You are on page 1of 14

Bird Study

ISSN: 0006-3657 (Print) 1944-6705 (Online) Journal homepage: https://www.tandfonline.com/loi/tbis20

Evaluating the Breeding Bird Survey for producing


national population size and density estimates

Stuart E. Newson , Rick J.W. Woodburn , David G. Noble , Stephen R. Baillie &
Richard D. Gregory

To cite this article: Stuart E. Newson , Rick J.W. Woodburn , David G. Noble , Stephen R. Baillie
& Richard D. Gregory (2005) Evaluating the Breeding Bird Survey for producing national population
size and density estimates, Bird Study, 52:1, 42-54, DOI: 10.1080/00063650509461373

To link to this article: https://doi.org/10.1080/00063650509461373

Published online: 29 Mar 2010.

Submit your article to this journal

Article views: 1808

View related articles

Citing articles: 70 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tbis20
Bird Study (2005) 52, 42–54

Evaluating the Breeding Bird Survey for producing


national population size and density estimates

STUART E. NEWSON1*, RICK J.W. WOODBURN1, DAVID G. NOBLE1, STEPHEN R. BAILLIE1


and RICHARD D. GREGORY2
1BritishTrust for Ornithology, The Nunnery, Thetford, Norfolk, IP24 2PU, UK and 2Royal Society for the
Protection of Birds, The Lodge, Sandy, Bedfordshire, SG19 2DL, UK

Capsule The BBS has potential for producing better estimates of habitat-specific densities and population
sizes for many UK bird populations than those available previously.
Aims To examine the use of the Breeding Bird Survey (BBS) in providing unbiased national population
and habitat-specific density estimates of British birds.
Methods Line transect data collected by volunteers in 1998 from 2287 1-km squares across the UK were
analysed using distance sampling methods to calculate habitat-specific density and abundance estimates.
For each species, the habitat-specific decline in detectability with distance from a transect line was
modelled and applied at a regional level to incorporate variation in sampling intensity in different areas
of the country.
Results National population and density estimates calculated here were at a magnitude expected for at
least seven species in this study. However, national population size estimates were higher than expected
for Starling Sturnus vulgaris, House Sparrow Passer domesticus, Blackbird Turdus merula, Greenfinch
Carduelis chloris, Jackdaw Corvus monedula, Whitethroat Sylvia communis, Woodpigeon Columba
palumbus and Linnet Carduelis cannabina and lower than expected for Skylark Alauda arvensis,
Dunnock Prunella modularis, Song Thrush Turdus philomelos and Corn Bunting Miliaria calandra. These
differences are likely to be related to differences in sampling design and survey methods. For example,
Starling, House Sparrow, Blackbird, Jackdaw, Greenfinch and Wood Pigeon, which have considerable
populations in urban areas, were undoubtedly underestimated by the Common Birds Census (CBC). The
counts of species that flock during the breeding season, or are not strongly territorial (e.g. Linnet, Jackdaw
and Wood Pigeon) could be biased if detectability is strongly related to flock size. National population
estimates of Skylark and Corn Bunting are lower than those based on CBC, but higher, or similar, to
targeted national surveys of these species. Possible reasons for the differences between these estimates
are considered.
Conclusions This study highlights the strength of the BBS over previous data sources in producing
national estimates of density and abundance at the habitat and national level. More research on the sex
ratio and status of birds counted during surveys, and on the reliability of the detectability functions derived
from distance sampling is needed to improve the interpretation of population estimates derived from BBS
data.

The British Trust for Ornithology (BTO)/Joint Nature reported annually (Noble et al. 2001, Baillie et al.
Conservation Committee/Royal Society for the 2002). In order to make the maximum use of the data
Protection of Birds Breeding Bird Survey (BBS) was collected, these analyses use all bird registrations.
introduced in 1994 to monitor population trends of a However, through the recording of birds in different
broad range of breeding birds in the UK. Annual distance categories, the BBS was also designed to
population indices are calculated for all species where enable an assessment of variation in detectability and
there is sufficient coverage, and population trends are hence to derive estimates of density and population
size.
*Correspondence author. Email: stuart.newson@bto.org For most common and widespread species in the UK,

© 2005 British Trust for Ornithology


Using BBS to estimate density and population size 43

the New Atlas of Breeding Birds in Britain and Ireland: ulations mainly occur on farmland and for which there
1988–91 (Gibbons et al. 1993) provides the most recent is sufficient coverage to produce annually updated
estimates of density and national population size. population trends.
However, these estimates have a number of limitations.
The Atlas uses the mean density of bird territories on
METHODS
woodland and farmland Common Birds Census (CBC)
plots surveyed in 1989 to derive habitat-specific density
Data collection
and hence population size estimates for these habitats.
Because the CBC allows volunteers to select their own We use BBS data for 1998 during which fieldwork was
plots in mainly farmland and woodland, there are too undertaken on 2287 1-km squares. Full details of the
few plots to calculate density estimates for habitats survey design can be found elsewhere (Gregory et al.
other than woodland and farmland, so density and pop- 1996, Gregory & Baillie 1998). In brief, volunteers
ulation size estimates for these other habitats are based were provided with squares following a stratified
on observed densities in woodland and farmland. CBC random sampling design. Initially, the number of
coverage is also biased to the south and east of Britain, squares allocated to each of the BTO’s 83 regions
although there is some attempt to correct for this in the (roughly counties or groups of counties) was a fixed
calculation of population size. A major advantage of the proportion of the number of potential volunteers in the
BBS over the CBC is that it employs a formal sampling region, estimated using BTO membership information.
design with 1-km survey squares of the National Grid Within each region, squares are selected randomly and
selected at random, within each of 83 regions of the UK allocated to volunteers through a network of voluntary
and is stratified by observer density to allow participa- regional organizers (ROs). Fieldwork involves three
tion to vary across regions without introducing bias into visits by a volunteer observer to each survey square.
the analysis. The first is to determine two parallel ideal 1-km line
Secondly, an analysis of BBS and CBC data since
1989 shows that the long-term declines reported for
Table 1. A list of species in the analyses and information of their
many widespread farmland birds (Noble et al. 2001) status within the UK.
have continued over this period of time. This suggests
that national population estimates such as those in Population
Stone et al. (1997) based on data from the last Breeding change
Species Scientific name Code 1970–98 (%)1
Bird Atlas (Gibbons et al. 1993), regardless of its
limitations, are no longer appropriate, and need to be Kestrel 2,5 Falco tinnunculus K. –17
updated. This is particularly true for species included in Grey Partridge 2,4,6 Perdix perdix P. –82
the national Biodiversity Action Plan (BAP) Lapwing 2,5 Vanellus vanellus L. –52
Wood Pigeon 2 Columba palumbus WP 201
programme, such as Grey Partridge, Skylark, Song Turtle Dove 2,4,6 Streptopelia turtur TD –77
Thrush, Tree Sparrow, Linnet, Corn Bunting and Reed Skylark 2,4,6 Alauda arvensis S. –52
Bunting. Table 1 lists all the scientific names. Blackbird 3 Turdus merula B. –29
Relative habitat-specific densities were previously Song Thrush 3,4,6 Turdus philomelos ST –55
Whitethroat 2 Sylvia communis WH 34
estimated for eight species monitored by the BBS Jackdaw 2 Corvus monedula JD 148
during 1995 (Gregory & Baillie 1998). The percent- Starling 2,4 Sturnus vulgaris SG –58
ages of the British populations of sparrows, finches and Dunnock 3,5 Prunella modularis D. –21
buntings (10 species) in broad habitat classes were also House Sparrow 4 Passer domesticus HS –58
Tree Sparrow 2,4,6 Passer montanus TS –87
estimated from BBS data for 1996 (Gregory 1999). Greenfinch 2 Carduelis chloris GR –2
However, neither of these studies explores the uncer- Linnet 2,4,6 Carduelis cannabina LI –38
tainties that may bias estimates of this type, or presents Bullfinch 3,4,6 Pyrrhula pyrrhula BF –40
absolute population size estimates. Yellowhammer 2,4 Emberiza citrinella Y. –43
Reed Bunting 2,4,6 Emberiza schoeniclus RB –52
We extend these studies to examine the potential of Corn Bunting 2,4,6 Miliaria calandra CB –85
BBS data for producing unbiased estimates of habitat-
specific densities and population size for the UK. We 1Data from Common Birds Census (CBC). 2A farmland species indi-
use as examples, data from 20 species that commonly cator as per Gregory et al. (1999). 3A woodland species indicator
as per Gregory et al. (1999). 4Red-listed in Gregory et al. (2002).
use farmland. Species (shown in Table 1) were selected 5Amber-listed in Gregory et al. (2002). 6Priority species in the UK
to cover all widespread species of birds whose pop- Government’s Biodiversity Action Plan.

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


44 S.E. Newson et al.

transects through the square and two 1-km line computerized data and subsequent analyses. Thus, the
transects that will actually be surveyed. These actual density and population estimates calculated in this
line transects should follow the ideal as closely as paper relate to breeding, or potentially breeding adults
possible, but allow for minor deviation due to impracti- only.
calities for the observer caused by land features or other
obstacles. The habitats of each 200-m section along
Distance sampling analysis
each of the ideal and actual line transects were also
recorded on this first visit using a hierarchical coding Distance sampling software developed by Buckland et al.
system (Crick 1992). 2001 (DISTANCE, Version 3.5; Thomas et al. 1998) was
Early morning bird counts are carried out on the used to produce estimates of density and population size
second and third visits, between early April and mid- using BBS data. This technique models the decline in
May and between mid-May and late June respectively. detectability of birds with distance from the transect line
It is recommended that the two visits should be at least in order to include an estimate of undetected individuals
4 weeks apart. Counts begin at 06:00–07:00 hours while estimating real density and hence population size
(GMT) so that they coincide with maximum bird of the species of interest. As recommended by Buckland
activity but avoid concentrated song activity at dawn. et al. (2001), birds recorded in the final distance band
Observers record all birds that they see or hear as they (100 m or more) were excluded from the analyses, as
walk along their transect routes within each 200-m counts within an unbounded category are difficult to
section in one of three distance categories from the line interpret. Birds in flight were also excluded. Count data
(0–25 m, 25–100 m, 100 m or more) or in flight. Flying from the two remaining bands (0–25 m, 25–100 m) were
birds obviously associated with the square (e.g. display fitted with a half-normal detection function in all cases,
flight of Skylark) are assigned to the appropriate as recommended for binomial data of this kind
distance band rather than recorded as in flight. (Buckland et al. 2001).
Distances are estimated at right angles to the transect For each species, the decline in detectability with
line. Juvenile and immature birds are recorded in the distance was modelled for each of 20 broad habitat
field, but as far as possible they are excluded from the classes (Table 2). This was done because one might

Table 2. Habitat classes used in the analyses. Habitat classes are based on Crick (1992).

Code Name Description

WOOD1 Broadleaved woodland Broadleaved woodland


WOOD2 Coniferous woodland Coniferous woodland
WOOD3 Mixed woodland Broadleaved and coniferous (min. 10% of each)
SCRUB Scrub Young regenerated woodland, downland scrub, heath scrub, young coppice, young
plantation and clear-felled woodland (<5 m in height)
GRASS1 Dry semi-natural grassland Chalk downland, grass moor (unenclosed), grass moor mixed with heather and other dry
grassland
GRASS2 Wet semi-natural grassland Machair, water meadow, grazing marsh, reed swamp, saltmarsh and other open marsh
HEATH Heath Dry heath, wet heath, mixed heath and breckland
BOG Bog Bog, drained bog and bare peat
FARM1 Improved grass farmland Grass that has been regularly treated with fertilizer and distinguished by its bright colour,
lush growth and even texture
FARM2 Unimproved grass farmland Grass that has not been treated with fertilizer but is usually grazed or mown and may be
rank and neglected
FARM3 Mixed grass/tilled farmland Mixture of grassland and tilled farmland
FARM4 Tilled farmland Arable farmland including standing crop, plough and bare earth awaiting cultivation
FARM5 Other farmland Orchard and other farmland not specified above
URBAN Urban human sites Land relating to human habitation in a city or town
SUBURB Suburban human sites Land relating to human habitation on the outskirts of a city or town
RURAL Rural human sites Land relating to human habitation in the countryside, e.g. a village or hamlet
WATER Freshwater bodies Freshwater pond, lake, reservoir, stream, river, canal and ditch
COAST Coastal Marine open shore, inlet, cove, loch, estuary, brackish lagoon and open sea
ROCK Inland rock Cliff, scree/boulder slope, limestone pavement, other rock outcrop, quarry, mine
spoil/slag heap and cave
MISC Miscellaneous Sites not categorized above

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


Using BBS to estimate density and population size 45

expect large differences in detectability between Table 3. A list of the UK level 1 NUTS (Nomenclature of Territorial
Units for Statistics) regions of the European Community used in the
habitats. For example, birds are likely to be detected at
analyses. The table also includes details of the number of squares
greater distances from the transect line in open country surveyed and the land-area in each region.
than they are in woodland or scrub. Habitat-specific
detectability curves were then applied to count data Squares Area % UK
Code Name surveyed (km2) by area
from BBS squares assigned to one of 12 NUTS
(Nomenclature of Territorial Units for Statistics) 71 North England 94 15 853 6.2
regions of the European Community (see Table 3) to 72 Yorkshire/Humberside 138 15 178 6.0
calculate a regional estimate of density for each habitat 73 East Midlands 154 16 112 6.3
74 East Anglia 130 13 091 5.1
and, by doing so, controlling for broad differences in
75 Southeast England 573 28 317 11.1
sampling intensity in different areas of the country. 76 Southwest England 295 24 408 9.6
BBS count data were pooled across bird recording visits 77 West Midlands 174 12 800 5.0
(early and late) to produce a mean density estimate 78 Northwest England 152 7 653 3.0
79 Wales 192 21 657 8.5
over these two visits. Counts were assumed to be a
7A Scotland 296 85 024 33.4
collection of sightings of individual birds rather than 7B Northern Ireland 82 13 991 5.5
taking the approach of modelling the detectability of 7C Channel Islands 7 261 0.1
flocks (see later for discussion of these assumptions). Total 2287 254 345 100
Although fitting habitat-specific detection curves is
preferable to the fitting of a single curve, this proved
difficult where there were few observations of birds in were calculated using a bootstrap resampling procedure
particular habitats. Since these species were considered of 400 iterations (Crowley 1992). For each iteration,
very rare in these habitats (e.g. Yellowhammer in urban squares were randomly resampled within region from a
habitat), it was considered appropriate to remove these list of squares in the original data with replacement. The
data (17 of 400 habitat–species combinations, habitat data and bird count data for the new sample of
Appendix) from the analyses. The resulting estimates squares forms the new data set of the iteration. The
of density and abundance were considered to be much entire process of analyses that was used to calculate the
more robust because of these exclusions. density and population estimates detailed above, is
Habitat-specific population estimates were calcu- repeated for each new sample to produce estimates by
lated by multiplying habitat-specific density estimates habitat across the UK and by NUTS region. We calcu-
for each NUTS region by the area represented by that lated 95% confidence intervals of the original estimates
habitat in each region. The area of each habitat within from the 400 new density and population size estimates
each NUTS region was calculated from the proportion for each habitat and for the total UK estimates. We cal-
that each habitat comprised of the total area of each culated 95% confidence intervals in a similar way for
region (see Table 3). The habitat data that the observer density and population size estimates by NUTS regions,
would have recorded had they walked the ideal transect but these are not presented here.
line through the square were used in these calculations Previous estimates of national population size in the
as this represents the true random sample of the region last Breeding Atlas or targeted national surveys were
(Field & Gregory 1998). based on the number of territories or pairs. To provide
National population estimates were calculated as the a broadly comparable measure of population size to the
sum of regional habitat-specific population estimates. numbers of adults estimated here, previous estimates
On the rare occasion that a habitat recorded by the based on territories or pairs are doubled.
ideal transect across all BBS squares was not surveyed
by the actual transect, a mean density estimate for that
RESULTS
habitat across NUTS regions was substituted as this
represents our best estimate for that habitat class. A
National population estimates
mean density of each species in each habitat across the
UK was estimated by dividing the habitat-specific National population size estimates with 95% confi-
national population estimates by the estimated total dence intervals for species derived from BBS data are
area of that habitat in the UK. Regional estimates of summarized in Table 4. We also present population size
density and population size are not considered here. estimates for these species from the literature. Of the 20
Confidence intervals for the estimates for each species species in this study, estimates for eight species

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


46 S.E. Newson et al.

Table 4. Comparison between previous national population size estimates (number of territories/pairs x 2) and new national population esti-
mates for 1998 calculated here with bootstrapped 95% confidence intervals. Figures are rounded to the nearest 10 individuals.

Estimates based on last Breeding Atlas


Species unless otherwise stated New estimates 95% CI

Kestrel 100 0001 77 420 62 040–156 130


Grey Partridge 300 000 265 040 229 630–746 320
Lapwing 370 000–476 000 436 310 325 390–574 290
Lapwing 265 450 2,3 436 310 325 390–574 290
Wood Pigeon 5 100 000 7 895 750 7 312 790–8 464 750
Turtle Dove 150 000 69 150 54 100–377 330
Skylark 4 000 000 2 649 650 2 350 650–2 939 730
Skylark 1 475 560–2 155 350 4 2 649 650 2 350 650–2 939 730
Blackbird 8 800 000 9 836 130 9 292 440–10 400 800
Song Thrush 1 980 000 1 233 160 1 124 520–1 351 030
Whitethroat 1 320 000 1,575,390 1 429 87–1 814 250
Jackdaw 780 000 2 504 340 2 211 950–3 016 970
Starling 2 200 000 9 231 150 8 181 740–10 458 580
Dunnock 4 000 000 3 350 010 3 138 410–3 632 590
House Sparrow 5 200 000–9 200 000 12 973 830 11 813 040–14 384 650
Tree Sparrow 220 000 283 980 218 490–760 400
Greenfinch 1 060 000 3 034 410 2 788 710–3 323 630
Linnet 1 040 000 2 258 410 2 015 950–2 658 200
Bullfinch 380 000 392 910 359 790–727 640
Yellowhammer 2 400 000 2 185 070 1 964 690–2 429 790
Reed Bunting 440 000 335 030 284 560–478 890
Corn Bunting 320 000 113 140 82 850–155 100
Corn Bunting 15 960–23 3205 113 140 82 850–155 100

1Gibbons et al. (1993) – based on an informed guess; 2Wilson & Browne (1999); 3Wilson et al. (2001); 4Browne et al. (2000); 5Donald &
Evans (1995).

(Kestrel, Grey Partridge, Lapwing, Turtle Dove, Tree defined by Crick (1998), was approximately 129 692
Sparrow, Bullfinch, Yellowhammer and Reed Bunting) km2, (equivalent to over 50% of the total land area).
gave estimates whose confidence intervals encom- Habitat-specific density and population size
passed the estimate based on the last Breeding Atlas. estimates for each of the twenty species in this study are
Estimates for eight species (Wood Pigeon, Blackbird, shown in Tables 6 & 7 respectively. We found that
Whitethroat, Jackdaw, Starling, House Sparrow, more than 60% of Turtle Dove, Dunnock, White-
Greenfinch and Linnet) were larger than Atlas throat, Greenfinch and Linnet and over 90% of Grey
estimates, whilst estimates for four species (Skylark, Partridge, Yellowhammer, Tree Sparrow, and Corn
Dunnock, Song Thrush and Corn Bunting) were Buntings occurred on farmland (Table 7 & Fig. 1).
smaller than Atlas estimates. Population size estimates Whilst Corn Bunting and Grey Partridge showed a very
for Lapwing, Skylark and Corn Bunting, for which high preference for arable land, Turtle Dove, Tree
targeted surveys had been carried out in recent years, Sparrow, Whitethroat, Linnet and Yellowhammer were
were consistently smaller than Atlas estimates, but in found on a number of farmland habitats, with scrub and
the case of Lapwing and Corn Bunting were still rural and urban habitats also holding important popu-
considerably larger than estimates based on targeted lations of these species. Although farmland was found
surveys of these species. to hold important populations of all 20 species in this
study, the importance of rural, suburban and urban
habitats for a number of species is obvious. Whilst an
Habitat-specific density and population estimates
estimated 10% of the UK was classified as containing
Of 22 330 200-m transect sections surveyed in 1998, habitat associated with humans (rural, suburban or
12 646 (57%) belonged to one of five farmland habitat urban habitats), a large proportion (17–62%) of the
classes (Table 5). By taking the variation in coverage populations of Blackbird, Song Thrush, Dunnock,
between different NUTS regions into account, the Greenfinch, Starling, House Sparrow, Wood Pigeon
total estimated area of farmland in 1998 in the UK, as and Jackdaw were found in this broad habitat class.

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


Using BBS to estimate density and population size 47

Table 5. Number of 200-m transects surveyed within each habitat class and the estimated area of each habitat in the UK (with bootstrapped
95% confidence intervals).

Name Transects surveyed Total habitat in UK (km2) 95% CI % UK by area

Broadleaved woodland 1125 10 715 9 577–11 954 4.2


Coniferous woodland 771 14 436 11 757–16 953 5.7
Mixed woodland 558 5 679 4 766–6 535 2.2
Scrub 487 6 929 5 502–8 365 2.7
Dry semi-natural grassland 860 13 681 11 442–16 102 5.4
Wet semi-natural grassland 164 1 850 1 297–2 522 0.7
Heath 1268 29 261 25 712–33 568 11.5
Bog 212 4 688 3 278–6 477 1.8
Improved grass farmland 4798 53 827 50 543–56 786 21.2
Unimproved grass farmland 1518 16 799 14 734–18 631 6.6
Mixed grass/tilled farmland 759 7 350 6 380–8 323 2.9
Tilled farmland 5259 49 356 46 446–52 779 19.4
Other farmland 312 2 360 1 873–2 956 0.9
Urban human sites 761 5 716 4 638–6 796 2.2
Suburban human sites 1670 12 988 11 371–14 697 5.1
Rural human sites 1068 9 205 8 302–10 240 3.6
Freshwater bodies 414 4 994 4 082–6 135 2.0
Coastal 163 2 094 1 343–2 963 0.8
Inland rock 121 2 107 1 443–3 000 0.8
Miscellaneous 42 309 173–461 0.1
Total 22 330 254 345 – 100

DISCUSSION is if a significant proportion of observers deviate from


their ideal transect route due to land features to follow
Limitations and assumptions of BBS data and the
linear or edge features. In this situation, an apparently
chosen approach
rapid fall in detectability may be related to the absence
Producing reliable population estimates for wild bird of the critical feature for that species at greater
species is important for a number of reasons. In distances (e.g. hedgerows), which might inflate the
addition to providing information for those interested resulting density estimates. Moreover, recording birds
in the numbers of a species and for directing academic in two distance bands (0–25 m and 25–100 m) means
research, population estimates are widely used for that it is not possible to test the data for violations of
setting conservation priorities and are important for these assumptions or to examine quantitatively the fit
influencing policy and promoting the conservation of of the model. It would therefore be useful to collect
birds. Within a country, the population status and more detailed distance and habitat data from a sample
trends of a species may be examined by comparing of BBS sites for each species, to look for possible
current population estimates with earlier estimates, as violations of our assumptions. The resulting detection
in the third assessment of the population status of birds functions could then be used to improve the precision
in the UK (Gregory et al. 2002). Population estimates and confidence in existing estimates.
may also be used to determine the national or inter- Our estimates of population size refer to the recorded
national importance of a particular site or to compare number of individuals, specifically adults. The assump-
the status and trends of a species in one country to tion that BBS observers correctly determine the age of
those in other countries within the species’ range. This all birds cannot be tested, but data included in
approach has been used in Europe to identify species of our analysis might include some juveniles for early
European conservation concern to prioritize those breeding, mainly resident, species, especially during the
species in most need of conservation action (Tucker & late visit (mid-May to late June). This would inflate
Heath 1994). estimates for these species. Conversely, for migratory
There is no reason to believe that the BBS violates species such as the Whitethroat which arrives in the
the main assumptions of distance sampling and there- UK relatively late in the spring, numbers on the first
fore our estimates are reliable (see Buckland et al. 2001 visit (April to mid-May) may be less than present later
for a discussion). However, one potential source of bias in the season. Given these potential biases, it is

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


48 S.E. Newson et al.

Table 6. Habitat-specific density estimates of each species within the UK with bootstrapped 95% confidence intervals in parentheses. Units:
birds/km2 quoted to one decimal place for the estimate and to the nearest whole number for confidence intervals. Habitat and species
codes are explained in Tables 2 & 1 respectively.

Species density estimate (birds/km2)

K. P. L. WP TD S. D. B. ST WH

WOOD1 0.2 0.3 0.1 64.5 0.3 0.9 11.5 57.1 11.9 3.5
(0–5) (0–6) (0–0) (55–75) (0–1) (0–1) (9–14) (51–65) (8–14) (2–5)
WOOD2 0.1 0 0.6 22.7 0 0.8 7.3 12.7 5.9 0.5
(0–0) (0–0) (0–1) (17–30) (0–0) (0–1) (5–10) (10–17) (4–8) (0–1)
WOOD3 0.1 0.3 0.1 79.2 0.1 1.3 7.9 42.7 9.7 2.1
(0–7) (0–5) (0–0) (53–113) (0–0) (1–3) (5–10) (35–50) (7–13) (1–4)
SCRUB 0.2 0.8 0.3 29.7 0.2 2.7 14.4 31.1 7.1 9.5
(0–4) (0–28) (0–4) (21–39) (0–0) (1–5) (9–20) (23–40) (5–10) (6–13)
GRASS1 0.1 0.2 6.9 3.9 0 28.9 2.3 5.4 1.1 1.5
(0–4) (0–6) (2–14) (2–7) (0–0) (21–36) (1–4) (3–8) (1–2) (1–2)
GRASS2 0.2 0.3 10.8 13.6 – 19.1 13.2 22 3.4 8.6
(0–1) (0–1) (4–19) (7–25) (–) (7–33) (7–24) (12–32) (1–7) (3–15)
HEATH 0.5 0 1.9 0.9 0 16.6 0.8 1.2 0.6 0.5
(0–1) (0–0) (1–4) (0–2) (0–0) (12–21) (0–1) (1–2) (0–1) (0–1)
BOG – 0 0 1.7 0 22.6 1.5 2.1 2.4 0
(–) (0–0) (0–0) (0–6) (0–0) (12–35) (0–26) (0–5) (0–6) (0–0)
FARM1 0.3 1 1.3 25.2 0.1 5.6 16.7 39.6 4.9 5.6
(0–1) (0–1) (1–2) (22–28) (0–0) (4–7) (15–19) (37–43) (4–6) (5–7)
FARM2 0.3 1.6 4.1 25.3 0.2 11.6 10.5 31.4 3.4 4
(0–0) (1–26) (2–8) (19–34) (0–0) (8–16) (8–13) (27–36) (2–5) (3–5)
FARM3 0.2 1.4 1.1 50.2 0.2 9.7 23.7 53.1 7.2 14.4
(0–1) (0–3) (1–2) (39–68) (0–0) (7–13) (19–28) (45–63) (4–11) (10–19)
FARM4 0.4 3.1 1.8 46.1 0.7 18.9 16.6 32.5 2.8 15.3
(0–1) (2–4) (1–3) (41–53) (0–1) (16–22) (15–19) (29–36) (2–4) (13–17)
FARM5 0.3 2.2 1 52.6 2 5.7 24.9 68.3 10.8 9.2
(0–9) (0–106) (0–2) (33–78) (0–4) (3–9) (16–37) (51–92) (5–18) (5–17)
URBAN 0.2 0 0.1 37.1 0 0.5 15.7 103.1 6.6 0.9
(0–7) (0–0) (0–0) (28–48) (0–0) (0–1) (11–22) (83–128) (4–10) (0–2)
SUBURB 0.2 0.1 0.1 46.9 0.2 0.6 23 132.1 9.7 2
(0–0) (0–0) (0–2) (40–54) (0–0) (0–1) (19–27) (120–145) (8–11) (1–3)
RURAL 0.6 0.1 0.5 55.1 0.5 3.5 29.1 103.8 14.6 4.3
(0–1) (0–0) (0–1) (48–63) (0–1) (2–6) (25–34) (95–114) (12–17) (3–6)
WATER 0.2 0.3 1 41 2 1.9 13.8 43.3 6.8 8.4
(0–0) (0–1) (0–2) (29–56) (1–59) (1–3) (9–20) (32–55) (4–10) (5–13)
COAST 0 0 0.5 4.7 0 2.1 16.1 46.6 1.3 15.6
(0–0) (0–0) (0–1) (1–14) (0–0) (1–5) (2–32) (13–84) (0–3) (3–34)
ROCK 0.4 – 0 6.8 0 3.2 2.6 12.6 0.1 2
(0–12) (–) (0–0) (1–23) (0–0) (1–7) (0–27) (3–32) (0–0) (0–89)
MISC 0.4 0 4 8.3 0 19.4 – 23.9 – 0
(0–1) (0–0) (0–9) (1–153) (0–0) (3–35) (–) (7–44) (–) (0–0)

continued

remarkable that the estimates are so similar to those use data from early visits for resident species and late
from the last Atlas. As an indication of the level of visits for migrant species and perhaps both visits for
bias, we can compare the counts of Whitethroat from ‘intermediate’ species. Counts for intermediate species
the early and late visits. During 1998, 2127 such as Collared Dove Streptopelia decaocto may not be
Whitethroat were recorded during the early visit biased in either of the two visits. Alternatively, it may
compared with 2651 during the late visit. Data from be most appropriate to treat data in relation to the
both visits were used in the analyses, meaning that the behaviour of each species, although this needs to be
resulting estimates are likely to be intermediate examined and decisions made at the species level.
between those calculated if early or late visit data only The BBS does not ask observers to record sexes sepa-
were used. In the future it may be more appropriate to rately, indeed for many species examined in this study,

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


Using BBS to estimate density and population size 49

Table 6 continued.

Species density estimate (birds/km2)

JD SG HS TS GR LI BF Y. RB CB

WOOD1 12.1 7.8 6.3 – 8.7 2 2.8 3 0.4 –


(7–19) (5–11) (4–10) (–) (6–12) (0–5) (2–4) (2–5) (0–6) (–)
WOOD2 2.6 0.3 3.5 – 3.7 0.9 1.6 0.9 0.1 0
(0–9) (0–1) (0–19) (–) (2–7) (0–6) (1–3) (0–2) (0–2) (0–0)
WOOD3 9 2.9 7.6 – 6 1.2 5.5 1.4 0.1 0
(5–15) (1–5) (3–14) (–) (3–10) (0–7) (3–9) (1–2) (0–0) (0–0)
SCRUB 2 9.4 6.6 0 6.6 9.9 5.3 6.4 2.1 0
(1–4) (4–17) (3–13) (0–0) (4–9) (5–16) (2–10) (4–10) (1–4) (0–0)
GRASS1 2.4 3.8 1.4 0 0.8 3.2 0.7 1.3 2.3 –
(1–4) (2–7) (0–3) (0–0) (0–2) (1–6) (0–2) (1–2) (1–5) (–)
GRASS2 5 13.9 12.5 0 7.7 12.4 – 1.6 17.3 0
(2–12) (6–29) (2–29) (0–0) (2–15) (5–23) (–) (0–5) (9–27) (0–0)
HEATH 0.4 1.3 0 0 0.4 3.6 0.2 0.7 0.5 0
(0–1) (0–3) (0–1) (0–0) (0–4) (1–9) (0–5) (0–1) (0–1) (0–0)
BOG 0.7 0.4 0.2 0 0.1 0 0.5 0 1.4 0
(0–20) (0–22) (0–0) (0–0) (0–0) (0–0) (0–19) (0–0) (0–29) (0–0)
FARM1 13.6 39.5 37.6 2 11.7 10.6 1.6 8 1 0.1
(11–17) (31–50) (32–43) (1–4) (10–14) (8–13) (1–2) (6–9) (1–2) (0–0)
FARM2 15.3 38.2 35.9 2.4 9.7 8.7 1.5 5.8 2.1 0.1
(10–23) (22–60) (24.7–47.4) (1–6) (7–12) (6–12) (1–2) (4–8) (1–4) (0–0)
FARM3 11.9 30.6 60.6 2.5 19.7 17.4 1.5 24.4 1.4 0.2
(7–18) (22–40) (44–77) (1–28) (14–27) (11–26) (1–3) (18–30) (0–4) (0–0)
FARM4 5.7 13.7 26.2 1.6 11.3 17.1 1.3 24.8 1.7 2
(4–8) (11–17) (21–32) (1–3) (9–13) (14–20) (1–2) (22–27) (1–2) (1–3)
FARM5 7.5 44.9 72.1 3.8 26.4 19.5 2.6 10 0.3 0.2
(3–14) (28–60) (47–100) (0–142) (18–36) (8–38) (1–46) (6–18) (0–1) (0–1)
URBAN 18.4 223.3 345.3 0 25.5 5.4 0.6 – 0.2 0
(10–29) (167–288) (275–440) (0–0) (19–34) (3–10) (0–4) (–) (0–1) (0–0)
SUBURB 26.9 214 324.4 0.1 40.1 5.4 1.2 0.2 0.1 0.2
(15–41) (176–261) (276–378) (0–1) (34–46) (3–8) (1–20) (0–0) (0–1) (0–0)
RURAL 34.7 99.7 194.8 2.2 53.1 10.6 2.8 5.9 0.8 –
(26–44) (79–132) (167–226) (1–4) (45–61) (7–15) (2–4) (4–9) (0–2) (–)
WATER 6.9 30.7 25.3 1.4 8.6 7 2.9 4 7.1 0
(3–13) (16–45) (15–36) (0–37) (6–14) (3–15) (1–33) (2–7) (3–12) (0–0)
COAST 8.1 17.2 33.3 0 5.6 3.8 – 2.2 1.9 0.6
(0–32) (5–43) (7–81) (0–0) (1–12) (1–10) (–) (1–5) (0–6) (0–2)
ROCK 8.4 0 – 0 1.4 1.7 0.3 2 – 0
(1–133) (0–0) (–) (0–0) (0–31) (0–42) (0–6) (0–73) (–) (0–0)
MISC 1.7 25.9 45 0 – 2.1 0 24.4 0.7 6.1
(0–4) (0–1395) (2–188) (0–0) (–) (1–5) (0–0) (3–59) (0–2) (2–17)

it would not be possible to distinguish sex in the field. However, if there is a considerable difference in
However, where there is a large difference in detect- detectability between different flock sizes, this could
ability between the sexes, the resulting figures could bias detection functions and hence the resulting
underestimate the true density and population size for estimates. A study that examines sex-specific detect-
that species. In the case of the Skylark for ability and importance of flocking is required to
example where singing males are highly detectable, but examine the extent to which these bias the resulting
females much less so, the resulting population size estimates.
estimates may be close to the number of adult males
only. We also assume that BBS data are a collection of
Comparison with existing estimates
sightings of individuals. It could be argued that the BBS
is designed to monitor breeding birds, which for the In the past, there has been a tendency for population
majority of species tend not to flock at this time of year. size estimates to be reported with little or no discussion

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


50 S.E. Newson et al.

Table 7. Habitat-specific population estimates of each species within the UK with bootstrapped 95% confidence intervals. Units: 1000
birds quoted to one decimal place for the estimate and to the nearest whole number for confidence intervals. Habitat and species codes are
explained in Tables 2 & 1 respectively.

Species population estimate (thousands of birds)

K. P. L. WP TD S. D. B. ST WH

WOOD1 1.6 3 1.1 691.3 2.9 9.7 122.9 611.8 127.5 37.1
(1–49) (0–69) (0–3) (590–821) (1–5) (7–13) (96–147) (528–703) (99–155) (24–55)
WOOD2 1.4 0 8.8 327.4 0.4 12 105.3 183.2 84.4 7.5
(0–2) (0–0) (0–12) (231–431) (0–1) (6–19) (65–156) (135–240) (54–116) (2–17)
WOOD3 0.7 1.6 0.4 449.9 0.8 7.2 44.8 242.4 55.2 12.1
(0–40) (0–27) (0–2) (291–712) (0–2) (4–16) (31–60) (187–294) (40–74) (6–23)
SCRUB 1.4 5.7 1.9 205.9 1.1 18.5 99.5 215.7 49.4 65.6
(0–25) (1–187) (0–30) (152–272) (0–2) (8–34) (64–135) (164–264) (33–68) (43–92)
GRASS1 1.9 2.4 94.6 53.7 0 394.7 31.3 73.2 15 20.5
(1–8) (0–73) (32–178) (30–90) (0–0) (264–530) (15–51) (49–104) (7–26) (10–33)
GRASS2 0.4 0.6 20 25.1 – 35.4 24.4 40.7 6.3 15.8
(0–1) (0–2) (7–37) (13–43) (–) (11–72) (11–42) (20–62) (2–14) (5–28)
HEATH 15.5 0 55.4 26.2 0.2 484.9 22 35.7 16.7 14.4
(0–24) (0–0) (21–108) (9–51) (0–0) (339–661) (10–38) (16–65) (7–30) (0–24)
BOG – 0 0 7.7 0 106.1 6.8 9.6 11.1 0
(–) (0–0) (0–0) (1–24) (0–0) (49–161) (1–128) (2–23) (1–30) (0–0)
FARM1 18 50.9 72.2 1355.9 5.8 299.3 899 2132.6 261.8 302
(11–29) (25–79) (49–100) (1180–1524) (2–12) (220–384) (809–1014) (1939–2349) (218–314) (247–371)
FARM2 4.5 27.4 68.2 425.7 3.5 194 176.7 526.8 56.8 67.5
(3–7) (12–42) (26–129) (319–563) (1–7) (123–264) (140–216) (448–604) (40–75) (48–87)
FARM3 1.7 10.2 7.7 368.8 1.3 71.5 174 390.4 52.6 106.2
(1–5) (3–23) (4–12) (272–498) (0–3) (47–100) (134–216) (323–469) (33–82) (74–144)
FARM4 19.1 154.4 89.5 2273.4 32.1 934.5 819.8 1604.8 136.6 755.8
(13–28) (116–197) (61–133) (1990–2650) (21–40) (793–1081) (712–923) (1429–1772) (108–169) (655–852)
FARM5 0.6 5.2 2.4 124.1 4.7 13.4 58.8 161.2 25.5 21.6
(0–20) (0–222) (0–6) (72–192) (1–10) (7–24) (35–92) (106–239) (11–47) (10–39)
URBAN 1.2 0 0.4 212 0 2.6 89.6 589.2 37.4 5
(0–43) (0–0) (0–2) (147–283) (0–0) (0–3) (56–128) (431–779) (22–58) (2–11)
SUBURB 2 1.2 1.8 609.6 2.3 7.3 298.7 1715.7 125.7 26
(1–5) (0–4) (1–23) (498–729) (0–5) (3–14) (241–364) (1458–2014) (97–152) (14–42)
RURAL 5.5 1.2 4.9 507.5 4.2 32.4 268.2 955.4 134.2 39.3
(3–10) (0–4) (2–11) (429–592) (2–8) (18–56) (220–317) (835–1089) (108–162) (29–52)
WATER 1 1.3 4.8 204.8 9.9 9.2 69 216.4 34.1 42.1
(0–2) (0–3) (2–10) (141–284) (3–314.9) (4–15) (43–108) (156–273) (18–52) (24–68)
COAST 0 0 1 9.9 0 4.3 33.6 97.5 2.7 32.7
(0–0) (0–0) (0–2) (2–31) (0–0) (1–10) (4–74) (25–201) (1–7) (6–73)
ROCK – – 0 14.2 0 6.6 5.5 26.5 0.2 4.1
(–) (–) (0–0) (2–48) (0–0) (2–15) (0–61) (6–62) (0–1) (0–195)
MISC 0.1 0 1.2 2.5 0 6 – 7.4 – 0
(0–0) (0–0) (0–3) (0–42) (0–0) (1–12) (–) (2–15) (–) (0–0)
Sum 77.4 265 436.3 7895.7 69.2 2649.6 3350 9836.1 1233.2 1575.4

continued

as to the limitations of the data or analyses on which to address these before further estimates of this type are
they are based. Obviously the greater number of differ- adopted for general use.
ing population estimates that are ‘in use’, the greater Habitat-specific density estimates presented here are
the potential confusion, and the potential damage to similar to those in Gregory & Baillie (1998) and
the credibility of population estimates in the wider Gregory (1999), suggesting population size estimates, if
public domain. By highlighting the limitations and they were to be derived from these studies would
uncertainties in the data and analyses presented here, it be similar to those presented here. However, this
is hoped that the reader is not only aware of these comparison is not particularly informative, because
issues, but that through this awareness steps are taken these studies are all based on BBS data, use similar

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


Using BBS to estimate density and population size 51

Table 7 continued.

Species population estimate (thousands of birds)

JD SG HS TS GR LI BF Y. RB CB

WOOD1 129.9 83.1 67.4 – 93.4 21.6 30.2 31.8 4.3 –


(72–207) (53–123) (39–107) (–) (66–130) (5–58) (18–42) (17–49) (0–69) (–)
WOOD2 37.6 4.3 50.6 – 52.9 12.8 23.3 13.2 0.8 0
(7–125) (1–13) (5–253) (–) (25–93) (1–91) (9–42) (5–26) (0–30) (0–0)
WOOD3 51.2 16.5 43.2 – 34.3 6.8 31.1 8.2 0.3 0
(28–85) (8–29) (15–80) (–) (20–55) (2–30) (16–53) (3–14) (0–1) (0–0)
SCRUB 13.7 65.4 45.4 0 45.6 68.8 36.8 44.6 14.3 0
(6–25) (28–122) (18–92) (0–0) (26–64) (36–106) (14–66) (26–66) (5–28) (0–0)
GRASS1 33.1 52.2 18.6 0 11.1 43.1 9.9 17.5 31.3 –
(17–55) (29–94) (5–37) (0–0) (5–19) (18–80) (1–29) (9–28) (12–64) (–)
GRASS2 9.3 25.7 23.2 0 14.2 22.9 – 3 31.9 0
(4–22) (12–52) (3–59) (0–0) (3–26) (8–41) (–) (1–8) (15–53) (0–0)
HEATH 11.3 37.6 0.5 0 11.7 103.8 6.7 19 14.8 0
(3–33) (8–101) (0–22) (0–0) (2–115) (35–249) (3–139) (4–37) (4–29) (0–0)
BOG 3.2 2 0.7 0 0.5 0 2.2 0 6.8 0
(0–77) (0–108) (0–2) (0–0) (0–1) (0–0) (0–79) (0–0) (1–141) (0–0)
FARM1 731.1 2127.2 2021.5 108.7 630 570.8 85.6 429.4 51.4 5.3
(568–895) (1600–2706) (1728–2342) (60–194) (525–728) (450–701) (60–115) (341–514) (31–82) (4–20)
FARM2 256.2 641 603.3 39.5 162.8 146.6 25.1 97.6 35.1 1.1
(170–390) (383–975) (427–793) (8–109) (118–204) (103–201) (15–38) (66–137) (18–61) (0–4)
FARM3 87.3 224.5 445.5 18.3 144.6 127.9 11.2 179.4 10.6 1.3
(52–133) (160–304) (313–585) (6–188) (99–201) (83–198) (6–20) (132–234) (3–28) (0–3)
FARM4 279.1 674.9 1294.1 80.2 558.1 841.9 64.9 1224.7 83.9 99.2
(184–399) (525–859) (1021–1565) (46–127) (453–669) (687–1009) (44–98) (1064–1365) (58–115) (64–134)
FARM5 17.7 105.9 170.2 9.1 62.3 46 6.1 23.6 0.7 0.6
(8–31) (64–146) (106–252) (0–327) (37–96) (17–97) (1–118) (12–46) (0–2) (0–1)
URBAN 105.1 1276.5 1973.7 0.1 145.9 31 3.4 – 1.1 0
(55–173) (903–1692) (1450–2637) (0–0) (96–208) (15–58) (1–22) (–) (0–3) (0–0)
SUBURB 349.9 2778.9 4212.8 1.2 520.4 70.2 15.8 2.1 0.9 2.6
(191–565) (2117–3513) (3326–5292) (0–16) (419–635) (39–110) (8–275) (1–5) (0–10) (0–3)
RURAL 319.1 918.1 1793.4 20 489.1 97.2 25.8 54.7 7.5 –
(237–412) (712–1202) (1511–2083) (7–37) (412–566) (67–137) (16–38) (34–85) (2–17) (–)
WATER 34.7 153.3 126.1 6.9 43 34.7 14.2 20 35.3 0.1
(13–63) (82–223) (74–185) (0–169) (25–75) (14–75) (5–15) (10–35) (17–63) (0–0)
COAST 16.9 36.1 69.7 0 11.6 8 – 4.6 4 1.1
(0–66) (9–101) (11–183) (0–0) (2–35) (2–19) (–) (1–10) (1-13) (0-4)
ROCK 17.6 0 – 0 2.9 3.6 0.6 4.3 – 0
(1–289) (0–0) (–) (0–0) (0–54) (1–78) (0–10) (0–162) (–) (0–0)
MISC 0.5 8 13.9 0 – 0.6 0 7.5 0.2 1.9
(0–1) (0–435) (0–46) (0–0) (–) (0–1) (0–0) (1–21) (0–1) (1–6)
Sum 2504.3 9231.1 12973.8 284 3034.4 2258.4 392.9 2185.1 335 113.1

analyses and the timing of data collection spans a the south and east of Britain (Fuller et al. 1985), it
relatively short period (1995–98), in which there is might be expected that CBC estimates for farmland
likely to be little population change. specialists such as Corn Bunting, Grey Partridge, Tree
One assessment of the feasibility of using the BBS to Sparrow, Yellowhammer, Whitethroat, Linnet and
derive densities and population size estimates for Turtle Dove occurring predominantly in lowland
species in this study, is to compare the estimates England might be least affected by geographical and
reported here with independent population estimates habitat bias.
in the Breeding Atlas based on CBC data for 1989 and This study suggests that at least 70% or more of all
targeted national surveys for Lapwing, Skylark and Corn Bunting, Grey Partridge, Tree Sparrow,
Corn Bunting. Because CBC farmland plots are Yellowhammer, Whitethroat, Linnet and Turtle Dove
believed to be representative of lowland farmland in occur in farmland habitat (Fig. 1), the majority of

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


52 S.E. Newson et al.

Other Grass Wood Human Farm


100

Proportion of population (%)


75

50

25

0
CB P. TS Y. WH LI TD D. WP S. K. L. JD RB GR BF B. ST SG HS
Species
Figure 1. Estimated proportions of the total British populations of 20 species in this study occurring within five major habitat classes in 1998.
Other habitat here includes SCRUB, HEATH, BOG, WATER, COAST, ROCK and MISC as defined in Table 2. A key to species codes is given
in Table 1.

which are found in lowland in the south and east of Species in our BBS analyses producing similar popu-
Britain. Assuming this to be correct and adjusted for lation size estimates to those expected from the Atlas
temporal change in these species populations between include Kestrel, Song Thrush, Dunnock, Bullfinch and
the Atlas and this study (see Baillie et al. 2002), species Reed Bunting. However, the previous population
estimates reported from BBS were similar to those estimate for Kestrel in the Breeding Atlas was based on
expected from previous studies, except those for Linnet an informed guess of density and hence population size
and Corn Bunting (Table 4). The national population (Gibbons et al. 1993), so we can probably have greatest
estimate for Linnet in this study is considerably higher confidence in the estimate presented here for this
than that based on the last Breeding Atlas. species. Although Song Thrush, Dunnock, Bullfinch
Linnets may travel several kilometres to communal and Reed Bunting all occur in a number of habitats
foraging sites (Cramp & Perrins 1994) and because poorly covered or not covered by the CBC, this study
communal foraging groups are likely to be more suggests that densities in habitat classes other than
detectable than individuals, this could upwardly bias farmland or woodland are either similar to densities in
estimates using distance sampling and BBS data if these habitats or are very different in rare habitat only.
detectability is strongly related to flock size. Using estimates for the other habitat category based on
In contrast, the national population size estimate for densities in woodland and farmland probably explains
Corn Bunting here is considerably lower than in the the similarity between estimates in the Atlas and this
Breeding Atlas, but considerably higher than an esti- study. Population estimates of Skylark from a national
mate produced by Donald & Evans (1995) as a result of survey in 1997 (Browne et al. 2000) are very similar to
the 1993 national survey. Donald & Evans (1995) the estimates presented here (Table 4), although both
suggested that the Breeding Atlas overestimates are considerably lower than the Atlas estimates. There
abundance of Corn Bunting because of a regional bias are a number of reasons why the CBC methodology
in the data towards regions subsequently shown to may suggest a high population size for this species. The
support the highest densities of birds. The reason why 1997 national survey of Skylark, and probably the BBS,
the estimate here is higher than in the national survey effectively record singing males only. The mapping of
is not known, but might be explained by the fact that territories by the CBC over a large number of visits
this, and other targeted studies discussed here, assume a (10–12) is therefore likely to provide a better estimate
full count is made of all individuals present, whilst of the real number of Skylark territories than the
our BBS study corrects for individuals present but national survey (four visits) and our use of BBS data
undetected using the distance-sampling approach (two visits) even if the maximum count over visits is
employed here. used, as in the national survey.

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


Using BBS to estimate density and population size 53

This study effectively uses the mean counts over two multi-species approach to provide such information for
visits and will almost certainly underestimate the true large numbers of bird species. The BBS methodology
population. This point needs to be examined before represents an improvement on earlier multi-species
BBS data can be used to routinely generate wide-scale approaches based on CBC data, such as Atlas estimates,
density and population estimates. However, the CBC because it takes explicit account of geographical and
methodology may still overestimate population size for habitat-specific variation in population densities and
two main reasons. The method by which data are sampling intensity. For these reasons we should have
extracted from CBC maps may bias towards higher greatest confidence in estimates derived from BBS data,
densities of birds because territories lying partly outside although there are still a number of limitations and
a CBC plot boundary will still be recorded as one uncertainties related to this approach. Probably the
territory rather than a proportion of that sighting main difficulty with the BBS is the interpretation of the
within the plot. The CBC tends to use field boundaries number of adults counted which cannot readily be
as the plot edge, so the bias may be greatest for connected to the number of breeding pairs as has been
hedgerow specialists, but may affect all species to some used previously. However, results of earlier studies based
degree. Also the fact that the CBC is concentrated in on territory mapping may also be difficult to interpret.
the southeast of Britain, where Browne et al. (2000) has More research on the sex ratio and status of birds
shown there to be the highest densities of Skylark in counted by BBS surveys, and on the reliability of the
Britain, is likely to increase the estimated number of detectability functions derived from BBS distance sam-
Skylarks across the country. In the case of the Lapwing, pling data is needed.
the population size estimate here was higher than the
national survey of this species carried out in 1998
ACKNOWLEDGEMENTS
(Wilson & Browne 1999, Wilson et al. 2001) but simi-
lar to the estimates in the last Breeding Atlas. We are grateful to the thousands of volunteers who have
The reason for the difference between the national contributed to the Breeding Bird Survey, which provided the
survey and estimate here is not known, although there data on which these analyses are based. The BBS is jointly
is the possibility that the national survey based on a funded by the British Trust for Ornithology, the Joint Nature
single survey visit could have missed a small proportion Conservation Committee (on behalf of the Countryside
of nesting birds present. Residential areas were found to Council for Wales, English Nature, Scottish Natural Heritage
support an estimated 54% of Starlings and 62% and the Department of the Environment for Northern
of House Sparrows (Fig. 1), which is likely to be the Ireland) and the Royal Society for the Protection for Birds. We
result of the high availability of nest holes and feeding thank Jeremy Wilson, Peter Vickery and John Marchant for
opportunities in this environment (Crick et al. 2002). helpful comments on an earlier draft of this paper
This may explain the considerably higher new popula-
tion size estimates for these species compared with past REFERENCES
estimates based on CBC plots, which under-surveyed
Baillie, S.R., Crick, H.Q.P., Balmer, D.E., Beaven, L.P.,
human habitats and, as Table 6 demonstrates, these
Downie, I.S., Freeman, S.N., Leech, D.I., Marchant, J.H.,
species occur at particularly high densities within these Noble, D.G., Raven, M.J., Simpkin, A.P., Thewlis, R.M. &
habitats. For a similar reason, population estimates for Wernham, C.V. 2002. Breeding Birds in the Wider Countryside:
other species that occur at high densities in human Their Conservation Status 2001. BTO Research Report No. 278.
BTO, Thetford. (http://www.bto.org/birdtrends).
habitats (Blackbird, Greenfinch, Jackdaw and probably
Browne, S., Vickery, J. & Chamberlain, D. 2000. Densities and
Wood Pigeon) are likely to have been underestimated population estimates of breeding Skylarks Alauda arvensis in Britain
by the CBC methodology. in 1997. Bird Study 47: 52–65.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L.,
Borchers, D.L. & Thomas, L. 2001. Introduction to Distance
Conclusions Sampling: Estimating Abundance of Biological Populations. Oxford
University Press, Oxford.
The approach described here demonstrates that the Cramp, S. & Perrins, C.M. 1994. The Birds of the Western Palearc-
BBS has the potential to produce more robust estimates tic, Vol. 8. Oxford University Press, Oxford.
Cramp, S. & Simmons, K.E.L. 1983. The Birds of the Western
of habitat-specific densities and population sizes for
Palearctic, Vol. 3. Oxford University Press, Oxford.
many species than those that have been previously Crick, H.Q.P. 1992. A bird-habitat coding system for use in Britain
available. While single-species surveys may provide the and Ireland incorporating aspects of land management and human
best estimates for individual species there is a need for a activity. Bird Study 39: 1–12.

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54


54 S.E. Newson et al.

Crick, H.Q.P., Robinson, R.A., Appleton, G.F., Clark, N.A. & Wilson, A.M. & Baillie, S.R. 1996. The Breeding Bird Survey
Rickard, A.D. 2002. Investigation into the causes of the decline of 1994–1995. British Trust for Ornithology, Thetford.
starlings and house sparrows in Great Britain. BTO Research Report Gregory, R.D., Wilkinson, N.I., Noble, D.G., Robinson, J.A.,
No. 290. British Trust for Ornithology, Thetford. Brown, A.F., Hughes, J., Procter, D., Gibbons, D.W. &
Crowley, P.H. 1992. Resampling methods for computation-intensive Galbraith, C.A. 2002. The population status of birds in the United
data analysis in ecology and evolution. Ann. Rev. Ecol. Systematics Kingdom, Channel Islands and Isle of Man: an analysis of conser-
23: 405–447. vation concern 2002–2007. Br. Birds 95: 410–448.
Donald, P.F. & Evans, A.D. 1995. Habitat selection and population Noble, D.G., Raven, M.J. & Baillie, S.R. 2001. The Breeding Bird
size of Corn Buntings Miliaria calandra breeding in Britain in 1993. Survey 2000. BTO Research Report No. 265. British Trust for
Bird Study 42: 190–204. Ornithology, Thetford.
Field, R.H. & Gregory, R.D. 1998. Habitat Bias in Actual and Ideal Stone, B.H., Sears, J., Cranswick, P.A., Gregory, R.D., Gib-
Transect Lines in Breeding Bird Surveys 1994–1997. BTO Research bons, D.W., Rehfisch, M.M., Aebischer, N.J. & Reid, J.B.
Report No. 214. British Trust for Ornithology, Thetford. 1997. Population estimates of birds in Britain and in the United King-
Fuller, R.J., Marchant, J.H. & Morgan, R.A. 1985. How repre- dom. Br. Birds 90: 1–22.
sentative of agricultural practice in Britain are Common Birds Census Thomas, L., Laake, J.L., Derry, J.F., Buckland, S.T., Borchers,
farmland plots? Bird Study 32: 56–70. D.L., Anderson, D.R., Burnham, K.P., Strindberg, S., Hed-
Gibbons, D.W., Reid, J.B. & Chapman, R.A. 1993. The New ley, S.L., Burt, M.L., Marques, F., Pollard, J.H. & Fewster,
Atlas of Breeding Birds in Britain and Ireland: 1988–91. Poyser, R.M. 1998. Distance 3.5. Research Unit for Wildlife Population
London. Assessment, University of St Andrews, St Andrews.
Gregory, R.D. 1999. Broad-scale habitat use of sparrows, finches and Tucker, G.M. & Heath, M.F. 1994. Birds in Europe: Their Conser-
buntings in Britain. Vogelwelt 120 (suppl.): 163–173. vation Status. Birdlife International, Cambridge.
Gregory, R.D. & Baillie, S.R. 1998. Large-scale habitat use of some Wilson, A.M. & Browne, S.J. 1999. Breeding population estimates
declining British birds. J. Appl. Ecol. 35: 785–799. for Lapwing, Oystercatcher and Curlew in Scotland; results of 1998
Gregory, R.D. & Gibbons, D.W. 1999. Generation of the Headline BTO Lapwing Survey. Scot. Birds 20: 73–80.
Indicator of Wild Bird Populations. BTO Research Report No. 221. Wilson, J.D., Vickery, J.A. & Browne, S.J. 2001. Numbers and
British Trust for Ornithology, Thetford. distribution of Northern Lapwings Vanellus vanellus breeding in Eng-
Gregory, R.D., Bashford, R.I., Balmer, D.E., Marchant, J.H., land and Wales in 1998. Bird Study 48: 2–17.

(MS received 20 April 2003; revised MS accepted 27 February 2004)

APPENDIX

Habitat–species combinations for which there were too few obser-


vations of birds to fit habitat-specific detection curves and so were
removed from the analyses.

Species Habitat

Kestrel BOG
Grey Partridge ROCK
Turtle Dove GRASS2
Dunnock MISC
Song Thrush MISC
House Sparrow ROCK
Tree Sparrow WOOD1, WOOD2 & WOOD3
Greenfinch MISC
Bullfinch GRASS2, COAST
Yellowhammer URBAN
Reed Bunting ROCK
Corn Bunting WOOD1, GRASS1, RURAL

© 2005 British Trust for Ornithology, Bird Study, 52, 42–54

You might also like