Content uploaded by Roger Kirkwood
Author content
All content in this area was uploaded by Roger Kirkwood on Sep 02, 2022
Content may be subject to copyright.
Polar Biol (2008) 31:153–162
DOI 10.1007/s00300-007-0342-7
123
ORIGINAL PAPER
Comparison of census methods for black-browed albatrosses
breeding at the Ildefonso Archipelago, Chile
Graham Robertson · Carlos A. Moreno ·
Kieran Lawton · Roger Kirkwood · José Valencia
Received: 11 November 2006 / Revised: 10 July 2007 / Accepted: 16 July 2007 / Published online: 4 August 2007
Springer-Verlag 2007
Abstract A breeding population of black-browed alba-
trosses has been known to exist at the Ildefonso Archipel-
ago, Chile, for >90 years but the population has never been
censused using scientiWcally defendable methods. To esti-
mate population size, and examine the accuracy and practi-
cality of various census methods, the population was
censused in the 2002/2003 breeding season using (a)
ground-truthed aerial photography, (b) yacht-based photog-
raphy, (c) ground counts, (d) quadrat sampling and (e)
point-distance sampling. Compared to ground-truthed aer-
ial photography (judged the most accurate) yacht-based
photography underestimated population size by 55%,
ground counts by 13%, quadrat sampling by 11% and
point-distance sampling by 9%. Ground-truthed air photog-
raphy revealed that in the 2002/2003 breeding season
47,000 pairs of black-browed albatrosses bred at the Ilde-
fonso Archipelago. A repeat aerial census in 2006 sug-
gested the size of the breeding population had not changed
in the 4 years between the two censuses. After the Falkland
Islands/Islas Malvinas, South Georgia and Diego Ramirez,
the Ildefonso Archipelago holds the fourth largest popula-
tion of black-browed albatrosses in the world.
Keywords Ildefonso Archipelago · Albatross census
methods · Photographic counts · Yacht-based counts ·
Ground counts · Quadrat sampling · Point-distance
sampling
Introduction
Albatrosses and petrels have been counted at their breeding
locations in the Southern Ocean since the early part of the
twentieth century (e.g., Murphy 1936). Early population
assessments were often made in the ethos of exploration
and discovery, to acquire knowledge of the components and
workings of the natural world. The imperatives of modern
times require a more focused approach. Seabird population
estimates are usually made to build food consumption and
trophic pathway models for Wsheries management purposes
(e.g., Croxall et al. 1984; de Brooke 2004; Hill et al. 2006)
and because of concerns about conservation status. With
respect to the latter, threats to populations mainly pertain to
the albatrosses and petrels which are especially vulnerable
to mortality in longline (Robertson and Gales 1998; Delord
et al. 2005) and trawl (González-Zevallos and Yorio 2006;
Sullivan et al. 2006) Wsheries, which have lead to popula-
tion decreases at many breeding locations in the past two–
three decades (Gales 1998; Poncet et al. 2006).
One species of seabird that has decreased markedly is
the black-browed albatross Thalassarche melanophrys, an
Endangered species according to IUCN criteria (BirdLife
International 2004). Most of the world’s black-browed
albatrosses breed at the Falkland Islands, South Georgia
and Chile (Brooke 2004). Populations at the Falkland
G. Robertson (&) · K. Lawton
Australian Antarctic Division, Channel Highway,
Kingston, TAS 7050, Australia
e-mail: graham.robertson@aad.gov.au
C. A. Moreno
Instituto de Ecología y Evolución, Casilla 567,
Universidad Austral de Chile, Valdivia, Chile
R. Kirkwood
Phillip Island Nature Park, PO Box 97,
Cowes, VIC 3922, Australia
J. Valencia
Instituto Antárctico Chileno,
Plaza Muñoz Gamero 1055, Punta Arenas, Chile
154 Polar Biol (2008) 31:153–162
123
Islands (Huin 2001) and South Georgia (Croxall et al.
1998) have been censused for many years but those in Chile
have rarely been censused using scientiWcally reliable
methods. In this study we determined the population size of
black-browed albatrosses at the Ildefonso Archipelago,
Chile. We compared the accuracy and practicality of sev-
eral census methods to determine the method most suitable
for Ildefonso, a group of islands that is diYcult to land and
work on. Assessment of accuracy also provided a measure
of the accuracy of previous attempts to census albatrosses
at Ildefonso. The census methods compared were (a)
ground-truthed aerial photography, (b) yacht-based photog-
raphy, (c) ground counts, and the area-based methods (d)
quadrat sampling and (e) point-distance sampling.
Materials and methods
Ildefonso
The Ildefonso Archipelago (55°48⬘S; 69°24⬘W) is a chain
of six rock stacks lying 111 km west of Cape Horn
(55°59⬘S; 67°16⬘W) and 93 km north of the Diego Ramirez
Archipelago (56°31⬘S; 68°43⬘W) oV the southern coast of
Chile (Fig. 1). The Archipelago is broken into northern and
southern groups, the northern group comprising Wve stacks
and the southern group comprising a single stack. The
southern-most stack—hereafter called Isla Grande—is the
largest in the group, being 1.6 km long, 150–400 m wide,
up to 143 m high and 20 ha in area (Fig. 2). It comprises
about 60% of the total land area of the archipelago and
holds the largest population of albatrosses. The north-east
face of Isla Grande is near-vertical and the north-western
face slopes sharply to the sea. The islets are ravaged by
wind and sea spray and soil on most of Isla Grande is
shallow and laden with salt. In the seabird-breeding season
virtually all surfaces of Isla Grande above the wave zone
are inhabited by seabirds. Seabirds nest in ‘pavement’ and
tussock slope habitats (Fig. 3). Pavements are areas of bare
rock with virtually no soil and tussock slopes are vegetated
with open stands of tussock grass Poa Xabellata. In addi-
tion to large numbers of black-browed albatrosses, in the
2002/2003 breeding season the archipelago held about
86,000 pairs of rockhhopper penguins Eudyptes chryso-
come, 5,600 pairs of macaroni penguins E. chrysolophus
(Kirkwood et al. 2007) and several hundred blue-eyed
cormorants Phalacrocorax sp.
The census involved estimating the number of nesting
pairs that had laid and egg. The black-browed albatross
population at the entire Archipelago was censused from the
air and the sea but the comparison of the ground-based
methods was conducted at Isla Grande only (Table 1).
Three people were deployed on Isla Grande from 24 October
to 1 November to conduct the ground-based censuses
and ground truth the air photographs. Activities on the
ground also involved monitoring the attendance patterns of
Fig. 1 Map showing the location of the Ildefonso Archipelago in rela-
tion to Cape Horn and the Diego Ramirez Archipelago
70
°
W
80
°
W
Ildefonso
Archipelago Diego Ramírez
Archipelago
50
°
S
60
°
S
50
°
W
Antarctic
Peninsula
Drake Passage
South
America
Is. Norte
Is. Cinclodes
Is. Square
Is. Spirit
Is. Sur
1 km
Ildefonso
Archipelago
Is. Grande
Falkland Islands/
Islas Malvinas
60
°
W
1000 km
Cape Horn
Fig. 2 Aerial photograph of the Ildefonso Archipelago showing the
sizes and shapes of southern (Isla Grande, foreground) and northern
groups of islets
Polar Biol (2008) 31:153–162 155
123
albatrosses to determine the preferred time-of-day for the
aerial census (see below).
Aerial census
Photography and counting
Ildefonso was photographed from a twin otter aircraft from
an altitude of 200 m on 26 October 2002. Flying speed was
minimised to facilitate the photography and was 90 knots
upwind and 140 knots downwind. During the overXights air
and ground parties maintained radio contact so Xights could
be modiWed or suspended if birds showed signs of distur-
bance. The photographs were taken from 1400 to 1500 h
local time, which included the period of the day (late morn-
ing to mid afternoon) when the ratio of nesting birds to total
birds was highest (see below). The photographs were taken
through an open cockpit window with a hand-held 35 mm
camera, 70–200 mm vibration reduction lens and ISO 100
transparency Wlm. Every eVort was made to ensure the pho-
tographs were taken from an angle perpendicular to the
land surface. To aid in the later construction of collages of
images in the computer the focal length of the lens was kept
constant. A total of eight circuits was made of the islets to
produce a complete series of overlapping images which
were used to compile a montage of all surfaces of the
islands occupied by albatrosses.
Following digitisation of the images a montage of the
complete island group was constructed from overlapping
images “stitched” together using the image program Adobe
Photoshop®. Albatrosses were counted individually on the
montage displayed on the computer screen in Photoshop.
To gain a measure of the repeatability of the counts and
diVerence between counters, the number of albatrosses in a
well-deWned area on the southern end of Isla Grande (with
habitat typical of the remainder) was counted twice by each
of two observers (this area was also used to assess the accu-
racy of ground counts). To determine if albatrosses near
stitch lines were omitted or counted twice, due to parallax,
the number of nests within about 50 m of the edges of 20
images (stitched together to form 10 images) were com-
pared to the number counted near the centre of photographs
taken on subsequent overpasses. The proportion of mon-
tages of albatross habitat falling on the stitch lines was esti-
mated using the grid tool in Photoshop.
On 4 October 2006 the aerial photography was repeated,
but only at Ilsa Grande, to gain a measure of population
change in the 4 years since the estimate in 2002. The pho-
tography was conducted from the same aircraft, same alti-
tude, similar time-of-day (1400–1430 h local time) and
used the same photographer, counter and computer processing
methods as in 2002. In 2006 the photographic gear included
the same lens as in 2002 and a 10 Mp digital camera. The
results of the 2006 census are given in “Appendix 2”.
Fig. 3 Aerial photographs of a pavement and b tussock slopes habitat
(with some pavement) at Isla Grande, Ildefonso, showing nesting
black-browed albatrosses
Table 1 Summary of areas,
habitat types and methods used
(indicated by ‘x’) to census
black-browed albatrosses at the
Ildefonso Archipelago
Location and habitat Census method
Air
photo.
Yacht
photo.
Ground
counts
Quadrat £
areaa
Point-distance £
areaa
Entire archipelago x x
Isla Grande
Tussock slopes, all x x
Pavement, all x
Sub area; both habitats x
aExcludes the 1,715 nesting
birds counted from air photo-
graphs on the inaccessible north-
east face
156 Polar Biol (2008) 31:153–162
123
Ground truthing
To simplify the counting process ground truthing was con-
ducted to identify the time of day when the ratio of nesting
birds to total birds was highest. Photographs taken during
this period of the day should maximise the number of birds
sitting on nests and minimise the number ‘loaWng’ in the
colony or sitting beside incubating partner birds. To iden-
tify this optimum time period the ratio of nesting-to-total
birds in Wve study sites, each of 100 birds, was recorded
daily at 2-h intervals from 0800 to 1600 h from 24 October
to 1 November.
Ground truthing also involved quantifying at the same
time as the aerial photography the proportion of albatrosses
not on nests (as described above) and the proportion of
birds sitting on nests that did not contain an egg. Knowing
these proportions permitted the number of albatrosses
counted from the air photographs to be corrected down-
wards to produce a more accurate estimate of the number of
breeding pairs. Estimates were made immediately before
and after the aircraft overXights and the values averaged.
To determine the proportions of birds loaWng in the colony
or partnering birds on nests, the behaviour of 100 alba-
trosses at each of the Wve study sites was assessed. To
determine the proportion of occupied nests with and with-
out an egg, a total of 923 occupied nests was examined
along 17 transects and the presence/absence of an egg, and
egg shells, noted. The transects were distributed at random
along the southern two-thirds of the islet. Transects started
at the edge of a group of birds and ended when a pre-
selected landmark was reached (usually a rock or tussock).
The chest of each nesting albatross encountered on transect
lines was gently raised and the presence/absence of an egg
noted.
Disturbance
To determine if the albatrosses were disturbed by the air-
craft overXights, counts in the Wve sites mentioned above
were compared before and after the overpasses. Birds were
also observed on the ground for visible signs of disturbance
during the overXights.
Yacht-based photography
A complete photographic record of the archipelago was
made in a series of 80 overlapping photographs taken from
100 to 150 m oVshore from the S/V ‘Spirit of Sydney’ when
the vessel circumnavigated the archipelago from 1300 to
1700 h on 24 October 2002. Ocean conditions during the cir-
cumnavigation were calm and the wind was <10 knots. The
photographic gear, image processing and computer count
methods were the same as for the aerial census.
Ground counts
The accuracy of ground counts was determined in a 2–3 ha
area comprising several thousand nesting albatrosses on the
southern end of Isla Grande. The area was Xanked on all
sides by well-deWned landmarks and included three areas of
pavement habitat with high-density albatross nests and an
area of open tussock with low-density albatross nests mixed
with rockhopper penguin nests. Albatrosses on nests in
pavement and tussock areas were counted separately using
hand-held talliers while traversing the area on foot. Two
counters were used and each person counted the area three
times to gain a measure of diVerences within and between
counters.
Quadrat sampling £area
Quadrat sampling involved counting the number of nests in
10 £10 m quadrats distributed along the island, determin-
ing the mean density in each habitat type and multiplying
up to the total area. All surfaces of Isla Grande were sam-
pled except the sheer north-east face. DiVerent methods
were used to count birds in pavement areas (higher density
nest distribution) and tussock slopes (lower density nest
distribution) habitats. In pavement habitats nest density was
estimated in 17 quadrats located about 50 m apart along the
axis of the islet. The location point of each quadrat was
selected by walking 30 paces into the pavement area in a
direction perpendicular to its boundary. At the location
point two 10-m tape measures were pegged out at right
angles to form the two boundaries of the quadrat with the
location point being at the apex of the boundaries. The
locations of the remaining two boundaries were estimated
by eye using the taped boundaries as a guide. The number
of birds in quadrats was counted twice by each of two
people.
In the tussock slope habitat nest density was determined
by counting nests in 27 10 £10 m quadrats distributed
along the islet at about 100 m intervals. Quadrats were
marked out using the 10 m tape measure and nests counted
as described above. The number of albatross nests was
counted twice by each of two people.
The total area occupied by albatrosses in each habitat
type was estimated by walking around the perimeter of the
entire nesting area and recording location and elevation
data with a hand-held GPS. Co-ordinates for the perimeter
were plotted using ArcGis software (ESRI, Redlands,
USA) and a three-dimensional (3D) version of the total
nesting area calculated. Information on slope was included
in the analysis using a network of 3D polygons between
GPS points on a Lambert Azimuthal projection. The pro-
portion of pavement to tussock slope habitats was estimated
by overlaying a Wne scale grid on the aerial photographs in
Polar Biol (2008) 31:153–162 157
123
Adobe Photoshop. The scale used resulted in small and
large areas of habitat (both types) being divided into about
50 grid squares and about 100 grid squares, respectively.
The proportion of each habitat type was calculated by
counting the number of grid squares covered by each habi-
tat type. Squares that included both habitat types were con-
sidered to be the habitat that dominated the square. These
proportions were then used to estimate the proportion of
each habitat type in the total nesting area.
Point-distance sampling £area
The point-distance method involves measuring distances
from deWned points to nesting birds that fall within line-of-
sight only (birds not observed are ignored) and analysis
using purpose-built software. Point-distance sampling was
used to estimate the density of albatross nests only in the
tussock habitat because in pavement habitats all nesting
birds could be seen and counted. Point-distance sampling
was used rather than line-distance sampling (see Bibby
et al. 1992; Buckland et al. 1993) to minimize disturbance
to nesting birds. The former method involves estimating
distances from a Wxed point, whereas the latter method
requires that observers move along a transect line to esti-
mate distances. Point-distance sampling was applied at the
same time as the quadrat sampling following the method of
Buckland et al. (1993). At the location points in each of the
27 quadrats, and using the tape measure as a guide, dis-
tances to albatross nests out to 20 m were estimated to the
nearest 1 m in a 360° area around the location points. Data
were later truncated at 11 m distance according to the
requirements of the goodness-of-Wt models (see below).
Nest density was estimated using the program DIS-
TANCE 4.0 (Laake et al. 1994). This program oVers a
range of models that describe the data and uses the model
projection to estimate the number of nests not detected. The
program also assesses how well each model describes the
data. Goodness-of-Wt of each model was examined using
Akaike’s Information Criterion (AIC; Burnham and Ander-
son 1998) and the model with the lowest AIC value chosen
to estimate nest density.
Results
Preferred time for air photography
During the period 25 October–1 November 2002, the 2-
hourly checks for the ratio of nesting birds-to-total birds
revealed a signiWcant interaction between ‘day’ and ‘hour’
(ANOVA: F42,279 =1.5; P= 0.035), suggesting that the
ratios for any given hour diVered among days (Fig. 4). In
general the ratios were highest from 1 h before solar noon
(mean: 0.96; 1300 h local time) up until about 3 h after
solar noon (mean: 0.95); this was the preferred time period
in which to conduct the air census. Over all days and hours
sampled the ratio averaged 0.94 §0.03.
Air photography
Stitching error
The number of albatrosses counted within 50 m of the joins
of the 20 stitched photographs ranged from 2 to 144 com-
pared to 2–147 in the ten separate (unstitched) images of
the same areas. Overall, 1.3% of birds were “lost” on stitch
lines. Given that the total area of albatross habitat aVected
by stitching was only 5.9%, the overall eVect of parallax
and stitching was minor.
Variation within-and-between counters
The duplicate counts of the number of nesting birds on the
southern end of Isla Grande were 4,529 and 4,589 by one
counter and 4,489 and 4,507 by the other. Within-counter
coeYcients of variation (SD ¥mean) were 0.95 and
0.28%, respectively. The mean values for both counters
diVered by only 1.3%, so the scores for both counters were
averaged, which gave a total of 4,528 nesting albatrosses in
the area examined. The variation between counters was
mostly due to imperfect image resolution of a small number
of birds on the montages, which meant positive species
identiWcation was not possible.
Disturbance
In the Wve nesting groups each of 100 albatrosses examined
for signs of disturbance, immediately before the Xights all
Fig. 4 Preferred time-of-day (highest proportion of birds on nests) for
aerial censusing of black-browed albatrosses at the Ildefonso Archipel-
ago in the period 25 October–1 November 2002. Air photography for
the census was conducted at 1500 h on 26 October
Ratio
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
0800 1000 1200 1400 1600 1800 2000
Hour of the da
y
Novembe
r
25
26
27
28
29
30
31
1
October
Preferred time
period for census
solar noon
158 Polar Biol (2008) 31:153–162
123
but 21 were sitting on nests. Immediately after the over-
Xights 29 of the 500 birds counted were not sitting on nests.
During the Xights, the birds were watched intently by two
people and no bird was observed to leave its nest or show
other signs of disturbance, such as neck craning or sitting
upright on nests. The slight increase in the number of ‘unat-
tached’ birds after the Xights is considered to be consistent
with the normal movements of unattached birds in the
colony and not considered to be the result of the aircraft
overXights.
Air census and ground truthing
Ground truthing conducted at the same time as the air pho-
tographs were taken revealed that the proportion of birds in
the colony on a nest averaged 0.95 §0.02 (SD), suggesting
that for each bird on a nest there was, on average, 0.05 (5%)
of birds not on a nest. Of the nests with albatrosses sitting
on them, on average 0.93 §0.03 contained an egg, reveal-
ing that for every occupied nest with an egg there were 0.07
(7%) occupied nests without an egg. Multiplying these two
sources of error, the diVerence between the total number of
albatrosses counted on the photographs and the number of
nests estimated to contain an egg was about 12%.
The results of the aerial photographic counts are shown
in Table 2. The Wrst correction column shows the number
of birds estimated to be sitting on nests during the census.
The second correction column shows the number of occu-
pied nests with an egg.
Since well-formed nests in healthy black-browed alba-
tross colonies are usually occupied immediately after egg
laying (the census was conducted about 3 weeks after the
commencement of egg laying), the number in the Wrst cor-
rection column is most likely to be indicative of the size of
the breeding population (occupied nests that did not have
eggs had likely already failed). Hence in the 2002/2003
breeding season there were an estimated 47,000 pairs of
black-browed albatrosses breeding at the Ildefonso Archi-
pelago. Of these Isla Grande and Isla Norte held 62 and
22%, respectively.
Comparison of methods
Given the high quality of the air photographs, the suitability
of Ildefonso for this method of censusing (see “Discus-
sion”) and the thoroughness of the ground truthing the
results of the air photography were used as the “control”
against which the other methods were compared.
Yacht-based census
The result of the census from the yacht was corrected
downwards to equal the number of birds sitting on nests as
for the air census. The yacht-based census revealed a total
of 22,376 pairs of albatrosses on the entire archipelago,
55% fewer than counted from the air.
Ground counts
The average of three counts of nesting albatrosses by each
of the two counters in the 2–3 ha area of the southern end of
Isla Grande (all pavement and tussock habitats combined)
was 4,000 §462 (CV = 11.5%) for one counter and
3,949 §365 (CV = 9.3%) for the other. Overall, the counts
by both people were statistically similar (paired t test:
t(2) =0.92; P= 0.45), so the estimates were pooled. The
ground counts yielded 3,973 (§374 SD) nesting albatrosses
compared to 4,528 (§57) nests counted from the air photo-
graphs, an underestimate of 12.3%. The cause of this dis-
crepancy was the ground counts in tussock slope habitat
(Table 3). The estimates from the air and ground in pave-
ment habitat were statistically similar, whereas in tussock
slopes the ground counts underestimated the number of
albatross nests by 30%.
Quadrats £area
The quadrat sampling on Isla Grande excluded the sheer
north-east face where 1,715 nests were photographed.
Quadrat sampling in both habitat types yielded 0.84
(CL95%: 0.77–1.0; n= 17 sites) nests/m2 and 0.10 (0.07–
0.13; n= 27) nest/m2 for pavement and tussock slopes hab-
itats, respectively. The total land area of Isla Grande
(except the north-east face) was estimated to be 20 ha, 30%
of which was in the wave zone and did not contain nesting
Table 2 Total number of black-browed albatrosses counted on air
photographs (total count), sitting on nests (Wrst correction) and sitting
on nests with an egg (second correction) at the Ildefonso Archipelago
on 26 October 2002
aTotal count £0.95 (=number of occupied nests; see text)
bNumber of occupied nests £0.93 (=number of nests with an egg; see
text)
Group/islet Estimates
Total
count
First
correctiona
Second
correctionb
Northern group
Isla Norte 10,920 10,374 9,648
Isla Cinclodes 775 736 685
Isla Square 488 464 431
Isla Spirit 1,383 1,314 1,222
Isla Sur 5,222 4,961 4,614
Southern group
Isla Grande 30,680 29,146 27,106
Total for Ildefonso 49,468 46,995 43,706
Polar Biol (2008) 31:153–162 159
123
habitat. Of the remaining 14 ha 10% (14,000 m2) was pave-
ment habitat and the remaining 126,000 m2 tussock slopes.
Multiplying these two areas by their respective nest densi-
ties yields a combined total of 26,360 albatross nests for
Isla Grande compared to 27,431 from the aerial photo-
graphs, a diVerence of 11%.
Point-distance sampling £area
The comparison of models with high goodness-of-Wt values
is shown in the “Appendix 1”. The model with the best
goodness-of-Wt (lowest AIC value) was the half-normal
model, which was used to estimate nesting density. Dis-
tance data was truncated at 11 m from reference points
because observations beyond this distance (to 20 m) did not
enhance the goodness-of-Wt of the models. Point-distance
sampling in tussock habitat (except the north-east face)
yielded 0.12 (CL95%: 0.08–0.17; n= 27 sites) nests/m2 or
15,120 nests for all tussock slope habitat on Isla Grande
(except the north-east face). This estimate is 9% less than
the 16,629 nests counted in the tussock slopes habitat from
the air photographs.
Discussion
Comparison of methods
In the comparison of methods air photography was used as
the standard against which other methods were compared.
Since it is not possible to validate the accuracy of the air
photography, conWdence in the results must be derived
from assessment of the suitability of the method and poten-
tial errors. Unlike some other albatross breeding islands,
which are irregular in shape and many sided, both northern
and southern groups in the Ildefonso Archipelago are long
and narrow with only two Xanks inhabited by albatrosses
(north-east and north-west faces), making the Xight path
required of the aircraft easy to follow. The full width of the
islets can be captured in overlapping photographs, thereby
avoiding the diYculty of having to capture (and stitch
together later) from a fast moving aircraft images in both
horizontal and vertical planes. Critically, there is no con-
cealing vegetation to obscure views of albatrosses from the
air. Consequently high quality montages of the landscape
can be produced with albatrosses clearly visible and easily
counted.
The main potential errors associated with the air photog-
raphy were parallax in the image stitching process and mis-
identiWcation of albatrosses. Parallax error occurred when
photographs were taken in quick succession at angles for-
ward of the position of the aircraft in relation to the position
on the ground. The eVect of parallax was minor overall
(only 1.3% of birds missed on stitch lines and <6% of the
albatross habitat aVected by stitching) but could be reduced
further by taking each photograph from a position as close
as possible to perpendicular to the landscape. With respect
to species misidentiWcation, separating albatrosses from
other seabird species on the computer screen was relatively
straightforward. Macaroni penguins breed in discrete colo-
nies on pavement habitat, were discernible on the air photos
and the locations of colonies known to ground parties. Sim-
ilarly, the locations of cormorant colonies were well known
and recognised on the photographs, as were individual
birds—cormorants are smaller than albatrosses and have a
diVerent shape. Black-browed albatrosses and rockhopper
penguins were identiWed by their diVerent size, shape and
colouration—rockhoppers are much smaller, their heads are
black (as against white) and they lack the distinctive alba-
tross bill shape. Separating albatrosses and rockhopper pen-
guins became diYcult when images were not critically
sharp. This was probably the main source of the 1.3%
diVerence between counters of the number of albatrosses on
the air photographs. However, since the great majority of
photographs were sharp this source of error is unlikely to
have signiWcantly aVected the accuracy of the counts.
The air photography required simultaneous ground tru-
thing to maximize accuracy. The ground truthing revealed
that 5% of the birds photographed were loaWng in the col-
ony and a further 7% were sitting on empty nests (the size
Table 3 Comparisons of the number of nesting albatrosses counted from the ground in pavement and tussock slopes habitats with the numbe
r
counted from the aerial photographs
Counts from air photographs are repeated counts of the same photographs
aEstimates multiplied by 0.95 (see text and Table 1)
Habitat Air censusa
(mean §SD)
Ground count
(mean §SD)
t Statistic df P
Pavement area A 912 §41 925 §115 ¡0.19 7 0.85
Pavement area B 515 §23 482 §49 1.07 7 0.32
Pavement area C 1,436 §91,401§129 0.45 7 0.66
Tussock slopes 1,665 §22 1,164 §120 6.97 7 <0.001
160 Polar Biol (2008) 31:153–162
123
of these errors would be increased if the census was con-
ducted other than during the preferred time period, see
Fig. 3). While ground truthing improved the accuracy of the
estimate of population size a sailing vessel had to be char-
tered to land the ground party, adding substantially to the
logistical diYculty and cost of the census. Ildefonso is not
an easy island on which to land and deploying a ground
party in future censuses may be neither practical nor aVord-
able Wnancially. Ildefonso is also laden with seabirds and
diYcult to work on without causing disturbance. In the
future, one possibility is to use digital photography and take
photographs at a larger scale to increase the size of alba-
trosses in the images. Larger scale photographs can be
magniWed on the computer screen with minimal loss of
quality, providing clear, detailed, images of each bird and
its behaviour/posture when photographed. This might
enable nesting and loaWng birds to be counted separately
(e.g., Arata et al. 2003), which would reduce/eliminate the
need for ground truthing.
The Wnal point about air photography is the archival
quality of the data. Air photography was the only method
tested where future researchers can re-examine the original
digitized air photographs, repeat the stitching process, re-
count the albatrosses and draw their own conclusions about
accuracy. This adds a level of objectivity to the assessment
that is not possible with other census methods.
The yacht-based photographic census was included in
the comparison because this was the method used by
Aguayo-Lobo et al. (2001), who counted albatrosses on a
video tape of the islands. The yacht-based photography
underestimated population size by 55%. This discrepancy is
not surprising given the large number of birds on the ridge
line of Ildefonso—the location of most of the pavement
habitat—not visible from sea level. Note, however, that this
Wnding does not question the suitability of yacht-based cen-
suses of islands where albatrosses breed on steep ocean-
facing cliVs, such as Chile’s Diego de Almagro
(51°11⬘¡51°39⬘S; 75°07⬘¡75°19⬘W, Lawton et al. 2003)
and South Georgia (53°30⬘¡55°00⬘S; 35°30⬘¡38°40⬘W,
Poncet et al. 2006), islands that were recently censused
using yacht-based photography. Rather, it highlights the
unsuitability of boat-based counts for islands with topo-
graphical features that prevent line-of-sight views of all
birds from sea level.
The ground counts yielded 12.3% fewer nesting alba-
trosses than air photography of the same area. Ground
counts in pavement habitat were statistically similar to
counts from the air, but counts in tussock slopes fell 30%
short of the number derived from the aerial counts. On each
of the three counts by both counters a similar number of
birds were missed. This result is surprising because the
counters had considerable prior experience in censusing
albatrosses and thought they had covered the area thoroughly.
Although accessible by foot, the area missed sloped away
from the remaining habitat, making it diYcult to observe,
and be aware of, from the ground. That so many albatross
nests were missed highlights the importance of air photo-
graphs or maps of albatross distributions as a guide to
ground parties, especially for islands with complex terrain
where the existence of albatross nesting sites may not be
easily detected from the ground.
Quadrat sampling underestimated population size in
pavement and tussock slopes habitats by 11% and point-
distance sampling underestimated population size in tus-
sock slopes by 9%. The accuracy of both techniques may
have been improved by classifying breeding areas into
more than two habitat types. This was the initial intention
but would have required substantially more habitat map-
ping and an increased number of albatross density estima-
tions. With virtually all available nesting spaces on Isla
Grande occupied by albatrosses and penguins a more
detailed approach would have been too disruptive to the
seabirds (a concern that also applies to ground counts).
These two techniques are likely more suited to albatross
breeding colonies where birds are more patchily distributed
than at Ildefonso.
Previous censuses
In the past 92 years there have been three other albatross
censuses at Ildefonso, which are listed in Table 4. It is
important to appreciate that these censuses were conducted
in a region of the world (Cape Horn) that experiences
extreme weather conditions and has a history of ships being
wrecked. Hence with these censuses it is conceivable more
attention was paid to crew and vessel safety than to count-
ing albatrosses. Murphy’s (1936) visit in 1914 simply con-
Wrmed the existence of albatrosses at Ildefonso. Clark et al.
(1992) visited the islands by yacht in 1985 in the middle of
the chick-rearing period, by which time considerable breed-
ing failure would have occurred. Nonetheless their estimate
of 8,500 chicks for Isla Norte, made by a ground party,
roughly equates to our estimate of 10,734 nesting pairs
(Table 2) for the same islet three months earlier in the
breeding cycle. These two estimates were made 17 years
apart and suggest the population size at Isla Norte during
Clark’s visit and our own visit was similar. There is little
doubt the boat-based census by Aguayo-Lobo et al. (2002)
is inaccurate. Albatrosses at Ildefonso cannot be censused
from a boat because most albatrosses cannot be seen from
positions oVshore at sea level.
Importance of Ildefonso
In addition to Ildefonso, black-browed albatrosses in Chile
breed at Diego de Almagro (15,594 pairs in 2001/2002;
Polar Biol (2008) 31:153–162 161
123
Lawton et al. 2003), the Evangelistas islets (4,670 pairs in
2002/2003; Arata et al. 2003), Diego Ramirez (55,300
pairs in 2002/2003; Robertson et al. 2003) and the newly
discovered colony (18 January 2003) of <50 pairs in the
Straits of Magellan (57°27⬘20 S; 69°01⬘12 W; Aguayo-
Lobo et al. 2003). The total breeding population in Chile is
about 123,000 pairs, of which Ildefonso contains 47,000
pairs or 38%. Ildefonso is not only important in the Chil-
ean context but globally. After the Falkland Islands/Malv-
inas (382,000 pairs in 2000/2001; Huin 2001), South
Georgia (74,296 pairs in 2003/2004; Poncet et al. 2006)
and Diego Ramirez the Ildefonso Archipelago holds the
fourth largest population of black-browed albatrosses in
the world.
Ildefonso also holds a small population of grey-
headed albatrosses T. chrysostoma. In February 1982
Clark et al. (1992) sighted six grey-headed albatrosses
from their yacht at the north-east end of Isla Grande. In
October/November 2001, the year before our census, we
searched Isla Grande by foot and found eight pairs of
grey-headed albatrosses nesting in the same vicinity as
recorded by Clark’s party, thereby conWrming the persis-
tence of this small colony of grey-headed albatrosses at
Ildefonso.
Acknowledgments We are grateful to the Armada de Chile for per-
mission to work at Ildefonso, to Roger Wallis, skipper of the yacht
‘Spirit of Sydney’ for his seamanship in landing us on Ildefonso, to
Sergio Cortes (in 2002) and Luis Utmann (2006) for their skill in pilot-
ing the twin otter, to Jay Watson for assistance on the ground and
Rachel Alderman for assistance with the counts on the air photos in
2002. Carlos A. Moreno was funded by Proyecto Fondo de Investiga-
cion Pesquera 2006–2030. Comments by Javier Arata, Barry Baker,
Sally Poncet and Barbara Wienecke improved a draft.
Appendix 1
Appendix 2
In 2006 a total of 31,770 black-browed albatrosses were
counted on the air photographs of Ilsa Grande. This Wgure
corresponds to the 30,680 albatrosses counted on Isla Grande
in 2002 (see Table 2). Both are raw (uncorrected) counts.
The estimate in 2006 is 3.6% higher than in 2002 but was
conducted three weeks earlier in the breeding season. There
is also considerable within-hour variation in the ratio of nest-
ing-to-total birds for any given days (see Fig. 4), which con-
founds the assessment of diVerences between years. For
these reasons the result for 2006 is not considered to indicate
an increase in the size of the breeding population at Ildefonso
in the 4 years between the censuses.
Table 4 List of known population estimates, dates and methods used to census the black-browed albatross population at the Ildefonso Archipel-
ago, Chile
Estimates are of the number of breeding pairs except for Murphy (1936) which most likely refers to the total number of albatrosses
aUsed by Gales (1998) to erroneously indicate 17,000 breeding pairs for the whole archipelago, an estimate subsequently cited by Tickell (2000),
Brooke (2004), Aguayo-Lobo et al. (2001) and (2003)
Source Census date Census method Data treatment Population size
Beck, cited in
Murphy (1936)
18 December 1914 Estimated by eye from boat;
whole archipelago
None ‘Several thousand’
Clarke et al. (1992) 28 January 1985 Estimate by eye from yacht;
Isla Norte only
Number of chicks (8,500a) extrapolated
to whole archipelago
50,000
Aguayo-Lobo
et al. (2001)
6 February 2001 Boat-based survey using
video recorder;
whole archipelago
9,039 chicks counted; survivorship data
in Prince et al. (1994) used to derive
population size
20,086
This study 26 October 2002 Aerial photography;
whole archipelago
Ground-truthed to derive # birds
on nests (see Table 2)
47,000
This study 3 October 2006 Aerial photography;
Isla Grande only
None As in 2002
(see “Appendix 2”)
Table 5 Models and associated number of Akaike’s Information
Criterion (AIC) values, goodness-of-Wt (GOF) P values for point-
distance estimates of the density of albatross nests
The half-normal model (lowest AIC value) was used in the analysis o
f
point-distance data
D
number of nests/m2, SE (D) standard error of D
Model Adjustments AIC GOF D SE (D) CL95%
Half-
normal
Simple
polynomial
1,210.91 0.64 0.117 0.023 0.080–0.172
Hazard
rate
Hermite
polynomial
1,210.93 0.65 0.108 0.023 0.071–0.165
Uniform Simple
polynomial
1,211.26 0.83 0.123 0.024 0.084–0.181
162 Polar Biol (2008) 31:153–162
123
References
Aguayo-Lobo A, Acevedo J, Valenzuela C, Venegas C (2001) Census
of black-browed albatross Diomedea melanophrys Temmink
1828 in Ildefonso Island, and comments on its breeding in Evout
Island. An Inst Patagonia Ser Cienc Nat (Chile) 29:165–172
Aguayo-Lobo A, Acevedo J, Acuña P (2003) Nuevo sitio de anidami-
ento del albatros ceja negra, Diomedea melanophrys (Temmink
1828), en el seno Almirantazgo, Tierra del Fuego, Chile. An Inst
Patagonia Ser Cienc Nat (Chile) 31:91–96
Arata J, Robertson G, Valencia J, Lawton K (2003) The Evangelistas
Islets, Chile: a new breeding site for black-browed albatrosses.
Polar Biol 26:687–690
Bibby CJ, Burgess ND, Hill DA (1992) Bird census techniques. Aca-
demic, London
Birdlife International (2004) Threatened birds of the world. Lynx
Ediciones and Birdlife International, Barcelona
Brooke M (2004) Albatrosses and petrels across the world. Oxford
University Press, New York, p 499
de Brooke LM (2004) The food consumption of the world’s seabirds.
Proc R Soc Ldn 271:246–248
Buckland ST, Anderson DR, Burnham KP, Laake JL (1993) Distance
sampling: estimating the abundance of biological populations.
Chapman and Hall, London
Burnham KP, Anderson DR (1998) Multiple model selection and
inference: an information theoretic approach. Springer, London
Clarke GS, Cowan A, Harrison P, Bourne WRP (1992) Notes on the
seabirds of the Cape Horn Islands. Notornis 39:133–144
Croxall JP, Ricketts C, Prince PA (1984) Impact of seabirds on marine
resources, especially krill, of South Georgia waters. In: Whittow
GC, Rahn H (eds) Seabird energetics. Plenum Press, New York,
pp 285–317
Croxall JP, Prince PA, Rothery P, Wood AG (1998) Population chang-
es in albatrosses at South Georgia. In: Robertson G, Gales R (eds) .
Albatross biology and conservation. Surrey Beatty & Sons, Chip-
ping Norton, pp 69–83
Delord K, Gasco N, Weimerskirch H, Barbraud C (2005) Seabird mor-
tality in the Patagonian toothWsh longline Wshery around Crozet
and Kerguelen Islands, 2001–2003. CCAMLR Sci 12:53–80
Gales R (1998) Albatross populations: status and threats. In: Robertson
G, Gales R (eds) Albatross biology and conservation. Surrey Be-
atty & Sons Pty Ltd, Sydney, pp 20–45
González-Zevallos D, Yorio P (2006) Seabird use of discards and inci-
dental captures at the Argentina hake trawl Wshery in the Golfo
San Jorge, Argentina. Mar Ecol Prog Ser 316:175–183
Hill SL, Murphy EJ, Reid K, Trathan PN, Constable AJ (2006) Mod-
elling Southern Ocean ecosystems: krill, the food-web, and the
impacts of harvesting. Biol Rev 81:581–608
Huin N (2001) Census of the bla ck-browed albatross populations o f the
Falkland Islands. Report to Falkland Conservation
Kirkwood R, Lawton K, Moreno C, Valencia J, Schlatter R, Robertson
G (2007) Estimates of southern rock-hopper and macaroni pen-
guin numbers at the Ildefonso and Diego Ramirez Archipelagos,
Chile, using quadrat and distance-sampling techniques. Water-
birds 30:169–316
Laake JL, Buckland ST, Anderson DR, Burnham KP (1994) DIS-
TANCE users guide. Version 4.0. Colorado Fish and Wildlife
Research Unit. Colorado State University, Fort Collins
Lawton K, Robertson G, Valencia J, Wienecke B, Kirkwood R (2003)
The status of black-browed albatrosses Thalassarche melanoph-
rys at Diego de Almagro Island, Chile. Ibis 145:502–505
Murphy RC (1936) Oceanic birds of South America. Macmillan, New
York
Poncet S, Robertson G, Phillips RA, Lawton K, Phalan B, Trathan PN,
Croxall JP (2006) Status and distribution of wandering, black-
browed and grey-headed albatrosses breeding at South Georgia.
Polar Biol 29:772–781
Prince PA, Rothery P, Croxall JP, Wood AG (1994) Population
dynamics of black-browed and grey-headed albatrosses Diome-
dea melanophrys and D. chrysostoma at Bird Island, South Geor-
gia. Ibis 136:50–71
Robertson G, Valencia J, Arata J (2003) Summary report on the status
of black-browed and grey-headed albatrosses breeding in Chile.
CCAMLR ScientiWc Abstracts. WG-FSA-03/11 28
Sullivan BJ, Brickle P, Reid TA, Bobe TG, Middleton DAJ (2006)
Mitigation of seabird mortality on factory trawlers: trials of three
devices to reduce warp cable strikes. Polar Biol 29:745–753
Tickell L (2000) Albatrosses. Pica Press, Sussex, p 448
