Kea Survival during aerial poisoning for rat and possum control DOC 5674512.pdf

Kemp et al.: Kea survival during aerial 1080 dr
New Zealand Jour
ops
nal of Ecology (2019) 43(1): 0-0 © 2018 New Zealand Ecological Society.
1
RESEARCH
Kea survival during aerial poisoning for rat and possum control
Joshua R. Kemp*, Corey C. Mosen, Graeme P. Elliott, Christine M. Hunter and Paul van Klink
Biodiversity Group, Department of Conservation, Private Bag 5, Nelson 7010, New Zealand
*Author for correspondence (Email: [email address])
Published online: 21 September 2018
Abstract: The kea (Nestor notabilis) is a highly intelligent and adaptable omnivorous New Zealand parrot.
These traits potentially put kea at risk of poisoning during vertebrate pest poisoning operations. However, as
kea fall prey to introduced pests, they also gain from pest control, creating a cost-benefit situation. Pest control
in kea habitat is mainly by aerial 1080, the distribution of sodium fluoroacetate poison pellets by helicopter.
Understanding the net outcome for kea of this pest control method is extremely important because kea are
endangered and aerial 1080 use is controversial. We use 222 monitoring cases of individually marked kea at 19
aerial 1080 operations to model kea survival of aerial 1080 operations with respect to five variables. Proximity
to human-occupied sites where kea scrounge human food was inversely related to survival; the odds of survival
increased by a factor of 6.9 for remote kea compared to those that lived near scrounging sites. High survival in
remote areas is explained by innate neophobia and a short field-life of prefeed baits, which together preclude
acceptance of poison baits as familiar food. Elevated risk to kea living near scrounging sites is explained by
learned neophilia, possibly exacerbated by lead poisoning. Survival was also related to the history of aerial 1080
treatment at a site; the odds of survival increased by a factor of 21.3 at sites with repeated operations compared
with first time treatments. This effect is possibly due to selection for neophobic phenotypes. We suggest that
1080 poisoning risk management for kea should focus on reducing human food availability through an advocacy
campaign. If most kea have not been fed by humans, then the long term outcome of the South Island aerial
1080 programme should be positive for the kea.
Keywords: aerial 1080; non-target risk; pest control; poisoning risk
Introduction
New Zealand is to apply a single ‘prefeed’ of non-toxic baits
1–6 weeks prior to poison baiting (Veltman & Westbrooke
The use of aerial 1080 in New Zealand and consequences
2011; Elliott & Kemp 2016) to improve kill rates (Coleman
for native birds
et al. 2007; Nugent et al. 2011).
Aerial 1080 (sodium fluoroacetate) operations are increasingly
Aerial 1080 operations can have both positive and negative
used in New Zealand to control mammalian pests. Pest control
effects on native birds. For example, Powlesland et al. (1999)
campaigns in 2014 and 2016 saw an unprecedented 554 000
found mortality rates of 9.7% and 55% for North Island
ha and 729 000 ha, respectively, treated with aerially applied
robins (Petroica australis longipes) following two aerial 1080
1080 baits (New Zealand Department of Conservation unpubl.
operations. However, overall robin abundance increased in both
data). These operations target brushtail possums (Trichosurus
cases one year after aerial 1080 baiting due to greatly improved
vulpecula) and ship rats (Rattus rattus) to reduce predation
recruitment in the absence of key mammalian predators. The
on native and endemic species and/or to reduce the incidence
Powlesland et al. (1999) study highlights the importance of
of bovine tuberculosis in livestock (Coleman & Caley 2000).
assessing the net effect of aerial 1080 for native species at the
Aerial 1080 baits are cereal-based pellets, weighing 6 g
population level over several years. Long term net outcomes
or 12 g, with a field life ranging from a few days to a few
are most often gauged by trends in relative abundance (e.g.
weeks depending on rainfall (Bowen et al. 1995; Frampton
O’Donnell & Hoare 2012; Greene et al. 2013; Van Vianen
et al. 1999). Poison pellets contain 1.5 mg g−1 of 1080 toxin;
et al. 2018) or by detailed studies of reproductive success and
hence a 6 g pellet contains 9 mg of active ingredient and a
survival (e.g. Powlesland et al. 1999, 2003). Understanding the
12 g pellet contains twice this amount. The pellets are flung
net effect of aerial 1080 for native species is of considerable
from a hopper slung under a helicopter at coverage rates of
conservation importance in New Zealand because it is one of
1–3 kg ha−1 (Elliott & Kemp 2016). Most pellets fall to the
few proven pest control methods that can be applied on a large
ground, where they lie until they are consumed by animals
scale in rugged terrain at reasonable cost (Wright 2011), and
or degraded by moisture. Rodents and possums find pellets at
because of public interest in the method (Green & Rohan 2012).
night by smell and directly consume them. Modern practice in
DOI: 10.20417/nzjecol.43.2

2
New Zealand Journal of Ecology, Vol. 43, No. 1, 2019
particularly stoats (Mustela erminea), which die from secondary
Anatoki
poisoning (Murphy et al. 1999). To balance this positive
Oparara
impact against non-target 1080 poisoning risk, and to find risk
mitigation pathways, requires improved quantification of risk
Mt Arthur
and an understanding of its spatial and temporal variability.
Wangapeka ●
●● Nelson Lakes
Rotoiti
Scrounging by kea as a potential influence on 1080
poisoning risk
Otira
An outcome of the kea’s adaptability that could possibly affect
●●
●
●
●
Hawdon
Okarito
Arthur's Pass
1080 poisoning risk is the exploitation of humans for food
●
●
●
Glacier Country
●
●
by some individuals. Human food is obtained by stealing,
Fox−Franz
●
●
●
●
●
Copland
soliciting, and scavenging food items, directly from people
Abbey Rocks
● Mt Cook
and indirectly from rubbish bins and open landfills. We use
the term ‘scrounging’ to describe this phenomenon. Young kea
Arawhata
are initially attracted to scrounging sites by the presence of
●
●●
Treble Cone
other kea and then learn to scrounge by emulation, particularly
●●● ● Milford Road
of adult males (Diamond & Bond 1991; Gajdon et al. 2004).
Scrounging by kea is patchy in the landscape and can only
be observed daily in three main areas of intensive tourist
activity: (1) Milford Road, (2) West Coast ‘Glacier Country’
(Fox and Franz Josef Glaciers), and (3) Arthur’s Pass; and at
some ski areas and high-use back-country huts, e.g. on the
Milford Track (Fig. 1). Scrounging was previously common
Figure 1.
at Mt Cook Village, but now is rare following changes in
Map of South Island, New Zealand, showing forests and
scrub inhabited by kea (light green), locations of our study sites
rubbish collection and disposal practices. In contrast, efforts
(blue labels, blue shading shows aerial 1080 treatment boundaries)
to prevent feeding of kea by visitors to the three places listed
and sites where kea scrounge food from humans (pink).
above have been unconcerted and, at present, unsuccessful;
kea continue to scrounge daily at these sites. In the remote
back-country, scrounging at tramping huts was common prior
to 1984, when open rubbish pits were provided by the forest
service. The closure of these pits and a new ‘pack it in, pack
The kea, a unique bird with positive and negative responses
it out’ ethic has reduced kea visitation rates at huts to very
to 1080
low levels. At scrounging sites kea are constantly exposed
A species for which a negative short term impact of aerial
to rewarding novel foods and become familiar with a wider
1080 has been detected is the kea (Nestor notabilis) (Veltman
range of food types than is normal in the wild. We hypothesise
& Westbrooke 2011), a large endangered parrot endemic to
that these experiences could lead to suppression of innate
the South Island. Kea are sparsely distributed across about
neophobic behaviours that might otherwise protect kea from
three million hectares of indigenous South Island forests and
1080 poisoning risk.
about 1.5 million hectares of adjoining sub-alpine shrublands,
We evaluated the effect of proximity to scrounging sites,
grasslands  and  herbfields  (Fig.  1)  (Robertson  et  al.  2007).
plus four other potentially important variables, on kea survival
Nearly all of the kea’s range is legally-protected montane
through aerial 1080 operations. Specifically, we investigated
wilderness, administered by the New Zealand Department of
whether 1080 poisoning risk was:
Conservation (DOC). Roughly one-third (c. 27%) of the kea
(1)  higher close to scrounging sites
species range had aerial 1080 applied between 2014 and 2016
(2)  lower after repeated aerial 1080 treatments (i.e. at sites
(DOC unpubl. data).
where aerial 1080 had been previously applied)
The kea is a distinctive member of the New Zealand
(3)  higher with larger bait size
avifauna (de Kloet & de Kloet 2005), possessing several
(4)  lower for adult kea than juvenile kea
traits that potentially elevate 1080 poisoning risk. Unlike kākā
(5)  different for males and females.
(Nestor meridionalis), kea’s sympatric congener, which nest
We could not evaluate whether prefeeding increased risk to kea
in tree cavities and forage predominantly in the forest canopy
because all aerial 1080 operations in the study used prefeed.
(Moorhouse 1997), kea nest underground (Jackson 1963)
and frequently forage on the forest floor (Young et al. 2012;
Greer et al. 2015). Kea are highly intelligent, omnivorous
and adaptable, the latter particularly during the juvenile stage
Methods
(Diamond & Bond 1991). Juvenile adaptability is a trait shared
by many omnivorous generalists (Rozin 1976) and also by the
We utilised planned aerial 1080 operations in the South Island
congeneric kākā (Wilson et al. 1998; Loepelt et al. 2016). As
to evaluate the research questions listed above. Between 2008
such, kea are more likely than arboreal birds to encounter 1080
and 2016, we monitored the survival of individual kea through
pellets and may be more likely than specialist feeders to eat
19 operations (n = 222 monitoring occasions) at 12 sites (Table
them, thus giving rise to a potentially high negative impact
1; Fig. 1). Seven of these sites were utilised once, three sites
from non target mortality. However, positive consequences
were utilised twice, and two sites were utilised three times. The
of aerial 1080 for kea have also been reported (Kemp et al.
number of kea monitored per operation (and per site) ranged
2018), attributable to the control of invasive exotic predators,
from 2 to 37 (Table 1; Fig. 2). We monitored 15 individuals

Kemp et al.: Kea survival during aerial 1080 drops
3
Table 1. Details of the aerial 1080 operations in the South Island, New Zealand, at which 1080 poisoning risk to kea was
measured. Number of previous aerial 1080 treatments and years since prior aerial 1080 are the basis for assignment of
1080 history category scores. Counts of monitored kea are given by age and sex with numbers of deaths attributed to 1080
poisoning in brackets. AF=adult female, AM=adult male, FYF=first year female, FYM=first year male, JF=second year
female, JM=second year male.
__________________________________________________________________________________________________________________________________________________________________

Number of

previous
Years since  1080

Bait

aerial 1080  prior aerial  history
Operation
Month  size (g)  Scrounge  treatments
1080
category  AF
AM
FYF  FYM  JF
JM
Total
__________________________________________________________________________________________________________________________________________________________________
Arawhata 2008
Jan
12
Remote
3
4
2
2 (0)
7 (0)
0 (0)  0 (0)  0 (0)  1 (0)  10 (0)
Fox-Franz 2008
May
12
Adjacent
1
NA
1
3 (1)
7 (4)
0 (0)  0 (0)  3 (1)  4 (1)  17 (7)
Mt Arthur 2009
Jun
12
Remote
1
NA
1
2 (0)
9 (0)
0 (0)  0 (0)  0 (0)  2 (0)  13 (0)
Hawdon 2009
Sep
6
Adjacent
2
3
3
3 (0)
3 (0)
1 (0)  1 (0)  1 (0)  1 (0)  10 (0)
Okarito 2011
Sep
12
Adjacent
2
13
2
13 (2)  13 (2)  2 (1)  3 (1)  3 (2)  3 (0)  37 (8)
Wangapeka 2011
Sep
12
Remote
2
5
2
4 (0)
7 (0)
0 (0)  0 (0)  1 (0)  1 (0)  13 (0)
Abbey Rocks 2011  Oct
12
Remote
3
2
3
3 (0)
4 (0)
0 (0)  0 (0)  1 (0)  0 (0)  8 (0)
Copland 2012
Sep
12
Adjacent
3
6
2
2 (0)
0 (0)
0 (0)  0 (0)  0 (0)  0 (0)  2 (0)
Hawdon 2012
Dec
6
Adjacent
3
3
3
6 (0)
0 (0)
0 (0)  0 (0)  0 (0)  0 (0)  6 (0)
Otira 2013
Jul
12
Adjacent
6
8
2
11 (2)  15 (3)  1 (0)  2 (0)  3 (0)  2 (0)  34 (5)
Mt Arthur 2014
Aug
12
Remote
2
5
2
1 (0)
3 (0)
0 (0)  1 (0)  0 (0)  2 (0)  7 (0)
Anatoki 2014
Oct
6
Remote
3
5
2
1 (0)
1 (0)
0 (0)  0 (0)  0 (0)  0 (0)  2 (0)
Wangapeka 2014
Oct
6
Remote
3
3
3
4 (0)
2 (0)
1 (0)  1 (0)  0 (0)  0 (0)  8 (0)
Abbey Rocks 2014  Nov
6
Remote
3
3
3
8 (0)
10 (1)  2 (0)  0 (0)  0 (0)  1 (0)  21 (1)
Oparara 2014
Nov
12
Remote
1
NA
1
2 (0)
1 (0)
1 (1)  1 (1)  0 (0)  0 (0)  5 (2)
Hawdon 2014
Dec
6
Adjacent
4
3
3
4 (0)
0 (0)
0 (0)  0 (0)  0 (0)  0 (0)  4 (0)
Rotoiti 2014
Dec
6
Adjacent
1
NA
1
1 (1)
1 (0)
0 (0)  0 (0)  0 (0)  0 (0)  2 (1)
Oparara 2016
Sep
6
Remote
2
2
3
3 (0)
2 (0)
0 (0)  0 (0)  0 (0)  0 (0)  5 (0)
Wangapeka 2016
Oct
6
Remote
4
2
3
8 (0)
8 (0)
2 (0)  0 (0)  0 (0)  0 (0)  18 (0)
Total

81 (6)  93 (10)  10 (2)  9 (2)  12 (3)  17 (1)  222 (24)
__________________________________________________________________________________________________________________________________________________________________
Figure 2. Sample sizes and distribution of
40
kea deaths with respect to sites at which kea
survival through aerial 1080 operations was
monitored in the South Island, New Zealand.
All of the recorded kea deaths are attributed to
direct consumption of 1080 baits. Blue = kea
not poisoned, pink = kea poisoned.
30
ea
Outcome
20
Kea poisoned
Kea not poisoned
Number of k
10
0
a
a
anz
ito
thur
ar
wdon
Otir
Rotoiti
y Rocks
Anatoki
awhata
Copland
Ha
Okar
Opar
be
Ar
Fox−Fr
Mt Ar
angapeka
Ab
W
Site

4
New Zealand Journal of Ecology, Vol. 43, No. 1, 2019
through a second aerial 1080 operation and then two of these
distance from there to the nearest scrounging site (shown in
were monitored through a third operation. Thus, the 222
Fig. 1). By this method the monitored kea clearly comprised
monitoring occasions were obtained from 205 different kea.
two distinct groups, (1) kea living adjacent to (<20 km) and
Monitoring was achieved using radio tags (n = 209 occasions)
(2) kea living remote from (>40 km) scrounging sites (Fig. 3),
or by observation of unique leg bands at active nest sites (n =
with 110 and 112 cases in each group, respectively (Fig. 4a).
13 occasions). Post aerial 1080 monitoring continued for at
No monitored kea had average locations between 20–40 km
least 3 weeks, by which time baits become non toxic due to
from scrounging sites and none were recorded moving between
rain leaching poison from the bait matrix (Bowen et al. 1995).
groups during monitoring. We assume that this grouping
Radio tags emitting a ‘mortality’ signal, indicating that the
correlates with the likelihood of kea with cryptic scrounging
transmitter was stationary, were retrieved as soon as possible,
histories having scrounged as juveniles because movements
usually within 24 hours of receiving this signal, to determine
of 5–10 km are common for kea, movements of 20 km are
whether the kea had died and the cause of death.
regular, but movements of >20 km are rare (DOC unpubl.
data, Wilson & Brejaart 1992). Some kea in the ‘adjacent’
Factors affecting 1080 poisoning risk
group likely had no history of scrounging, and some long-
Scrounging
distance dispersers in the remote group probably scrounged as
We surmised that a history of scrounging might increase 1080
juveniles, but these exceptions make our results conservative.
poisoning risk due to learned neophilia and/or suppressed
The qualitative nature of this variable necessitates cautious
neophobia. Scrounging histories could not be quantified for
use of any quantitative predictions derived from it.
individual kea due to variability in the ages and locations at
Assignment to a scrounging group of two kea monitored
which birds were enrolled in the study. The scrounging histories
at the Oparara 2014 aerial 1080 operation was complicated
of kea first caught as adults are cryptic at >5 km from scrounging
by the erection of a research hut in the study area partway
sites because only adult male kea living <5 km from scrounging
through the study. These two first-year kea, one male and one
sites display an ongoing daily scrounging habit (scrounging
female, were initially enrolled (i.e. captured and radio tagged)
by adult females is rare) (pers. obs., Jackson 1960; Wilson &
as ‘remote’ study birds in April 2014, at c. 4 months of age.
Brejaart 1992). Adult females at all distances and adult males
During spring 2014, the research hut was erected in the alpine
at >5 km from a scrounging site were never directly observed
zone about 800 metres from where these kea were caught. In
scrounging but may have scrounged as juveniles.
the months leading up to this first-time aerial 1080 operation,
We addressed the problem of cryptic individual scrounging
these two kea were frequently observed at the hut, investigating
histories by dividing our study kea into two groups, ‘adjacent’
the building and associated objects, materials and people. The
and ‘remote’ based on distance to the nearest scrounging site.
occupants of the hut intentionally avoided feeding the kea, but
To assign each kea to a group we calculated an average location
unguarded food was possibly stolen. In the analysis presented
(latitude and longitude) as the mean of all positions recorded
here, these two kea were scored as ‘remote’. We explored the
for that kea during its monitoring history and measured the
consequences of alternative treatment of these two kea with
Figure 3. Histogram of distances between
Adjacent
Remote
mean kea positions (obtained for each kea
40
during monitoring history) and the nearest
scrounging site where kea obtain food from
humans (Fig. 1), showing a clear division
of the sample into two scrounging groups
‘adjacent’ and ‘remote’. All of the recorded
kea deaths are attributed to direct consumption
30
of 1080 baits. Blue = kea not poisoned, pink
= kea poisoned.
ea
20
Number of k
10
0
0
10
20
30
40
70
100
Distance to nearest scrounging site (km)

Kemp et al.: Kea survival during aerial 1080 drops
5
a. Scrounging
b. 1080 History
Figure 4. Numbers of kea monitored and
the number that died with respect to the five
100
modelled variables Scrounging, 1080 History,
ea 90
ea 75
Toxic Bait Size, Age Class and Sex. Blue =
kea not poisoned, pink = kea poisoned. Kea
60
50
were assigned to a scrounging group based on
30
25
distance to nearest scrounging site (‘adjacent’
Number of k
Number of k
≤20 km, ‘remote’ ≥40 km, Fig. 3). 1080
0
0
History is a three-level categorical variable
Adjacent
Remote
1
2
3
for which 1 = first time aerial 1080 treatment
Scrounge group
1080 History
at site, 3 = repeated treatment 2–3 years after
c. Toxic Bait Size
d. Age Class
previous treatment, and 2 = intermediate
150
history including partial block treatments and
ea
150
longer intervals between treatments.
ea
100
100
50
50
Number of k
Number of k
0
0
6
12
> 2 years
1−2 years
<1 years
Bait size (g)
Class
e. Sex
100
ea
75
50
25
Number of k
0
Female
Male
Sex
a. Study kea
Figure 5. Kernel densities (smoothed
histograms) of distance to scrounging
sites for the 222 kea monitoring cases
in this study (a), the 4.5 million hectare
kea species range (b), and the 27% of the
kea species range treated with aerial 1080
Density
in the years 2014–2016 (c). Histograms
were constructed using GIS generated
distances between (1) mean locations
0
10
20
40
80
120
200
for the individual study kea and the
Distance (km) from scrounging sites
nearest scrounging site, and between (2)
b. Entire species range
500 000 and (3) 100 000 random points,
generated within the relevant polygons,
and the nearest scrounging site.
Density
0
10
20
40
80
120
200
Distance (km) from scrounging sites
c. Recent aerial 1080 programme
Density
0
10
20
40
80
120
200
Distance (km) from scrounging sites

6
New Zealand Journal of Ecology, Vol. 43, No. 1, 2019
alternative analyses, in which the birds were (1) excluded
Kea can be accurately assigned to yearly age classes until 2
from the dataset, or (2) classified as ‘adjacent’.
years of age based on cere and eyelid colouration and plumage
We assessed the representativeness of our sample by
condition. Ceres and eyelids are bright yellow at hatching and
visual comparison of kernel density (smoothed histograms)
start fading to grey after 2 years. The rate of fading is variable
of distance to the nearest scrounging site for (1) our sample
after 2 years of age and some adult females retain faint yellow
of monitored kea, (2) the kea species range, and (3) recent
markings, making age cryptic beyond 2 years. We modelled
aerial 1080 use (2014–2016) in kea habitat (Fig. 5a–c). Plots
age as a three-level categorical variable comprising ‘first-year’
for (2) and (3) were constructed using distances from the
(0–1 years), ‘juvenile’ (1–2 years), and ‘adult’ (3+ years) (Table
nearest scrounging site to 500 000 and 100 000 random points,
1). More adult than young individuals were monitored, with
respectively, generated within the relevant polygons. Random
174 adults : 29 juveniles : 19 first-years (Fig. 4d).
points and distances to scrounging sites were generated in the
ESRI ArcMap geographic information platform. Comparing
Sex
plots showed that our study kea poorly represented the South
Sex was determined using body measurements (Bond et al.
Island kea distribution (Fig. 5a, b). A representative sample
1991) and/or by observing breeding behaviour (only female
would have 14% of kea within 20 km of scrounging sites,
kea incubate eggs). We expected that risk might vary between
19%  20–40  km  from  scrounging  sites,  and  67%  >40  km
the sexes due to dimorphic body size (females 750–800 grams,
away. Hence, kea adjacent to scrounging sites were greatly
males 900–1050 grams), and/or to the breeding biology of kea
overrepresented in our study, and kea living 20–40 km were
in which males forage more than females (adult males provide
not represented. Similarly, our study kea do not reflect the
food to females) (Table 1). The sex ratio of monitored birds
full coverage of recent aerial 1080 use (Fig. 5a, c). During
was approximately even, at 119 females : 103 males (Fig. 4e).
the period 2014–2016, 83% of aerial 1080 was applied >40
km from scrounging sites (Fig. 5c).
Statistical analysis
Aerial 1080 history
After 3 weeks from the day of poison application, each
monitored kea was scored as having either survived (0) or
We expected that 1080 poisoning risk at a given site might
died (1). We modelled this binary response using mixed-effects
decrease with repeated aerial 1080 operations through selection
models on the logit scale, with the predictor variables described
for innately cautious phenotypes and/or the creation of bait
above as fixed effects (Scrounging, 1080 History, Bait Size,
shy individuals via sublethal poisoning (conditioned aversion).
Age and Sex). Balance within the dataset was visually assessed
We devised a three-tier classification system based primarily
using pairwise plots (see Appendix S1 in Supplementary
on the time since previous aerial 1080 treatment. Kea at sites
Material). Non-independence of observations within sites was
treated with aerial 1080 for the first time were assigned to
category 1. Kea at sites that had been comprehensively treated
addressed by specifying site as a random effect. We did not
(entire block) within the previous 3 years were assigned to
address the additional non-independence of the 17 monitoring
category 3. Six operations did not clearly fall into either of
occasions involving birds that we had previously monitored
these categories due to extended inter-treatment intervals and/
because their 1080 history was modelled as a fixed effect. Social
or partial treatments of blocks. Kea in these operations were
interactions among kea may also cause non-independence, but
assigned to category 2. Four of the 19 operations in the study
we believe this effect will be limited because most of our birds
were first time treatments. Eight of the operations were at
were adult kea, which are largely solitary and unlikely to be
sites treated within the previous 3 years. The remaining seven
influenced by the behaviour of other kea. We could not test
operations were at sites previously partially treated and/or
for overdispersion in our data because established statistical
treated 4 or more years prior (Table 1). Our sample comprised
methods applicable to binary data are lacking.
37, 105 and 80 kea in categories 1–3, respectively (Fig. 4b).
Our model set comprised all possible additive combinations
of the predictor variables, with no interactions due to limited
Bait size
sample size. Modelling was implemented in R (R Core Team
2016) using package MuMIn version 1.40.4. (Barton 2018). We
The aerial 1080 operations used either 6 g or 12 g poison
used Information Theoretic methods (AIC
pellets (Table 1) with equal toxin concentration (1.5 mg
c) to rank models and
assess the relative importance of the five modelled variables
g−1). Because a 12 g bait contains twice as much toxin we
(Akaike 1973; Burnham & Anderson 2002). For supported
expected that these larger pellets might elevate risk to kea.
variables we use the odds ratio to express the magnitude of
Our sample comprised 76 and 146 records for 6 g and 12 g
the effect, calculated as exponentiated model coefficients. For
pellets, respectively (Fig. 4c).
example, the odds ratio showing the effect of variable y in a
Age
two-factor additive model is:
Kea have an extended juvenile phase, in which independence is

𝑂𝑂𝑅𝑅
(1)
attained at 1–6 months of age, followed by sexual maturity at
𝑦𝑦 = 𝑒𝑒𝛽𝛽𝑖𝑖+𝛽𝛽𝑥𝑥+𝛽𝛽𝑦𝑦
𝑒𝑒𝛽𝛽𝑖𝑖+𝛽𝛽𝑥𝑥
3–4 years (JRK unpubl. data). Pre-adult kea tend to congregate
where  β
into dynamic mixed-age flocks and can range over tens of square
i,  βx  and  βy  are  the  model  coefficients  for  the
intercept, variable x, and variable y, respectively. We used
kilometres (Jackson 1960; Wilson & Brejaart 1992). Once
the or_glm function in the oddsratio package version 1.0.2.
mature, kea mate for life. Adult kea are relatively sedentary,
for R (Schratz 2017) to calculate odds ratios and associated
with core ranges spanning only a few square kilometres and
95% confidence intervals.
social interactions are primarily with the mate and any recent
offspring (JRK pers. obs.; Jackson 1960; Wilson & Brejaart
1992). We expected that younger kea might be more susceptible
to 1080 poisoning because they are in a phase of intensive
learning and heightened behavioural flexibility.

Kemp et al.: Kea survival during aerial 1080 drops
7
Table 2. Ranking of the top ten alternative models by AICc of factors affecting kea survival through aerial 1080 operations.
Scrounging = 2-level categorical variable (adjacent versus remote). 1080 History = 3-level categorical variable, scored per
aerial 1080 operation based on time since previous aerial 1080 treatment (Table 1). k = number of parameters in the model.
__________________________________________________________________________________________________________________________________________________________________
Model
k
Delta AICc Weight
__________________________________________________________________________________________________________________________________________________________________
Scrounging + 1080 History
5
0
0.352
Scrounging + 1080 History + Age Class
7
1.359
0.179
Scrounging + 1080 History + Sex
6
2.366
0.108
Scrounging + 1080 History + Bait Size
6
2.466
0.103
Scrounging + 1080 History + Age Class + Sex
8
3.86
0.051
Scrounging + 1080 History + Age Class + Bait Size
8
4.045
0.047
Scrounging + 1080 History + Sex + Bait Size
7
4.938
0.03
Scrounging + 1080 History + Age Class + Sex + Bait Size
4
5.037
0.028
1080 History
3
6.51
0.014
Scrounging
9  6.663 0.013
__________________________________________________________________________________________________________________________________________________________________
Figure 6. Survival rates for kea exposed
Adjacent
Remote
to aerial 1080, estimated using the
best-supported model, with respect to
100
●
●
Scrounging  (‘remote’  ≥40  km  from
●
scrounging sites, ‘adjacent’ ≤20 km) and
●
1080 History (1 = first time aerial 1080
90
treatment at site, 3 = repeated treatment
●
2–3 years after previous treatment and 2
80
= intermediate history including partial
block treatments and longer intervals
between treatments). Error bars are 95%
70
binomial confidence intervals.
60
●
al (%) 50
viv
Sur 40
30
20
10
0
1
2
3
1
2
3
1080 History
Results
strongly supported by AICc as a negative influence on the
chance of a kea surviving an aerial 1080 operation (Table 2;
We recorded 24 kea deaths, all within six operations (n = 222
Fig. 6). The odds of survival was 6.9 times higher for kea in
kea monitoring cases; Table 1). Poison-pellet cereal matter
the ‘remote’ group than for the ‘adjacent’ group (95% CI =
present in the digestive tracts of all necropsied corpses (n =
2.1–31.6). An effect of 1080 History was also supported (Table
17) indicated that primary poisoning was the cause of death.
2). The odds of survival was 21.3 times higher at sites with
Necropsies were not conducted in seven cases owing to
a recent aerial 1080 treatment than at first time treatments
advanced decomposition that resulted from logistical delays
(95% CI = 3.6–406.9). The best-supported model contained
retrieving  corpses  from  the  field.  The  seven  deaths  were
both Scrounging and 1080 History. The Age Class variable
considered for the study as due to 1080 poisoning.
also had moderate support. However, because this result may
Membership of the ‘adjacent’ scrounging group was
be an artefact of small sample sizes of young kea, we present

8
New Zealand Journal of Ecology, Vol. 43, No. 1, 2019
further results for the best-supported model only. There was
Therefore, we recommend further monitoring of young kea
little to no support for Bait Size and Sex as predictors. The
through aerial 1080 operations to improve our understanding
random effect of Site was non-zero for all models that did not
of how kea age influences 1080 poisoning risk.
include both Scrounging and 1080 History, but was near-zero
for all models that included both.
Utility of estimates and extrapolation to other sites
Support for Age Class was absent in two alternative
While primarily aiming to evaluate for a qualitative relationship
analyses in which two first-year kea poisoned during the
between scrounging and 1080 poisoning risk, our study has
Oparara 2014 operation were either excluded or scored as
generated a useful estimate of survival of remote kea through
‘adjacent’ for scrounging (see Methods). The survival rates
a repeated aerial 1080 operation. With only one death out of
predicted by the best-supported model (Scrounging + 1080
60 remote monitoring occasions in repeated operations, the
History) for remote first time treatments improved considerably
lower 95% confidence limit for kea survival in this context
in the alternative analyses (Appendix S2), but neither alternative
is 95.6% (Fig. 6). Given the high magnitude of benefits from
suggested different conclusions or recommendations.
predator control measured for kea and other species with
similar nesting ecology, such as kākā and whio (Hymenolaimus
malacorhynchos) (Moorhouse et al. 2003; Whitehead et al.
Discussion
2008; Kemp et al. 2018), we expect to see strong kea population
growth resulting from long term aerial 1080 programmes in
Mechanisms behind 1080 poisoning risk
remote areas.
Our finding of lower poisoning risk from aerial 1080 operations
The estimated survival rates derived from our analysis for
for remote kea than for kea living adjacent to scrounging sites
kea adjacent to scrounging sites should be treated with caution
can be explained if scrounging causes suppression of innate
due to the qualitative nature of our method for assigning kea to
neophobic behaviours that protect against poisoning from novel
the ‘adjacent’ and ‘remote’ groups (see Methods). In addition,
foods (Rozin 1976; Galef et al. 1998; Galef & Whiskin 2001).
aerial 1080 operations near scrounging sites vary in size, shape
Suppression of neophobia could potentially be learned through
and location relative to scrounging sites, and the intensity
frequent exposure to rewarding novel foods, and/or a direct
of scrounging activity varies from place to place. Further,
result of sublethal lead poisoning. Lead poisoning is prevalent
our samples possibly contained disproportionate numbers of
among kea near scrounging sites (McLelland et al. 2010; Reid
scrounging kea – these being presumably easier to catch for the
et al. 2012) and this could potentially contribute to suppressed
same reason that they are more prone to ingesting 1080 pellets
neophobia through effects on brain function, thermoregulation,
– hence possibly leading to over-estimated risk. Interpolation
immune system function and body condition (Newth et al.
of survival rates to kea living 20–40 km from scrounging sites
2016). We could not assay our study birds for blood lead levels
is especially to be avoided, for the reasons given above and
for logistical reasons, so we could not distinguish the relative
because the shape of the relationship between kea risk and
importance of learned neophilia versus lead poisoning. Thus,
distance is unknown. Notwithstanding these cautions, we
in remote sites, initial interactions with 1080 pellets may
conclude that 1080 poisoning risk near scrounging sites is cause
comprise a mixture of neophilia and neophobia, in which kea
for concern. Benefits from predator control would need to be
may approach novel objects (neophilia), but initial interactions
high and of extended duration (i.e. more than two breeding
with the novel item are cautious (neophobia). The short field
seasons) to achieve a positive net outcome under our lowest
life of the 1080 pellets used in kea habitat appears to preclude
mean survival estimate of 56.8% survival at first time aerial
their acceptance as a familiar rewarding food in the remote
1080 treatments adjacent to scrounging sites (Fig. 6). Should
context. Thus, the potential for high 1080 poisoning risk to kea
survival fall between the point estimate and the lower 95%
does not necessarily manifest under current baiting practices.
confidence limit of 35.9%, achieving a positive net outcome
Our finding of lessened risk where aerial 1080 had been
in these circumstances is beyond reasonable expectations.
used in recent years could be explained by (1) learned aversion
Only 33% of the kea species range lies within 40 km of
(bait shyness) induced by sublethal 1080 doses, and (2) selection
scrounging sites and only 17% of recent (2014–2016) aerial
for neophobic individuals due to mortality of neophilic ones
1080 applications have been within this range (Fig. 5). Thus,
from 1080 poisoning. However, bait shyness can be precluded
costs to the kea population from aerial 1080 use adjacent to
by using prefeed, at least in some animals (Ross et al. 2000),
scrounging sites might have been offset by benefits accrued
and all the operations in this study were prefed. Therefore,
in remote areas. However, the limitations of our study and the
we favour the latter explanation that strong selection pressure
potential to rapidly reduce the prevalence of scrounging by kea
applied by aerial 1080 results in an increasing prevalence of
are such that we recommend immediate mitigation action for
innate neophobic behaviours. Possibly the effect of 1080 history
the benefit of kea population recovery and to address concern
differs between ‘remote’ and ‘adjacent’ groups, but we had
about kea welfare.
insufficient data to test for such an interaction. To enable this
idea to be tested, we recommend further radio-tag monitoring
Managing aerial 1080 poisoning risk
of kea at remote first-time treatments, if DOC’s and/or TbFree
Approaches to managing 1080 poisoning risk to kea near
New Zealand’s South Island aerial 1080 programme expands.
scrounging sites potentially include (1) reducing scrounging
Our study found moderate support for an effect of Age
by removing the rewards that kea obtain from anthropogenic
Class on 1080 poisoning risk, which may reflect the enhanced
sources, (2) not prefeeding aerial 1080 operations near
behavioural plasticity observed for juveniles within the genus
scrounging sites, (3) using low-strength 1080 pellets (0.8 mg
Nestor (Wilson et al. 1998; Loepelt et al. 2016). We caution
g−1 pellets are available for use in New Zealand), (4) adding
that our data set contained relatively few young kea, and that
bird repellents to baits, (5) aversion training of individual
two poisoned first-year kea were scored as ‘remote’ although
kea at scrounging sites, and (6) spatial solutions such as
they are known to have scrounged around a newly built hut.
attracting kea out of treatment blocks during aerial 1080

Kemp et al.: Kea survival during aerial 1080 drops
9
operations, and capture and holding of kea during aerial
Summary
1080 operations. However, not all these options are currently
In summary, we are confident that aerial 1080 application is not
viable. Not prefeeding will almost certainly compromise pest
cause for concern for kea in repeated aerial 1080 operations at
control effectiveness (Coleman et al. 2007; Nugent et al.
remote sites, particularly when the operations are designed to
2011), reducing benefits to native species other than kea. Bird
maximise predator control benefits to kea, as risk to kea is easily
repellents and aversion training are potentially viable solutions,
offset by these benefits. Proximity to scrounging sites is strongly
but both require research and development (Orr-Walker et al.
associated with elevated 1080 poisoning risk, but this risk is
2012; Cowan et al. 2016; Crowell et al. 2016). Spatial solutions
not well quantified. We recommend (1) the implementation of
involving catching and holding kea would be stressful for
a scrounge-management programme, (2) research to improve
the birds and some will always avoid capture. Attracting kea
our understanding of the links between scrounging, neophobia
away from 1080 blocks may not succeed, especially for adult
and 1080 poisoning risk, (3) further measurement of risk to
females due to their more sedentary nature. The remaining
young kea, and (4) further measurement of risk at first treatments
two options – reducing scrounging and using lower strength
in remote sites. If the South Island aerial 1080 programme
1080 pellets – are potentially viable and could be immediately
comprises mostly remote repeated operations, then a positive
implemented. Managers should focus first on preventing the
long term net outcome for kea is likely.
establishment of new scrounging sites, which is a real risk
in the context of rapidly expanding tourism in New Zealand.
Acknowledgements
Second, the attractiveness of established scrounging sites should
be actively reduced through careful management of waste and
This research was carried out by DOC with co-funding from
of human behaviour. The effectiveness of any such programme
TBfree NZ, a subsidiary of OSPRI New Zealand. The Kea
should be tested by re-evaluating aerial 1080 poisoning risk near
Conservation  Trust  contributed  significantly,  through  the
managed scrounging sites after 5 to 10 years of management.
funding of ongoing field studies in the Hawdon and Wangapeka
Scrounging may never be eliminated, but the current level of
Valleys. Radio tagging and banding of kea was conducted in
effort to prevent scrounging is very low (JRK pers. obs.) and
accordance with Department of Conservation Animal Ethics
the recent reduction of scrounging rates at back-country huts
Guidelines. For their significant contributions to the field effort
suggests similar reductions are achievable at Arthur’s Pass,
that went into this study we thank Francesca Cunninghame,
Glacier Country, and Milford Road. A successful scrounge-
Joseph Fraser, Troy Makan, Dan Palmer, Brent Barrett, Josh
reduction programme would have the additional benefit of
Adam, Jacinda Amey, Kirsty Moran, Robyn Blyth, Tom
reducing mortality arising from other hazards such as lead
Goodman, Lucy Rossiter and Liam Bolitho. We thank the
and cars. Using lower strength 1080 pellets is not current best
staff of the various DOC Offices that facilitated our field work.
practice, but the recent change from 0.8 mg g−1 to 1.5 mg g−1
Comments from Susan Timmins, Clayson Howell, Ximena
was based on theory (Frampton et al. 1999) and benefits to
Nelson and Alex Taylor greatly improved the manuscript, for
pest control efficacy of the higher concentration have not been
which we are grateful. We also thank Luis Ortiz-Catedral and
experimentally demonstrated in the field.
Deb Wilson (NZJE referee and associate editor), plus one other
anonymous referee for their valuable input to the manuscript.
Potential for prefeed to affect 1080 poisoning risk
We were unable to quantify the effect of prefeeding on 1080
poisoning risk to kea because all of the operations in the study
References
were prefed. However, risk to remote living kea is clearly not
elevated to concerning levels by a single application of prefeed
Akaike H 1973. Information theory and an extension of the
at the sowing rates used in this study. Changes to 1080 baiting
maximum likelihood principle. In: Petrov BN, Csaki F
practices, such as using two rounds of prefeeding (e.g. Nugent
eds. Budapest: Akademiai Kiado, Proceedings of the
et al. 2011) or longer life baits such as Wanganui Number
2nd International Symposium on Information Theory.
7 pellets (Bowen et al. 1995), could elevate risk to kea and
Pp. 267–281.
therefore should be accompanied by kea monitoring. The
Barton K 2018. MuMIn: Multi-Model Inference; R
high survival of remote adult kea is consistent with studies
package version 1.40.4. https://CRAN.R-project.org/
of kākā, in which adults exhibited a disinclination to explore
package=MuMIn (accessed May 2018).
potential new food sources (Wilson et al. 1998; Loepelt et al.
Bond AB, Wilson KJ, Diamond J 1991. Sexual dimorphism
2016).  Juvenile  kākā,  conversely,  learned  to  exploit  novel
in the kea Nestor notabilis. Emu 91: 12–19.
foods more readily. For kea, which learn socially (Diamond
Bowen LH, Morgan DR, Eason CT 1995. Persistence of
& Bond 1991; Gajdon et al. 2004), 1080 poisoning risk to
sodium monofluoroacetate (1080) in baits under simulated
juveniles may be naturally mitigated through these learning
rainfall. New Zealand Journal of Agricultural Research
processes. However, young kea sometimes forage and explore
38: 529–531.
independently of adults, at which time they may investigate
Burnham KP, Anderson DR 2002. Model selection and
baits. If so, their investigation of baits may be coupled with a
multimodel inference: a practical information-theoretic
cautious testing strategy (Rozin 1976) within which a single
approach. 2nd edn. New York, Springer-Verlag. 488 p.
prefeed  application  may  provide  sufficient  familiarisation
Coleman JD, Caley P 2000. Possums as a reservoir of bovine
time to elevate 1080 poisoning risk to a concerning level.
TB. In: Montague TL ed. The brushtail possum. Biology,
We recommend behavioural studies of the process by which
impact and management of an introduced marsupial.
foods become accepted by kea to improve understanding of
Lincoln, Manaaki Whenua Press. Pp. 92–104.
the role of prefeeding in determining 1080 poisoning risk, and
Coleman JD, Fraser KW, Nugent G 2007. Costs and benefits
the potential implications of multiple prefeed applications.
of pre-feeding for possum control. New Zealand Journal
of Zoology 34: 185–193.

10
New Zealand Journal of Ecology, Vol. 43, No. 1, 2019
Cowan P, Booth L, Crowell M 2016. Repellents with potential
Fraser I, August A, August C 2003. Control of introduced
to protect kea and other native birds from aerial poisoning
mammalian predators improves kaka Nestor meridionalis
for possum and rat control. New Zealand Journal of
breeding success: reversing the decline of a threatened
Ecology 40: 29–41.
New Zealand parrot. Biological Conservation 110: 33–44.
Crowell M, Martini M, Moltchanova E 2016. Effect of the
Murphy EC, Robbins L, Young JB, Dowding JE 1999.
addition of bird repellents to aerially applied 1080 baits
Secondary poisoning of stoats after an aerial 1080 poison
on rat and possum abundance. New Zealand Journal of
operation in Pureora forest, New Zealand. New Zealand
Ecology 40: 49–59.
Journal of Ecology 23: 175–182.
Diamond J, Bond AB 1991. Social behavior and the ontogeny
Newth JL, Rees EC, Cromie RL, McDonald RA, Bearhop
of foraging in the kea (Nestor notabilis). Ethology 88:
S, Pain DJ, Norton GJ, Deacon C, Hilton GM 2016.
128–144.
Widespread exposure to lead affects the body condition
Elliott G, Kemp J 2016. Large-scale pest control in
of free-living whooper swans Cygnus cygnus wintering
New Zealand beech forests. Ecological Management &
in Britain. Environmental Pollution 209: 60–67.
Restoration 17: 200–209.
Nugent G, Warburton B, Thomson C, Sweetapple P, Ruscoe WA
Frampton CM, Warburton B, Henderson RJ, Morgan DR 1999.
2011. Effect of prefeeding, sowing rate and sowing pattern
Optimising bait size and 1080 concentration (sodium
on efficacy of aerial 1080 poisoning of small-mammal
monofluoroacetate) for the control of brushtail possums
pests in New Zealand. Wildlife Research 38: 249–259.
(Trichosurus vulpecula). Wildlife Research 26: 53–59.
Orr-Walker T, Adams NJ, Roberts LG, Kemp JR, Spurr EB
Gajdon GK, Fijn N, Huber L 2004. Testing social learning in a
2012. Effectiveness of the bird repellents anthraquinone
wild mountain parrot, the kea (Nestor notabilis). Learning
and D-pulegone on an endemic New Zealand parrot, the
& Behavior 32: 62–71.
kea (Nestor notabilis). Applied Animal Behaviour Science
Galef BGJ, Rudolf B, Whiskin EE, Choleris E, Mainardi M,
137: 80–85.
Valsecchi P 1998. Familiarity and relatedness: effects on
O’Donnell C, Hoare J 2012. Quantifying the benefits of long-
social learning about foods by Norway rats and Mongolian
term integrated pest control for forest bird populations in
gerbils. Animal Learning and Behavior 26: 448–454.
a New Zealand temperate rainforest. New Zealand Journal
Galef BGJ, Whiskin EE 2001. Interaction of social and
of Ecology 36: 131–140.
individual learning in food preferences of Norway rats.
Powlesland RG, Knegtmans JW, Marshall I 1999. Costs and
Animal Behaviour 62: 41–46.
benefits of aerial 1080 possum control operations using
Green W, Rohan M 2012. Opposition to aerial 1080 poisoning
carrot baits to North Island robins (Petroica australis
for control of invasive mammals in New Zealand: risk
longipes), Pureora Forest Park. New Zealand Journal of
perceptions and agency responses. Journal of the Royal
Ecology 23: 145–159.
Society of New Zealand 42: 185–213.
Powlesland RG, Wills DE, August ACL, August C 2003.
Greene TC, Dilks PJ, Westbrooke IM, Pryde MA 2013.
Effects of a 1080 operation on kaka and kereru survival
Monitoring selected forest bird species through aerial
and nesting success, Whirinaki Forest Park. New Zealand
application of 1080 baits, Waitutu, New Zealand.
Journal of Ecology 27: 125–137.
New Zealand Journal of Ecology 37: 41–50.
R Core Team 2016. R: a language and environment for statistical
Greer A, Gajdon GK, Nelson XJ 2015. Intraspecific variation
computing. R Foundation for Statistical Computing.
in the foraging ecology of kea, the world’s only mountain-
Vienna, Austria.
and rainforest-dwelling parrot. New Zealand Journal of
Reid C, McInnes K, McLelland JM, Gartrell BD 2012.
Ecology 39: 254–261.
Anthropogenic lead (Pb) exposure in populations of a
Jackson JR 1960. Keas at Arthurs Pass. Notornis 9: 39–58.
wild parrot (kea Nestor notabilis). New Zealand Journal
Jackson JR 1963. The nesting of keas. Notornis 10: 319–326.
of Ecology 36: 56–63.
Kemp JR, Mosen C, Elliott GP, Hunter CM 2018. Effects of
Robertson CJR, Hyvonen P, Fraser M, Pickard CR 2007. Atlas
the aerial application of 1080 to control pest mammals
of bird distribution in New Zealand. Wellington, The
on kea reproductive success. New Zealand Journal of
Ornithological Society of New Zealand. 296 p.
Ecology 42: 158–168.
Ross JG, Hickling GJ, Morgan DR, Eason C 2000. The role
de Kloet RS, de Kloet SR 2005. The evolution of the spindlin
of non-toxic prefeed and postfeed in the development and
gene in birds: sequence analysis of an intron of the
maintenance of 1080 bait shyness in captive brushtail
spindlin W and Z gene reveals four major divisions of the
possums. Wildlife Research 27: 69–74.
Psittaciformes. Molecular Phylogenetics and Evolution
Rozin P 1976. The selection of food by rats, humans, and other
36: 706–721.
animals. Advances in the Study of Behaviour 6: 21–76.
Loepelt J, Shaw RC, Burns KC 2016. Can you teach an old
Schratz P 2017. R package ’oddsratio’: Odds ratio calculation
parrot new tricks? Cognitive development in wild kaka
for GAM(M)s & GLM(M)S. R package version 1.0.2.
(Nestor meridionalis). Proceedings of the Royal Society
https://CRAN.R-project.org/package=oddsratio (accessed
B 283: 20153056.
May 2018).
McLelland JM, Reid C, McInnes K, Roe WD, Gartrell BD 2010.
Van Vianen J, Burge OR, MacFarlane AT, Kelly D 2018. The
Evidence of lead exposure in a free-ranging population
effects of single aerial 1080 possum-control operations on
of kea (Nestor notabilis). Journal of Wildlife Diseases
common forest birds in the South Island, New Zealand.
46: 532–540.
New Zealand Journal of Ecology 42: 169–178.
Moorhouse R 1997. The diet of the North Island kaka
Veltman CJ, Westbrooke IM 2011. Forest bird mortality and
(Nestor meridionalis septentrionalis) on Kapiti Island.
baiting practices in New Zealand aerial 1080 operations
New Zealand Journal of Ecology 21: 141–152.
from 1986 to 2009. New Zealand Journal of Ecology
Moorhouse RJ, Greene TC, Dilks PJ, Powlesland RG, Moran
35: 21–29.
L, Taylor GE, Jones A, Knegtmans JW, Wills D, Pryde M,
Whitehead AL, Edge K-A, Smart A, Hill G, Willans M 2008.

Kemp et al.: Kea survival during aerial 1080 drops
11
Large scale predator control improves the productivity of a
rare New Zealand riverine duck. Biological Conservation
141: 2784–2794.
Wilson K-J, Brejaart R 1992. The kea: a brief research review.
In: Joseph L ed. Issues in the conservation of parrots
in Australasia and Oceania: challenges to conservation
biology. Royal Australian Ornithologists Union Report
no. 83. Victoria, Royal Australian Ornithologists Union.
Pp. 24–28.
Wilson PR, Karl BJ, Toft RJ, Beggs JR, Taylor RH 1998. The
role of introduced predators and competitors in the decline
of kaka (Nestor meridionalis) populations in New Zealand.
Biological Conservation 83: 175–185.
Wright J 2011. Evaluating the use of 1080: predators, poisons
and silent forests. Wellington, Office of the Parliamentary
Commissioner for the Environment. 85p.
Young LM, Kelly D, Nelson XJ 2012. Alpine flora may
depend on declining frugivorous parrot for seed dispersal.
Biological Conservation 147: 133–142.
Received 19 December 2017; accepted 27 June 2018
Editorial board member: Deb Wilson
Supplementary Material
Additional supporting information may be found in the
supplementary material file for this article:
Appendix S1. Pairwise plots of modelled predictor variables
of kea survival of aerial 1080 operations.
Appendix S2. Survival rates for kea exposed to aerial 1080,
estimated using the top model, with respect to Scrounging and
1080 History, with respect to three alternative categorisations of
two juvenile kea monitored through the Oparara 2014 operation.
The New Zealand Journal of Ecology provides supporting
information supplied by the authors where this may assist
readers. Such materials are peer-reviewed and copy-edited
but any issues relating to this information (other than missing
files) should be addressed to the authors.