Anthony D. Griffiths, B. Ed. (Env. Sci.) Faculty of Science Northern Territory University Darwin

Transcription

1 The effect of a seasonal environment and fire on the ecology of frillneck lizards, Chlamydosaurus kingii, in the wet-dry tropics of northern Australia. by Anthony D. Griffiths, B. Ed. (Env. Sci.) Faculty of Science Northern Territory University Darwin A thesis submitted to the Northern Territory University in fulfilment of the requirements for the degree of Master of Science. November, 1994

2 Statement of Sources This thesis is my original work and has not been submitted, in whole or in part, for a degree at any other University. Nor does it contain, to the best of my knowledge and belief, any material published or written by another person, except as acknowledged in the text. Signed Tony Griffiths 11

3 This work is dedicated to the memories of my mother Lou Griffiths and Alicia Johnson.

4 Acknowledgments I would like to thank Dr Keith Christian for his supervision, technical support, and constructive comments throughout the course of this project. His thoroughness and patience during the writing of this thesis is of special note, and hopefully this will assist me in my future endeavours. My family and friends provided tremendous support and understanding. In particular, Ben Scambary helped in so many small but important ways. Marisa Fontes played an instrumental role in the maintenance of my sanity. Logistical support in the form of fues and workshop space was provided by Kapalga Research Station in Kakadu National Park. Robert Eager's generosity and helpful advice were especially helpful, particularly during those long wet season days. My thanks to all the fue crews for lighting the fires, and for keeping out all the unwanted ones. Many volunteers ably assisted with fieldwork, too many to list here. Ajio Pereira and Gavin Bedford gave considerable help in the laboratory constructing radio transmitters and answering questions. Owen Price and Zhoa Yuen explained the intricies of GLIM. Dick Braithwaite's generous nature and flexibility enabled me to keep myself financially solvent during the course of this project. John Woinarski must also be thanked for putting the idea of post-graduate study into my head in the fust place. Valuable comments on early drafts of this thesis were provided by Keith Christian, John Woinarski, Dick Williams, Gavin Bedford, Dave Bowman, Grant Farrell and Dick Braithwaite. The project was partly funded by student travel grants from Australian Nature Conservation Agency (ANCA), North Australia Research Unit student assistance scheme, and Australian Geographic. Permission to work on frillnecks was given by Tropical Ecosystems Research Centre and Northern Territory University Animal Ethics Committees. Permits were granted by Conservation Commission of the Northern Territory and ANCA. iii

5 Abstract A population of Chlamydosaurus kingii, in Kakadu National Park, Northern Territory, was studied between April 1991 and April This species is a large, arboreal and insectivorous agamid lizard which inhabits open forests and woodlands throughout northern Australia. The principal aims of the study were to examine the ecological relationships between this species and seasonal variation experienced in the wet-dry tropics, and the frequent annual fires in this region. Additional information on the population dynamics and the previously undocumented dry season ecology was also collected. The study involved the telemetry and mark-recapture of C. kingii in Eucalypt dominant tall open forest during the dry and wet seasons. All research was conducted within Kapalga Research Station, where a landscape scale fire experiment was in progress. Seasonal variation has a significant effect on the ecology of Chlamydosaurus kingii. Frillneck lizards show a clear response to the prolonged dry season. This includes a decrease in the volume of food taken, reduced activity and reduced seasonal growth. This coincides with the selection of large Eucalyptus trees. Termites are a substantial part of their diet during the dry season. The presence of termites in their diet is initially perplexing given the "sit and wait" foraging behaviour of frillneck lizards, but the unique foraging behaviour (above the ground during the day) exhibited by Australian Harvester termites (Drepanotermes) allows frillneck lizards access to them when other food resources are generally low. Growth is reduced in the dry season as a result of the lower volume of food ingested. There is a general trend of energy conservation in the dry season. Male frillneck lizards are able to maintain their body condition throughout the dry season, but their lv

6 condition decreases at the commencement of the reproductive period. This presumably is due to the increased energy expenditure associated with defending home ranges and sexual activity. Females show considerably more variation in body condition, apparently related to the production of eggs. Fire is common throughout northern Australia. Frillneck lizards are able to survive fues lit in the first few months of the dry season by remaining perched in trees. A higher level of mortality occurs in the late dry season fues, along with a change in the behavioural response to fire by sheltering in either larger trees or in hollow termite mounds. Food is more accessible after fues due to the removal of ground vegetation. This is reflected in greater volume and diversity of prey in stomach contents after fues. This increase is more pronounced after late dry season fues, possibly due to the greater amount of ground vegetation removed by these more intense fues. During the dry season frillneck lizards prefer trees with a medium canopy cover located in an area with a low density of grass. Fire has an effect on this relationship. Habitat which has not been burnt for a number of years develops a denser layer of grass and possibly a more continuous tree canopy cover, compared to annually burnt habitat Frillneck lizards in habitats unburnt for a number of years tend to perch in trees with a smaller canopy. Diets are generally similar among the three fire treatments. In general, fire influences the habitat use of frillneck lizards by creating more heterogenous habitats, with more open areas. v

7 Frillneck lizards restrict their reproductive cycle to the wet season (summer months), as is common with other agamids. The reproductive season begins prior to the onset of monsoonal rains and the related increase in food availability. Hatchlings begin to emerge during the period of highest food availability, although some late hatchings occur, and they may experience greater difficulty in obtaining food. Male frillneck lizards grow faster than females and have a larger body size. Males also maintain larger home ranges than females in the dry season, which may ensure the presence of a number of females within their home ranges. Only large males actively defend their home ranges. The low densities of frillneck lizard in one site may be related to the absence of fire, and this may affect the long-term viability of populations. Vl

8 Relevant Publications (copies attached at the end of the thesis) Bedford, G.B., Christian, K.A. and Griffiths, A.D. (1993). Preliminary investigations on the reproduction of the frillneck lizard (Chlamydosaurus kingii) in the Northern Territory. pp /n Lunney, D. and Ayers, D. (eds.), "Herpetology in Australia: A Diverse Discipline". Transactions of the Royal Zoological Society of New South Wales. Surrey Beatty and Sons, Sydney. Attachment l Vll

9 Table of Contents Statement of Sources (ii) Acknowledgments (iii) Abstract (iv) Relevant Publications... (vii) Table of Contents (viii) List of Figures (x) List of Tables (xvi) Chapter 1 General Introduction Chapter 2 Seasonal ecology of the frillneck lizard, Chlamydosaurus kingii, in the wet-dry tropics of Australia INTRODUCTION METHODS. RESULTS Rainfall Diet Food availability Body condition Seasonal growth rate and change in body mass Habitat use Seasonal activity index DISCUSSION Summary Chapter 3 The short-term and longer-term effects of annual fue on the behaviour, diet, growth, and habitat use of the frillneck lizard, Chlamydosaurus kingii INTRODUCTION METHODS Vlll

10 RESULTS Short-term effects of fire Longer-term effects of flre DISCUSSION Short-term effects of fire Longer-term effects of fire Summary Chapter 4 The demography, population dynamics and dry season home range of frillneck lizards, with reference to the effects of three different ftre regimes INTRODUCTION METHODS RESULTS Reproduction biology Growth and age Population dynamics Home ranges DISCUSSION Summary Chapter 5 Synopsis and management implications. 118 References lx

11 List of Figures Figure 2.1. A map showing the location of Kapalga Research Station, Kakadu National Park within the Northern Territory, Australia. 7 Figure A map showing the location of permanent study sites within Kapalga Research Station, Kakadu National Park. 9 Figure 2.3. Monthly rainfall (mean of 3 sites ) recorded at Kapalga Research Station (data from CSIRO's Division of Wildlife and Ecology). 15 Figure 2.4. The mean total volume of stomach contents (ml) over four seasonal periods. Closed circles represent males, open circles represent females and numbers are sample sizes. Error bars are one standard error. 16 Figure 2.5. Relationship between total invertebrate abundance collected from sweepnetting at each site and the corresponding total rainfall for the previous three months. 23 Figure 2.6. Mean residuals of linear regression of log-body mass with log-svl for adult frillneck lizards. Closed circles are adult males, open circles are adult females, and numbers are sample sizes. Error bars are one standard error. Line through zero represents the best least squares fit. 24 X

12 Figure 2.7. Seasonal growth rate (a,b) and change in body mass (c,d) relative to initial body length. Closed circles are wet season values and open circles are dry season values. 26 Figure 2.8. The mean height of trees (m) used by frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are nontelemetered lizards, numbers represent sample sizes. Error bars are one standard error. 32 Figure 2.9. The mean trunk diameter (em) of trees used by frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are nontelemetered lizards, numbers represent sample sizes. Error bars are one standard error. 33 Figure The mean perch height (m) of frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are non-telemetered lizards, numbers represent sample sizes. Error bars are one standard error. Figure Monthly variation in number of lizards sighted or captured per kilometre driven during routine censussing of sites. Numbers are the number of censuses each month, and error bars are one standard error. 36 Figure 3.1. Location of three permanent fire treatment sites within Kapalga Research XI

13 Station. The shaded areas represent each site. 49 Figure 3.2. Frequency distribution of the number of invertebrate orders and prey size classes in the stomach contents of frillneck lizards, one week before and one week after early dry season fire. Open bars represent before fire and closed bars represent after fire. 59 Figure 3.3. Relative abundance and relative volume of prey taxa in stomach contents, one week before and one week after early dry season fires. Open bars represent before fire and closed bars represent after fire. Abbreviations for prey taxa are:!so.- lsoptera; Orth.- Orthoptera; Hem. - Hemiptera; Col. - Coleoptera; Dip. - Diptera; Lep. - Lepidoptera; Hym. - Hymenoptera; Blat. - Blattodea; Mant.- Mantodea; Odon.- Odonata; Phas.- Phasmotodea; Aran.- Aranea; Chi/. - Chilopoda. 60 Figure 3.4. Frequency distribution of the number of invertebrate orders and prey size classes in the stomach contents of frillneck lizards, one week before and one week after late dry season fires. Open bars represent before fire and closed bars represent after fue. 62 Figure 3.5. Relative abundance and relative volume of prey taxa in stomach contents of frillneck lizards, one week before and one week after late dry season fires. Open bars represent before fire and closed bars represent after fire. Abbreviations of prey taxa names follows Figure xu

14 Figure 3.6. The effect of significant variables in a logistic regression model of the occupancy of trees by frillneck lizard during the dry season. The graphs show changes in estimated probability of occurrence of frillneck lizards with changes in (a) tree canopy cover; {b) the ground vegetation density at 25 em; (c) trunk diameter; and (d) tree height. For each relationship the values of other variables in the model are fixed at the mean value. For graphs {a) and {b), solid lines represent the unburnt site, dashed lines are the early dry season fire site and the dotted lines are the late dry season fire site. 69 Figure 3.7. Relative abundance of the number of prey size classes and invertebrate orders present in the stomach contents of frillneck lizards for each fire regime in the dry season. Open bars represent unbumt treatment, hatched bars rising right are early ftre treatment and hatched bars rising left are the late fire treatment. 72 Figure 3.8. Relative abundance of the number of prey size classes and invertebrate orders in the stomach contents of frillneck lizards for each fire regime during the wet seasons. Open bars represent unbumt, hatched bars rising right are early fire treatment and hatched bars rising left are the late fire treatment. J Figure 3.9. Mean residuals of linear regression of log-body mass with log-svl for adult male frillneck lizards in each of the fire treatments, during the dry and wet Xlll

15 seasons. Numbers are sample sizes and error bars are one standard error. Line through zero represents the best least squares fit. 79 Figure Mean residuals of linear regression of log-body mass with log-svl for adult female frillneck lizards in each of the fire treatments, during the dry and wet seasons. Numbers are sample sizes and error bars are one standard error. Line through zero represents the best least squares fit. 80 Figure 4.1. The percentage of home range estimated from the cumulative number of locations collected in the dry season. Approximately eight locations are needed to describe 80% of a frillneck lizards' dry season home range. 94 Figure 4.2. The number of gravid and non-gravid female frillneck lizards captured each month during the wet season. Both reproductive seasons are combined. 96 Figure 4.3. Growth rate of male and female frillneck lizards. The mid-point of SVL between first and last recapture is used. Open circles represent males and closed circles represent females. 99 Figure 4.4. Number of lizards in the different size cohorts within each fire treatment site. Open bars represent females, closed bars represent males and lined bars represent juveniles and hatchlings. 102 Figure 4.5. Mean monthly capture rates of males (circles) and female (squares) frillneck XIV

16 lizards. All sites and years have been combined. 104 Figure 4.6. Proportion of adult male frillneck lizards with scars for four size classes in each of the fire treatment sites. Triangles represent the late fire site, squares represent the early ftre site and circles represent the unburnt site. 108 XV

17 List of Tables Table 2.1 Occurrence, relative abundance and relative volume of prey taxa present in stomach samples of C. kingii during the dry and wet seasons. 18 Table 2.2. The number and relative abundance (lk) of invertebrate orders from sweep- netting during the dry and wet seasons. 21 Table 2.3. Mean and standard error of seasonal growth rate as measured by changes in SVL (mm day"1) and change in body mass (% g day"1) for male and female frillneck lizards in the wet and dry seasons. Numbers in parentheses are sample sizes. 27 Table 2.4. The relative abundance (%) of tree species used by telemetered and nontelemetered frillneck lizards during the wet and dry seasons, and the relative abundance (%) of tree species available in the field as determined by a random walk technique. 30 Table 2.5. A summary of tropical insectivorous lizard's foraging strategy, habitat, and relative abundance of ants and termites in their diets. 40 Table 3.1. Post-fire selection of habitat by telemetered frillneck lizards. All lizards were monitored adjacent to burnt or unburnt habitat after the early and late dry season fires during XVI

18 Table 3.2. Means ± one standard error of habitat structure variables used for the logistic regression model. Data describing habitat structure of open forests were collected from each flre treatment site. Occupied trees were known to contain frillneck lizards, and unoccupied trees randomly selected trees were assumed not to contain frillneck lizards. 66 Table 3.3. Logistic regression model for occupancy of trees by telemetered frillneck lizards during the dry season. X2 values refer to difference in scaled deviance from total deviance of the null model. The 'k total deviance is the proportion of scaled deviance from the total deviance of the null model. Treatment (site) factors are represented in the model by: (1) early dry season site; (2) late dry season site; and (3) unburnt site. 67 Table 3.4. Selectivity indices for tree species used by frillneck lizards within each site during the wet and dry seasons. 71 Table 3.5. Dry season stomach contents of frillneck lizards from the three ftre treatment sites. 75 Table 3.6. Wet season stomach contents of frillneck lizards from the three fire treatment sites. 76 xvu

19 Table 3.7. Relative abundance of invertebrate orders obtained from sweep-netting in each of the three fire treatment sites, during the dry and wet seasons. Samples from both years are combined. Taxa denoted with (--) were not recorded from sweep-netting but were recorded in the stomach contents of frillneck lizards. 78 Table 4.1. Observed and expected capture frequencies, population estimates and "goodness of fit" test for the early fire site. 105 Table 4.2. Observed and expected capture frequencies, population estimates and "goodness of fit" test for the late fire site. 105 Table 4.3. Observed and expected capture frequencies, population estimates for the unburnt site. The "goodness of fit" test was not possible due to the low number of recaptures. 106 Table 4.4. Dry season home range and movement for telemetered male and female frillneck lizards in each of three fire sites. 109 XVlll

20 Chapter 1 General Introduction Approximately 270 terrestrial reptile species are known from the northwestern region of Australia (Woinarski and Braithwaite 1991). Much of this region is sparsely populated and relatively undisturbed by European settlement. The terrestrial reptiles of this region experience consistently high ambient temperatures, an annual cycle of high rainfall in the summer months and a prolonged dry period. This creates a unique set of environmental conditions to study terrestrial reptiles. Lizards comprise a large component of this terrestrial reptile fauna. Research on the population ecology of terrestrial lizard species in this region is sadly lacking. Previous population ecology studies have made important contributions in establishing a range of ecological relationships (Shine 1986, James and Shine 1988, Shine and Lambeck 1989). However, the population ecology of the majority of lizard species and their relationships with current ecological theories remain unknown. The frillneck lizard, Chlamydosaurus kingii (Sauria: Agamidae), is possibly the most recognisable member of the lizard fauna of this region. Its presence on the now defunct two cent coin, innumerable post cards and tourist brochures has made it into an unofficial faunal emblem of northern Australia. It has an extensive distribution in northern Australian extending from the west to the east coast, and into the southern part of Papua New Guinea (Cogger 1992). Shine and Lambeck ( 1989) described the general ecology of this species. 1

21 This thesis is presented in four chapters. Chapters two, three, and four and are written as individual papers and examine a range of ecological relationships between C. kingii, its environment and other taxonomically related lizard species. The fifth chapter presents a synopsis of the three previous chapters, along with some management implications arising from this research. A brief description of each chapter is given below. Chapter two examines the seasonal variation in the ecology of C. kingii. Dry season ecology of this species was previously undocumented, mainly due to the cryptic nature of its behaviour during this period. By monitoring lizards throughout the year, it is possible to determine and document the seasonal ecological changes. This information will provide comparative information for future studies of seasonal variation in terrestrial reptile ecology in the wet-dry tropics of Australia. Chapter three investigates the short-term and longer-term effects of annual fires on the ecology of C. kingii. Much of the open forest inhabited by frillneck lizards is regularly burnt during the dry season. Questions pertaining to how lizards survive these fires and the effects of ftre on their diet and habitat selection in the subsequent weeks are examined in the ftrst half of this chapter. The longer-term effects on diet, habitat use and body condition are considered in the second half of this chapter. By examining these aspects of the frillneck lizard ecology, it is then possible to consider the relative importance of fire on these factors, and whether this may effect the management of this species in northern Australia. 2

22 Chapter four presents data pertaining to the demographics, population dynamics and dry season home ranges of frillneck lizards. This chapter also compares the information from frillneck lizards to the life history of other large Australian agamids. The effect of fire on the population dynamics is also examined. Chapter 5 presents a brief synopsis of the conclusions presented in the previous chapters. This will include both general and noteworthy results of the ecology and life history strategies of C. kingii in the wet-dry tropics of Australia. Management of this species, in the light of information presented in this thesis, is briefly discussed. 3

23 Chapter 2 Seasonal ecology of the frillneck lizard, Chlamydosaurus kingii, in the wet-dry tropics of Australia. INTRODUCTION Seasonal variation in temperate and arid habitats has a significant effect on the activity, reproduction, resource acquisition, growth and habitat use of lizards (Dunham 1978, Ballinger and Congdon 1980, Rose 1981, Andrews 1982, Paulissen 1988). Comparative studies of lizard ecology and life history at the family level (predominantly lguanids) in these habitats have contributed significantly to current theories of life history strategies and general ecology (see Huey et a/. 1983, Pianka 1986). The understanding of the seasonal ecology of reptiles in the seasonal tropics is poor. Constant high ambient temperatures and high periodic annual rainfall combine to create a unique environment for reptiles. Studies of tropical lizard ecology have encompassed a variety of habitats (tropical savannas to closed rainforest) in several families, and they have revealed a surprising diversity of responses to seasonal variation (Sexton et a/. 1972, Aeming and Hooker 1975, Huey et a/. 1977, Christian et a/. 1983, Floyd and Jenssen 1983, Vitt 199la, Vitt 1991b, Vitt and Blackburn 199 1, Bullock et a/. 1993, van Marken Lichtenbelt 1993, van Marken Lichtenbelt et a!. 1993). This diversity and the relatively small number of studies limit tests of current ecological theories in tropical habitats. Previous studies of lizard ecology in the wet-dry tropics of northern Australia are few 4

24 considering the high species diversity of the region. Seasonal variation in diet (James et a/. 1984, Shine 1986), reproductive strategy (James and Shine 1985, James and Shine 1988), and thermoregulation and energetics (Christian and Green 1994, Christian and Bedford 1995) have provided a number of hypotheses relating to the evolution of certain life history strategies for lizards in this region. Frillneck lizards inhabit open forests and woodlands of the wet-dry tropics of Australia, and on the east coast they extend into the temperate climate zone (Cogger 1992). The general ecology of this species was studied by Shine and Lam beck ( 1989), but their field data is exclusively from the wet season. They revealed a diurnal, arboreal, "sit-and-wait" insectivorous lizard. Reproduction occurs exclusively during the wet season, and the lizards reduce activity during the dry season. They are sexually dimorphic in body size with adult males considerably larger than adult females. Frillneck lizards exhibit a significant reduction in body temperature, field metabolic rate and water turnover during the dry season (Christian and Green 1994, Christian and Bedford 1995). The principal aim of this chapter is to establish how the ecology of the frillneck lizard changes between the wet and dry seasons, in light of the changes in their physiology and the environment. The specific aspects of their ecology considered in this chapter are: diet and food availability; body condition and growth rate; habitat use; and their seasonal activity. The relative importance of the effect of each environmental factor (e.g. temperature, relative humidity) on the ecology of frillneck lizards is difficult to determine due to the interdependent relationships of these factors. However, an assessment of the strategies used by C. kingii in relation to other lizard species is possible. 5

25 METHODS Site All field work was done between April 1991 and April 1994 at Kapalga Research Station (CSIRO Division of Wildlife and Ecology) in Kakadu National Park (12Q 43' S, 132Q 26' E), 250 km east of Darwin, Northern Territory, Australia (Figure 2.1). This region experiences a distinct seasonal wet-dry climate. The wet season in Kakadu National Park (October to April) is characterised by high maximum (35"C) and minimum (24.C) air temperatures, high relative humidity (60% at 0900 hrs) and high rainfall (1480 mm) (13 year averages from the Jabiru airport, Bureau of Meteorology). Rainfall is irregular and relatively low during the first months of the wet season (early wet season; October to December), becoming more consistent and relatively higher during the second half (late wet season; January to April). The dry season (May to September) is characterised by high maximum (33.C) but lower minimum (l9.c) air temperatures, lower relative humidity, and little or no rainfall (Bureau of Meteorology). The dry season may also be divided into an early dry season (May to July) when the soil and vegetation begin to dry out after the cessation of monsoonal rains, and a late dry season (August to September) when conditions are extremely dry. The study was conducted in open forest dominated by Eucalyptus tetrodonta, E. miniata, E. porrecta and Erythrophleum chlorostachys, with an understorey dominated by annual grass Sorghum spp. The study area was based around a network of roads to facilitate the censussing of lizards. Data were collected at three separate sites and they ranged in 6

26 0 KAPALGA RESEARCH STATION ARNHEM LAND kilometres Figure 2.1. A map showing the location of Kapalga Research Station, Kakadu National Park within the Northern Territory, Australia. 7

27 area from 400 to 500 ha (Figure 2.2). Each site was subjected to a different fire regime. but this will be discussed in Chapter 3. For this Chapter, the three sites are examined together as one site. Sampling All frill neck lizards were originally sighted by driving slowly ( < 30 km hr-1) along the roads, and either caught by hand or by a noose attached to a telescoping pole. At the initial capture, each lizard was given a unique identifying number with a transponder (Destron IDI) implanted sub-cutaneously under a loose fold of skin in the neck area. At each capture the following data were recorded: date; time; mass to the nearest g using an electronic balance (Bonso ); and the snout-vent length (SVL) to the nearest mm using a perspex ruler. Lizards were released at the exact point of capture within two hours of capture. All field work was done between 0700 and 2000 hours. A sub-sample of adult lizards (n = 55) was fitted with small radio-transmitters to assist in data collection. Location transmitters (Biotrack SS-1 and Biotel TX-1) weighing approximately 15 g (2-6% of lizard body mass) were attached to the tail using a small amount of glue and adhesive bandage. After fitting of transmitters, lizards were released at the exact point of capture within two to twelve hours. Telemetered animals were relocated a minimum of twice a month, and transmitters were changed every three months to replace the batteries. Animals were monitored for varying periods over the study period due to movement out of transmission range, battery failure and predation. 8

28 N c:::::i Permanent study sites 0 5 km Figure 2.2. A map showing the location of study sites within Kapalga Research Station, Kakadu National Park. 9

29 The mark-recapture and telemetry study provided data on diet, body condition, seasonal growth, habitat use and activity. Rainfall data were collected fortnightly at each site by the CSIRO Division of Wildlife and Ecology. Diet At initial capture (after April 1992) lizards with a SVL greater than 150 mm were stomach flushed. Lizards were restrained on a wooden board using velcro strips, and a padded plastic ring was placed in their jaws to hold their mouth open. One end of a plastic tube (diameter = 3 mm, length = 25 em) was positioned into the stomach via the mouth, and the other end was fitted to a 60 ml syringe filled with freshwater. The stomach was then filled with approximately 40 ml of freshwater, while the abdomen was palped, and then the animal was inverted. The stomach contents were collected with a permeable cloth stretched over a bucket. Stomach contents were stored in 70% ethyl alcohol for later analysis. Lizards recaptured within less than 6 months of the previous capture were not stomach flushed. Stomach contents were classified to order using a dissecting microscope. Each prey item was assigned to one of five size classes by length (0-5 mm; 6-10 mm; mm; nun; > 20 mm). The volume for each prey taxa was estimated by the volumetric displacement (in a graduated measuring cylinder± 0.1 ml) of a representative sub-samples from each of the five size classes, and then all sizes classes were added together. The total volume (ml) of stomach contents were measured by volumetric displacement of the whole wet sample in a graduated measuring cylinder (± 0.1 ml). 10

30 Food availability was measured by sweep-netting the ground foliage (0-2 m above the ground) for invertebrates (Stamps and Tanaka 1981) every three months at each site. Ten samples of 30 sweeps were collected for each sampling period (total of 900 sweeps per sampling period). Samples were collected along line transects, and no foliage was swept twice. The contents of the net were placed in resealable plastic bags and preserved in 70% ethyl alcohol. Samples were later sorted to the level of order, and into five size classes (as with stomach contents). Habitat Data from the following habitat variables were collected at each capture or sighting of C. kingii; tree species; tree height (m) and perch height (m) of the lizard, using a clinometer (Suunto ); trunk diameter (em) at breast height using a calibrated tape measure. An estimate of the frequency distribution of tree species in the study area was determined using a random walk sampling method (Goldsmith and Harrison 1976). A minimum distance of 30 m between each sampling point was set to avoid concentration of samples within a confined area. Analyses Some seasonal analyses were done by dividing the data into dry (May to September) and wet seasons (October to April), but when sample sizes were large and evenly spread across the year, four seasons were used: late wet (January to March); early dry (April to June); late dry (July to September); and early wet (October to December). 11

31 Diet Diets were analysed using four measures: (1) the total volume (ml) of stomach contents; (2) the relative abundance(%) of total prey items; (3) the relative volume(%) of prey taxa; and (4) the occurrence (%) of one or more items of a particular prey taxon in the stomach contents. Differences in the total volume of stomach contents between sexes and in different seasons were compared using two-way ANOV A. Tukey's comparison of means test was used to determine significant differences among seasons. A non-parametric Mann Whitney U-test was needed to test for seasonal differences in the number of prey orders and size classes due to the non-normal distributions. Seasonal differences in the abundance of prey taxa in the stomach contents, and the availability of prey in the field were analysed using contingency tables. The relative volume of prey taxa was analysed using a Kolmogorov-Srnirnov non-parametric test. Simpson's Diversity index (D) was used to determine seasonal changes in dietary diversity: D = 1 - I,( <pi ) 2 where <pi is either relative abundance or relative volume of each prey taxa in the stomach contents. This index ranges from 0 (low diversity) to a maximum of (1-1/S), where S is the number of prey taxa (Krebs 1985). Body condition Body condition was determined by plotting log10 transformed body mass against log1o transformed SVL of adult male (SVL > 230 mm) and non-gravid adult female (SVL > 175 nun). A linear regression was applied to these data, giving a mean body condition for the total sample. The residual deviation from the regression is indicative of an individual's body condition, corrected for different body lengths (James 1991a). Differences between 12

32 sexes were determined using an unpaired t-test, and a one-way ANOV A was used to test for seasonal differences in the mean residual deviations. Seasonal growth and change in body mass Seasonal growth rates and change in body mass were calculated from lizards caught more than twice within a single dry (May to September) or a single wet season (October to April) period. Seasonal growth rates were determined by dividing the change in SVL by the number of days between the fust and last capture (within a single season). Recaptures of less than one week were excluded. The concurrent change in body mass over the same period was also determined. The change in body mass was divided by the number of days between the first and last capture within a single season, and this number was divided by the initial body mass and expressed as a percentage. Seasonal differences in both seasonal growth and change in body mass were tested using ANCOVA with SVL as covariate. Habitat use Contingency table analysis tested for differences in the relative frequency of tree species used by frillneck lizards between seasons. Seasonal differences in the habitat use of lizards (tree height, trunk diameter and lizard perch height) were analysed using one-way ANOVA. Seasonal activity index An index of activity was calculated by dividing the total number of lizards either sighted or captured in daily car censuses by the number of kilometres driven during each field census. This index is an indirect measurement of the population's general activity, not 13

33 an individual animal's activity. Censuses were pooled into four seasons and tested by Kruskal-Wallis one-way ANOV A. All means are presented with one standard error unless specified. RESULTS Rainfall Rainfall from Kapalga Research Station over the course of this study is illustrated in Figure 2.3. The onset of monsoonal rains occurred in November of each year, apart from the wet season where the first rains were delayed until December. The highest monthly rainfall was recorded in January or February of each year (Figure 2.3). The cessation of rainfall was similar among the consecutive wet seasons, with minimal rainfall occurring after April in each year. Total rainfall for the , , wet seasons were 1116 mm, 1410 mm, and 1307 mm, respectively. Diet A total of 226 stomach content samples were collected b tween May 1992 and April A distinct reduction in total volume of stomach contents in both sexes occurred in the early dry season, after the relatively high total volume recorded in the early and late wet seasons (Figure 2.4). The total volumes of stomach contents of females is greater than males in the first six months of the year (late wet and early dry season), whereas this relationship is reversed in the later half of the year (late dry and early wet season) with males recording a greater total volume than females (Figure 2.4). A two-way ANOV A for sex and seasons, 14

34 E E - 4oo ns o T --. Jan. July 1991 Dec. July Dec Months July 1993 Dec. Figure 2.3. Monthly rainfall (mean of 3 sites) recorded at Kapalga Research Station (data from CSIRO's Division of Wildlife and Ecology). 15

35 9 (/)... c Q) c 8 0 u -5 7 ro E (/) 0 5 :J 0 > 4 ro... 0 c ro Q) \ \ \ T 35 T r la1 / / t/ _--r---, late wet early dry late dry early wet Seasons _ Figure 2.4. The mean total volume of stomach contents (ml) over four seasonal periods. Closed circles represent males, open circles represent females and numbers are sample sizes. Error bars are one standard error. 16

36 was significant for seasons only (seasons, F3 194 = 3.38, P = 0.02). Neither sex or the interaction between sex and season was significant (sex, F1 194 = 0.01, P = 0.92; sex x season, F3.194 = 0.31, P = 0.81). Tukey's comparison of means test on seasons indicated the early dry season mean total volume of the stomach contents was significantly lower than both early and late wet season means, although the late dry season mean was not significantly different from any of the other seasonal periods. The correlation between SVL and total volume for dry season stomach samples was non-significant (r = -0.06, P = 0.48, n = 144). However, the correlation between SVL and wet season total volume indicated a weak but significant relationship for both sexes combined (r = 0.23, P = 0.03, n = 82). Preliminary analyses revealed no significant differences between sexes in the number of prey taxa, the number of size classes of prey or the taxonomic composition of the diet, therefore the results presented here are for both sexes combined. Table 2.1 summarises the occurrence, relative abundance and relative volume of prey taxa for stomach contents of C. kingii for the dry and wet seasons. A total of 15 invertebrate orders were recorded from 144 dry season stomach samples. Stomachs contained a mean of 2.85 ± 0.12 invertebrate orders and a mean of ± 9.15 items. A total of 15 invertebrate orders were recorded from 82 wet season stomach samples. Stomachs contained a mean of 3.96 ± 0.19 prey taxa and a mean of ± prey items. The number of prey orders per stomach sample is a broad indicator of the diversity of prey taxa in C. kingii diet, and the wet season diet contained significantly more orders (Mann-Whitney U-test: z = 4.79, P < ). There was no significant 17

37 Table 2.1. Occurrence, relative abundance and relative volume of prey taxa present in stomach samples of C.kingii during the dry and wet seasons. Dry season Wet season Prey taxa Occurrence Abundance Volume Occurrence Abundance Volume (%) (%) (%) (%) (%) (%) Isoptera Onhopetra Hemiptera Coleoptera Diptera Lepidoptera Hymenoptera Blanodea Mantodea Odonata Phasmotodea Aranea Plecoptera Chilopoda Gastropoda other totals n = 144 n = (ml) n = 82 n = (ml) Simpson's Diversity Index

38 difference in the total number of items per stomach between wet and dry season samples (t = 0.73, DF = 226, P = 0.43). The relative abundance of prey taxa in the dry season was dominated by the order isoptera, comprising 73.3% of all items present. Only one species of isoptera was identified from these stomach samples, Drepanotermes rubriceps (A. Anderson, personal communication), and most of these were soldiers. There is some taxonomic confusion within this genus, and these termites will be referred to as Drepanotermes. Drepanotermes occurred in 36.8% of the dry season stomachs samples, representing 33.6% of the relative volume. Another important dry season prey taxon in the stomach samples was chilopoda (centipedes). The low relative abundance of chilopoda (< 1 %), is misrepresentative of its importance as a food item for frillneck lizards. Most chilopoda present in stomach samples (77%) had a large body length (> 20 mm), and therefore the relative volume of 21.8% is more indicative of the importance of centipedes in the dry season diet. Similarly, orthoptera comprised 14.9% of the relative volume, although relative abundance was low (0.5%). Hymenoptera exhibited a high relative abundance, comprising 23.4% of the total number of prey items taken and was the most corrunon prey taxon in the dry season stomach samples (66.7% occurrence). However, due to the small body length of ants (< 5 rrun), they comprised only 5.17% of the relative volume. Dietary composition during the wet season was dominated by three prey taxa: Iepidoptera, isoptera and chilopoda (Table 2.1). Lepidoptera were present in a high proportion of stomachs (59.7%), and contributed substantially to the relative volume (40.4%). All Iepidoptera in the stomach samples were larvae, of which 37.2% were between rrun 19

39 in length and 29.3% were longer than 20 mm. The relative volume of isoptera was reduced in the wet season samples, compared to the dry season, probably due to the high volume of Iepidoptera. Hymenoptera occur in a high proportion of samples in all seasons, but the contribution to the relative volume of the diet remained small. The relative abundance of prey taxa differed significantly between the wet and dry seasons (X2 = , DF = 7, P < ). The relative volume of various prey taxa also differed significantly between the wet and dry seasons (Kolmogorov-Smirnov: D = 0.23, DF = 14, P = ). Inspection of Table 2. 1 indicates that the variation in relative abundance and relative volume of hymenoptera, Iepidoptera and isoptera accounts for much of the difference between the two seasons. Simpson's diversity index of the relative volume suggests a similar diversity of prey taxa present in the dry and wet seasons (Table 2. 1 ). The diversity index of the relative abundance of prey taxa suggests that stomach samples collected during dry season contain a higher diversity of prey taxa than wet season samples. Food availability Table 2.2 shows the total number and relative abundance of invertebrate orders collected from sweep-netting over two years. Each sample period is the sum of three pennanent sites. A total of 12 invertebrate orders were collected using this method. It should be noted that this method of sampling (sweep-netting) failed to sample some important invertebrate orders that were present in stomach samples of Chlamydosaurus, namely isoptera and chilopoda. There were fewer orders of invertebrates in the dry season samples, compared to wet season samples. Dry season samples contained a high relative abundance of 20

40 Table 2.2. The number anu relative abunuance (%) of invertebrate oruers from sweep-netting during the ury anu wet seasons. Invertebrate Dry season Wet season Dry season Wet season orders Mt y August Decem her March June 1993 August 1993 N ovemher February No. % No. % No. % No. % No. % No. % Nn. % No. % Isopter< Orthopetra R R Hemiptera II. I ( R(i X.3 Coleoptert I 0. () IM (1 7X IX Diptera Lepidoptera (i II I ("'I H yrnenoptera I Blattodea I Mantodea I 0.3 I 0.5 I II. I I 0.1 Ouonata (> X I.9 I 11.4 I Phasmotouea () 2.0 I 0/ I I I 11.4 Aranea X (> ( l.(l Plccoprcra O.X I Chilopodt Gastropod< Total No. items XI

41 hymenoptera, orthoptera, aranea, hemiptera and coleoptera (Table 2.2). Wet season samples contained a high relative abundance of orthoptera, coleoptera and hymenoptera (Table 2.2). Invertebrates were most abundant in the wet season samples. A two-way ANOV A of total abundance of invertebrates by site and seasons ( wet and dry) gave a non-significant result for the interaction site x season (F2.18= 1.75, P = 0.203) and among the three sites (F2_18= 2.49, P = ). However, the total abundance of invertebrates collected from sweepnetting was significantly greater in the wet season than during the dry season (Fu8= 27.19, p < ). There was a strong correlation between total invertebrate abundance at each of the three sites for the 8 sampling periods and total rainfall for the previous three months at each site (r = 0.83, P < , n = 24) (Figure 2.5). The discontinuous nature of this relationship is due to the intense rainfall experienced during the three wet season months (January, February and March), and the relatively low rainfall in the remaining months. Body condition The body mass at a given SVL of adult males was significantly larger than adult females (t = 3.27, OF = 372, P = 0.001). Therefore, body condition was analysed in separate linear regressions for males and females. Male body condition differed significantly among the four seasonal periods (F3 273 = 3.38, P = 0.019) (Figure 2.6). Tukey's comparison 22

42 400 0) (..) c c..0 ro _ (J) c ro _ Rainfall (mm) Figure 2.5. Relationship between total invertebrate abundance collected from sweep-netting at each site and the corresponding total rainfall for the previous three months. 23

43 T CJ) 35 - ro = lag "'0 - CJ) (1) cc l late wet early dry late dry early wet Seasons Figure 2.6. Mean residuals of linear regression of log-body mass with log-svl for adult frillneck lizards. Closed circles are adult males, open circles are adult females, and numbers are sample sizes. Error bars are one standard error. Line through zero represents the best least squares fit. 24

44 of means test indicated that the early wet season sample mean was significantly lower than other three periods. Male body condition remained constant throughout both dry season periods. Female body condition also differed significantly among the four seasonal periods (Figure 2.6) (F3.110 = 3.14, P = 0.028). Tukey's comparison of means test indicated two groups of significantly different mean residuals, group one included the early wet and early dry seasons with relatively high residuals, while the second group included the late wet and late dry seasons with low residuals. Seasonal growth rate and change in body mass The seasonal growth rates of male and female lizards were small for individuals recaptured during the dry season (Figure 2.7, Table 2.3). Individuals recaptured within the wet season exhibited higher seasonal growth rates (Figure 2.7, Table 2.3). However, seasonal growth rates of large adult lizards (males SVL > 240 mm, females SVL > 210 mm) during the wet season were negligible (Figure 2.7). These individuals have either reached or are close to their maximum body size, and growth thereafter is reduced. There was no significant difference between the sexes in seasonal growth (Mann-Whitney U -test: z = 0.58, P = 0.599). Seasonal growth rates were significantly higher in the wet season than in the dry season (z = 3.05, P = 0.002). Males and females exhibited an overall negative change in body mass during the dry season, but there was a positive change in body mass by both sexes in the wet season (Table 2.3 ). Some males recorded a positive change in body mass in the dry season, whereas most females exhibited a negative change in body mass in the dry season (Figure 2.7). Wet season values of change in body mass were significantly higher than dry season 25

45 ' >. 8 rn "0 E E.6..._ <1) - <a '-..c.4-0 '- C).2 ro c:: 0 en roo.o <1) (/) (a) Males I e 0 O O 0 <D Snout to vent length (mm)... (b) Females '>. <a.4 "0 E E.3..._ <D - ro '-..c.2-0 '- C).1 ro c:: 0 en c roo.o o cc -c:: co <1) (/) Snout to vent length (mm) -..-' >- <a 1.0 "0 C) I.8 o 0'..._ en.6 C/) ro E.4 >- "C ').0 Q) Ol c:: -.2 ro..c (.) c 180 (c) Males ;:( -,.. o 0 c.,...:: 0 'XC C) 0,..., c-::; 6!...; 0 '-' -' :, " Q 280 Snout to vent length (mm)..- ' >- <a.0 "0 C) _ en.6 C/) ro E.4 >- " <D C) c:: -.2 (d) Females -<' --,..,...;,, " '-' :-.._, c <a..c 300 (.) Snout to vent length (mm) Figure 2.7. Seasonal growth rate (a,b) and percentage change in body mass (c,d) relative to initial body length. Closed circles are wet season values and open circles are dry season values. 26

46 Table 2.3. Mean and standard error of seasonal growth rate as measured by changes in SVL (mm day-1) and change in body mass (% g day-1) for male and female frillneck lizards in the wet and dry seasons. Numbers in parentheses are sample sizes. Dry season Wet season Males Growth rates ± ± (mm day-1) (36) (14) Body mass ± ± (% g day-1) (36) (14) Females Growth rates ± ± (mm day'1) (9) (10) Body mass ± ± (% g day-1) (9) (10) 27

47 values for both males and females using the mid-point SVL as the covariate, but this was marginal (ANCOVA: males, F1 46= 3.92, P = 0.047; females, Fu6 = 4.55, P = 0.049). Habitat use Two methods were used to acquire data relating to habitat use and these methods have important implications for the interpretation of these results. The ftrst method involved locating frillneck lizards from a moving vehicle, and the second method involved relocating frillneck lizards fitted with radiotransmitters. The first method of locating lizards may be biased in three ways, and the bias is related to the fact that most lizards were perched on vertical tree trunks: ( 1) different sizes of the tree trunks may alter the probability of sighting an individual; (2) agamid lizards tend to move to the opposite side of a tree trunk when approached or disturbed (Greer 1989), and this may influence the probability of sighting; and (3) lizards perched on vertical trunks may be involved in foraging, social interactions, and predation avoidance and/or detection (Stamps 1977 a, Shine 1990). Relocation of individuals using telemetry should not be influenced by these three factors. Therefore, the data set has been divided into two groups, data from lizards with radio transmitters (telemetered) and data from lizards without radio transmitters (nontelemetered). Preliminary analysis indicated no differences between the sexes in the frequency distribution of tree species used, therefore, males and females were combined in the subsequent analyses. Sightings of lizards located on the ground were not included in the statistical analysis. To determine if the data were biased by repeated observations, the freq uency distribution of trees species of five observations from each of six adult males 28

48 with transmitters, was compared to the single observations of tree species used by 30 individual adult males. Chi-square test revealed no difference in the frequency distribution of tree species for repeated and single observations (X 2 = 0.37, DF = 2, p = 0.832). There was no difference in the frequency distribution of tree spectes occupied by telemetered lizards between the dry and wet seasons (X2 = 8.87, DF = 7, P = 0.262). Telemetered lizards frequently occupied Eucalyptus tetrodonta, with highest use of this tree species occurring in the dry season (Table 2.4). During the wet season, telemetered lizards increased their use of the Sand Palm (Livistona humilis) and dead trees, with a higher proportion of lizards located on the ground. Compared to the random sample of tree species, telemetered lizards used a significantly diffe rent distribution of tree species in both dry and wet seasons (dry season, X2 = 63.52, DF = 7, P < ; wet season, X 2 = 26.87, DF = 7, P = ). The telemetered lizards used a much higher proportion of Eucalyptus tetrodonta than was available, and they under-used E. miniata (Table 2.4). The telemetered lizards generally used the other tree species as would be expected from a random selection. There was no difference in the frequency distribution of tree species used by nontelemetered lizards between the dry and wet seasons (X2 = 13.45, DF = 7, P = 0.062). The frequency distribution of tree species used by these lizards during the dry season closely reflected the availability of tree species (X 2 = 6.55, DF = 7, P = 0.477). However, during the wet season, they did not select trees according to availability (X2 = 44.55, DF = 7, P < ). Inspection of Table 2.4 reveals an under-usage of E. miniata, and a higher than random use of the Sand Palm, Livistona humilis, by non- 29

49 Table 2.4. The relative abundance (%) of tree species used by telernetered and nontelemetered fr illneck lizards during the wet and dry seasons, and the relative abundance (%) of tree species available in the field as determined by a random walk technique. Telernetered Non-telernetered Random Dry Wet Dry Wet Tree species season season season season (%) (%) (%) (%) (%) Eucalyptus tetrodonta E. porrecta E. miniata Erythrophleum chlorosrachys Terminal ia ferdinandiana Livisrona humilis Planchonia careya Buchanania obvata Petalostigma pubescens Xanthostemon paradoxus Eucalyptus tectifica E. bleeseri E. clavigera E. confertiflora Owenia vemicosa Acacia mimula Terminalia latipes Terminalia grandiflora Syzygium suborbiculare Planchonella pohlmaniana Melaleuca nervosa Cochlospermum fraseri Grevillea pteridifolia dead tree other tree species ground Total number of trees (n)

50 telemetered lizards during the wet season. There was a highly significant difference in the frequency distribution of tree species occupied by telemetered lizards compared to nontelemetered lizards in the dry season, but not in the wet season (dry season, x 2 = 44.88, DF = 7, P < ; wet season, X2 = 13.75, DF = 7, P = 0.056) (Table 2.4). The sampling bias described in the previous section was evident in the analysis of structural habitat use. Telemetered lizards were located on significantly larger trees than non-telemetered lizards (tree height, t = 11.06, DF = 737, P < ; trunk diameter, t = 14.69, DF = 737, P < ). Therefore, it was necessary to analyse the two groups separately. Comparing the seasonal use of trees by telemetered lizards revealed that they used bigger trees and were perched higher in the early and late dry seasons (Figures 2.8, 2.9, 2.10). Trunk diameter was the only variable that was significantly different over the four seasonal periods CANOVA: F3A41= 8.7 1, P < ). Tukey's comparison of means test indicated that the mean trunk diameter of the trees used in the early dry season was significantly larger than the other three periods. The height and trunk diameter of trees used by non-telemetered lizards remained constant throughout the four seasonal periods (Figures 2.8, 2.9, 2.10). However, the perch heights used by these lizards were significantly different CANOVA: F3.308= 9.09, P < ) over the four seasonal periods (Figure 2.10). Tukey's comparison of means test indicated that the mean of the early dry season period was significantly lower than the late dry and early wet season periods, but not significantly different from the late wet season period. 31

51 E _.....c 0) - 12 Q)..c Q) Q) '... c: co 9 Q) 22 r- L r late wet early dry late dry early wet Seasons Figure 2.8. The mean height of trees (m) used by frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are non-telemetered lizards, nwnbers represent sample sizes. Error bars are one standard error. 32

52 - 25 E (.) -....r:. 0> U) ro Q) L-.0 -ro 15 L- Q)... Q) E ro "'0 10 c: ro Q) :E 155 t-- --f--4---q L r ,-- late wet early dry late dry early wet Seasons Figure 2.9. The mean tnmk diameter (em) of trees used by frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are non-telemetered lizards, numbers represent sample sizes. Error bars are one standard error. 33

53 E..._... C).. Q)...c: (.) Q) c.. c ro Q) Q o A- --o ,----r--- late wet early dry late dry early wet Seasons Figure The mean perch height (m) of frillneck lizards in four seasonal periods. Closed circles are telemetered lizards, open circles are non-telemetered lizards, numbers represent sample sizes. Error bars are one standard error. 34

54 Seasonal activity index The number of lizards captured or sighted per kilometre during censussing changed substantially throughout the year (Figure 2.11). This index showed that the largest number of lizards were seen during the months of November through February, and the lowest number of lizards were seen from June through August. The number of lizards sighted or captured increased sharply between October and November (Figure 2.11), and this period coincides with the beginning of the frillneck lizard reproductive season (Shine and Lambeck 1989). The pooling of monthly data into four seasonal periods produced unequal variances between the four sample periods, therefore a Kruskal-Wallis one-way ANOVA was used to test for differences among seasons. There was a significant difference among the four seasons (z = 31.73, P < ). Inspection of the activity index means indicated that the two wet season periods were higher than the two dry season periods. 35

55 X Q).3 "'0 c.2 > -, u <(.1 J F M A M J J A S 0 N D Months Figure Monthly variation in number of lizards sighted or captured per kilometre driven during routine censussing of sites. Numbers are the number of censuses each month, error bars are one standard error. 36

56 DISCUSSION Frillneck lizards undergo seasonal changes with respect to diet, condition, growth, habitat use and activity. The dry season is characterised by a decrease in the activity of frillneck lizards, selection of large Eucalyptus trees when perched in the canopy, and a relatively small amount of food in their stomachs. This reduction in food is reflected in reduced growth rates and a small decrease in body mass, although general body condition remains relatively stable. The lizards continue to feed on a diverse array of arthropods despite overall low prey abundance. Termites, centipedes and ants are common prey items. The wet season is characterised by an increase in activity of the population, selection of shorter trees with small diameters, and an increased amount of food in their stomachs. This increase in activity coincides with four ecologically significant events: (1) an increase in ambient temperatures; (2) the onset of rainfall; (3) an increase in food availability; and (4) the onset of the reproductive period. The relative importance of these four factors is unclear in determining the behaviour of C. kingii, but they are probably all important. The previous dietary results of Shine and Lam beck ( 1989) are consistent with this study, particularly with respect to wet season data. Both studies found a diverse diet with a high relative abundance of isoptera, Iepidoptera and hymenoptera. One notable difference between the two studies of diet is the low proportion of centipedes reported by Shine and Lambeck ( 1989). One of the most interesting aspects of the prey taken by Chlamydosaurus is the high incidence of the harvester termite, Drepanotermes, in the diet throughout the year. 37

57 Drepanotermes is a widespread endemic genus (of at least 23 species) which inhabits arid and tropical regions of Australia. They differ from other harvester termites by foraging in the day and at night, whereas other harvester termites forage only at night (Watson and Perry 1981). Over 19 species of Drepanotermes occur in the arid regions of Western Australia, with between four and eight species occurring in the Top End of the Northern Territory (Watson 1982). The effect of seasonal conctitions in the wet-dry tropics on the foraging behaviour of harvester termites is unclear, because sweep-netting from this study ctid not sample termites. Foraging and reproduction in Drepanotermes are restricted to summer months in arid and temperate habitats (Watson 1974, Watson and Perry 1981, Park et a/. 1993). The presence of large quantities of Drepanotermes in the stomach contents of Ch/amydosaurus in the dry season, suggests these harvester termites are active throughout the dry season months in the study area. Dry grass is the preferred food item of Drepanotermes (Watson and Perry 1981), and dry grass (Sorghum spp.) is more abundant during the dry season in northern Australia. Termites are an important food source for many species of lizards in arid environments (Pianka 1986, Morton and James 1988, James 1991b, Abensperg-Traun 1994). The presence of termites in the diet of lizards has been related to foraging mode in arid habitats (Huey and Pianka 1981). Lizards that are primarily sedentary in locating prey or "sit-andwait" foragers (in arid habitats) tend to encounter and eat fairly mobile prey (e.g. ants, centipedes), whereas more actively foraging lizards consume less active prey (e.g. termites, caterpillars) (Huey and Pianka 1981). There is some difficulty in categorising invertebrates into these functional groups, but they have been generally accepted in previous research. 38

58 Data from lizards in tropical Brazilian forests support this general hypothesis, with widely foraging lizards consuming a greater proportion of termites (less mobile prey) than other less actively foraging lizards (Magnusson et al. 1985). Table 2.5 summarises studies of tropical lizards with respect to foraging mode and termite prey in tropical environments. The relative volume of prey were not provided in a number of studies, therefore only relative abundance is presented. The proportion of termites in the diets of six species of tropical "sit-and-wait" lizards (excluding Chlamydosaurus) in open and closed forests was neglible, whereas the proportion of ants was generally high. In contrast, the diets of active foraging lizards in tropical habitats included a substantial proportion of termite prey, and relatively fewer ant prey. This relationship supports the foraging strategy hypothesis of Huey and Pianka (1981 ). The saxicoline Tropidurus group displays a high numbers of both social insect prey groups. Comparisons of this group are problematic as they may use a range of foraging strategies, and they also consume some plant material. The presence of a large proportion of both termites and ants in the diet of Chlamydosaurus is clearly distinct Considering the almost exclusive "sit-and-wait" foraging strategy used by Chlamydosaurus (Shine and Lambeck 1989), they appear to be an exception to the generalisation (Huey and Pianka 1981). However the presence of other mobile prey (orthoptera and chilopoda) in the diet of Chlamydosaurus, broadly supports the pattern of a sedentary predator consuming mobile prey. Much of this hypothesis is based on the assumption that the spatial distribution of termites is unpredictable (Wilson and Clark 1977 as cited in Huey and Pianka 1981). A possible explanation for the relationship between frillneck lizards and harvester termites, is that Drepanotermes have a more even spatial distribution or higher density than is generally 39

59 Table 2.5. A summary of tropical lizard's fo raging strategy, habitat, and the relative abundance of ants and termites in their diets. Species Climate Foraging Habitat type Diurnal? Ants Termites Reference mode' (%) (%) Chlamydosaurus kingii wet-dry sw open forest yes This study C. kingii wet-dry sw open forest yes Shine and Lambeck 1989 Anolis auratus wet-dry sw open forest yes Magnusson et al A.opalinus wet-dry sw open forest yes Floyd and Jenssen 1983 A. cupreus wet-dry SW open forest yes Fleming and Hooker 1975 A. aeneus wet-dry SW closed forest yes Stamps et al A. oculatus wet sw closed forest yes Bullock et al Corytophanes cristatus wet sw closed forest yes 0 0 Andrews :t Plica plica wet-dry sw closed forest yes Vitt l991b Uranoscodon superciliosum wet-dry sw closed forest both Howland et al Tropidurus spp (4 species) wet-dry? open forest yes Vitt 1993 Kentropyx calcarata wet-dry WF open forest yes Vitt 1991a K. striatus wet-dry WF open forest yes Magnusson et al Ameiva ameiva wet-dry WF open forest yes Magnusson et al Cnemidophorus lemniscatus wet-dry WF open forest yes Magnusson et al C. deppii wet-dry WF coastal yes Vitt et al Mabuya bistriata wet-dry WF closed forest yes Vitt and Blackburn 1991 a. foraging mode either SW = sit-and-wait or WF = widely foraging.

60 considered to be the case for termites. The density and spatial distribution of Drepanotermes in tropical open forests are unknown. Density estimates of Drepanotermes in arid habitats may be as high as 200 mounds ha 1 (Watson and Perry 1981 ), although no estimates are available for Kapalga. A high density of termites would increase their predicability and availability to both "active" and "sit-and-wait" foragers. The above ground activity during day by harvester termites allows Ch/amydosaurus to access a large, relatively constant food resource. This may be an important factor in the diet of ChlamydosG;urus, especially during the dry season when other food sources are relatively low. This food supply may possibly be available to other "sit-and-wait" insectivorous lizards, as well as actively foraging insectivorous lizards. A reduction in the volume of food taken during dry conditions is a commonally observed pattern for lizards inhabiting arid and temperate seasonal environments, and it is a direct response to decreased food availability (Ballinger and Ballinger 1979, Rose 1981, Stamps and Tanaka 1981, James 1991b). Unfortunately, information on the total volume of stomach contents is available from only two dietary studies of tropical lizards listed in Table 2.5. Bullock et a/. ( 1993) reported no seasonal change in number of items or total volume of stomach contents for Anolis oculatus, despite large differences in food availability. Another tropical iguanid, Anolis oplinus, showed no change in the volume of stomach contents between wet and dry seasons (Floyd and Jenssen 1983). Reduced growth during periods of low food and water availability has been well documented in lizards (Stamps 1977b, Dunham 1978, Ballinger and Congdon 1980, Stamps and Tanaka 1981, Andrews 1982), and Ch/amydosaurus follows this pattern. Feeding continues throughout the dry season, but the volume of food taken is apparently not sufficient to support rapid 41

61 growth. The reduced field metabolic rate during the dry season (Christian and Green 1994 ), suggests that the lizards are generally conserving energy. A small decrease in body mass (mean = 1.35% per month) of the lizards reflects this. Male frillneck lizards were able to maintain a relatively constant level of body condition throughout the dry season. However, the body condition of males decreased during the early wet season, which is the beginning of the reproductive season for this species. Even though there is a marked reduction in food availability during the dry season, a decrease in activity levels and the reduction in field metabolic rate (Christian and Green 1994) enables males to minimise the loss of body condition during this period of limited food resources. This seasonal pattern in body condition is consistent with males of other tropical lizard species (Fleming and Hooker 1975, Floyd and Jenssen 1983, Howland et a!. 1990). Female Chlamydosaurus exhibit a more variable pattern of body condition across the seasons, with low body condition in the late wet season and the late dry season. Expenditure of energy for reproduction may explain the drop in body condition in the late wet season because this corresponds with the period of oviposition. The late dry season is the period with the lowest availability of prey (Table 2.2, Churchill 1994 ), and is the season prior to the reproductive season. However, it is not clear why female body condition is lowest at this time but males is lowest in the early wet season (three months later), especially with a similar volume of food in their stomachs. This may be related to the energy used by males in territorial defense and mate acquisition at this time, whereas females may spend this time feeding. 42

62 Trees are an important component of the habitat used by C. kingii. Over 95% of all lizards in this study were located in trees, and this agrees with previous observations (Shine and Lambeck 1989). The use of structural habitat by frillneck lizards changes with rainfall, food availability, activity and reproductive season. The way habitat data were collected should be considered for the interpretation of the data. If only one of these sampling methods was used (telemetry or vehicle) then only one conclusion would be possible. If only non-telemetered lizards were used to study habitat use then the conclusion would be that frillneck lizards randomly selected the tree species available, and the size of trees selected remains constant throughout the year. If only telemetered lizards were used, then the conclusion would have been that frillneck lizards show a strong preference for Eucalyptus tetrodonta and a seasonal change in size of trees selected (trunk diameter and tree height). Therefore, it is important to consider the behaviour of lizards with respect to the reasons they perch on vertical tree trunks. The perching by lizards on vertical tree trunks has been related to feeding and social interaction (Scott et a/. 1976, Stamps 1977a, Shine 1990). It is suggested here that frillneck lizards that were seen from the vehicle were engaged in either foraging or social behaviour, and that this behaviour corresponds to the use of relatively small trees at all times of the year. This behaviour on low perches takes place on tree species in proportion to their abundance. Frillneck lizards fitted with telemetry devices are generally perched higher in trees, and their selection of tree size changes significantly throughout the year. This seasonal change is related to a general decrease in activity, and the lizards apparently use the large trees as a dry season refuge (Christian and Green 1994). This is also reflected in the behaviour of some telemetered lizards that 43

63 remained in the same tree for prolonged periods (up to three months). During the wet season, however, frillneck lizards regularly changed trees (Shine and Lambeck 1989, pers. obs.). Summary In summary, frillneck lizards show considerable seasonal change in their behaviour and ecological relationships in the wet-dry tropical environment. No specific environmental variable was identified as being responsible for these changes because of the interdependence of these environmental variables. Frillneck lizards continue to feed on a diverse range of invertebrates during low food availability in the dry season. The presence of termites in the diet of frillneck lizards indicates a fundamental difference from other tropical "sit-and-wait" lizards. It also offer a constant food supply in the dry season. The volume of stomach contents is reduced in the dry season, which in turn reduces the growth of frillneck lizards. Lower levels of activity in the dry season coincide with lizards selecting large Eucalyptus trees, in which they become very inconspicuous until the beginning of the reproductive season in October. Male lizards held their body condition during the dry season to a greater extent than females. Differences in reproductive roles of the two sexes may be responsible for seasonal differences in body condition. 44

64 Chapter 3 The short-term and longer-term effects of annual fire on the behaviour, diet, growth, and habitat use of the frillneck lizard, Chlamydosaurus kingii. INTRODUCTION Fire is an important environmental factor for many vertebrate communities within Australia (e.g. Newsome et al. 1975, Fox 1982, Braithwaite 1987, Woinarski 1990) and elsewhere (e.g. Gillon 1983, Clark and Kaufman 1990, Mushinsky 1992). Fire creates a mosaic of habitats within environments, and this may be crucial in the maintenance of species diversity. Previous studies of lizards and fire have documented changes in species abundance and community composition related to the seral changes in the habitat structure after fire (Lee 1974, Lilywhite and North 1974, Barbault 1977 as cited in Gillon 1983, Simovich 1979, Fyfe 1980, Means and Campbell 1981, Patterson 1984, Caughley 1985, Mushinsky 1985, Bamford 1986, Braithwaite 1987, Woinarski 1989, Lunney et at. 1991, Masters 1991, Mushinsky 1992, Trainor and Woinarski 1994). Changes to the habitat structure may alter the relationship between the lizards and their thermal environment, food resources and predators. In contrast, relatively few studies have investigated the short or longer-term effect of fire on the ecology of individual lizards or a single species of lizard (Kahn 1960, Lilywhite and North 1974, Bamford 1986, Mushinsky 1992). The response to fire of individual species shows wide variation in this literature, which possibly reflects the diverse life histories of 45

65 these lizard species. Additional information is needed about the relationship of individual lizard species and flre to make management and conservation decisions. The majority of research on lizards and fire has been done in temperate and arid regions where fire is generally of low frequency. Important differences in the effects of frre on lizard communities exists between tropical and temperate environments based on the differences in the frequency of fire (Braithwaite 1987). Habitat succession after fire is less important to lizard communities in tropical savannas because the higher frequency of fr re does not allow habitat succession to occur. Rather, it is the time of year and related intensity of the fire that are important in determining composition and abundance in lizard communities (Braithwaite 1987). Trainor and Woinarski (1994) support this, although they suggest that moisture availability is a more dominant factor in determining the composition of lizard communities in tropical savannas. Fires occur on an annual or biennial basis in open forest and woodlands of northern Australia (Braithwaite and Estbergs 1985). There is evidence that, subsequent to settlement of Europeans in northern Australia, the fire regime has changed from a relatively high frequency of low intensity burns early in the dry season to one of more intense burns later in the dry season (Haynes 1985, Braithwaite and Estbergs 1985, Press 1988). This alteration in timing and intensity of frres in northern Australia increases the need for ecologists and land managers to be able to understand the effects of different fire regimes in this environment. The distribution of Ch/amydosaurus kingii extends throughout open forests and woodlands 46

66 of northern Australia (Cogger 1992). The general ecology of this species was reported by Shine and Lam beck ( 1989). During the dry season frillneck lizards decrease their field metabolic rate, water turnover and body temperatures (Christian and Green 1994, Christian and Bedford 1995). During the dry season C. kingii have reduced activity and feeding rates, and they select larger Eucalyptus trees than during other seasons (Chapter 2: Diet and Seasonal activity). This period of relative inactivity coincides with the majority of fues in the savannas of northern Australia. This chapter examines two aspects of the effect of fire on the ecology of Chlamydosaurus kingii: ( 1) the short-term effect of two different fire treatments on mortality, behaviour, diet and post-fire habitat selection; and (2) the longer-term effects of three different fire regimes on habitat selection, diet, body condition and seasonal growth. METHODS Sites For a full description of the site and climate see Chapter 2 (Methods: Site). Fire treatments Prescribed fues were lit by CSIRO's Division of Wildlife and Ecology at Kapalga Research Station (Kakadu National Park), as part of a landscape-scale, replicated fire experiment. Three different fire regimes were used for this experiment that represent fue regimes experienced in open forests and woodlands in northern Australia. The fire treatments were: 47

67 Early dry season fires. These fires are lit annually at the beginning of the dry season (May/June), approximately three months after cessation of monsoonal rains. They are characterised by low intensity, low flame height (1-2 m), and a slow rate of spread (< 0.5 m sec 1). They invariably leave substantial areas (30-70lk) of unburnt ground vegetation, and seldom scorch the upper tree canopy. The fire treatment had been applied annually to one study area (Early fire, Figure 3.1) since Late d1y season fires. These fires are lit annually at the end of the dry season (September), approximately six months after cessation of monsoonal rains. They are characterised by high intensity, high flame height (2-4 m), and a faster rate of spread (> 1 m sec 1). All ground vegetation is usually consumed during the fire, and a large proportion of the tree canopy (> 80lk) is scorched. The increased intensity of these fires, compared to the early dry season flies, is due to higher fuel loads plus increased wind speeds, and lower relative humidity. The treatment had been applied annually to one study site (Late fire, Figure 3.1) since Unbumt. The treatment involved the total exclusion of fire, which had been applied to one study area (Unburnt, Figure 3.1) since The large amount of work and time involved in the collection of data prevented replication of treatments. Also, the large area of sites ( ha) required to sample a sufficient number of lizards in each flie treatment made replication unrealistic. 48

68 - Floodplain N 0 5 km Figure Location of three penn anent fire treatment sites within Kapalga Research Station. The shaded areas represent each site 49

69 Sampling Frillneck lizards were monitored using both telemetry and mark-recapture methods, which are described in Chapter 2 (Methods: Sampling). Both of these sampling methods provided data on behaviour, habitat use, food availability, diet and body condition of the three populations. The intensity of individual fires (kw m 1) were measured by the CSIRO. Short-term effe cts of fire Individual lizards (fitted with radio transmitters) were located in trees immediately before a prescribed fire was ignited. The trees were marked, and the tree height (m), trunk diameter at breast height (em), and perch height (m) of the lizard was recorded. Lizards were relocated immediately after the fire front had passed in order to determine mortality or subsequent movements. Changes in the location of lizards during a fue were recorded. Stomach contents of lizards were obtained by stomach flushing large lizards (SVL > 150 mm) within two hours of capture (Chapter 2: Methods). This was done one week before and one week after fires, allowing a minimum of two days after fires to minimise the overlap with pre-fue stomach contents. Lizards that were stomach flushed before fues were not re-sampled after fires. Data concerning post-fire habitat selection by lizards were collected during the fust year of this study ( 1991 ), prior to the establishment of permanent study sites. Telemetry was used to observe lizards over a large area, to determine whether they preferred or avoided burnt or unburnt habitat. Lizards were captured at the boundaries of fire treatment sites before prescribed fires were lit, so that after fires there was both burnt and unburnt habitat 50

70 available nearby. They were monitored immediately before fires and at regular intervals up to two weeks after fires. Longer-tenn effe cts of fire An important objective of this study was to determine whether the use of the structural habitat (occupancy of trees) by frillneck lizards was influenced by a particular factor (or combination of factors) of the structural habitat, and whether different fire regimes influenced this relationship. This was done by comparing the structural characteristics of the trees (and surrounding ground vegetation) occupied by frillneck lizards against a randomly selected sample of trees and surrounding vegetation within the same habitat. Lizards were assumed to be absent from this randomly selected sample. Data on structural habitat use by frillneck lizards were collected specifically during dry season months (prior to annual fires in early dry and late dry season fire sites) for two reasons: ( 1) the habitat use of telemetered lizards changes between wet and dry seasons with regard to height and diameter of trees selected (Chapter 2: Habitat use); and (2) the habitat structure undergoes dramatic change after fire, and again during the wet season. Therefore, it was necessary to collect data during the dry season (prior to fire treatments) to minimise the influence of these factors before comparisons of the different fire regimes could be made. Data from trees occupied by frillneck lizards fitted with telemetry devices were used in the analysis of structural habitat because there was as a sampling bias in the mark-recapture method (Chapter 2: Habitat use). Tl}e random sample of trees was selected using a random 51

71 walk method (Goldsmith and Harrison 1976). A minimum distance of 30 m was set between randomly selected trees to increase the independence of each sample. The following structural habitat variables of trees were recorded: Tree height: Tree height (m) was measured using a clinometer. Diameter: The diameter (em) of each tree trunk was measured at breast height using a calibrated measuring tape. For trees with more than one trunk, a mean of all trunks was taken. Tree canopy: The amount of canopy cover of individual trees was recorded using a five -point ordinal scale (1, l -5%; 2, 6-25%; 3, 26-50%; 4, 51-75%; 5, %), that was estimated visually. The following structural habitat variables were collected within a three metre radius from each tree trunk: Ground vegetation density: The ground vegetation density was recorded in circular quadrats (0.4 m 2 in area). Within each quadrat, the total number of contacts of ground vegetation against a piece of string rotated horizontally were measured at four heights (25, 50, 100, 150 em). Three quadrats were collected per tree, and an average of number of hits at each height was taken. Means were converted to a density (hits m 2 ) at each height. Litter cover: The litter cover was measured by the total number of leaves pierced by a steel pin in 10 sample points per tree. Bare ground: The percentage of bare ground was estimated visually within a three metre radius of the tree. 52

72 Data on the selection of tree species by frillneck lizards were collected separately from structural habitat data. Tree species was recorded at all captures and sightings of lizards throughout both wet and dry seasons at each site. Sampling of tree species available to lizards was collected at each site using a random walk method (Goldsmith and Harrison 1976). This is described in Chapter 2 (Methods: Sampling). Stomach samples were collected from large lizards (SVL > 150 mm) captured within each site. Stomach contents of lizards were collected by stomach flushing and later sorted and total volume measured (Chapter 2: Methods). Regular sampling was done from April 1992 to April Analyses Short-term effects of fires Data from the three fires per fire treatment were combined for analyses. Student's t-test was used to analyse changes before and after fires with respect to total volume, number of items in stomach contents, and the height of trees used before and after fire. Difference between early and late fire behavioural responses during fire were analysed using Kolmogorov-Smirnov two sample test. Longer-term effects of fire Six different measures were used to analyse the short and longer term effects of fire on the stomach contents of frillneck lizards: (l) the total number of items; (2) the total volume (rnl); (3) the relative abundance (%) of each prey taxa; (4) the relative volume (%) of each prey taxa; (5) the occurrence (%) of one or more prey taxa in stomach samples; and (6) 53

73 the number of prey taxa and size classes present in each stomach sample. Stomach samples from the dry and wet seasons were analysed separately for longer-term variation among fire regimes. A single factor analysis of variance (ANOV A) was done to test for differences in total volume among sites, for wet and dry seasons separately. A non-parametric ANOV A (Kruskal-Wallis) was used to test for differences in the frequency distribution of prey size classes and prey taxa among fire treatments. Logistic regression analysis was used to determine the factors that affected the occupancy of trees by frillneck lizards during the dry season. This technique uses a logistic probability curve through binary data (occupied trees versus unoccupied trees). The logistic curve has the form P(y= l) = EXP(G)/(1 + EXP(G)] where P(y= 1) is the probability that an individual will occupy a certain tree; G is analogous to a standard regression equation and is where A is a constant; xil xin = the first and ith independent terms in the model; Bl, BN = the coefficients for the first and nth terms of the model; and i = 1, 2,.....N (individual trees). 54

74 The terms of the logistic model were estimated using maximum likelihood, termed 'scaled deviance' in this analysis. Scaled deviance is analogous to the F ratio in standard regression analysis, but has an approximately chi-square distribution. Degrees of freedom are equal to the difference in the number of degrees of freedom from an original null model, when each term is added to the model (McCullagh and Neider 1983 ). Terms and interactions were added to the model using a forward selection method where all single terms are added to an original null model individually, with the term producing the greatest significant decrease in scaled deviance being selected into the model. This process was continued with each set of remaining independent terms until no significant decrease was achieved by any of the remaining terms. Quadratic terms (XixXi) were tested with each independent term to determine the presence and significance of non-linear relationships within the model. Interaction terms (denoted by multiplication sign between two terms) were then tested in the same manner, after the analysis of single terms. A final model of occupancy of a tree was evaluated using two methods: (1) testing the significance of each of the accepted terms, by deleting each term individually from the full model, with non significant terms being rejected from the model; and (2) examination of the relationship between coefficient values and standard errors of accepted terms to test the significance of approximate t values (Nicholls 1985 ). Selectivity of tree species by frillneck lizards was calculated using Strauss's (1979) electivity index. This index is calculated for each tree species: L = ri - Pi where ri is the relative abundance of each tree species used by frillneck lizards in both wet 55

75 and dry seasons and Pi is the relative abundance of the same trees species in each site as determined using the random walk method. Differences among fire treatment sites in lizards' body condition was determined by ANOVA of mean residuals from a linear regression of log-body mass with log-svl (James 1991 a). Samples were pooled into wet and dry seasons. All means are presented with one standard error unless stated otherwise. RESULTS Short-term effects of fire A total of 17 frillneck lizards were monitored using telemetry during three early dry season ftres. The intensity of the fires at ground level were approximately 5000, 1000 and 800 kw m ' for 1991, 1992 and 1993 respectively (Williams et a/. unpublished data). All lizards were relocated after the fires, and no direct mortality was.recorded during early dry season fires. A possible indirect mortality may have occurred after one animal was scorched during a ftre, losing at least 12 toes and therefore being unable to climb. This would reduce its chances of survival dramatically, but the fate of the animal was not determined. During early dry season fires lizards tended to remain in the trees they occupied before fire (59%), with a smaller proportion changing trees (35%), and one lizard (6%) sheltered in a disused hollow termite mound. The height of trees used by lizards immediately before fire was not significantly different (paired t-test: t = 1.06, OF = 5, P = 0.34) compared to trees to which 56

76 lizards moved during a fire. A total of 24 frillneck lizards were monitored using telemetry during three late dry season fires. The intensity of the fires at ground level were approximately 10,000, 9000 and 8000 kw m for 1991, 1992 and 1993 respectively (Williams et a!., unpublished data). All lizards were relocated after the fires. Six were killed as a direct result of the fires, and another lizard died two weeks after a fire because of the severe scorching it received while perched in the tree canopy. This represents 29% mortality during late dry season fires. Compared to the early dry season fires, lizards exposed to high intensity late dry season fires were less inclined to remain in the same tree during a fire (16%), although a similar proportion changed trees (25%), and most notably seven lizards (30%) found shelter in hollow termite mounds on the ground. Those lizards that changed trees during a late dry season fr re selected significantly taller trees (paired t-test: t = 3.79, DF = 11, P = 0.013) compared to the trees they occupied before the fire. The behavioural responses of frillneck lizards to late dry season fires differed significantly from their responses to early dry season fires (Kolmogrov-Smirnov two-sample test: D = 0.53, P = 0.008). The stomach contents of frillneck lizards contained a mean of 9.5 ± 4.23 (n = 10) items during the week before early dry season fires, and this increased to ± 4.03 (n = 15) items one week after fr res, but this increase was not significant (unpaired t-test: t = 1.01, DF = 24, P = ). The mean total volume of stomach contents increased from 2.08 ± 0.94 ml before frre to 3.10 ± 0.67 ml after fire, but this increase was not significant (t = 0.91, DF = 24. P = 0.371). Figure 3.2 illustrates the frequency distribution for the number of insect orders and number of prey size classes present in each stomach, one week before 57

77 and one week after fire. Stomach contents after fire tended to contain a greater number of prey orders and prey size classes. However, only the number of size classes was significantly different for before and after fire comparisons (Mann-Whitney U-test: z = p = ). Hymenoptera was the most commonly taken prey taxon before and after the early dry season fires (Figure 3.3). The relative abundance of hymenoptera decreased after fires, whereas orthoptera, hemiptera and coleoptera showed a small increase in relative abundance (Figure 3.3). Although hymenoptera were numerically abundant, the relative volume in the stomach contents of lizards was small. The relative volume of blattodea, chilopoda and aranea decreased after fires, but the relative volume of orthoptera, hemiptera, coleoptera and Iepidoptera increased (Figure 3.3). The distribution of the relative abundance of prey taxa did not change significantly after fires (Kolmogorov-Smimov two sample test: D = 0.08, P = 0.805). However, the distribution of the relative volume of prey taxa was significantly different after fires (D = 0.51, P < ). A total of 53 stomach contents were collected for the analyses of the short-term effects of late dry season fire, and they contained a mean of ± (n = 20) items before fires, and this decreased marginally to ± (n = 33) after fires. This decrease was not significantly different (unpaired t-test: t = 0.07, DF = 52, P = 0.93). The mean 58

78 5 c.. E C0 4 (J)..c () C03 E 0.. (J)2-0 1 E ::J zo Number of prey size classes (J) <1>8 c.. E co (J) 6..c () co E (J) - 0 <1> 2..0 E ::J Z o II Number of invertebrate orders Figure 3.2. Frequency distribution of the number of invertebrate orders and prey size classes in the stomach contents of frillneck lizards, one week before and one week after early dry season fire. Open bars represent before fire and closed bars represent after fire. 59

79 ... - :::R e._. 80 Q) u c: ro 60 "'0 c ::::J.c 40 ro Q) > ro Q) 0:: 0 ci..!!! [ e ci ci c ui c -.. c :E 0 Cll Cll 0....J iii "0.s:: X X :E C(... Cll.s::. 0 Invertebrate orders - :::R 0-60 Q) 45 E ::::J 0 30 > Q) > ro Q) 0:: 15 0 ci e Cll 0 X...: 'ci - ci 0 0 Cll..J [ X 7;; c ui c c iii "0.s::.... :E 0 c. C( :E 0... Cll.s::. 0 Invertebrate orders Figure 3.3. Relative abundance and relative volume of prey taxa in stomach contents, one week before and one week after early dry season fire. Open bars represent before fire and closed bars represent after fire. Abbreviations for prey taxa are: I so.- Isoptera; Orth. - Orthoptera; Hem. - Hemiptera; Col. - Coleoptera; Dip. - Diptera; Lep. - Lepidoptera; Hym. -Hymenoptera; Blat. - Blattodea; Mant. - Mantodea; Odon. - Odonata; Phas. - Phasmotodea; Aran. - Aranea; Chi/. - Chilopoda. 60

80 total volume increased from 3.28 ± ml before fires, to 6.01 ± 0.84 ml after fires, and this increase was significant (t = 2.08, OF = 52, P = 0.042). The number of invertebrate orders and the number of size classes present in stomach contents increased after fue (Figure 3.4). Stomach contents collected after late dry season fires contain a significantly greater number of prey size classes (Mann-Whitney U-test: z = 2.85, P = 0.04). This result helps explains how the total volume differed significantly between the two periods, but the number of items did not. Lizards consume a greater range of prey size classes after fires, which increases the total volume of stomach contents. There was also a significant increase in the number of prey taxa in stomach samples after late dry season fires (Figure 3.4, z = 3.34, p < 0.00 l ). Isoptera were the most commonly taken prey taxa before late dry season fues, accounting for approximately 65% of relative abundance, and 30% of relative volume (Figure 3.5). This was reduced considerably after fires, when isoptera accounted for 23% of relative abundance, and only 6.5% of relative volume. Hymenoptera, however, increased in relative abundance and relative volume after the late dry season fires (Figure 3.5). Orthoptera, hemiptera, coleoptera, and chilopoda also increased in relative abundance and relative volume immediately following late dry season fires (Figure 3.5). Both the distribution of the relative volume and relative abundance of prey taxa were significantly different after late dry season fires (Kolmogorov-Smirnov two sample test: relative volume, D = 0.22, P = 0.022; relative abundance, D = 0.41, P < ). 61

81 en Q) 14 a. 12 en..c 10 (.) m E en 6.._ 0 4 Q).c E 2 :::s z ' Number of prey size classes en Q) 10 a. E m en 8..c (.) E 6.9 en.._ 4 0 Q).c 2 E :::s z 0 n 0,- I I 7 8 Number of invertebrate orders Figure 3.4. Frequency distribution of the number of invertebrate orders and prey size classes in the stomach contents of frillneck lizards, one week before and one week after late dry season fire. Open bars represent before fire and closed bars represent after fire. 62

82 :::R _ Q) g 60 ro "'C c: :l ro Q) > 20 ro Q) 0:: 0 0.!!... E 0 ci. ci. E i;j c c IIi c :E Ql Ql 0 Ql 0.s::. 0 (J..1 >- iii "C.s::.... (J :z: :z: 0 c. 5 Invertebrate orders 60 - :::R 0.._ Q)' E :l 0 > Q) > ro Q) 0:: !!... E 0 ci. ci. E i;j c c IIi c :E Ql Ql 0 Ql 0.s::. 0 (J..1 >- iii "C.s::.... (J :z: :z: 0 c. 5 Invertebrate orders Figure 3.5. Relative abundance and relative volume of prey taxa in stomach contents, one week before and one week after late dry season fire. Open bars represent before fire and closed bars represent after fire. Abbreviations of invertebrate groups follow Figure

83 The data presented in Table 3.1 summarises the response by frillneck lizards for post-fire habitat selection two weeks after early and late dry season fires. Frillneck lizards either remained in freshly burnt areas or moved into burnt areas from the adjacent unburnt areas after early dry season fires. Only one individual (of 21 observed) left a recently burnt area, while two remained in the unburnt habitat next to a burnt area. The movements following late dry season fires are more variable, with some lizards leaving burnt areas, and some moving into the freshly burnt areas. However, most of the areas adjacent to the late fire plots had been burnt three months previously for early dry season prescribed fires, so there was less distinction between areas because the individuals outside late fire areas were already in burnt habitat. There was a significant difference in the distribution of responses to the two fue treatments (Kolmogorov-Smirnov two sample test: D = 0.32, P < ). Longer-term effects of fire Data on habitat structure were collected from 177 trees occupied by frillneck lizards (with radio transmitters) during dry season months. Data from a further 104 randomly selected trees (unoccupied) were collected concurrently. Table 3.2 summarises the data for each of the structural habitat variables for both occupied and unoccupied trees at each site. Logistic regression analysis identified four structural habitat variables that influenced the occupancy of trees by frillneck lizards during the dry season (Table 3.3). The amount of tree canopy cover, and the density of the ground vegetation at (25 em above the ground) were the two most significant terms present in the regression model. Both diameter and height of trees were significant terms, although they were only a minor influence on the 64

84 Table 3.1. Post-flre selection of habitat by telemetered frillneck lizards. All lizards were monitored adjacent to burnt or unburnt habitat after the early and late dry season fires during Remain in Move out of Remain in Move into Total Fire burnt area burnt area unburnt area burnt area treatment No. % No. % No. % No. % No. Early dry season fire Late dry season fue Total

85 Table Means ± one standard error of habitat structure variables used for the logistic regression model. Data describing habitat structure of open forests were collected from each fire treatment site. Occupied trees were known to contain frillneck lizards, and unoccupied, randomly selected trees were assumed not to contain frillneck lizards. Variable Unburnt Early fr re Late fire occupied unoccupied occupied unoccupied occupied unoccupied Tree height (m) 12.2 ± ± l.o 13.7 ± ± ± l.o 16.0 ± 1.3 Trunk diameter 17.3 ± ± ± ± ± ± 2.0 (em) Liner cover 8.5 ± ± ± ± ± ± 0.3 Bare ground (%) 1.8 ± ± ± ± ± ± 2.6 Tree canopy 2.7 ± ± ± ± ± ± 0.2 Ground 26.5 ± ± ± ± ± ± 3.6 vegetation density (25 em) Ground 9.3 ± ± ± ± ± ± 2.1 vegetation density (50 em) Ground 1.7 ± ± ± ± ± ± 0.5 vegetation density (100 em) Ground 0.4 ± ± ± ± ± ± 0.1 vegetation density (150 em) No. trees (n)

86 Table 3.3. Logistic regression model for occupancy of trees by telemetered frillneck lizards during the dry sea_son. x2 values refer to difference in scaled deviance from total deviance of the null model The % total deviance is the proportion of scaled deviance from the total deviance of the null model Treatment (site) factors are represented in the model by: (1) early dry season site; (2) late dry season site; and (3) unburnt site. Term Coefficient SE x2 DF p % total deviance n = 282 Constant (1) Tree canopy (Tree canopy) < Vegetation density (25cm) < Trunk di ameter (diameter) < Tree height (Tree height) < Tree canopy x Treatment < (2) n.s. (3) <0.01 Vegetation density ( < em) X Treatment (2) n.s. (3) <

87 overall model. Figure 3.6 illustrates the relationship between the probability of occupancy and the fined values of these four variables. The probability of occupancy of a tree by C. kingii was highest in trees with a canopy cover of 25-50% in the unburnt site. The probability of occupancy of trees by C. kingii in both the early and late fire sites was highest in trees with 50-75% canopy cover. Importantly, trees with a very light canopy in the unbumt site had a high probability (P = 0.5) of occupancy by lizards, whereas trees with a light canopy from both the burnt sites showed a low (P = 0.1) probability of occupancy by lizards (Figure 3.6). The similarity of the relationship between both burnt sites should be noted considering the large differences in the intensity of the early and late dry season fires. The probability of occupancy of trees increased with decreasing density of ground vegetation (Figure 3.6) at all sites. However, this relationship was stronger in both burnt sites, in which the probability of occupancy is small (P < 0.5) for trees with a medium density (> 75 stems per m2) layer of grass at 25 em above the ground. The probability of occupancy by lizards is high (P > 0.5) for trees in the unburnt site that have a moderately dense grass layer (> 165 stems per m 2 ). The relationship between the probability of occupancy of trees and the variables tree height and trunk diameter were similar for all three sites. A complex relationship appears to exist between these two variables. Inspection of probability curves for these variables (Figure 3.6) shows no clear pattern, except that the probability of occupancy is relatively high (P > 0.5) for all fitted values from either variable. 68

88 .9 (Q) 1.0 > > (.).8 (.) c c co :J ::l (.) (.) (.) (.) ro.8 (\v)... \ >.2; \..0..c..c.2 ro co.3 e.2 0 a \... a.. ' c ro.8 0. ::l (.) (.) OS 120 Canopy cover Vegetation density (stems per m2 ) 1.0 ) 1.0 Cd) >. (.) c ro.8 0. ::::s.6 (.) Z'.4.?: co I- 0 a (.)..c ro..c.2 a so I Tree diameter (em) Tree height (m) Figure 3.6. The effect of significant variables in a logistic regression model on the occupancy of trees by frillneck lizard during the dry season. The graphs show changes in estimated probability of occurrence of frillneck lizards with changes in (a) tree canopy; (b) the ground vegetation density at 25 em; (c) trunk diameter; and (d) tree height. Eor each relationship the values of other variables in model are fixed at the mean value. For graphs (a) and (b), solid lines represent the unburnt site, dashed lines are the early dry season fire site and dotted lines are the late dry season fire site. 69

89 There were no significant differences among the fire treatment sites with respect to the selectivity indices of tree species during either the wet or dry seasons (Kruskal-Wallis: dry season, z = 0.42, P = 0.809; wet season, z = 0.55, P = 0.774). The selection of Eucalyptus tetrodonta was consistently high in all three sites (wet and dry season), except in the unburnt site during the wet season (Table 3.4). Selectivity of E. tetrodonta in the unburnt site was reduced in the wet season, and this was offset by an increase in selection of the sand palm, Livistona humilis. A total of 226 stomach samples were collected from all three sites during dry and wet seasons, including 78 stomach samples used in the analysis of the short-term effects of fire. The total volume and the number of items in the dry season stomach samples were not significantly diffe rent among the three treatments (ANOVA: total volume, F = 0.37, P = 0.699; number of items, F2.J.H = 1.58, P = 0.226). The distribution of the number of invertebrate orders and the number of prey size classes in dry season stomach samples were also not significantly different among the three fire treatment sites (Kruskal-Wallis: number of orders, z = 4.18, P = 0. 12; number of size classes, z = 3.35, P = 0.12) (Figure 3.7 and Figure 3.8). Wet season stomach samples also showed a high degree of similarity among the three treatments. The total volume and number of items in the stomach contents were not significantly different among the sites (ANOVA: total volume, F2 81 = 0.71, P = 0.499; number of items, F2.81 = 0.47, P = 0.630). The distribution of the number of prey size classes in wet season stomach samples (Figure 3.8) did not differ significantly among the three treatments (z = 5.23, P = ). There was, however, a significant difference in 70

90 Table 3.4. Selectivity indices for tree species used by frillneck lizards within each site during the wet and dry seasons. Unburnt Early fire Late fire Tree species Dry Wet Dry Wet Dry Wet Eucalyptus tetrodorua E. porrecta E. miniara Erythrophleum chlorostachys Te rminalia ferdirumdiana Livistona humilis Dead" Other' a. all dead trees m sample regardless of species b. species with low relative abundance ( < 0.05) were combined into this group 71

91 en Q) 40 c. E ro en (.) ro E 0 +J (/) Dry season Number of prey size classes Dry season en 40 Q) c. 30 en 20 E 0 U Number of invertebrate orders Figure 3.7. Relative abundance of the number of prey size classes and invertebrate orders present in the stomach contents of frillneck lizards for each fire regime in the dry season. Open bars represent unbumt treatment, hatched bars rising right are early fire treatment and hatched bars rising left are the late fue treatment. 72

92 Cl) <1) a. E ro Cl). (.) ro E 0 CJ) 40 Wet season Number of prey size classes Cl) <1) a. E ro Cl). (.) ro E 0 CJ) Wet season Number of invertebrate orders Figure Relative abundance of the number of prey size classes and invertebrate orders in the stomach contents of frillneck lizards for each fire regime in the wet season. Open bars represent unburnt, hatched bars rising right are early fire treatment and hatched bars rising left are the late fire treatment. 73

93 the number of invertebrate orders (Figure 3.7) present in stomach samples among the sites (z = 9.57, P = 0.022). Inspection of Figure 3. 7 indicates that the stomach samples from the early fire site have a greater proportion of six or more invertebrates orders present. The taxonomic composition was broadly similar among the three fire treatments during the dry season (Table 3.5). Isoptera (termites), hymenoptera (ants) and chilopoda (centipedes) were the most common prey taxa at all sites. Stomach samples collected in the unburnt site recorded a higher occurrence of termites than the early and late fire stomach samples. The relative volume of termites was also high in stomach samples from the unburnt site, compared to the samples from the early and late dry season fire sites. Ants formed a large proportion of late fire site stomach samples in terms of relative abundance, compared to the other two sites. Orthoptera (grasshoppers) were frequently present in late fire site stomach samples, and constituted most of the relative volume from this site. Wet season taxonomic composition of stomach samples showed some variation among the three fi re treatments (Table 3.6). Termites, caterpillars and ants were the most commonally taken prey items at all sites. Termites frequently occurred in stomach samples at all sites, and the relative volume of termites was again much higher in the unbumt site. Termites were regularly present in stomach samples from both late and early fire sites, but the relative volumes were lower than the unburnt site. The relative volume of lepidopteran larvae (caterpillars) is partly responsible for this difference with both fire sites (early and late dry season) recording a high relative volume of caterpillars, whereas the unburnt site recorded a substantially lower relative volume of caterpillars (Table 3.6). 74

94 Table 3.5. Dry season stomach contents of frillneck lizards from the three fire treatment sites. Prey taxon Unburot Early fire Late fire Occurrence Abundance Volume Occurrence Abundance Volume Occurrence Abundance Volume (%) (%) (%) (%) (%) (%) (%) (%) (%) lsoptera Orthopetra Hemiptera Coleoptera Diptera Lepidoptera Hymenoptera Blattodea tn r- Mantodea Odonata Phasmotodea Aranea Chilopoda Gastropoda other totals n = 24 n = (ml) n = 61 n = (ml) n = 59 n = (ml) particular prey taxon, expresseo as a percentage or tne numoer or stomach samples examined. Abundance is the percentage of the number of items in each particular prey taxon. Volume is the percentage of the total estimated volume of all items. - denotes prey taxa not recorded in sample.

95 Table 3.6. Wet season stomach contents of frill neck lizards from the three fire treatment sites. Prey taxon Unburnt Early fire Occurrence Abundance Volume Occurrence Abundance Volume (%) (%) (%) (%) (%) (%) Isoptera Orthopetra Hemiptera Coleoptera Late fire Occurrence Abundance (%) (%) Volume (%) Diptera Lepidoptera Hymenoptera Blattodea \0 t- Mantodea Odonata Phasmotodea Aranea l.o Chilopoda Gastropoda other totals n = 20 n = (ml) n = 31 n = (ml) n = 30 n = (ml)

96 A total of 4227 invertebrates were collected by sweep-netting from April 1992 to February 1994 from the three fire treatment sites (Table 3.7). The abundance of invertebrates was greater for the wet season sample, comprising 71.8% of total items. The invertebrate orders and their proportion in the sample were broadly similar among the three fire treatment sites. Importantly, two prey taxon (termites and centipedes) which represent over 80% of the relative abundance and over 50% of the relative volume of stomach contents (Table 3.5 and 3.6), were not sampled using sweep-netting. Therefore, the prey availability data presented in Table 3.7 represents only a portion of the food available to the frillneck lizards. Hymenoptera was the most common invertebrate order at all three sites in the wet and the dry seasons. The relative abundance of orthoptera increased during wet season samples at both early and late fire sites, but remained constant between seasons in the unburnt site. Conversely, the decrease in relative abundance of hymenoptera was small in the unbumt site compared to the relatively larger decrease of hymenoptera in both fire sites, in the wet season samples. Body condition was analysed separately for males and females because males are significantly heavier at a given body length than females (Chapter 2: Body condition). Adult male body mass at a given length was significantly different among the three populations, but only during the wet season (dry season, F,82 = 2.65, P = 0.075; wet season, F1 56 = 4. 15, P = 0.018) (Figure 3.9). Tukey's comparison of means test indicated that males from the unburnt site had significantly lower body condition than either of the burnt sites during the wet season. Adult female body mass at a given body length was significantly different during the wet season, (F.. 65 = 4.33, P = ) (Figure 3.10). Tukey's comparison of means test indicated that females from the unburnt site have 77

97 Table 3.7. Relative abundance of prey taxon obtained from sweep-netting in each of the three fire treatment sites, during the dry and wet seasons. Samples from both years are combined. Taxa denoted with (--) were not recorded from sweep-netting but were recorded in the stomach contents of frillneck lizards. Prey taxon Unburnt Early fire Late fire Dry Wet Dry Wet Dry Wet Abundance Abundance Abundance Abundance Abundance Abundance (%) (%) (%) (%) (%) (%) Isoptera Onhopetra Hemiptera Coleoptera Diptera Lepidoptera Hymenoptera Blanodea M.antodea Odon.a.ta Pb.asmotodea Aranea Chilopoda Gastropoda other total no. items

98 .10 Dry season.05 en - ro ::l " en Q) 0::: f 39! Unburnt Early fi re Late fire.10 Wet season.05 en - 56 ro 32 ::l 0.00 "0 en Q) ::: I Unburnt Early fire Late fire.fire regimes Figure 3.9. Mean residuals of linear regression of log-body mass with log-svl for adult male frillneck lizards in each of the fire treatments, during the dry and wet seasons. Numbers are sample sizes and error bars are one standard error. Line through zero represents the best least squares fit. 79

99 .10,... Dry season en - ro =:J "'C en Q) 0::: 18.OS t t-- -.OS f- 3 f -.1 s 1----,-----,.-----r--- Unburnt Early fire Late fire.10 Wet season en - ro =:J "'C en Q) 0:::.OS OS I f -. 1 S Unburnt Early fire Late fire Fire regimes Figure Mean residuals of linear regression of log-body mass with log-svl for adult female frillneck lizards in each of the fire treatments, during the dry and wet seasons. Numbers are sample sizes and error bars are one standard error. Line through zero represents the best least squares fit. 80

100 significantly lower mean residuals than the late fire site, but not the early fire. Low samples sizes from the dry season population precluded statistical analysis of these data. DISCUSSION The results presented here show that dry season fires have a substantial influence on the ecology of frillneck lizards, and this effect was different between the two types of fires. The short-term effects of fire on frillneck lizards were considerably greater than the longerterm effects. Short-term effects of fire The mortality of frillneck lizards during early dry season fires was low compared to the late dry season fires, in which a substantial proportion of marked lizards died. Remaining perched in the tree canopy offers sufficient shelter for lizards during early dry season fires. Late dry season fires force lizards to take a more evasive response, with most lizards leaving their pre-fire location and either sheltering in larger trees or in hollow termite mounds on the ground. The use of termite mounds offers sufficient protection from fue, as all frillneck lizards that used them survived the high intensity fires. Many other vertebrates and invertebrates shelter in termite mounds during fires (Braithwaite 1990, pers. obs.). They are an important refuge during fires for Ch/amydosaurus, because there are no other refuges in this habitat which are large enough for adult lizards. Frillneck lizards are exposed to a greater risk of mortality by remaining in the tree canopy during late dry season fires because these fires regularly scorch the canopy. 81

101 Direct mortality of lizards during fire has generally been assumed to be low (Erwin and Stasiak 1979, Means and Campbell 1981, Bamford 1986). This is based on the assumption that the lizard fauna can escape fire by sheltering in burrows, crevices, or other available shelter. These conclusions are based on changes in abundance of lizard species after single prescribed or wild fires, and may be influenced by other factors such as predation and migration. The results from this study clearly show that mortality during fire can be either high or low within a single species of lizard depending on the intensity of fire. This study also documents that a range of behaviours are used by a single species of lizard when sheltering from fire, and the behavioural response is affected by fire intensity. Frillneck lizards move into or remain in open areas produced by fire. Population density over the two year period was higher in both burnt sites than the unburnt site (Chapter 4: Population size and density). This suggests that frillneck lizards remain in the burnt habitat. The selection of open habitat after fires has been documented for other species of agamid lizards in a variety of environments within Australia. The small agamid Diporiphora bilineata, which occurs in sympatry with Chlamydosaurus kingii in Kakadu National Park, increased in abundance after early dry season fires (Braithwaite 1987, Trainor and Woinarski 1994). Ctenophorus ine1mis, which inhabits hummock grasslands of arid regions exhibited a dramatic increase in abundance following fire (Pearson unpublished data, Dell and How unpublished data). However, other agamid species present in hummock grassland showed a decrease in abundance following fire (e.g. C. isolepis; Pearson unpublished data). In recently burnt mallee, Amphibo/urus pictus is a common species (Caughley 1985 ). Another study in mallee habitat found twice as many Amphibolurus fordi on burned sites as unburned sites (Cogger 1984). 82

102 The family lguanidae is analogous to the Australian agamids in many respects (Stamps 1983, Pianka 1986 ). Some species of iguanids also select open habitat resulting from f1 res. Sce/oporus occidentalis shows a preference for perching on burned shrubs, whereas they are usually found on the ground in unburnt habitat (Lilywhite and North 1974). A separate investigation of this species concluded that burnt habitat was preferred (Kahn 1960) although only a small sample was studied. The selection of open areas resulting from fires is not restricted to these two reptile families. Lizard species from other families also show a preference for recently burnt areas (Fyfe 1980, Caughley 1985, Mushinsky 1985). Numerous bird species in northern Australia colonise recently burnt areas because of increased access to food resources (Braithwaite and Estbergs 1987, Woinarski 1990). The volume of stomach contents in frillneck lizards increased in the week after early and late dry season fires, although this was only significant after the late dry season fires. However, there was only a small change in number of items in the stomach contents after both fire treatments. The number of prey size classes in the stomach contents increased one week after both fire treatments, but again only the late dry season fire was significant. Thus, even though lizards captured an equal number of prey items before and after fire, they are able to capture a greater range of prey sizes, which increases the volume of their stomach contents. The abundance and diversity of invertebrate prey taxa were not sampled directly after fr re in this study. However, research from a tropical savanna in Africa revealed an overall 83

103 decrease in abundance and a minor decrease in diversity of invertebrates immediately after ftre (Gillon 1983 ). The high intensity of late dry season fires would possibly have a similar effect on the overall abundance of invertebrates. Frillneck lizards increase the volume of food taken after ftres, whether or not there is a change in invertebrate abundance related to fire. The fact that this increase in volume is greater after the late dry season ftres, suggests that removal of ground vegetation after these fires increases the accessibility to food resources. An understanding of the foraging behaviour of this species supports this argument. They are "sit-and-wait" predators that perch on vertical tree trunks (Shine and Lambeck 1989, Chapter 2: Diet, Habitat use). From this position they visually locate invertebrate prey, descend to the ground to capture the prey, and then return to their vertical position on the tree trunk. A decrease in the density of ground vegetation would assist them in visually locating prey and increase the mobility of lizards while feeding on the ground. The change in relative abundance of invertebrate orders present in stomach contents was small after early dry season fire, but late dry season ftres resulted in a greater increase. The relative abundance of termites decreased by over 40% after late dry season frres and this was offset by a large increase in ants. The foraging activity of termites is decreased by the removal of grass which is their main food resource. Ants are prevalent in recently burnt areas within this habitat (Andersen 1991). This is possibly reflected in the diet of Chlamydosaurus. It should be noted that ants actually decreased in relative abundance in the early fire treatment stomach samples, and only increased after the late dry season ftres. This suggests that ants may be more active or abundant after the late dry season fires 84

104 because of the more open habitat, which is favoured by some species, i.e., Iridomyrmex. Longer-term effects of fire The effect of fire on the vegetation structure of open Eucalypt forests in northern Australia is complex, and remains unclear and contentious among researchers. Previous authors suggest that because of the many complex relationships between the vegetation and edaphic factors, the effect of fire on vegetation is difficult to determine (Stocker and Mott 1981, Bowman et a/ Lonsdale and Braithwaite 1991 ). Exclusion of fire does not result in closed forest community in the open Eucalyptus forests of northern Australia (Bowman et a/. 1988). Preliminary data obtained from the Kapalga fire experiment suggest unburnt forests maintain a homogenous canopy cover, whereas annually burnt forest have a heterogenous canopy cover (R.J. Williams pers. comm.). Fire generally creates a heterogenous habitat, which may create a wider choice of habitats for frillneck lizards to use. The probability of occupancy of trees by frillneck lizards increased with a thicker canopy and less ground vegetation. The amount of canopy cover of trees selected by frillneck lizards may influence their thermal relationships. Frillneck lizards thennoregulate carefully in their environment, reducing their midday body temperature by 4oC during the dry season (Christian and Bedford 1995). Assuming a heterogenous canopy cover in annually burnt areas, the selection of trees with thicker canopies suggests that these trees are possibly preferred by frillneck lizards. Lizards in habitat unburnt for several years select trees with a more open canopy, which may be associated with more open habitat The suitability of these trees for thermoregulation is unclear and would warrant further investigation. 85

105 The preference for trees surrounded by a low density of ground vegetation possibly relates to the foraging strategy of frillneck lizards. As suggested in the analysis of short-term effects of fire, a decrease in ground vegetation may enable lizards to access a greater variety of different sized prey. Even before fires, the lizards prefer areas with less grass in all three fire regimes (Figure 3.6). The avoidance of densely vegetated areas around creeks and rivers by frillneck lizards (Shine and Lambeck 1989) supports this conclusion. The variation in overall stomach contents (i.e. total volume, number of items, number of orders and prey size classes) among the three fire treatments during both wet and dry season was small. However, there were some important differences in the type of prey taxa eaten by frillneck lizards. Termites remain a primary food source for frillneck lizards in the unburnt site in both the wet and dry seasons. The proportion of termites in the stomach contents was lower in the early and late dry season fire sites. Harvester termite (Trinervitemzes and Cubitermes) abundance in a tropical African savanna habitat decreased following fire due to the decrease in food availability for termites (grass and leaf litter) (Benzie 1986). A similar relationship in this savanna habitat is possible. The removal of dry grass by dry season fires may reduce the abundance of harvester termites, and therefore reduces their availability to frillneck lizards over the dry season and the following wet season. The predominance of lepidopteran larvae (caterpillars) in the wet season stomach contents exhibited substantial difference among the fire treatments. The stomach contents from the late fire site in the wet season were dominated by caterpillars. Caterpillars were less important in early fire samples, and caterpillar abundance was lowest in the stomach 86

106 samples from the unburnt site. Sweep-netting for invertebrates during the wet season months failed to reveaj any major differences among sites with respect to the abundance of caterpillars. Therefore, the reason for the difference in abundance of caterpillars among the fire treatments remains unclear. Given the similarity in total volume of stomach contents among the populations in the three fire regimes, it is surprising to find a significant difference in the body condition for both majes and females among the three sites during the wet season. The wet season is a period of increased activity and high field metabolic rates, as well as being the reproductive season. Possible differences in the types of invertebrate orders eaten in the wet season may account for differences in body condition. The time spent foraging was not investigated in this study, but may be an important factor in influencing body condition. Generally, lizards use areas that have a lower density of grass. Fires create more of these areas and the open areas may increase the ease of capture of prey. There was no difference in body condition during the dry season. Access to larger prey may only last a short time after the dry season fires, therefore dry season body condition may not be affected. Fire is important in maintaining the relatively open habitat that is preferred by Chlamydosaurus kingii. The exclusion of fire has been implicated as a possible reason for reduced size and growth rate in a peripheral population of a North American lizard, Crotaphytus collaris (Sexton er a/. 1992). Complete fire exclusion is unrealistic in northern Australia, where fire is extremely common. Early dry season fires are possibly the most beneficial to Chlamydosaurus as mortality is low and a more heterogenous habitat (containing preferred open areas) is created. The long-term viability of populations in areas 87

107 that are repeatedly burnt during the late dry season is unclear, although the 30% direct mortality suggests that a population will decrease if these fires are repeatedly applied over an extended period (greater than 5 years). The relatively high densities of frillneck lizards within the late fire sites (Chapter 4: Population dynamics) and numerous adult females suggests that these populations are viable at the present time. Migration is a crucial factor in assessing the viability of populations, and it has not been discussed in this paper. The distances moved by these lizards can be substantial during the wet season (Shine and Lambeck 1989, pers. obs.). However, the greater the area burnt during a late dry season fi re, the less migration will occur into the centre of the area. Therefore, restricting the area burnt during late dry season fires will ensure migration is relatively easy for this species. Summary In summary, fire has a clear effect over a short period of time, and a less obvious effect on the ecology of Chlamydosaurus over a number of years. The intensity of fires influences the level of the behaviour, mortality and diet. The general diet shows no clear differences among different fire regimes. Overall, frillneck lizards occupy trees with relatively thick canopy cover surrounded by a low density of grass and shrubs. Lizards inhabiting fire excluded habitat show differences in the occupancy of trees, which is related to fire induced changes to the vegetation structure. Differences in condition among the three fue regimes may be related to differences in prey taxa taken, or potential differences in energy budgets related to foraging or reproduction. 88

108 Chapter 4 The demography, population dynamics and dry season home range of frillneck lizards, with reference to the effects of three different fire regimes. INTRODUCTION The study of species population dynamics is essential to understanding life-history strategies. The primary cause of similarities in general life-history strategies among species are their phylogenetic relationships, particularly at the family level (Stearns 1984, Dunham and Miles 1985). Variation of life history within families is largely based on phenotypic responses to environmental conditions (Ballinger 1983 ). The subject of this study, Ch/anzydosaurus kingii, belongs to the family Agamidae. A total of 64 species of agamids inhabit Australia. Nine species, including Chlamydosaurus, have large body sizes (snout-vent length > 150 mm) and represent some of the most spectacular members of this continent's diverse reptile fauna. The taxonomic relationships among the large agamids suggest differences in their biogeographic origins. Hypsilurus and Physignathus are restricted to moist habitats along the east coast of Australia (Cogger 1992). The existence of other species in the same genus (with a similar karyotype) throughout southern Asia, indicates that they originated from Asia (Witten 1982, 1983). Pogona and Chlanzydosaurus represent the large Australian agamids which have an extensive distribution throughout Australia, but predominate in the arid and tropical climatic regions (Cogger 1992). They share many morphological and genetic characteristics with the other smaller Australian agamids, and it is most probable that their evolution has taken place entirely within Australia (Witten 1983, Greer 1989). 89

109 The population dynamics of large agamid lizards in Australia is poorly understood. A recent demographic study on the eastern water dragon, Physignarhus lesueurii represents the first published data on the population ecology and demography of the Asian lineage of Australian agamids (Thompson 1993 ). The only published information on the population dynamics of the lineage of large agamids that evolved in Australia is for C. kingii. The population dynamics of the genus Pogona is even more poorly understood. This lack of demographic information for large agarnids in Australia makes analyses of familial trends in life history strategies difficult. Shine and Lambeck's (1989) study of the ecology of C. kingii established some basic demographic information for this species. The reproductive cycle of males and females, and body size at sexual maturity were determined using museum specimens. Additional reproductive data is available on the incubation periods, hatchling size, clutch size and clutch frequency (Bedford et a/. 1993). This study has two principal aims: (1) to further document the demography and population dynamics for this species; and (2) to examine some possible effects of fue on aspects of the population dynamics. This information will facilitate future research concerning the evolution of life history strategies of large agamids in Australia. Specifically, data on the general reproductive cycle, growth, age, population structure, density, home ranges and intraspecific interaction will be examined for Chlamydosaurus kingii. Where sample sizes permit, analyses on the possible effects of fue have been included. 90

110 METHODS Sampling For a full description of the site and climate see Chapter 2 (Methods). Frillneck lizards were monitored using both telemetry and mark-recapture, and both methods are described in Chapter 2 (Methods). Reproductive biology The presence of oviductal eggs in females was determined by palping the abdomen. This method is sufficiently accurate for determining whether the lizards were gravid or not. Data over the two reproductive seasons sampled were combined to establish an annual reproductive cycle. The small sample size prevented comparison of variability in the reproductive cycle on a site and yearly basis. The age and size at sexual maturity were determined from the recapture of permanently marked hatchlings. Minimum longevity of C. kingii in the field was determined from the recapture of permanently marked and fully grown adult lizards. Growth and age The growth rates of frillneck lizards were measured by the change in snout-vent length (SVL) between the initial capture and the last recapture of an individual. Growth rates are expressed in mm day" 1 The mid-point of the SVL between the initial and last recapture was also calculated for further analyses between sexes and sites. The age of lizards within a certain SVL size class were estimated using growth rates and the SVL at hatching. 91

111 Population dynamics The SVL at initial capture of all individual lizards were used to determine the population structure of C. kingii at each fire treatment site. Lizards were grouped into one of ten size classes using 25 mm SVL intervals. Sex ratio (number of males/number of females) was calculated using these data. Population estimates of C. kingii were calculated using recapture data for the three sites. Lizards monitored using telemetry were omitted from these analyses because their recaptures were non-random. An underlying assumption of most population estimate models is the equal catchability of individuals within the population (Caughley 1977, Krebs 1989). Behavioural differences in C. kingii between the wet and dry seasons make this assumption difficult to meet (Chapter 2: Results). A variety of population estimate models were applied to the mark-recapture data because of this seasonal variability. Three frequency of capture distribution models were used: poisson; negative binomial; and geometric distribution (Caughley 1977). An additional jackknife estimate was also used due to the unequal distribution of some of the recaprure data (Chao 1988). The following assumptions were made for each of the three populations: the populations were open (birth, deaths and migration occurring throughout time); that marked and unmarked lizards were equally catchable; and sexes were equally catchable. To determine the 'most' accurate population estimate, a 'goodness of fit' test was applied between the observed recapture data and the expected values calculated by each of the three frequency distribution models (not applied to the jackknife estimate). A model that is significantly different to the observed data will produce a poor estimate of population size (Caughley 1977). 92

112 Densities of C. kingii were calculated by dividing the population size estimate (N) of each fire treatment site by the area (ha) of each site. James (1991 a) describes this as a "naive density", as it is not strictly an absolute density. Sites were based on a network of roads and were irregular in shape. The area of each site was estimated by multiplying the total length of the roads by the width of the area surveyed. The length of each road was measured using a vehicle's tachometer. The width of the area was measured as the distance between the centre of the road and the furthermost point from the road at which a lizard was captured, and this distance was multiplied by two for total width. Home ranges Adult lizard home ranges were determined from locations of telemetered individuals during a single dry season period (May to September). Home range analysis was restricted to the dry season because the primary purpose of this exercise was to study the effect of prescribed fires on home range size. Wet season home ranges have been reported previously (Shine and Lambeck 1989). Lizards were located between two week and monthly intervals, except during periods prior to and after prescribed fires when several locations were recorded. The minimum number of locations needed to confidently predict an animal's home range was determined by plotting the percentage of home range area against the cumulative number of locations (Rose 1982). An asymptote is reached at 8 locations, which accounts for approximately 80% of an adult frillneck lizard's home range in the dry season (Figure 4. 1 ). Therefore, lizards with less than a total of eight locations in the dry season were omitted from this analysis. Home ranges were calculated by the minimum convex polygon method, using the computer program Wildtrak (Todd 1993). Home ranges were not corrected for unequal number of locations. 93

113 100 I - - I : I I I 80 I -..! I I i 0._... l Q.) 0') 60 I c I l I rn I... Q.) 40 E I 0 I l I 20 I I I I I A A. I. 16 Number of locations Figure The percentage of home range estimated from the cumulative number of locations collected in the dry season. Approximately eight locations are needed to describe 80% of a frillneck lizards' dry season home range. 94

114 An index of movement within a home range was calculated by taking the mean distance (m) of all relocation sites from the geometric centre of each individual's home range. This produces a weighted measure of movement for each individual within their home range, and reduces the effect of uneven intervals between relocation of transmittered lizards. Analysis Differences in growth rates between sexes and fire treatments were analysed by AN COY A with mid-point SVL as a covariate. Contingency table analysis was used to test for differences in the population structure of adult frillneck lizards among the three fire treatments. Differences in home ranges between sexes were analysed using unpaired t-tests, and one-way ANOV A was used to test for differences among fire treatments. All means are presented with one standard error unless otherwise stated. RESULTS Reproduction biology The mean SYL of gravid females was 202 mm ± 2.45 (n = 24). Gravid females represented 39% of all sexually mature females caught during both reproductive seasons sampled (n = 60). There was no difference between the proportion of gravid females in each of the two years sampled, with 42% (n = 25) gravid in the season and 37% (n = 35) gravid in season. Gravid females were flrst captured during early November in both years, and this was also the month when the largest number of gravid females were caught (Figure 4.2). Capture effort did vary among months and was highest 95

115 en (1) cu E... ;:j ""0 cu 0 (1)..0 E ;:j z I l Non-gravid n Gravid Oct Nov Dec Jan F b M e ar Apr Months Figure 4.2. The number of gravid and non-gravid female frillneck lizards captured each month during the wet season. Both reproductive seasons are combined. 96

116 during November. When capture effort (number of kilometres censussed per month) is used to weight monthly samples, December recorded the highest number of gravid females. Gravid females continued to be captured until March, where a slight increase in the proportion of gravid females occurred. Eight adult females were caught in both reproductive seasons. From this sample, three were gravid in both seasons, three were gravid for only one of two seasons, and two adult females were not gravid in either season. Bedford et a/. ( 1993) recorded a minimum incubation period in the laboratory of 54 days at 33 c and days at 30"C. If an average of 60 days incubation period is used, the earliest date of hatching is approximately the 1st of January each year, given the 1st of November as the earliest date of oviposition. The latest possible date of hatching would be approximately the 1st of June, given the 1st of April is the latest recorded date of a gravid female in this population. Therefore, a considerable amount of variation in the age of hatchlings (up to 6 months) is possible in any one year. The earliest record of a hatchling C. kingii from Kapalga was April 1992, with a SVL of 55 nun. Bedford et a/. ( 1993) recorded a mean SVL of 48 mm (range: mm) of hatchlings hatched in the laboratory. Therefore, i would be reasonable to assume that a hatchling with an SVL of 55 mm would have emerged the previous month (March). Hatchlings have a distinctly different color pattern compared to juveniles and adults. They are grey with distinct uneven dark bands across their body. The ventral surface is white and the frill is short In contrast, juveniles and adults develop a dark ventral surface, have a predominant brown or reddish color and a relatively long frill. Only a small number of hatchlings were captured throughout this study, possibly due to their small size, cryptic 97

117 behaviour and excellent camouflage. A female hatchling was marked in August 1992, with a SVL of 94 mm. It was re-captured in November 1993 with a SVL of 180 mm, and it was gravid. This recapture establishes two important population parameters for C. kingii: (1) sexual maturity in females is attained at an S VL of 180 mm; and {2) sexual maturity in females may occur during the second year of a frillneck lizard's life. Shine and Lam beck (1989) estimated size at sexual maturity of females to be approximately 175 mm SVL. Growth and age Growth rates for male and female frillneck lizards are illustrated in Figure 4.3. The growth rates were calculated over a mean of 174 ± (n = 73) days. Juvenile frillneck lizards (SVL <175 mm) grow significantly faster than larger adult lizards (juveniles, mean = ± nun day, n = 6; adults, mean = ± mm day, n = 67; t = 6.94, DF = 71, P < ) (Figure 4.3). Males grow significantly faster than females (AN COY A: slopes, Fu = , P = 0.476; intercepts, F1.70 = 66.25, P < ). Female growth rate decreases at a SVL of approximately 200 nun, after that growth is neglible. Males, however, continue to grow until they reach a SVL of approximately 240 mm, after that growth is neglible. Analyses of growth rates of male lizards among the fire treatment sites showed no significant difference (ANCOVA: slopes, F2.38 = 0.755, P = 0.477; intercepts, F2.40 = 0.1 1, P = ). A comparison of female growth rates was not possible among the three 98

118 I>-. rn "0.3 E E 0..._... Q).2.. rn - 0..c CJ t SVL (mm) Figure 4.3. Growth rate of male and female frillneck lizards. The mid-point of SVL between first and last recapture is used. Open circles represent males and closed circles represent females. 99

119 fire treatments because of the low sample size in the unbumt site (n = 3). However, a comparison of female growth rates between the early and late dry season frre sites showed no significant differences in growth rates (ANCOVA: slopes, F1.'1A = , P = 0.879; intercepts, F us = 1.899, P = ). By knowing the earliest date of emergence, the mean growth rate of lizards in each size class and the date of capture, it is possible to estimate the age of an individual up to three years. Individuals in the first year age cohort (hatchlings) range between a SVL of mm. This estimate is based on a mean growth rate of sexually immature juveniles of 7.29 mm month 1 (n = 6). Therefore, hatchlings are able to grow approximately 90 mm in their frrst 12 months. The SVL of individuals in their second year (juvenile) ranges between mm, based on the same mean growth rate of the hatchling size class. Males grow to a greater SVL than females in the second year because of their greater growth rate (Figure 4.3). The differences in age classes becomes unclear at this point, but it is reasonably safe to assume females with a SVL of 200 mm or greater, would be at least in their third year of growth. Males do not reach an asymptote in SVL until 240 mm, which would require a further year of growth. Therefore, males with SVL > 200 mm and < 240 mm are approximately three years of age, and males above 240 mm SVL would have a minimum age of four years. Minimum longevity of frillnecks in the field can be estimated from recapture data of adults. An adult male with a SVL of 260 mm was marked in September 1991 and was periodically recaptured until October Therefore, the approximate minimum age of this individual is 6 years, given that it takes males four years to reach a SVL of 240 mm. 100

120 An adult female with a SVL of 205 mm was captured in December 1992 and recaptured in March 1994 with a SVL of 215 mm. The approximate minimum age of this female is 4.3 years, given it takes females two years to reach a SVL of 180 mm. Population dynamics The initial capture of all individual lizards was used to determine the population structure. A total of 245 C. kingii were captured within the three fire treatment sites from April 1992 to April 1994 (Figure 4.4). A notable feature of all three sites is the general absence of hatchling and juvenile lizards. This was due to their small body size and cryptic behaviour, which increased the difficulty in sighting them. The relative number of adult lizards (SVL > 175 nun) in each size class did not differ significantly among the three populations (X2 = 5.92, DF = 6, P = 0.435). Lizards in the mm SVL size class were the most numerous relative to the other size classes, in all three populations (Figure 4.4). Only seven adult females were captured in the unbumt site compared to 34 and 39 in early and late sites, respectively (Figure 4.4). Therefore, the potential reproductive output of the population in the unburnt site is limited. The number of lizards with a SVL between nun in the late fire treatment site was low compared to the early fire treatment. This may represent selective mortality of this size class during late fires. The sex ratio of all captures from all sites was heavily biased towards males (m/f: 1.73 ). This is possibly a result of greater visibility and mobility of large males as they defended territories. The number of male and female frillneck lizards captured each month over two 101

121 Early fire CJ) "'0,_ 45 ro 40 N 35 0,_ Late fire Q) 15.c 10 E 5 :::s 0 z Unburnt , SVL size classes (mm) Figure 4.4. Number of lizards in the di[f-.;rent size cohorts within each fire treatment site. Open bars represent females, closed bars represent males and lined bars represent juveniles and hatchlings. 102

122 years were tabulated to allow direct monthly comparisons of the sex ratio (Figure 4.5). Considerable monthly variation is evident in the sex ratio which possibly reflects differences in reproductive behaviour. The sex ratio of adult frillneck lizards among the three fire treatments varied considerably. The unburnt site exhibited the largest male bias (2.85), with early fire (1.76) and late fire ( 1.85) treatment populations being slightly less biased towards males. Population estimates from each site were derived from four frequency of capture models (Tables 4.1, 4.2 and 4.3 ). The geometric distribution model produced the best 'goodness of fit' of the three frequency of capture models for each fire treatment site. The negative binomial model was the least accurate showing highly significant differences with the observed data. A 'goodness of fit' test was unable to be used for the unburnt site because of insufficient recaptures. The jackknife estimate consistently produced estimates within the range produced by Caughley's (1977) "triple" frequency distribution of captures method, and the jackknife estimate was selected for the calculation of frillneck lizard density. The population estimate for the unburnt site is likely to be under-estimated due to the large proportion of adult lizards (65%) 'removed' from the mark-recapture population for monitoring by telemetry, compared with a lower proportion of 'removal' in the early (20%) and late ( 17%) fire treatment sites. Using the jackknife population estimate, the density of C. kingii in the early ftre site, late fire site and the unbumt site were 0.65 ha I, 0.78 ha I and ha 1, respectively. 103

123 50 CJ) 40 '1::3 l... ro N 30 0 l... /.\ <D..c I \ E 20 ::J I I z I I 10? 0 I j\ \ \ I I b "'-- rej 0 T J F M A M J J A s 0 N D Months Figure 4.5. Mean monthly capture rates of males (circles) and female (squares) frillneck lizards. All sites and years have been combined. l04

124 Table 4.1. Observed and expected capture frequencies, population estimates and "goodness of fit" test for the early fire site. Number of Number of Poisson Negative Geometric Jackknife estimate captures individuals binomial (95% confidence intervals) x z = DF = 2 2 P = < Estimate N = ( ) Table Observed and expected capture frequencies, population estimates and "goodness of fit" test for the late fire site. Number of Number of Poisson Negative Geometric Jackknife estimate captures individuals binomial (95% confidence intervals) ll = x z DF = 2 2 P= Estimate N = ( ) 105

125 Table 4.3. Observed and expected capture frequencies, population estimates for the unburnt site. The "goodness of fit" test was not possible due to the low number of recaptures. Number of Number of Poisson Negative Geometric Jackknife estimate caprures individuals binomial (95% confidence intervals) = x.2 Invalid Invalid Invalid OF = P= Estimate N = (27 to 238) 106

126 From a total of 79 females caught in the three fire treatment sites, none had any evidence of scars (fresh or old). For adult males, 27% (n = 120) were recorded with scars. Analysis of the proportion of males with scars in four different size classes within this sample revealed that only 2% of males with an SVL < 225 mm (n = 50) had scarring (Figure 4.6). The proportion of males with scars increased with increasing SVL (Figure 4.6). The cause of scars is presumed to be from intraspecific aggression, not predation, as females did not have any scars. The data suggest that male-male fighting (territorial defense) is greater among large males. The proportion of scars from each of the three sites was not constant. No males in the unburnt site had scars. This suggests less interaction among males. Fresh scarring on adult males was recorded only during the fu st months (October to December) of the reproductive season. Home ranges Dry season home ranges are presented in Table 4.4. They represent data from a period of no greater than six months for each individual lizard. Adult male home ranges during the dry season were significantly larger than females (males, mean = 1.96 ± 0.57 ha, n = 16; females, mean = ± ha, n = 7; t = 2.9, DF = 20, P = 0.009). This suggests either: ( 1) adult males require larger home ranges to find more food resources; or (2) adult males establish larger home ranges to include the maximum number of females. There was no significant correlation between SVL and home range for either sex (male, r = 0.406, P = ; female, r = 0.488, P = 0.531). Analysis of differences in home range among the fire treatment sites was only possible for males, as female sample size was too small. Although the mean values for both burnt plots were greater than the unburnt plot (early, mean = 2.91 ± 1.08 ha; late, mean = 1.94 ± 1.04 ha; 107

127 en.8 (]) ro E.. :::J.6 / D. "'C / ro / '+- 0 / 0.4 c 0 t 0!/ a..2 0 /I l.. _a! I 0.0 j/ fr..,...- c..< Adult male size classes (mm) Figure 4.6. Proportion of adult male frillneck lizards with scars fo r four size classes in each of the fire treatment sites. Triangles represent the late fire site, squares represent the early fire site and circles represent the unbumt site. 108