We use cookies to collect anonymous data to help us improve your site browsing experience.

Click 'Accept all cookies' to agree to all cookies that collect anonymous data. To only allow the cookies that make the site work, click 'Use essential cookies only.' Visit 'Set cookie preferences' to control specific cookies.

Set cookie preferences

Your cookie preferences have been saved. You can change your cookie settings at any time.

Publication - Research and analysis

Production of Seabird and Marine Mammal Distribution Models for the East of Scotland

Published
18 November 2022
Directorate
Energy and Climate Change Directorate
Topic
Energy, Environment and climate change, Marine and fisheries
ISBN
9781804359389

This report describes temporal and spatial patterns of density for seabird and marine mammal species in the eastern waters of Scotland from digital aerial surveys. This is important in order for the Government to make evidence-based decisions regarding the status of these species and management.

View supporting documents Supporting documents

6. Results

Distribution models were produced for ten out of eleven seabird species and three out of four marine mammals. The list of covariates used in the initial models for each species in given in Table 7 and the selected covariates are listen in Table 8 for seabirds and Table 9 for marine mammals.

For seabirds for which it was possible to fit a separate model for breeding and non-breeding season, the model for non-breeding season retained more environmental variables than for breeding season. There was no single environmental covariate which would characterise distribution and density of all seabirds or all marine mammals.

Below we present the results each species: distribution patterns with point estimates and associated uncertainties (confidence intervals and coefficient of variation, CV)

6.1 Adjustments for non-recognition

Table 5 gives the adjustments made for each species due to non-recognition or species grouping. For species such as guillemot, 94% of observed numbers were assigned to species group, rather than species. For a small review of species grouping see Appendix 1.

Table 5. Allocation of vaguely identified animals.

Species

Adjustments

Non-recognised animals allocated

Proportion of total* numbers recorded

Northern fulmar

No adjustments made

0

0%

Northern gannet

No adjustment made

0

0%

Great skua

No adjustment made

0

0%

Common gull

Adjusted based on MERP database

41

30%

Lesser black-backed gull

Adjusted based on MERP database

57

63%

Herring gull

Adjusted based on MERP database

1347

42%

Great black-backed gull

Adjusted based on MERP database

603

28%

Black-legged kittiwake

No adjustment made

0

0%

Common guillemot

Adjusted based on MERP database

1583

94%

Razorbill

Adjusted based on MERP database

30

0.04%

Atlantic puffin

Adjusted based on MERP database

104

19%

Minke whale

Adjusted based on MERP database

6

16%

Common dolphin

Adjusted based on MERP database

0

0%

White-beaked dolphin

Adjusted based on MERP database

306

58%

Harbour porpoise

Adjusted based on MERP database

545

56%

*Knowns and estimated unknowns

6.2 Realised effort

The total effort (as areas) per year and per month is given in Figure 3. The total area surveyed was 16,882.47 km2. The distribution of effort by survey is shown in Table 6 and Figure 6. The was no surveys taking place in May, August and December and February and March were surveyed both in 2020 and 2021.

Figure 3. Total area covered (in km 2) by (left) surveyed year (2020 and 2021) and (right) by calendar month (numbered, Jan. = 1 etc.), including months when survey did not occur. March had the largest cover but January the lowest but note that March was included in 2020 and 2021 survey. N.B no data for May, August or December were recorded.
Table 6. Total area covered (in km 2) by each survey.

Survey

Time period

Covered area, as (km2)

1

February/March 2020

2181

2

April 2020

1608

3

June 2020

2180

4

July 2020

2192

5

September 2020

2185

6

October 2020

2176

7

November 2020

1884

8

February/March 2021

2534

Total

16940

Table 7. Initial models including list of environmental covariates for each species. All covariates were fitted as continuous variables (indicated by s()).

Most Birds

breeding season

Initial Model (on scale of link function)

s(Easting, Northing) + s(MeanSSTbymonth) + s(MeanRangeSSTbyMonth) + s(MeanSalinitybyMonth) + s(MeanRangeofSalinitybyMonth) + s(Current) + s(Depth) + s(Seabed Roughness) + s(Simpson Hudson Stratification Index )+s(Colony Index) + offset(log(Km2)),

Non- breeding season

Initial Model (on scale of link function)

s(Easting, Northing) + s(MeanSSTbymonth) + s(MeanRangeSSTbyMonth) + s(MeanSalinitybyMonth) + s(MeanRangeofSalinitybyMonth) + s(Current) + s(Depth) + s(Seabed Roughness) + s(Simpson Hudson Stratification Index ) + offset(log(Km2)),

Birds (reduced dataset (n = 5093)

Initial Model (on scale of link function)

s(Easting, Northing by Season) + s(MeanSSTbymonth by Season) + s(MeanRangeSSTbyMonth by Season) + s(MeanSalinitybyMonth by Season) + s(MeanRangeofSalinitybyMonth by Season) + s(Current by Season) + s(Depth by Season) + s(Seabed Roughness by Season) + s(Simpson Hudson Stratification Index by Season )+s(Colony Index by Season) + offset(log(Km2)),

Minke whale
Initial Model (on scale of link function)

s(Easting, Northing) + s(Dayofyear) + offset(log(Km2))*

Common dolphin
Initial Model (on scale of link function)

s(Easting, Northing) + offset(log(Km2))

White-beaked dolphin
Initial Model (on scale of link function)

s(Easting, Northing) + s(MeanSSTbymonth) + s(MeanRangeSSTbyMonth) + s(MeanSalinitybyMonth) + s(MeanRangeofSalinitybyMonth) + s(Current) + s(Depth) + s(Seabed Roughness) + s(Simpson Hudson Stratification Index ) + offset(log(Km2)),

Harbour porpoise
Initial Model (on scale of link function)

s(Easting, Northing) + s(MeanSSTbymonth) + s(MeanRangeSSTbyMonth) + s(MeanSalinitybyMonth) + s(MeanRangeofSalinitybyMonth) + s(Current) + s(Depth) + s(Seabed Roughness) + s(Simpson Hudson Stratification Index ) + offset(log(Km2))

*s(Depth) was subsequently tried.

Table 8. Selected models for seabirds.

Northern fulmar
Season

Breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + MeanMonthlySST + MonthlySalinityRange

No of segments

29291

Correlation structure*

AR1

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + MeanMonthlySST + SSTMonthlyRange + MeanMonthlySalinity

No of segments

67799

Correlation structure*

AR1

Assumed error

Negative binomial

Northern gannet
Season

All year

Final Model (on log scale, offset not shown)

s(Easting) + s(MeanMonthlySST)

No of segments

5093

Correlation structure*

ARMA(1,1)

Assumed error

Negative binomial

Great skua
Season

All year

Final Model (on log scale, offset not shown)

s(Easting, Northing by Season) + s(SSTMonthlyRange by Season)

No of segments

5093

Correlation structure*

ARMA(1,1)

Assumed error

Negative binomial

Common gull
Season

Breeding

Final Model (on log scale, offset not shown)

s(Depth)

No of segments

2723

Correlation structure*

ARMA(1,1)

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

None

No of segments

2370

Correlation structure*

ARMA(1,1)

Assumed error

Negative binomial

Lesser black-backed gull
Season

Breeding

Final Model (on log scale, offset not shown)

None

No of segments

29291

Correlation structure*

AR1

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

None

No of segments

67799

Correlation structure*

AR1

Assumed error

Negative binomial

Herring gull
Season

All year

Final Model (on log scale, offset not shown)

s(Easting, Northing by Season) + SSTmonthlyRange + s(Dayofyear) + s(Depth by Season)

No of segments

5093

Correlation structure*

AR1

Assumed error

Negative binomial

Great black-backed gull
Season

Breeding

Final Model (on log scale, offset not shown)

MeanMonthlySalinity + Seabed roughness

No of segments

29291

Correlation structure*

AR1

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + MeanMonthlySST + Depth

No of segments

67799

Correlation structure*

AR1

Assumed error

Negative binomial

Black-legged kittiwake
Season

All year

Final Model (on log scale, offset not shown)

s(Easting, Northing by Season) + s(MeanMonthlySalinity by Season)

No of segments

5093

Correlation structure*

AR1

Assumed error

Negative binomial

Common guillemot
Season

All year

Final Model (on log scale, offset not shown)

s(MeanMonthlySST) + s(Depth) + s(Dayofyear) + Seabed roughness

No of segments

5093

Correlation structure*

ARMA(0,1)

Assumed error

Negative binomial

Razorbill
Season

Breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + s(SSTMonthlyRange)

No of segments

29291

Correlation structure*

AR1

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + MeanMonthlySST + MeanSSTMonthlyRange + MeanMonthly Salinity

No of segments

67799

Correlation structure*

AR1

Assumed error

Negative binomial

Atlantic puffin
Season

Breeding

Final Model (on log scale, offset not shown)

s(Easting, Northing) + s(MeanMonthlySST) + MeanSSTMonthlyRange + MeanMonthly Salinity

No of segments

29291

Correlation structure*

AR1

Assumed error

Negative binomial

Non-breeding

Final Model (on log scale, offset not shown)

Current

No of segments

67799

Correlation structure*

AR1

Assumed error

Negative binomial

*The assumed structure of the correlation in the error ARn: autoregressive of lag n, ARMA Autoregressive (of lag n) and moving average (m)

Table 9. Selected models for marine mammals (offsets not given).

Species

Final Model (on scale of link function)

No of segments

Correlation structure*

Assumed error

Minke whale

s(Easting, Northing) +s(Dayofyear)

5093

AR1

Logistic

Common dolphin

None

52549

AR1

Negative binomial

White-beaked dolphin

s(Easting, Northing) + Month

5093

ARMA(1,1)

Negative binomial

Harbour porpoise

S(Easting, Northing) + MeanMonthlySalinity

5093

ARMA(1,1)

Negative binomial

*The assumed structure of the correlation in the error. ARn: autoregressive of lag n, ARMA Autoregressive (of lag n) and moving average (m)

6.3 Seabird Species

6.3.1 Northern Fulmar

The separate fitted models for breading and non-breeding season are given in Table 8. The estimates of numbers of fulmars during the survey period is given in Figure 4. They indicate peak abundance after the end of the breeding season, in September declining thereafter.

Point estimates of fulmar density for the sampling months along with the confidence bounds are given in Figure 5, Figure 6, and Figure 7. They show highest densities offshore in the northernmost part of the North Sea east of Shetland and Orkney south to the Moray Firth. The CVs are shown in Figure 8 and are largest at the southern areas of the study site. The mean point estimates for breeding and non-breeding season in shown in Figure 9. The breeding distribution is closer to the shore than in the non-breeding season.

The effect of the non-location variables, sea surface temperature (SST) and salinity, in the breeding season model, is given in Figure 10. They indicate a general increase in fulmar densities with increasing SST and decreasing salinity during this season.

The effect of the same non-location variables as well as salinity range in the non-breeding season model, is given in Figure 11. They showed positive relationships of fulmar density with both SST and salinity during the non-breeding season.

Note that these are the effects of these variables given the presence of location in the model, so they may be very different from the actual biological effect.

Figure 4. A graph showing estimated numbers of northern fulmars over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of fulmars peaked in September and had lowest values in winter months (January to April).
Figure 5. A graph showing point estimates of northern fulmar densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 6. A graph showing lower confidence bound estimates (2.5%) of northern fulmar densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 7. A graph showing upper confidence bound estimates (97.5%) of northern fulmar densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 8. A graph showing coefficients of variation ( CV, in %) in estimated densities of northern fulmars for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the peripheries of the study area, especially in the south.
Figure 9. A graph showing mean fulmar density surfaces for breeding (April – August) and non-breeding (September – March) seasons.
Figure 10. Graphs showing effect of (left) monthly sea surface temperature and (right) mean monthly salinity range on northern fulmar observed density assuming the middle of the survey area during the breeding season.
Figure 11. Graphs showing effect of (upper left) monthly sea surface temperature, (upper right) mean monthly salinity and (lower left) mean monthly salinity range on northern fulmar observed density assuming in the middle of survey area outside of the breeding season.

6.3.2 Northern Gannet

The best fit single model for the whole year is given in Table 8. Estimated numbers during the study period are depicted in Figure 12. They show a strong peak during the main breeding season between June and October.

Point estimates of gannet density for the sampling months along with the confidence bounds are given in Figures Figure 13, Figure 14 and Figure 15. They indicate a wide offshore distribution between June and October, with smaller numbers between November and March occurring closer inshore in areas such as the Moray Firth. The CVs are shown in Figure 16 and are largest along the east and west border of the study area. The mean point estimates for breeding and non-breeding season in shown in Figure 17. The breeding distribution of birds is closer to the shore than in the non-breeding season.

The effect of sea surface temperature range on density is given in Figure 18, and shows a general positive trend.

Figure 12. A graph showing estimated numbers of northern gannets over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to October) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of gannets peaked in July and September and had lowest values in winter months (January to April).
Figure 13. A graph showing point estimates of northern gannet densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 14. A graph showing lower confidence bound estimates (2.5%) of northern gannet densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 15. A graph showing upper confidence bound estimates (97.5%) of northern gannet densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 16. A graph showing northern gannet coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the eastern and western part of the study area.
Figure 17. A graph showing mean northern gannet density (birds/km 2) surfaces for breeding (April – October) and non-breeding (November – March) seasons.
Figure 18. A graph showing effect of mean monthly sea surface temperature on northern gannet observed density assuming the middle of the survey area during the breeding season.

6.3.3 Great Skua

The model based on the amalgamated data set is given in Table 8. The estimates of numbers of great skuas during the survey period is given in Figure 19. They indicate peak abundance during the breeding season. The results of this model should be treated with caution as the diagnostics of the model were not ideal.

Point estimates of great skua density for the sampling months along with the confidence bounds are given in Figures Figure 20, Figure 21 and Figure 22. They show highest densities in the far north-western part of the survey region east of Shetland. The CVs are shown in Figure 23 and are largest at the southern areas of the study site during the breeding season and at the peripheral areas of the study site outside the breeding season. The mean point estimates for breeding and non-breeding season in shown in Figure 24 The breeding distribution is concentrated at the northern site of the study area.

The effect of the non-location variable, sea surface temperature monthly range, in (red) and outside (black) the breeding season, is given in Figure 25.

Note that these are the effects given the presence of location in the model, so they may be very different from the actual biological effect.

Figure 19. A graph showing estimated numbers of great skuas over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of great skuas peaked in June and had lowest values in late autumn and winter months (November to March). The results of this model should be treated with caution as not all model assumptions were met.
Figure 20. A graph showing point estimates of great skua densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals. The results of this model should be treated with caution as not all model assumptions were met.
Figure 21. A graph showing lower confidence bound estimates (2.5%) of great skua densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. The results of this model should be treated with caution as not all model assumptions were met.
Figure 22. A graph showing upper confidence bound estimates (97.5%) of great skua densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. The results of this model should be treated with caution as not all model assumptions were met.
Figure 23. A graph showing great skua coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the peripheries of he study area outside the breeding season and in the southern art during the breeding season. The results of this model should be treated with caution as not all model assumptions were met.
Figure 24. A graph showing mean great skua density (birds/km 2) surfaces for breeding (April – July) and non-breeding (August – March) seasons. The results of this model should be treated with caution as not all model assumptions were met.
Figure 25. A graph showing effect of (mean monthly sea surface temperature range in (red) and out (black) the breeding season on great skua observed density assuming the middle of the survey area during respective season.

6.3.4 Common Gull

Predictions based a breeding season model is given in Table 8. No model was fitted to non-breeding season. Estimates of numbers in the area over the survey period are given in Figure 26. They show a general increase in abundance in the study area in the autumn. Results are constant within season.

Point estimates of common gull density for the sampling months along with the confidence bounds are given in Figure 27, Figure 28, and Figure 29. They indicate higher densities mainly off the coast during the breeding season. As the spatial pattern was uniform and consistent over the survey month for both breeding and non-breeding season, seasonal spatial patterns are not presented for this species as it can be deduced from Figure 27, Figure 28, Figure 29. The CVs are shown in Figure 30 and are largest at the central and norther part of the study site.

The effect of depth during the breeding season is shown in Figure 31.

Figure 26. A graph showing estimated numbers of common gulls over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of common gulls was lowest during the breeding season.
Figure 27. A graph showing point estimates of common gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals. As the spatial pattern in density was consistent and uniform outside the breeding season, the graphs show mean estimates for non-breeding season for each surveyed month within this season.
Figure 28. A graph showing lower confidence bound estimates (2.5%) of common gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. As the spatial pattern in density was consistent and uniform outside the breeding season, the graphs show mean estimates for non-breeding season for each surveyed month within this season.
Figure 29. A graph showing upper confidence bound estimates (97.5%) of common gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. As the spatial pattern in density was consistent and uniform outside the breeding season, the graphs show mean estimates for non-breeding season for each surveyed month within this season.
Figure 30. A graph showing common gull coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. As the spatial pattern in density was consistent and uniform outside the breeding season, the graphs show mean estimates for non-breeding season for each surveyed month within this season.
Figure 31. A graph showing the effect of depth on common gull observed densities assuming the middle of survey area. The effect of depth is estimated breeding season only.

6.3.5 Lesser Black-backed Gull

No variables were found that could predict the density of lesser black-backed gulls when the autocorrelation in the data was considered. The best estimate for abundance was 640 (95% CI: 280-1490). This species was only observed during five surveys and in very low numbers.

6.3.6 Herring Gull

The single (all year) fitted model is given in Table 8. Estimates of numbers in the area over the survey period are given in Figure 32. They show a general increase in abundance in the study area between September and March.

Point estimates of herring gull density for the sampling months along with the confidence bounds are given in Figure 33, Figure 34, and Figure 35. They indicate higher densities in the vicinity of the Moray Firth in October and November. The CVs are shown in Figure 36 and are largest at the peripheries of the study area. The mean point estimates for breeding and non-breeding season in shown in

Figure 37. The breeding distribution is closer to the shore in the non-breeding season.

The non-location spatial effects in all year model for herring gull are given below. They show greater densities in winter, and at greater depths during the breeding season but no relationship in the non-breeding season (Figure 38).

Figure 32. A graph showing estimated numbers of herring gull over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of herring gulls peaked in November.
Figure 33. A graph showing point estimates of herring gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 34. A graph showing lower confidence bound estimates (2.5%) of herring gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 35. A graph showing upper confidence bound estimates (97.5%) of herring gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 36. A graph showing herring gull coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the peripheries of the study area.
Figure 37. A graph showing mean herring gull density (birds/km 2) surfaces for breeding (April – July) and non-breeding (August – March) seasons.
Figure 38. Graphs showing effect of day of year (upper left), depth (upper right) and mean monthly sea surface temperature range (lower left) on herring gull density assuming the middle of survey area outside of the breeding season. The effect of depth differs dependent on whether it is the breeding (red) or non-breeding period (black).

6.3.7 Great Black-backed Gull

The separate fitted models for the breading and non-breeding season are given in Table 8. Estimated numbers through the study period are shown in Figure 39 and indicate relatively low numbers throughout the year but with a small peak between October and January.

Point estimates of gull density for the sampling months along with the confidence bounds are given in Figure 40, Figure 41, and Figure 42. There is high uncertainty in some peripheral regions of the survey region leading to the high upper bounds in Figure 42. Areas with higher densities occur east of Orkney and around the Moray Firth between October and February. The CVs are shown in Figure 43 and are largest at the southern areas of the study site outside the breeding season and in the centre of the study area during the breeding season. The mean point estimates for breeding and non-breeding season in shown in Figure 44. The breeding distribution is further off shore than in the breeding season.

The effect of salinity and seabed roughness in the breeding season model is given in Figure 45. Any salinity effect is weak compared to the effect of greatest seabed roughness. The effect of depth in the non-breeding season model is given in Figure 46 and shows higher densities at greater depths. The relationship with sea surface temperature is difficult to interpret and the model is quite possibly overfitting.

Figure 39. A graph showing estimated numbers of great black-backed gulls over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of gulls peaked in November.
Figure 40. A graph showing point estimates of great black-backed densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 41. A graph showing lower confidence bound estimates (2.5%) of great black-backed gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 42. A graph showing upper confidence bound estimates (97.5%) of great black-backed gull densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 43. A graph showing great black-backed gull coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at southern part of the study area during non-breeding season and at the centre during the breeding season.
Figure 44, A graph showing mean great black-backed gull density (birds/km 2) surfaces for breeding (April – July) and non-breeding (August – March) seasons.
Figure 45. Graphs showing effect of (left) monthly range of salinity and (right)seabed roughness on great black- backed gull observed density assuming the middle of survey area in the breeding season.
Figure 46. Graphs showing effect of (left) mean monthly sea surface temperature and (right) depth on great black-backed gull observed density assuming the middle of survey area outside of the breeding season.

6.3.8 Black-legged Kittiwake

The model for black-legged kittiwakes is given in Table 8. Estimated abundances of kittiwakes through the study period are given in Figure 47. They show a slight increase during the breeding season (April to July) but little change outside that period.

Point estimates of kittiwake density for the sampling months along with the confidence bounds are given in Figure 48, Figure 49, and Figure 50. They indicate greatest densities nearer to the coast around the Moray Firth and Firth of Forth. The CVs are shown in Figure 51 and are largest at the peripheries of the study site. The mean point estimates for breeding and non-breeding season in shown in Figure 52 and is comparable for these two seasons with highest estimates at the south-western part of the study area.

The effect of monthly mean salinity on kittiwake density is shown in Figure 53 and shows opposite trend for breeding (red) and non-breeding season (black) with increase in kittiwake density with decrease in salinity during the breeding season.

Figure 47. A graph showing estimated numbers of kittiwakes over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of kittiwakes peaked in June.
Figure 48. A graph showing point estimates of black-legged kittiwake densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 49. A graph showing lower confidence bound estimates (2.5%) of black-legged kittiwake densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 50. A graph showing upper confidence bound estimates (97.5%) of black-legged kittiwake densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 51. A graph showing kittiwake coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the peripheries of the study area.
Figure 52. A graph showing mean black-legged kittiwake density (birds/km 2) surfaces for breeding (April – August) and non-breeding (September – March) seasons.
Figure 53. Graph showing effect of monthly mean salinity on black-legged kittiwake observed density assuming the middle of survey area within (red) the breeding season and (black) outside the breeding season.

6.3.9 Common Guillemot

The single model (i.e. for the full sampling period) based on the amalgamated data is given in Table 8. The predicted numbers over the period of the survey are given in Figure 54. No stable model with a location effect could be found so the model consisted of main environmental effects only.

Point estimates of common guillemot density for the sampling months along with the associated confidence bounds are given in Figure 55, Figure 56, and Figure 57. They show greatest densities nearer to the coast in all months, but particularly between June and September from east of Orkney to the Moray Firth, and down the East Grampian coast as far as the Firth of Forth. The CVs are shown in Figure 58 and are largest at different locations over the study period. The mean point estimates for breeding and non-breeding season in shown in Figure 58. The breeding distribution is closer to the shore than in the non-breeding season.

The effects of the non-location variables are given in Figure 59. Note that these are the effects of these variables in the model given the other variables, so they may be very different from the actual marginal effect in the absence of the other variables. This is especially the case for SST which is correlated with Dayofyear (not shown).

Figure 54. A graph showing estimated numbers of common guillemots over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of guillemots peaked in June-September and had lowest values throughout non-breeding season.
Figure 55. A graph showing point estimates of common guillemot densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of birds with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 56. A graph showing lower confidence bound estimates (2.5%) of common guillemot densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 57. A graph showing upper confidence bound estimates (97.5%) of common guillemot densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 57. A graph showing common guillemot coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the north eastern part of the study area.
Figure 58. A graph showing mean common guillemot density (birds/km 2) surfaces for breeding (April – July) and non-breeding (August – March) seasons.
Figure 59. Graphs showing effect of (upper left) depth, (upper right) mean monthly sea surface temperature and (lower left) seabed roughness on common guillemot observed density.

6.3.10 Razorbill

The separate fitted models for breeding and non-breeding season are given in Table 8. The estimated numbers of razorbills in the survey region is given in Figure 60. It shows a strong peak during the breeding season between April and July.

Point estimates of razorbill density for the sampling months along with the confidence bounds are given in Figure 61, Figure 62, and Figure 63. They indicate greatest densities nearer the coast around the Moray Firth, east Grampian region and Firth of Forth, between April and July. There appears to be a hotspot of higher density offshore east of the Moray Firth in late summer and autumn, particularly September and October. The CVs are shown in Figure 64 and are larger in the non-breeding than breeding season. The mean point estimates for breeding and non-breeding season in shown in Figure 65 and shows higher densities in the breeding season at the south western part of the study site.

The effect of monthly mean sea surface temperature (SST) in the breeding season model is given in Figure 66, but note that there may be some overfitting of the model here. The effect of SST, salinity and salinity range in the non-breeding season model, is given in Figure 67. They show little obvious relationships although densities are possibly higher at greater sea surface temperatures and salinities. Note that these are the effects of these variables given the presence of location in the model, so may be very different from the actual biological effect.

Figure 60. A graph showing estimated numbers of razorbills over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of razorbills peaked in June and had lowest values in winter months (January to March).
Figure 61. A graph showing point estimates of razorbill densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of razorbill with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 62. A graph showing lower confidence bound estimates (2.5%) of razorbill densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 63. A graph showing upper confidence bound estimates (97.5%) of razorbill densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 64. A graph showing razorbill coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the northern and eastern part the study area.
Figure 65. A graph showing mean razorbill density (birds/km 2) surfaces for breeding (April – July) and non-breeding (August – March) seasons.
Figure 66. Graph showing effect of mean monthly sea surface temperature range on razorbill observed density assuming the middle of the survey area during the breeding season.
Figure 67. Graphs showing effect of (upper left) monthly sea surface temperature,(upper right) mean monthly salinity and (lower left) mean monthly salinity range on razorbill density assuming the middle of the survey area outside of the breeding season.

6.3.11 Atlantic Puffin

The separate fitted models for breading and non-breeding season are given in Table 8. Estimated numbers through the study period are given in Figure 68. They show a peak during the breeding season between April and July. Point estimates of puffin density for the sampling months along with the confidence bounds are given in the Figure 69, Figure 70 and Figure 71, For this species there were predicted areas of high density near the coast in regions with little survey effort. The CVs are shown in Figure 72 and are higher in breeding than non-breeding season. The mean point estimates for breeding and non-breeding season in shown in Figure 73 and indicate higher densities in the south-western part of the study area during the breeding season.

Greatest densities occurred between April and June particularly in the southern part of the study area, around the Firths of Forth and Tay and east Grampian coast, although high densities are predicted east of Caithness and the Northern Isles where survey effort was low.

The effect of the non-location variables in the breeding season model is given in Figure 74. The effect of current in the non-breeding season model is given in Figure 75. Note that these are the effects of these variables given the presence of location in the model, so may be very different from the actual biological effect.

Figure 68. A graph showing estimated numbers of puffins over the duration of the study from February 2020 to March 2021. Red points indicate the breeding season (April to July) and the dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of puffins peaked in June. High uncertainty is generated in peripheral regions in the non-breeding season.
Figure 69. A graph showing point estimates of puffin densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of fulmars with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 70. A graph showing lower confidence bound estimates (2.5%) of puffin densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 71. A graph showing upper confidence bound estimates (97.5%) of puffin densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 72. A graph showing Atlantic puffin coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the northern and eastern of the study area.
Figure 73. A graph showing mean puffin density (birds/km 2) surfaces for breeding (April – August) and non-breeding (September – March) seasons
Figure 74. Graphs showing effect of (upper left) monthly sea surface temperature, (upper right) mean monthly SST range and (lower left) mean monthly salinity on Atlantic puffin observed density assuming the middle of survey area during the breeding season.
Figure 75. Graph showing effect of current on Atlantic puffin observed density assuming the middle of the survey area outside of the breeding season

6.4 Marine Mammal Species

6.4.1 Minke Whale

There were only 35 datums out of the 5,036 reduced data set locations with minke whale presences, so a number based spatial model could not be fitted. A binomial presence-absence model was fitted and then numbers were estimated based on the mean number seen per presence. This model is not ideal. Availability was estimated at 0.04. Peak numbers occurred in June followed by a steady decline (Figure 76 Table 9).

Point estimates of minke whale density for the sampled months along with the confidence bounds are given in Figure 77, Figure 78, and Figure 79. The higher numbers observed in June occurred in two areas: in the north-western North Sea east of the Grampian region and Firth of Forth, and further east in the middle of the North Sea. No CVs were produced for this species.

Figure 76. A graph showing estimated numbers of minke whales over the duration of the study from February 2020 to March 2021. Dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of minke whales peaked in June.
Figure 77. A graph showing point estimates of minke whales densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of whales with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 78. A graph showing lower confidence bound estimates (2.5%) of minke whale densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 79. A graph showing upper confidence bound estimates (97.5%) of minke whale densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.

6.4.2 Common Dolphin

In the case of common dolphins, there were only 18 datums with common dolphin presences. A binomial model of presence was attempted on a reduced dataset from March to July (as no animals were seen outside this time) but no spatial signal could be found (Table 9). The estimates of abundance are therefore 0 outside March to July, otherwise 6170 (95% confidence interval 3530 – 10790), assuming an individual availability at the surface of 0.05.

Locations of sightings for common dolphins across the sampled months are given in Figure 80 but no model fits are done. Most sightings occurred in the month of June offshore in the middle of the North Sea although the species was also recorded further north, east of Caithness in March 2020.

Figure 80. A graph showing locations of sightings of common dolphins. No model was fitted to the data; hence the distribution model outputs are not presented. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of common dolphin with areas proportional to number.

6.4.3 White beaked Dolphin

The fitted model of all data is given in Table 9. A model with Dayofyear could not be fitted so instead Month was used as a factor variable. The estimated numbers from the model are given in Figure 81. Although there were sightings in several months of the year, they show a strong seasonal peak in July. Surface availability was assumed to be 0.06.

Point estimates of white-beaked dolphin density for the sampled months along with the confidence bounds and CVs are given in Figures Figure 82, Figure 83, Figure 84, and Figure 85. The models indicate a general distribution across the North Sea with highest densities further offshore.

Figure 81. A graph showing estimated numbers of white-baked dolphins over the duration of the study from February 2020 to March 2021. Dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of dolphins peaked in July.
Figure 82. A graph showing point estimates of white-beaked dolphins densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of dolphins with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals
Figure 83. A graph showing lower confidence bound estimates (2.5%) of white-beaked dolphins densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 84. A graph showing upper confidence bound estimates (97.5%) of white-beaked dolphin densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 85. A graph showing white-beaked dolphin coefficients of variation ( CV, in %) in estimated densities of dolphins for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month.

6.4.4 Harbour Porpoise

No model with a correlated error structure could be fitted to the n= 97090 data so the data were further amalgamated into 3-min chunks. This gave a total of 5093 datums. The model is given in Table 9. The predicted numbers over the period of the survey are given in Figure 86. These show broadly similar abundances through the year but with a peak between April and June.

The spatial density surfaces for the sampled months are given in Figure 87, and Figure 89 along with associated uncertainty. They indicate broad distributions for the species in most months, with highest offshore densities east of the Moray Firth and Grampian region.

The effect of salinity on density is given in Figure 91, suggesting there may be higher density where freshwater has a greater influence.

Figure 86. A graph showing estimated numbers of harbour porpoises over the duration of the study from February 2020 to March 2021. The dashed lines represent upper and lower bounds of the 95% confidence intervals. Numbers of porpoises peaked between April and June.
Figure 87. A graph showing point estimates of harbour porpoise densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month. Red dots indicate observed numbers of porpoises with size proportional to observed number. Note that scale is matching the following graphs depicting lower and upper confidence intervals.
Figure 88. A graph showing lower confidence bound estimates (2.5%) of harbour porpoise densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 89. A graph showing upper confidence bound estimates (97.5%) of harbour porpoise densities for each surveyed month from February 2020 to March 2021. Colours represent estimated densities per km 2. Black lines indicate sampling locations in that month.
Figure 90. A graph showing harbour porpoise coefficients of variation ( CV, in %) in estimated densities of birds for each surveyed month from February 2020 to March 2021. Black lines indicate sampling locations in that month. The largest CVs are at the peripheries of the study area.
Figure 91. A graph showing effect of mean monthly salinity on surface porpoise observed density assuming the middle of the survey area.

Contact

Email: REEAadmin@gov.scot

Back to top