Journal of the Royal Society of Western Australia, 86(3), September 2003 
been burnt for at least 12 years, and probably for much 
longer than this. Climate is typical Mediterranean with 
an annual average rainfall of 514 mm. Monthly rainfall 
records were obtained from the closest official Bureau of 
Meteorology rainfall recording station 009865 at Warra 
Jarra AMG reference zone 50, 673303E 6191882N. Warra 
Jarra is approximately 3.5 km from the Tomlinson's site 
and 8 km from Parsons'. Annual rainfall during the trial 
period was 447 mm in 1999, 441 mm in 2000, and 598 mm 
in 2001. 
Short-term effects of rabbits after fence construction 
A 10.5 ha (350 x 300 m) patch of native mallee-heath 
vegetation on Parsons' property was used to determine 
the short-term effects of confined rabbits on remnant 
vegetation (Experiment 1). The impact of rabbits on the 
remnant vegetation was assessed, as described below, 
within each of thirty 10 x 10 m plots, 15 in each half of 
the site. The site was then divided into two halves and 
we surrounded one half with a rabbit-proof fence (Fig 1). 
At the same time, we also surrounded five of the 10 x 10 
m areas (chosen at random) in both the treatment and 
experimental control areas with rabbit-proof fencing to 
exclude rabbits. These exclosures were equivalent to 
removing the effect of rabbits on the vegetation 
completely and thus enabled comparison with and 
without the effects of rabbit grazing. The remaining open 
(i.e. subject to 'normal' rabbit grazing) 10 x 10 m plots (10 
in each of the fenced and unfenced areas) enabled the 
comparison to be made between the areas where rabbits 
were confined and where they had ready access to 
surrounding pastures and crops (i.e. 'rabbits free to move 
out'). Vegetation assessments, as described below, were 
repeated at fixed marked positions at two-monthly 
intervals for the 14 months immediately after the 
treatment vegetation was fenced. 
Because the treatment in the comparison 'rabbits 
confined' versus 'rabbits free to move out' was the 
erection of the external rabbit-proof fence, the open plots 
(vegetation assessment areas) did not provide true 
replication, and so the experiment was not suitable for 
strict statistical analysis. However, this experiment was 
carried out because information about the initial effect of 
rabbits is critical in the decision making process about 
the overall effects of the rabbit-proof fencing, and the 
documentation of the process was valid, even without 
rigorous statistics. Means (± standard error) for each 
vegetation parameter were calculated and plotted for 
each monitoring period. 
Long-term effects of rabbits on remnant vegetation 
The long-term effects on vegetation of rabbits confined 
to bush remnants (Experiment 2) was assessed at two 
sites (i.e. patches of native vegetation) situated on the 
Tomlinson's "Pallinup Park" property. One site ("Site 1"; 
11 ha; 550 x 200 m) had been rabbit-proof fenced 4-5 
years previously, and the other ("Site 2"; 8.5 ha; 500 x 170 
m) had been similarly fenced less than a year previously. 
Within each of these sites, 10 locations were randomly 
selected. At each location, an area of visually uniform 
vegetation was chosen, and in each of these a 10 x 10 m 
exclosure was constructed (preventing rabbit access) and 
a 10 x 10 m open plot, where rabbit access was 
unimpeded, was marked. In all exclosures and open 
plots, the vegetation was assessed within five 1 m 2 
quadrats as described below. Assessments were made 
before the exclosures were constructed in 1999, and again 
at the same time of year in 2000 and 2001 once the 10 m x 
10 m exclosures had been erected. 
In this assessment, the exclosures constituted the 
treatment, so on each site there were 10 replicates, and 
the data were analysed by analysis of variance (Zar 1984). 
The analysis of variance for each measurement compared 
the grazed and ungrazed plots in 2000 and 2001 using 
the measurements from 1999 as a covariate. A split plot 
analysis of variance was used with the year as the sub¬ 
treatment. Residual plots were used to check that the 
assumptions underlying the analysis of variance were 
valid. Analysis of variance was also used to compare 
grazed and ungrazed plots in 1999, prior to any treatment 
effects. These analyses were made using Genstat (v 6, 
Lawes Agricultural Trust, Rothamsted). 
Vegetation Measures 
A point quadrat method (36 points; Gilfillan 1999) was 
used for assessing percentage cover within 10 m x 10 m 
permanently marked, randomly selected, plots (n = 5-10) 
within each of the study sites. Each plot contained five 
permanently marked 1 m 2 quadrats. One quadrat was 
located near each corner, and the fifth quadrat was 
situated in the middle, of each 10 m x 10 m plot. Care 
was taken to avoid edge effects caused by the fence 
construction and any perching birds. The following 
vegetation measures were taken for all quadrats: percent 
cover of small (<0.5 m high) understorey shrubs (SS), 
percent cover of sedges and grasses (SG), percent cover 
of other monocotyledonous plants (OM), number of 
quadrats (n = 5 per plot) in which seedlings or reshoots 
of rootstocks were visible (SR Quadrats; score, 0 (none 
present) to 5 (present in all 5 quadrats), and an 
abundance index of seedlings and reshoots combined as 
single category (SR Score: 1 = 1-5 seedlings/reshoots, 2 = 
Figure 1. Schematic representation of the experimental design 
for the short-term study undertaken at the Parsons' site. 
98 
