Journal of the Royal Society of Western Australia, 87(3), September 2004 
Study area and methods 
Study area 
Barrow Island (20°43’S, 115°28'E) is a continental 
island, 56 km from the Western Australian coast between 
Onslow and Port Hedland, and approximately 1300 km 
north of Perth. It has been isolated from the mainland for 
7500 to 8000 years. It was formed from Tertiary 
limestone, and consists of high coastal cliffs on the west, 
sloping down through central uplands to coastal plains 
on the east. The climate is classified as tropical arid, or as 
a northern monsoonal climate (Gentilli 1972), with 
seasonally extremely high temperatures (> 45°C), and 
sporadic and highly variable rainfall. The rainfall 
averages 330 mm, but in four out of five years, the 
rainfall averages just 20% of this value. Most of the 
annual rainfall occurs during cyclones between February 
and June. Creeks and claypans will hold fresh water after 
a cyclone, but otherwise freshwater is extremely limited. 
Heavy dews, however, occur throughout the year 
(Buckley 1983) and provide a source of fresh water for 
birds, as do shoreline seeps. Additionally, facilities 
associated with the oil-field development now contain 
standing fresh water. 
Vegetation on Barrow Island has been studied by 
several investigators. Butler (1970) identified six major 
vegetative zones, and Buckley (1983) identified nine 
major zones, with 29 subtypes. On contract from 
WAPET, Matiske Consulting Group produced a 
detailed vegetation map of the island, identifying six 
major terrestrial vegetation types and 31 subtypes 
(excluding marine and two tidal vegetation zones). Our 
work and analyses were based on this latter map. A 
digital version of this map was available from National 
Geographic Information Systems (NGIS, Perth, Western 
Australia), and under contract, NGIS produced a field 
map of the vegetation zones for our use and calculated 
the area occupied by each zone, as well as areas 
occupied by various types of development (roads, oil 
wells, etc.). 
Briefly, the six land forms/vegetation units were: 
coastal complexes and dune systems (type C in Table 1, 
creeks or seasonal drainage lines (D), flats (F), limestone 
ridges (L), clay pans (S), and valley and escarpment 
slopes (V). Triodia hummock grassland is the dominant 
vegetation type and Triodia sp. occurs over 93% of the 
island. A complete list and description of each vegetation 
type, and subtype, is available on request from the 
authors. 
Because we wanted to calculate total population 
densities for the common land birds, we needed to 
calculate the total habitat area (land area) available to 
birds on Barrow Island. To do this, we corrected each 
vegetation zone for the area in each zone that was taken 
up by the presence of roads, oil pads, etc. We corrected 
the area measurements as follows: the total area of 
Barrow Island is 23 452.9 ha (measured from the high 
tide mark). Of this area, 161.3 ha (0.69%) consist of bare 
sand, marine and tidal habitats (which were not censused 
and in which there were no disturbances), or large areas 
with no vegetation, such as the airport and main 
Chevron camp. This leaves a total area of 23 291.6 ha 
encompassing the six major vegetation zones surveyed 
in this study, and also the area in which all of the 
disturbances associated with the oil-field development 
were located (excluding the airport and camp). The total 
area of disturbed areas (again excluding the airport and 
camp) is 915.7 ha or 3.93% of the total area. This includes 
the area occupied by 821 oil wells (as of 1997), and 
approximately 550 km of roads. In correcting the areas of 
each vegetation zone, we reduced the area by 3.93% to 
account for the disturbance in that zone. We recognize 
that this correction makes the assumption that 
disturbances are distributed randomly across vegetation 
zones, which is not correct. It would be logistically 
impossible to precisely calculate how much of each 
vegetation zone is disturbed, however we believe that 
our correction provides as accurate a calculation of the 
available habitat area as is possible. We also fully 
recognize that land birds do use roads and other man¬ 
made structures (although no birds were seen to use oil 
wells as 'habitat'). Nevertheless, our correction of the 
area of each vegetation zone by the amount of disturbed 
area within that zone seemed to us as the most 
conservative solution in order to calculate overall 
population estimates. 
Methods 
Fieldwork was conducted from 14 September 2000 to 
12 October 2001. A census was carried out on birds via 
transects in quadrats in each of the 31 vegetation zones. 
Our initial goal was to carry out a census in six quadrats 
within each vegetation zone, but in some cases this was 
not possible because some habitats were limited in area. 
Although the quadrats were located in specific vegetation 
zones, we made a concerted effort to census birds in all 
areas of the island. A total of 178 quadrats were 
censused. 
The transect counts were conducted each morning, 
weather permitting, from 0545 to 1100 and again from 
1500 to 1700. The quadrats were 2 ha in size, and in most 
cases were 200 m long X 100 m wide. The length of the 
quadrat was determined using a portable GPS unit 
(Garmin 45; accuracy = 10 m). The width of the quadrat 
was estimated visually. In order to maintain accuracy, 
both observers practiced estimating 50 m lengths and 
then checked these distances with a tape measure. We 
surveyed each quadrat together. One of us walked slowly 
in a straight line down the middle of the quadrat, and 
the other person zigzagged through the quadrat, 
intersecting the path of the person walking straight every 
50 m. All birds seen or heard inside the quadrats were 
counted. Each individual bird that was recorded was 
noted by both observers to ensure that it was only 
recorded once. A census was carried out in each quadrat 
only once, and all censuses took approximately 10 min to 
conduct (range 9-13 min). 
The exceptions to our methods as described above 
were as follows: first, in narrow, and linear vegetation 
zones, e.g., coastal habitats, the transects were 400 m long 
and 50 m wide; second, in habitats with dense and large 
spinifex both observers walked straight lines through the 
quadrats in parallel lines 40 m apart. 
As most vegetation zones were characterized by low 
plant height, and discontinuous distributions of plants, 
all bird species were equally detectable and we believe 
102 
