Decision letter for "Bird community recovery following removal of an invasive tree"

HandlingEditor:MaxLambert Abstract 1. Faunal responses to plant invasions and their managed removal can expand our understanding of the nature of disturbance and the success of restored plant communities. 2. We examined how bird communities responded to the presence and removal of the invasive understorey tree Pittosporum undulatum Vent. (sweet pittosporum) in matched woodland areas in temperate south-eastern Australia that were free of P. undulatum invasion, were invaded, or had been cleared up to 14 years prior to our sampling. 3. Overall bird species richness and individual abundance were insensitive indicators, as neither were significantly affected by the presence or removal of P. undulatum. However, richness and abundance were sharply lower in and beneath the P. undulatum canopies compared to the forest overstorey, pointing to a large structural modification by the invader. Bird community composition changed in fairly consistent ways at multiple sites upon invasion by P. undulatum, changes that were partly but not completely reversed by removal of P. undulatum. The suite of functional traits of the birds present at the sites was disrupted in idiosyncratic ways at sites invaded by P. undulatum and only very weakly restored upon clearing of P. undulatum. Functional and diversity indices are dependent on the type of management implemented. 4. We propose that a more nuanced approach to management such that some of the invaded forest in neighbouring areas is retained while new trees become established in the cleared areas, providing access to suitable habitat for birds during the transition phase. Suchmeasures are challenging in terms of management and funding but are necessary to maintain avian diversity during and after restoration processes.

Here we investigate the effects of the invasive weed tree, Pittosporum undulatum (sweet pittosporum), and its removal on the richness, abundance and functional traits of resident bird communities.
There has been some consideration of the interaction of P. undulatum with particular bird species (Gleadow, 1982;McNabb & McNabb, 2011), but not with bird communities more broadly. P. undulatum is a shade-tolerant tree native to coastal forests of south-eastern Australia that has become invasive throughout temperate Australia, including some regions of its natural territory where it is considered to perform the function of an invader (Gleadow & Ashton, 1981;Gleadow et al., 1983). The presence of a dense P. undulatum canopy in forest understorey severely alters the structure and floristic composition of invaded habitats, leading to understorey areas with a high proportion of bare ground and a loss of plant species diversity (O'Leary et al., 2018). As a result, we expect invaded areas less suitable for a number of bird species due to low habitat heterogeneity (Stirnemann et al., 2014).
The ecological role of bird species can be characterized by their functional traits. Such traits reflect environmental tolerances and capacities for resource capture, reproduction, dispersal (McGill et al., 2006;Reich et al., 2003;Westoby et al., 2002) (Table 1). Species trait values can therefore be used to characterize the functional diversity of a community. Despite the traditional focus on species richness to assess ecological communities, greater attention should be given to functional diversity (Lindenmayer et al., 2015). Our hypotheses are (1) that given the density and uniformity of the P. undulatum canopy, invaded areas will have a reduction in the range of functional types of birds and (2) given that the removal of P. undulatum allows substantial recovery of native plant communities (O'Leary et al., 2018), bird communities would be restored to the types of communities seen in neighbouring remnant vegetation including richness, density and functional diversity.
Changes in bird functional diversity following restoration might provide a different perspective on recovery of community processes, one that emphasizes community resilience, long-term stability and ecosystem functioning (Fischer et al., 2007;Karp et al., 2011;Lindenmayer et al., 2015), which in turn will lead to the need for more nuanced restoration programmes. We propose that restoration programmes may need to be staggered in space and time to allow bird communities to move into new areas before the habitat provided by the coalesced P.

Site selection
Ten sites across peri-urban areas of Melbourne, in south-eastern Australia, were selected to evaluate their bird communities (  (Guido & Pillar, 2017).
As we are assuming a space-for-time substitution, we used Ecological Vegetation Class (EVC) mapping supported by on-ground obser-vations to ensure that the vegetation patches within each site supported similar vegetation (DELWP, 2017). Sites ranged in size from 1 to 12 ha. The management area at each site was characterized as having a severe P. undulatum infestation (30-70% canopy cover) prior to removal work (Table 1). With the exception of the Wonga Park site, which was cleared in 2016, all sites had experienced some degree of follow-up weed control within 12 months of the initial P. undulatum removal, and all sites cleared for three years or more had received maintenance at least twice. The reference controls at approximately half of the sites had been exposed to control burns within the past 15 years as a means of maintaining the natural disturbance regime and stimulating biodiversity (Penman et al., 2011). Unlike some other studies (e.g. , fire was not used in the initial con- Information on the density of P. undulatum at the cleared areas, weed management practices and disturbance regime was included in the analysis.

Bird surveys
Bird surveys were conducted on three separate mornings at each site from mid-May to late June of 2017. Surveys were conducted within the first 3 h after sunrise, following a modified version of the process established by Loyn (1986) and outlined in Loyn et al. (2007), that is 10 min of surveying time was implemented for each hectare of sampling area at each site, to a maximum of 20 min. This timeframe has been observed as appropriate to survey bird communities within south-eastern Australian forests, whilst reducing the risk of bias towards conspicuous species with distinctive and/or frequent calls. One of us (BO'L) was present for each survey at each site, supported by volunteer surveyors experienced with local bird communities. All birds observed by sight and call within and below the vegetation canopy were identified to species level. Birds flying overhead were not included in the study. Bird species relative abundance was determined by dividing the total number of birds observed over the three surveys at each site.
Surveys distinguished bird use of the habitat within or below the P.
undulatum canopy (PU) from use of the overstorey (NPU, not PU) due to the predicted large effects of dense P. undulatum canopies. Density was determined by dividing the numbers of birds by the area surveyed.

Functional traits
Values of five traits reflecting ecological functionality (life history, habitat preference and feeding guild) were extracted for each bird species identified in our surveys from the Handbook of Australian, New Zealand and Antarctic Birds (Higgins et al., 2006). The traits and brief descriptions of their ecological relevance are listed in Table 2.

Statistical analysis
Differences in bird species richness and individual abundance among reference invaded and cleared vegetation types and between the P.
undulatum understorey and overstorey in invaded sites were tested with one-way ANOVAs. To compare species richness in individual feeding guilds across the three vegetation types, we used Kruskal-Wallace tests, due to the smaller numbers of species involved in each guild.
The effect of time since P. undulatum removal was tested with linear regression on relativized measures of species richness and individual abundance (richness or abundance in cleared patch divided by richness or abundance in the corresponding reference control). To summarize the differences and similarities among bird communities in the three vegetation types, we employed principal components analysis (PCA) of bird species presence/absence data using the prcomp function in R (R Core Team, 2017). Loadings on the first two principal components were examined to account for the contribution of individual species to the community configuration. We repeated the PCA using bird abundance data using nonmetric multi-dimensional scaling and checked that the results were biologically meaningful as suggested by Björklund (2019).
Since all results were in qualitative agreement, we present PCA of the presence-absence data here. Scores on the first two principal components were compared among reference, invaded, and cleared vegetation types using Wilcoxon signed-rank tests.
The PCA was also used to investigate the functional response of bird communities to P. undulatum infestation and removal. Mean functional trait values in the species assemblage were calculated from the traits of individual species weighted by the abundance of individuals of the species. All data were centred and scaled to unit variance prior to analysis. All analyses were conducted using the R statistical platform (R Core Team, 2017).

RESULTS
Overall species richness and abundance of birds varied widely among sites and vegetation conditions, but no significant differences occurred among reference, invaded, and cleared vegetation (ANOVA test of the In total, 47 different bird species were observed across all study sites. The remnant section of the Greens Bush site provided the greatest diversity with 22 separate species, whilst only four species were observed in the area invaded by P. undulatum at Woods Point. The brown thornbill (Acanthiza pusilla) was the most abundant species across all sites, whilst a number of birds were observed only once throughout the study. Few exotic species contributed to these numbers: only three non-native species were observed across all sites. The only threatened species recorded was the powerful owl, Ninox strenua (Webster et al., 1999): two individuals were observed in reference controls of the Woods Reserve site and a further two birds observed within the cleared area of the Birdsland Reserve site.
Within patches invaded by P. undulatum, both richness and abundance differed significantly between the bird communities in and under the P. undulatum canopy and those in the Eucalyptus overstorey above (species richness: F 1,14 = 7.1, p = 0.018; abundance: F 1,145 = 10.6, p = 0.005). There were an average of 49% fewer species and 64% fewer birds in and under P. undulatum canopies compared to the overstorey.
At invaded sites, the Eucalyptus overstorey remains until either the trees die, or in some cases is overtopped by the P. undulatum, but does not reform. No Eucalyptus seedlings have ever been observed growing under a canopy of P. undulatum (Gleadow & Ashton, 1981;Gleadow & Walker, 2014;Naryan et al.;O'Leary et al., 2018). F I G U R E 2 Effect of time since P. undulatum removal on (a) relative species richness and (b) relative abundance (richness or abundance in cleared patch divided by richness or abundance in the corresponding reference control). Numbers within symbols refer to site numbers in Table 2 toe bird, Dicaeum hirundinaceum (Dh), were unique to site 4 and also had strong influence (Figure 3b). The shifts from remnant controls to invaded plots to cleared plots were strongly influenced by two types of birds: ground-feeding and/or nesting specialists, such as the white-browed scrubwren, Sericornis_frontalis (Sf), superb fairy-    Table 2 (FunctionalTrait.loadings.Fig4b.csv). Symbols and site numbers as in Figure 3 the plane of the first two principal components, while removal brings further pronounced change, but only at site 8 does the post-removal assemblage of functional traits resemble that of the control (Figure 4).

F I G U R E 4 Principal components analysis based on functional traits of bird species (see
The PCA loadings suggest that all traits contribute roughly equally to the pattern, although a ground-arboreal distinction in both nesting and foraging had correlated effects (Figure 4b). Feeding guilds were a distinctive contributor to these changes. Invasion by P. undulatum tended not to greatly affect the species richness in three of the F I G U R E 5 Effect of P. undulatum on abundance of species in four feeding guilds present at the ten sites listed in Table 1 Table 1, which had no Pittosporum-invaded vegetation four guilds we delineated, but sharply reduced the number of carnivorous species present at most sites ( Figure 5). Species numbers did not differ significantly among control, invaded and cleared habitats in the nectarivore-frugivore-granivore guild (Kruskal-Wallace χ 2 = 1.04, d.f. = 2, p = 0.59), the omnivore guild (χ 2 = 2.16, d.f. = 2, p = 0.34) or the insectivore guild (χ 2 = 2.12, d.f. = 2, p = 0.34). However, the loss of carnivorous bird species in P. undulatum-invaded patches and rebound in cleared patches (Figure 5d) produced a significant difference in richness (χ 2 = 11.01, d.f. = 2, p = 0.004), an effect detectable also in the loadings of the powerful owl (Ns) and laughing kookaburra (Dn) on the second principal component in Figure 3b, which tended to separate the bird communities in control and cleared vegetation.

DISCUSSION
P. undulatum profoundly alters the structure of vegetation in the landscapes it invades (Gleadow & Ashton, 1981;O'Leary et al., 2018;Rose & Fairweather, 1997). We found here that bird communities are also different in areas invaded by P. undulatum, although the effects are not readily apparent at the level of species richness or bird abundance ( Figure 1) and could be overlooked with less systematic observations.

Cleared, remnant and invaded areas differed in species composition
and the functional traits of the present bird assemblages (Figures 3   and 5), the most obvious being the decrease in carnivorous species in the invaded sites. Some bird species were present only in the remnant or cleared plots, while others, such as the white-throated treecreepers, grey fantails and white-browed scrubwren, were relatively resilient to P. undulatum invasion. The latter group are all relatively small, insectivorous species, and there may be trade-offs between the protection afforded by the dense canopy and the increased floristic diversity in the treated site. In contrast, species such as the rainbow lorikeet (Th, a nectarivore), little raven (Cm, an omnivore), varied sittella (Dc, a branchgleaning insectivore) and mistletoebird (Dh, a frugivore) were more sensitive to the long-term effects of P. undulatum invasion even after clearance and contributed strongly to the differences between communities in control and cleared habitats (as indicated by strong loadings on PC2 in Figure 3b).

Overall bird richness and abundance did not differ among vegetation types
It is striking that neither bird species richness nor abundance differed significantly among reference, invaded and cleared areas, given that a relationship between habitat structure and avian richness is well established (Ikin et al., 2012;MacArthur & MacArthur, 1961;Recher, 1969;Stirnemann et al., 2014). Foliage, flowers, bark, ground plants, air spaces and tree hollows have been identified as important habitat features for birds (Antos et al., 2008;Ikin et al., 2012;McElhinny et al., 2006;Stirnemann et al., 2014). In particular, ground vegetation, woody debris and logs support an invertebrate fauna that acts as a foraging substrate for many bird species (Antos & Bennett, 2005;McElhinny et al., 2006). Invasion greatly reduces plant growth beneath the dense P. undulatum canopy (Gleadow & Ashton, 1981;Mullett & Simmons, 1995;O'Leary et al., 2018;Recher et al., 2002;Stirnemann et al., 2014), and both species richness and abundance of birds were sharply reduced in and under P. undulatum canopies relative to the overstorey at invaded sites. But other than the near absence of a shrub/ground layer in invaded patches and a regenerating layer after clearance, many of the habitat components required by native birds are likely to remain, particularly in the intact Eucalyptus overstorey.
Invaded sites seem, therefore, to be able to support abundant bird communities.
Two important considerations constrain this optimistic conclusion, however. The first is that ground foraging birds, which form some of the most abundant bird communities in temperate Australia (Antos et al., 2008;McElhinny et al., 2006), may have suffered long-term decline apart from any effect of P. undulatum (Antos et al., 2008;Ford, 2011;Ford et al., 2001;Stirnemann et al., 2014). Fragmentation of landscapes along with urbanization and the introduction of invasive mammalian predators, especially foxes and cats, has resulted in the gradual reduction of ground-dwelling bird species from many temperate Eucalyptus forest and woodlands (Antos & Bennett, 2005;Antos et al., 2008;Ford, 2011). Recovery of ground vegetation following removal of P. undulatum may not attract a large avian fauna if ground-dwelling bird species richness is compromised at a regional scale.
The second consideration is that even if a Eucalyptus overstorey provides suitable habitat for many native species despite a high P. undulatum density, invasive populations of this species dramatically limit the germination of Eucalyptus seedlings, resulting in the virtual absence of younger Eucalyptus age classes able to replace the mature stock (Gleadow & Walker, 2014). The Eucalyptus overstorey persists in the invaded area until the trees die and does not reform. Eucalyptus seedlings have never been observed to grow under a canopy of P. undulatum (Gleadow & Ashton, 1981;Gleadow & Walker, 2014;O'Leary et al., 2018).
Therefore, the quality of habitat structure is likely to become increasingly diminished at invaded sites as the overstorey trees age and eventually die.

4.2
Speed of recovery of the shrub-and ground-level habitat limits the recovery of bird communities Most bird communities experienced similar changes when P. undulatum invaded a site and when it was removed. The trajectories between reference control patches and invaded patches along the first principal component of a PCA were similar among most sites, as were reversals when patches become cleared (Figure 3). However, the restoration of bird community composition following the removal of P. undulatum was usually incomplete. Sites 5, 6 and 9 (Montrose, Birdland Reserve and Glenfern Valley Bushlands) had the most similar communities in reference and cleared patches, but differences between reference and cleared patches at other sites were considerable. These differences may persist because the speed of recovery of the shruband ground-level habitat limits the recovery of bird communities, especially ground-dwelling birds. We found some indication of such a rate-  (Table 2).
One notable change in the functional traits present in a community is the reduction in the number of carnivorous bird species when P. undulatum is present ( Figure 5). A carnivore's view of prey is likely to be restricted by the dense foliage of a P. undulatum canopy (Stirnemann et al., 2014), impeding the foraging of larger species such as carnivorous magpies (Gymnorhina tibicen) and pied currawongs (Strepera graculina), as well as large territorial honeyeaters such as bell miners (Manorina melanophrys) and noisy miners (M. melanocephala) (Ford, 1979(Ford, , 2011. Smaller birds are more commonly associated with complex, heterogeneous and fine-grained vegetation (Holling, 1992;Fischer et al., 2008;Stirnemann et al., 2014). Thus, the discontinuity in vegetation texture created by P. undulatum appears to strongly limit the kinds of behavioural and life-history traits that can be supported. The key factor appears to the provision of canopy cover for roosting rather than foraging (McNabb & McNabb, 2011). While forest trees can also supply this need, there is a danger that lack of cover after clearing the invasive P. undulatum could be detrimental to these birds, creating some hesitancy around restoration programmes (McNabb & McNabb, 2011). It is noteworthy that, in the present study, we did not observe any powerful owls in areas invaded by P. undulatum. Our results, albeit based on limited observations, suggest that powerful owls find suitable habitat in areas where invasive tree control has been implemented. The retention of established tall canopy and subcanopy trees suitable for roosting in neighbouring areas is thus likely to important to ensure continuity of habitat for this species.
Only three out of the 47 bird species observed (6.38%) throughout the study were exotic species and all at relatively low abundance.
Despite reports suggesting European blackbirds may be a common seed dispersal vector of P. undulatum (Gleadow & Rowan, 1982), no difference in presence or abundance of this or other exotic species was found across the three vegetation types. This supports the hypothesis that it is the juxtaposition of habitats preferred by blackbirds, such as lawns and open areas, that promotes dispersal by these introduced avian vectors, rather than the presence of P. undulatum alone (Gleadow & Ashton, 1981;Gleadow, 1982). The role of exotic and native birds in seed dispersal and continuing invasion and re-invasion by P. undulatum warrants further study (Gleadow, 1982).

Birds as indicators in restoration management and implications for practitioners
Birds offer insights into restoration management distinct from those uncovered by monitoring vegetation alone (Munro et al., 2011). The rate and scale of restoration programmes can affect the types of bird communities that are able to either remain or recolonize particular areas, affecting density, diversity, and functional groups. Restored sites often vary in vegetation composition and structure independently of their time under management (Batisteli et al., 2018). While the outcomes from this study have potential to improve other restoration management programmes focused on invasive trees, the effect of invasive tree removal will likely differ depending upon the habit and ecology of the trees in question, together with their relationship with the surrounding environment. Attention to bird communities and focal species can improve our capacity to monitor the effects of invasive tree removal and other restoration efforts.
Information gathered here offers details distinct from those uncovered from a monitoring programme focusing specifically on vegetation alone, and that the richness, abundance and functionality of birds in invaded sites should be taken into consideration when designing and evaluating the efficacy of ecological restoration programmes. In our case, it is essential that habitat is maintained for the threatened powerful owl. In order to do this, management must be slow and sensitive and carried out in patches over a 5-10 year period, or even longer. This retains some of the invaded forest in adjacent areas to provide habitat in the transitional phase while new trees become established and become tall enough. Leaving dead trees left from 'drill and fill' strategies also provides habitat while the vegetation recovers. Such approaches are extremely difficult given the nature of funding for management that tend to drive short-term interventions on a wide scale. It is also challenging from the perspective of amenity and safety: leaving many dead trees could be perceived as ugly, they may become a fire hazard and may even be in danger of falling on people in areas that are accessed by humans.
While the outcomes from this study have potential to improve the management of other restoration programmes focused on invasive trees, the effect of invasive tree removal will differ depending upon the habit and ecology of the trees in question, together with their relationship with the surrounding environment. We recommend that a monitoring programme similar to that presented here is conducted to improve the capacity of invasive tree removal projects, to help in the design of the programme to ensure there is transitional habitat to help maintain biodiversity values and to assess the effects of restoration efforts.

CONCLUSION
In this study, we compared the functional diversity, density and species richness of bird communities in areas invaded by P. undulatum with areas that had been managed to remove this woody weed, and adjacent remnant areas. We found that while clearing tended to shift the community back towards that found in the remnant sites, this recovery seldom left the post-clearing assemblage similar to the control assemblage. We conclude that the richness, abundance and functionality of birds in invaded sites should be taken into consideration when designing and evaluating the efficacy of ecological restoration programmes.