encounter rates (<0.02±0.002groups nm -1 ) were found in shallow waters (0 400 m) and waters over 2,000 m (significant at the 5% level).Encounter rates were highest (0.057±0.011 groups nm -1 ) between 400 and 1,000 m followed by 1,000 2,000 m (0.038±0.006 groups nm -1 ).A similar pattern was found for encounter rates of animals (Fig. 11c).Average group sizes showed a significant increasing trend with depth, from 8.0±1.44 animals in waters <400 m up to 18.6±2.76animals in waters >2,000 m (Fig. 8b).

Sea surface temperature(SST)
Fig 12a shows the encounter rate of groups stratified into 4 SST bins.As noted above, SST data were only available for some of the waters north of 52° N (Fig. 4).This has the effect of both limiting the sample size (see Table 2) and restricting the analysis to SSTs between 8° and 16° C.Only 1 group was seen in the coolest bin (8°-12°C) and that was a group of 3 animals in water of 11.6°C at 54°10' N, 35°26' W. Encounter rates of groups in waters between 12°-15° C were stable at around 0.03±0.005groups nm -1 but increased sharply to 0.07±0.01groups nm -1 for the warmest (15°-16° C) bin.A similar gener- al increasing trend with temperature was found for both encounter rate of animals (up to a peak of 1.28±0.41animals nm -1 ) in the warmest bin (Fig. 9c), and for group size (Fig. 9b) with around 8-10 (SEs=0.78-1.15) animals for waters <15° C increasing up to around 21.8±3.38 animals for waters of 15°-16° C (significant at the 5% level).

Abundance estimates
A total of 74 primary sightings of common dolphins were recorded, 25 in block E and 49 in block W (see Table 3 and Fig. 13).The results of the analysis including bootstrapping with 1,000 resamples are shown in Table 3 and the mean abundance estimates obtained from the bootstrapping are close to the point estimates (Cañadas et al. 2004).The limitations and implications of these estimates are discussed in the following section.For the reasons discussed there, we believe that only the abundance estimate for the western block can be considered reliable.

General limitations of the NASS datasets in the context of this study
The advantage of the datasets examined is that they were all dedicated cetacean surveys following recognised design principles for line transect surveys.However, it should be noted that, for most of the NASS surveys, the primary target species were large cetaceans (see Table 1).Whilst the instructions to observers would of course be to record all cetacean sightings, the different priorities have the potential to affect the obtained results for non target species such as the common dolphin in a number of ways e.g.
(1) survey areas and blocks may be non-optimal for common dolphins; (2) surveys may continue in weather conditions that are sub optimal for common dolphins (possibly leading to an over-representation of larger and/or more visible groups);  et al. (2004).Where appropriate, values of the mean (Mean bs ) and coefficients of variation (cv bs ) after bootstrapping are given, together with the 95% confidence intervals (CI) calculated using a simple percentile approach (Cañadas et al. 2004).
The truncation distance was 0.3° nm.Note: for reasons given in the discussion, only the abundance estimate for Block W is considered satisfactory.(3) time may not be allocated to confirming species identity and/or group size, particularly where animals are not seen close to the vessel (possibly leading to bias in the perpendicular distances to confirmed schools); (4) data reporting (e.g. of angle or distance to sighting) may be less reliable if observers are told or believe that small cetacean sightings are of low priority; (5) in high density 'target species areas', nontarget species may be overlooked or not recorded, in order to maximise high quality data for the target species.
The importance of consideration of these factors to any conclusions drawn will, of course, depend on the use to which the data are being put.A brief examination of the group size data by sea state (see Table 4) did not suggest that (2) above will be particularly problematic for discussions of general distribution and group size considerations; over 90% of the sightings were made in BSS 4 or less with just over half being made at BSS 2 or less.With respect to item (3), some indication of this can be given by examining the numbers of unidentified dolphins reported: 77 sightings were classified as 'maybe common dolphin' or 'unidentified dolphin'; 298 confirmed sightings were made.Item (4) is clearly only of importance if abundance is being investigated (for example, a problem was encountered with the distance measurements in the 1989 dataset from the vessel Investigador, where the reticule readings were wrongly stored) but in general, it is difficult to assess how reliable are the data on angle and distance for non target species without data.Similarly, it is difficult to examine whether item ( 5) is indeed a problem and so this possibility should be borne in mind in the discussion below.Given inter alia the preliminary nature of the analyses regarding distribution, we believe that it was appropriate to use all of the data for this purpose.
After considering the potential for bias arising out of consideration of the above factors, we believe that only the Faroese NASS-95 double platform data are suitable for abundance estimation.The primary target species of that vessel was the pilot whale rather than large whales and thus the potential for bias arising from the above points is minimised.Any limitations of the data with respect to abundance estimation were also considered in Cañadas et al. (2004).Despite the limitations of the available SST data (the reduced sample size of dolphins due to the large areas where no SST data were available also meant that a less than optimal averaging of the SST data for 2 different years was needed), the results (Fig. 12) suggest that temperature (presumably via its effect on prey species) may be an important factor in at least determining the northerly limit of distribution.The absence of SST data for the more southerly (and presumably warmer) part of the area does not allow investigation of how far the trend of increasing encounter rates and average group size with temperature may continue.This would be a valuable exercise and inclusion of SST data would also strengthen any future spatial analyses (see below).
Inspection of Figs 6 and 7 shows that shortbeaked common dolphins are found across the North Atlantic between about 50°-57°N at least as far as around 35° W. Effort west of 35° W was relatively low in the present study although some extended out to around 40° W.These data are not in accord with an apparent hiatus shown at 20° W by Perrin (2002).The apparent westerly limit or hiatus may be related in some way to temperature since the westernmost sightings were also in the coolest (ca 11° C) waters.The apparent 'gap' in distribution between around 53°-57° N and 25°-34° W may be a function of the relatively low effort.However the 'gap' between 42-48° N and 12°-18°W appears real given the high level of effort there.The reason for this gap is unclear but does not appear to be related to any obvious depth feature.
A simple interpretation of the preliminary analysis presented here is that the primary areas for groups of short-beaked common dolphins were in waters over 15° C and depths of 400-1,000 m (there does appear a link with shelf features), between around 49°-55° N especially between 20°-30° W. Group sizes interestingly showed a strong increase with depth (over 2,000 m) that may imply social strategies related to feeding or social behaviour.

Spatial analysis
Spatial analysis is an increasingly powerful and flexible method for examining density, distribution and abundance of cetaceans (e.g.Hedley et al. 1999, Marques 2001, Cañadas et al. 2005, Cañadas and Hammond, 2006) that can cope better with non-systematic survey data and 'filling in' gaps in effort; however, it is 'data' hungry and the simple analysis given in Appendix 1 suffers for that reason.At this stage, the limitations of the data preclude a full analysis that properly takes into account a wider vari-ety of factors (e.g. a spatial trend in detectability with sea state, further examination of the mean group size surface and predictive relative group density surface).As noted above, it was presented for illustrative reasons as a possible future approach rather than as a serious study at this stage.Despite this, a comparison of the results of the simple analysis presented in the Appendix is interesting.In general, the predicted relative densities (Appendix Fig. 3) from the spatial modelling show similar patterns to the more 'traditional' plots (e.g.see Figs 6 and  7), with relatively high densities in similar areas between 50°-55° N.However, in particular it does not allocate as much weight to depth (see Appendix Figs 1-2) as one might have expected from Fig. 11.This is most noticeable along the shelf edge from the Bay of Biscay to the southwest of Ireland where encounter rates of both groups and animals were relatively large but are not incorporated into the predicted densities.

Abundance estimate(s)
The abundance estimates presented here are the first for this part of the North Atlantic that incorporate corrections for g(0) and responsive movement; as shown by Cañadas et al. (2004), the latter is particularly important and failure to consider it can result in severe negative bias in some surveys.Although we have presented the results for block E, we do not believe that this can be considered a reliable abundance estimate for a number of reasons: (1) the fact that the realised transects lie in the middle of the stratum giving inadequate spatial coverage; (2) the large differences between the designed and the realised cruise tracks; and (3) the realised cruise tracks lay roughly parallel to the depth contours (Cañadas et al. 2004).In addition, there were relatively few sightings.Given that, we believe that it is only appropriate to present a final abundance estimate for the western block, where although the completed tracks were largely in the east of that block, they seemed representative of the block as a whole and the sample size was high.Thus the estimate of abundance of common dolphins in block W (52°-57.5° N, 18°-28° W) is 273,000 (cv 0.26, 95% CI 153,000-435,000).
There have been several other estimates of abundance from dedicated surveys in this area (Fig. 14).However, they are not strictly compa-rable.Only the NASS-95 and SIAR (O'Cadhla et al. 2004) surveys corrected for g(0) not allowing for this would lead to a negative bias.Only the NASS-95 estimate accounts for responsive movement; if similar movement occurred in the other surveys their estimates would be biased upwards.Both of these biases will be ship specific.
During 2005, a new survey was carried out on the shelf waters of the north-eastern Atlantic, the SCANS II project (Hammond and MacLeod (MS) 2004).This survey accounts both for g(0) and for responsive movement.Data from this survey are being now analysed and the results will be then comparable with the NASS-95 estimate.

CONCLUSIONS
This paper has extended our knowledge of the distribution of short-beaked common dolphins in an area previously poorly studied.It has also highlighted the need for more data both in terms of sightings data and appropriate explanatory variables if a more sophisticated spatial modelling analysis of distribution is to be carried out.It has also presented the first abundance estimate a previously unknown area based on NASS-95 data for the summer period.
Valuable as this is, from a management perspective (e.g.see Hall and Donovan 2002), it is important that future work focuses on: (1) obtaining a better understanding of stock structure; (2) the undertaking of additional surveys that take advantage of the available information presented here and elsewhere with a more appropriate stratification and appropriately focused on small cetaceans; (3) better estimates of anthropogenic removals.

Fig. 1 .
Fig.1.Boundaries of the study area for the NASS surveys and bathymetry.A dark line has been plotted over the 200 m depth contour, and a dark dashes line over the 1000 m depth contour.The blocks E and W of NASS-95 and the blocks for MICA-93 and SCANS are also shown.

Fig. 2 .
Fig. 2. Cruise tracks for all surveys considered in this paper.Thick line corresponds to effort carried out with BSS 0 to 2. Thin dashed line corresponds to BSS 3 to 5.

Fig. 5 .
Fig. 5. Cruise tracks and sightings of all surveys considered in this paper

Fig. 7 .
Fig. 7. Schematic plot of encounter rate of animals by 0.5° square.Solid lines indicate effort with BSS 0 to 2. Dashed lines indicate effort with sea states of 3 or more.

Fig. 8 .Fig. 9 .
Fig. 8. Schematic plot of group sizes by 0.5° square.Solid lines indicate effort with BSS 0 to 2. Dashed lines indicate effort with BSS of 3 or more.Encounter rate of groups

Fig. 10 .Fig. 11 .
Fig. 10.Encounter rates of groups (a) and animals (c) and average group sizes (b) of common dolphins for 4 bins of longitude.Vertical bars show standard errors.

Fig. 12 .
Fig. 12. Encounter rates of groups (a) and animals (c) and average group sizes (b) of common dolphins for 4 bins of sea surface temperature.Vertical bars show standard errors.
Shapes of the functional forms for smoothed co-variables used in the model of density of groups.Zero on the vertical axis corresponds to no effect of the co-variables on the estimated response.The locations of the observations are plotted as small ticks along the horizontal axis.Shapes of the functional forms for smoothed co-variables used in the model of group sizes.Zero on the vertical axis corresponds to no effect of the co variables on the estimated response.The locations of the observations are plotted as small ticks along the horizontal axis.

Table 1 .
Vessel characteristics, primary target species, survey dates and effort by BSS for the surveys considered in this paper (where available).

Table 2 .
A summary of the data available for examining the encounter rates and explanatory variables by bin size.Sightings refers to sightings of groups.The total number of cells under 'All Effort' is usually greater than the sum of those for BSS 0-2 and 3+ since not all of the datasets included information on sea state.The number of potential cells for the temperature analysis is lower because data are only available north of 52° N.

Table 3 .
Estimates of the various parameters required to estimate abundance for the blocks, E, W and the combined area covered by Miđvingur in 1995.The density estimates, accounting for responsive movement and g(0) are from Cañadas

Table 4 .
Average group size by recorded BSS