Chapter 1 LITERATURE REVIEW 1.1 Nectarivore Communities

Chapter 1 LITERATURE REVIEW

1.1 Nectarivore communities

There is considerable debate regarding the role of competition in producing patterns of niche partitioning in ecological communities (e.g. Connell 1983; Roughgarden 1983; Simberloff 1983; Stong Jr. 1983; Mac Nally 1995; Walter and Paterson 1995; Wisheu 1998; Gordon 2000; Mac Nally 2000). Competition is thought to refine the niche of a species when two or more species compete for some resource that is limiting. Natural selection favours conspecifics that compete less with individuals of other species, tending to widen any niche separation between the competing species. There are a number of lines of argument against niches being competitively structured: individualistic, logical structure and lack of field evidence (Gordon 2000). The individualistic argument suggests that species have individual tolerances, preferences and limitations in relation to environmental gradients that govern their distribution and abundance, not pressure from sympatric species (e.g. Stong Jr. 1983). The logical structure argument attacked the idea of competitively structured communities, suggesting that studies had failed to adequately discount a ‘null’ hypothesis that communities are actually assembled at random from species in a regional pool (e.g. Simberloff 1983). Finally, reviews of field evidence suggested that other factors are more important in limiting the growth of natural populations, such as predation, parasites and environmental heterogeneity (e.g. Connell 1983).

Despite the arguments of the previous paragraph, studies of nectarivore communities usually invoke competition as the driving force behind community organisation (e.g. Keast 1968a; Ford and Paton 1976b; Lyon 1976; Ford 1977; Ford and Paton 1977; Ford 1979; Feinsinger et al. 1985; McFarland 1986a). In the case of nectarivore communities, competition does appear justified as the primary factor in structuring communities. This is partly due to obvious interference competition, particularly by the larger species within the community, and partly due to the fact that nectar is often a resource in demand by all the component species, and has been shown to be inadequate to meet the requirements of the local nectarivore community at least temporarily (Ford 1977; 1979; Feinsinger et al. 1985; McFarland 1986a; Armstrong 1991). Particularly in Australian systems there are likely to be a number of different nectarivore species

1 attempting to access nectar from the same plant at any one time. Two factors contribute to this shared preference for nectar: the flowering seasons of Australian plants provide a series of different nectar sources at different times of the year (Ford 1977; 1979; Paton 1979; 1986a); and there is no specificity of nectarivore species to plant species (Paton and Ford 1977; Collins and Briffa 1982). Any fluctuations in nectar availability are therefore likely to cause nectarivores to compete for resources during times of shortage. Alternative sources of renewable carbohydrate (Section 1.5.3.2) are likely to be used by nectarivore communities in similar ways to nectar resources (Paton 1980). A further piece of evidence suggesting the importance of competition in honeyeaters are the consistently male skewed sex ratios recorded by studies on honeyeaters (Paton 1979; Pyke et al. 1989; Foster 2001). Male honeyeaters are usually larger than female honeyeaters (Collins and Paton 1989; Paton and Collins 1989), and have been implicated in aggressively excluding females from the best resources, thus forcing them to both move further and forage from inferior resources leading to increased mortality (Paton 1979).

Studies of nectarivore communities suggest a number of ways through which competition drives community organisation: size (Ford and Paton 1977; Ford 1979; Paton 1979; Wykes 1985; McFarland 1986a; Collins and McNee 1991), beak length (Ford 1977; Ford and Paton 1977; Paton 1986a; Paton and Collins 1989), habitat (Ford and Paton 1976b; Ford 1977; Recher 1977; Ford 1979; Loyn 1985; Wykes 1985) and behaviour, which includes a mix of social and feeding strategies. The best documented of behavioural strategies is the dominance of an area by Manorina (miners) through group territorial defence (Dow 1977; Loyn et al. 1983; Poiani et al. 1990; Pearce et al. 1995; Grey et al. 1997; Clarke and Schedvin 1999) but also includes concepts that have received only passing comments in the literature (e.g. 'prostitution', Wolf 1975; Paton 1979). Two behaviours, termed here swamping and stealth, have been reported by a number of authors. Stealth behaviour is the use of secretive behavioural techniques to access resources that are being protected (Lyon 1976; Paton 1979; McFarland 1996). Flocking behaviour, or swamping, is the use of a combined direct approach by a number of individuals to access resources that are being protected (Paton 1979; 1980; McFarland 1986a; Slater 1994; Timewell 1997).

2 Ford (1979) suggested size as the most important niche axis dividing honeyeater communities; the largest species aggressively exclude the smaller species from the best resources. Size has been criticised as a niche axis on the basis that it really includes a variety of actual niche axes such as resource harvesting efficiencies, metabolic rates and home range size (Gordon 2000). However, in the case of nectarivores it appears justified, as it plays a direct role in giving the largest species access to the best resources. Large honeyeaters require access to the best available nectar resources to meet their energy requirements and due to their size are able to aggressively dominate those resources (interference competition). The smaller honeyeaters, will gladly also use the best nectar supplies if they can access them (shared preference for resources) but are often compelled to use inferior nectar resources due to the aggression of larger honeyeaters. Smaller honeyeaters are able to use inferior nectar supplies as their overall energy requirements are less (Ford 1979). Collins and McNee (1981) supported this hypothesis, finding the largest honeyeaters relied on the most productive plants, while the smaller honeyeaters were forced to use less rewarding plants, as did Paton (1979). McFarland (1986a) provided further support, finding that a spatial gradient in nectar richness enabled several species to coexist. The largest species dominated the richest areas and the smaller species exploited the poorer areas. However, temporal variation was also found to be important, with most species, ‘unhindered in terms of where, and on what resources, they can forage’ during times of either very low or very high nectar resources (McFarland 1986a).

Nectar resources are also divided amongst honeyeaters based on beak length. The long- beaked species take nectar from the flowers of all plant species, while the short-beaked honeyeaters may have trouble reaching the nectar from tubular or gullet-shaped flowers (e.g. Astroloma or Epacris) and are therefore primarily limited to open flowers (e.g. Eucalyptus) (Ford and Paton 1977; Paton 1986a).

Dow (1977) documented indiscriminate interspecific aggression of dense colonies of a species of Manorina (miners) leading to its domination of an area. Studies have since shown that the removal of Manorina colonies results in an influx of other honeyeaters and insectivorous birds (Loyn et al. 1983; Pearce et al. 1995; Grey et al. 1997; Clarke and Schedvin 1999). A long term study (7 years) at one site in south-eastern Australia showed a decrease in honeyeater populations corresponding to an increase in a

3 Manorina melanophrys (Bell Miner) population (Poiani et al. 1990). Similar studies have not been published with other suspected aggressive species, such as Anthochaera spp. (Wattlebirds), Philemon spp. (Friarbirds) (Higgins et al. 2001) or New Holland Honeyeaters (Phylidonyris novaehollandiae). However, one study did demonstrate an increase in small honeyeater species after a decrease in abundance of the dominant Anthochaera chrysoptera (Little Wattlebird) from a site following removal of nectar sources (Pyke 1989). There is also some evidence that removal of P. novaehollandiae from an area that they previously dominated (in which flowers were grown commercially) resulted in an influx of other smaller honeyeaters (D. Paton, unpublished data.).

Besides the arguments against community organisation being competitively structured, outlined in the first paragraph, there is also an argument that on the Australian mainland, the scale at which many birds move and the heterogeneity of the landscape preclude the conditions necessary for competitive interactions to develop patterns of community organisation (Mac Nally 1995). Mac Nally (1995, pg. 378) suggests that, ‘Local diversity at any time appears to be determined by a complex relation between the available regional pool of species potentially able to occupy a location, idiosyncratic habitat requirements and large-scale dynamics of individual species, resource irruptions and habitat architectures.’ This complex relation supposedly leads to a situation in which, ‘it is not surprising that similar species should frequently co-occur. To the contrary, it would be surprising were ecological differentiation to emerge from small- scale competitive interactions (1995, pg. 378).’ Mac Nally (2000) provides data outlining a situation in which three similar species (insectivorous birds) co-occur in relatively high densities without substantial differences in foraging, providing one example of how competitive interactions are not structuring the bird community. However, while communities may not always be organised through competitive interactions, the importance of competition in refining a species niche should not be ignored, especially in nectarivore communities which are reliant on an overlapping subset of resources. For such communities, movement and spatial heterogeneity are unlikely to provide refuge from competition, both exploitative and interference, at any scale when nectarivore resources are reduced regionally. Thus, any nectarivore which is able to reduce the effects of competition through a combination of attributes related to

4 size, beak length, habitat or behaviour, or other attributes, will be advantaged through evolutionary time.

1.2 Declining woodland birds

There is now a large body of evidence supporting an apparent continual decline in bird species of woodlands and associated habitats across agricultural Australia (Ford and Howe 1980; Garnett 1992a; Garnett 1992b; Paton et al. 1994a; Robinson 1994; Robinson and Traill 1996; Paton et al. 1999; Recher 1999; Garnett and Crowley 2000; Ford et al. 2001). While loss of woodland habitat is undoubtedly the primary cause of the decline in woodland bird species, there are a plethora of other hypotheses put forward to account for the continued decline in many woodland bird species. Robinson and Traill (1996) initially summarised the known threatening processes to woodland birds, but not until Ford et al. (2001) has an attempt been made to present hypotheses based on the list of threatening processes (although this is based to some extent on Walters et al. 1999). Any attempt to clearly present the threatening processes as a set of hypotheses is hampered by the inherent inter-dependence of many of them, and the fact that many interact, exacerbating each other.

In recent years a number of studies have investigated the cause(s) of decline in individual species. Some of this work has suggested a single primary cause of decline, enabling better targeted management. Walters et al. (1999) looked at three hypotheses for the decline of Climacteris picumnus (Brown Treecreeper) in fragmented woodlands of north-eastern New South Wales and suggested that disrupted dispersal rather than reduced fecundity or reduced food availability was responsible for the decline. Translocation of female birds then confirmed that dispersal rather than any other habitat degradation process was the cause of decline (Cooper and Walters 2002). Short-term management of C. picumnus can therefore focus on moving individuals around the landscape. However, long-term recovery options must focus on providing a less fragmented landscape, if the species is to survive without continued management.

There have also been a number of studies on ground foraging insectivores. Weeds were found to limit access to food by Petroica multicolor (Scarlet Robin) (Heddle 1999). The interaction of perches and open ground were found to be important in determining use of areas within the home range of Melanodryas cucullata (Hooded Robin) and Myiagra

5 inquieta (Restless Flycatcher) (Rogers and Paton, unpublished data). In these cases weed management is likely to benefit the species, and also the prevention of firewood collecting and other practices that reduce the abundance of perches at low heights. Eopsaltria australis (Eastern Yellow Robin) was found to suffer food shortage in small forest fragments (Zanette et al. 2000). Small fragments had about half the invertebrate biomass of large fragments, suggesting that re-establishment of native vegetation was the primary recovery requirement.

Research on declining honeyeater species has been less successful at determining specific causes of decline and therefore management options. Ford et al. (1993) postulated the decline of Xanthomyza phrygia (Regent Honeyeater) to be the result of being forced to spend a disproportionate amount of time in aggressive acts during breeding. They suggested that in the current landscape, X. phrygia, a nomadic species which nested in loose aggregations at rich flows of nectar, was no longer able to reach sufficient numbers in nesting aggregations to share the efforts of excluding other birds. The current fragmented landscape also focused the largest honeyeaters and lorikeets into the best nectar resources where X. phrygia was attempting to breed. However, two studies carried out in different areas over periods of at least two years found the reproductive success of X. phrygia was not limiting their ability to maintain current populations (Geering and French 1998; Oliver 1998b). There was also no evidence of a problem with lack of, or access to, food (Oliver 2001). Despite not finding any specific causes of decline and associated management possibilities, autecological work on X. phrygia has been able to highlight the importance of mature trees and non-nectar carbohydrates, especially when breeding (Oliver 1998a; 2000). Previous work had suggested that X. phrygia was more dependent on key eucalypt species (Franklin et al. 1989; Webster and Menkhorst 1992).

Lichenostomus melanops cassidix (Helmeted Honeyeater) has also received some attention. Again, there was no evidence for reproductive performance limiting the population in a study lasting nearly ten years (Franklin et al. 1995). However, again, competition was thought to be important to L. melanops cassidix populations, with competition for space demonstrated between L. melanops cassidix and Manorina melanophrys (Bell Miner). Lichenostomus melanops cassidix re-occupied former habitat after removal of invading M. melanophrys (Pearce et al. 1995). Lichenostomus.

6 melanops cassidix was also restricted in its use of plant species compared to other honeyeater species occurring in sympatry, being associated with only one (Eucalyptus camphora open forest) of six vegetation communities. Pearce and Minchin (2001) postulated that this floristic restriction may lead to competition for habitat with the sympatric, habitat generalist honeyeaters, both of which occurred in all six vegetation communities. Finally, intraspecific variation in vertical resource use during the non- breeding season was demonstrated in L. melanops cassidix, with males occupying the canopy of Eucalyptus camphora and females the lower strata of E. camphora and tea- tree (Moysey 1997). This study used colour-banded birds to enable recognition of individuals which had been previously sexed, as L. melanops cassidix is monomorphic (males and females appear alike). While not providing evidence of competition, the results are consistent with the idea that competition is important within honeyeater communities and in this case intraspecifically. Recovery suggestions highlighted the importance of various plant species and structures in the re-establishment of L. melanops cassidix habitat (Moysey 1997). There is also good evidence of sexual niche partitioning as a result of intraspecific competition in one species of declining woodland honeyeater in the Mt Lofty Ranges. Phylidonyris pyrrhoptera females have been shown to alter their behaviour after the removal of P. pyrrhoptera males by foraging more in the upper stratum. No corresponding change in behaviour followed the removal of P. pyrrhoptera females (Foster 2001).

Of the hypotheses suggested by Ford et al. (2001), the most likely to effect honeyeaters are the disproportionate loss of some habitats and interspecific competition. Nectar feeding birds often use resources from a range of habitats, depending on which plant species flower well in a particular year (Keast 1968b; Mac Nally and McGoldrick 1997; Paton et al. 1999; Mac Nally and Horrocks 2000; Paton 2000; Paton et al. 2004b). The disproportionate clearance of one or more of these habitats is likely to increase the chance of a lack of resources occurring. Xanthomyza phrygia is suggested as one species which historically visited a succession of plant species in different areas but is now limited by the clearance of habitats containing those plant species (Ford et al. 1993; Ford et al. 2001). The probable importance of interspecific competition has already been discussed, and is well documented regarding miners (Section 1.1).

7 There is a need for further research to clarify the primary causes of decline in woodland birds, a process that will out of necessity have an autecological focus, at least in the initial stages. While this is an important step, it is by itself not enough. Despite studies lasting at least two years finding that populations are not limited by breeding success in both X. phrygia (Geering and French 1998; Oliver 1998b) and L. melanops cassidix (Franklin et al. 1995), the effects of the recent long drought in eastern Australia, has resulted in considerably lower breeding success in both species (Birds Australia 2004a; b). This sort of ecosystem perturbation, normal in Australian systems and most relevantly measured in time scales approaching decades, does devalue somewhat the results of most studies investigating the ecology of declining species due to the more limited time frame of the studies. How would the results of, say, Oliver (2001) change if the study had occurred over a time frame including a severe drought, and, more importantly, what are the long term effects on X. phrygia populations of such droughts? The studies outlined above investigating the autecology of declining woodland birds are important in highlighting the continual, ongoing factors facing declining populations but they must be undertaken in concert with appropriately designed, regional scale monitoring of populations. Without such monitoring, the effects of large scale perturbations, such as prolonged drought will not be detected, as demonstrated by Birds Australia (2004a; 2004b). It is probably a combination of continual, ongoing factors causing decline and occasional large scale perturbations that lead to so called ‘relaxation’ or ‘time lag’ effects (e.g. Diamond 1975; Mac Nally and Horrocks 2002; Paton et al. 2004b) of regional population extinction following large scale habitat clearance.

1.3 Restoration ecology

There is an increasing effort devoted to restoring ecosystems throughout Australia (e.g. Recher 1993; Saunders et al. 1993; Sisk and Margules 1993; Yates and Hobbs 1997), including woodland habitats in the Mt Lofty Ranges (Ellis 2000; Paton et al. 2004b). This restoration takes many forms and has many perceived goals from agricultural management (farmers planting salt tolerant, non-local species as stock fodder in salinity degraded areas) to biodiversity conservation (attempts to re-establish an area with local native species in a way that reflects the original habitat diversity and structure). However, there is little to no information on how to undertake restoration projects that will best achieve biodiversity conservation goals (e.g. planting densities, species

8 diversity). Partly due to this lack of information, and partly due to the land available, restoration projects are typically small, and are located on the poorest soils and consist of plantings of limited numbers of species (Paton 2000; Paton et al. 2004b). The sort of autecological studies advocated in the previous section are necessary to enable better understanding of what is required in restoration projects to best meet the requirements of declining woodland bird species.

1.4 The Mount Lofty Ranges

The Mt Lofty Ranges occur in South Australia between 138°0’ and 139°0’ E and 34°20’ and 35°40 S. The region experiences a temperate Mediterranean climate of hot, dry summers and cool, wet winters. Annual rainfall varies from about 300 mm on the eastern edges of the ranges near the Murray River up to over 800 mm in the highest elevations (Figure 1.1). The area of the region is about 350,000 ha, about 80% of which is used for primary production and less than 15% is remnant native vegetation (Long 1998; Paton et al. 1999).

Due to the altitude of the ranges themselves (Figure 1.1), the region experiences quite a different climate to similar latitudes elsewhere in South Australia and because of this are surrounded by quite different vegetation. To the east, southeast and north in the past there was mostly mallee but it is now mostly agricultural land. To the west and south there is sea. Although nearby Kangaroo Island has some relatively large areas of woodland remaining, its importance to mainland bird species has been debated (Abbott 1974; Ford and Paton 1975; Abbott 1976). The Mt Lofty Ranges (± Kangaroo Island) therefore form an isolated patch of woodland and forest in a generally dry part of Australia.

9 Figure 1.1: Annual mean rainfall isohyets and altitude map of the Mt Lofty Ranges

Scale: 1cm = 30 km. Both maps from Paton et al. (1994b).

1.4.1 Vegetation Today the dominant eucalypt species in the Mt Lofty Ranges consist of stringybark (messmate stringybark E. obliqua and brown stringybark E. baxteri), South Australian blue gum (E. leucoxylon), pink gum (E. fasciculosa), red gum (E. camaldulensis), manna gum (E. viminalis), cup gum (E. cosmophylla) and boxes (long-leaved box E. goniocalyx, grey box E. microcarpa and peppermint box E. odorata) (Armstrong et al. 2003). There are also a few mallee species that are mainly peripheral to the Mt Lofty Ranges, but do occur within the region, particularly coastal white mallee (E. diversifolia), narrow-leaf red mallee (E. leptophylla) and ridge-fruited mallee (E. incrassata) (Armstrong et al. 2003).

Originally, the region was dominated by savannah woodlands, a formation most common on fertile soils and having a grassy understorey. Dominant savannah woodland eucalypts included E. leucoxylon, E. fasciculosa, E. viminalis, E. odorata, E. microcarpa and E. camaldulensis. All these species require the most fertile soils within the Mt Lofty Ranges and now occur only in remnants that are few in number, small and usually degraded (Paton et al. 1999; Armstrong et al. 2003; Paton et al. 2004b), or in sub-optimal habitat (particularly E. fasciculosa). Because of their occurrence on fertile

10 soils, these areas are particularly susceptible to weed invasion. Eucalyptus viminalis, E. leucoxylon and E. camaldulensis woodlands were associated with the greatest number of weeds in a recent survey of the region (Armstrong et al. 2003). Blackberries (Rubus spp.) are particularly important in the understorey of areas dominated by E. viminalis, and olives (Olea europaea) are important in many areas dominated by E. leucoxylon (pers. obs., Armstrong et al. 2003). Other widespread weeds include bridal creeper (Myrsiphyllum asparagoides), pine (Pinus spp.), cape broom (Genista monspessulana) and gorse (Ulex europaeus). Despite their already widespread occurrence, all the above weeds are not thought to have reached their maximum possible distribution in the Mt Lofty Ranges (Armstrong et al. 2003).

Today, clearance of the savannah woodlands for agriculture and urban areas has seen the dry sclerophyll forests become the dominant vegetation formations in the region. These areas are dominated by E. fasciculosa, E. obliqua, E. cosmophylla, E. baxteri and E. goniocalyx. Most of this vegetation occurs on infertile soils and/or high rainfall areas in the higher elevations of the region on land least suited to agriculture (Long 1998; Paton et al. 1999; Armstrong et al. 2003; Paton et al. 2004b). These areas often contain a dense sclerophyllous understorey of relatively constant composition, with Xanthorrhoea semiplana, Acacia pycnantha, Lepidosperma semiteres, Platylobium obtusangulum, Leptospermum myrsinoides, Hakea rostrata, Acacia myrtifolia, Pultenaea daphnoides and Acrotriche serrulata particularly common and widespread (Armstrong et al. 2003).

The plants producing nectar taken by birds in the Mt Lofty Ranges are listed by Ford (1977; 1979). Of the eucalypts, E. leucoxylon, E. cosmophylla and E. fasciculosa are probably the most widely used by honeyeaters. The mistletoes, Amyema miquelii and Lysiana exocarpi are also important as they provide nectar resources during the late summer and autumn period when nectar is most likely to be limiting. Banksia ornata and Banksia marginata produce copious amounts of nectar, are important seasonally to honeyeaters, and occur as understorey in stringybark forest and woodland and as dominant species in heath on sandy soils. A series of smaller shrubs also produce nectar and are visited frequently by birds including Epacris impressa (forest), Brachyloma ericoides, Astroloma conostephioides, Grevillea lavandulacea, G. ilicifolia, Callistemon spp. and some Correa spp. (all heath and woodland).

11

The Mt Lofty Ranges have suffered some of the highest rates of native vegetation clearance in Australia. Overall the region retains less than 15% of its native vegetation (Long 1998; Paton et al. 1999; Armstrong et al. 2003; Paton et al. 2004b). The problems of habitat loss are further exacerbated by the preferential clearance of woodlands in areas best for agriculture, leaving the remaining woodlands on rocky ridges, steep gullies and poor soils (Robinson and Traill 1996; Paton et al. 1999; Armstrong et al. 2003; Paton et al. 2004b). In the Mt Lofty Ranges, below 150 metres, there is only 2.2% of native vegetation remaining but this increases to 16.6% above 450 metres (Paton et al. 2004b). Overall, about 13% of the Mt Lofty Ranges remains uncleared (Armstrong et al. 2003).

1.4.2 Birds Due to changes in the habitats and vegetation types of the region, predominantly as a result of vegetation clearance since European settlement, the Mt Lofty Ranges have one of the largest concentrations of threatened bird taxa in Australia. Paton et al. (1999) summarise the woodland bird species for which there is documentation of a decline in abundance and/or distribution (Table 1.1). Nine species of bird are no longer seen in the Mt Lofty Ranges (Ford and Howe 1980). A further eight have undergone a decline in distribution and abundance (Paton et al. 1994a; Chapman 1995) while another six have undergone a decline in abundance (Paton et al. 1994a). Besides those birds that have undergone a documented decline in distribution and/or abundance, there are about another 35 bird species that are recognised as declining woodland bird species (Attwood and Cale 2002), based on the experiences of local ornithologists. This class of declining woodland birds are suspected of undergoing a continual decline despite cessation of vegetation clearance in 1983 (with the first regulation of native vegetation clearance being implemented at that time under the Planning Act 1982). Therefore, simple vegetation clearance is not the whole story and it appears likely some other process(es) is acting to cause the continued decline of birds within the region. The loss of habitat may result in loss of a certain percent of the component species as the local biota ‘relaxes’ to a new equilibrium (e.g. Diamond 1975; Mac Nally and Horrocks 2002). In the Mt Lofty Ranges this relaxation has been predicted to cause the regional extinction of a further 25-40 bird species after an unknown lag period (Ford and Howe 1980; Paton et al. 2004b). While it seems likely that these predictions will be fulfilled if no

12 action is taken, there is the chance to reverse the declines with careful monitoring, research and reconstruction of woodlands. It is therefore important to undertake autecological studies of declining woodland bird species in order to manage their remaining populations and prevent their regional extinction, but also to acquire the knowledge necessary to provide declining birds with additional habitat through reconstructing large areas of habitat (Paton et al. 2004b).

Table 1.1: Declining woodland birds of the Mt Lofty Ranges

Summary from Paton et al. (1999). Species Common Name No longer seen Alcedo azurea Azure Kingfisher Burhinus grallarius Bush Stone-curlew Calyptorhynchus lathami Glossy Black-Cockatoo Cinclosoma punctatum Spotted Quail-thrush Coturnix chinensis King Quail Coturnix ypsilophora Brown Quail Lathamus discolor Swift Parrot Turnix pyrrhothorax Red-chested Button-quail Xanthomyza phrygia Regent Honeyeater Decline in distribution and abundance Acanthiza nana Yellow Thornbill Aphelocephala leucopsis Southern Whiteface Climacteris picumnus Brown Treecreeper Melanodryas cucullata Hooded Robin Melithreptus gularis Black-chinned Honeyeater Microeca fascinans Jacky Winter Myiagra inquieta Restless Flycatcher Stagonopleura guttata Diamond Firetail Decline in abundance Artamus cyanopterus Dusky Woodswallow Falcunculus frontatus Crested Shrike-tit Hirundo nigricans Tree Martin Pachycephala rufiventris Rufous Whistler Petroica multicolor Scarlet Robin Psephotus haematonotus Red-rumped Parrot

1.4.3 Honeyeaters Table 1.2 provides a summary of the honeyeater species found in the Mt Lofty Ranges. The relative abundance (% of records within a study) is taken from two studies. Study 1 is banding records of honeyeaters from three sites within the Mt Lofty Ranges; the Monarto State Forest area, Newland Head Conservation Park area and an area encompassing Hale & Cromer Conservation Parks (Paton and Bradley, unpublished data). Study 2 is the study of Ford and Paton (1977) in which sixteen sites were visited

13 twice a month for a year to count honeyeaters. For Study 1, the resident honeyeaters are separated from those that can be considered to have their primary habitat peripheral to the Mt Lofty Ranges (together these ‘peripheral’ species total 2% of all the honeyeater records from the three sites). The beak lengths, weights and residency status are also given and Manorina melanocephala (Noisy Miner) is included in the table as it is locally common in degraded woodland areas of the Mt Lofty Ranges, especially roadsides.

The two smaller Melithreptus are both short-beaked (true culmen about 14 mm). Melithreptus brevirostris is recorded consistently at about 3% of the honeyeater population, while the recording rate for M. lunatus is less consistent (between 4% and 8% of the honeyeater population). Melithreptus gularis is now recorded very infrequently, and is similar to L. penicillatus in being short-beaked (true culmen about 16.5 mm) and about 20 grams. The recording rate for L. penicillatus is not consistent between the two studies in Table 1.2.

The fate of the resident Mt Lofty Ranges honeyeater species has been varied since European settlement (Table 1.3). Both P. novaehollandiae and Anthochaera carunculata remain widespread and abundant in the region. There is a suite of declining woodland honeyeater species, including two of the Melithreptus (Attwood and Cale 2002). Melithreptus gularis is listed by the South Australian National Parks and Wildlife Act 1972 as Vulnerable and X. phrygia is regionally extinct. In the Mt Lofty Ranges, M. gularis deserves a more critical status than Vulnerable. The current population of M. gularis in the Mt Lofty Ranges is estimated at between 50 and 100 individuals, with two studies in the last ten years arriving at similar estimates of the population (Chapman 1995; Paton 2002). The status of the other two Melithreptus species as declining woodland birds is probably appropriate.

14 Table 1.2: Honeyeaters of the Mt Lofty Ranges; relative abundance (% of records), weight and bill length

% Records are provided from two studies to give an indication of relative abundance. Study 1 is an unpublished database (Paton and Bradley) of banding records from three sites: Monarto State Forest; Cromer and Hale Conservation Parks; and Newland Head Conservation Park. ‘Peripheral species’ records in total were 2% of resident records. Study 2 is Ford and Paton (1977). Weight is from Paton and Bradley (unpublished database) and bill length (true culmen) is from Higgins et al. (2001), except M. melanocephala which is from Paton and Collins (1989). The Paton and Bradley database also contains one record of a Fuscous Honeyeater (Lichenostomus fuscus) at the three sites summarise here.

Study Weight Bill length Species Common Name 1 2 (grams) (mm) Long-billed Resident Species Acanthorhynchus tenuirostris Eastern Spinebill 5.3 2.3 10.5 25.4 Anthochaera carunculata Red Wattlebird 3.8 11.0 106.4 26.2 Anthochaera chrysoptera Little Wattlebird 0.5 0.0 63.3 29.9 Gliciphila melanops Tawny-crowned Honeyeater 1.0 0.5 19.0 22.1 Phylidonyris novaehollandiae New Holland Honeyeater 64.4 39.1 20.2 22.8 Phylidonyris pyrrhoptera Crescent Honeyeater 6.9 7.4 14.4 21.4 Short-billed Resident Species Lichenostomus chrysops Yellow-faced Honeyeater 8.7 13.6 16.4 16.2 Lichenostomus penicillata White-plumed Honeyeater 1.1 13.9 20.0 16.2 Manorina melanocephala Noisy Miner 0.0 0.0 68.0 22.1 Melithreptus brevirostris Brown-headed Honeyeater 3.6 3.1 14.2 13.6 Melithreptus gularis Black-chinned Honeyeater 0.0 0.6 20.8 16.7 Melithreptus lunatus White-naped Honeyeater 4.6 8.4 13.8 14.7 Peripheral Long-billed Species Acanthagenys rufogularis Spiny-cheeked Honeyeater 9.2 46.2 26.2 Phylidonyris albifrons White-fronted Honeyeater 5.3 17.4 20.9 Peripheral Short-billed Species Lichenostomus cratitius Purple-gaped Honeyeater 2.7 19.7 18.4 Lichenostomus leucotis White-eared Honeyeater 0.4 18.7 18.1 Lichenostomus ornatus Yellow-plumed Honeyeater 43.2 17.7 15.6 Lichenostomus virescens Singing Honeyeater 39.2 27.9 20.2

15 Table 1.3: Current status of resident honeyeater species in the Mt Lofty Ranges

The status listed for each honeyeater is based on a number of sources including survey data, personal communication, personal observations and references (Paton et al. 1994a; Chapman 1995; Paton et al. 1999; Attwood and Cale 2002; Paton 2002). Status Species increasing Manorina melanocephala widespread and abundant Anthochaera carunculata Phylidonyris novaehollandiae locally common Lichenostomus penicillatus declining Acanthorhynchus tenuirostris Anthochaera chrysoptera Gliciphila melanops Lichenostomus chrysops Melithreptus brevirostris Melithreptus lunatus Phylidonyris pyrrhoptera endangered Melithreptus gularis extinct Xanthomyza phrygia

1.4.3.1 Seasonal fluctuation A seasonal fluctuation in reporting rate of honeyeater numbers within the Mt Lofty Ranges has been recorded. An increase occurred in April and May, with a decrease in September to October (Ford 1977; Ford and Paton 1977). Across sixteen sites, surveyed twice each month for a year, up to 93% of records for some species were taken in winter, which was defined as April to September (Table 1.4, Figure 1.2). The observed seasonal fluctuation may reflect: • partial migration of honeyeaters into and out of the Mt Lofty Ranges; • partial migration of honeyeaters between localities within the Mt Lofty Ranges; • an actual seasonal fluctuation in numbers; and/or • a seasonal fluctuation in detectability. There is some evidence of a partially migrating M. lunatus population, but little evidence for any migration in M. brevirostris or M. gularis (Section 1.5.6).

16 Table 1.4: Percent of year records occurring in April-September (winter) for each species of honeyeater recorded by Ford and Paton (1977)

Presented in increasing order of percent of records taken in winter. Species % of records occurring in winter Lichenostomus penicillata 65 Acanthorhynchus tenuirostris 67 Phylidonyris pyrrhoptera 71 Phylidonyris novaehollandiae 73 Melithreptus brevirostris 74 Anthochaera carunculata 75 Melithreptus gularis 78 Gliciphila melanops 81 Melithreptus lunatus 81 Lichenostomus chrysops 93

Figure 1.2: Number of individuals within ten species of honeyeater recorded in October-March (summer) and April-September (winter) at sixteen sites in the Mt Lofty Ranges by Ford and Paton (1977)

4000 3500

ed 3000 d r 2500 co Summer

Re 2000

er Winter

b 1500

Num 1000 500 0

s s s s s a s p u ri u ri ae ps ri ta o t t a er t la s la st a l di pt no s u ry l ro un o a ro c h ici i l gu lan h i n ev . l u u . c en r M M. ho rr mel en r L p e t . b a . py G. c. L. M ov P A n An. ca P. Species

1.5 Melithreptus

The Melithreptus genus is an Australasian endemic genus with six species that between them occupy most of Australia, except for the very far north of Cape York and Arnhem Land and the centre of Western Australia (Table 1.5). Keast (1968a) published a study of the genus and suggested that it consisted of three species groups: 1) M. lunatus – M. affinis – M. albogularis

17 2) M. brevirostris 3) M. gularis – M. validirostris

The members of the first two groups are small compared with the last group. Keast’s reasoning for the distinction between the first two groups is not so clear. Generally, Keast (1968a) found that two species of Melithreptus occurred together, with one of the species being from group 3 and the other species from group 1 or 2, i.e. a larger and a smaller species. The larger species was suggested to be less specialised in their choice of feeding zone, while the smaller species was suggested to be primarily a foliage or bark gleaner. However, as even Keast (1968a) pointed out, there is a large area of south eastern Australia over which M. brevirostris and M. lunatus are sympatric breeders, stretching from Toowoomba to the Mt Lofty Ranges. While Keast (1968a, pg 767) suggested that the, ‘overlap is sufficiently great to make ecological separation advantageous as a means of restricting competition’, he was unable to elaborate what differences did occur to enable ecological separation other than the statement, ‘Where the smaller M. lunatus and M. brevirostris cohabit in eastern Australia they differ in the latter having a shorter wing, and hence larger bill/wing and tarsus/wing ratios (Keast 1968a, pg 767)‘. Keast does not suggest what such a difference would mean for the two species in sympatry. Thus, the most important work on Melithreptus fails to adequately explain the sympatry of two species that are very closely related in morphology, feeding behaviour and distribution. Other studies have addressed the similarity between M. brevirostris and M. lunatus, but without the focus of Keast (1968a) (particularly Ford and Paton 1977; Recher et al. 1985; and Ford et al. 1986). The findings of these studies are discussed in the following sections where appropriate.

Melithreptus have a patch of coloured, bare skin above the eye, also known as a wattle (Higgins et al. 2001). The wattle has been suggested to have an important role in preventing hybridisation within members of Melithreptus (Keast 1961), especially within members of Keast’s (1968a) group 1.

18 Table 1.5: Distribution of the species of the Melithreptus genus

From Keast (1968a). Species Common Name Distribution Melithreptus gularis Black-chinned Honeyeater Eastern Australia including South Australia. includes M. gularis laetior Golden-backed Honeyeater Northern Australia M. validirostris Strong-billed Honeyeater Endemic to Tasmania and Bass Strait Islands M. brevirostris Brown-headed Honeyeater Southern & Eastern Australia including South Australia M. albogularis White-throated Honeyeater New Guinea (Ford, pers. comm..) and Northern Australia M. lunatus White-naped Honeyeater East, Southeast & Southwest Australia including South Australia M. affinis Black-headed Honeyeater Endemic to Tasmania and Bass Straight Islands

1.5.1 Melithreptus in the Mt Lofty Ranges Historically all three species were frequently encountered in a number of areas, such as Para Wirra RP (Ford and Paton 1976a), Sandy Creek Conservation Park (Rix 1976b), Scott Conservation Park (Paton and Paton 1980) and, in the case of M. lunatus and M. gularis, even across the Adelaide Plains (White 1914; Crompton 1915; South Australian Ornithological Association 1915).

1.5.1.1 Melithreptus brevirostris Melithreptus brevirostris (Figure 1.3) is about the same size as M. lunatus, but can be distinguished from M. lunatus by the lack of a black ‘cap’ on the head and lack of red wattle above the eye. Melithreptus brevirostris occurs throughout the Mt Lofty Ranges and also in the surrounding areas of mallee (Figure 1.4).

Rix (1976a; 1976b) visited Sandy Creek Conservation Park 71 times over a ten year period from 1962, making notes on all the bird species encountered during each visit. At Sandy Creek Conservation Park, M. brevirostris was thought to be resident, with a party of eight to ten birds seen on most visits to the park, although in at least some years, numbers appeared reduced over the period December to April. There were also some years in which a marked influx occurred (Rix 1976b). Melithreptus brevirostris was thought to be a common resident of Para Wirra Recreation Park, although was not recorded breeding in the park (but was recorded breeding just outside the park), and appeared to have maintained its status in the park between the mid 1960’s and mid 1970’s (Ford and Paton 1976a). At Belair Recreation Park, M. brevirostris was considered transient, with irregular sightings during winter in the woodland of the northwest corner of the park (Baxter 1980). In Scott Conservation Park M. brevirostris

19 was considered probably resident and breeding (with several birds re-trapped over periods of one to three years), but was uncommon within the park (Paton and Paton 1980). In the higher elevations of the Mt Lofty Ranges, M. brevirostris appears to be present in only small numbers (e.g. Laybourne-Smith 1989).

Some of the most specific information on the natural history of M. brevirostris comes from a colour-banding study of M. brevirostris near Armidale, New South Wales (Noske 1983). The birds in this study both bred and roosted communally over two years. Roosting behaviour was described in which as many as eight individuals huddled together along a thin twig and/or petiole amongst the foliage usually near the tip of a eucalypt. Often, following a struggle, the birds would ‘dismantle’ before reforming at the same site. Individuals each faced different directions and Noske (1983) thought this may aid in packing tightly into a limited space while allowing the group a wide field of vision for predator avoidance. The roosting behaviour noted is also likely to provide improved thermal regulation for the group over roosting individually.

1.5.1.2 Melithreptus lunatus Melithreptus lunatus (Figure 1.3) can be distinguished from other similar honeyeaters by their white nape which does not extend to the eye and their orange wattle above the eye (white in western subspecies, redder in the eastern subspecies) which is also present in juveniles. Melithreptus lunatus occurs throughout the Mt Lofty Ranges (Figure 1.4).

There have been few studies noting natural history information on M. lunatus. Most information comes from annotated bird lists for various regions/parks. Rix (1976b) made notes on M. lunatus at Sandy Creek Conservation Park. He found that the extent of flowering of Astroloma influenced the numbers using the park, with numbers of 40 to 50 in average years but with up to 200 or more in years of extensive Astroloma flowering. Melithreptus lunatus was also a seasonal visitor to the park, with the highest numbers of M. lunatus occurring between May and October. The majority left the area by spring, with rarely more than two or three pairs remaining. One nest was found in the park in mistletoe on pink gum (E. fasciculosa). Melithreptus lunatus was thought to be a frequent winter visitor to Para Wirra Recreation Park, and appeared to have maintained its status in the park between the mid 1960’s and mid 1970’s (Ford and Paton 1976a). In Belair Recreation Park, M. lunatus was considered a breeding resident most common in

20 the stringybark forest areas of the park (Baxter 1980). At Scott Conservation Park, M. lunatus was considered a resident breeder, more common in winter than summer with numbers building up around April and declining around September (Paton and Paton 1980). Melithreptus lunatus appears to be found in high numbers in the higher elevations of the Mt Lofty Ranges (e.g. Laybourne-Smith 1989).

1.5.1.3 Melithreptus gularis Melithreptus gularis (Figure 1.3) can be distinguished from other similar honeyeaters by its black chin, white nape that extends all the way to the eye and blue wattle above the eye. There is a northern Australian subspecies of M. gularis which has a green wattle. Melithreptus gularis is larger than both M. brevirostris and M. lunatus, weighing about 20g (Ford 1979). Melithreptus gularis occurs mainly around the foothills of the Mt Lofty Ranges (Figure 1.4, Figure 1.5).

In the 1960’s M. gularis was considered resident at Sandy Creek Conservation Park. Rix (1976b) noted that the call of M. gularis was considerably more prominent than either M. brevirostris or M. lunatus and because of this was often recorded within a few minutes of entering the park. Rix also noted a song and deep notes which were soft in comparison to the prominent calls. The population appeared to vary from year to year between one or two pairs to about ten pairs. One individual was recorded in Sandy Creek Conservation Park in 1992, two individuals in 1994 (Chapman 1995) and one individual since (P. Paton, pers. comm.). Melithreptus gularis was thought to be a frequent, breeding resident of Para Wirra RP in the mid 1970’s, apparently improving in status from occasional to frequent within the park between the mid 1960’s and mid 1970’s (Ford and Paton 1976a). However, the last (unconfirmed record) of M. gularis in Para Wirra RP was in 1994 and the last breeding record was in 1989 (Chapman 1995). Ashby (1936) recorded M. gularis as ‘almost as common’ as M. lunatus within Belair National Park, but M. gularis was not recorded in Belair at all by Baxter (1980). In the late 1970’s, M. gularis was considered scarce but breeding at Scott Conservation Park with a small flock of four to six occasionally seen, particularly when E. leucoxylon was in flower (Paton and Paton 1980). Scott Conservation Park remains one of the few sites in the Mt Lofty Ranges where M. gularis can be seen with some regularity. These records and others (particularly Chapman 1995; Paton 2002) clearly document the decline of M. gularis across the Mt Lofty Ranges.

21

During 2000 and 2001 a project using Natural Heritage Trust funds was run to: • monitor numbers and sub-populations of the Black-chinned Honeyeater Melithreptus gularis in the Mt Lofty Ranges (MLR); and • restore native vegetation and revegetate in known stronghold areas for the Black-chinned Honeyeater along the lower Inman River and near Scott Conservation Park (Paton 2002).

One outcome of this project was the production of a map of currently known extant, extinct and uncertain M. gularis populations (Figure 1.5). This map also suggests that M. gularis is currently using the foothills of the Mt Lofty Ranges and points to four confirmed remaining populations of M. gularis within the Mt Lofty Ranges. These four populations are in the vicinity of Scott Conservation Park, Victor Harbor township, suburban Morphett Vale and Altona (Paton 2002). There is also possibly an ‘eastern’ population between Woodside and Tepko.

22 Figure 1.3: a) M. brevirostris Brown-headed Honeyeater. b) M. lunatus White-naped Honeyeater. c) M. gularis Black-chinned Honeyeater

a) M. brevirostris. Photo: Nigel Willoughby

b) M. lunatus. Photo: Nigel Willoughby

c) M. gularis. Photo: David Paton

23 Figure 1.4: Distribution of a) M. brevirostris, b) M. lunatus and c) M. gularis in the Adelaide Region

Black dots are breeding records; white dots are all other records. From Paton et al. (1994b). Scale: 1 cm = 30 km.

a) M. brevirostris b) M. lunatus

c) M. gularis

24 Figure 1.5: Results from a 2000 and 2001 project to document the current distribution of M. gularis in the Mt Lofty Ranges

From Paton (2002). ‘Uncertain’ status refers to locations fairly close to core areas where only one or two records of a small number of birds were made.

1.5.2 Breeding It is possible that Melithreptus in the Mt Lofty Ranges breed, if conditions are suitable, at any time of the year, but breeding is most likely to occur from mid-winter through to

25 mid-summer (Table 1.6). Nests are usually built in the drooping, outer foliage of, primarily eucalypts, but also mistletoes and occasionally other species. Clutch size is usually two or three eggs, and incubation is approximately 14 days (Higgins et al. 2001). While no studies have been made on Melithreptus breeding, studies on other declining honeyeater species in south-eastern Australia have provided good evidence that reproductive performance is generally not limiting the ability of current populations to maintain themselves (Franklin et al. 1995; Geering and French 1998; Oliver 1998b).

One interesting aspect of Melithreptus nest building is the use of fur from live animals, apparently with a preference for white hair (summarised in Higgins et al. 2001). In the southern suburbs of Adelaide, M. gularis was observed collecting wool from a rug and dog fur pegged to a clothesline (Elsworth 1997). Perhaps the decline in M. gularis can even be partially attributed to this habit, if the fur collecting behaviour noted by White (1914) is (was) at all common, ‘They (M. gularis) persecuted the cat so severely that it often cried out with rage, and may be pain’.

Harrison (1969) provides evidence that M. brevirostris, M. lunatus and M. gularis often have helpers at the nest other than the parents. This ‘communal’ breeding or cooperative breeding has been noted by a number of other authors. Noske (1983) provides evidence of the occurrence of communal breeding in M. brevirostris, although only on a single nest. Observations on the nest suggested that only the female directly participated in nest building and maintenance, but other birds fed the female on a number of occasions. There were five birds in the group, the ‘primary pair’ (presumably the parents), two other adults and a juvenile. Colour-banded birds were sexed by behaviour during copulation. All birds except the juvenile were seen to incubate the nest. The primary pair incubated for 73% of the time, their contributions being nearly equal, while the least contribution by an individual was 3.5% (in 5 sittings). The female of the primary pair was fed on the nest by other birds a number of times, but no other birds were observed being fed while sitting. The juvenile rarely visited the nest tree but usually fed with the other birds. One other nest was briefly observed, during which time four separate adult birds were seen to feed nestlings. There is also evidence of cooperative breeding in M. lunatus (Dow 1980; Noske 1983; Higgins et al. 2001). Higgins et al. (2001) summarise records of five, ‘most of a flock of twelve’ and three adults feeding a fledgling. However, it appears likely that only female birds build and incubate nests

26 (Noske 1983; Higgins et al. 2001), although there is one record of two adults building a nest (Higgins et al. 2001). Cooperative breeding is said to occur regularly in M. gularis, with up to five adults assisting (Higgins et al. 2001). However, there have been no studies focussing on Melithreptus to determine the actual incidence of communal breeding, only anecdotal reports and their summary. Given this limitation, it appears that M. brevirostris, M. lunatus and M. gularis are facultative cooperative breeders, although each to a different extent. It appears likely that in all Melithreptus, the female is generally responsible for nest building, although she is often accompanied by the male (Noske 1983; Elsworth 1997; Higgins et al. 2001). However, in M. lunatus, it appears the female is also usually responsible for incubating eggs and nestlings, while in M. brevirostris, it appears that many group members assist the female in incubating. In each of the species, feeding nestlings and juveniles is done by both parents and sometimes other group members. The incidence of communal breeding has not been established.

While it has been suggested that helping at a nest is generally a second rate investment compared to immediate independent breeding (Hatchwell and Komdeur 2000), there are numerous hypotheses to account for helping behaviour in cooperatively breeding birds including: increased production of collateral kin; payment-of-rent, allowing access to other benefits; access to shared mating opportunities; improvement of local conditions; establishment of strategic alliances; improved skills (Cockburn 1998); and potential inheritance of the breeding territory. Possible reasons for cooperative breeding in Melithreptus have not been addressed.

Table 1.6: Approximate breeding season of Melithreptus occurring in the Mt Lofty Ranges

Source 1: Simpson & Day (1999). Source 2: Higgins et al. (2001) summary of the Birds Australia Nest Record Scheme. Species Start Finish Source Notes M. brevirostris July December 1 Throughout range August January 2 Mt Lofty Ranges only M. lunatus July January 1 Throughout range August January 2 Mt Lofty Ranges only M. gularis April January 1 Throughout range July November 2 Also April. Mt Lofty Ranges only

27 1.5.3 Food All honeyeaters are likely to consume both nectar and insects. Some species of honeyeater rely heavily on nectar, and follow the flowering of key food plants across the landscape, while other species use nectar when available locally. Studies on foraging of honeyeaters have found that, where available, honeyeaters use nectar resources, but in the absence or limitation of nectar, spend much of their foraging time gleaning leaves or bark and probing under bark (e.g. Paton 1980; Collins and Briffa 1982; Ford et al. 1986; Mac Nally 1996; Mac Nally and McGoldrick 1997; Wilson and Recher 2001). This foraging behaviour is predominantly used to collect non-nectar carbohydrates which are available in a number of forms: lerp, manna and honeydew (Paton 1980). The importance of these non-nectar carbohydrate foods to honeyeaters (and other birds) has since been supported by a number of further studies (Ford and Paton 1985; Recher et al. 1985; Ford et al. 1986; Woinarski et al. 1989; Recher et al. 1991; Recher et al. 1996; Moysey 1997; Oliver 1998a; 2000). The rewards available to P. novaehollandiae from various food resources are given in Table 1.7 and it is clear that non-nectar carbohydrates provide favourable rewards, especially later in a day when nectar can become depleted. While the Melithreptus are often referred to as insectivorous (e.g. Keast 1968a; Ford and Paton 1977; Higgins et al. 2001), many observations of insect feeding are perhaps more likely to have been feeding on non-nectar carbohydrates.

Nectar is at least occasionally depleted to levels inadequate to meet the energetic requirements of the resident honeyeater population (Ford 1977; 1979; McFarland 1986a; Armstrong 1991). Similar patterns of competition for and depletion of non- nectar carbohydrate resources have also been found (Paton 1980; Moysey 1997).

Pollen is not considered important in the diet of most honeyeaters. It is not digested easily by the non-specialist digestive systems of most birds (Wooller et al. 1988) and even if fully digested, the amounts ingested would provide only a small proportion of the protein requirements of the birds and a negligible amount of energy (Paton 1981).

28 Table 1.7: Net energy gain obtained by foraging P. novaehollandiae from various food sources

Units are Joules/min, except pollen (Joules/day). Net energy gain Resource (J/min) Max/Min Relates to: Source Max Min : Insects 29 -13 Dawn/Dusk Paton (1982b) Manna 11 Only 1 estimate Paton (1980) Nectar 2300 30 Dawn/Dusk Paton (1982b) Pollen 170 J/day Paton (1981) Psyllid secretions 23 7 Two territories Paton (1980) Nectar of: Amyema pendulum 41 10 Dawn/Dusk Paton (1980) Astroloma conostephioides 63 58 Dawn/Dusk Paton (1980) Banksia marginata 75 5 Dawn/Dusk Paton (1980) Callistemon macropunctatus 50 3 Dawn/Dusk Paton (1980) Epacris impressa 9 4 Dawn/Dusk Paton (1980) Eucalyptus leucoxylon 133 9 Dawn/Dusk Paton (1980) Eucalyptus melliodora 27 0 Dawn/Dusk Paton (1980) Eucalyptus obliqua 26 1 Dawn/Dusk Paton (1980) Grevillea aquifolium 81 16 Dawn/Dusk Paton (1980)

1.5.3.1 Nectar Many species of Australian plant provide nectar for birds, particularly honeyeaters. In return, the plants receive a pollination service, with many birds carrying pollen between flowers (Ford 1977; Paton and Ford 1977; Ford et al. 1979; Paton 1982a; 1986a; b). The genera most frequently visited by honeyeaters include Eucalyptus, Banksia, Callistemon, Adenanthos, Grevillea, Dryandra, Amyema, Lysiana, Correa, Astroloma, Brachyloma, Epacris, Anigozanthos, Eremophila and Xanthorrhoea (Ford et al. 1979). Nectar is generally regarded as containing very little protein (Recher and Abbott 1970; Paton 1980). Extra-floral nectaries of Acacia species are also important to some bird species (Knox et al. 1985; Vanstone and Paton 1988), providing nectar with up to 50% sugars, and perhaps of more importance to honeyeaters, providing a number of amino acids, especially glutamine and phenylalanine (Knox et al. 1985).

There have been numerous studies examining the relationship between nectar and honeyeater density and abundance (e.g. Keast 1968b; Ford 1977; Ford and Paton 1977; Paton 1979; 1980; Collins and Briffa 1982; Paton 1982a; Pyke 1982; 1983; Collins et al. 1984; Ford and Paton 1985; Paton 1985; Collins and Newland 1986; McFarland 1986b; c; Pyke and Recher 1986; Pyke and Recher 1988; Armstrong 1991; Pyke et al. 1993; Mac Nally 1996; Mac Nally and McGoldrick 1997; McFarland 2002). While

29 studies have generally concluded that there is a relationship between nectar availability and honeyeater abundance, this is not always the case. For example, Pyke (1983) found no relationship between variation in honeyeater density in 20m radius circles and energy production or insect biomass within the circles. Collins et al. (1984) found honeyeaters ignored a profitable nectar source (Banksia prionotes inflorescences) at certain times of the year and Pyke (1989) found no effect on honeyeater nesting or total abundance from removal of almost all honeyeater nectar sources from a 5.6 ha area of heathland during the breeding season. There are also cases of intensive flowering that are virtually ignored by honeyeaters (Higgins et al. 2001). Attempts to find a relationship between nectar and honeyeater density or abundance are undoubtedly complicated by honeyeater use of non-nectar carbohydrate resources (Ford and Paton 1985). For example, one study found a negative relationship between numbers of all honeyeaters, M. lunatus in particular, and nectar productivity. However, M. lunatus presence was strongly correlated with the presence of manna (Ford and Paton 1985).

In the Mt Lofty Ranges, nectar was most likely to be limiting in summer and autumn (Ford 1977; 1979). There was more nectar per flower and more flowers in winter and spring than in summer and autumn. Nectar was often depleted, especially in summer and autumn, by honeyeaters and sometimes other visitors (silvereyes, lorikeets and insects) to a level at which it was uneconomical for some species to exploit (Ford 1979). Nectar availability insufficient to satisfy nectarivore community requirements, probably resulting in competition, has been documented elsewhere (Kodrick-Brown et al. 1984; Feinsinger et al. 1985; Paton 1985; McFarland 1986a; Armstrong 1991).

Melithreptus, being short beaked honeyeaters, are most commonly recorded taking nectar from plant species with open flowers, particularly Eucalyptus (Ford 1977; Ford and Paton 1977). Melithreptus lunatus has been recorded feeding at extra-floral nectaries of Acacia pycnantha (Vanstone and Paton 1988).

1.5.3.2 Non-nectar carbohydrates Non-nectar carbohydrates are available in a number of forms. Lerp, manna and honeydew are the most important, although to different extents in different areas (Paton 1980).

30 Manna is an exudate from the injured leaves and branches of certain eucalypts and angophoras. The secretion of manna takes place throughout the year but appears to be most abundant in the spring and summer, when the growth of new leaves is most rapid. Only about one in one hundred injured leaves will secrete manna, suggesting the need for a set of specific conditions to enable production of manna (Basden 1965). Manna from three different plant species consisted of approximately 60% sugar, 16% water and a small amount of ash (oxides) (Basden 1965).

Honeydew and lerp are produced by some members of the suborder Sternorrhyncha of the Hemiptera order of insects (this suborder is one of two now recognised in what used to be combined in the Homoptera). The most important families within the Sternorrhyncha for producing honeydew and lerp are the Psyllidae (psyllids), Aphididae aphids), Eriococcidae (eriococcids) and Coccidae (coccids) (Zborowski and Storey 1998). The food source of each is sap which is ingested through long feeding stylets inserted into the phloem of the host plant (Phillips 1996). In order to meet their protein requirements, a volume of sap is ingested that creates an excess of carbohydrate for the insect. It is this excess carbohydrate which is excreted, usually with considerable alteration, as honeydew or lerp (Gray 1952; Basden 1966). Due to the insect removing protein from the phloem before excretion, honeydew and lerp usually contain very little protein (Basden 1966; 1970), although honeydew with up to 4% protein has been recorded for a South African psyllid (Hodkinson 1974). Many of the Sternorrhyncha live in the shelters they excrete (lerp), which help them to maintain a less variable climatic environment and hide them from predators and parasites (Hodkinson 1974; Phillips 1996; Zborowski and Storey 1998). Those that excrete only honeydew are often attended by other insects, especially ants. Ants have evolved close symbiotic relationships with some Sternorrhyncha, including protecting them from predators, moving them to fresh food sources and taking them into the ants’ nest at night (Zborowski and Storey 1998).

Basden (1968) found the honeydew secreted by Eriococcus coriaceus contained over 30% sugar, although only just over 50% of the dehydrated honeydew was analysed for its chemical composition. Eriococcus coriaceus is widely distributed and found on many different trees but favours eucalypts with blue-green juvenile foliage such as the

31 Tasmanian blue gum (Eucalyptus globulus), the shining gum (E. nitens) and the South Australian blue gum (E. leucoxylon) (Phillips 1996).

Lerp is secreted as honeydew but hardens on contact with air to form the lerp and consists of starch or polymers of glucose (Basden 1966; 1970; White 1972; Woinarski et al. 1989). Some lerp can be made entirely of substances not found in phloem that have been synthesised in the insect from phloem sap (Basden 1966; 1970). Lerps vary enormously, with each species having its own characteristic size, shape and design, such as simple cones, univalves, bivalves or intricately woven basket or fan shapes (Phillips 1996). Honeyeaters often remove lerp without removing the psyllid insect (Paton 1980; Woinarski et al. 1989). However, ‘the gizzard contents of birds shot between 1953 and 1963 revealed that pardalotes, notably Pardalotus ornatus Temminck and Laugier, the White-plumed Honeyeater Meliphaga penicillata Gould, and the Yellow-tailed Thornbill Acanthiza chrysorrhoa (Quoy and Gaimard) fed frequently both on the nymphal and adult stages of C. albitextura’ (Clark 1964a). Cardiaspina albitextura is a lerp forming psyllid common on Eucalyptus blakeyi and was part of a twenty-five year study of the lerp forming psyllids. Predation of the various life stages of C. albitextura by the above birds was thought to have played a dominant role in preventing outbreaks of this psyllid (Clark 1964a; b).

1.5.3.3 Invertebrates Almost all honeyeater species take some invertebrates through either hawking or gleaning from bark and foliage (Higgins et al. 2001). However most honeyeaters, including Melithreptus, take invertebrates primarily to provide protein, not energy, as most other honeyeater foods are poor in protein (Recher and Abbott 1970; Paton 1980). The protein from insects consumed is necessary for maintenance of adult bird health, but also is especially important for growth of nestlings (Pyke 1983). Honeyeaters often expend considerable energy in obtaining invertebrates, which is commonly achieved through hawking in many species (Recher and Abbott 1970; Paton 1982b; 1986b). The energetic rewards for honeyeaters taking insects by hawking are probably in the order of 3 J per minute. At this rate, feeding on insects by hawking is unlikely to meet an individuals daily energy requirements and net energy gains from insect feeding by hawking were sometimes found to be negative (Paton 1979; 1982b; 1986b).

32 One study found no relationship between variation in honeyeater density in 20m radius circles and insect biomass within the circles, suggesting that honeyeaters do not respond to insect density at a fine scale (Pyke 1983) and one study found that breeding was carried out in response to carbohydrate resources, not invertebrate resources (Paton 1979).

1.5.4 Feeding Behaviour Melithreptus honeyeaters feed in the canopies of eucalypts by probing flowers for nectar, probing bark for honeydew and/or insects, gleaning bark for honeydew and/or insects, gleaning foliage for honeydew, lerp, manna and/or insects and occasionally hawking for insects (Keast 1968a; Ford and Paton 1977; Ford 1979; Paton 1980; Recher et al. 1985; Ford et al. 1986; Higgins et al. 2001; Craig 2002).

Given that Melithreptus, and M. brevirostris and M. lunatus in particular, are often recorded in sympatry at a local scale throughout south-eastern Australia, any attempt to explain their sympatry should be applicable at a local scale. In a number of studies the authors note the high degree of ecological similarity between M. brevirostris and M. lunatus and then look to their results to explain the two species’ sympatry (e.g. Keast 1968a; Ford and Paton 1977; Recher et al. 1985; Ford et al. 1986). While, the similarity in feeding behaviour between the two smaller Melithreptus species is often striking, particularly when compared with feeding behaviours of other species, they often appear most different in their use of feeding substrate, as summarised below and in Table 1.8.

Ford and Paton (1977) found the overlap between M. brevirostris and M. lunatus to be about 75%, with the only closer pairing being Crescent Honeyeater (Phylidonyris pyrrhoptera) and Eastern Spinebill (Acanthorhynchus tenuirostris). The latter pairing also had very similar overlap in habitat, but their different morphologies allowed different access to floral resources, P. pyrrhoptera feeding more from Astroloma, while Acanthorhynchus tenuirostris fed more on Epacris. Ford and Paton (1977) found that the foraging behaviour of M. brevirostris and M. lunatus differed with M. lunatus being less of a bark feeder than M. brevirostris.

Recher et al. (1985) found a larger difference in feeding behaviour between M. brevirostris and M. lunatus. Melithreptus brevirostris grouped most closely with

33 Silvereyes (Zosterops lateralis) and M. lunatus most closely with Yellow-faced Honeyeaters (Lichenostomus chrysops), based on data given on their page 414 (used to generate the dendogram in Figure 1.7). The main differences were that M. brevirostris was found to feed more on bark, while M. lunatus fed more on foliage.

Ford et al. (1986) found a high overlap in most aspects of feeding ecology between M. brevirostris and M. lunatus (90% overall similarity, Figure 1.6). There were only two other species pairs with higher similarity in feeding behaviour, although one of those pairs, Gymnorhina tibicen (Australian Magpie) and Stagonopleura guttata (Diamond Firetail), both ground gleaners (Ford et al. 1986), are clearly not competitors. The other pair was Hirundo neoxena (Welcome Swallow) and Hirundo nigricans (Tree Martin). The aspect of feeding behaviour in which the most difference between M. brevirostris and M. lunatus was observed was use of plant species; M. brevirostris were recorded more often in stringybark (Eucalyptus caliginosa) and yellow box (E. melliodora), while M. lunatus were recorded more often in manna gums (E. viminalis). They noted that, ‘in this study the latter species (M. lunatus) fed rather more on bark’.

Mac Nally (1994) found M. brevirostris and M. lunatus were similar in their foraging behaviour, but did not examine more closely any differences. Both M. brevirostris and M. lunatus were considered part of a foliage searching guild, with that guild’s mean foraging by leaf gleaning being 40% and twig gleaning 25%. There were 11 other species in the foliage searching guild (Mac Nally 1994).

Thus, the ratio of bark feeding to foliage feeding exhibited by M. brevirostris and M. lunatus is cited often, and while not explicitly stated in the studies, a review of the literature (Table 1.8) suggests foraging substrate is an important aspect of the species occurrence in sympatry. Taking the mean of all studies and areas (Figure 1.8), M. lunatus spends a greater proportion of time foraging from foliage than bark, while the opposite is true for M. brevirostris. However, a closer examination shows weaknesses in a theory of M. brevirostris and M. lunatus sympatry based on different foraging substrates. In the broad scale studies listed in Table 1.8, the question answered within each study is ‘As a species, what feeding behaviour does M. lunatus/M. brevirostris exhibit?’. While these are legitimate questions in wider studies of a bird community, the question that is of more interest in the areas of sympatry is ‘When feeding in the same

34 areas at the same time, does the feeding behaviour of M. lunatus differ from that of M. brevirostris?’ Perhaps of even more importance is a limitation on food resources while trying to answer this question, something no studies have determined. Melithreptus brevirostris and M. lunatus may be equally efficient while foraging from foliage or bark, depending on which substrate has the most rewarding resources. One study that did look at feeding substrate in relation to available carbohydrate found that M. lunatus changed its feeding substrate in relation to where carbohydrates were available – when manna was the primary carbohydrate, M. lunatus fed on foliage, and when nectar and honeydew were the primary carbohydrates, M. lunatus fed mostly on nectar and under bark (honeydew). Likewise, M. brevirostris fed more on foliage when manna was the primary source of carbohydrate available and more under bark when honeydew was the primary source of carbohydrate available (Paton 1980). Also, the paper with the most limited spatial scale found the most similarity in feeding substrate use by M. brevirostris and M. lunatus (Ford et al. 1986). Therefore it appears that Melithreptus may change their feeding substrate in relation to available resources (Ford et al. 1986). This substantially weakens any theory that where M. brevirostris and M. lunatus occur in sympatry they use different feeding substrates to prevent competition.

An interesting note on this matter is that the results of one study that may have been able to shed some light on the bark versus foliage feeding of M. brevirostris and M. lunatus, Cullen (1983), are not available. This work was presented at a conference some time in the 1980’s, and the abstract for the talk gives a tantalising taste of the work, but gives no specifics. Ford et al. (1986) suggested that Cullen’s work found M. brevirostris was more adept at extracting prey from a bark-like substrate, while M. lunatus was better at stretching for prey on leaf tips. If this is the case, it may explain the summary of results given in Figure 1.8.

35 Figure 1.6: Proportion of observations in various feeding classes for M. brevirostris and M. lunatus

From Ford et al. (1986). Data were collected in eucalypt woodland near Armidale, New South Wales. Sample size for M. brevirostris (BHH) was 146 individuals and for M. lunatus (WNH) 340 individuals. The identity of individuals was not known, so actual numbers of individuals may have been overestimated. Classes were different for each attribute, the point here is the similarity in use of classes within each attribute between the two species.

100%

90%

80%

ns 70%

o Class 7 i t a 60% Class 6 erv

s Class 5 b 50% Class 4 O f Class 3 40% Class 2

ercent o Class 1

P 30%

20%

10%

0% BHH WNH BHH WNH BHH WNH BHH WNH foraging method height plant species substrate Species & Foraging Behaviour Attribute

36 Figure 1.7: Dendogram grouping species of honeyeaters and Silvereye based on similarity in feeding behaviour

Figure generated using data provided in Recher et al. (1985). SIL – Silvereye (Zosterops lateralis), BHH – Brown-headed Honeyeater (M. brevirostris), WNH – White-naped Honeyeater (M. lunatus), YFH – Yellow-faced Honeyeater (Lichenostomus chrysops), WEH – White-eared Honeyeater (Lichenostomus leucotis), ESP – Eastern Spinebill (Acanthorhynchus tenuirostris), CRH – Crescent Honeyeater (Phylidonyris pyrrhoptera), RWB – Red Wattlebird (Anthochaera carunculata). Similarity matrix made using untransformed Bray-Curtis similarity and dendogram made using group averaged clustering in PRIMER (Clarke and Warwick 1994).

SIL

BHH

WNH

YFH

WEH

ESP

CRH

RWB

20 40 60 80 100

Similarity

37 Table 1.8: Review of literature relating to proportion of feeding observations on various substrates for M. brevirostris, M. lunatus and M. gularis

Ford and Paton (1977) data were generated from Figure 4 of that paper. All other data are as presented by the cited authors, although some combining has occurred to fit the categories. In some cases the ratios do not sum to one, as in the original papers. The actual sample size in each case may have been overestimated, as individuals were not known. 90% of observations in Ford et al. (1986) were at one site, with the remaining 10% at another (nearby) site. Species n Proportion of observations at: Area n refers to Sites Years Season Source Bark Foliage Flower Other M. brevirostris 481 0.37 0.41 0.22 0.00 M. gularis 94 0.14 0.51 0.23 0.12 Mt Lofty Ranges Feeding Actions 16 1 All Year Ford and Paton (1977) M. lunatus 565 0.23 0.36 0.33 0.08 M. brevirostris 146 0.02 0.90 0.08 0.00 Near Armidale, NSW Feeding Actions 1 3 All Year Ford et al. (1986) M. lunatus 340 0.09 0.79 0.10 0.02 M. brevirostris 61 0.65 0.35 0.00 0.00 Kangaroo Island M. brevirostris 39 0.80 0.20 0.00 0.00 Mt Lofty Ranges M. lunatus 75 0.18 0.76 0.00 0.04 M. brevirostris 54 0.25 0.70 0.00 0.00 NSW mallee M. gularis 47 0.39 0.59 0.00 0.00 Individuals 2 Sep-Jan Keast (1968a) Southern Victoria M. lunatus 67 0.17 0.83 0.00 0.02 M. brevirostris 109 0.82 0.18 0.00 0.00 M. gularis 53 0.38 0.62 0.00 0.00 Sydney M. lunatus 207 0.17 0.83 0.00 0.00 M. brevirostris 0.78 0.22 0.00 0.00 Eastern & southern Not given 2 Keast (1976) M. lunatus 0.17 0.80 0.00 0.03 Australia M. brevirostris 235 0.49 0.48 0.01 0.02 Southern NSW Feeding Actions 3 1 Oct-Jan Recher et al. (1985) M. lunatus 2805 0.19 0.77 0.04 0.00

38 Figure 1.8: Mean proportion (± s.e.) of feeding substrate used by each Melithreptus species over the areas and studies given in Table 1.8

Areas and studies for M. brevirostris = 8, M. gularis = 3, M. lunatus = 6.

1.2

1 tions

c 0.8

M. brevirostris eding A e 0.6 M. lunatus M. gularis tion of F r 0.4 opo r P

0.2

0 Bark Foliage Flower Other Feeding Substrate

1.5.5 Habitat In South Australia, Melithreptus occupy forest, woodland and mallee habitat. Melithreptus brevirostris is a woodland and mallee species, M. lunatus a woodland and forest species, and M. gularis a woodland species (Keast 1968a; Ford and Paton 1977; Chapman 1995; Higgins et al. 2001). Keast (1968a) considered M. brevirostris and M. gularis dry sclerophyll forest and temperate savannah woodland specialists and M. lunatus a wet sclerophyll forest specialist, although in his own words, ‘Differences between wet and dry sclerophyll forest are somewhat arbitrary (pg 769)’. Ford and Paton (1977) suggest that congeneric short-beaked honeyeaters in the Mt Lofty Ranges occupy different habitats. They found that M. brevirostris prefers drier sites while M. lunatus prefers some forest (wetter) sites, although the two species still overlapped by about 31% in distribution across their sixteen sites (data inferred from Figure 3 in Ford and Paton (1977)). It appears likely that M. brevirostris is more of a habitat generalist than the other Melithreptus. Ford and Paton (1977) found M. brevirostris present at fifteen of their sixteen (large; > five hectares) sites, with only P. novaehollandiae being present at more (all) sites. In Victoria, Loyn (1985) found M. brevirostris present in all

39 habitats surveyed and did not find M. brevirostris associated with the bird community of any particular habitat type.

There is evidence that E. viminalis (Ford et al. 1986; Carpenter and Reid 1988; Westphal et al. 2003) and E. leucoxylon (Carpenter and Reid 1988) are important plant species to M. lunatus. No specific links with a plant species have been suggested for either M. brevirostris or M. gularis.

1.5.6 Movement The geographic range of all three Mt Lofty Ranges Melithreptus are reasonably well documented (e.g. Blakers et al. 1984). The best information on the life-time range of individuals comes from banding data, which suggests Melithreptus movements are restricted to a local area (Table 1.9); M. brevirostris and M. gularis are particularly restricted, with M. lunatus a bit less so. As a comparison, the proportion of the M. lunatus population recaptured greater than 10 km from its banding site is over twice as large as M. brevirostris. The larger (>10 km) movements of M. lunatus may be slightly inflated relative to the situation in the Mt Lofty Ranges, as the data include records from the eastern states where a proportion of the population is definitely migratory (Keast 1968b; Higgins et al. 2001). Keast (1968b) stated M. lunatus is a resident or local nomad in the Mt Lofty Ranges, the southern race of M. gularis is a resident or undertakes minor local movements associated with flowering and that M. brevirostris is resident in parts of its range and nomadic elsewhere. The following definitions were used: resident – a species or population that remains throughout the year in its breeding area; local nomad – a population or species where such movements are restricted to an amplitude of a few miles or to the general district or area of breeding; nomad– a non- repetitive, or only partly repetitive, form of seasonal movement. Thus, the banding data appears to support the work of Keast (1968b). However, there is other evidence which suggests both sedentary and migratory elements in the Mt Lofty Ranges M. lunatus population. The phenomenon in which some individuals of a species are migrants while others are residents is known as partial migration. Two summaries of the literature describe M. brevirostris, M. lunatus and M. gularis as resident, migrant and partial migrant (Chan 2001; Higgins et al. 2001). The evidence for partial migration in the Mt Lofty Ranges comes from two sources. Firstly, apparently migratory flocks of M. lunatus have been noted on the southern Fleurieu Peninsula (Paton 1988). Secondly,

40 there are seasonal fluctuations in all Melithreptus numbers, with the reporting rate increasing in April and May and decreasing from September to October (Ford 1977; Ford and Paton 1977). Across sixteen sites, surveyed twice each month for a year, M. brevirostris were recorded about twice as often, M. lunatus about four times as often and M. gularis three times as often in winter as in summer (Table 1.4, Figure 1.2).

The short term movements of individual Melithreptus are not well known. No published studies have looked at the movement of individuals over short time frames, or investigated their home ranges, although McFarland and Ford (1991) state that Melithreptus live in family groups which are not strongly aggressive and which may move locally or migrate.

Table 1.9: Banding recoveries of selected honeyeaters July 1984 to March 1999

Data from Higgins et al. (2001). <10km 10-49km 50-99km Species Recoveries 1984-1999 % M. brevirostris 1,295 99.8 0.2 0.00 M. lunatus 1,095 99.4 0.5 1 record – 0.10 M. gularis 35 100.0 0.0 0.00 P. novaehollandiae 11,260 99.1 0.9 4 records – 0.04

1.5.7 Ectethmoid-mandibular articulation Melithreptus are the only genus of honeyeater to have evolved a well developed ectethmoid-mandibular articulation, in addition to the normal quadrate articulation of the jaw (Bock and Morioka 1971). The ectethmoid-mandibular articulation may be considered a brace, against posterior forces, of the mandible (lower jaw) against the ectethmoid (a bone in the skull). In Melithreptus the mandible has a dorsal process (bit that sticks up) which fits into a ventral fossa (hole) in the ectethmoid plate. Combined with specialised musculature, this allows Melithreptus to hold the mandible closed against the skull but still raise the maxilla (upper jaw) independently with minimal musculature expenditure (Bock and Morioka 1971). Some species of Manorina have a similar, but less well developed ectethmoid-mandibular articulation, as does a species of Ptiloprora. A number of other diverse bird groups also possess an additional articulation, the basitemporal articulation, which serves a similar function (Bock 1960). While the basitemporal articulation has evolved a number of times in diverse bird groups, an ectethmoid-mandibular articulation has evolved only in the three genera of honeyeaters, Melithreptus, Manorina and Ptiloprora (Bock and Morioka 1971).

41

The ectethmoid-mandibular articulation was detailed by Bock and Morioka (1971), including the skull, musculature of the tongue and jaw and position of salivary glands and ducts. Following this account of the feeding apparatus, they suggested two possible situations in which the ectethmoid-mandibular articulation may be useful: under an external posterior force acting on the mandible; and under contraction of the mandibular adductor muscles. Firstly, normally under an external posterior force acting on the mandible, a muscle group prevents disarticulation of the jaw if the force is excessive, but the muscle group also tightly closes the jaw under these circumstances. The ectethmoid-mandibular articulation would both prevent disarticulation and allow the upper jaw to be opened under such circumstances. Secondly, normally under contraction of the mandibular adductor muscles, the mandible exerts a compressive force on the quadrate, rotating the quadrate back and opposing the force of the protractor muscle which is used to raise the maxilla. With the ectethmoid-mandibular articulation, the mandible may be held fully adducted and fixed in position without compressing the quadrate, which in turn allows the maxilla to be raised without any loss of force.

Bock and Morioka (1971) felt that the second of their two suggestions was the most likely explanation for the evolution of an ectethmoid-mandibular articulation in Melithreptus. They were unaware of any normal feeding circumstances in which an external posterior force acted on the mandible, and the structure of the normal quadrate articulation, which did not appear to differ from other Meliphagidae, was capable of preventing disarticulation of the jaw under normal circumstances. Thus, to them, the first suggestion lacked any real biological function. However, they were able to provide a biological explanation for their second suggestion. This explanation postulated that Melithreptus captured insects with a mucous coated tongue and required that during feeding the tongue must contact the duct opening of the maxillary gland on the tip of the upper jaw in order to be coated with mucous. If the jaw was opened using the normal quadrate articulation, the lower jaw dropped and the tongue does not pass over the ducts. However, this would not be an adaptation of Melithreptus that would enable them to access resources not available to other honeyeaters. Other honeyeaters, such as Phylidonyris spp., with decurved beaks would also have the tongue pass over the maxillary gland on the tip of the upper jaw (Figure 1.9).

42

The work of Paton (1980) suggests another possibility for the functional significance of the ectethmoid-mandibular articulation and associated modifications of the musculature and salivary glands. Perhaps Melithreptus are using their modified skull to access honeydew under decorticating bark. Both the possibilities for the functional significance of the modified skull suggested by Bock and Morioka (1971) would be useful in such a situation. The ability to withstand large posterior forces would allow Melithreptus to insert the beak under bark with greater force than other honeyeaters and they would then also be able to raise their maxilla slightly, allowing them access to honeydew that other honeyeaters may not be able to reach. The maxillary gland may assist in dissolving highly concentrated or crystallised honeydew, further enhancing their ability to access a resource inaccessible to other honeyeaters. It is hard to know if being short-beaked would be a disadvantage in this case. Long-beaked species may be able to get their beaks further in under bark, but in cases where it is necessary to force the beak in and then open it Melithreptus would have an advantage.

Figure 1.9: Comparison of the beaks of a) Melithreptus (M. gularis) and b) Phylidonyris (P. novaehollandiae)

The beak of Phylidonyris is decurved, forcing the tongue to contact the duct opening of the maxillary gland on the tip of the upper jaw when extended. An adaptation specifically to do this in Melithreptus would therefore not confer on them any advantage not available to some other honeyeaters. Photos: David Paton.

a) b)

1.6 Summary

Despite considerable debate in the literature regarding the role of competition in producing patterns of niche partitioning, its status as an important factor in structuring nectarivore communities appears justified. In Australia, where numerous nectarivore species are sympatric and where variability is the norm, at any one time, there are likely to be a number of nectarivores accessing the same resources.

43

Since European settlement of Australia, there have been large scale changes to habitats and vegetation, particularly in south-eastern Australia. This has resulted in the decline of numerous species of woodland birds, including honeyeaters. At the same time, a number of species are increasing in abundance, with some honeyeaters remaining widespread and abundant if not actually increasing.

The Mt Lofty Ranges woodland bird community has a particularly large number of declining species, possibly as a result of its isolation but particularly as a result of its large proportion of vegetation clearance. Included in the declining woodland birds of the Mt Lofty Ranges are three Melithreptus species, M. brevirostris, M. lunatus and M. gularis. Due to the importance of competition in the structuring of nectarivore communities, competition has been highlighted as a potential and likely cause of decline for those declining woodland honeyeater species. Competition within the genus Melithreptus is already an unresolved issue. Perhaps the most important question in Melithreptus ecology is the sympatry of M. lunatus and M. brevirostris over a large area of south-eastern Australia. A number of authors have drawn attention to the similarities between M. brevirostris and M. lunatus, particularly compared to the rest of the bird community they are examining, but given their wider focus have not been able to tease apart any differences. Keast (1968a) attempts to answer this question, but his ‘answer’ is phenomenological and he suggests no biological implications.

In most studies of coexistence, both within Melithreptus and within nectarivore communities in Australia, feeding ecology is given prominence. One area of agreement is that Melithreptus get their food from the canopy of eucalypts, generally gleaning either bark or foliage and taking nectar from flowers. In studies contrasting M. brevirostris and M. lunatus the proportion of bark and foliage feeding by the two species is often commented on, giving the indication in a review of the literature that this is important in accounting for the large area of sympatry between M. brevirostris and M. lunatus. However, due to the wider focus of many of these studies, a closer look at their results suggests that this question is not yet satisfactorily resolved.

The ectethmoid-mandibular articulation is an interesting aspect of Melithreptus biology that has received no attention since the original paper (Bock and Morioka 1971). Such a

44 unique adaptation is likely to be of importance in the ecology of Melithreptus but has not been studied in practical situations by subsequent studies. One study that may have been able to shed some light on the topic was the work of Cullen (1983), although the results of the study were not published.

Other aspects of the Melithreptus niche have received varying levels of investigation. At a broad scale there is agreement that M. brevirostris is found in drier habitats (mallee), M. lunatus in wetter habitats (forest), and M. gularis in intermediate habitats. However, at smaller scales all three species are (were) likely to occur in sympatry. Behavioural aspects of the Melithreptus niche, other than foraging behaviour, have been recorded through anecdote. It appears from the literature that there are large scale movements of Melithreptus within (and possibly into and out of) the Mt Lofty Ranges. Seasonal fluctuations suggested by Ford and Paton (1977) are generally supported by the longer term reports on the birds of various conservation parks.

45 Chapter 2 A SURVEY TO COMPARE FACTORS POTENTIALLY

INFLUENCING MELITHREPTUS DECLINE

2.1 Introduction

In order to manage a declining species a good knowledge of its requirements and the causes of its decline are necessary. However, determining the appropriate causes to address for any individual species or group of species can be difficult, given the range of possible causes for the decline of woodland bird species (e.g. Robinson 1994; Robinson and Traill 1996; Ford et al. 2001). This chapter presents work that aimed to rank the most likely causes of decline for each of the Mt Lofty Ranges Melithreptus. Information-theoretic methods were used in order to test between a range of possible hypotheses for decline in Melithreptus. This paradigm of data analysis differs from ‘traditional’ statistical techniques in allowing for the direct ranking of multiple working hypotheses.

The set of hypotheses for decline of Melithreptus outlined below were determined from the intersection of hypotheses for the decline of woodland birds in general and what is known of the biology and ecology of Melithreptus. In keeping with the information theoretic paradigm, the candidate set was kept to an adequate minimum of hypotheses that could be supported from the known biology of the situation. The hypotheses for decline tested were: • Loss of specific structural habitat; • Loss of specific floristic habitat; • Reduced food abundance; • Increased competition with aggressive species; and • Fragmentation.

No studies have attempted to determine the primary factors influencing the abundance and possible causes of decline in Melithreptus. Most studies including Melithreptus in their results are from either much wider studies which are examining the entire bird community or smaller scale studies with a focus on a particular area (usually a conservation park). From these studies, the most consistent factor correlated with Melithreptus is the presence of a eucalypt overstorey and the broad preference of

46 individual species for a particular structural formation. For example, structure was seen as important by Ford and Paton (1977) when investigating the comparative ecology of ten honeyeater species in South Australia; M. brevirostris preferred drier, open areas, while M. lunatus preferred wetter, denser areas. Floristics have also been reported as important to M. lunatus, with an association between M. lunatus and Eucalyptus viminalis (Ford et al. 1986; Carpenter and Reid 1988; Westphal et al. 2003) and M. lunatus and E. leucoxylon (Carpenter and Reid 1988) reported. However, no quantitative data have been presented to support any associations. Loss of areas with appropriate structure and/or floristics (eucalypt overstorey) fall under the hypothesis of disproportionate loss of specific habitats put forward by Ford et al. (2001) and were two of the hypotheses tested in the work presented here. If Melithreptus were sufficiently specialised in any structural or floristic habitat, and that habitat was now limiting their abundance, the data collected should support most strongly the habitat model.

Ford et al. (2001) suggested reduced food abundance as a possible cause of decline of woodland birds. Adequately quantifying all food resources of honeyeaters is remarkably difficult (Section 1.5.3) but by quantifying the available feeding substrate (foliage and peeling bark) for Melithreptus, a substitute for food abundance was established and included in the set of hypotheses tested. Numerous studies have established gleaning on and under bark and gleaning foliage as the main mode of foraging in Melithreptus. If food resources were limiting Melithreptus, the data collected should support this model above any of the others.

The final hypothesis suggested by Ford et al. (2001), thought a potential cause of decline in Melithreptus, was that of increased competition with aggressive species. They suggested examining bird communities for species that rarely occur together or that tend to be negatively correlated (forbidden combinations) as a test of this hypothesis. In the Mt Lofty Ranges, two species of honeyeater remain abundant and widespread, Phylidonyris novaehollandiae and Anthochaera carunculata. If competition with either or both of these species is contributing to the decline of Melithreptus, the data collected should support a negative relationship between Melithreptus and the widespread, abundant and aggressive honeyeaters.

47 Finally, the remaining pattern of vegetation in the landscape may limit the ability of Melithreptus to meet its requirements effectively. While not explicitly suggested as a hypothesis by Ford et al. (2001), it is closely related to habitat fragmentation reducing food availability. In a homogenous environment a species may decline through the fragmentation of its habitat if habitat loss is of such an extent that individuals must move excessively to meet their requirements. A number of landscape metrics were used to characterise each site in order to test the effect of fragmentation on Melithreptus abundance. If fragmentation was contributing to the continued decline of Melithreptus, the data collected should indicate a preference of Melithreptus for areas of native vegetation with the lowest degree of fragmentation.

2.2 Methods

Ninety sites were established in the Mt Lofty Ranges confined to three main areas; a northern area encompassing Para Wirra Recreation Park and Hale and Warren Conservation Parks; a central area encompassing Belair Recreation Park and Mark Oliphant and Scott Creek Conservation Parks; and a southern area encompassing Scott, Cox Scrub and Bullock Hill Conservation Parks. All areas also included some sites in vegetation outside these parks, mainly road reserves. There were 30 sites in each area, each site being a circle of radius 56 metres to give an area of 1 hectare. Logistical constraints saw sites located along paths, tracks, roads and areas between 200 m and 500 m apart, allowing efficient use of time during the survey periods. The random number generator in Microsoft Excel was used to determine distances between sites. After an initial reconnaissance, areas, paths, tracks or roads were chosen that crossed a number of habitat types. Dominant woodland habitats present within the study area were stringybark (Eucalyptus baxteri and/or E. obliqua), box (E. microcarpa, E. odorata or E. goniocalyx), blue gum (Eucalyptus leucoxylon), manna gum (E. viminalis or Eucalyptus rubida), pink gum (E. fasciculosa) and red gum (E. camaldulensis).

Part of the southern area can be seen in Figure 2.1 including fifteen survey sites. For the landscape fragmentation analysis, each site was buffered by a 10 hectare area which has also been plotted in Figure 2.1. Three of the sites shown are located on road reserves in vegetation that was not mapped on the base layer used for vegetation cover, two of which are visible towards the southwest of the Figure, and one just to the east of Cox

48 Scrub Conservation Park (Section 2.2.3 describes the buffering process and how sites that did not occur within mapped vegetation were treated).

Figure 2.1: Example of fifteen survey sites from the southern area

The numbered grid is AMG coordinates (GDA94), with each square on the grid representing 1 km2. The six sites enclosed within one polygon are within Cox Scrub Conservation Park. Some sites do not fall within mapped vegetation. The method for dealing with this situation and the buffering process is outlined in Section 2.2.3.

291000 292000 293000 294000 295000 296000 297000 298000 299000 300000 301000 302000

# 6090000 6090000 #

6089000 6089000

6088000 6088000

# # # # # 6087000 6087000

# #

#

6086000 6086000

#

6085000 6085000

#

6084000 6084000

6083000 6083000

#

6082000 6082000 #

# 6081000 6081000

291000 292000 293000 294000 295000 296000 297000 298000 299000 300000 301000 302000

# Survey Sites 10 Ha Buffer clipped with Vegetation Cover Vegetation Cover N

2.2.1 Honeyeater abundance Each of the ninety sites were visited on three separate days in February, May, August and November 2001. During each visit, a ten minute area search was carried out and all birds seen or heard were recorded. Caution was taken not to count the same bird twice within a single count. Birds heard or seen outside the site were not included in any analyses (except M. gularis, see results). Sites were surveyed throughout the day from about half an hour after sunrise until half an hour before sunset. The order of visits to areas and parks was changed between surveys. The sum of honeyeaters seen over all visits to a site was used as an estimate of the abundance of that species at that site. The

49 total number of Melithreptus of each species counted at a site was the dependent variable modelled by the various explanatory variables.

Three models of aggressive honeyeater abundance were used (where Phylidonyris novaehollandiae is NHH and Anthochaera carunculata is RWB): • RWB; • NHH; and • RWB + NHH. Manorina melanocephala was originally planned as an explanatory variable, but was left out of the analysis due to low numbers of sites with miners.

2.2.2 Structural, floristic and feeding substrate attributes A 50 metre tape measure was run from the centre of the site in a random direction between 0 and 179° and another from 180 and 359° (i.e. two transects at each site). Percent eucalypt canopy cover was taken along these lines by recording distances at which an imaginary vertical line intersected the outer edge of the eucalypt canopy. The (up to) ten eucalypts closest to the centre of the site whose canopy overlapped this transect were selected for the following measurements: species, tree height, height of the top and the bottom of the canopy, maximum and minimum diameter of the canopy projected onto the ground, projected foliage cover and amount of peeling bark. All height and width measurements were taken with a tape measure and/or clinometer as appropriate. A transparency of examples of projected foliage cover (Figure 2.2) was used when recording projected foliage cover (Heard and Channon 1997). Peeling bark was recorded using a scale of 0 (none) to 5 (super-abundant). A list of eucalypt species was taken for each site, and an estimate was made of their relative abundance.

The following variables were determined from the above measurements: • projected foliage cover of eucalypt overstorey for the site (PFC) from the percent eucalypt canopy cover for the site multiplied by the projected foliage cover averaged over the 10 individual trees; • overstorey height from average canopy top measurement (HT); • percent cover of each eucalypt species (E. leucoxylon BLU, E. microcarpa &/or E. odorata &/or E. goniocalyx BOX, E. cosmophylla CUP, E. viminalis MAN, E. fasciculosa PIN, E. camaldulensis RED, E. obliqua &/or E. baxteri STR)

50 from the relative abundance of each eucalypt species multiplied by the overstorey percent cover recorded for the site; • amount of foliage (AFO) from the average canopy volume (4/3×π×(canopy top – canopy bottom/2) ×(maximum canopy width/2)×(minimum canopy width/2)) of the ten trees multiplied by projected foliage cover; and • amount of peeling bark (ABA) from the average canopy volume (as measured above) multiplied by the average amount of peeling bark of individual trees and the percent cover of eucalypt overstorey.

One model for structure was tested, in which projected foliage cover and height were used. Five models for floristics were tested, one containing all eucalypt species, one containing the ‘gum’ eucalypts common on the more productive areas in the Mt Lofty Ranges (Eucalyptus leucoxylon blue gum, E. fasciculosa pink gum and E. viminalis manna gum) and three models each containing the common gum species separately. Originally, E. camaldulensis was to be included in these models as well, but it became clear during survey work that E. camaldulensis was not an important species to Melithreptus. One model for feeding substrate was tested, in which the amount of foliage and the amount of peeling bark were used. Thus, the following seven models were established from the explanatory variables: • PFC + HT; • BLU + BOX + CUP + MAN + PIN + RED + STR; • BLU + MAN + PIN; • BLU; • MAN; • PIN; and • AFO + ABA.

51 Figure 2.2: Examples of projected foliage cover

These examples were photocopied to a transparency sheet for use in the field.

2.2.3 Landscape attributes Using the buffering feature of the Animal Movement SA v2.04beta extension for ArcView 3.x (Hooge et al. 1999) a radius of 178.41 metres (to give an area of 10 hectares) was created around each survey point. The layer of buffers was then intersected with a layer of native vegetation (provided by Planning South Australia, only vegetation with an area greater than 1 hectare was mapped) using the Patch Analyst extension for ArcView 3.x. (see Westphal et al. 2003). Patch Analyst was then used to analyse the buffered landscape around each survey point for the following statistics:

52 • the distance from the survey point to the nearest vegetation edge with a monoculture (DTE); • proportion of the 10 ha which was vegetation (PV); and • the edge density (ED), calculated from the total distance of edge / the total area of native vegetation within the 10 ha buffer. If a survey point occurred in an area not mapped as native vegetation DTE was set at 10m and the area of native vegetation within the corridor was set at (2×10m×178.41m) 3568.2m2. This area of native vegetation was then added to any other native vegetation mapped within the 10ha buffer. One model incorporating landscape variables was tested, in which distance to edge, proportion of vegetation in 10ha and edge density was used (DTE + PV + ED).

2.2.4 Analysis A total of 13 models were tested, including a null model in which Melithreptus was hypothesised to be distributed evenly across the landscape regardless of other variables and a global model which included all variables.

An information-theoretic approach was used to determine the best model from the set, given the data collected (Anderson et al. 2000; Anderson and Burnham 2001; Burnham and Anderson 2001; 2002). Each a priori model was ranked using Akaike’s Information

Criteria (AIC) corrected for small sample size (AICc). Anderson et al. (2000) suggest using AICc when n/K is less than about forty, where n is sample size and K is the number of parameters in the model. In this case, n was 90 and the global model contained 17 parameters, giving n/K=5.3. AICc was modified by including a variance inflation factor (ĉ) to allow for overdispersion, common in modelling count data due to the large number of zero values (Burnham and Anderson 2001; 2002). The variance inflation factor was estimated from the ratio of residual model deviance to the residual degrees of freedom of the global model and the number of parameters was increased by 1 to include the estimation of the variance inflation factor (Burnham and Anderson 2001; 2002). The differences in AIC score between each model and the model with the lowest AIC score were determined (∆i, i.e. the model with the lowest AIC score has

∆i=0 and is the model best supported by the data). Burnham and Anderson (2001) suggest that models with ∆i values of 0-2 have a substantial level of empirical support, 4-7 considerably less support and >10 essentially no support. In this way, it is possible

53 to have more than one model with a substantial amount of support from the data. As a further aid in model selection, Akaike weights (ωi) are able to give the relative probability of each model being the best supported by the data (Burnham and Anderson 2001; 2002). These were calculated for each model and in order to determine the relative importance of predictor variables Akaike weights for all models containing a predictor variable were summed. This allows a more thorough investigation of the importance of each explanatory variable than simply taking each variable from the best model(s) (Burnham and Anderson 2001).

Models were fitted to the data by generalised linear modelling (GLM) in R (Ihaka and Gentleman 1996). The original variance inflation factor obtained from the global model fitted assuming a Poisson error structure was greater than four, suggesting the model did not adequately fit the data (Burnham and Anderson 2002). Assuming a negative binomial error structure (variance > mean) rather than Poisson error structure (variance = mean) (Ludwig and Reynolds 1988; Venables and Ripley 1999; Crawley 2002) gave a variance inflation factor closer to 1, suggesting adequate model fit (Burnham and Anderson 2002). Models were fitted using the glm.nb function in the MASS library (Venables and Ripley 1999) of R.

2.3 Results

Melithreptus brevirostris were recorded at 56% of the sites, with a mean of 3.3 birds per site over all counts (total M. brevirostris recorded divided by 90 sites) and 0.5 birds per hectare where they were present (total M. brevirostris recorded divided by 50 sites at which they were present divided by 12 counts per site) (Table 2.1). Melithreptus lunatus were recorded at 41% of the sites with a mean of 5.9 birds per site and 1.2 birds per hectare. Melithreptus gularis were recorded at only four sites. Phylidonyris novaehollandiae was by far the most widespread and abundant honeyeater. It was present at 84% of sites and the mean total per site was 33 birds. This is one and a half times as widespread and over ten times as abundant as M. brevirostris and over two times as widespread and over five times as abundant as M. lunatus. Further examination of the data revealed that only 1.67% of the M. brevirostris recorded and 0.75% of the M. lunatus recorded occurred at sites without any P. novaehollandiae. The incidence, and mean, maximum and minimum values, of each of the explanatory variables is given in Table 2.2.

54 Table 2.1: Honeyeater incidence, mean count, maximum count and birds per hectare at 90 sites counted 12 times in the Mt Lofty Ranges during 2001

Sites is the number of sites at which the species was present. Mean is the total number of birds counted divided by 90. Maximum is the maximum number of birds recorded at one site over all counts. Birds per hectare is calculated only using sites at which birds were present, and taking the number of counts per site into account. Species Sites Mean Maximum Birds per hectare Acanthorhynchus tenuirostris 67 3.4 20 0.4 Anthochaera carunculata 80 19.2 104 1.8 Lichenostomus chrysops 71 6.9 31 0.7 Lichenostomus penicillatus 16 1.6 24 0.7 Manorina melanocephala 7 2.3 85 2.2 Melithreptus brevirostris 50 3.3 21 0.5 Melithreptus gularis 4 0.1 2 0.2 Melithreptus lunatus 37 5.9 61 1.2 Phylidonyris novaehollandiae 76 33.4 189 3.3 Phylidonyris pyrrhoptera 59 6.0 51 0.8

55 Table 2.2: Explanatory variable incidence and mean, maximum and minimum values at 90 sites in the Mt Lofty Ranges

Sites is the number of sites at which the variable was present. The mean is taken across all sites (including absences). Explanatory variables – PFC: projected foliage cover; HT: overstorey height; BLU: percent cover of E. leucoxylon; BOX: percent cover of E. microcarpa/E. odorata/E. goniocalyx; CUP: percent cover of E. cosmophylla; MAN: percent cover of E. viminalis; PIN: percent cover of E. fasciculosa; RED: percent cover of E. camaldulensis; STR: percent cover of E. obliqua/E. baxteri; AFO: amount of foliage; ABA: amount of peeling bark; NHH: total records of P. novaehollandiae; RWB: total records of A. carunculata; DTE: distance to monoculture; PVA: proportion of surrounding 10ha that were vegetated; and ED: edge density Variable Sites Mean Maximum Minimum units % eucalypt cover× PFC 90 21.6 50.3 1.8 average foliage density HT 90 11.6 27.0 3.1 metres BLU 41 13.4 80.0 0 % BOX 26 5.2 53.6 0 % CUP 12 1.9 71.1 0 % MAN 20 5.3 68.0 0 % PIN 49 8.2 69.3 0 % RED 10 3.2 63.3 0 % STR 43 16.3 81.3 0 % NHH 76 33.4 194 0 total birds RWB 80 19.2 104 0 total birds pfc× average canopy volume (m3)× AFO 90 13,025,767 112,806,060 38,365 average foliage density pfc× average canopy volume (m3)× ABA 90 601,666 7,136,710 1,839 average peeling bark DTE 90 118.7 178.4 10.0 metres PVA 90 0.8 1.0 0.03 ratio ED 90 0.02 0.11 0.01 (veg edge m) /(veg m2)

2.3.1 Melithreptus brevirostris

QAICc model selection criteria indicated that NHH (abundance of P. novaehollandiae) was the model with the best fit, having a 65% probability of being the best model of M. brevirostris abundance in the candidate set (as indicated by Akaike weights Table 2.3). There was also substantial support for the model containing NHH + RWB (P. novaehollandiae and A. carunculata). The other models had essentially no support. Adding together the Akaike weights for models including NHH gives 0.9877 (0.3349 +0.6512 +0.0016) making it by far the most important predictor variable. The next closest is RWB with predictor weight of 0.3368.

56 The coefficient of NHH in both models was positive, while the coefficient of RWB was negative. Melithreptus brevirostris responded positively to P. novaehollandiae abundance and, to a lesser degree, negatively to Anthochaera carunculata abundance.

Table 2.3: Model selection statistics for M. brevirostris data

See Table 2.2 for explanatory variable codes. Presented in decreasing order of probability of being the best model of M. brevirostris abundance, as indicated by Akaike weights.

Model -log(L)/ĉ K QAICc ∆i ωi NHH 157.45 2 319.05 0 0.6512 NHH + RWB 157.05 3 320.38 1.33 0.3349 AFO + ABA 160.63 3 327.53 8.49 0.0093 global 145.83 17 331.12 12.07 0.0016 MAN 164.25 2 332.65 13.60 0.0007 PIN 164.29 2 332.72 13.67 0.0007 null 165.80 1 333.64 14.60 0.0004 RWB 165.02 2 334.17 15.13 0.0003 BLU + MAN + PIN 162.97 4 334.42 15.37 0.0003 PFC + HT 164.32 3 334.91 15.86 0.0002 BLU + BOX + CUP + MAN + PIN + RED + STR 158.93 8 335.63 16.58 0.0002 BLU 165.76 2 335.65 16.60 0.0002 DTE + PVA + ED 165.78 4 340.03 20.98 0.0000

2.3.2 Melithreptus lunatus

QAICc model selection criteria indicated that BLU + MAN + PIN (percent cover of E. leucoxylon, E. viminalis and E. fasciculosa) was the model with the best fit, having a 52% probability of being the best model of M. lunatus abundance in the candidate set (as indicated by Akaike weights Table 2.4). There was also substantial support for the model containing only E. viminalis and one other model had some support, (BLU + BOX + CUP + MAN + PIN + RED + STR). Adding together the Akaike weights for models including MAN gives 0.9547 (0.1505 +0.5176 +0.2445 +0.0421) making it by far the most important predictor variable. The next closest are BLU and PIN, both with predictor weights of 0.7102 (0.1505 +0.5176 +0.0421).

The coefficients for MAN, BLU and PIN in each of the models were positive. Melithreptus lunatus responded positively to E. viminalis cover but E. leucoxylon and E. fasciculosa cover were also important.

57 Table 2.4: Model selection statistics for M. lunatus data

See Table 2.2 for explanatory variable codes. Presented in decreasing order of probability of being the best model of M. lunatus abundance, as indicated by Akaike weights.

Model -log(L)/ĉ K QAICc ∆i ωi BLU + MAN + PIN 173.55 4 355.57 0 0.5176 MAN 176.47 2 357.07 1.50 0.2445 BLU + BOX + CUP + MAN + PIN + RED + STR 170.13 8 358.04 2.47 0.1505 PFC + HT 177.08 3 360.44 4.87 0.0453 global 159.04 17 360.59 5.02 0.0421 PIN 185.78 2 375.70 20.12 0.0000 null 188.07 1 378.18 22.61 0.0000 AFO + ABA 186.07 3 378.43 22.85 0.0000 NHH 187.28 2 378.69 23.12 0.0000 BLU 187.91 2 379.97 24.39 0.0000 RWB 188.03 2 380.20 24.62 0.0000 NHH + RWB 187.26 3 380.81 25.23 0.0000 DTE + PVA + ED 187.43 4 383.33 27.76 0.0000

2.3.3 Melithreptus gularis While M. gularis was only recorded at four sites during survey work, records of M. gularis from seven sites were made (including records outside survey time and records of M. gularis made from a site but outside the 1 ha area). These data were included in the analysis for M. gularis. Due to the small number of sites at which M. gularis was recorded the analysis for M. gularis was carried out using uni-variate non-parametric correlation of M. gularis against the variables included in the information theoretic approach (Table 2.5). The closest non-parametric correlation with M. gularis was P. novaehollandiae, and as with M. brevirostris the correlation was positive. The next closest correlation was a negative correlation with stringybark (E. baxteri and E. obliqua).

58 Table 2.5: Results of non-parametric correlation of M. gularis numbers against each explanatory variable

See Table 2.2 for variable codes. Sorted on order of P value. P is the probability of that non-parametric association occurring by chance. Variable Spearman Rho P NHH 0.301 0.004 STR -0.259 0.014 PIN 0.198 0.062 BOX -0.181 0.087 ABA 0.178 0.093 DTE -0.173 0.103 AFO 0.160 0.133 RWB 0.151 0.155 BLU 0.151 0.156 ED 0.128 0.231 HT 0.123 0.247 CUP -0.113 0.287 RED -0.102 0.337 PVA -0.102 0.339 MAN -0.060 0.576 PFC 0.024 0.822

2.4 Discussion

2.4.1 Summary Two important themes came through in the results for Melithreptus. Firstly, at least two of the species (M. brevirostris and M. gularis) appear to be responding positively to the most widespread and abundant honeyeater within the ranges, P. novaehollandiae. Secondly, the type of woodlands that occur on the more productive soils, E. viminalis, E. leucoxylon and E. fasciculosa appear important to at least two of the species (M. lunatus and M. gularis). There was essentially no support for the models which included amount of foraging substrate (bark and foliage) or landscape metrics.

2.4.2 Melithreptus brevirostris While the result for a negative response of M. brevirostris to A. carunculata was as expected, the positive correlation between M. brevirostris and P. novaehollandiae abundance was a surprise, given the initial prediction for a negative response, if any. At face value then, the work presented here failed to find support for the forbidden combinations suggested by Ford et al. (2001). Unlike the postulated negative response of Melithreptus to P. novaehollandiae, there are no realistic biological explanations for

59 a positive response to P. novaehollandiae itself. It does not seem credible to conclude that M. brevirostris may be limited in the Mt Lofty Ranges by a lack of P. novaehollandiae. Melithreptus brevirostris also occurs in parts of Australia in which P. novaehollandiae does not occur, further suggesting that a direct, positive response of M. brevirostris to P. novaehollandiae is unlikely.

A more likely explanation of the result is a shared preference for resources, with both M. brevirostris and P. novaehollandiae found in higher numbers at a one hectare site containing those resources, most likely food resources. If the postulated shared preference for resources is correct, M. brevirostris responded most strongly to resources (with P. novaehollandiae being a surrogate for resources), and was possibly limited by the availability of, or access to, those resources. If correct, this is an interesting result, given P. novaehollandiae generally forages from floral resources (e.g. Recher 1977; Paton 1979; 1982b; 1985), while M. brevirostris generally forages from bark and foliage (Table 1.8, pg 38). However, Paton (1980) showed that at times P. novaehollandiae relied on non-floral sources of carbohydrate, which were generally available from bark and foliage. Therefore, a shared preference for carbohydrate resources available from bark and foliage is possible.

If there is a shared preference for resources, then competition between M. brevirostris and P. novaehollandiae should not be discounted, despite the result obtained here. The scale at which observations occurred (1 ha) may have masked any impact of competition between M. brevirostris and P. novaehollandiae by including enough diversity in resource levels to provide areas in which the less dominant honeyeaters can avoid harassment and/or feed on resources not protected/depleted by the more dominant honeyeaters (both numerically and aggressively). One interesting prediction can then be made from this suggestion; in areas outside the Mt Lofty Ranges a positive relationship between M. brevirostris and locally widespread and abundant honeyeaters should occur, at a scale of one hectare, due to a shared preference for resources. One probable exception to this prediction would be areas dominated by species of the genus Manorina. M. melanocephala was originally a species to be included in the analysis, but it was found to be present at too few sites (eight in total). Melithreptus brevirostris was only recorded at one of the eight sites at which M. melanocephala was recorded.

60 2.4.3 Melithreptus lunatus While M. brevirostris apparently responded to resources, M. lunatus responded to floristics, or more specifically, the eucalypt overstorey. Melithreptus lunatus was most abundant at sites with, most importantly E. viminalis, but also E. leucoxylon and E. fasciculosa. Given that the floristics model was the best supported by the data, it suggests that M. lunatus is limited by the availability of this habitat. Recher et al. (1996) found that bird species selected between species of plant as foraging substrates based on the kinds of arthropods available and their abundance. The richest and most abundant arthropod faunas occurred on tree species with the highest levels of foliage nutrients, corresponding to where soil fertility was highest. Psyllids were well represented on each of the four species of eucalypt sampled. Recher et al. (1996) also suggested that bark arthropod communities are likely to follow a similar pattern. It is likely that M. lunatus was most abundant in areas with gum eucalypts, and specifically E. viminalis, as these eucalypts grow on the most productive soils within the ranges (Armstrong et al. 2003) and therefore probably have the most abundant and diverse arthropod communities, but more importantly, probably have abundant and diverse psyllid communities, providing honeydew and potentially manna (hence the common name of E. viminalis, manna gum) on which M. lunatus forage.

In their comparative study of honeyeaters in the Mt Lofty Ranges, Ford and Paton (1977) found that the niches of M. brevirostris and M. lunatus could be separated most obviously by vegetation structure, or habitat. Melithreptus lunatus was most abundant at forest sites while M. brevirostris was most abundant at drier sites. The results presented here suggest that vegetation structure per se is not the driving force behind this apparent separation. In the Mt Lofty Ranges, E. viminalis often occurs as part of the structural formation forest, suggesting that the results of Ford and Paton (1977) may in fact reflect a response of M. lunatus to floristics rather than forest. An E. viminalis association is not mentioned in Ford and Paton (1977) but Ford et al. (1986) did find that M. lunatus fed in particular from the loose bark of E. viminalis. In mentioning this association, they state, ‘… Interestingly the site in South Australia where White-naped Honeyeaters were most abundant also had many manna gums...’, presumably referring to unpublished results from the study of Ford and Paton (1977). A response to floristics rather than structure also explains why M. lunatus did not respond to the large areas of E. obliqua forest remaining in the Mt Lofty Ranges.

61 2.4.4 Melithreptus gularis If the results for M. gularis are accepted despite small sample size, M. gularis showed some similarity to both M. brevirostris (responded to resources) and M. lunatus (responded to floristics). The close correlation between M. gularis and P. novaehollandiae was almost certainly due to a shared preference for resources, as discussed previously for M. brevirostris. The next closest correlation was negative and with stringybark (E. baxteri and E. obliqua). If M. gularis preferred the box and gum woodlands that tend to occur at lower altitudes within the Mt Lofty Ranges this may be reflected in a negative relationship with stringybark which occurs on the higher elevations within the Mt Lofty Ranges, generally on poorer soils. There is some weak support for this argument, with the next closest correlations being E. fasciculosa and the box eucalypts. If there is a preference of M. gularis for the box and gum woodlands of the lower altitudes of the Mt Lofty Ranges, the decline of this species is undoubtedly in large part a response to the clearance of its preferred habitat. It has been well documented that the more productive land in the Mt Lofty Ranges was preferentially cleared for agriculture, leaving the less productive, rocky and steep terrain with the most remaining vegetation (Long 1998; Paton 1999; Paton et al. 1999; Paton 2000; Paton et al. 2004b). However, clearance in the Mt Lofty Ranges theoretically stopped in 1983 with the first regulation of native vegetation clearance being implemented at that time under the Planning Act 1982, whereas the decline of M. gularis really only started to become precipitous around that time. Therefore, it appears that M. gularis is one species currently in the ‘lag’ period between habitat clearance and loss of the component species, and it is likely to become regionally extinct if any process(es) acting to cause the continued decline of M. gularis within the region are not reversed. Besides the loss of habitat, a positive correlation of M. gularis with P. novaehollandiae at a 1 ha scale does not discount competition between the two species.

2.4.5 Criticism One aspect of this study which could be improved was site selection. Random distances between 200 and 500 metres were used after roads, tracks, paths and areas were chosen that crossed a number of habitats. In order to improve this site selection method, each road, track, path and area could be loaded into a GIS program and buffered to, say, 100m. An extension which randomly assigns points within a layer could be then be used to place the required number of survey sites randomly. Stratifying the sites within

62 mapped vegetation types is also a possibility, although at the present time, the vegetation mapping of the Mt Lofty Ranges is not of sufficient accuracy to warrant this.

2.4.6 Conclusion The data collected provided essentially no support for models of Melithreptus abundance based on habitat fragmentation or food substrates. Melithreptus abundances were best modelled by certain floristics and abundance of P. novaehollandiae.

The positive correlation between Melithreptus abundance and P. novaehollandiae abundance suggests a shared preference for resources. However, the work leaves unresolved the issue of competition between Melithreptus and widespread and abundant honeyeaters, particularly P. novaehollandiae. Phylidonyris novaehollandiae is known to defend only a very small volume of space, immediately surrounding either its feeding or breeding territory (Paton 1979; McFarland 1986b; Paton 1993; Pyke and O'Connor 1993). Therefore, at the scale of 1ha, a considerable amount of structural space is likely to be undefended, allowing less dominant honeyeaters into the area. The results here showed that Melithreptus are not excluded from a 1 ha area through competition with P. novaehollandiae, but in order to investigate the issue of competition more closely, the relationships between the species need investigation at scales less than 1 ha. Such smaller scale interactions will be addressed in Chapter 3.

The importance of gum species (in particular E. viminalis to M. lunatus) was not unexpected given anecdotal reports in the literature. However, it is an important finding that M. lunatus abundance is most closely linked to these species. If the importance of gum woodlands can be extrapolated to Melithreptus, one major cause of their decline – loss of habitat – becomes evident, given the preferential clearance of such habitat in the Mt Lofty Ranges (e.g. Armstrong et al. 2003; Paton et al. 2004b).

63 Chapter 3 SMALL SCALE STUDIES RELATED TO COMPETITION AND

MELITHREPTUS ECOLOGY

3.1 Introduction

The results of Chapter 2 were not as expected if competition was occurring between Melithreptus and Phylidonyris novaehollandiae, although they were consistent with competition between Anthochaera carunculata and M. brevirostris. However, the results did not exclude competition as a possibility between Melithreptus and P. novaehollandiae if they have a shared preference for resources at the scale of 1 ha. In the Mt Lofty Ranges, previous work (Ford 1977; 1979) has shown that it is in the Melithreptus non-breeding season (roughly January to June) that resources are most likely to be limiting to honeyeaters. Do the Melithreptus compete with widespread and abundant honeyeaters during this time? The results of Chapter 2 indicated that P. novaehollandiae was present at 84% of sites and the mean total P. novaehollandiae per site was 33 birds. This is one and a half times as widespread and over ten times as abundant as M. brevirostris and over two times as widespread and over five times as abundant as M. lunatus. Further, only 1.67% of the M. brevirostris recorded and 0.75% of the M. lunatus recorded occurred at sites without any P. novaehollandiae. Being so widespread and abundant, P. novaehollandiae may not have to spend much of its time foraging from similar substrates to compete with Melithreptus for food through depleting resources (exploitation competition due to being widespread and abundant). Behaviour data collected (Chapter 6) indicated that there was no statistical difference in the proportion of time P. novaehollandiae and Melithreptus spent foraging from bark or foliage, although in both cases the trend was for P. novaehollandiae to spend less time on both. Besides this potential exploitative competition, aggression data from the behavioural study showed that P. novaehollandiae accounted for 73% of aggressive interactions involving M. brevirostris (n = 65 observations), 56% of aggressive interactions involving M. lunatus (n = 102) and 61% of aggressive interactions involving M. gularis (n = 18). This suggests that for Melithreptus, P. novaehollandiae is also the most important interference competitor. Thus, competition between P. novaehollandiae and Melithreptus seems likely.

64 This chapter is in two parts. Part one (Section 3.2) investigates the possibility of interspecific competition limiting the distribution and/or abundance of M. lunatus through an experimental removal of P. novaehollandiae. Part two (Section 3.3), having confirmed the likelihood of such competition, investigates two behavioural strategies, termed swamping and stealth, used by Melithreptus to improve their access to any resources they share with P. novaehollandiae. This investigation was done by watching individual trees and timing visit lengths of honeyeaters. Both the experiment (Section 3.2) and the tree watches (Section 3.3) were undertaken to provide results meaningful at a small spatial scale. While the experimental removal was not able to distinguish small temporal scale interactions, the tree watches were designed to investigate both small scale spatial and temporal interactions.

3.2 Response of M. lunatus to removal of P. novaehollandiae

3.2.1 Introduction If competition is occurring through either interference or exploitation competition with P. novaehollandiae, or most likely a combination of both, the hypothesis can be made that Melithreptus will not occur in certain areas of the landscape due to the presence of P. novaehollandiae. These areas are likely to occur at a scale less than 1 ha. In order to test this hypothesis, a study able to discern associations between P. novaehollandiae and Melithreptus at a scale of tens of metres was undertaken and combined with the removal of P. novaehollandiae from the site.

3.2.2 Methods The experiment was carried out during late February and early March 2003 at Charleston Conservation Park. The main overstorey species on the grid were Eucalyptus leucoxylon, E. viminalis and E. camaldulensis. The understorey was open with a mix of Banksia marginata, Xanthorrhoea semiplana and low shrubs. The experimental area was a square of nine hectares with transects running 25 metres apart from east to west and from north to south, making a total of 26 transects (each line of the grid in Figure 3.2). Only honeyeaters 25 metres either side of the transect were counted. It was necessary to limit data collection to the transect method used, in order to obtain accurate and independent spatial information on the distribution of honeyeaters within the study site. A GPS (Garmin 12XL) was used to keep track of the transect lines and to record

65 the position of each individual honeyeater encountered. For each honeyeater, the plant species and activity were also recorded, with any aggression noted. Each transect was walked once to make a single count of the area (thus, each 25m×25m grid cell was counted four times). In order to maintain equal effort over all grid cells, records outside the grid were taken while walking the outside transects but these records were not included in analyses. A count was undertaken twice a day for three days before removal of P. novaehollandiae (before treatment), twice a day for the three days after removal (removal) and twice a day for three days after release of captured P. novaehollandiae back to the site (release), totalling 140 km of transects over the course of the experiment. The first count of the day started about half an hour after sunrise and took about five hours. The second count of the day started about five and a half hours before sunset and always finished before sunset. Adjacent transects were not walked consecutively.

Sixteen P. novaehollandiae were removed from the study area and subsequently released at their capture sites three days later. There was no difference in the weights of these honeyeaters between the time of their capture and their release, suggesting that the birds did not lose condition during their three day detention (paired t-test, t = -1.0105, df = 15, P = 0.3283).

Three results were used to summarise the results of the experiment. Firstly, the abundance of each honeyeater species was compared between each treatment. Secondly, the distribution of the three most abundant honeyeater species were compared (P. novaehollandiae, M. lunatus and Lichenostomus penicillatus). Finally, the abundance of M. lunatus within each 25×25m cell before and during removal treatments was plotted against the abundance of P. novaehollandiae within each cell before removal.

3.2.3 Overall summary of results The total records, treatment records and estimated grid population of all honeyeater species recorded on the grid during the study period are listed in Table 3.1. Phylidonyris novaehollandiae was by far the most recorded species, followed by M. lunatus and then L. penicillatus.

66 Figure 3.1 shows the mean number of records for P. novaehollandiae, M. lunatus and L. penicillatus during each of the treatments. There was clearly a drop in records of P. novaehollandiae after the removal of 16 individual P. novaehollandiae. However, there was little change in the numbers of M. lunatus and L. penicillatus recorded within the study area. Also, there was little increase in the records of P. novaehollandiae after the release of the 16 captured individuals.

Phylidonyris novaehollandiae was the only species recorded being aggressive, and it was always directed at M. lunatus, except in one case where it was directed against a pardalote (Pardalotus striatus). However the overall number of records of aggression was too low for analysis (3.4% of M. lunatus records involved the individual being chased or displaced – in total 6 records).

Table 3.1: Total numbers, numbers during each treatment and estimated grid population for all honeyeater species recorded

Estimated grid population = Treatment Records/(6 ×4 counts of each cell). Presented in decreasing order of total records. Treatment Records Estimated Grid Population Species Total Records Before Removal Release Before Removal Release Phylidonyris novaehollandiae 479 281 84 114 11.71 3.50 4.75 Melithreptus lunatus 175 49 50 76 2.04 2.08 3.17 Lichenostomus penicillatus 83 33 32 18 1.38 1.33 0.75 Phylidonyris pyrrhoptera 53 28 17 8 1.17 0.71 0.33 Lichenostomus chrysops 26 7 12 7 0.29 0.50 0.29 Acanthorhynchus tenuirostris 9 4 1 4 0.17 0.04 0.17 Anthochaera carunculata 5 4 1 0.17 0.04 0.00 All Honeyeaters 830 406 197 227 16.9 8.2 9.5

67 Figure 3.1: Mean records (± s.e.) per grid count for P. novaehollandiae, M. lunatus and L. penicillatus during the three treatments

There were six counts in each treatment.

60

50

s 40 d cor

Re Before

of 30 Removal Release age # r

Ave 20

10

0 P. novaehollandiae M. lunatus L. penicillatus Species

3.2.4 Results of association of P. novaehollandiae with M. lunatus and L. penicillatus The distributional data for P. novaehollandiae, M. lunatus and L. penicillatus within the study area are shown in Figure 3.2. The data were analysed using Spatial Analysis by Distance Indices or SADIE (Perry 1998; Perry et al. 1999), and an extension of SADIE in which two different counts can be analysed for spatial association (Winder et al. 2001). SADIE can detect and measure the degree of non-randomness in two dimensional patterns by determining the minimum effort needed to move to a completely regular arrangement in which abundance is equal in each sample unit. Areas within the sample site are identified as patches (neighbourhoods of relatively high density) or gaps (neighbourhoods of relatively low density). SADIE is well suited to analysing the count data common in ecology as it is designed for situations in which species are distributed patchily in discrete aggregations with relatively well defined boundaries and it has the ability to test the patterns observed (Winder et al. 2001). An extension of SADIE enables the quantifying of similarity between two patterns generated using the SADIE technique. Originally designed to compare counts at two

68 times (Winder et al. 2001), it has been used here to compare both counts of the same species at two times, and also counts of different species at the same time.

SADIE analysis on each species at each treatment level found that the counts were clumped with respect to an even distribution (P<0.05 in each case).

A number of comparisons were then made for various Treatment/Species associations: • within treatment comparisons between species: an observational test on associations between species (i.e. is the distribution of M. lunatus within the site associated with the distribution of P. novaehollandiae at any time?); • the before vs. removal treatments between species: an experimental test of the effect of one species on the other (i.e. does the distribution of M. lunatus within the site become more associated with the original distribution of P. novaehollandiae after their removal?); and • between treatments (within and between species): observational tests of changes in distribution over the course of the experiment (i.e. does the distribution of, say, M. lunatus change throughout the course of the experiment due to, for example, changes in resource distribution within the study area across the experimental time period?).

The results (Table 3.2) showed for: • within treatment comparisons: P. novaehollandiae and M. lunatus were always associated, while P. novaehollandiae and L. penicillatus were always distributed randomly with respect to each other (or in one case a weak dissociation); • before vs. removal comparisons: the distribution of P. novaehollandiae within the site did not change after removal of 16 individuals, M. lunatus distribution continued to be associated with P. novaehollandiae distribution and L. penicillatus continued to be distributed randomly with respect to P. novaehollandiae; • between treatments, within species comparisons: the distribution of each species remained the same across the site throughout the experiment; and • the between treatments, between species comparisons: confirmed the results of the within treatments, within species. Phylidonyris novaehollandiae and M.

69 lunatus were always associated, while P. novaehollandiae and L. penicillatus were always distributed randomly with respect to each other.

Overall these results demonstrate that at the Charleston grid site, the distribution of M. lunatus records is highly associated with the distribution of P. novaehollandiae records, while the next most common honeyeater, L. penicillatus, is not. The temporary removal of 16 P. novaehollandiae did not change this in any way. There was no change in the distribution of each species through time at the site during the study period.

Figure 3.2: Raw distributional data from the P. novaehollandiae removal experiment at Charleston Conservation Park

Each square on the grid is 25m×25m. The experimental area (9 ha) extended from 54/ 312475/ 6134400 (NW corner of the grid in AMG coordinates, datum = GDA94) to 54/ 312775/ 6134100 (SE corner). Data presented are for P. novaehollandiae (squares), M. lunatus (circles) and L. penicillatus (triangles) before the removal of P. novaehollandiae (white), during removal (grey) and during release (black).

312450 312500 312550 312600 312650 312700 312750 312800

S# %U %U%U %U%US# S#S# S# S# 6134400 %U %U $T%U%U$T%U# S#S#S# %US#%U S# 6134400 $T #S#%U #%U # %U%U $T %U%U%US#S#%US# # %U %U %U%U%S#%U#%U%US %U S S %U# %U%U%U%US%U%U S# S %U%U%U %U%U S%U%U %U %U S %U %U S# S#S# %U%U %U %U %U %U %U S%U S# S# S# %U%U %U %U%U%U%U%U S# %U %U %U %U S# %U%U %U%U %U %U %US# %U %U%U%US#%U%U %U %U %U%U %U %U%U%US#S#%U%U %U %U%U%U%U %U%U%U S# %US%U#S%U#S#S#S#S%U#%U %U%U %U S# %U %U %U %U %U%U%U%U%U%U%US#S#%U%U%U %U%U%U%US%U%U#%U%U%U%U# # %U S# %U %U%U%U %U%U %U%U %U #%U%USS#S# %U %U %U %U%U S#%USS#%U%U S# 6134350 $T %U S#S# S# %U %U %U%U 6134350 # S# %U S %U S# %U %U%U %U # S# S# %U%%UU%%U S# S %U S%U %U %U %U %U%U S#%U%U%U %US#%U %U %U %U S# %U%U%U%U%U%U %U%U%U %U %U ## %U $T %U$T$T%U %U %U%U%U%U#%U%U%U%U %U S# $T SSS#%U S# %U%U%U %U %U %U%U%U%S#%UU%U%U%U%U%U%U $T %U %U%U %US#%U%U#%U%U%U%U%U%U%U%U%U%U%U%U %U%U %U $T # $T %U %U%U%U%U S%U%U%US#%U%U%US#S%U#%U%U%U%U%U $T $T S %U%U %U %U%U%U%US %U %U 6134300 $T %U #S%U%U#%U %U %U%U%U 6134300 S# %U%U %U %U S %U%U S# S#%U %U S# %U %U%U $T %U S# %U %U S#$T $T S# $T %U%U %U%U%U %U $T $T %U%U $T $T $T$T$T$T $T %U%U S# %U $T$T$T $T $T S%U# %US#S# %U %U $T $T$T # #%U %U %U%U $T S#$T %U S S# S %U%U%U%U%U%U%US#S# $T %US#S# %U %U%U S#S# %U %U%US#%U%US%U%U#S#S%U#%U%U%U%US# %U 6134250 $T S# %U %U%U %U %U %U%U%U%US%U#%U%U%US#%U%U 6134250 $T $T $T%U %U %U%U%U%U%U %U%U %U %U $T$T$T $T%U # %U %U %U %U%U %U $T $T S# S %U%U $T $T S#S# %U%U # S# %U $T %US#$T S# S %U %U $T# S# # %U%U%U%U#S%U%U%U SS# S S#%U S# %U # S%U# S# S 6134200 S# S# S# %U 6134200 $T %U $T $T S# $T $T S# $T $T S# $T $T $T %U$T $T$T $T %U 6134150 6134150 S#

S# $T $T $T $T %U %U%U 6134100 %U 6134100 %U 312450 312500 312550 312600 312650 312700 312750 312800

70 Table 3.2: Results of spatial association tests

NHH = P. novaehollandiae, WNH = M. lunatus, WPH = L. penicillatus. Χ is the overall SADIE measure of spatial association between the two sets of cluster indices. P is the probability of the observed X occurring by chance. 1st Treatment 2nd Treatment Χ P Relationship Within Treatments, between species NHH Before WNH Before 0.4915 <0.0001 association NHH Before WPH Before -0.0456 0.6842 random NHH Removal WNH Removal 0.6298 <0.0001 association NHH Removal WPH Removal -0.1478 0.9602 weak dissociation NHH Release WNH Release 0.4871 <0.0001 association NHH Release WPH Release -0.0332 0.6037 random Before vs. Removal NHH Before NHH Removal 0.7929 <0.0001 association NHH Before WNH Removal 0.6521 <0.0001 association NHH Before WPH Removal -0.1064 0.8896 random Between treatments, within species NHH Before NHH Release 0.6212 <0.0001 association NHH Removal NHH Release 0.7651 <0.0001 association WNH Before WNH Removal 0.4206 <0.0001 association WNH Before WNH Release 0.4276 <0.0001 association WNH Removal WNH Release 0.5829 <0.0001 association WPH Before WPH Removal 0.7457 <0.0001 association WPH Before WPH Release 0.5171 0.0021 association WPH Removal WPH Release 0.5103 0.0068 association Between treatments, between species NHH Before WNH Release 0.4777 <0.0001 association NHH Before WPH Release -0.0444 0.6743 random NHH Removal WNH Release 0.5505 <0.0001 association NHH Removal WPH Release -0.0121 0.5412 random

3.2.5 Results of P. novaehollandiae and M. lunatus abundance at grid cell scale The total cell counts of P. novaehollandiae during the before treatment were divided into three categories: no P. novaehollandiae; between one and three P. novaehollandiae; and greater than three P. novaehollandiae. Within these categories, the mean cell count of P. novaehollandiae during both the before and removal treatments were determined, as were the mean cell counts of M. lunatus during both treatments. Figure 3.3 shows mean cell counts of P. novaehollandiae during the before and removal treatments in the three cell categories and clearly demonstrates the effect of removal. The numbers of P. novaehollandiae in both cell categories which originally had any P. novaehollandiae before removal were greatly reduced during the removal treatment (there is also one record of a P. novaehollandiae in a ‘0’ cell during the removal period).

Figure 3.4 shows mean cell counts of M. lunatus during the before and removal treatments, using the cell categories based on before P. novaehollandiae counts. The

71 mean M. lunatus count in cells with >3 P. novaehollandiae before removal increased from 0.88 M. lunatus before to 1.46 M. lunatus during removal. There was a drop in M. lunatus counts within cells that originally had 1 to 3 P. novaehollandiae and no change in M. lunatus counts within cells that originally had no P. novaehollandiae. Thus, M. lunatus altered its distribution within the experimental area to reflect more closely the original distribution of P. novaehollandiae during a time of reduced P. novaehollandiae presence.

72 Figure 3.3: Mean P. novaehollandiae counts (± s.e.) in the before and removal treatments within three cell categories reflecting counts of P. novaehollandiae in the before treatment

Mean P. novaehollandiae counts ± standard error. 90 cells in ‘0’ P. novaehollandiae, 28 cells in ‘1-3’ P. novaehollandiae and 26 cells in ‘>3’ P. novaehollandiae.

14

12 e ia nd

lla 10 o h e a v o

n 8 . P

Before of

t Removal n

u 6 o c l l e

e c 4 ag er Av 2

0 01-3>3 Average cell count of P. novaehollandiae before removal

Figure 3.4: Mean M. lunatus counts (± s.e.) in the before and removal treatments within three cell categories reflecting counts of P. novaehollandiae in the before treatment

Mean M. lunatus counts ± standard error. 90 cells in 0 P. novaehollandiae, 28 cells in 1-3 P. novaehollandiae and 26 cells in >3 P. novaehollandiae. Statistically, there was an interaction effect of P.

novaehollandiae category and treatment on M. lunatus counts (F2,282=3.1366; P=0.0449). There was no

effect of treatment (F1,282=1.0315; P=0.3107) but an effect of P. novaehollandiae (F2,282=32.0083; P<0.0001). Post hoc contrasts on treatment within P. novaehollandiae category showed no change in M. lunatus counts within P. novaehollandiae category 0, no change in M. lunatus counts within P. novaehollandiae category 1-3 and an increase in M. lunatus category within P. novaehollandiae class >3.

2

1.8

1.6

1.4 natus lu . M

1.2 of t Before 1 Removal ll coun

e 0.8 c

age

r 0.6 Ave 0.4

0.2

0 01-3>3 Average cell count of P. novaehollandiae before removal

73 3.2.6 Discussion The result here extends the findings of the wider surveys. Melithreptus lunatus is ‘associated’ with P. novaehollandiae in a similar way to M. brevirostris and M. gularis. While vegetation limits the abundance of M. lunatus at a landscape scale, the results of the removal experiment suggest that P. novaehollandiae may be competing with M. lunatus within a site.

This experiment was carried out as P. novaehollandiae was thought to be influencing the distribution and/or abundance of Melithreptus through competition (interference and/or exploitation). The results clearly showed no effect of removal of P. novaehollandiae on the mean counts of M. lunatus over the 9 ha despite a decrease in the mean counts of P. novaehollandiae across the same area. There are a number of possible reasons for this: • lack of M. lunatus in neighbouring areas to move into the study area; • a hesitancy by neighbouring resident M. lunatus to move and/or lack of time for non-resident M. lunatus to find the area; • Phylidonyris novaehollandiae numbers were not reduced to a low enough level for a numerical response to be seen; and/or • Melithreptus lunatus in the area were using different resources and that the removal of P. novaehollandiae was never going to effect the abundance of M. lunatus within the study area (although this assumes no interference competition).

The last possibility appears unlikely given the results here for the distribution data and the results for other Melithreptus (Chapter 2). Perhaps there is some evidence for the second point, in that during the release treatment, mean counts of P. novaehollandiae did not increase back to before levels, but mean counts of M. lunatus numbers did increase – perhaps non-resident birds had moved into the area. There is little evidence in the results or previous observations to discriminate between the other possibilities.

The fact that the distribution of M. lunatus was highly associated with the distribution of P. novaehollandiae during all treatment combinations, while the next most common honeyeater, L. penicillatus, was distributed randomly or dissociated from P.

74 novaehollandiae supports the theory that M. lunatus and P. novaehollandiae are likely to compete for similar resources, as the most probable explanation for their association is a shared preference for resources. This is in contrast to L. penicillatus, which is certainly not responding to the same resources as P. novaehollandiae and M. lunatus. However, the high association between P. novaehollandiae and M. lunatus at all treatment levels made it difficult to determine if, when P. novaehollandiae was removed, the association between the distribution of M. lunatus and the original distribution of P. novaehollandiae became stronger.

The best test of any possible competition between P. novaehollandiae and M. lunatus comes from cell based results. Melithreptus lunatus abundance increased in cells in which P. novaehollandiae had been in high numbers before removal of 16 individuals. This suggests that M. lunatus were able to change their distribution within the site from one reflecting sub-optimal P. novaehollandiae habitat to one reflecting optimal P. novaehollandiae habitat. This result supports the original hypothesis that Melithreptus are unable to use certain areas of the landscape due to competition with P. novaehollandiae. When the competition was lessened through removal of P. novaehollandiae, M. lunatus were recorded more frequently in the areas used most by P. novaehollandiae. If this experiment were to be repeated, it is this interaction that should be the primary target of analysis. Due to the low incidence of aggression, it appears likely that any competition is mostly exploitative, not interference.

The failure of P. novaehollandiae to return to original levels during the release treatment may be indicative of the released individuals having difficulty re-establishing territories on the grid, but could also be due to decreased resources within the grid across the experimental time period, or both. It is interesting to note that if resources did decline, it is unlikely that there was any change in their spatial distribution, as each species had a stable spatial distribution throughout the experiment. Colour-banding individuals that were removed before their release would have facilitated determining if the released individuals re-established on the study site.

Finally, the conclusions drawn by the experiment would be strengthened through replication. The methods used here were extremely labour intensive, but were necessary to obtain enough data points for M. lunatus, which was present in low numbers on the

75 study site (estimates of grid population were between 2 and 3 individuals), despite being an area in which M. lunatus was the second most abundant honeyeater. Perhaps an improvement on this experimental design to make the intensity of effort more manageable would be to use a smaller area, say 1ha, with replicated 1 ha areas at different sites. Further improvements such as reciprocal removal of M. lunatus to determine the effect on P. novaehollandiae and paired control sites with no removal would also improve confidence in the conclusions drawn from the results obtained here.

3.3 Comparison of behavioural techniques typical of M. brevirostris and M. lunatus in accessing defended resources

3.3.1 Introduction Distinct preference for resources occurs when two species prefer different ranges along a niche gradient. Each species has its highest fitness within its preferred range on the niche axis. In this case, the removal of one of the species does not necessarily result in any change to the fitness of other species (Wisheu 1998). Shared preference for resources occurs when a number of species prefer the same range along a niche gradient. In the absence of other species, each species has its highest fitness along the same range of the niche axis. However, when occurring in sympatry, each species has its own competitive abilities based on other aspects of its niche and, ‘intolerant, competitive dominants occupy the preferred segment (range) of the environmental gradient (niche axis) while tolerant, subordinate species occupy regions with suboptimal levels of resources’ (Wisheu 1998). Melithreptus brevirostris, M. lunatus and P. novaehollandiae (and probably M. gularis) appear to have a shared preference for resources when occurring in sympatry.

Given the dominance of P. novaehollandiae numerically and aggressively over M. brevirostris and M. lunatus, the two Melithreptus species are probably the ‘tolerant, subordinate’ species of the above quote. Is there an aspect of the ecology of these two smaller species that allows them access to resources in limited supply due to the abundance and aggression of P. novaehollandiae? Melithreptus brevirostris and M. lunatus are shown in other parts of this study to have very similar ecology. Areas of similarity include weight, functional morphology, feeding substrate, and to a lesser extent, home range size and habitat. The most consistent difference between M.

76 brevirostris and M. lunatus is their flock size, with M. brevirostris moving in larger groups during both breeding and non-breeding seasons. Based on their flock size differences, one theory for the coexistence of M. lunatus and M. brevirostris is that they are competing for the same resource through two different behavioural techniques.

Two behavioural techniques are suggested here to provide M. brevirostris and M. lunatus access to resources in limited supply due to the abundance and aggression of P. novaehollandiae. These techniques will be referred to as stealth and swamping. In the Mt Lofty Ranges, M. lunatus generally use stealth to access resources defended by P. novaehollandiae. They move quietly in small groups or as individuals and spend as long as possible feeding in good areas before being found and chased out (thus, M. lunatus are less likely to move over short time frames). Melithreptus brevirostris use swamping to access resources defended by P. novaehollandiae. They move in large groups, calling a lot and descend on the territory(ies) of other honeyeaters, so that even if a bird gets chased out immediately, others in the group get time to feed before either being chased out themselves or leaving to follow other birds in the flock that have already been chased out (thus, M. brevirostris are more likely to move over short time frames). An alternative way to look at this theory is that M. lunatus find ‘holes’, while M. brevirostris make ‘holes’, in a layer of aggressive honeyeater species that cover the resources within their home range. Community organisation including a behavioural niche has already been demonstrated by other authors for various miner species (Section 1.1). The suggestion here is that swamping and stealth behaviours enable partitioning of resources in honeyeater communities. These behaviours have received little attention in the literature regarding community organisation.

Stealth behaviour has been documented in a montane hummingbird association in Mexico by the White-eared Hummingbird (Hylocharis leucotis) (Lyon 1976). Hylocharis leucotis showed the most flexible feeding behaviour of the six species of hummingbird present in the study area and included a behaviour consisting of secretive, low approaches to the territories of larger species. This behaviour allowed H. leucotis to feed for longer before it was discovered in the other bird’s territory and it was the only species within the hummingbird community to exhibit the behaviour. The behaviour was most effective at medium levels of floral resources, with flowers widely dispersed. As flower density increased, the territories of large, aggressive species became smaller

77 and more numerous, with a corresponding increase in the efficiency of territorial defence. Lyon (1976) thought that the flexibility in behaviour of H. leucotis was important in the success of the species in montane Mexico where it was reported to have wide geographic, elevation and habitat tolerance. A further point of interest is that H. leucotis was the only abundant year-round resident in the study area (Lyon 1976). Paton (1979) documented an occasion in which an individual P. novaehollandiae learnt to raid part of an Anthochaera chrysoptera territory without being detected immediately. McFarland (1996) also noted that the, ‘small solitary spinebills may not be as readily noticed as larger species’, in a discussion of the low percentage of Acanthorhynchus tenuirostris attacked compared with other species when they intruded to feed on a P. novaehollandiae territory. But are these cases in which the territorial owner tolerates the presence of individuals calculated to rob little nectar, and therefore not worth chasing out? This is a possibility, but in each case the authors specified that the behaviour appeared to result in the invader remaining on the territory for longer before being detected, with detection then resulting in a chase or displacement and ejection (Lyon 1976; Paton 1979; McFarland 1996). That detection is an important part of territorial defence has been documented (Paton and Carpenter 1984), further suggesting that exploiting weaknesses in defence through stealth is a legitimate technique for accessing limiting resources.

Swamping as a method of accessing resources held by aggressive species is perhaps more commonly recognised than stealth (one would hope so…) and has been mentioned by a number of authors. Paton (1979; 1980) noted that P. novaehollandiae had difficulty driving off flocks of M. brevirostris that had ‘swamped’ their territories. McFarland (1986a) thought flocking reduced the interspecific dominance of larger honeyeaters, allowing L. chrysops, M. lunatus and Myzomela sanguinolenta (Scarlet Honeyeater) to find, swamp and feed on clumped resources with ‘reduced individual risk’. Slater (1994) suggested that flocking by Melithreptus in Tasmania enabled the two species to overrun the territories of larger honeyeaters and ‘temporarily plunder its resources’. Finally, Timewell (1997) discussed the ability of flocks of M. lunatus to swamp a flowering patch and feed on nectar before being aggressively ejected by Anthochaera carunculata.

78 An attempt to test the theory of stealth and swamping was made by watching individual (flowering) trees and recording the number of honeyeaters moving into and out of the trees and the time at which these movements occurred. Individual trees were used as they provided a discrete sampling unit which could be (fairly) easily watched and honeyeater visits documented. Further, on completion of the watch period, a measure of the resources and volume represented by the sampling unit could be evaluated, and a reasonable assumption made that resources were equally available within the sampling unit. Three study sites were used, Monarto State Forest, Scott Creek Conservation Park and Scott Conservation Park. All observations occurred during the Melithreptus non- breeding season.

The following predictions were made from the flocking and stealth theory: • swamping should allow access to trees with better resources than stealth (The swamping advantage should allow access the best resources that are probably too highly defended for stealth to be an effective means of access); • stealthy birds should feed until forced out or for a very short time (If a stealthy bird accesses a tree with good resources it should remain in the tree until forced out. If the resources are poor the bird should leave anyway after a brief sampling period to determine resources are poor); and • the feeding time of swamping birds should be unrelated to reason for leaving, or only slightly longer for just leaving (If enough swamping birds are forced out of a tree, other members of the group will follow regardless of whether or not they are forced out. It is this behaviour that maintains the group and hence the swamping advantage).

3.3.2 Methods Individual trees at Monarto State Forest, Scott Creek Conservation Park and Scott Conservation Park were watched for between 30 minutes and 1 hour (if Melithreptus were in the tree at half an hour, the watch continued until all Melithreptus left or 1 hour had passed). Tree watches occurred over several days during the Melithreptus non- breeding season in 2003. Honeyeaters in the tree at the start of the watch were estimated and the number, species and time of any exits or entries of honeyeaters to the tree were recorded. A final estimate of honeyeaters in the tree at the end of the watch period was made to provide a check on precision of the data collected. When an individual left the

79 tree, the reason (‘just left’ or ‘forced out’) was also recorded. These data were used to determine the mean visit length for each species from the total amount of time spent in the tree by all individuals of a species divided by the number of individuals recorded leaving the tree.

For each tree watched, the height of the canopy top and bottom, maximum and minimum canopy width and level of flowering were measured. Flowering was measured on a scale of 0 to 5; 0 = none to 5 = super-abundant. Together with honeyeater data, data on trees were used to determine the ‘standardised protection of resources’ by aggressive species for each tree observed using:

n ∑ w× si i=1 S (Equation 1), 4 ×π × r × r × r × F 3 ch max min where si was the number of seconds individual i of n individuals spent in the tree, w is the mean weight of the species, S is the total seconds the tree was observed for, rch is half the vertical canopy spread (canopy top – canopy bottom), rmax is the maximum radius of the canopy, rmin is the minimum radius of the canopy and F is the intensity of flowering of the tree. The ‘standardised protection of resources’ (SPR) by aggressive species (P. novaehollandiae and A. carunculata) was the sum of the above formula for each species. The term above the line in Equation 1 is a measure of the protection of a tree by aggressive honeyeater species, being the proportion of time that the tree is occupied by aggressive honeyeater species multiplied by the mean weight of that species. For example, a tree which had 2 P. novaehollandiae (mean weight 21 grams) remaining in the tree throughout the watch period (say, 30 minutes) would score (21× 30× 60 + 21× 30× 60) = 42 g of aggressive honeyeater (g ah). The term below the 30× 60 line in Equation 1 is a measure of the ‘attractiveness’ of the tree to a honeyeater, being 4 the volume of the canopy (from ×π × r 3 , the volume of a sphere) multiplied by a 3 score of the flowering intensity. For example, if the above tree had a flowering intensity of 4 (abundant), a canopy height of 10m, a maximum canopy width of 8 m and a

80 4 10 8 6 minimum canopy width of 6 m, the attractiveness would be ×π × × × × 4 = 418 3 2 2 2 m3.F. Thus, overall the SPR for the tree would be 42/418 = 0.10 g ah/(m3.F).

Statistical tests on the effects of species and reason for leaving on visit time and SPR were carried out on ln transformed data after the Shapiro-Wilk W test for normality using JMP IN software (SAS Institute Inc. 1997).

3.3.3 Results 58 different trees were watched for a total of 35 hours at the three sites. Phylidonyris novaehollandiae and A. carunculata dominated the sites both numerically (Table 3.3) and aggressively. Melithreptus brevirostris were the next most recorded species. Melithreptus brevirostris as a species accumulated more time per visit than other species of small honeyeater (Figure 3.5) and moved into trees in larger groups than other small honeyeaters (Table 3.3).

Melithreptus brevirostris were forced out sooner than M. lunatus, but when just leaving there was very little difference between the two species (Table 3.4, Figure 3.6). There was little difference in the time spent in a tree by M. brevirostris, whether or not the reason for leaving was forced out or just left.

The SPR of trees visited by M. brevirostris was higher than that of trees visited by M. lunatus and the SPR of trees from which both species were forced out was higher than when they just left (Table 3.5, Figure 3.7).

81 Table 3.3: Sample size (trees, arrivals and entries) and mean arrival size of all species recorded entering focal trees

Sorted by total number of individuals recorded entering a tree over all visits. Trees is the number of watched trees in which the species was recorded. Arrivals is the total number of arrivals by the species. Individuals is the total number of individuals in all arrivals during all visits. Arrival size is the mean number of individuals present per arrival. The actual number of individuals was probably lower, as some individuals undoubtedly revisited trees after leaving them. Group size for the species will be larger as not all members of a flock entered the same tree at the same time when landing in an area. Vertebrate Trees Arrivals Individuals Mean arrival size Phylidonyris novaehollandiae 30 183 240 1.3 Anthochaera carunculata 31 172 225 1.3 Melithreptus brevirostris 26 83 191 2.3 Lichenostomus ornatus 19 53 65 1.2 Melithreptus lunatus 12 34 48 1.4 Glossopsitta porphyrocephala 8 28 44 1.6 Glossopsitta concinna 2 7 18 2.6 Lichenostomus chrysops 5 9 11 1.2 Lichenostomus virescens 3 6 6 1.0 Zosterops lateralis 2 5 6 1.2 Phylidonyris albifrons 1 3 3 1.0 Pardalotus punctatus 2 2 2 1.0 Psephotus haematonotus 1 1 2 2.0 Lichenostomus penicillatus 1 1 1 1.0 Pachycephala pectoralis 1 1 1 1.0 Rhipidura fuliginosa 1 1 1 1.0

Figure 3.5: Mean cumulative time (± s.e.) for each species in focal trees

Order of presentation as in Table 3.3. Mean cumulative time is the number of individuals of a species × seconds those individuals were present in a watched tree. Sample size for each species is the number of trees given in Table 3.3. ) s d

n 6000

5000

4000 er tree (seco p

e 3000 tim

e 2000 tiv

la 1000 mu 0 cu

n e s s s s a a a ta ri us na ns li us u e la st at ops e a rons otus at ali u natus cin s sc er f ll r inos landia iro lun ephal n ry on ci to M re lbi c lig unc v s or oc o ch lat a i e u re u us r vi punctat mat en hol b ta c s ps s e p p ae car m pt hy t us u o ri us la a f v a to re p i m m r y ha to o te ot us dur ptus ith or ops st s us m pha i s no e l p os idon dal t to e ip i haer enos a ss n no Zo l r yc h thr ch Me itt e hy pho os R hoc li Glo che h P Pa ch nt Li ps i c se hen donyr A Me so L Li P c Pa yli h os Li P Gl Species

82 Table 3.4: Mean visit times (seconds) for small Melithreptus and reason for leaving

Statistically, there was no interaction of species and reason for leaving (F1,27=0.0686; P=0.7953), no effect of species (F1,27=0.1591; P=0.6931) but an effect of reason for leaving (F1,27=5.4226; P=0.0276) on visit time. Melithreptus brevirostris fed for a statistically similar amount of time in trees whether or not they were chased out, while M. lunatus fed for statistically longer in trees from which they were chased out. Both species fed for a statistically similar amount of time when they chose to just leave. Mean visit time (seconds) M. brevirostris M. lunatus Both Species Forced Out 116.1 321.0 210.6 Just Left 132.2 139.3 134.5 All exits 126.2 230.1 166.4

Table 3.5: Mean SPR for small Melithreptus and reason for leaving

Statistically, there was no interaction of species and reason for leaving (F1,27=0.6603; P=0.4235), but

there was an effect of species (F1,27=5.2729; P=0.0296) and reason for leaving (F1,27=7.0880; P=0.0129) on SPR. Melithreptus brevirostris visited trees with a higher SPR regardless of reason for leaving. Both species just left trees with a low SPR and were forced out of trees with a high SPR. See text for explanation of SPR units. Mean SPR (((g ah)/(m3.F))/100) M. brevirostris M. lunatus Both Species Forced Out 3.72 1.24 2.58 Just Left 1.29 0.31 0.96 All exits 2.18 0.78 1.64

83 Figure 3.6: Mean visit time (± s.e.) for M. brevirostris and M. lunatus and reason for leaving

Total sample size is trees visited by the species, as given in Table 3.3.

700

600 )

s 500 d n

400 me (seco i M. brevirostris t T i

s M. lunatus i 300 e V g a r e v

A 200

100

0 Forced Out Just Left Species

Figure 3.7: Mean SPR (/100) (± s.e.) of trees visited by M. brevirostris and M. lunatus and reason for leaving

Total sample size is trees visited by the species, as given in Table 3.3.

6 0 /10 )) .F

3 5 )/(m h a

) ((g 4 R P (S

n M. brevirostris tio

c 3 e

t M. lunatus o r

urce P 2 so e R ised

rd 1 a d n a t S 0 Forced Out Just Left Species

84 3.3.4 Discussion The results of this study support the predictions made from the theory of swamping and stealth: • birds using swamping should access trees with better resources than those using stealth; • birds using stealth should feed until forced out or for a very short time; and • for birds using swamping their feeding time should be unrelated to reason for leaving.

Melithreptus brevirostris were able to use trees with a high level of SPR, especially compared with M. lunatus and moved in larger groups, entering a tree with the largest group size of any honeyeater species. However, M. brevirostris were forced to move more frequently than M. lunatus.

Melithreptus lunatus did remain in a tree longer when they were forced out than when they chose to just leave and were forced out of trees with a higher level of SPR than trees where they chose to just leave. This suggests that M. lunatus was able to ‘sneak’ into trees that were part of the territory of an aggressive honeyeater and remain there for a while before being found and chased out. The large standard error (visible in Figure 3.6) is likely due to real differences experienced by M. lunatus when accessing defended resources; sometimes they are able to feed for long periods of time before being found, while at other times they may be chased out almost immediately. Presumably, individual M. lunatus would be more likely to revisit the trees at which they were allowed to feed for longer before being chased off. Thus, a good further test of the theory could be that an individual M. lunatus should revisit trees at which it succeeds in feeding for longer more frequently than an individual M. brevirostris should revisit trees at which it is allowed to feed for longer. It is also clear that M. lunatus visited trees for much less time when the individual chose to just leave the tree. Trees which M. lunatus chose to just leave had a very low SPR, suggesting that these trees had poor resources anyway.

Finally, M. brevirostris leaving time was unrelated to reason for leaving. Melithreptus brevirostris were chased out after a shorter time than M. lunatus, but left soon after,

85 anyway. Thus, those individuals that are not chased out immediately do get to feed for a small amount of time before leaving to follow other individuals that have already been chased out and moved on.

What about an effect of M. brevirostris depleting resources faster due to its larger group size and having to move frequently because of this? While, this may play some role in M. brevirostris spending less time in a tree, it does not explain the different flock sizes used by two closely related (morphologically and congeneric) species. The different flock sizes are likely to serve some ecological function, not simply cause M. brevirostris to move more frequently (i.e. Why would an individual M. brevirostris remain in the flock if being forced to continually move when resources are depleted? Why not simply forage alone (as M. lunatus often does), reap the rewards of resources less quickly depleted and move less often?). Swamping and stealth have been shown here to provide one explanation for the different flock sizes.

Some further pieces of evidence support the theory that swamping and stealth are effective behavioural techniques. Presumably stealth is most successful when coupled with a knowledge of levels of both resources and aggressive defence of those resources. Therefore, the prediction could be made that if M. lunatus leave an area they have knowledge of, they may switch to a swamping technique. While there is no published evidence of swamping per se, there is a well documented migration of M. lunatus in flocks in the eastern states of Australia (e.g. Keast 1968b; Higgins et al. 2001). Migrating birds are unlikely to have good knowledge of either resource levels or defence of those resources in most of the areas they pass through, suggesting that the flocks of migrating M. lunatus may indeed use a swamping technique. Further, at times during this study, M. lunatus were seen in flocks, both conspecific and mixed (including at times M. brevirostris), and at times M. brevirostris were seen in pairs and as individuals (Chapter 5). A crossover of the two behaviours was not limited to the overall approach. Upon landing to feed, a flock of M. brevirostris will often become very quiet, with just the odd ‘chip’ until the whole flock leaves the area, at which time they return to constant calling. This is perhaps an attempt to exploit any hole(s) they have made by landing en masse in the territory(ies) of aggressive honeyeaters in much the same way that M. lunatus does. If not chased out immediately, an individual M. brevirostris will have the best chance of feeding if it remains quiet while doing so. [As

86 an aside, I have noticed that the contrast in calling between a foraging M. brevirostris flock and a moving M. brevirostris flock can give the impression that a flock has left the area, when in fact it has not (perhaps increasing the flocks swamping effectiveness)]. In the Mt Lofty Ranges it appears that swamping and stealth are typical of M. brevirostris and M. lunatus respectively, but not exclusive to M. brevirostris and M. lunatus respectively.

The results of this study at a small spatial and temporal scale support the hypothesis that when accessing defended resources, M. brevirostris and M. lunatus generally use two different behavioural techniques. These two techniques, termed swamping and stealth are essentially a trade off between group size and resource quality. Melithreptus brevirostris move in larger groups which enable access to better resources, albeit for a shorter time, while M. lunatus move in smaller groups allowing longer access to resources that are not as good. While no resource shortage was demonstrated here, it is likely that the widespread and abundant honeyeaters, particularly P. novaehollandiae and Anthochaera carunculata, cause a limitation at least during certain times of the year. At these times M. brevirostris and M. lunatus are able to make use of two different behavioural techniques, or behavioural niches, to access resources in limited supply.

87 Chapter 4 MORPHOLOGY OF SMALL HONEYEATERS IN THE MT LOFTY

RANGES

4.1 Introduction

Koehl (1996) reviews the importance of morphology to various disciplines of biology, including ecology. A number of cautionary cases are reviewed, including cases in which: morphology does not affect performance (allowing diversity of form); cases in which small or simple changes in morphology lead to quite novel functions; and cases where the effect of morphology is dependent on other considerations such as habitat or other characteristics of the organism’s body. However, Koehl’s main concern in outlining surprising and complex relationships between morphology and function is to emphasise the importance of mechanistic information, ‘although there are instances when the function of an organism has been inferred successfully from its structure alone, many other cases exemplify the problems of trying to read function from morphology without the aid of mechanistic information.’ The conclusion for ecologists is that they should care about the mechanisms by which morphology affects performance as it can provide insights about processes affecting the structure of populations, communities and ecosystems.

Two studies have examined morphology in detail within Melithreptus. Slater (1994) looked at numerous morphological attributes of Melithreptus affinis, M. validirostris and Lichenostomus flavicollis (Yellow-throated Honeyeater), three species of short- billed, sympatric honeyeaters in Tasmania. While there was no overlap between the three species in weight or bill length, the additional morphological attributes measured did reflect the differences in foraging ecology.

Keast (1968a) also examined Melithreptus morphology in detail. He suggested his Melithreptus groups 1 and 2 (Section 1.5) have smaller members than group 3 and that the overall difference in linear dimensions of 10% was enough to prevent interspecific competition ‘under mainland conditions’. However, this did not explain the large area of sympatry of M. lunatus and M. brevirostris in south-eastern Australia, which Keast suggested was ‘sufficient to make ecological separation advantageous as a means of restricting competition’. Keast’s results (Mt Lofty Ranges results in Table 4.1) showed

88 that in the area of sympatry M. lunatus had longer wings than M. brevirostris which gave M. brevirostris a larger bill/wing ratio and tarsus/wing ratio. It is unclear whether Keast felt this was sufficient to prevent competition, and the biological implications of such differences were certainly not made clear. A serious impediment to further interpretation of this work is the lack of any published methods, which were referred to in Keast (1968a) as Keast (1969), a work apparently not published.

There were two main aims to this morphological study. Firstly, the morphological data were used to sex individual honeyeaters. There is no accurate method for sexing individual honeyeaters of all species in the field. Most honeyeaters exhibit sexual dimorphism, with males tending to be larger than females, for example, male nectarivores generally have greater wing lengths than females, with only a few known exceptions (Collins and Paton 1989). However, there is usually some level of overlap between the sexes, preventing confident sexing of all individuals based on only one attribute (Higgins et al. 2001). A combination of morphological attributes is therefore likely to provide more accurate sexing than a single attribute. Secondly, the work presented in this chapter aimed to examine the morphology of M. brevirostris and M. lunatus to determine any differences which could explain, or suggest factors that may explain, their sympatry despite otherwise similar niches. In determining differences between M. brevirostris and M. lunatus it was useful to compare similarity between these two species with the rest of the small honeyeater community, in which the effects of differing morphology on ecology are better known.

Table 4.1: Percent differences in various morphological attributes between M. brevirostris and M. lunatus in the Mt Lofty Ranges as determined by Keast (1968a)

Data are percent differences in mean values based on adult males (between M. lunatus and M. brevirostris, i.e. M. brevirostris have shorter wings than M. lunatus). The only attribute for which non- overlap exceeded 75% was wing length in which all male M. lunatus had longer wings than any M. brevirostris males. Sample size was 11 for M. lunatus and 13 for M. brevirostris. The actual data (lengths) were not provided. Attribute % difference between M. lunatus and M. brevirostris Wing length -14 Bill length -3 Tarsus length 2.5 Hallux + Claw length 3

89 4.2 Methods

4.2.1 Measurements Birds were caught in mist nets at, or in the vicinity of, Newland Head Conservation Park, Scott Conservation Park, Bullock Hill Conservation Park, Scott Creek Conservation Park, Kaiser Stuhl Conservation Park, Charleston Conservation Park and Monarto State Forest. The methods used for measuring morphology were chosen to provide the most precise measurements in the field while limiting the handling time and discomfort for the birds. Birds of all ages were measured, but measurements on juveniles were excluded from any analysis. I was the only observer to take measurements, except for the three M. gularis caught by David Paton (see below).

Total body, wing and tail lengths were measured to the nearest 1 mm with a ruler. Wing length was measured using maximum chord; the wing was flattened and straightened as much as possible for measurements of the maximum distance between the carpal joint and the tip of the longest primary (Higgins et al. 2001). Maximum chord was used to avoid the variability that can be generated by differences in stretching effort exerted when measuring wing span, although maximum chord is still subject to seasonal variation through wear of primaries (Collins and Paton 1989). Tail length was measured from the point of emergence of the central rectrix from the skin to the end of the longest rectrix (Higgins et al. 2001). Total body length was measured both dorsally (originally found to be easier) and ventrally (more commonly used elsewhere) from the tip of the bill to the end of the longest rectrix with the bird stretched as far as practicable. Any variation in measures of wing and tail (and hence total length) associated with moult were ignored.

Callipers were used to measure lengths of total head (THL), tarsus, hallux, claw (on rear toe), bill gape and bill top to the nearest 0.1 mm. THL was measured from the back of the skull to the tip of the bill. There are numerous measures of beak length available, the most common being true culmen, exposed culmen, tip to nares and tip to gape (Paton and Collins 1989). The beak length most common in Higgins et al. (2001), based on museum specimens was true culmen. The use of beak gape (tip to gape) and beak top (exposed culmen) achieved best the aim of precision and minimum handling.

90 Weights were measured with a Sartorius PT 120 portable electronic balance to the nearest 0.1g. Time of day and amount of subcutaneous fat (related to time of year) have both been suggested as a source of variability in weight of nectar feeding birds (Collins and Paton 1989), but were not included in the analyses as it is unlikely there was a bias in time (of day or year) of capture for any particular species or sex.

After an investigation of data collected early during the project, a field method of determining M. lunatus sex, based on THL, was used to determine the sex of most individual M. lunatus. Those individuals with THL greater than 31 mm were sexed as male, less than 30.5 mm were sexed as female and those with THL in-between were classed as indeterminate. No field based method of sexing M. brevirostris or M. gularis was determined.

4.2.2 Analysis The following attributes were available for inclusion in analysis: lengths of total head, wing, tail, tarsus, hallux, claw, beak gape, beak top and total (measured dorsally) and weight. Not all individuals had all attributes measured, so each analysis was a balance of sample size (individuals) and available attributes. For example, three of the M. gularis were not measured as part of this project and had no tail, hallux or claw measurements, so were left out of the overall analysis, but included in the analysis of sex in M. gularis. Once the attributes to be used in an analysis were determined, all individuals with each of those attributes measured were included, with the exception of juveniles. A mean value of all captures was used where individuals had been caught on more than one occasion during the study, thus no individual was included more than once in the analysis.

Principle components analysis (PCA) reduces the dimensions of a multivariate dataset by producing a series of smaller conglomerate variables which are linear combinations of the original variables. These linear combinations of the original variables are the principal components. The coefficients in the linear combination making up each PC are known as eigenvectors. Generally the first principal component (PC1) explains the maximum variability of the original data, and successive principal components explain progressively less variability. Usually the original multidimensional dataset can be displayed in 2 or 3 dimensions with minimal loss of information (James and McCulloch

91 1990; Clarke and Warwick 1994). PCA was used for two main purposes. Firstly, to improve ability to sex individual honeyeaters. Secondly, to determine which attributes of closely morphologically related species were most different between the two species. PCA was run using PRIMER (Clarke and Warwick 1994) on normalised data. Once important attributes were identified, the absolute data were used to clarify what differences did occur with traditional statistical tests used on each attribute (one-way ANOVA comparing each attribute between species/sexes).

4.3 Results

Table 4.2 lists the number of individuals from each species which had all appropriate data available for inclusion in analyses. Sex in Table 4.2 was determined either from plumage (P. pyrrhoptera and Acanthorhynchus tenuirostris) or field measurement of total head length (M. lunatus).

Table 4.2: Number of individual honeyeaters measured in each species/sex group and number of sites at which that species was measured

Species Number of individuals Number of sites Melithreptus brevirostris 84 6 Melithreptus lunatus ♀ 32 5 Melithreptus lunatus ? 5 5 Melithreptus lunatus ♂ 32 5 Melithreptus gularis 4 1 Lichenostomus chrysops 12 4 Lichenostomus penicillatus 19 3 Phylidonyris novaehollandiae 209 5 Phylidonyris pyrrhoptera ♀ 21 4 Phylidonyris pyrrhoptera ♂ 26 4 Acanthorhynchus tenuirostris ♀ 14 3 Acanthorhynchus tenuirostris ♂ 15 3

4.3.1 Sexing The results using morphological features to sex individuals was mixed. For some species, as little as one attribute successfully sexed all individuals (M. gularis and L. penicillatus). One species required a large number of attributes to successfully sex all individuals (M. lunatus) and two species (M. brevirostris and L. chrysops) could not be sexed from the attributes measured. Phylidonyris novaehollandiae was able to be sexed once the site at which they were caught was taken into account.

92 4.3.1.1 Melithreptus brevirostris An ordination of the results for M. brevirostris (Figure 4.1) shows no groups of individuals based on the morphological features measured, suggesting that there is very little difference between the sexes (sexually monomorphic). The lack of differentiation is also evident in the cumulative variation explained by each of the principal components (Table 4.3), with the first three principal components only explaining about 60% of the original variation. The most highly correlated variable(s) with PC1 was THL, with PC2 were lengths of tail and tarsus and with PC3 were lengths of beak top and tarsus again (Table 4.4).

In an attempt to reduce any explicable variation, site was used as a factor to see if differences between the sexes could be detected on a site basis. No differences between the sexes could be determined from the site based analysis. An ordination of the results for the Scott Conservation Park area are shown in Figure 4.2.

Table 4.3: Cumulative percent variation explained by the first five principal components from analysis of 84 individual M. brevirostris

PC Eigenvalues %Variation Cumulative % Variation 1 3.37 33.7 33.7 2 1.43 14.3 48.0 3 1.18 11.8 59.8 4 1.05 10.5 70.3 5 0.85 8.5 78.8

Table 4.4: Eigenvectors from PCA of 84 individual M. brevirostris

Variable PC1 PC2 PC3 PC4 PC5 Weight -0.261 -0.314 0.234 -0.052 -0.742 THL -0.402 -0.298 -0.133 -0.160 0.080 Total -0.320 0.252 0.147 0.462 -0.040 Wing -0.342 0.360 0.282 -0.265 0.123 Tail -0.251 0.573 0.134 -0.233 -0.288 Tarsus -0.155 -0.506 0.505 -0.149 0.250 Hallux -0.276 -0.071 -0.021 0.757 -0.005 Claw -0.362 0.020 0.259 -0.031 0.420 Beak gape -0.371 -0.170 -0.465 -0.175 -0.197 Beak top -0.345 0.022 -0.519 -0.089 0.255

93 Figure 4.1: Ordination of PCA results on morphological attributes of M. brevirostris, demonstrating lack of differentiation between the sexes based on the morphological features measured

a) PC1 and PC2

5

4

3

2

1 2

PC 0

-1

-2

-3

-4 -4-3-2-101234 PC1

b) PC1 and PC3

3

2

1

3 0 PC

-1

-2

-3 -4-3-2-101234 PC1

c) PC2 and PC3

3

2

1

3 0 PC

-1

-2

-3 -4 -3 -2 -1 0 1 2 3 4 5 PC2

94 Figure 4.2: Ordination of PCA results on morphological attributes of M. brevirostris demonstrating lack of differentiation between the sexes at the level of site (Scott area) based on morphological features measured

2 1.5 1 0.5 0 -0.5 PC2 -1 -1.5 -2 -2.5 -3 -5 -4 -3 -2 -1 0 1 2 3 4 PC1

4.3.1.2 Melithreptus lunatus PC1 and PC2 explained nearly 70% of the original variation, while together with PC3 they explained just under 80% of the variation (Table 4.5). No attributes were highly correlated with PC1, but the most important attributes on that component were THL, total length, beak gape length and weight (Table 4.6). Tarsus and hallux were the most highly correlated with PC2, while beak top was by far the most highly correlated with PC3, but tail length was also important. The ordination of M. lunatus results (Figure 4.3) showed that most individuals were correctly sexed based on THL measurements, with two exceptions. One ‘male’ fell on the left of the cloud of female points in Figure 4.3a, but another falls high on PC1, almost on the right of the female cloud of points. Both these ‘males’ had THL of 31.5 mm, making them male on the field based criteria, but the combination of other attributes on PC1, saw them both fall in the range of female values along PC1. Displayed graphically in Figure 4.4, the problem with relying on only THL to sex M. lunatus is clear. Besides the incorrectly sexed ‘males’ it was possible to sex the individuals previously sexed as ‘indeterminate’ based on THL only measurements. Three of these individuals fell among the male values on PC1, while two fell among the female values.

95

The field based method sexed 3% of individuals incorrectly (2/69), 90% correctly (62/69) and was unable to determine the sex of 7% of individuals. The fact that so many attributes were important on PC1 suggests that any accurate method of sexing individuals based on morphological data does require a number of different attributes to be successful.

Table 4.5: Cumulative percent variation explained by the first five principal components from analysis of 69 individual M. lunatus

PC Eigenvalues %Variation Cumulative % Variation 1 5.90 59.0 59.0 2 1.00 10.0 69.0 3 0.89 8.9 77.9 4 0.52 5.2 83.2 5 0.47 4.7 87.9

Table 4.6: Eigenvectors from PCA of 69 individual M. lunatus

Variable PC1 PC2 PC3 PC4 PC5 Weight -0.342 0.019 -0.218 0.505 0.233 THL -0.359 0.082 0.142 -0.098 -0.378 Total -0.353 0.260 -0.182 0.222 -0.059 Wing -0.327 0.344 -0.189 0.009 0.451 Tail -0.290 0.171 -0.503 -0.503 -0.167 Tarsus -0.285 -0.534 -0.149 0.437 -0.221 Hallux -0.276 -0.590 0.144 -0.231 0.199 Claw -0.322 -0.230 0.124 -0.404 0.399 Beak gape -0.342 0.081 0.252 -0.098 -0.540 Beak top -0.245 0.292 0.700 0.123 0.171

96

Figure 4.3: Ordination of PCA results from data on morphological attributes of 69 individual M. lunatus

Sex on the ordination was determined using measurements of THL only. a) PC1 and PC2

2.5 2 1.5 1 0.5 0 Male Female PC2 -0.5 Indeterminate -1 -1.5 -2 -2.5 -3 -5-4-3-2-1012345 PC1

b) PC1 and PC3

4

3

2

1 Male Female PC3 0 Indeterminate

-1

-2

-3 -5 -4 -3 -2 -1 0 1 2 3 4 5 PC1

c) PC2 and PC3

3

2

1 Male 0 Female PC3 Indeterminate -1

-2

-3 -3 -2 -1 0 1 2 PC2

97 Figure 4.4: THL for each individual M. lunatus with sex determined by score on PC1 and lines indicating THL for field sexing

33 32.5 32 31.5 31 M. lunatus ♂ M. lunatus ♀ 30.5 Min. THL for ♂ 30 THL (mm) Max. THL for ♀ 29.5 29 28.5 28 0 10203040506070 Individual

4.3.1.3 Melithreptus gularis Only one M. gularis was handled in the field as part of this project, with the other three records made earlier by David Paton. There were no measurements of hallux, claw or total lengths for those three individuals. Measurements of each attribute for each individual are given in Table 4.7 and a graphical representation of the differences are given in Figure 4.5. The results showed one individual (034-77616) with larger measurements for each attribute and therefore this individual was sexed as male. No other individual was consistently larger than the others in their measurements, suggesting they were all female.

98 Table 4.7: Morphological data collected on four individual M. gularis

Sex is based on measurement of THL only. Weight THL Wing Tail Tarsus Beak Gape Beak Top Band Sex grams (mm) 034-77616 Male 22.5 36.0 90 73 23.8 18.4 15.7 034-77617 Female 22.5 33.9 85 69 23.1 16.4 14.3 035-89715 Female 20.0 34.8 85 67 23.5 16.3 14.4 035-90184 Female 20.3 34.1 84 68 22.6 16.9 14.9

Figure 4.5: Graphical representation of differences between each individual M. gularis for morphological data

1 ent

urem 0.95 as e 034-77616

m 034-77617

mu 035-89715 xi 035-90184 ma

f 0.9 o o ti Ra

0.85 THL Weight Wing Tail Tarsus Beak Gape Beak Top Attribute

4.3.1.4 Phylidonyris novaehollandiae PC1 and PC2 explained just over 60% of the original variation, while together with PC3 they explained about 70% of the variation (Table 4.8). As with the other species, there were a number of attributes negatively correlated with PC1 including weight and lengths of THL, wing, tail, hallux, beak gape and total (Table 4.9). Tail length was by far the most highly (negatively) correlated attribute on PC2 and claw length was the most highly (positively) correlated on PC3. When the results for all 210 individuals were plotted on the same ordination, there was no clear separation of the points for each individual into two ‘clouds’, suggesting little difference between the sexes (Figure 4.6). However, the site information included in Figure 4.6 suggests that there may be sex differences evident on a site basis. Therefore, the results were run through the same

99 analysis on a site basis, with an example of the resulting ordination for Scott Conservation Park data provided in Figure 4.7. Similar results were obtained for other sites allowing all individuals except two to be sexed with confidence and included in further analyses.

Following the site based analysis, it was clear that nearly two-thirds of the individual P. novaehollandiae caught were males. Re-examining Figure 4.6 in light of that information, the differences between the sexes are more evident, with the separation occurring at about 1 on PC1. The two individuals which could not be confidently sexed on the site based analyses appear in Figure 4.6 just above 1 on PC1 and both were from Scott Creek Conservation Park.

Table 4.8: Cumulative percent variation explained by the first five principal components from analysis of 210 individual P. novaehollandiae

PC Eigenvalues %Variation Cumulative % Variation 1 6.13 61.3 61.3 2 0.90 9.0 70.3 3 0.82 8.2 78.5 4 0.48 4.8 83.3 5 0.47 4.7 88.1

Table 4.9: Eigenvectors from PCA of 210 individual P. novaehollandiae

Variable PC1 PC2 PC3 PC4 PC5 Weight -0.319 -0.020 0.269 -0.636 -0.236 THL -0.370 0.155 -0.202 -0.166 -0.142 Total -0.362 -0.314 -0.083 0.145 0.025 Wing -0.304 -0.430 0.034 -0.329 0.253 Tail -0.302 -0.575 -0.140 0.359 0.104 Tarsus -0.285 0.385 0.246 -0.036 0.709 Hallux -0.300 0.209 0.266 0.299 0.167 Claw -0.269 0.063 0.607 0.361 -0.492 Beak gape -0.337 0.230 -0.336 -0.147 -0.265 Beak top -0.299 0.335 -0.496 0.258 -0.067

100 Figure 4.6: Ordination of PCA results on morphological attributes of 210 individual P. novaehollandiae showing PC1 and PC2

4

3

2

1 Bullock Charleston 2 0 Newland PC Scott -1 Scott Creek -2

-3

-4 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 PC1

101 Figure 4.7: Ordination of PCA results from data on morphological attributes of 128 individual P. novaehollandiae caught in the vicinity of Scott Conservation Park

Sex marked on the ordination was determined from PCA. a) PC1 and PC2

3

2

1

0

2 Male

PC Female -1

-2

-3

-4 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 PC1

b) PC1 and PC3

3 2.5 2 1.5 1

3 Male 0.5

PC Female 0 -0.5 -1 -1.5 -2 -5-4-3-2-10123456 PC1

c) PC2 and PC3

3 2.5 2 1.5 1

3 Male 0.5

PC Female 0 -0.5 -1 -1.5 -2 -4 -3 -2 -1 0 1 2 3 PC2 c

102 4.3.1.5 Lichenostomus penicillatus PC1 and PC2 explained about 70% of the original variation, while together with PC3 they explained just over 80% of the variation (Table 4.10). Weight and lengths of wing, tail, tarsus, hallux, beak top and total were all negatively correlated with PC1 (Table 4.16). THL was by far the most highly (positively) correlated attribute on PC2 and claw length was the most high ly (positively) correlated on PC3. The ordination of L. penicillatus results (Figure 4.8) show that PC1 is the important axis in sexing individuals, but unlike the results for M. lunatus, there are a number of single attributes that successf ully sex each of the individuals. Wing, tail and hallux lengths by themselves correctly sex each of the nineteen individuals.

The ordination also shows three individuals particularly low on PC2, but falling within the range of values for males on PC1. THL was by far the most important aspect of PC2, thus these three individuals have low measures of THL. The reasons for this are not obvious, w ith one of the individuals coming from Charleston Conservation Park and two from Sco tt Conservation Park (i.e. it is not due to geographic variation) and only one o f the ind ividuals wa s aged as immature with the other two being adults (i.e. it is not due to age differences: note, juveniles were removed from the analysis).

Table 4.10: Cumulative percent variation explained by the first five principal components from analysis of 19 individual L. penicillatus

PC Eigenvalues %Variation Cumulative % Variation 1 5.8 57.9 57.9 2 1.3 12.5 70.4 3 1.0 10.0 80.4 4 0.7 7.4 87.8 5 0.5 4.7 92.5

Table 4.11: Eigenvectors from PCA of 19 individual L. penicillatus

Variable PC1 P C2 PC3 PC4 PC5 Weight -0.330 -0.297 -0.29 2 0.002 0.199 THL -0.069 0.788 0.05 6 0.368 0.313 Total -0.350 0.262 -0.07 8 -0.440 0.032 Wing -0.364 0.153 -0.31 4 0.066 -0.383 Tail -0.356 0 .107 -0.21 0 -0.490 0.139 Tarsus -0.333 -0.149 0.170 0.462 -0.366 Hallux -0.360 -0.322 0.201 0.071 0.113 Claw -0.286 -0.100 0.607 -0.024 0.490 Beak gape -0.281 -0.062 -0.420 0.450 0.296 Beak top -0.320 0.218 0.387 -0.066 -0.470

103 Figure 4.8: Ordination of PCA results from data on morphological attributes of 19 individual L. penicillatus

Sex marked on the ordination was determined by PCA. a) PC1 and PC2

2 1.5 1 0.5 0

2 Male -0.5

PC Female -1 -1.5 -2 -2.5 -3 -3-2-101234 PC1

b) PC1 and PC3

3

2

1

3 Male 0

PC Female

-1

-2

-3 -3 -2 -1 0 1 2 3 4 PC1

c) PC2 and PC3

3

2

1

3 Male 0

PC Female

-1

-2

-3 -3 -2 -1 0 1 2 PC2 c

104 4.3.1.6 Sex Ratio The sex of all individuals of each species (most other individuals with less than the full complement of attributes used in PCA were able to be sexed on the basis of available attributes) is given in Table 4.12. Melithreptus lunatus is the closest to a 1:1 sex ratio, while P. novaehollandiae is the furthest, with nearly two-thirds of the population being male. Looking at differences on a site basis (Table 4.13), M. lunatus is consistently close at all sites, while P. novaehollandiae has considerable site based differences. In the Scott Conservation Park area, over twice as many males as females were caught, while in the Scott Creek Conservation Park area, the reverse was true. Another interesting result is that Acanthorhynchus tenuirostris has a nearly 1:1 sex ratio overall, but on a site basis there are about twice as many males as females at two sites, while at the third site, no males were caught at all. This reversal of the male biased sex ratio trend in the Scott Creek Conservation Park area was the same for P. novaehollandiae. However, in most cases the sample size was small.

Table 4.12: Numbers of individuals of each sex for those honeyeater species which could be sexed

Presented in increasing order of percent of overall capture that was sexed as male. Species Female Male % male Melithreptus gularis 3 1 25.0 Melithreptus lunatus 48 45 48.4 Acanthorhynchus tenuirostris 19 20 51.3 Phylidonyris pyrrhoptera 23 27 54.0 Lichenostomus penicillatus 9 12 57.1 Phylidonyris novaehollandiae 81 132 62.0

105 Table 4.13: Numbers of individuals of each sex at each site for those honeyeater species which could be sexed, and mean weights of sexes at each site

Captures Weight (grams) Female Male % male Female Male Melithreptus lunatus Charleston 1 1 50.0 13.6 15.8 Kaiser 7 9 56.3 13.0 14.3 Newland 10 11 52.4 13.1 14.3 Scott 22 17 43.6 12.6 14.1 Scott Creek 8 7 46.7 13.3 15.9 Melithreptus gularis Scott 3 1 25.0 20.9 22.5 Lichenostomus penicillatus Charleston 2 3 60.0 17.0 22.0 Monarto 2 0 17.9 Scott 5 9 64.3 18.3 20.8 Phylidonyris novaehollandiae Bullock 4 3 42.9 20.6 22.4 Charleston 4 8 66.7 18.8 23.2 Newland 15 19 55.9 19.4 23.1 Scott 40 94 70.1 18.7 21.7 Scott Creek 18 8 30.8 20.5 23.0 Phylidonyris pyrrhoptera Charleston 1 3 75.0 13.4 15.3 Newland 4 7 63.6 13.3 17.4 Scott 10 9 47.4 12.2 15.2 Scott Cre ek 8 8 50.0 13.4 16.7 Acanthorhynchus tenuirostris Newland 3 6 66.7 9.5 11.8 Scott 8 14 63.6 9.1 11.1 Scott Creek 5 0 10.6

4.3.2 Morphology After being sexed via PCA (previous section), a mean for all individuals in each species/sex with available data was included in a PCA to determine which morphological features were most important in distinguishing species. Mean values of attributes and num bers of individuals and sites for each species/sex are given in Table 4.14. Figure 4 .9 shows the o rdination of this PCA. The variation explained by the first two principal components is nearly 95% and the first three is nearly 98% (Table 4.15). There are numerous attributes negatively correlated with PC1, including weight and wing, tail, tarsus, hallux, claw and total lengths (Table 4.16). On PC2 each of the measures of beak length (including THL) are the most important features, and each is negatively correlated. On PC3, claw and hallux are negatively correlated, while wing, tail and total lengths are positively correlated.

106

The combination of PC1 and PC2 separated the long-beaked guild very well (Figure 4.9a). However, PC1 and PC2 were not successful in separating the short-beaked guild.

The M. gularis female fell between L. penicillatus males and females on PC1 and M. brevirostris fell between male M. lunatus and females on PC1. In both cases the ‘species in the middle’ fell only just below the other on PC2. The M. gularis was most different on PC3 where it fell slightly below female L. penicillatus (Figure 4.9 b & c). Looking at the absolute values, the M. gularis had a longer claw & longer beak top and shorter total tail & body length than either sex of L. penicillatus (Table 4.14). Sample size was not adequate to test differences between L. penicillatus and M. gularis statistically. Of all the measures, beak top is perhaps the most different, appearing longer for M. gularis, but with a sample size of 1 for M. gularis this result should be treated with caution.

Melithreptus brevirostris plotted very close to the male M. lunatus on the ordination of the PCA. Melithreptus brevirostris fell between the M. lunatus sexes and very close to male M. lunatus on PC1, and only slightly below the M. lunatus sexes and again very close to male M. lunatus on PC2. The component on which M. brevirostris and M. lunatus are most different is PC3, with M. brevirostris falling lower on that component than any of the other species, suggesting that lengths of claw, hallux, wing and tail are important in distinguishing M. brevirostris from other species, especially M. lunatus. The statistical tests (Table 4.17) agree quite well with the attributes suggested by the PCA, with M. brevirostris having statistically different wing, hallux, claw and total lengths to both M. lunatus sexes (i.e. different to the PCA in not suggesting a difference in tail). However, as in the PCA, the differences are not in a consistent direction (Table 4.17). It is also dubious as to the biological significance of some of the statistical differences (e.g. a 0.3 mm difference in THL (<1%) between M. brevirostris and male M. lunatus). To make some sense of the differences that are likely to be important, the percent differences between each species/sex pair are given in Table 4.18. These are presented in descending order of percent difference between M. brevirostris and male M. lunatus, the most similar pair according to the PCA. From this it is clear that wing length is the attribute most different between this closely morphologically related pair. The other major difference is that of claw length, with M. brevirostris having a longer

107 claw than both M. lunatus sexes, although it is much closer in length to male M. lunatus than females.

108 Table 4.14: Mean attribute values (± s.e.) and sample size for 13 species/sexes of small honeyeaters in the Mt Lofty Ranges

Ind.: number of individuals. Sites: number of sites. THL: Total head length. All measurements are in mm except for weight (grams) Spp n Weight THL Total length Wing Tail Tarsus Hallux Claw Beak Gape Beak Top Ind. Sites grams mm M. brevirostris 84 6 14.7 ± 0.2 31.5 ± 0.4 142 ± 2 68 ± 1 56 ± 1 20.3 ± 0.2 9.5 ± 0.1 6.1 ± 0.1 15.2 ± 0.2 12.8 ± 0.1 M. lunatus ♀ 34 5 13.0 ± 0.4 29.8 ± 0.9 140 ± 4 70 ± 2 56 ± 2 19.8 ± 0.6 8.6 ± 0.2 5.2 ± 0.1 14.4 ± 0.4 11.7 ± 0.3 M. lunatus ♂ 35 5 14.6 ± 0.4 31.8 ± 0.9 148 ± 4 76 ± 2 59 ± 2 20.5 ± 0.6 9.2 ± 0.3 5.8 ± 0.2 15.6 ± 0.4 12.8 ± 0.4 M. gularis ♀ 1 1 20.3 34.1 167 84 68 22.6 10.3 6.7 16.9 14.9 L. chrysops 12 4 15.9 ± 1.4 33.1 ± 3.0 167 ± 15 77 ± 7 70 ± 6 21.2 ± 1.9 8.7 ± 0.8 5.9 ± 0.5 17.1 ± 1.6 12.1 ± 1.1 L. penicillatus ♀ 8 3 17.9 ± 2.2 33.8 ± 4.2 171 ± 21 77 ± 10 69 ± 9 23.0 ± 2.9 9.8 ± 1.2 6.4 ± 0.8 15.1 ± 1.9 12.0 ± 1.5 L. penicillatus ♂ 11 2 21.1 ± 1.9 34.5 ± 3.1 183 ± 17 87 ± 8 80 ± 7 24.4 ± 2.2 10.7 ± 1.0 6.9 ± 0.6 18.0 ± 1.6 14.0 ± 1.3 P. novaehollandiae ♀ 81 5 19.3 ± 0.2 39.5 ± 0.5 178 ± 2 74 ± 1 74 ± 1 24.1 ± 0.3 10.5 ± 0.1 6.8 ± 0.1 21.9 ± 0.3 18.8 ± 0.2 P. novaehollandiae ♂ 126 5 22.1 ± 0.2 41.8 ± 0.3 190 ± 1 79 ± 1 81 ± 1 25.3 ± 0.2 11.2 ± 0.1 7.3 ± 0.1 23.4 ± 0.2 20.2 ± 0.2 P. pyrrhoptera ♀ 21 4 12.9 ± 0.6 35.2 ± 1.7 149 ± 7 64 ± 3 57 ± 3 21.0 ± 1.0 9.2 ± 0.4 6.0 ± 0.3 19.4 ± 0.9 16.0 ± 0.8 P. pyrrhoptera ♂ 26 4 16.2 ± 0.6 38.5 ± 1.5 167 ± 6 72 ± 3 66 ± 3 22.2 ± 0.9 10.0 ± 0.4 6.6 ± 0.3 21.2 ± 0.8 17.5 ± 0.7 Acanthorhynchus tenuirostris ♀ 14 3 9.7 ± 0.7 38.4 ± 3.0 144 ± 11 60 ± 5 53 ± 4 19.3 ± 1.5 8.1 ± 0.6 5.4 ± 0.4 25.3 ± 1.9 20.9 ± 1.6 Acanthorhynchus tenuirostris ♂ 15 2 11.3 ± 0.8 42.7 ± 2.8 159 ± 11 66 ± 4 60 ± 4 20.5 ± 1.4 8.7 ± 0.6 5.8 ± 0.4 28.0 ± 1.9 24.0 ± 1.6

109 Table 4.15: Cumulative percent variation explained by the first five principal components from analysis of 468 individual honeyeaters from eight species in the Mt Lofty Ranges

PC Eigenvalues %Variation Cumulative % Variation 1 6.31 63.1 63.1 2 3.17 31.7 94.8 3 0.31 3.1 97.9 4 0.13 1.3 99.3 5 0.03 0.3 99.6

Table 4.16: Eigenvectors from PCA of data on 468 individual honeyeaters from eight species in the Mt Lofty Ranges

Variable PC1 PC2 PC3 PC4 PC5 Weight -0.382 0.138 -0.08 0.149 0.223 THL -0.127 -0.53 0.02 -0.056 -0.27 Total -0.379 -0.1 0.371 -0.326 -0.289 Wing -0.321 0.263 0.438 0.71 -0.091 Tail -0.385 0.007 0.381 -0.282 0.178 Tarsus -0.392 -0.033 -0.089 -0.327 0.45 Hallux -0.382 -0.038 -0.421 0.045 -0.643 Claw -0.377 0.03 -0.546 0.148 0.219 Beak gape -0.014 -0.556 0.174 0.14 0.062 Beak top -0.02 -0.555 -0.064 0.366 0.293

110 Figure 4.9: Ordination of PCA results from data on 468 individual honeyeaters from eight species (13 species/sexes) in the Mt Lofty Ranges

a) PC1 and PC2

2 M. brevirostris M. lunatus ♀ 1 M. lunatus ♂ M. gularis ♀ 0 L. penicillatus ♀ L. penicillatus ♂ 2 -1 L. chrysops PC P. pyrrhoptera ♀ -2 P. pyrrhoptera ♂ P. novaehollandiae ♀ P. novaehollandiae ♂ -3 A. tenuirostris ♀ A. tenuirostris ♂ -4 -5-4-3-2-101234 PC1

b) PC1 and PC3

1.5 M. brevirostris M. lunatus ♀ 1 M. lunatus ♂ M. gularis ♀ L. penicillatus ♀ 0.5 L. penicillatus ♂ 3 L. chrysops PC 0 P. pyrrhoptera ♀ P. pyrrhoptera ♂ P. novaehollandiae ♀ -0.5 P. novaehollandiae ♂ A. tenuirostris ♀ A. tenuirostris ♂ -1 -5 -4 -3 -2 -1 0 1 2 3 4 PC1

c) PC2 and PC3

1.5 M. brevirostris M. lunatus ♀ 1 M. lunatus ♂ M. gularis ♀ L. penicillatus ♀ 0.5 L. penicillatus ♂ 3 L. chrysops PC 0 P. pyrrhoptera ♀ P. pyrrhoptera ♂ P. novaehollandiae ♀ -0.5 P. novaehollandiae ♂ A. tenuirostris ♀ A. tenuirostris ♂ -1 -4 -3 -2 -1 0 1 2 PC2

111 Table 4.17: Statistical differences between morphological attributes of M. brevirostris, female M. lunatus and male M. lunatus

BHH = M. brevirostris, WNF = M. lunatus ♀, WNM = M. lunatus ♂. Tukey tests are presented as largest to smallest mean from left to right. Melithreptus brevirostris falls between M. lunatus ♂ and M. lunatus ♀ for 6 of the 10 attributes, but only has the smallest value for one attribute, wing length. Mean n ANOVA Attribute Unit BHH WNF WNM BHH WNF WNM df F P Tukey test Weight grams 14.7 12.9 14.6 109 46 42 2,194 64 <0.001 BHH=WNM>WNF THL mm 31.5 29.8 31.8 109 47 45 2,198 142.49 <0.001 WNM>BHH>WNF Total length mm 141 139 147 102 47 43 2,189 24.66 <0.001 WNM>BHH>WNF Wing mm 68 70 75 107 47 44 2,195 100.6 <0.001 WNM>WNF>BHH Tail mm 56 56 59 106 47 43 2,193 29.61 <0.001 WNM>BHH=WNF Tarsus mm 20.3 19.8 20.5 107 47 44 2,195 17.98 <0.001 WNM=BHH>WNF Hallux mm 9.2 8.6 9.1 107 47 44 2,195 49 <0.001 BHH>WNM>WNF Claw mm 6.1 5.3 5.8 107 46 44 2,194 88.81 <0.001 BHH>WNM>WNF Beak Gape mm 15.2 14.4 15.6 100 41 38 2,176 47.3 <0.001 WNM>BHH>WNF Beak Top mm 12.7 11.6 12.8 99 41 38 2,175 27.84 <0.001 WNM=BHH>WNF

112 Table 4.18: Percent difference in mean value of morphological attributes in the small Melithreptus guild

BHH = M. brevirostris, WNF = female M. lunatus, WNM = male M. lunatus. Presented in descending order of magnitude of % difference in BHH & WNM. % difference between mean Attribute BHH & WNM BHH & WNF WNM & WNF Wing 9.33 2.86 -7.14 Claw -5.17 -15.09 -9.43 Tail 5.08 0.00 -5.36 Total 4.08 -1.44 -5.76 Beak Gape 2.56 -5.56 -8.33 Hallux -1.10 -6.98 -5.81 Tarsus 0.98 -2.53 -3.54 THL 0.94 -5.70 -6.71 Beak Top 0.78 -9.48 -10.34 Weight -0.68 -13.95 -13.18

4.4 Discussion

4.4.1 Sex Individuals that fell in the cloud of points (within each species) that weighed more (or had longer wings etc.) were sexed as male. Was the use of morphological features to determine sex a circular argument? A number of studies, working with individuals of known sex, have shown that honeyeater males do weigh more (Collins and Paton 1989; Paton and Collins 1989; Schodde and Mason 1999; Higgins et al. 2001), including M. lunatus and M. gularis males (Schodde and Mason 1999; Higgins et al. 2001), so the sexing results presented here are almost certainly legitimate.

All the honeyeater species except M. brevirostris and L. chrysops demonstrated a clear separation of the sexes based on the attributes measured (sexually dimorphic). In each of the species showing difference between the sexes, males were larger in each of the measured attributes than females, in accord with the summaries of Collins and Paton (1989) and Paton and Collins (1989). However, the inability to sex M. brevirostris based on any of the attributes or combinations of the attributes suggests M. brevirostris is particularly similar in terms of sex related morphology. It may still be the case that male M. brevirostris are larger than female M. brevirostris but the mean differences must be very small and the overlap is great.

113 Higgins et al. (2001) found very few statistical differences in morphological measurements between males and females of various M. brevirostris subspecies based on measurements of skins in museums, where the birds had been sexed. They found no statistical differences between males and females of the subspecies pallidiceps, which occurs in the Murray-Darling Basin (and the Mount Lofty Ranges: Schodde and Mason 1999), in lengths of wing, tail, bill top and tarsus. They also found that within each of the subspecies of M. brevirostris there was no difference in weight of the sexes, based on labels of museum specimens. Schodde and Mason (1999) found similar results.

The difference between M. brevirostris and M. lunatus with regards sexual dimorphism was likely due to two related factors. Firstly, M. brevirostris appeared to be non- territorial, while M. lunatus, particularly males, defended breeding territories (Chapter 6). Secondly, M. brevirostris appeared to share the incubating of eggs and nestlings between a number of individuals, while in M. lunatus the responsibility fell mainly to the female (pers obs., Noske 1983; Higgins et al. 2001). If females are responsible for most of the incubating in M. lunatus, it is likely males are responsible for most territorial defence. Individuals of the sex to which the responsibility falls for territorial defence are likely to face selection pressure for larger size, while incubating is likely to select for smaller individuals with less overall energy requirements (Paton 1979).

4.4.2 Sex ratio A skewed sex ratio in honeyeater populations has been shown elsewhere. Dow (1977) found some M. melanocephala colonies in which 78% of the population was male. Dow did not suggest a cause for this skewed sex ratio, but he did note that intraspecific aggression was maintained at a high level throughout the year. Paton (1979) suggested that a difference in sex ratio in favour of males developed from differences in the survival of males and females based on several lines of evidence. Most importantly, there was no skewed sex ratio in juveniles. Dominance and territorial results indicated that males were dominant to females (they were larger) and held feeding territories on the richest food sources from which adult females, juveniles and immatures were excluded when resources began to decline, forcing them to feed on poorer resources. This combination of factors strongly suggest that P. novaehollandiae females have lower survival rates than P. novaehollandiae males. Implicit in this argument is the idea that the sites were saturated with P. novaehollandiae, with males excluding females

114 from the best resources, at least during times of food resource shortage, thus exhibiting intersexual competition with an effect on survival. This is likely to lead to the observed sex related differences in morphology, with males being larger for more effective dominance and females being smaller for more effective exploitation of poorer resources (Paton 1979). Similar patterns of male skewed sex ratio have been reported in other studies. Phylidonyris pyrrhoptera in the Mt Lofty Ranges were found to have a population consisting of 65% males (Foster 2001). In the resident P. novaehollandiae population of a heathland near Sydney 74% of the population were males (Pyke et al. 1989).

While higher female mortality is undoubtedly a factor in the recorded sex ratio of honeyeater populations, other factors may contribute: males may be easier to catch; and/or females may be more cryptic in their behaviour. Both explanations may also contribute to a portion of the skewed sex ratio, although no studies have examined the potential for each to contribute. Phylidonyris novaehollandiae are known to be territorial, defending feeding and breeding territories (Paton 1979; 1980; 1985; McFarland 1986b; Pyke and Recher 1986; Pyke and O'Connor 1993; McFarland 1996; 2002) and it is likely that male P. novaehollandiae defend territories more often than females (Paton 1979; 1993). This has the potential to introduce a bias towards male P. novaehollandiae in mist-net catch rate. Phylidonyris pyrrhoptera females at Cromer Conservation Park in the Mt Lofty Ranges were found to feed within a lower stratum to males, almost certainly as a result of male aggression (Foster 2001). This may force them to feed secretively as has been observed for some other nectarivores (Lyon 1976; McFarland 1996), probably decreasing their records during a census but having an unknown (probably small to nil) effect on their trap rate in mist-nets. When nesting they were especially subtle, ‘females usually arrived by furtively and quietly moving to the nest via a backdoor route’ (Foster 2001).

No studies on honeyeaters have reported sex ratios skewed in favour of females. The evidence for a female skewed sex ratio at some sites is quite good, particularly for P. novaehollandiae at Scott Creek. Is this a case where females are better exploitative competitors than their male conspecifics? Scott Creek occurs within a large (the largest) expanse of intact native vegetation in the Mt Lofty Ranges, although the vast majority of the area is E. obliqua, the understorey in most of these areas contains a mix of shrubs

115 (Banksia marginata, Correa spp., Epacris impressa, Astroloma conostephioides) that provide honeyeaters with small rewards from flowers most suited to long-beaked species (Paton 1986a). It is possible that the females in these areas are able to out- compete the males through exploitation of these sources of nectar. The female skewed sex ratio warrants further investigation. Given the number of times it appears a male skewed sex ratio in honeyeaters has resulted from intersexual interference competition (Dow 1977; Paton 1979; 1985; Pyke et al. 1989; Foster 2001), a logical argument would be that intersexual exploitative competition is a cause of female skewed sex ratio. However, there are further alternative explanations. There may be higher predation rates on a less secretive (see below) male population, although this is unlikely given the more commonly recorded male biased sex ratio in honeyeaters (Paton 1979; 1985; Pyke et al. 1989; Foster 2001). An Area such as Scott Creek may also act as a refuge for female P. novaehollandiae during the non-breeding season, with females moving into the area to exploit small reward floral resources that they are better at exploiting than their male conspecifics.

One important aspect of population sex ratio not tackled here is seasonal patterns. At least one other study found consistent seasonal trends in sex ratio, with the ratio being closest to one soon after breeding periods but becoming increasingly male biased through time since the last breeding season (Paton 1979). This pattern was established through intensive trapping regimes at two sites in Victoria (Paton 1979). Here, each site had markedly different trapping regimes, with the Scott Creek area being trapped during only one period in the Melithreptus non-breeding season, while the Scott area was trapped over two years in most months of the year. It is therefore unlikely that the recorded sex ratios are representative of each of the areas, but the results do demonstrate a spatial variability in sex ratio that is consistent with competition being an important aspect of honeyeater community organisation.

In summary, the sex ratio of a number of honeyeater populations in the Mt Lofty Ranges appears to be skewed in favour of males, a trend most apparent for P. novaehollandiae. Competition is likely to explain at least a portion of the skewed sex ratio. If sites within the Mt Lofty Ranges were saturated with P. novaehollandiae, it has implications for other species of honeyeater in that they are much more likely to find themselves in competition with P. novaehollandiae. There was no evidence of an

116 overall skewed sex ratio in any of the Melithreptus (assuming too small a sample size in M. gularis to be representative).

4.4.3 Morphology As could be expected from the major niche axes determining the structure of nectarivore communities (Section 1.1), size related attributes (PC1) and beak length (PC2) explained most of the variation between honeyeater species/sexes and were able to separate most species, especially for the long-beaked species. However, these axes were unable to differentiate the Melithreptus species from themselves in the case of the smaller species, and L. penicillatus in the case of the larger M. gularis. In both these cases, a further dimension was necessary to separate the species, and even then the separation was not particularly good, especially when compared to the long-beaked species. One possible explanation for this is that interference competition is more important in resource partitioning within the long-beaked guild (size separates species) than it is within the short-beaked guild (size does not separate species).

Looking at the absolute values between pairs of species that were plotted closely on the PCA ordination suggests a few attributes in which there were some differences. Melithreptus gularis had a longer beak and claw, but shorter tail and total length than L. penicillatus. A longer bill, especially in association with the ectethmoid-mandibular articulation, may allow M. gularis to access resources deeper within bark substrates. The longer claw may enable M. gularis to perch more effectively on vertical surfaces, further enhancing its ability to probe under bark. This is in contrast to L. penicillatus, which if the differences are legitimate (i.e. discounting sample size for M. gularis), appears slightly better suited to gleaning foliage than M. gularis, with shorter claws for clasping petioles and a shorter beak, which has been associated with gleaning foliage (Slater 1994).

The primary aim of this work was to suggest further differences in niche separation between M. brevirostris and M. lunatus. Given this aim, I have ignored the range of values for the various attributes within a species/sex (other than the statistical tests, such as carried out in Table 4.17 which do give some indication of the range of overlap between the averages being compared) and concentrated on the mean values which should more appropriately reflect any overall differences in ecology and function

117 between species/sexes. The overall similarity in morphology documented here between M. brevirostris and M. lunatus extends the similarities found by other studies. However, the overall similarity of M. brevirostris and male M. lunatus, compared with female M. lunatus, has not been previously documented. Melithreptus brevirostris and male and female M. lunatus will hereon be refe rred to as the small Melithreptus ‘guild’, the term guild being used in the same sense as Burton and Olsen (2000) in describing two species of sympatric, sexually d imorphic goshawk.

Keast (1968a) compared wing, bill and tarsus lengths of male M. lunatus and male M. brevirostris within th e Mt Lofty Ranges. The results obtained in this study are similar to those of Keast, with some minor exceptions. Keast found no overlap in wing length between th e species i n his sam ple, whereas only the 18 longest wing lengths in my sample of male M. lunatus were larger than all the M. brevirostris wing lengths. The percent difference in overall wing length between the two species from this study was slightly lower than found by Keast (Keast: 14%, this study 9.3%) despite the wing length from this study being ‘dragged down’ by any females included in the M. brevirostris sample. The percent difference in bill length is nearly identical between the two studies (Keast: 3%, this study 2.7%), while the percent difference in tarsus length is also very close despite a changed sign of the difference (Keast: -2.5%, this study 1.0%). Again, this is likely due to the inclusion of females in the M. brevirostris sample. Thus, the results of this study confirm the resul ts of Keast (1968a) tha t wing length is the most different morphological attribute between M. brevirostris and M. lunatus. However, this is not as straightforward a result as suggested by Keast (1968a). Wing length has been singled out in this study a fter examining a large number of attributes and including both M. lunatus sexes in the analysis. In most cases, the morphological attributes of M. brevirostris fall between those of male M. lunatus and female M. lunatus and usually closer to male M. lunatus. Wing length has been singled out in this study as it is the one attribute of M. brevirostris that is diff erent to both male and female M. lunatus and not close to male M. lunatus. The functional significance of a difference in wing length is likely to be related to moveme nt. The relative differences in wing length also suggest that if differences in movements are found, M. brevirostris should be more similar to female M. lunatus than male M. lunatus. Perhaps this prediction could be extended to suggest that in aspects of behaviour and ecology other than movement, male M. lunatus should be more similar to M. brevirostris than female M. lunatus.

118

A difference between M. brevirostris and M. lunatus not highlighted before is that of claw length. As discussed above, a shorter claw may enable more efficient clasping of petioles, while longer claws may enable better perching on vertical surfaces. This lends support to the suggestion that M. brevirostris feeds more from bark, while M. lunatus feeds more from foliage (Section 1.5.4).

In summary, this study measured the most comprehensive number of morphological attributes from sympatric populations of M. brevirostris and M. lunatus. There were two main differences found between the species: wing length; and claw length. However, more interesting was the finding that M. brevirostris and male and female M. lunatus form a continuum in a number of morphological attributes. The greatest differences between the small Melithreptus guild were found in wing length and leg morphology. The direction of the differences suggested that M. brevirostris would be more similar to female M. lunatus in movements and more similar to male M. lunatus in foraging behaviour. These hypothesised differences were investigated further as part of the wider studies into home range and movement (Chapter 5) and behaviour (Chapter 6).

119