Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 



 

 

 

Investigating the habitat suitability of 

Maungatautari Ecological Island for the 

reintroduction of kākāpō (Strigops habroptilus) 

 

 

 

 

 

A thesis submitted in partial fulfilment of the requirements for the degree of  

Master of Science in Conservation Biology  

Massey University, Palmerston North 

New Zealand 

 

 

 

 

 

Alexandra J. Hurley 

2017



 

 

“They are our national monuments. They are our Tower of London, our Arc de Triomphe, 

our pyramids. We don’t have this ancient architecture that we can be proud of and swoon 

over in wonder, but what we do have is something that is far, far older than that. No one else 

has kiwi, no one else has kākāpō. They have been around for millions of years, if not 

thousands of millions of years. And once they are gone, they are gone forever. And it’s up to 

us to make sure they never die out." 

 

Don Merton  



 

Abstract 
 

 

The kākāpō (Strigops habroptilus) is a large, flightless parrot endemic to New Zealand which 

was once abundant across mainland New Zealand. However, this nocturnal bird species is now 

listed as critically endangered with a population of approximately 154 individuals. Kākāpō are 

currently only found on four offshore, predator-free islands where kākāpō were not found 

historically - Whenua Hou/Codfish Island, Hauturu-o-Toi/Little Barrier Island, Anchor Island 

and an unnamed island in Fiordland. However, there is hope to have kākāpō living on mainland 

New Zealand again with the potential reintroduction of kākāpō to Maungatautari Ecological 

Island in the near future. Kākāpō breeding is heavily dependent on environmental factors, 

specifically in that breeding coincides with the mast fruiting of specific plant species, 

particularly rimu (Dacrydium cupressinum). Therefore, kākāpō breed only once every two to 

five years which significantly constrains the potential population growth of the species. 

However, with a record breeding season in 2016 and expectations for kākāpō breeding to 

continue successfully, kākāpō populations on Whenua Hou and Anchor Island are considered 

to be nearing capacity. Therefore, identifying sites that contain the environmental factors that 

favour kākāpō survival and reproduction is an important task. Additionally, finding new habitat 

sites will help mitigate catastrophe risks and may be used for kākāpō advocacy. The purpose 

of this research is to assess the habitat suitability of Maungatautari Ecological Island as a 

potential site for the reintroduction of kākāpō, specifically assessing the density of selected tree 

species known to induce nesting; and modelling habitat suitability based on key habitat features 

known to influence kākāpō distribution. 

 

A total of 260 adult trees were identified during a distance line transect survey DISTANCE 

analysis and were used to estimate the density of key tree species across Maungatautari. A 

maximum distance a priori was set at 100 m and so only trees observed within 100 m of the 

transect line were recorded. An a priori for the minimum diameter at breast height (DBH) was 

also set at 30 cm. The results of this analysis found density estimates of 1.113 stems/ha for 

adult rimu and 2.310 stems/ha for other key adult tree species across the entire study area. 

These findings are not at all comparable to the stem densities estimated on Whenua 

Hou/Codfish Island where kākāpō already inhabit and have successfully bred.  However, 

although rimu and other key tree species occur at higher densities on Whenua Hou, the Whenua 

Hou tree population are most probably much smaller in size in comparison to those on 

Maungatautari. This suggests that a comparison of basal areas or size distributions between the 

two sites would indicate that rimu may be closer in biomass to Whenua Hou than in density. 



Abstract 

Therefore, based on these comparisons of density, we should not discredit that rimu and other 

key tree species may occur on Maungatautari at sufficient levels to at least induce breeding 

attempts by female kākāpō if they were to be reintroduced. 

This research also combined GIS spatial analysis tools with multi-criteria evaluation (MCE) 

methods to create a model of habitat suitability which can be used to identify the areas of 

Maungatautari most likely to sustain kākāpō and support their breeding. The computed habitat 

suitability map predicted that suitable and moderately suitable breeding habitat for kākāpō 

occupied 5% and 3% of the mainland island’s area respectively. These areas of suitable and 

moderately suitable habitat occurred predominantly in the southern regions and in some central 

regions of the mountain at moderate to high altitude levels. I predict that these areas of the 

mountain are likely to provide the most adequate sustenance and support for kākāpō survival 

and breeding, particularly in low podocarp mast years. Habitat located at low altitudes, around 

the outer regions of the mountain, and in gullies in the central regions are predicted to be 

unsuitable for breeding, particularly in years when podocarp and rimu fruit supply is limited. 

Areas predicted to be unsuitable for kākāpō occupied 77% of the total area on Maungatautari. 

To increase the likelihood of a successful reintroduction to Maungatautari, it is necessary to 

release kākāpō into areas likely to support survival and breeding. Therefore, I recommended 

that the first cohort of reintroduced kākāpō should be released in the southern and central 

regions of the mountain at moderate to high altitude levels prior to any other region of 

Maungatautari. Additionally, modifications need to be made to the existing Xcluder™ fence 

prior to a reintroduction to prevent kākāpō from climbing outside of the sanctuary boundaries. 



 

Acknowledgements 
 

 

I would like to acknowledge a number of people for their continuous support during the various 

stages of this study, without whom this research would not have been possible. Firstly, I would 

like to thank my supervisor, Prof. Doug Armstrong, for his academic advice, expertise and 

constructive feedback throughout this thesis. I would also like to acknowledge the input and 

support from my co-supervisors, John Innes and Andrew Digby.  

 

I would like to thank those that helped extensively with fieldwork, namely Brooke Donoghue 

and Ash Rowe for their assistance and long hours with me in the field. A special thanks to Rod 

Miller for his advice on Maungatautari’s monitoring tracks and for organising transport for 

myself and my field assistants. Also, a big thank you to his team, Warren, Steve, Richard and 

the rest of the MEIT team for their 4WD skills and transport to my survey track start points 

each day. I would also like to thank Matthew Lark for assisting with fieldwork permit approval 

and for coordinating the start date of my fieldwork with the MEIT team.  

 

A special thank you to the Department of Conservation’s Kākāpō recovery team and the Waipa 

District Council for their financial support that helped make this study possible.  

 

Finally, a heartfelt thank you to my family and friends for their continuous support over the 

years, particularly during my Masters study. A special thank you to my parents, Vince and 

Carol for always supporting all of my endeavours and encouraging me in every aspect of life. 

 



 

Table of Contents 
 

 

 

Chapter 1 General Introduction ....................................................................................... 1 

1.1 Statement of research problem and stimulus for this research ...............................................2 
1.2 Habitat ....................................................................................................................................4 
1.3 Current knowledge - Kākāpō .................................................................................................7 
1.4 Study Site - Maungatautari Ecological Island ......................................................................18 
1.5 Research objectives ..............................................................................................................23 
1.6 Thesis structure ....................................................................................................................24 

 

Chapter 2  Estimating the density of key tree species on Maungatautari ................... 25 

2.1 Introduction ..........................................................................................................................26 
2.2        Methods ................................................................................................................................28 
2.3 Results ..................................................................................................................................36 
2.4 Discussion ............................................................................................................................43 

 

Chapter 3  Predictive habitat suitability modelling of Maungatautari as a 

reintroduction site .......................................................................................... 48 

3.1 Introduction ..........................................................................................................................49 
3.2 Methods ................................................................................................................................51 
3.3 Results ..................................................................................................................................61 
3.4 Discussion ............................................................................................................................73 

 

Chapter 4  General Discussion ......................................................................................... 79 

4.1 Summary of findings ............................................................................................................80 
4.2 Management recommendations ............................................................................................82 
4.3 Future research .....................................................................................................................83 

 

References ................................................................................................................................ 86 

Appendices ............................................................................................................................. 101 

 

 

 



 

 1 

 

Chapter 1 General Introduction 

 

 

 

Sirocco, an adult male kākāpō on Maud Island. Photo credit: Chris Birmingham 

 

 



Chapter 1  General Introduction

 

 2 

1.1 Statement of research problem and stimulus for this research  

 

Reintroductions are attempts to restore locally extinct populations to parts of their historical 

ranges where they were extirpated (Armstrong & Seddon, 2008; Mihoub et al., 2013). Whilst 

many reintroduction attempts have failed, (Armstrong & Seddon, 2008) a number have 

succeeded and resulted in the restoration of a number of high profile New Zealand bird species 

from the brink of extinction, including the Chatham Island black robin (Petroica traversi) and 

the South Island saddleback (Philesturnus carunculatus), to predator free islands (Sutherland 

et al., 2010).  

 

International reintroduction guidelines have been under development for some time now, and 

the most fundamental guideline is that it is essential to evaluate the ecological features of a 

potential release site and the habitat use of a species prior to its release (IUCN, 2013).  Habitat 

plays a crucial role in population persistence and regardless of the strategy used to establish a 

population, a reintroduction will fail if the habitat at the release site cannot support the species 

(Armstrong & Seddon, 2008). Therefore, establishing what habitat conditions are needed for 

the persistence of the reintroduced population is fundamental to any successful reintroduction. 

Assessing ecological features of an environment prior to a reintroduction can facilitate 

management planning and actions in a number of ways. These include determining the habitat 

suitability of the proposed release site and the potential distribution of the species, as well as 

being able to project the long-term viability of reintroduced populations (Mihoub et al, 2013). 

Furthermore, assessment of habitat prior to release can help in the design and cost-effectiveness 

of reintroduction release strategies (Sutherland et al., 2010; Mihoub et al., 2013).  

 

The kākāpō (Strigops habroptilus) is a large, flightless parrot endemic to New Zealand. This 

nocturnal bird species was once abundant across mainland New Zealand but is now classified 

as critically endangered (IUCN) with an adult population of approximately 154 individuals 

recorded in 2017 (Elliott, 2017). Kākāpō are currently only found on four offshore, predator-

free islands where kākāpō were not found historically – Whenua Hou/Codfish Island, Hauturu-

o-Toi/Little Barrier Island, Anchor Island and an unnamed island in Fiordland. However, there 

is hope to again have kākāpō living on mainland New Zealand in the near future with the 

potential reintroduction of kākāpō to Maungatautari Ecological Island, located 18 km south of 



Chapter 1  General Introduction

 

 3 

Cambridge in the Waikato Region of New Zealand (McQueen, 2004; Smuts-Kennedy & 

Parker, 2013).  

 

Kākāpō breeding has been found to coincide with mast fruiting of specific tree species, 

predominantly rimu (Dacrydium cupressinum) trees. Therefore, kākāpō typically only breed 

once every two to five years (Elliott et al., 2006). This relationship between mast seeding and 

nesting in kākāpō is speculated to be due to a cognitive trigger rather than a nutritional or 

chemical trigger (Harper et al., 2006). Harper et al., (2006) hypothesised that female kākāpō 

nest in response to unripe fruit crops because the presence of these unripe fruit predicts an 

abundant supply of nutritious ripe fruit during autumn, the period in which kākāpō raise their 

young. Rimu seed is believed to be the predominant nesting trigger for kākāpō as the ripe fruit 

of rimu is higher in protein, fat and vitamin D as well as being easier to collect than many other 

plant species in the kākāpō diet (Eason & Moorhouse, 2006; von Hurst et al. 2015). It has also 

been found that when rimu fruit is sufficiently abundant it is the sole food source provided to 

nestlings by the female kākāpō (Cottam et al., 2006). However, other plant species have also 

been found to be selected for by breeding kākāpō. A study conducted on Hauturu-o-Toi by 

Stone (2013) found that female kakapo that attempted to breed on Hauturu preferred kauri 

(Agathis australis) dominated vegetation over any other vegetation type. Therefore, other plant 

species should be considered as potential food sources and nesting triggers for kākāpō when 

looking at the vegetation present at Maungatautari Ecological Island. 

 

The purpose of this research was to assess the habitat of Maungatautari Ecological Island as a 

potential site for the reintroduction of kākāpō. Given the low breeding frequency of the kākāpō 

and the relationship that exists between kākāpō nesting and food abundance, careful assessment 

of Maungatautari as a potential habitat must be carried out to determine: (i) if the vegetation 

present at this site is suitable to support kākāpō survival and breeding by quantifying the density 

of selected tree species known to induce nesting in relation to those as other sites where kākāpō 

breed; and (ii) to infer the suitability of Maungatautari as habitat by modelling habitat 

suitability based on key habitat features known to influence kākāpō distribution. 

  



Chapter 1  General Introduction

 

 4 

1.2 Habitat 

1.2.1 General overview of habitat 

 

Habitat is often most simply defined as the place where an organism lives (Davis, 1960; 

Krausman, 1999). However, the meaning in biological science goes further than this and a 

number of other habitat related concepts must be considered when discussing habitat in relation 

to wildlife management (Krausman, 1999). The word habitat, as used in English, stems from 

the Latin habitare, meaning to live or dwell. The term also has the same origins as the word 

habit, which is the past participle of the Latin verb habire, meaning to have or hold. From these 

origins the word incorporates both the concept of a space or place having suitable conditions 

where life can dwell, and the occurrence of particular living organisms in such places and 

spaces. (Davis, 1960; Wallace, 2007). Habitat has been more comprehensively defined by 

Krausman (1999) and Morrison, Marcot, & Mannan, (2006) as the resources (food, cover, 

water) and environmental conditions (temperature, precipitation, and presence or absence of 

predators and competitors) present in an area that produce occupancy by individuals of a given 

species (or population) and that allow those individuals to survive and reproduce. Therefore, 

wherever an organism is provided with such resources and conditions that permit survival and 

reproductive success, that area is considered a habitat. Thus, even migration and dispersal 

corridors and the land that animals occupy during breeding and nonbreeding seasons are 

considered components of habitat. A number of other habitat-related terms are important to 

consider when discussing the concept of habitat in its entirety including habitat use, habitat 

selection, habitat preference, habitat availability, habitat quality, and critical habitat. These 

terms are briefly summarised below. 

 

Habitat Use 

 

Habitat use is the way an organism uses the physical and biological resources in a habitat (Hall 

et al., 1997; Krausman, 1999). Habitat use is typically measured as the relative amount of time 

spent in different areas within a habitat. Therefore, more time spent in a given area means more 

“use” of resources or conditions at that location (Johnson, 1980).  

Habitat may be used for various behaviours such as foraging, cover, nesting, escape and 

denning and patterns of habitat use can often be observed and vary with these described 

behaviours. Habitat use is also subject to temporal variability as some animal behaviours and 



Chapter 1  General Introduction

 

 5 

activities require specific environmental components that may vary on a seasonal or yearly 

basis (Krausman, 1999).   

 

Habitat Selection 

 

Habitat selection is the hierarchical process, involving a series of both innate and learned 

behavioural decisions, by which animals choose which habitat components (conditions, 

resources) to use at different scales of the environment (Hutto 1985; Krausman, 1999). 

Decisions in habitat selection are believed to be driven by inherited behaviours and experiences 

typically associated with factors such as foraging, cover availability, food quality and quantity, 

and resting or nesting sites (Krausman, 1999). A number of external constraints can also have 

an influence on habitat selection for an individual such as competition and predation. 

Competition is involved in the intraspecific and interspecific relationships between individuals 

that partition the available resources within an environment. Consequently, competition may 

result in a species failing to select a habitat that is otherwise suitable in all other resources 

(Krausman, 1999). The existence of predators may also prevent an individual from occupying 

an otherwise suitable area. Survival of the species and its future reproductive success are the 

ultimate driving forces that influence an individual to select for or against a habitat. Therefore, 

a high occurrence of competition and predators may influence an individual to choose a 

different site with less optimal resources (Krausman, 1999).  

 

Habitat Preference 

 

Habitat preference is the result of habitat selection, describing the disproportional use of some 

resources over others when offered alternative choices on an equal basis (Johnson 1980; 

Krausman, 1999). Habitat preference can be quantified by the statistical comparison of samples 

of habitat use and availability. Therefore, preference is contingent on both habitat use and 

availability (Beyer et al., 2010). Habitat preferences are most noticeable when animals spend 

a high proportion of time in habitats that are not very abundant on the landscape (Krausman, 

1999).   



Chapter 1  General Introduction

 

 6 

Habitat Availability 

 

Habitat availability is the accessibility of physical and biological components of a habitat to an 

organism (Krausman, 1999). Available habitat is usually defined from the total study area; 

however, not all of the area may be available to an animal as it may be constrained by other 

factors (e.g. the presence of other animals) that restrict accessibility (Aebischer et al., 1993; 

Krausman, 1999). Availability differs from the abundance of resources, which refers only to 

their quantity in the habitat, irrespective of their accessibility to an organism. Measuring actual 

resource availability is important for understanding wildlife habitat, but in reality it is rarely 

measured because of the difficulty of determining what is and what is not available from an 

animal’s perspective. Consequently, quantification of habitat availability usually consists of a 

priori or a posteriori measure of the abundance of resources in an area used by an animal, 

rather than true availability (Krausman, 1999). 

 

Habitat Quality 

 

Habitat quality refers to the ability of the environment to provide conditions appropriate to 

promote individual fitness and population growth and persistence (Krausman, 1999; Johnson, 

2007). Habitat quality as a quantifiable measure is considered a continuous variable ranging 

from low to high  (e.g. depending on the level of resources available to sustain survival and 

reproduction) (Hall et al., 1997; Krausman, 1999). Habitat quality should be explicitly linked 

with demographic features of the species of interest rather than density or vegetative 

characteristics if it is to be used as a useful measure. While population density can be equated 

with some level of habitat quality, it can also be a misleading indicator of habitat quality. For 

instance, high density can cause animals to congregate in, or be forced into, areas of low habitat 

quality whilst a low population density may mean some areas of habitat remain unused or 

unoccupied even though they are of high quality. Additionally, vegetative characteristics are a 

poor indicator of habitat quality as a particular plant association may promote high fitness in 

one animal species but not another (Hall et al., 1997; Krausman, 1999). 

  



Chapter 1  General Introduction

 

 7 

Critical Habitat 

 

The term critical habitat is primarily used as a legal term to describe the physical or biological 

features essential to the conservation of a species and which may require special management 

consideration or protection (Hall, Krausman, & Morrison, 1997; Krausman, 1999). However, 

Hall, Krausman, & Morrison (1997) suggest that critical habitat should be specifically linked 

with the concept of high-quality habitat, which equates to an area's ability to provide resources 

for population persistence. This makes it an operational and ecological term rather than a 

political term (Hall, Krausman, & Morrison, 1997; Krausman, 1999). 

1.3 Current knowledge - Kākāpō 

1.3.1 Distribution and Status 

 

Prior to human settlement, subfossil records indicate that the kākāpō was widespread and 

abundant throughout New Zealand’s three main islands and inhabited a range of regions from 

sea level to alpine areas (Powlesland et al., 2006; Millener, 1981). By 1880 a significant decline 

in kākāpō populations was evident with populations being restricted to Stewart Island, and 

forested areas of the North and South Island. This was predominantly due to predation by 

humans and introduced mammals (Butler, 1989; Lloyd & Powlesland, 1994). Remaining 

kākāpō populations in the North and South Island suffered further rapid declines, and by the 

early 20th century the kākāpō was extinct in the North Island. By 1976 kākāpō seemed 

effectively extinct with only a small, male-dominated population known to remain in remote 

subalpine valleys of Fiordland (Powlesland et al., 1995; Lloyd & Powlesland, 1994). In 1977 

a breeding population of kākāpō was rediscovered in southern Stewart Island (Powlesland et 

al., 1995; Elliott et al., 2006). However, evidence of severe cat predation on adult birds became 

apparent which prompted the decision to transfer the remaining birds to cat and mustelid-free 

offshore islands to ensure their survival (Lloyd & Powlesland, 1994; Elliott et al., 2006; 

Powlesland et al., 2006). Thereafter between 1980 and 1997 all known kākāpō from Fiordland 

and Stewart Island were transferred to alternative offshore islands (Figure 1.1) and by 2002 

kākāpō inhabited five different islands; Hauturu-o-Toi, Maud, Whenua Hou, Anchor and Pearl 

(Powlesland et al. 2006). 



Chapter 1  General Introduction

 

 8 

 

Figure 1.1 Islands around New Zealand where kākāpō have been translocated to for 

conservation efforts. Map compiled in ArcMap GIS. 

 

Whilst annual adult survival on these offshore islands had been high, low productivity caused 

the population to become reduced to a nadir of 51 birds in 1995 (Elliott et al., 2006; Powlesland 

et al., 2006). A number of intensive management schemes have since been employed raising 

the population to 154 adult birds as of June 2017 (Table 1.1) (Elliott, 2017). The kākāpō is 

listed as nationally critical by the New Zealand Department of Conservation, the highest 

conservation ranking available (Hitchmough, 2002; Powlesland et al., 2006).  

  



Chapter 1  General Introduction

 

 9 

Table 1.1 The current total kākāpō population across New Zealand offshore Islands (modified 

from Kakapo Recovery, 2016).  

*Juveniles are classed as being individuals less than 5 years old.  

1.3.2 General biology 

 

Kākāpō are a flightless, nocturnal and secretive parrot endemic to New Zealand (Higgins, 

1999) and belong to the endemic New Zealand subfamily Strigopinae (Powlesland et al., 2006). 

They are most distinguishable by their characteristic mottled yellowish-green plumage and 

large bulky stature (Higgins, 1999). That characteristic and cryptically coloured plumage keeps 

the kākāpō well camouflaged as it blends perfectly with foliage so that even at close range 

detection is difficult (Powlesland et al., 2006). Adult birds have a distinctive owl-like facial 

disc; forward pointing eyes; hair-like feathers; a broad, pale grey beak; robust, fleshy legs and 

feet; short, rounded wings; and a relatively short, decurved tail (Higgins, 1999; Powlesland et 

al., 2006). The kākāpō is the heaviest parrot in the world and has pronounced sexual 

dimorphism in body size (Higgins, 1999; Eason et al., 2006; Powlesland et al., 2006). Males 

typically weigh 30 - 40% more than females with weights averaging around 2kg for adult males 

and 1.5kg for adult females. However, the weight of adult males can range from 1.6 to 4.0 kg 

whilst adult females range from 1.3 to in excess of 2.0 kg (Higgins, 1999; Eason et al., 2006; 

Powlesland et al., 2006).  

1.3.3 Breeding biology and behaviour 

 

The kākāpō is the only parrot species known to have a lek-breeding system and possibly the 

only avian lek species to have evolved in an environment lacking mammalian predators 

(Merton et al., 1984). In lek mating systems no permanent bonds between sexes are formed, 

and breeding is driven by the females’ pursuit of mates rather than that of the males. When in 

breeding condition, adult males will congregate in loose association in courtship areas forming 

 Adults (breeding age) Juveniles* Total 

Females 57 17 74 

Males 59 21 80 

Total 116 38 154 



Chapter 1  General Introduction

 

 10 

a ‘track and bowl’ system (Higgins, 1999). A track and bowl system is comprised of excavated 

shallow depressions or ‘bowls’ linked by tracks cleared by the birds (Merton et al., 1984; 

Higgins, 1999). Distances between track and bowl systems of neighbouring birds can vary 

between 15 and several hundred metres, with groups of up to 50 associated bowl systems 

extending over several square kilometres (Merton et al., 1984; Higgins, 1999). Male kākāpō 

use these bowls to perform a range of courting postures and calls, which include a characteristic 

low pitched, resonant booming call and a high pitched chinging call (Merton et al., 1984; 

Powlesland et al., 1992; Higgins, 1999). The male produces the booming call by progressively 

lowering its head and inflating its chest whilst simultaneously producing repeated booms. At 

maximum chest inflation the kākāpō begins to produce a loud rhythmic boom which can be 

heard at distances of up to 5 km in ideal conditions (Merton et al., 1984; Eason et al., 2006). 

The booming season commences in late November and lasts until March. The booming and 

chinging serenade can last for 6 to 8 h per night, starting 1 h after dark and stopping 1 h before 

first light (Merton et al., 1984; Higgins, 1999). Additional behaviours associated with booming 

include side-to-side rocking movements; walking backwards while slowly raising and lowering 

fully extended wings; vigorous wing-flapping and static postures between booming sequences 

(Merton et al., 1984). 

 

Males have no further parental role after mating, and thus females brood, forage and feed their 

young alone (Higgins, 1999). Females nest within their own home ranges and have been found 

to nest at different sites in different years (Powlesland et al., 1992). Nesting sites generally 

occur in natural cavities at ground level. Nests can be found in holes in banks or rotten trees, 

under thick vegetation, or in small caves and are typically located within several hundred 

metres of prime feeding areas (Higgins, 1999). Egg laying occurs from late January to mid-

March with two to five eggs laid per clutch however, broods are usually comprised of only 1 

or 2 chicks (Williams, 1956; Powlesland et al., 1992; Eason et al., 2006). Egg incubation 

begins immediately after the first egg is laid and the typical incubation period is 28 to 31 days 

(Eason et al., 2006). Young are born blind and totally helpless and are brooded by day until 

approximately 30 days old. During this brooding period, the female leaves her young for long 

periods at night in order to forage. From about 9 weeks after hatching the young may spend 

increasing periods away from the nest until they fledge at around 10 weeks (Powlesland et al., 

1992, Higgins, 1999). Fledglings typically remain close to the nest for a further month and may 

associate with, and solicit food from,  their mothers until 9 months old (Higgins, 1999). 



Chapter 1  General Introduction

 

 11 

1.3.4 Diet and foraging behaviour 

 

Kākāpō are exclusively herbivorous (Best, 1984; Higgins, 1999; Atkinson & Merton, 2006; 

Butler, 2006). Their diet is extremely diverse in both the number of species eaten and in the 

part of the plant eaten (Butler, 2006). For example, Gray (1977) found that approximately 80 

species of plants were eaten by kākāpō in Fiordland. Plant material that is eaten by the kākāpō 

can include leaves, twigs, bark, nectar, fruit, seeds, fern pinnae, rhizomes, fungi, ripe sporangia, 

tussock-grass tillers and roots of herbaceous plants (Higgins, 1999; Atkinson & Merton, 2006). 

Kākāpō have evolved as opportunistic feeders with highly variable feeding patterns. This 

flexible feeding pattern allows them to utilise a broad range of seasonal foods, many of which 

are only available for short periods or in intermittent years (masting species) (Best, 1984; 

Higgins 1999).  

 

Kākāpō have an unusual method of feeding. Their short, powerful beak and broad, thick tongue 

are well adapted for browsing, crushing and grinding to extract nutrient rich juices from fibrous 

plant tissues (Powlesland et al., 2006). The chewed fibrous material is compressed in the bill 

to form a tight wad of fibre which is expelled from the bill with the aid of the tongue as a pellet 

or ‘chew’. This formation of ‘chews’ is characteristic of kākāpō feeding habit (Higgins, 1999; 

Butler 2006; Powlesland et al., 2006). Kākāpō rely on this specialised feeding method as their 

gizzard lacks the muscular development suitable to digest more fibrous plant material (Higgins, 

1999; Butler 2006; Powlesland et al., 2006).  

 

Kākāpō are typically solitary foragers and mostly forage within their own home range (Higgins, 

1999). Whilst foraging occurs mainly on or near the ground, where species diversity is greatest 

(Butler, 2006), kākāpō are also skilled climbers and are able to reach canopy heights of up to 

30 metres (Powlesland et al., 2006). Kākāpō use their bill and powerful feet to climb and move 

from tree to tree through the canopy. During such arboreal foraging, periods of silence while 

the bird is feeding can be observed, interspersed with noisy movement of foliage and ample 

wing flapping as the bird changes position (Higgins, 1999; Butler, 2006; Powlesland et al., 

2006).  

 

Kākāpō are nocturnal feeders (Best, 1984; Higgins, 1999; Butler, 2006). However, 

occasionally females with dependent young will forage during daylight, at dawn or at dusk. 

Foraging activity is usually interspersed with long periods of inactivity (up to 60 minutes) 



Chapter 1  General Introduction

 

 12 

(Higgins, 1999; Powlesland et al., 2006). Whilst the kākāpō has forward-orientated eyes which 

provide some degree of binocular vision, their sight is considered poorly developed (Corfield 

et al., 2011). Instead, they use their keen sense of smell, which is well developed (Gsell, 2012), 

to locate their food. The kākāpō also employs a tactile method when locating their food. When 

foraging, birds will adopt a near-horizontal posture which brings the lower rictal bristles of 

their facial disk into contact with the ground. Such sensory perception is considered to be 

important not only when traversing unfamiliar terrain in the dark, but also when feeding at 

night on certain foods, such as Aciphylla spp. which have long, rigid, leaves with spiny tips 

(Higgins, 1999; Powlesland et al., 2006). 

 

Breeding has a significant influence on the kākāpō diet as there are differences in diet between 

breeding and non-breeding years (Wilson, 2004; Wilson et al., 2006). Breeding is known to 

coincide with the mast fruiting years of a number of tree species. Mast fruiting or seeding is 

defined as the synchronous production of a heavy seed crop in some years (mast years) by a 

population of plants or trees, while in other years there is no, poor or only moderate seed 

production (Norton & Kelly, 1988). In these mast fruiting years, kākāpō predominantly eat 

podocarp fruits and the incidence of other food sources in their diet significantly declines. 

Additionally, diets differ between female and male kākāpō in both breeding and non-breeding 

years. In breeding years, females are most likely to eat podocarp fruit as well as leaves of trees 

and shrub of Dracophyllum (D. longifolium, D. pearsonii, D. politum) (Wilson, 2004; Wilson 

et al., 2006). Blechnum fern fronds (B. novae-zealandiae, B. procerum and unidentified 

species) also appears to be prevalent in the diet of female kākāpō during breeding years 

(Wilson, 2004; Wilson et al., 2006) suggesting that understorey vegetation may be important 

during this time (Whitehead, 2007). In contrast, in non-breeding years the incidence of Hall’s 

totara (Podocarpus laetus) leaf in the diet of females increases. Males are more likely to eat 

podocarp fruit, fern, Lycopodium rhizomes and monocots (L. ramulosum, L. varium and 

unidentified species) during breeding years and Manuka (Leptospermum scoparium) fruit in 

non-breeding years (Wilson, 2004; Wilson et al., 2006). These difference in diets between 

males and females may be explained by differences in foraging behaviour between the birds, 

particularly in breeding years. During breeding years females gather food for their chicks while 

males are predominantly active on the ground in lek breeding areas (Powlesland et al., 1992; 

Wilson et al., 2006).  

 



Chapter 1  General Introduction

 

 13 

Podocarp fruits, particularly from rimu, are the preferred food in the diet of breeding females 

and for provisioning chicks (von Hurst et al., 2016). A study is currently investigating whether 

this preference is attributed to the high calcium and vitamin D content of rimu berries (von 

Hurst et al., 2016). Calcium is essential for both egg shell production and the growing skeleton 

of chicks. Vitamin D is also critical for these processes as well as to utilise dietary calcium and 

for the maintenance of calcium homeostasis (von Hurst et al., 2016). Ripe rimu berries also 

provide protein, fatty acids and a range of both digestible and non-digestible carbohydrates 

which are important for chick growth and adult maintenance (Cottam et al., 2006; von Hurst 

et al., 2016).   

1.3.5 Conservation management  

 

The need for conservation management efforts began in the late 1800s following the steady 

decline in kākāpō population numbers from the mid-1800s onwards as a result of the spread of 

introduced mammalian predators and human disturbance. Heavy predation of native bird 

species prompted the New Zealand Government to recognise the need to preserve native bird 

populations on offshore islands, leading to the purchase of Hauturu-o-Toi, Kapiti, and 

Resolution Islands as wildlife reserves (Cockrem, 2002). In 1894, the first translocation of 

kākāpō to a predator-free offshore island was undertaken with the translocation of more than 

300 birds from Fiordland to Resolution Island. Unfortunately, stoats were still able to swim to 

the island from the mainland. However, this endeavour paved the way for using translocation 

of threatened populations as an important conservation strategy in preserving native bird 

species in New Zealand (Cockrem, 2002). 

 

No further major conservation efforts were undertaken until 1949 when the newly established 

New Zealand Wildlife Service began searching for kākāpō, predominantly in Fiordland where 

the last remaining kākāpō were believed to be located (Lloyd, & Powlesland, 1994). These 

searches led to the discovery of an all-male population near Milford Sound.  In the 1960s The 

Wildlife Service established a captive breeding programme. However, most birds died within 

a few months and thus captivity was deemed an unviable method for kākāpō management and 

the programme was abandoned (Lloyd, & Powlesland, 1994; Cockrem, 2002). 

 



Chapter 1  General Introduction

 

 14 

In 1972, the Wildlife Service adopted a new conservation strategy which concentrated on 

establishing safe populations by translocating wild-caught kākāpō to offshore islands free of 

mammalian predators (Lloyd, & Powlesland, 1994). In 1977 a population of 100 to 200 birds 

was discovered on Stewart Island including the first female kākāpō to be seen in over a century. 

The discovery of this population prompted a revision of this conservation strategy to include 

maintenance of the Stewart Island population and it became the subject of intensive research 

and management during the 1970s and early 1980s (Lloyd, & Powlesland, 1994; Elliot et al., 

2001; Cockrem, 2002). A major programme of feral cat control begun in 1982 in an attempt to 

reduce the rate of predation on kākāpō.  However, high rates of predation were still evident, 

and it became clear that maintaining the kākāpō population on Stewart Island was 

impracticable due to the population's low productivity and the high cost of predator control. 

Consequentially, all surviving kākāpō were translocated to three relatively predator-free 

islands: Whenua Hou, Hauturu-o-Toi, and Maud Island during the 1980s and 1990s (Lloyd 

and Powlesland, 1994; Cockrem, 2002). The translocation of kākāpō to suitable predator-

free offshore islands successfully halted the decline in adult population numbers. However, 

productivity was found to be low, with only three kākāpō reared to independence from the time 

the birds were transferred to the islands until 1995 (Lloyd, & Powlesland, 1994; Elliot et al., 

2001; Cockrem, 2002). The failure of kākāpō to thrive led to a review of the kākāpō 

conservation programme in 1995. This in turn led to a new kākāpō recovery plan which 

included a more intensive and intrusive management of every individual bird to maximise the 

chances of producing fledged young from a fertile kākāpō egg (Elliot et al., 2001; Cockrem, 

2002). This new recovery plan saw significant improvements in kākāpō productivity, with the 

support of nesting females through supplementary feeding, intensive monitoring and 

subsequent intervention when necessary (Powlesland et al., 2006). Consequentially, these 

conservation efforts have seen population numbers increase from 51 birds in 1995 (Elliott et 

al., 2006; Powlesland et al., 2006) to 149 birds in March 2018. 

1.3.6 Kākāpō Habitat 

 

Historical habitat use 

 

Prior to human and introduced predator disturbance and their consequent decline, kākāpō were 

known to have occurred throughout the three main islands of New Zealand, from the far north 



Chapter 1  General Introduction

 

 15 

of the North Island to southern Stewart Island (Powlesland et al., 2006). Kākāpō were believed 

to be habitat generalists, with historic reports indicating they inhabited a range of vegetation 

types, altitudinal and climatic zones (Higgins, 1999; Powlesland et al., 2006). Kākāpō formerly 

occurred from near sea-level to the subalpine zone (> 1200 m) and in rolling to extremely steep 

landforms (Higgins, 1999; Butler, 2006; Powlesland et al., 2006). Kākāpō frequently occurred 

in temperate forests from lowland podocarp to upland beech. However, they were not an 

exclusively forest-dwelling species with accounts indicating they frequently associated with 

grassland habitats as well as other scrubland, herbfields, and tussock grasslands. Kākāpō 

tended not to penetrate far into tall forest but instead favoured ecotones where forest 

transitioned into grassland and where they had access to varied food resources (Powlesland et 

al., 2006). 

 

The kākāpō population in Fiordland mostly occupied the subalpine zone at the edge of beech 

forest, among scrub, and in tussock grasslands on steep slopes of valleys, glaciers or avalanche 

and alluvial fans (Higgins, 1999; Butler, 2006). Kākāpō on Stewart Island inhabited podocarp 

forest and subalpine forests and scrub on rolling hilly peatlands (Higgins, 1999; Powlesland et 

al., 2006). Kākāpō that were translocated to offshore islands from Fiordland and Stewart Island 

had to quickly adapt to unfamiliar habitat types and food resources, including exotic pastures 

(Higgins, 1999; Powlesland et al., 2006). 

 

Current utilised vegetation and topography 

 

The translocation of all kākāpō to predator-free offshore islands has meant they have been 

exposed to very different vegetation types and topography than what they used historically. 

Habitat studies have been conducted on Hauturu-o-Toi (Moorhouse, & Powlesland, 1991; 

Stone, 2013), Maud Island (Walsh et al., 2006) and Whenua Hou (Whitehead, 2007) and 

describe vegetation and topographical selection by kākāpō on these islands. On Hauturu-o-Toi 

kākāpō were found in most vegetation types. However, they appeared to have preferred high-

altitude vegetation on the island's cooler, wetter southern side. Kākāpō also showed seasonal 

use of low-altitude forest (Moorhouse, & Powlesland, 1991). On Maud Island, kākāpō showed 

considerable individual variation in the use of habitats and plant species. Kākāpō fed largely 

in the treeland community dominated by five-finger (Pseudopanax arboreus) in autumn 

months, whilst exotic pines (Pinus radiata) were used extensively in spring and summer. 

Pasture communities were mostly avoided by kākāpō with the exception of boundary areas 



Chapter 1  General Introduction

 

 16 

between pasture and other habitats (Walsh et al., 2006). On Whenua Hou, foraging locations 

were positively correlated with high-abundance rimu forest with a tall canopy (> 20 m) 

(Whitehead et al., 2012). Whitehead (2007) found that these areas with higher abundances of 

mature rimu trees were also optimal breeding habitat. Historically, kākāpō inhabited a range of 

altitudes. However, kākāpō translocated to islands have been found to have a more limited 

topographical range. Males translocated to islands have tended to establish home ranges on the 

upper slopes, high plateaus and summit regions in the warmer months, whilst females have 

generally settled at slightly lower altitudes on the mid slopes (Powlesland et al., 2006). 

 

Climate 

 

Kākāpō have historically endured a range of climatic zones with variations in rainfall and 

temperature. Historical reports indicate that kākāpō have occurred in areas of high rainfall in 

the Milford region (> 6000 mm per annum) and in areas of low to moderate rainfall in parts of 

Otago, Canterbury, and Marlborough (< 800 mm per annum) (Powlesland et al., 2006). In 

Fiordland, some kākāpō lived year-round in subalpine habitat enduring severe winter frosts, 

snow and ice for up to four to six months each year rather than descending to snow-free valley 

floors (Higgins, 1999; Powlesland et al., 2006). Kākāpō were also able to withstand high 

summer temperatures (upwards of 30°C) and dry conditions in parts of Otago, Marlborough, 

Nelson and the northern North Island. Kākāpō that have been translocated to offshore islands 

have been found to mostly inhabit relatively cool, moist and shaded slopes (Powlesland et al., 

2006). 

 

Roost sites 

 

Kākāpō are nocturnal feeders and roost during the day (Butler, 2006). The birds roost within 

their own home range, typically in natural cavities such as small caves, hollow tree stumps or 

logs, under dry rock overhangs, at the base of trees, or under low hanging branches or ferns 

(Higgins 1999; Butler, 2006). Individuals often display a preference for roosting either above 

or on the ground and some sites may be used repeatedly or irregularly over many weeks or 

even years. Favoured roost sites are typically dark, dry, sheltered from strong winds, and large 

enough to allow the bird to turn (Higgins 1999; Powlesland et al., 2006). Studies in Fiordland 

and Stewart Island suggest that kākāpō would wander considerable distances using many 

different roost sites, each for brief periods of time (Butler, 2006). 



Chapter 1  General Introduction

 

 17 

 

Home range 

 

An animal’s home range is described as the area traversed by an individual in its normal 

activities of foraging, mating and parental care that contains the resources required by the 

individual for survival (Farrimond et al., 2006). Both male and female kākāpō are solitary and 

generally remain within their individual home range for most of the year however, some 

overlap of home ranges occurs (Powlesland et al., 2006). The sizes of kākāpō home ranges 

have been studied on a number of islands they have inhabited, including Hauturu-o-Toi, 

Whenua Hou, Stewart, Maud, and Pearl Islands. These studies have shown a variation in home 

range sizes among individuals with variations from 15 - 50 ha on Stewart Island (Best & 

Powlesland, 1985); 21 - 38 ha on Hauturu-o-Toi (Moorhouse & Powlesland, 1991); 2 – 145 ha 

on Maud Island (Walsh et al., 2006); 0.8 – 11.4 ha on Pearl Island (Trinder, 1998); and a mean 

home range of approximately 15 ha on Whenua Hou (Farrimond et al., 2006). On Hauturu-o-

Toi and Maud Island, kākāpō home ranges have been seen to vary seasonally in location and 

size, with smaller home ranges in winter recorded on Maud Island (Moorhouse & Powlesland, 

1991; Walsh et al., 2006). 

 

These observations of home range size have been determined for remnant kākāpō populations 

which most likely exist at lower population densities than the species naturally would have in 

the same habitat types prior to the introduction of mammalian predators (Powlesland et al., 

2006). 

 

Movements 

 

Both male and female kākāpō typically use the same home range area for most of the year. 

However, individuals of either sex occasionally move up to several kilometres from their core 

home range to sites where they can remain for several days, weeks or even months (Higgins, 

1999; Powlesland et al., 2006). These long distance movements are often related to the 

availability of abundant food, breeding activity or climatic factors. Adult males and females 

commonly move beyond their core home ranges during late December to early February to 

visit leks. These visits can entail movements of a few hundred metres to several kilometres 

(Higgins, 1999; Powlesland et al., 2006; Joyce, 2009). 

  



Chapter 1  General Introduction

 

 18 

1.4 Study Site - Maungatautari Ecological Island 

1.4.1 General background 

 

Maungatautari Ecological Island is an isolated andesitic volcanic cone located 18 km south of 

Cambridge in the Waikato Region of the North Island of New Zealand (38°03′08″S, 

175°33′58″E) (McQueen, 2004; Doerr et al., 2017). The mountain is a dominant landform in 

the region encompassing a large area (3363 ha) of largely intact dense, mature podocarp-

broadleaved forest (Smuts-Kennedy & Parker, 2013; Doerr et al., 2017). Maungatautari is 797 

m in height at its highest point and has three peaks – Te Akatarere (727 m), Pukeatua (753 m) 

and Maungatautari (797 m) (Waipa District Council, 2005). 

Figure 1.2 Map of Maungatautari Ecological Island showing its location within the Waikato 

region, New Zealand. Map compiled in ArcMap GIS. 

 

The topography ranges from strongly rolling slopes at the base of the mountain to steep and 

very steep slopes near the peaks and in the gullies (Smuts-Kennedy & Parker, 2013). The lower 

margins of the forested mountain are primarily dominated by scattered large rimu and northern 

rata (Metrosideros robusta) over a canopy of abundant tawa (Beilschmiedia tawa), mangeao 



Chapter 1  General Introduction

 

 19 

(Litsea calicaris), hinau (Elaeocarpus dentatus), miro (Prumnopitys ferruginea), rewarewa 

(Knightia excelsa) and pukatea (Laurelia novae-zelandiae) (Burns and Smale, 2002; Innes et 

al., 2011). The upper margins of the mountain (> 600 m a.s.l.) are dominated by tawari (Ixerba 

brexioides) and quintinia (Quintinia serrata) (Baber et al., 2008). Large-scale logging on parts 

of the lower slopes of the mountain during early European settlement times has seen the 

removal of some mature podocarps, particularly rimu and to a lesser extent tawa. Totara 

(Podocarpus totara) was also logged for farm fencing, and northern rata  and other species 

were taken for firewood (Ewen et al., 2011; Smuts-Kennedy & Parker, 2013; Doerr et al., 

2017). The landscape surrounding Maungatautari is dominated by pasture land used 

predominantly for dairy production. This farmland environment offers little to no habitat for 

native animal species, essentially making Maungatautari a habitat island (Ewen, 2011; Doerr 

et al., 2017).  

 

Average rainfall on the mountain is between 1,400 and 1,600 mm per annum, compared to 

1,100 to 1,200 mm per annum on the surrounding flat pastures. All of the streams originating 

on Maungatautari flow into the Waikato River system and these streams have high water 

quality where they leave the forest (Waipa District Council, 2005; Smuts-Kennedy and Parker, 

2013). The Waikato Basin receives about 2,000 h of sunshine a year and 30 to 50 days of fog. 

The average temperature is around 14°C (de Lisle, 1967; McQueen, 2004). 

 

Prior to human occupation, Maungatautari was a high point in a large area of conifer-broadleaf 

forest rising above the wetland areas of the Waikato Basin (Leathwick et al. 1995; McQueen, 

2004). Tall, mature forest covered most of the region, except for extensive areas of bogs and 

deep swamps (Nicholls 2002; McQueen, 2004). The arrival of Maori to the region in 1500 AD 

coincided with extensive fires which saw the destruction of large expanses of tall forest in the 

area. These fire ravaged areas were in time replaced by scrub and fern lands. However, 

Maungatautari’s current vegetation state suggests that the mountain largely escaped burning 

(Clayton-Greene 1976; McQueen, 2004). By 1873, the landscape was in the process of being 

modified into pasture lands for agricultural use, and now Maungatautari stands as an isolated, 

forested mountain among surrounding pasture lands and represents nearly half of the remaining 

forest in the region (McQueen, 2004).   

 

In the 1970s, the forest of Maungatautari was assigned a national habitat ranking of high to 

outstanding by the Wildlife Service of the Department of Internal Affairs based on the recorded 



Chapter 1  General Introduction

 

 20 

presence of North Island kokako (Callaeas wilsoni), long-tailed bat (Chalinolobus 

tuberculatus), possibly short-tailed bat (Mystacina tuberculata) and a wide variety of more 

common forest birds. However, anecdotal evidence suggests that Maungatautari was even 

more abundant in indigenous fauna in pre-European times, with species including saddleback 

(Philesturnus rufusater), North Island robin (Petroica longipes), hihi (Notiomystis cincta), kiwi 

(Apteryx spp.), kaka (Nestor meridionalis) and kakariki (Cyanoramphus spp) (MacGibbon, 

2001; McQueen, 2004). It is believed that kokako remained on Maungatautari until the early 

1980s. However, saddleback, North Island robin, hihi and other sensitive bird species 

disappeared from the forest in the 1800s or the early decades of the 1900s. Kiwi, kaka and 

kakariki are also believed to have disappeared from the forest by the mid 1900 (MacGibbon, 

2001; McQueen, 2004).  

 

Introduced mammals at Maungatautari were noted from the mid-1900s as a consequence of 

European settlement in the area. Anecdotal information indicates the presence of possums 

(Trichosurus vulpecula), goats (Capra hircus), red deer and fellow deer (Cervus elaphus, 

Dama dama,), pigs (Sus scrofa), stoats (Mustela erminea), ferrets (Mustela putorius furo), 

weasels (M. nivalis) and rodents (ship rat Rattus rattus, Norway rat R. norvegicus and mice 

Mus musculus). Vegetation damage from possums became apparent in the mid to late 1970s 

but was not considered a major problem until the 1980s and 1990 when possum densities 

substantially increased and caused the significant decline of many plant species (MacGibbon, 

2001; McQueen, 2004). The presence of goats and pigs during the 1940s and 1950s has caused 

damage in both the upper areas (from goats) and lower areas (from pigs) which is still evident 

today (MacGibbon, 2001; McQueen, 2004).  

1.4.2 Conservation management on Maungatautari 

 

As early as 1912, the environmental value of Maungatautari Mountain was observed and it was 

proposed that it be set aside as a reserve for climatic and water conservation purposes. In 1927, 

Matamata, Waipa and Waikato County Councils together with the Cambridge Borough 

Council and the Leamington Town Board purchased 1558 ha of land on the mountain and it 

was gazetted as a scenic reserve. The Matamata and Waipa County Councils jointly managed 

the reserve until they were replaced by District Councils in 1989 from which time the Waipa 

District Council took over management responsibilities. Since 1927, the various managing 



Chapter 1  General Introduction

 

 21 

bodies have continuously added further pieces of land to the reserve so that the protected area 

now covers approximately 3363 ha of private land, Maori land and Department of Conservation 

estate and council land. (McQueen, 2004).  

 

In the late 1990s poisoning programmes were introduced in order to reduce the growing 

population of possums and the damage they were imposing on the vegetation. This action 

significantly reduced possum abundance (MacGibbon, 2001; McQueen, 2004). Furthermore, 

in 2006 a 47 km Xcluder™ pest-proof fence was installed around the entire base of the 

mountain, and all pest mammals inside were targeted for eradication, making it a mainland 

ecological island (Innes et al. 2011). Mainland ‘islands’ are a conservation management 

concept which refers to defined areas that are isolated by fencing, geographical features or, the 

intensive management of pests from other areas not managed intensively for conservation 

purposes. (Saunders & Norton, 2001). The installation of this specialised fence has allowed for 

the eradication of all introduced mammalian species, except for mice (Mus musculus) across 

the main mountain (Ewen et al., 2011; Richardson & Ewen, 2016; Doerr et al., 2017). Within 

this fenced reserve there are two smaller pest-proof enclosures at the northern and southern 

ends. The northern enclosure comprises 35 ha while the southern enclosure comprises 65 ha, 

and both would act as ecological islands in their own right if the main perimeter fence failed 

(McQueen, 2004). Both of these smaller enclosures are actively managed by the Manugatautari 

Ecological Island Trust (MEIT) with the aim of keeping them free from all mammalian pest 

species, including mice. 

 

The installation of the Xcluder™ fence has paved the way for the reintroduction of a number 

of endemic species to Maungatautari’s forest. Between 2005 and 2012, seven locally extinct 

native bird species were reintroduced to Maungatautari (Table 1.2), and an eighth species, the 

New Zealand falcon, appears to have self-reintroduced as a breeding species (Smuts- Kennedy 

& Parker, 2013). Most recently in 2016, the North Island kokako (Callaeas wilsoni) was 

reintroduced to Maungatautari with 40 birds being released. There are now 21 native forest 

bird species present compared to the 12 species that were noted prior to the commencement of 

the restoration project. This number is expected to eventually exceed 30 species, with many 

being listed as endangered or vulnerable on the IUCN Red List (Smuts- Kennedy & Parker, 

2013; Richardson & Ewen, 2016). These bird species will be part of a functioning ecosystem 

that is likely to include at least 50 indigenous vertebrate species (birds, bats, lizards, tuatara, 

frogs and fish) (Smuts- Kennedy & Parker, 2013).  



Chapter 1  General Introduction

 

 22 

Table 1.2 Bird reintroductions to Maungatautari from 2005-2012 (Smuts-Kennedy & Parker, 

2013) 

† Takahe have since been reclassified as NV (Robertson et al., 2016).  



Chapter 1  General Introduction

 

 23 

1.5 Research objectives 

 

This research will address the third key goal outlined in the current Kākāpō Recovery Plan 

(2006 - 2016) which states the need for research and management action to “secure, restore or 

maintain sufficient habitat to accommodate the expected increase in the kākāpō population” 

(Neill, 2008). The recent breeding success of kākāpō makes this need to select a suitable large 

island an even more imperative goal to achieve. This research specifically aims to determine if 

Maungatautari Ecological Island is a viable reintroduction site for kākāpō by investigating if 

the habitat provides “sufficient availability of natural food for kākāpō which could also 

facilitate breeding” (Cresswell, 1996). Additionally, this research aims to determine the spatial 

distribution of kākāpō habitat on Maungatautari to potentially guide reintroduction strategies. 

 

Objectives of this study are to: 

 

1. Estimate the density of five selected tree species, in particular rimu, on Maungatautari that 

are believed to be important in kākāpō breeding and as food resources.  

 

2. Investigate the approximate spatial distribution of tree species important for kākāpō survival 

and breeding across Maungatautari. 

 

3. Investigate the particular vegetation and topographical characteristics that are important in 

determining the potential distribution of kākāpō on Maungatautari Ecological Island. 

 

4. Produce a predictive habitat suitability map of Maungatautari Ecological Island showing 

areas suitable for kākāpō survival and breeding. 

  



Chapter 1  General Introduction

 

 24 

1.6 Thesis structure 

 

This thesis is divided into four chapters as follows: 

 

Chapter 1. General Introduction 

This chapter outlines the stimulus for this research and also provides a brief literature review 

of habitat as a general overview, a description of kākāpō and its habitat, and a description of 

Maungatautari Ecological Island. This chapter also includes the research objectives and 

provides an outline of the thesis structure. 

 

Chapter 2. Estimating the density of key tree species on Maungatautari 

This chapter focuses on estimating the density of five selected tree species important for kākāpō 

breeding on Maungatautari through the analysis of the distance line transect data.  

 

Chapter 3. Predictive habitat suitability modelling of Maungatautari as a reintroduction site  

This chapter focuses on assessing the spatial distribution of key tree species across 

Maungatautari from the distance line transect data, and aggregating particular vegetation and 

topographical characteristics of Maungatautari to produce a predictive habitat suitability map 

and infer the overall suitability of Maungatautari and predict the potential distribution of 

kākāpō. 

 

Chapter 4. General discussion  

This chapter provides a final discussion on the habitat characteristics of Maungatautari 

Ecological Island and its suitability as a reintroduction site for kākāpō. This chapter also 

provides some concluding remarks as well as future research and management 

recommendations. 

 

Each chapter that presents research results (Chapters 2 and 3) has been arranged as a separate 

paper, with a methods, results and discussion section. Both Chapters 2 and 3 refer to data 

collected in the distance line transect sampling survey. In order to avoid repetition, a complete 

description of the methods used for the distance line transect data collection is outlined in 

Chapter 2. Therefore, refer back to Chapter 2 for the data collection methods in Chapter 3. 



25 

Chapter 2  Estimating the density of key tree 

species on Maungatautari 

Maungatautari Ecological Island shown in the background. Photo credit: Sarah MacDonald 



Chapter 2  Estimating density of key tree species 

 

 26 

2.1 Introduction 

 

Understanding the environmental factors that influence habitat use and selection by organisms 

is fundamental to ecology and conservation biology (Manel et al., 2001) and is particularly 

important when making management decisions for rare and endangered species (Balbontín, 

2005). For example, knowledge of environmental factors and habitat features that favour the 

survival and reproduction of a particular species can be used to guide conservation decisions 

and species recovery planning (Manel et al., 2001; Balbontín, 2005; Mayor et al., 2009). 

Habitat quality is believed to be the main reason for the success or failure of reintroduction 

projects. Therefore, extensive evaluation of the habitat quality of candidate locations should be 

a fundamental requirement prior to any species reintroduction (Osborne & Seddon, 2012).  

 

Basic habitat selection theory predicts that individuals will select habitat types with high 

quality resources rather than occurring randomly across a given landscape or in habitats with 

low quality resources (Alcock, 1989). As mentioned in Chapter 1, a number of environmental 

factors can have an influence on habitat selection, one of which is food availability which has 

long been considered one of the fundamental factors underlying the distribution and abundance 

of bird populations and furthermore in driving habitat use and selection by birds (Lack, 1954; 

Strong & Sherry, 2000; Douglas et al., 2004; Champlin et al., 2009). Selecting a habitat with 

limited food resources can have significant consequences on the persistence and reproductive 

success of bird populations. For example, limited food abundance may result in delayed 

nesting, fewer nesting attempts and reduced nest provisioning and parental care (Champlin et 

al., 2009).  

 

Quantifying abundance of food resources in a given area in a way that accurately reflects food 

availability from an animal’s perspective can be difficult to achieve. However, fruit and plant 

food resource abundance has been found to be relatively easy to estimate as they are often more 

conspicuously displayed compared to other common bird food resources such as insects (Kwit 

et al., 2004). Plant food resource numbers can be estimated in a given area by a measure of 

their density in that specified area. Density information is among the most fundamental and 

sought after data in ecology and conservation biology (Williams et al., 2002; Nomani et al., 

2012). More specifically, plant species density data is particularly useful for effective forest 

assessment (Williams et al., 2002; Kissa & Sheil, 2012) and for the conservation of endangered 



Chapter 2 Estimating density of key tree species 

27 

species whose survival and reproductive success may be dependent on those plant species. 

Total counts of plant individuals in a forest environment are typically expensive to apply with 

satisfactory results and so researchers and resource managers often employ sampling 

methodologies to obtain reasonable estimates of density and population size whilst being more 

cost effective (Kissa & Sheil, 2012; Nomani et al., 2012). Distance sampling is a popular 

method for estimating density of organisms whilst being cost effective and minimising field 

effort (Buckland et al., 2001; Williams et al., 2002; Nomani et al., 2012).  

Food availability is a particularly important factor in driving habitat selection by kākāpō 

(Strigops habroptilus) in New Zealand. Previous studies have found that the only plants that 

produce fruit crops that are known to induce kākāpō nesting are podocarp trees, particularly 

rimu (Dacrydium cupressinum) (Elliott et al., 2006; Harper et al., 2006). However, this is not 

the case on Hauturu-o-Toi/Little Barrier Island which has very few rimu trees (Stone, 2013). 

The kākāpō on Hauturu-o-Toi have different breeding triggers which are believed to be driven 

by abundant beech seeding (Harper et al., 2006). On some islands adult female kākāpō will 

only breed when there is an exceptionally abundant supply of rimu fruit available, and in poor 

rimu fruiting years no females will attempt to breed (Elliott et al., 2006). On Whenua 

Hou/Codfish Island the foraging locations of breeding, female kākāpō were usually areas 

containing a high abundance of mature rimu, and such areas were described as optimal breeding 

habitat for kākāpō (Whitehead et al., 2012).  

One of the key goals for the management of the kākāpō outlined in the current Kākāpō 

Recovery Plan is to secure, restore or maintain sufficient habitat to accommodate the expected 

increase in the kākāpō population (Neill, 2008). With a record breeding season in 2016 and 

expectations for kākāpō breeding to continue successfully, kākāpō populations on Whenua Hou 

and Anchor Island are considered to be nearing capacity (Stone et al., 2017). Finding new 

habitat in the near future is therefore an important task. Given Maungatautari is one of the only 

large (>300 ha) fenced sanctuaries, and it is already cleared of introduced predators, it is 

beneficial for the future management of kākāpō to determine if the vegetation present at this 

site is suitable to support survival and breeding of this species. The aim of this study was to 

investigate the availability of potential food sources for kākāpō on Maungatautari by estimating 

the density of five selected tree species known to induce or have the potential to induce nesting 

in relation to other sites where breeding occurs. 



Chapter 2  Estimating density of key tree species 

 

 28 

2.2 Methods 

 

There are a number of sampling methods available to estimate animal and plant species 

densities and sizes and each has its own advantages and limitations. This study uses visual 

detection based distance line-transect sampling to make density estimations of the tree species 

of interest to this study.  

2.2.1 An overview of distance sampling  

 

Distance sampling is a widely used technique for estimating population density (Thomas et al., 

2010). Distance sampling involves a set of methods in which the distance from a line or point 

to detections are recorded and further used to estimate density and/or abundance of the study 

object (Thomas et al., 2010). Objects sampled are usually animals or animal groups (termed 

clusters), but can also be plants or inanimate objects (Thomas et al., 2010). Detections are 

usually of the animal, plant or object itself, but may be of cues (such as bird song bursts) or a 

sign (such as dung or nests) (Thomas et al., 2010). Distance sampling has numerous advantages 

for estimating the absolute density of biological populations which include (Buckland et al., 

2001): 

- The ability to estimate the absolute density for a population, even when not every 

individual is detected per unit area.  

- The same estimation of density for a population can be calculated from data collected 

by two different observers, even if one of these observers fails to detect a lot of objects 

away from the line or point. This is because the difference in observations can be 

accounted for in the model detection curve.  

- Only a relatively small percentage of individuals needs to be detected within the sample 

area (possibly as few as 10-30%).  

- The size of the sample area can be unknown. 

 

The most widely used form of distance sampling is line-transect sampling (Thomas et al., 

2010). In line-transect sampling, a survey region is sampled by placing lines at random or, if 

the terrain allows, systematically placing a series of equally spaced parallel lines on the survey 

site. An observer walks along each line, recording any objects detected within a distance (w) 

of the line (Buckland et al., 2010; Thomas et al., 2010). The perpendicular distance from the 

line to the object is recorded. In some instances, the distance of detected objects from the 



Chapter 2  Estimating density of key tree species 

 

 29 

observer (so-called radial or object-to-observer distance), together with the angle from the line 

of the detection, are recorded, from which the perpendicular distance from the line can be later 

calculated using basic trigonometry (Buckland et al., 2010). These perpendicular distances are 

used to estimate a detection function, which is the probability that an object is detected, as a 

function of distance from the line (Buckland et al., 2010). For the standard method, it is 

assumed that this probability is one at 0 distance from the line; i.e., that objects on the line are 

detected with absolute certainty, and that detection probability then decreases with increasing 

distance from the line (Buckland et al., 2010; Thomas et al., 2010). From the distribution of 

distances observed, a detection function can be fitted and this function used to estimate the 

proportion of objects detected within a strip extending a distance w from the line on either side. 

Assuming an adequate number of lines are placed through the survey region, this density 

estimate is representative of the whole survey region, allowing abundance within that region 

to also be estimated (Buckland et al., 2010). 

 

Distance line-transect sampling is already popular in estimating animal species densities due 

to its efficiency and practicability (Buckland et al., 2000). However, it has been less readily 

used for the assessment of vegetation such as forest trees (Kissa, & Sheil, 2012). Good density 

estimates, particularly for low abundance tree species, are costly to achieve, especially in 

rugged forest landscapes. This has prompted investigation into more cost effective means to 

assess forest composition and make density estimations. For this reason, distance line-transect 

sampling is being employed more frequently in a number of vegetation studies (Kissa, & Sheil, 

2012; Hessenmoller et al., 2013; Mirzaei, & Bonyad, 2016) with the aim of minimising the 

amount of field measurements and to produce accurate estimations of tree species composition 

and density. (Kissa, & Sheil, 2012).  

 

For distance to produce reliable estimates of density, it is essential that the assumptions of this 

sampling method are met (Buckland et al., 2010). If these assumptions are not met, estimates 

of density can have substantial bias (Buckland et al., 2010). The fundamental assumptions on 

which distance sampling is based are described below. 

  



Chapter 2 Estimating density of key tree species 

30 

Assumptions 

1.) Objects on the line or point are detected with certainty. 

In practice, objects at zero distance from the line i.e., objects right above the observer, on or 

close to the line or point should be detected with near certainty. If the observer fails to detect 

objects on or close to the line or point this causes underestimation of density (Buckland et al., 

2001; Thomas et al., 2010). In this study trees are unlikely to be missed on the transect line. 

2.) Objects are detected at their initial location before any response to the observer. 

Theoretically, distance sampling is a ‘snapshot’ method as the objective of a point or line 

transect is to record the number of objects present at a single moment in time, and the position 

of these objects in relation to a random point or line (Buckland et al., 2001; Thomas et al., 

2010). In practice, non-responsive movement in line transect surveys is not problematic 

provided it is slow relative to the speed of the observer. However, responsive movement before 

detection is problematic because subjects are assumed to be located independently of the 

position of the line or point (Thomas et al., 2010). This assumption can be disregarded when 

sampling plants, inanimate objects or physical signs (e.g. dung or nests), immobile animals 

(e.g. barnacles) or dead animals (e.g. after disease outbreak) given their stationary nature. 

Therefore, in this study, where the observed objects are trees, this assumption can be 

disregarded.  

3.) Measurements are exact. 

In practice, it is essential that the measurement of distances from the line to the centre of each 

detected object is accurate for the data to be effective in making estimations of density 

(Buckland et al., 2001; Buckland et al., 2010). Wherever possible, training and technology 

(e.g. laser rangefinders) should be used to ensure accuracy. It is also assumed that species are 

not misidentified (Thomas et al., 2010). In this study, training for correct sampling protocol, a 

laser rangefinder with a clinometer feature, and a tape measure were used to ensure accurate 

measurements were taken. 

4.) There is an adequate sample of randomly distributed lines, or an adequate grid of lines, in 

the survey region. 

An adequate sample of lines or points at randomised locations ensures that object locations are 

independent of the positions of the lines or points. This assumption becomes critical if transects 



Chapter 2 Estimating density of key tree species 

31 

are placed along roads or tracks (Buckland et al., 2010; Thomas et al., 2010). In this study a 

total of 21 transect lines were used which were well distributed over the entire area of 

Maungatautari. 

2.2.2 Line transect sampling survey design in this study 

To study the vegetation on Maungatautari, line transects, which covered areas that reflected 

the true forest composition, were first established (Hardy & Sonké, 2004). Large study areas 

with variable topography, such as Maungatautari, generally contain extensive diversity in 

forest conditions and therefore, a large number of sampling lines are often required in order to 

accurately quantify the structural parameters of interest (Sandmann & Lertzman, 2003). 

However, financial and logistical constraints often dictate the spatial location of sampling sites 

and the intensity of sampling and thus some trade-offs are often required (Sandmann & 

Lertzman, 2003). As previously described, the topography of Maungatautari ranges from 

strongly rolling slopes to very steep slopes near the peaks and in the gullies and the vegetation 

cover is dense across the entire mountain. These conditions make establishing new randomly 

distributed line transects difficult as creating new transect tracks would require cutting through 

vegetation and traversing across difficult terrain. Therefore, pre-existing monitoring lines on 

Maungatautari were used. These monitoring lines are frequently used for pest-monitoring and 

bird surveying and their distribution across the mountain has been purposely determined to 

cover the range of variability in the ecological and structural components of the forest and 

topography. Therefore, these pre-existing monitoring tracks are believed to be appropriate for 

the requirements of this study.  

There is a strong correlation between landforms and vegetation composition patterns (Parker, 

& Bendix, 1996). Therefore, to ensure that the monitoring lines were located in areas that 

accurately reflected the variation in vegetation composition on Maungatautari, the landform 

types were differentiated using a digital terrain model. This digital terrain model was used to 

differentiate between three different landform types: ridges, faces, and gullies. The total 

number of monitoring lines were then randomly distributed across the mountain, occupying 

landform types in the same proportion that they occur in on Maungatautari.  



Chapter 2  Estimating density of key tree species 

 

 32 

A total of 21 monitoring lines were selected (Figure 2.1). The lengths of these transect lines 

vary from 1.32 km to 4.46 km, and each transect line was walked once between August and 

October 2016. Given the sampling objects in this study are immobile, the transect lines were 

only walked one time as observations are not expected to vary significantly between occasions. 

Figure 2.1. Locations of monitoring lines used for the distance line-transect survey at 

Maungatautari Ecological Island.  

 

Along each transect, the specified trees of interest to this study were searched for by two 

observers at the same time walking along the transect line together. The same observers were 

used for each transect line and were taught correct sampling protocols prior to sampling to 

ensure consistency between observers. An a priori for the minimum diameter at breast height 

(DBH) of sampled trees was set at 30 cm. The perpendicular distance from the transect to each 

tree was recorded using a tape measure for short distances and a laser range finder when 

distances were greater than 10 m so could not be measured easily using a tape measure. Tree 

diameter measurements of each sighted tree was also taken at breast height from the upper side 

of the slope using a tape measure. The height of each sighted tree was recorded and measured 

using the clinometer feature on the laser rangefinder. A maximum distance a priori was set at 

100 m and so only trees observed within 100 m of the transect line were recorded.  



Chapter 2 Estimating density of key tree species 

33 

Five particular tree species were of interest because they were known to be reasonably 

abundant on Maungatautari and have the potential to induce nesting in kākāpō. These include 

the following podocarp species: rimu, matai (Prumnopitys taxifolia), miro (Prumnopitys 

ferruginea), and Hall’s totara (Podocarpus laetus), as well as the southern beech species, silver 

beech (Lophozonia menziesii). The data to be analysed were categorised into two groups, rimu 

only and all key tree species combined; which includes all tree species identified as being 

important to kākāpō combined (matai, miro, Hall’s totara, silver beech). Analysing rimu as a 

separate category was done because this species plays a critical role in stimulating kākāpō 

nesting in southern New Zealand and it is the only breeding trigger for which there is strong 

contemporary evidence (A. Digby, pers. comm.). Therefore, it is useful to identify its density 

across the mountain exclusively.  

2.2.3 Data analysis 

The data collected from the line transect surveys were analysed using DISTANCE 6.2 Release 

1 software to estimate densities of the tree species of interest to this study. DISTANCE uses 

the equation  

to estimate density, where n = the total number of trees recorded, W and L = the total width and 

length of the transect, and Pa = the estimated probability of observing an object within a defined 

area (Buckland et al., 2000; Thomas et al., 2002: Kissa, & Sheil, 2012). Estimating density in 

DISTANCE involves selecting a model that suitably represents the influence of distance on 

detection probability (Kissa, & Sheil, 2012). The standard guidelines were followed for model 

selection when using DISTANCE software (Buckland et al., 2000). Four detection functions 

(Half-normal, Hazard-rate, Uniform and Negative-exponential), each with cosine, simple 

polynomial and hermite polynomial adjustments, were considered (Kissa, & Sheil, 2012). The 

data were truncated to only include trees observed within 50 m of the transect line to reduce 

the influence of outlier observations. Truncation also prevents extra adjustment terms from 

needing to be fitted for data a long way from the transect line. Such data contribute little to the 

abundance estimate and can reduce estimate precision (Buckland et al. 2001). At distances 

beyond 50 m correct identification of the tree species became increasingly more difficult, 

WLP

n
D

a 2ˆ
ˆ =



Chapter 2  Estimating density of key tree species 

 

 34 

particularly in distinguishing between similar looking species (such as matai and miro). 

Therefore, removing observations beyond this distance also ensured greater accuracy in species 

identification.  

 

Observed trees with a diameter at breast height (DBH) of less than 30 cm were also removed 

from the data to be analysed. Growth rates of New Zealand podocarp species have previously 

been studied (Smale & Kimberley, 1986) and stem sizes of less than DBH 30 cm have been 

classified as saplings and poles whilst stem sizes of greater than or equal to DBH 30 cm are 

classified as being adult trees. Vegetation surveys conducted at Maungatautari between 1947 

and 1955 to produce a vegetation map in 1963 (Figure 3.1) also classified tree stems as having 

a minimum DBH of 30 cm (McKelvey, 1963). Therefore it seems appropriate to use the same 

measure of stem size in this study. Only mature adult podocarp trees have the capability to be 

fruit bearing (Nanami et al., 2005). Podocarp fruit, which as previously mentioned in chapter 

1, directly influences kākāpō life processes by inducing nesting and providing suitable food 

resources (Wilson, 2004; Wilson et al., 2006). Therefore, sapling and pole sized podocarps, 

which do not bear fruit, have no immediate impact on kākāpō vital rates (survival and 

reproduction) and thus can be excluded from data analysis. In addition to this, sapling trees 

have very variable spatial distribution patterns compared to adult trees. For example, spatial 

patterns shift from high clumping to looser aggregation or random, independent distribution 

when moving from saplings to adults for the same tree species (Fangliang et al., 1997; Nanami 

et al., 2005; Jensen, & Meilby, 2012). This high variability in patterns of spatial distribution 

among sapling/pole trees would make it difficult to apply an appropriate detection function and 

estimates of density for sapling and pole sized trees are likely to be unreliable. 

2.2.4 Selection of best Distance model 

 

DISTANCE software enables one of four basic detection functions, with various standard 

adjustments, to be selected for (Kissa, & Sheil, 2012). Selection of the most appropriate 

detection function was based on the minimum Akaike Information Criterion (AIC) (Buckland 

et al., 1997; Thomas et al., 2010) in conjunction with meeting the criteria for the utility of 

model classes as outlined by Buckland et al. (2001) (see below) (Miller & Thomas, 2015). 

Detection function models should be:  

1. flexible, so that they can take a wide variety of shapes;  



Chapter 2 Estimating density of key tree species 

35 

2. efficient, many plausible shapes can be represented using few parameters;

2. flat at zero distance (g0 (0; θ) = 0), indicating that objects in the immediate vicinity of the

observer are detectable at near certainty. The rate of diminution should be slow initially and 

present as a ‘shoulder’; and,  

4. monotonic non-increasing with increasing distance (g0 (y; θ) 0 for 0 < y w), as it is usually

unrealistic for detection to increase with increasing distance. 

Goodness of fit (GoF) was also tested by chi-square for each model to determine the absolute 

performance of each model compared to the data. GoF is used to ensure that the model selected 

as being the best does actually have a reasonable fit to the data. 

The Negative exponential model was excluded from model selection for all DISTANCE 

analyses. This is because the estimated probability of detection for this model falls off more 

quickly with distance than is consistent with the sightings made by the observers (Thomas et 

al., 2010). This negates the criterion that the detection function model should be flat at zero 

distance and this model also always yields significantly higher tree density estimates than other 

detection function models. The use of the negative exponential model is also heavily 

recommended against as it is biologically less realistic than other models and represents cases 

where detection decreases rapidly with increasing distance from the line (Burnham, et al., 

1985; Thomas et al., 2010). 

2.2.5 Comparison of density across the whole mountain and within the Southern Enclosure 

Density estimates of key tree species were made for the entire ecological island and also within 

the boundaries of the Southern Enclosure exclusively. Investigating density of key tree species 

within the Southern Enclosure is of particular interest to kākāpō conservation as this location 

is where the majority of initial reintroduction releases occur at Maungatautari (MEIT, pers. 

comm.). The Southern Enclosure has previously been proposed by the Maungatautari 

Ecological Island Trust (MEIT) as a potential initial release site if kākāpō are to be reintroduced 

(MEIT, pers. comm.). Therefore, density estimates of key tree species within the Southern 

Enclosure are useful to identify whether this site contains sufficient food sources for kākāpō 

and whether it could be a suitable release site. However, alterations to the current predator-

proof fence would need to be made for this option to be explored further. 



Chapter 2  Estimating density of key tree species 

 

 36 

2.3 Results 

2.3.1 Locations of key tree species 

 

A total of 260 adult trees were identified during the distance line transect survey. Rimu and 

miro were found to be well distributed across the entire mountain whilst matai and Hall’s totara 

were found to occur only in one patch in the north (Figure 2.2). Rimu (Dacrydium cupressinum) 

was the dominant species identified in this survey, accounting for 56.5% of the total number 

of adult trees observed. Miro (Prumnopitys ferruginea) accounted for 38.1% of adult trees 

surveyed followed by Hall’s Totara (Podocarpus laetus) (3.1%) and Matai (Prumnopitys 

taxifolia) (2.3%). No adult beech (Nothofagus spp.) were identified during the distance line 

transect survey. 

Figure 2.2. Locations of the adult tree observations made in the distance line transect sampling 

survey on Maungatautari of tree species identified as being important to kākāpō habitat. 

Different tree species are differentiated by different shapes and colours as indicated in the map. 

Map compiled in ArcMap GIS 



Chapter 2 Estimating density of key tree species 

37 

2.3.2 Estimates of density of all adult key tree species combined and adult rimu only across 

Maungatautari 

DISTANCE software enables the user to select from one of four basic detection functions 

(Kissa, & Sheil, 2012). The Hazard-rate model (with variable series adjustments) was found to 

be most suitable for both groups (all key tree species combined, and rimu only) (Table 2.1), as 

judged by low AIC and consistency with Buckland et al.’s (2001) criteria for selecting an 

appropriate detection function model (Miller & Thomas, 2015). 

For the ‘all key tree species combined’ group, the hazard-rate model (series adjustment 

insignificant as all produce congruent results) was selected as the most appropriate detection 

function with the lowest AIC value relative to the other models (Table 2.1). Additionally, the 

detection probability graph (Figure 2.4) shows that the selected detection function curve fits 

the data well and has a slightly wider shoulder more consistent with the data compared to some 

other detection function models. The GoF p-value of 0.868 for the selected detection function 

model indicates that it does have a reasonable fit to the data and is therefore appropriate to use. 

The hazard-rate model estimated a density of 2.310 key trees/ha across the entire mountain 

with a 95% confidence interval ranging between 2.247 and 4.876 trees/ha. The total abundance 

of key trees was estimated as 7,769 trees with a 95% confidence interval between 7,557 and 

16,398 trees (Table 2.1). 

The hazard-rate model (series adjustment insignificant as all produce congruent results) was 

also selected as the most suitable detection function for estimating the density of adult rimu 

trees across the entire mountain. This model has the lowest AIC value relative to the other 

models (Table 2.1). Additionally, the detection probability graph (Figure 2.5) shows that this 

detection function fits the data well with a shoulder consistent with the rate of diminution of 

the observed data. The GoF p-value of 0.825 for this selected detection function model 

indicates a reasonable fit to the data and it is therefore appropriate to use. The hazard-rate 

model estimated a density of 1.113 adult rimu trees/ha across the entire mountain with a 95% 

confidence interval ranging between 0.662 and 1.873 rimu trees/ha. The total abundance of 

adult rimu over the entire mountain was estimated at 3,743 trees with a 95% confidence interval 

between 2,226 and 6,299 trees (Table 2.1). 



Chapter 2  Estimating density of key tree species 

 

 38 

Table 2.1. Comparison of density estimates between different detection function models and 

series adjustments for adult trees of ‘all key tree species’ and ‘rimu only’ across Maungatautari. 

Tree 

population 

Model/ 

Adjustment 

series 

AIC 
GoF 

(p-value) 
�̂� 

�̂�  

95% CI 
N 

N  

95% CI 

All key tree 

species 

U/C 1554.30 0.067 2.218 1.572-2.731 7459 5287-9184 

U/SP 1704.69 0.000 0.928 0.662-1.301 3121 2226-4375 

U/HP 1658.21 0.000 1.135 0.805-1.602 3817 2707-5388 

HN/C 1548.53 0.353 2.465 1.740-2.493 8290 5852-8384 

HN/SP 1548.53 0.353 2.465 1.740-2.493 8290 5852-8384 

HN/HP 1622.22 0.000 1.454 1.038-2.036 4890 3491-6847 

HR* 1539.53 0.868 2.310 2.247-4.876 7769 7557-16398 

Adult rimu 

U/C 931.40 0.558 0.915 0.570-1.470 3077 1917-4944 

U/SP 939.82 0.049 0.717 0.449-1.143 2411 1510-3844 

U/HP 941.71 0.044 0.718 0.002-290.992 2415 7-978606 

HN/C 928.73 0.678 0.947 0.590-1.521 3185 1984-5115 

HN/SP 932.42 0.483 0.889 0.553-1.427 2990 1860-4799 

HN/HP 944.65 0.004 0.719 0.451-1.145 2418 1517-3851 

HR* 927.03 0.825 1.113 0.662-1.873 3743 2226-6299 

U/C = Uniform/Cosine, U/SP = Uniform/Simple Polynomial, U/HP = Uniform/Hermite 

polynomial, HN/C = Half-normal/Cosine, HN/SP = Half-normal/Simple Polynomial, HN/HP 

= Half-normal/Hermite polynomial, HR/C = Hazard-rate/Cosine, HR/SP = Hazard-rate/Simple 

Polynomial, HR/HP = Hazard-rate/Hermite polynomial; AIC = Akaike Information Criterion; 

GoF = goodness of fit test probability value; 95% confidence interval (CI); �̂� = Density 

Estimate; 𝑁  = Total Abundance Estimate. 

*Replicate results not shown; these detection function models produce identical results 

regardless of the series adjustment applied.  



Chapter 2  Estimating density of key tree species 

 

 39 

 

Figure 2.4. Detection probability for all adult key tree species across Maungatautari. The 

histogram represents the number of trees observed at different distances from the transect line. 

The smooth curve is the detection probability predicted by the best detection function (Hazard-

rate model with any series adjustment) Figure produced in DISTANCE. 

 

Figure 2.5. Detection probability for adult rimu across Maungatautari. The histogram 

represents the number of trees observed at different distances from the transect line. The 

smooth curve is the detection probability predicted by the best detection function (Hazard-rate 

model with any series adjustment) Figure produced in DISTANCE. 

  



Chapter 2 Estimating density of key tree species 

40 

2.3.3 Estimates of density of all adult key tree species combined and adult rimu only within 

the Southern Enclosure 

All of the detection function models produced fairly similar estimates of density within both 

groups (all key tree species combined, and rimu only) for inside the Southern Enclosure. 

However, the half-normal model was found to be most suitable for both groups within the 

Southern Enclosure (Table 2.2) as judged by the criteria described above.  

For the ‘all key tree species combined’ group, the half-normal model (series adjustment 

insignificant as all produce congruent results) was selected as the most appropriate detection 

function as it has the lowest AIC value relative to the other models (Table 2.2). Additionally, 

the detection probability graph for the selected model shows that the detection function curve 

fits the data well and the rate of diminution appears to be initially slow, presenting as a shoulder 

at proximate distances (Figure 2.6). The GoF p-value of 0.557 for the selected detection 

function model also signifies that it fits the data well. The half-normal model gives an estimated 

density of 2.454 key trees/ha within the Southern Enclosure with a 95% confidence interval 

ranging between 1.978 and 2.644 trees/ha. The total abundance of key trees within the Southern 

Enclosure was estimated at 160 trees with a 95% confidence interval between 129 and 172 

trees (Table 2.2). 

The half-normal model (series adjustment insignificant as all produce congruent results) was 

also selected as the most suitable detection function for estimating the density of rimu trees 

within the Southern Enclosure. This model has the lowest AIC value relative to the other 

models (Table 2.2). Furthermore, the detection probability graph (Figure 2.5) shows that this 

detection function fits the data well and is consistent with the criteria for the utility of model 

classes outlined by Buckland et al. (2001) and described above. The GoF p-value of 0.694 for 

this selected detection function model indicates a reasonable fit to the data and making it 

suitable to use. The half-normal model produces an estimate for density of 2.206 rimu trees/ha 

within the Southern Enclosure with a 95% confidence interval ranging between 1.749 and 

2.784 rimu trees/ha. The total abundance of rimu within the Southern Enclosure was estimated 

to be 143 trees with a 95% confidence interval between 114 and 181 trees (Table 2.2). 



Chapter 2 Estimating density of key tree species 

41 

Table 2.2. Comparison of density estimates between different detection function models and 

series adjustments for adults of ‘all key tree species’ and adult rimu within the Southern 

Enclosure of Maungatautari.  

Tree 

population 

Model/ 

Adjustment 

series 

AIC 
GoF 

(p-value) 
�̂� 95% CI N 95% CI 

All key 

tree 

species 

U/C 176.89 0.490 2.702 1.595-4.578 176 104-298

U/SP 174.43 0.493 2.420 1.704-2.438 157 111-158

U/HP 182.93 0.025 1.618 0.016-161.687 105 1-10510

HN* 172.73 0.557 2.454 1.978-2.644 160 129-172

HR* 172.84 0.616 2.609 1.784-2.814 170 116-183

Adult rimu 

U/C 166.58 0.491 2.073 1.260-2.410 135 82-157

U/SP 170.13 0.033 1.375 1.244-1.521 89 81-99

U/HP 171.98 0.052 1.498 0.029-78.047 97 2-5073

HN* 162.21 0.694 2.206 1.749-2.784 143 114-181

HR* 162.78 0.616 2.298 1.545-2.416 149 100-157

U/C = Uniform/Cosine, U/SP = Uniform/Simple Polynomial, U/HP = Uniform/Hermite 

polynomial, HN/C = Half-normal/Cosine, HN/SP = Half-normal/Simple Polynomial, HN/HP 

= Half-normal/Hermite polynomial, HR/C = Hazard-rate/Cosine, HR/SP = Hazard-rate/Simple 

Polynomial, HR/HP = Hazard-rate/Hermite polynomial; AIC = Akaike Information Criterion; 

GoF = goodness of fit test probability value; 95% confidence interval (CI); �̂� = Density 

Estimate; N = Total Abundance Estimate 

*Replicate results not shown; these detection function models produce identical results

regardless of the series adjustment applied. 



Chapter 2 Estimating density of key tree species 

42 

Figure 2.6. Detection probability for all adult key trees within the Southern Enclosure of 

Maungatautari. The histogram represents the number of trees observed at different distances 

from the transect line. The smooth curve is the detection probability predicted by the best 

detection function (Half-normal model with any series adjustment). Figure produced in 

DISTANCE. 

Figure 2.7. Detection probability for adult rimu within the Southern Enclosure of 

Maungatautari. The histogram represents the number of trees observed at different distances 

from the transect line. The smooth curve is the detection probability predicted by the best 

detection function (Half-normal model with any series adjustment). Figure produced in 

DISTANCE. 



Chapter 2  Estimating density of key tree species 

 

 43 

2.4 Discussion  

 

The aim of this chapter was