Systematic definition of threatened fauna communities is critical to their conservation

Most terrestrial ecological communities are defined primarily on their constituent flora. We aimed to develop a repeatable approach to defining a community, both intact and degraded, based on its fauna. We demonstrate how the approach can be used to guide conservation—for example, determining whether a multispecies community of animals is threatened.


| INTRODUC TI ON
The management of biodiversity is often focussed on protection of threatened species, exemplified by widely used tools such as the IUCN Red List of Threatened Species. There is good reason for such a focus, with species decline and extinction approaching catastrophic levels (Ceballos, Ehrlich, & Dirzo, 2017). However, given that multiple species within different taxonomic groups, geographical locations and communities are under threat (Pimm et al., 2014), it is well accepted that a species-focussed approach is insufficient (Mace, 2014), and increasingly, ecosystems and ecological communities are being recognized as principal entities needing protection.
For example, in 2015, the IUCN endorsed a new set of criteria for listing ecosystems as threatened entities (Keith et al., 2015), in recognition that the concept of ecosystem collapse is analogous to that of species extinction (Keith et al., 2013).
Focussing conservation action on ecological communities presents pragmatic benefits. A community-wide approach recognizes explicitly that numerous species contribute to ecosystem function via a complex array of biotic and abiotic interactions (Hooper et al., 2005;Ings et al., 2009). It follows that changes in elements of these communities-even if those elements do not individually qualify as "threatened"-can cause unforeseen or cascading effects (Newsome & Ripple, 2015). From a management perspective, a focus on communities allows for a more holistic approach to addressing anthropogenic threats because it accounts for the array of ecological processes, interactions and species-specific responses that characterize multispecies assemblages (Keith et al., 2015). Furthermore, perverse or unintended consequences of management actions may be avoided, as management actions are designed in consideration of community-level responses and the cascading interactions in ecological and socioeconomic systems, a central tenet of network theory (Dee et al., 2017). Lastly, from a perspective of cost-effectiveness, simultaneously managing multiple species that share ecological traits, threats and habitat characters can help to better prioritize and target limited funding for conservation (Gonthier et al., 2014).
Ecological communities can be defined as biotic assemblages delineated by a characteristic composition of species, associated abiotic environmental variables, and their ecological functions and interactions, occurring within a bounded space (adapted from Keith et al., 2013). In practice, however, most ecological communities documented and formally protected under policy are vegetation communities, defined and described based largely on geology, hydrology and floristic composition (Australian Government, 1999;IUCN-CEM, 2018;Mucina et al., 2016). Vegetation communities and their associated landforms are relatively stable in time and space. Because of this, their detection, description and mapping of their spatial distribution-along with assessment of their decline-are conceptually relatively straightforward. For example, Australia's national environmental legislation allows for protection of ecological communities, of which 76 are currently listed as threatened (Australian Government, 1999). However, like the IUCN Red List of Ecosystems, the emphasis of these formally recognized Australian ecosystems is firmly on vegetation communities. Only two include non-plant biota (microbes and invertebrates) in their description (Australian Government, 1999).
Importantly, animal communities can be degraded even in ecosystems that appear, superficially at least, to have relatively intact vegetation. While habitat loss is a major driver of biodiversity declines, a multitude of other threats acts on animal species, including many insidious threats operating within apparently intact systems (Maxwell, Fuller, Brooks, & Watson, 2016). For example, the depletion of wild animals through overexploitation is recognized from tropical forests across the world ("empty forest syndrome") (Wilkie, Bennett, Peres, & Cunningham, 2011). Several other examples occur from different Australian ecosystems, including (a) the ongoing depletion of small mammals in northern Australia due to feral predators and altered fire management regimes, in a biome generally considered to be largely uncleared and minimally modified (Fisher et al., 2014;Lawes et al., 2015); (b) the sustained high levels of predation of birds by introduced cats in natural environments (>250 million per year), particularly in Australia's least-modified regions (the dry, arid interior) (Woinarski et al., 2017); and (c) the degradation of bird communities even in continuous, intact woodland due to the dominance of a hyper-aggressive native competitor (Howes & Maron, 2009).
Declines in animal species can therefore precede-and indeed, precipitate-ecosystem collapse. This is because a depauperate or degraded animal community can itself be a threat to the persistence and health of vegetation communities, given the key role performed by intact animal assemblages in maintaining vegetation through services such as pollination, predation, seed dispersal and preventing invertebrate outbreaks (Sekercioglu, 2006;Wilkie et al., 2011). For example, small insectivorous birds play an important role in ecosystem health and provide vital ecosystem functions such as pollination, seed dispersal, insect control and decomposition (Şekercioğlu, Daily, & Ehrlich, 2004;Wenny et al., 2011). These functions are economically important in both natural and agricultural landscapes (Wenny et al., 2011;Whelan, Şekercioğlu, & Wenny, 2015). Similarly, the local extinction of co-evolved pollinators and effective pollination processes threatens the long-term persistence of entire forest and woodland ecosystems (Woinarski, Connors, & Franklin, 2000). As such, failing to protect animal communities may undermine conservation actions that focus on individual threatened species and/or vegetation communities.
Here, we emphasize that defining and, where necessary, protecting whole animal communities are crucial complements to efforts focussed on conserving individual threatened species, plant communities and ecosystems. Yet, it is rarely done in terrestrial settings. In this paper, we present an approach to first defining, and then estimating trends in the extent and condition of, the woodland bird community of Australia's highly modified temperate and subtropical zone.
This large group of birds has been the focus of substantial research for many decades (e.g., Recher, Holmes, Schulz, Shields, & Kavanagh, 1985, Barrett, Ford, Recher, & Barrett, 1994, Ford, 2011b, not least because of prevailing long-term concerns regarding the declining status of the community, its exposure to ongoing threats and evidence that many species that comprise this community are on the brink of calamitous decline (Ford, 2011a;McGinness, Arthur, & Reid, 2010;Szabo, Vesk, Baxter, & Possingham, 2011;Vesk & MacNally, 2006). Degradation of the vegetation communities in which the community is found does not appear reliably to indicate a degraded woodland bird community is present, nor does high-condition vegetation necessarily obviate bird community degradation (Howes & Maron, 2009;Kath, Maron, & Dunn, 2009;Maron, 2005;). Yet consensus on what comprises this "woodland bird" community and where it occurs is lacking. Further, differences in how this community is defined may affect the extent to which it meets criteria for being threatened (Fraser, Garrard, Rumpff, Hauser, & McCarthy, 2015;Fraser, Pichancourt, & Butet, 2017). Therefore, despite widespread concern about the community, its threat status has never been formally assessed.
Here, we detail a method to define and evaluate a widespread, continental-scale fauna community, in this case the Australian Temperate and Subtropical Woodland Bird Community. Using structured expert elicitation, we describe which species comprise the community then we examine how community characteristics relate to variation in expert-judged community condition and distinguish it from other bird communities. We apply a method developed for plant communities to create a metric of the condition of the community, where it occurs (Sinclair, Griffioen, Duncan, Millett-Riley, & White, 2015). Last, we evaluate the community against the Australian Government's criteria for listing as a threatened ecological community under national biodiversity conservation law.

| ME THODS
There were two main components to this work (see Figure 1): (a) defining and describing a recognizable, consistent community of woodland birds, and (b) evaluating the community against the criteria for listing as a threatened ecological community (TEC) under Australia's Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act). This is the pathway for officially identifying a threatened ecological community in Australia; we made no a priori supposition that the community will meet the criteria, though as a starting point, there is recognition that the community as an entity is diminishing (Barrett et al., 1994;Ford, 2011aFord, , 2011bRecher et al., 1985). Our approach is outlined below and described in further detail in Supporting Information Appendix S1.

| Expert workshops
We invited woodland bird experts from across Australia to attend Within these regions, experts considered the potential habitat for the TSWBC to be restricted to areas of woody vegetation with a canopy cover less than 70% (i.e., excluding wetter, forest ecosystems: Specht, 1970), excluding mallee (low woody vegetation of a particular, multi-stemmed growth form of various species of Eucalyptus) as it was associated with more arid climates and distinct bird species.
Beyond these broad parameters delineating the relevant vegetation formation and climatic zones, descriptions of the community and its condition were based on bird species composition alone.   NCAS extent 1980NCAS extent , 1998NCAS extent , 2015NCAS extent (1977NCAS extent -1981NCAS extent , 1998NCAS extent -2002NCAS extent , 2011NCAS extent -2016 For eastern Australian states, experts were then presented with a list of all Australian land birds and asked to individually respond to a series of questions for each bird species: 1. "Is this species a part of most woodland bird communities (degraded or intact)?"; and then if yes, two further questions: 2. "Would you expect to see this species more often in intact than degraded communities?"; and 3.
"Would you expect to see this species more often in degraded than intact communities?". This process was designed to identify the full suite of species comprising the woodland bird community, including generalist as well as specialist species. Crucially, participants were instructed to focus on their concept of intact versus degraded bird communities in their answers, not intact versus degraded vegetation. This distinction is important, because there is often limited direct relationship between vegetation community condition (typically measured in relation to floral and structural composition) and the species richness or composition of woodland bird communities (Fraser, Rumpff, Yen, Robinson, & Wintle, 2017).
We compiled information from all experts and identified the species included by 70% or more of experts under each question.
For Question 1, this threshold was identified by the experts as distinguishing between a clear member of the TSWBC and one only tenuously associated with woodlands. We excluded introduced species from lists generated by Question 1 and Question 2 but retained them for Question 3 because an increased richness or abundance of introduced species might be an indication that the bird community is degraded. We circulated, for further comment, the compiled information to all the contributing experts as well as a range of experts who were unable to attend on the day. This process is broadly consistent with the process of deciding on objectives and measures of performance used in structured decision making (Gregory et al., 2012) and allowed us to reach a consensus on which species should be included in a TSWBC.
Because of the narrower geographical coverage and lower num-

| Community definition
In workshop discussions, experts agreed that beyond a list of species, there was a characteristic taxonomic and functional profile of F I G U R E 2 Subcommunity variants of the Temperate and Subtropical Woodland Bird Community the bird community that varied across the temperate and subtropical ecoregions of Australia primarily through species-or genus-level substitutions. We therefore considered the TSWBC to cover the entire extent depicted in Figure 2, but in conjunction with expert opinion provided in the workshops, we developed descriptions of six subcommunities, split by their ecoregional affinity. Each subcommunity had its own species list (described in Supporting Information Appendix S3) and threat was assessed separately for each group.

| Community composition
Following the workshops, participating experts provided published and unpublished bird survey data from 20 projects collected using a 20-min, 2-ha survey method in woodland, forest, heathland and grassland habitats. For each survey, we recorded bird species richness and the proportion of species that were TSWBC species (determined through the expert workshop process described above); species more typically associated with intact woodland bird communities (determined through the expert workshop process described above); species more typically associated with degraded woodlands (determined through the expert workshop process described above); water birds; hollow nesters; small birds (<50 g), ground insectivores; bark insectivores; foliage insectivores; granivores; carnivores; frugivores; and nectarivores (trait data from G. Luck unpublished data).
Species could belong to multiple categories. We compared the proportion of species in each of these groups among the four habitat types by looking at the 95% confidence intervals around the proportions, to identify the extent to which the TSWBC was distinct from assemblages in other habitats.

| Community condition
For each site for which bird data were submitted, the submitting expert noted whether they considered the bird community to be intact or degraded for a subset of bird surveys from woodland sites. We to the species assemblages more commonly seen in those areas. We selected bird data for the online survey from the submitted data provided by our experts. Examples of survey data were selected to represent different levels of each of the variables that had differed between the communities that the submitting experts had nominated as intact and degraded (i.e., low or high species richness, and the proportion of species that are small, associated with intact communities or associated with degraded communities). In the survey, experts were asked to assign absolute condition values on a scale of 0-100 to each of five woodland bird community calibration sites, based on species lists (from 2-ha 20-min bird surveys), where 0 represents the worst possible condition and 100 represents the best possible condition of the TSWBC. These five calibration sites were chosen to include sites that we expected to be in very good (intact: high species richness, a high proportion of species that are small or associated with intact communities, and a low proportion of species associated with degraded communities) or very bad (degraded: low species richness, a low proportion of species that are small or associated with intact communities, and a high proportion of species associated with degraded communities) condition based on our initial analysis of expert-specified intact and degraded communities.
Next, participants were asked to answer a series of questions, rating the relative value of a number of other sites based on species lists using swing weighting (Von Winterfeldt & Edwards, 1986). In each question, we included one of our calibration sites so that the relative weights given by each expert could be calibrated to a common condition scale of 0 and 100.
We took these values for expert-judged community condition and investigated how they related to the key variables identified as being likely to align with condition: (a) species richness, (b) the proportions of species that were small (<50 g), (c) the proportion of species primarily associated with intact communities and (d) the proportion of species primarily associated with degraded communities. We divided the expert-judged condition score by its maximum to rescale the data between 0 and 1. Using the package glm2 (Marschner, 2014)  Appendices 1 and 2). We compared the model coefficients and significance of relationships between the models for each of the community variants to determine whether there are ecoregional differences in the way that experts perceive bird community condition.

| EPBC listing criteria
The EPBC Act has six criteria governing the threat status of ecological communities (

| Quantifying declines in extent
Data on bird occurrence and distributions across suitable habitat types within the relevant ecoregions are not available from prior to 1977. This is unfortunate because the most profound declines in the TSWBC are thought to have occurred before then, when large swathes of their habitat was cleared for agriculture (Bradshaw, 2012;Evans, 2016 Relates to communities distributed over less than 10,000 km 2 3. Loss or decline of functionally important species Critically Endangered: very severe (>80%) decline and restoration cannot be achieved in the immediate future (three generations max 60 years) Endangered: severe (>50%) decline and restoration cannot be achieved in the near future (five generations, max 100 years) Vulnerable: substantial (>20%) decline and restoration cannot be achieved in the medium term future (10 generations, max 100 years)

Reduction in community integrity
Critically Endangered: very severe (>80%) decline and restoration cannot be achieved in the immediate future (three generations max 60 years) Endangered: severe (>50%) decline and restoration cannot be achieved in the near future (five generations, max 100 years) Vulnerable: substantial (>20%) decline and restoration cannot be achieved in the medium term future (10 generations, max 100 years)

Rate of continuing detrimental change
Critically Endangered: an observed, estimated, inferred or suspected detrimental change of at least 80% over the immediate past or projected for the immediate future (three generations max 60 years) Endangered: an observed, estimated, inferred or suspected detrimental change of at least 50% over the immediate past or projected for the near future (five generations, max 100 years) Vulnerable: an observed, estimated, inferred or suspected detrimental change of at least 30% over the immediate past or projected for the medium term future (10 generations, max 100 years)

Quantitative analysis showing probability of extinction
Critically Endangered: at least 50% in the immediate future (three generations max 60 years) Endangered: at least 20% in the near future (five generations, max 100 years) Vulnerable: at least 10% in the medium term future (

| Evaluating declines in condition
We obtained BirdLife Australia data from 1-km 2 , 20-min bird surveys  Therefore, we determined that a bird community would qualify as the woodland bird community if 70% or more of species were included in the appropriate regional list of woodland bird species (Supporting Information Appendix S3).

| Community condition
For all six community variants of the woodland bird community, the condition of the community as judged by experts was best characterized by species richness combined with either the proportion of small species or proportion of intact species (Supporting Information Appendix S2). Because the three eastern mainland ecoregions (subtropical Queensland, temperate south-east mainland Australia and temperate South Australia) are contiguous and the coefficients were similar among ecoregions, we used averaged coefficients to calculate community condition for all three associated community variants ( Table 2).  We used averaged coefficients to calculate the condition of the two Western Australian woodland bird communities (Table 2).

| Evaluating declines in extent
We estimated substantial declines in the amount of suitable habitat for the TSWBC since 1750. In 1750, there was approximately 168 million hectares of potential woodland habitat available, and in 2016, only 92 million hectares remained (Figure 4). This decline has not been evenly distributed across Australia (

| Evaluating declines in bird community condition
Based on our estimates, the average condition of the TSWBC declined severely since 1750. For all community variants except Tasmania, most of this decline happened between 1750 and 1981 ( Figure 5). In Tasmania, there was also a dramatic decline between 2002 and 2016.
Similarly, the TSWBC in the two Western Australian community variants declined in condition between 1981 and 2016 ( Figure 5).

| EPBC Act listing criteria
Our evaluation suggested that the newly described TSWBC meets the criteria for Endangered or Critically Endangered across its entire extent. This is based on the evidence presented here regarding reduction in both the extent and condition of community, supported by information in Supporting Information Appendices S3 and S4 with regard to the loss of functionally important species and continued threats. Our analyses showed that the TSWBC had reduced in extent substantially enough since 1750 to qualify as Vulnerable (>50% decline) in temperate south-east mainland Australia and Western Australian Banksia Woodland community variants and Endangered (>70% decline) in temperate South Australia (Tables 1 and 4). Our  Woodland Bird Community (TSWBC)-based on bird species composition, and a method for evaluating its condition based on existing data sources and expert opinion. This allowed us to describe quantitatively the oft-cited declining "woodland bird community" (Rayner, Lindenmayer, Gibbons, & Manning, 2014), and formally nominate the TSWBC for protection as a threatened ecological community under Australian environmental law. The definition of any community is by necessity arbitrary, and subject to interpretation driven by scale, biome or evolutionary history (Mittelbach & Schemske, 2015). Nevertheless, our approach was explicit and transparent; we focussed on TSWBC due to a long-standing expert consensus that it was a coherent and recognizable community undergoing decline; we used a range of regional experts to define the community via consensus and test their judgements for consistency; and we focussed on co-occurrence of species and condition, rather than a simple list.

| D ISCUSS I ON
The final point is the most important emphasis of our approach; communities of species such as birds, co-occur, respond to multiple threats together and should be managed as entities for their conservation (Tulloch, Chadès, & Lindenmayer, 2018).
The decline of species globally and the increasing number of species that are threatened across multiple ecosystems (Ceballos et al., 2017), in an era of dwindling focus on environmental management and stewardship (Waldron et al., 2013), suggests that protecting communities may be more effective (and financially viable) than purely focussing on preserving individual species (Keith et al., 2013).
Compared to spatially explicit components of the landscape such as vegetation (Keith et al., 2015), the description and evaluation of the state of terrestrial animal communities is less tractable, due to the challenges of defining the condition and composition of communities that are mobile and span multiple vegetation types (Ferenc, Sedláček, & Fuchs, 2014). However, the process we developed led to a widely accepted, quantitative description of a faunal community. TA B L E 3 Change in extent of potential habitat for the Temperate and Subtropical Woodland Bird Community from pre-European settlement (pre-1750) to 2016 and for the periods 1980-1998, 1998-2015 and 1980-2015   The NVIS dataset and Landsat datasets provide different estimates of habitat extent due to differing resolutions and inclusion criteria. See detailed methods for more information.

| Defining an animal community
We distinguished potential woodland bird habitat from the definition of the TSWBC itself. Based on the input of contributing experts, we determined a set of criteria that were mostly independent of characteristics of vegetation. This is a key distinction of our process-recognizing that the occurrence of an animal community is not necessarily tied to a discrete set of abiotic and/or vegetation associations (Ives  McAlpine et al., 2016). Indeed, in Australian woodlands, vegetation condition is often not strongly associated with bird species richness or composition . In such cases, animal communities are most appropriately characterized directly, rather than bundled with a type or condition of vegetation.

| Broader applications
The approach described in this article engages experts in the process of describing a community. This capitalizes on available knowledge, creates support for the definition of the community and mutes inconsistencies that appear in the broader literature (Fraser et al., 2015;Fraser, Pichancourt, et al., 2017

| Evaluating threat status of the community
Our evaluation clearly identified that the TSWBC met criteria for Endangered or Critically Endangered status across four of the Australian threatened ecological community criteria, two of which we report here, across all six community variants. Declines in community integrity provided the strongest evidence that the TSWBC is threatened. The EPBC Act states that a community qualifies as Endangered if the change in integrity is severe and unable to be reversed soon (20 years or five generations). We found that the integrity of the woodland bird community has undergone severe or very severe declines that are unlikely to be restored in 20 years (due to the absence of nesting hollows etc). These declines were based on average values calculated across large numbers of 20-min surveys.
In assessing the condition of a particular example of a bird community at a site, the condition value calculated from a single 20-min survey is unreliable on its own (Maron, Lill, Watson, & Nally, 2005).
Therefore, site-level assessments of the presence and integrity of the TSWBC will require multiple surveys, ideally across different seasons, to derive an understanding of the community's condition at that site.
Bird species richness and the proportion of species that were small (<50 g) and those which experts judged were typically more common intact bird communities were the most influential factors determining the condition of the TSWBC. This makes intuitive sense, as the proportion of small or intact woodland specialist TA B L E 4 The threat status of the TSWBC, by ecoregion and against each of the Australian Government EPBC Act threatened ecological community criteria  Thomson et al., 2015). Both processes are listed as Key Threatening Processes under the EPBC Act. The small-bodied species that form the majority of the TSWBC play vital roles in ecosystem health through complementary pollination, seed dispersal, insect control and decomposition services (Şekercioğlu et al., 2004;Wenny et al., 2011), and their loss can result in declines in vegetation condition (Grey, Clarke, & Loyn, 1998;Maron et al., 2013).
Our evaluation of the long-term declines in the condition of the woodland bird community was limited by the difficulties of comparing data between the two BirdLife Australia Atlases. Surveys in each were conducted using somewhat different methodologies, which may have affected our estimates of declines in community integrity between 1981 and 2002, despite our efforts to minimize the differences by restricting our analyses to 20-min surveys. About one-quarter of species were likely to be found less commonly in the second atlas based purely on differences in survey methodology (Barrett et al., 2004), potentially resulting in lower species richness estimates in 2002. However, most of the species that Barrett et al. (2004) identified as less likely to be detected due to survey methodology are large-bodied birds. So, while species richness may be higher as an artefact of survey methodology, the proportion of small birds may also be artificially higher, having the opposite effect on condition scores.
The TSWBC did not meet the criteria for threatened status based on decline in geographical distribution alone in all ecoregions, and this reflects the history of land clearing in eastern Australia, where there is a gradient of vegetation loss from temperate to tropical woodlands. Nevertheless, extensive clearing of woody vegetation is continuing, and even increasing, in Australia (Reside et al., 2017) and globally (Richards & Friess, 2015). Based on our analyses of changes in extent of potentially suitable habitat, we estimate that the loss of woodland bird habitat since pre-1750 is 76,000,000 ha across Australia, approximately 45%. In temperate South Australian and temperate south-eastern mainland regions, little woodland remains and clearing has all but ceased (Bradshaw, 2012;Evans, 2016). Other areas (especially subtropical Queensland) are now the focus of agricultural expansion. The extent of clearing has recently been increasing with a significant surge in Queensland with predicted and intersecting climate change effects on forests, woodlands and agriculture (Lawrence & Vandecar, 2015), suggests that the geographical distribution of the TSWBC will continue to decline.

| CON CLUS ION
The consequences of global change affect entire communities, not just single species. However, defining ecological communities based solely or predominately on plant species composition risks missing the point of community-focussed conservation.
While a focus on the protection of vegetation communities shines a light on key proximate threats like deforestation, it often fails to illuminate the complex biotic associations that are both sustained by, and crucially, sustain the vegetation.
Focussing conservation on an interacting community of animal species might be more likely to prevent an unnoticed gradual attrition of individual species, perhaps considered more "common," but vital component for the function of both the vegetation community and the animal community (Birdlife Australia, 2015;Gaston, 2010;Gaston & Fuller, 2008). As such, we advocate that biodiversity conservation must extend beyond protection of vegetation communities to also consider animal communities as an essential complement to the protection of species habitat.

DATA ACCE SS I B I LIT Y
All data to which we have rights are provided in appendices or on the Open Science Framework (https://osf.io/dhqb6/). All other data sources are either cited in text and available from other sources (e.g., BirdLife data) or provided by woodland bird experts for purely exploratory purposes.