Poplar box woodlands of Eastern Australia: an assessment of a threatened ecological community within the IVC framework

Aims: Ecosystems nationally at risk in Australia are listed under the Environmental Protection and Biodiversity Act (EPBC Act), and many cross State jurisdictional boundaries. The determination of these ecosystems across the State boundaries are based on expert knowledge. The International Vegetation Classification has the potential to be useful as a cross-jurisdictional hierarchy which also gives global perspective to ecosystems. Study Area: All bioregions that include Eucalyptus populnea as a dominant or major component of woodlands across the species known distribution. Methods: We use plot-based data (455 plots) from two states (Queensland and New South Wales) in eastern Australia and quantitative classification methods to assess the definition and description for the Poplar Box Woodland ecosystem type (hereafter “ecological community” or “community”) that is listed as endangered under the EPBC Act. Analyses were conducted using kR-CLUSTER methods to generate alliances. Within these alliances, analyses were undertaken to define associations using agglomerative hierarchical clustering and similarity profile testing (SIMPROF). We then explore how assigning this community into the IVC hierarchy may provide a mechanism for linking Australian communities, defined at the association and alliance levels, to international communities at risk. Results: We define three alliances and 23 associations based on the results of floristic analysis. Using the standard rule-set of the IVC system, we found that the IVC hierarchy was a useful instrument in correlating ecological communities across jurisdictional boundaries where different classification systems are used. It is potentially important in giving a broader understanding of communities that may be at risk continentally and globally. Conclusions: We conclude that the IVC hierarchy can incorporate Australian communities at the association level into useful units at higher levels, and provides a useful classification tool for Australian ecosystems. Taxonomic reference: PlantNET (http://plantnet/10rbgsyd.nsw.gov.au/) [accessed June 2019]. Abbreviations: EPBC Act = Environmental Protection and Biodiversity Act; IVC = International Vegetation Classification; NMDS = non-metric multidimensional scaling; NSW = New South Wales; PCT = Plant Community Type; QLD = Queensland; RE = Regional Vegetation Community; SIMPER = similarity percentage analysis; SIMPROF = Similarity profile analysis.


Introduction
One of the core methods for tackling the loss of biodiversity is the listing of threatened ecological communities on international, national and regional lists (IPBES 2019). However, this necessarily requires that such communities are defined and are identifiable. Without clear definitions of inclusion or exclusion we risk conservation priorities being misdirected (Hunter 2021a;Saunders et al. 2021).
One key impediment to the process of listing threatened ecological communities is a lack of jurisdictional conformity in typology (Gellie et al. 2018;Muldavin et al. 2021;Saunders et al. 2021). Only through the unification of terminology and procedure, at least with some critical components of survey and naming across jurisdictions, can a clearer understanding of the distribution and threats to communities occur (De Cáceres et al. 2015;Gellie et al. 2018;Luxton et al. 2021).
A lack of jurisdictional conformity is a global issue within many regions and concerted efforts are being made to unify classificatory procedures at all levels to allow greater regional, continental and global understandings (Faber-Langendoen et al. 2014;De Cáceres et al. 2018;Luxton et al. 2021;Muldavin et al. 2021). Though many early attempts at classifying vegetation within Australia were continental in focus (e.g. Carnahan 1976;Beadle 1981;Walker and Hopkins 1990;Specht et al. 1995), classification within Australia has become strongly State and Territory led, each with their own individualistic approaches (Gellie et al. 2018;Luxton et al. 2021). In most instances, intuitive qualitative supervised methodologies have been used to create typologies, often with minimal hierarchical structures that are used primarily for mapping (Gellie et al. 2018). As such, difficulty arises when a threatened ecological community is listed at the continental scale on the Federal Environmental Protection and Biodiversity Act 1999 (EPBC Act; https:// www.environment.gov.au/epbc) and is known to occur across jurisdictional boundaries within Australia. An intent of threatened community listings is to channel and prioritise limited resources towards those systems that are in urgent need of immediate protection, however, listings are often constrained by limited knowledge, outdated taxonomy and jurisdictional differences (Wallace and Fluker 2015; Dovey and Walker 2018; Saunders et al. 2021). Currently the EPBC Act contains 92 threatened ecological communities (4 Nov 2021). Any organisation or community member can nominate a listing which goes to a scientific committee for discussion. Potential listings are then refined and placed on public exhibition for comment before finally being presented to the federal minister for acceptance or rejection. Although guidelines suggest that communities should be defined based on numerical classification this has not been applied to many currently listed, some of which are clearly defined based on geomorphological features with only a generalised concept of a floristic assemblage (see, e.g., Hunter and Hunter 2020;Hunter 2021a). Without a full comprehen-sion of all floristic and ecological components and interrelationships with co-occurring types, a real understanding cannot be gained of threats and persistence (Franklin et al. 2016;Jansen et al. 2016).
Although adjacent to each other and sharing approximately 1,500 km of border the vegetation classification methodologies between New South Wales (NSW) and Queensland (QLD) (Gellie et al. 2018) are highly divergent. Within QLD communities are defined as regional ecosystems (RE) that are classified at a thematic level considered equivalent to association. Unlike traditional concepts of an association, which strongly emphasize floristics, REs in QLD are named based firstly on the bioregion (IBRA7; Thackway and Cresswell 1995) in which they occur, secondarily by geology, landform and soils and only thirdly by the most dominant stratum in terms of biomass (not height) and then dominant floristics within strata (Gellie et al. 2018;Addicott et al. 2021). The approach is mapping based and created predominantly through expert opinion, with more than 1300 types currently defined (Gellie et al. 2018), although recently quantitative classification approaches are being implemented (Addicott et al. 2018;Addicott et al. 2021). In NSW, the vegetation classification has three hierarchical levels, of which the Plant Community Type level (PCT) was derived under a separate process to the other thematic levels of class and formation (Keith 2004;Benson 2006;Gellie et al. 2018). PCTs are based on floristics, unlike REs, and thus are closer to the traditional concept of association sensu Braun-Blanquet (Benson 2006). Un-supervised, semi-supervised, and, more rarely, fully supervised methods were used to define PCTs, depending on the density of qualitative data (Benson 2006). In contrast to REs, the PCT approach was not mapping based. Currently, approximately 1500 PCTs are defined for NSW. Independently developed classes and formations have also been defined for NSW through largely supervised and semi-supervised methods, with the relationships between the thematic levels based on expert opinion (Gellie et al. 2018). Overall NSW and QLD typologies have been developed through expert opinion; rarely do plot-based analyses underpin the circumscription of units.
Plot-based techniques are needed to better circumscribe communities within and across jurisdictions for greater consistency. Several tests have been completed within select vegetation types (e.g. Hunter and Lechner 2017;Addicott et al. 2018;Hunter 2020;Hunter and Hunter 2021a;Muldavin et al. 2021). Here we introduce an additional test based on the Poplar Box Woodland dominated by Eucalyptus populnea. Eucalyptus populnea is a widespread species with a wide edaphic tolerance but is generally restricted to annual rainfalls between 300 and 500 mm (Beeston et al. 1980;Beadle 1981) with a distribution almost equally divided across NSW and QLD and is restricted to these two jurisdictions. Beeston et al. (1980) subjectively defined 31 Eucalyptus populnea communities based on structure primarily for mapping purposes. Beadle (1981) defined a Eucalyptus populnea alliance with seven sub-alliances. These subjective cross jurisdictional works have been replaced by the Qld RE and the NSW PCT classifications. Within QLD, 34 REs have been circumscribed that are either dominated by, or have Eucalyptus populnea, as a characteristic canopy species, and they are found in 4 bioregions and a number of land zones (Suppl. material 1). Within NSW, 56 PCTs have Eucalyptus populnea as the dominant overstorey species or listed as a diagnostic element. These PCTs are placed within 23 Classes and 10 Formations (see Suppl. material 1).
In 2013, Poplar Box (Eucalyptus populnea) Grassy Woodland on Alluvial Plains was nominated as a nationally threatened ecological community within Australia and was accepted as such in 2019 under the Commonwealth EPBC Act 1999. At the time of listing no independent numerical classification was undertaken but existing state-based classifications were used as a guide for what should be included within listing advice for identification. Although there are 90 PCTs and REs types across both states that have Eucalyptus populnea as a defined diagnostic component, the current conservation listing advice for the endangered community only lists four PCTs and five REs as being characteristic of the endangered community. The listing advice was based on expert opinion and no cross jurisdictional analyses were performed to justify the conclusions made or to assess the interrelationships of the types incorporated. The differences between classification systems and methodologies in NSW and QLD and a lack of plot-based analysis limits our understanding of communities dominated by Eucalyptus populnea across its range. To address conservation priorities and to better place limited management resources, the interrelationships of these communities need to be better understood from a local, continental and global perspective. Hierarchical classification schema allow for a better understanding of interrelationships between communities and the conceptualisation of different ranks allows the scale of management to be applied at appropriate scales (Faber-Langendoen et al. 2018;Luxton et al. 2021).
One such hierarchical classification schema is the International Vegetation Classification (IVC) system, which is based on the EcoVeg approach (Faber-Langendoen et al. 2014) and was developed to characterise the world's vegetation. Due to its hierarchical structure, which includes eight thematic levels (indigenous and anthropogenic), the IVC enables vegetation types to be defined locally, regionally, and globally, without regard to jurisdictions (Gellie et al 2018;Muldavin et al. 2021).
Here we propose to resolve the issues of differences between state-based classification schema and the lack of knowledge of their interrelationships by using plot based analytical techniques and defining types using the IVC criteria and structure across the full range of systems in which Eucalyptus populnea is a characteristic dominant. The results are used to assess the current circumscription of the listed endangered Poplar Box Grassy Woodlands on Alluvial Plains.

Study region
The study region incorporates the full range of environments across NSW and QLD in which Eucalyptus populnea is found to be a dominant or a characteristic species. This includes the eastern Australian bioregions of: Brigalow Belt North, Brigalow Belt South, Desert Uplands, Darling Riverine Plains, Nandewar, Mulga Lands, Cobar Peneplains, NSW South Western Slopes and the Murray Darling Basin (Figure 1) covering over 960,000 sq km and 14 degrees of latitude (Beeston et al. 1980).

Data and statistical analysis
Different Australian jurisdictions (States and Territories) have different protocols for plot-based vegetation sampling, using different sized plots and scoring systems (Gellie et al. 2018). There currently is no Australian national vegetation database system, although data exchange protocols for incorporating data from individual databases are under development (TERN AEKOS). Thus, vegetation data from the different databases were used to cover the extent of Eucalyptus populnea dominated communities within eastern Australia. These databases included the QLD government 'CORVEG' database, which is the most comprehensive database covering QLD, and a private database curated by one of the authors (JTH; listed in GIVD as Au-Au-003 -https://www/givd.info/databses.xhtml), which primarily covered NSW but includes some parts of QLD. Use of the private database was considered appropriate as it contained much of the data already incorporated in state-based databases and had the additional benefit of having a single surveyor providing consistency in identification and scoring of species.
Floristic data was extracted from plots in which Eucalyptus populnea was a dominant or co-dominant from CORVEG and Au-Au-003. From each database, plots were extracted where Eucalyptus populnea had >10% canopy cover. Within the Australian context, woodlands are defined as having a canopy cover of between 10-30% and thus at minimum the plots chosen for analysis had to have Eucalyptus populnea occupying a third of the canopy cover. Plots where less than six taxa were recorded within plots were removed. Plots where a misidentification with the closely related Eucalyptus brownii was made were also removed. Misidentification was determined by knowledge of the distribution and habitat preferences of the two species. Taxa not identified to species level were removed. The final dataset incorporated 455 plots (151 from CORVEG) and 1326 species (native and introduced) (see Figure 1 for distribution). IVC protocols specify using percentage cover of all species in all strata for the description of types (Jennings et al. 2009).
Within the CORVEG protocol, species cover can be recorded differentially across strata and there is a standard plot size of 50 × 10 m. This plot size has been shown to adequately capture species richness in Eucalypt woodlands in Queensland (Neldner and Butler 2008). Within QLD plots, species were recorded using percent cover down to fractional percentages (0.1%). Plots surveyed within NSW most commonly were recorded using a modified six-point Braun-Blanquet cover abundance method (Westhoff and van der Maarel 1980) or percent cover and are of a 20 × 20 m dimension. Later protocols within NSW were changed to record percent cover down to 1%. Differences in recognised nomenclature were noted between jurisdictions. In order to assist compatibility across datasets, the following protocols were used; a) Braun-Blanquet scores were rescored to the mid-percent of each category, b) all fractional percentage scores were increased to a minimum of 1%, c) cover scores between strata of the same taxa were summed, d) nomenclature was standardised.
Primer E (ver. 7.0.11; Quest Research Limited; Ivybridge, Devon, UK) was used for data exploration, as commonly utilised within the target jurisdictions (e.g. Hunter and Lechner 2017;Addicott et al. 2018;Hunter and Hunter 2020;Muldavin et al. 2021). Due to the size of the dataset, an initial analysis was performed using kR-CLUSTER to generate major groups based on lowest stress (R = 0.77188). From this analysis three groups were defined, which were visually assessed secondarily via projection in 3-D using non-metric multidimensional scaling ordination (nMDS). The three groups were then separated for within group analysis. Removing sparse species from a dataset is also recommended (McCune and Grace 2002; Clarke et al. 2014). To avoid removing species which may occur infrequently but contribute a large component to the cover, species occurring only once and contributing 1% to the total cover across each of the major groups were removed.
Each of the major groups was analysed using the Bray-Curtis similarity co-efficient after square root transformation of cover values, and agglomerative hierarchical clustering was applied using group averaging. The similarity was profile tested using similarity profile analysis (SIMPROF) permutation tests (9999 iterations) in order to assess a relevant statistically significant cut-off dissimilarity for defining vegetation types at the association level. 3-D ordinations were generated using nMDS and defined groups were further assessed based on group projection and associated ordination stress. Where plots were found to be outliers within the group analyses, they were removed and placed within analyses of other groups to assess if the original analyses had caused a misallocation. Occasionally individual plots were reallocated to different proposed associations based on nMDS 3-D projection and visual assessment of species occurrence if they were deemed to have been misallocated during initial clustering. Once preliminary associations were determined, all plots within each association were combined and their scores averaged to form a single sample. A further cluster, SIMPROF, and ordination was performed against all associations to determine higher level relatedness between groups.
Similarity percentage analysis (SIMPER) identifies the species that drive differences between selected types. SIMPER uses the Bray-Curtis similarity measure to identify positively and negatively diagnostic taxa across vegetation types. Taxa with combined high frequency and cover were also identified and listed for diagnostic purposes and type delineation.

Alignment within the IVC hierarchy
The IVC schema is based on a hierarchy of natural physiognomic-ecological types at the upper levels, physiognomic-biogeographic-floristic characteristics at the middle levels and floristic-ecological characteristics at the lower level (Faber-Langendoen et al. 2016). For incorporation into the IVC hierarchy, expert knowledge and qualitative application of the criteria is often used at upper level, whereas quantitative analysis of plot-based data is used to distinguish vegetation types at the mid to lower levels (Faber-Langendoen et al. 2014). For the current study, allocation of proposed vegetation types into the IVC hierarchy was achieved by combining the key to IVC formation classes and brief definitions provided by Faber- Langendoen et al. (2016), the criteria of the IVC (Jennings et al. 2009;Faber-Langendoen et al. 2014) and expert knowledge with reference to environmental datasets and existing sub-continental scale vegetation classification systems. Sources of expert knowledge include publications by other authors, including Beadle (1981), Beeston (1980), Keith and Tozer (2017) and Neldner et al. (2019). In applying the key to IVC formation classes (Faber-Langendoen et al. 2016), we included scleromorphic trees in the mesomorphic tree concept, as the descriptions of Forest and Woodland (C01) and Shrub and Herb Vegetation (C02) formations include scleromorphic growth forms.

Crosswalk of Plant Community Types and Regional Ecosystem types to associations
In order for the IVC to provide a link between classification systems used by different jurisdictions, we crosswalked existing PCTs from NSW and REs from QLD to the associations recognised in this study. To do this we did two things: (i) allocated REs to associations using the RE attribution in the metadata of CORVEG plots from QLD and allocated PCTs from NSW to associations based on the metadata held within BioNET (https://www.environment.nsw.gov.au/research/Visclassification.htm) (see Suppl. material 1), and (ii) listed REs and PCTs that would make up the E. populnea woodlands based on the descriptions given online (see Suppl. material 1). In addition to providing a cross-walk table between jurisdictional classifications, this enabled us to indicate REs and PCTs that are most likely to be part of the E. populnea woodlands. PCTs and REs are maintained on a searchable databases by the respective state governments (https://apps. des.QLD.gov.au/regional-ecosystems; https://www.environment.nsw.gov.au/research/Visclassification.htm; both accessed 27 June 2021). Eucalyptus populnea was used as a key search term to find all REs and PCTs where this species was used in describing types.

Alignment with the IVC hierarchy
The E. populnea woodlands range in height from 8-16 metres and from 12-38% in cover and are dominated by scleromorphic trees. This puts it into the IVC formation class 1. Forest and Woodland. The E. populnea woodlands are referred to as occurring in the subtropical and sub-humid climate zones of Australia (Fensham et al. 2017;Keith and Tozer 2017) and both climate zones are included in the Warm Temperate climatic zone of the IVC (Faber-Langendoen et al. 2016). We therefore suggest they be placed within the formation 1.B.1 Warm Temperate Forest and Woodland of the IVC. This is supported by Eucalypt woodlands of Australia having been specifically identified as part of the Temperate Forest and Woodlands formation by Faber-Langendoen et al. (2016). This contrasts with Keith and Tozer (2017)'s placement of subtropical woodlands in Savanna, which they have aligned with 1.A.1 Tropical dry forest and/or woodland and 2.A.1 Tropical lowland, grassland and savanna IVC formations. Although the Eucalypt woodlands of Australia have been referred to in formation level descriptions of the IVC types (Faber-Langendoen et al. 2014), there is currently no formal recognition of the eucalypt dominated woodlands at the division and lower levels of the IVC hierarchy within the Warm Temperate Forest and Woodlands formation. There is, however, informal recognition of the woodlands suggesting an Australian division of 1.B.1.La.4 Australian Warm Temperate Subhumid Woodland which would accommodate the E. populnea woodlands (Faber-Langendoen pers comm 2020). Although the IVC protocols recommend quantitative analyses to determine the mid-levels of the hierarchy, based on the criteria and descriptions given for the mid-level IVC types (Faber-Langendoen et al. 2014) we suggest the "Brigalow Forests and Associated Eucalypt Woodlands of Subtropical Eastern Australia" (Fensham et al. 2017) would be placed as a 'macrogroup' within this division. This 'macrogroup' is identified by the diagnostic species of Acacia harpophylla -Eucalyptus populnea -Eucalyptus crebra/melanophloia occurring on deep soils formed predominantly on sedimentary rocks on the western side of the Great Dividing Range of eastern Australia. Within this the E. populnea woodlands match the criteria of a 'group' , in having a limited set of diagnostic species Table 1. Circumscription of Poplar Box Woodlands (Eucalyptus populnea) of New South Wales and Queensland within eastern Australia. Descriptions of the 3 alliances and 23 associations include positive and negative diagnostic and negatively associated species, common dominant taxa (based on cumulative frequency and cover) and notes for each unit. Positive diagnostic species are listed in order of decreasing contribution to group identity. Negative diagnostic taxa are those not found within plots and should not occur or only occasionally within the defined type.s Common taxa are listed in decreasing order of cumulative frequency and cover within each identified group. Non-native taxa are indicated by '*'.

Hierarchy
Positive diagnostic (SIMPER) Negative diagnostic (SIMPER) (E. populnea, Callitris glaucophylla and Acacia aneura), a diagnostic growth form (trees) with broadly similar composition, and a distribution that reflects a regional mesoclimate and soil characteristics (sub-humid / subtropical climate and largely on soils with sodic sub-soils; Fensham et al. 2017). We propose that the major vegetation types within this E. populnea woodlands group are alliances and describe the vegetation types within those alliances as associations. Confirming these proposed mid-levels of the hierarchy using plot-based data remains to be done.

Vegetation types
Analysis of our data of 455 plots in which Eucalyptus populnea was a major component of the canopy enabled us to define three interim alliances and 23 associations. We propose the types as interim and refrain from adding proper formal and colloquial names that are generally provided for alliances and associations within the IVC as we would prefer standardised naming to be provided based on a wider decision-making process than the authors alone. Table 1 highlights for each community type the positive and negative diagnostic taxa, along with their most common taxa (i.e., those with high summed cover) (Suppl. material 2 and 3). The Eucalyptus populnea -Eremophila mitchellii -Carissa spinarum / Heteropogon contortus -Eragrostis lacunaria alliance (Figure 2), primarily of the Brigalow Belt (IBRA7; Thackway and Cresswell 1995), was prominent in QLD and incorporated most of the plots from this state. It was generally widespread across the whole geographic range of Eucalyptus populnea and contains nine associations. The Eucalyptus populnea -Callitris glaucophylla -Casuarina spp. / Geijera parviflora -Eremophila mitchellii alliance (Figure 3) contains seven associations, and it was primarily restricted to southern QLD, though also found in the most southern locations sampled within the range of Eucalyptus populnea. This alliance was commonly found within the Brigalow Belt South and the Darling Riverine Plains Bioregions and thus had general south easterly distribution (IBRA7; Thackway and Cresswell 1995). The Eucalyptus populnea -Acacia aneura -Eucalyptus intertexta / Enteropogon acicularis -Austrostipa verticillata alliance (Figure 4) also includes seven associations and while occurring across the entire geographic range sampled, was primarily found in the most western semi-arid districts of southwestern QLD and northwestern NSW and the only alliance distributed in these areas (Figure 4).
Although the listing advice for the endangered Poplar Box Grassy Woodlands on Alluvial Plains only includes the six REs 11.3. 2, 11.3.17, 11.4.7, 11.4.12, 12.3.10, and the four PCTs 56, 87, 101, and 244 (https://www.environment.gov.au/cgi-bin/sprat/public/publicshowcommunity. pl?id=141&status=Endangered), there are fifteen PCTs and 34 REs that have Eucalyptus populnea as a diagnostic species within the title or detailed descriptions of the type (Suppl. material 1). All of these types were found to correspond to our associations directly or in part within our classification. Thus, all described Eucalyptus populnea dominant PCTs or REs were sampled and incorporated within our analyses (Table 2). However, a few of our defined associations had no direct correlates and thus could not be placed within the current state-based classifications (association 19, 22 and 23; Tables 1 and 2) and thus may require new RE and PCT designations. Many of the defined PCTs had a 1:1 or a 2:1 relationship with our defined types. Only association 20 appeared to incorporate multiple PCTs (6 in total) suggesting this PCT maybe overly split at the association level. There was less correlation found between the NSW classes and formations compared to that found for PCTs and there is little direct relationship between REs and our proposed types, with most associations having multiple REs (up to 13), as potentially synonymous. Additionally, REs were found to occur across multiple associations. RE 11.3.2 in particular was found to be attributed to nearly half of our associations (9 in total) and to all three alliances, and it is listed as an assemblage that typifies the listed endangered Poplar Box Grassy Woodlands on Alluvial Plains ( Table 2). Thirteen of the associations defined here are synonymous with the nine REs and PCTs contained in the listing advice for the endangered Poplar Box Woodlands. Based on our analysis the listing of the endangered Poplar Box Grassy Woodland on Alluvial Plains does not correspond to any particular level of a classification hierarchy and incorporates multiple associations and crosses alliances but not in a consistently applicable way. We also found that at the RE and PCT diagnostic level some areas that could be included or excluded as part of the endangered community in one state would not in the other if based purely on the listed REs considered synonymous. Thus, from a floristic perspective there is a lack congruence within the current definition of the listed endangered community and plot-based analyses but also between jurisdictions if using PCTs and REs. Basing listed communities on plot-based classifications could present a better approach and allow for greater cross jurisdictional alignment when categorising what is and isn't included in the definition on ground.
No equivalent in QLD munities at local, continental and global levels as opposed to classification systems which rely on correlative environmental gradients or cross-walked map-based systems (ESCAVI 2003;Keith and Tozer 2017;Luxton et al. 2021). The congruence between our associations and the types in existing classifications varied between the different jurisdictions. Most PCTs types (NSW) were found to form a closer relationship with our proposed associations than REs (QLD). This may not be surprising as the methods used to define PCTs were either based on previous published and unpublished un-supervised analyses or, where fully supervised means were used, types were defined based on floristic composition and dominance, whereas the REs in the bioregions included in this study have been derived by fully supervised means and incorporate historical units derived from disparate studies. There are some notable exceptions within the PCTs, in particular those generally listed for the Cobar Peneplain Bioregion, where association 20 was potentially synonymous with six PCTs suggesting these PCTs are over-split at the association level. The lack of correlation on the Cobar Peneplain may be due to previous limited plot data within this bioregion. A lack of congruence was more apparent between our types and the NSW class and formation types. The situation was much more complicated for REs, where we also found little congruence between our associations and REs. Under the RE classification system, similar plant associations are divided by geomorphological categories, reflecting the assumption that there will be different biodiversity values associated with different substrates which are not necessarily reflected in plant diversity (Sattler and Williams 1995). This means that ideally, there should not be plots from one RE occurring in multiple associations, such as found in this study; for example, all plots attributed to RE 11.3.2 should match only one association, rather than nine (Table 2). When this mismatch does occur, it is likely reflecting the qualitative nature of the current classification of REs within each bioregion of QLD. The lack of hierarchical quantitative delineation of the NSW classes and formations and their relationship to PCTs is also likely to be the reason for their lack of congruence between our alliances and associations. One use of the results of this study, and future associations recognised under the IVC hierarchy, is to provide feedback into the individual jurisdictional classification systems to improve the delineations of individual vegetation types. Conversely, in identifying a possible new division, macrogroup, group, alliances, and associations within the IVC, analysis such as in this study feed back into the flexible design of the IVC, modifying it to include new levels in the hierarchy which accurately reflect the diversity of vegetation globally.
Under the EPBC Act 1999 an ecological community is defined as "The extent in nature in the Australian jurisdiction of an assemblage of native species that inhabits a particular area in nature" and is defined by the co-occurrence and interactions of species with overlapping distributions (Threatened Species Scientific Committee 2017). Furthermore, listing guidelines state that threatened communities should be defined based on classification of (dis-) similarities between vegetation types preferably based on composition (Threatened Species Scientific Committee 2017). Thus, the intent is to include in the classification vegetation types that are defined by composition. Our analysis indicates that the endangered community listing is largely based on a landscape element with an emphasis on alluvial plains, excluding types that were not predominantly grassy, reflected in its title and the REs and PCTs characterising this landscape element and structural type, rather than plant associations, to which it bears little relationship. It thus cannot be placed directly within a hierarchical classification scheme. Although low lying floodplain landscapes are commonly the most highly impacted within the Australian landscape, the emphasis on this landscape element over floristic coherence raises a number of important questions regarding conservation targets, with consideration of the whole distribution of the plant association required rather than one particular element of its distribution. Concentration on one landscape element does not help to increase our understanding of these communities or their interrelationships. Furthermore, restriction to a predominantly grassy understorey can be complicated in systems where this is transitory in nature due to natural climatic variation, disturbances both natural and human induced (Hunter 2021b;Saunders et al. 2021). It is possible that consideration of the threatened community at the alliance level may provide a more useful level of protection for the Poplar Box Grassy Woodlands than disparate sections of numerous associations.
Our relationship of synonymous types (Table 2) with the associations in this study highlights an important function of using a consistent national classification system, such as one based on the EcoVeg approach and integrated with the IVC. Adherence to the rules and processes of quantitative classification systems such as the IVC provides a clear and repeatable process when defining vegetation units and also allows for interrelationships to be recognised across jurisdictions. This is obscured within both the current NSW and QLD systems from a purely floristic-ecological classification perspective, and compounded when comparing across jurisdictions. For instance, our comparison table shows that the RE types 11.3.2 and 11.3.17, which are included in the definition of the listed endangered community description, align in part with PCT 35 yet this PCT is not one listed as defining the endangered community. The strength of using a national classification system based on quantitative plotbased analysis is in showing the relationships between floristic assemblages across jurisdictions. These may not show up in classification systems that are mapping oriented and not quantitatively based, such as the National Vegetation Inventory System, which is the current Australian national classification system (ESCVAI 2003). The strength of the IVC is that it also puts the individual threatened ecological community in a global perspective. If many of the plant associations within any given level of the IVC are listed as threatened communities it helps pro-vide a continental and global perspective for communities within any level of the hierarchy.

Conclusion
This investigation highlights how a rigorous rule-based hierarchical classification system, where the lower schematic levels are based on plot-based vegetation analyses of floristic and ecological data, should underpin our understanding of Australian vegetation. Such processes allow for a better understanding of distribution, interrelatedness, rarity, and threat of ecological communities at lower levels and inform mid to broad levels of vegetation pattern. Our study also suggests that state-based systems should not, in and of themselves, be the only basis for the listing of endangered ecological communities. Lack of clear guidelines and a similar process applied across state and territory borders only adds further confusion leaving practitioners to rely on intuition and opinion. Using a classification system such as the IVC allows an understanding of the threats to, and status of, communities both at local and regional levels and within a continental and global perspective.

Data availability
The NSW data is contained within Version 3 of sPlot (https://www.idiv.de/?id=176&L=0) (Bruelheide et al. 2019) and is listed on GIVD as AU-AU-003 (https://www. givd.info/databases.xhtml). The Queensland data is contained within the Queensland government QBEIS database and is publicly available on request.

Author contributions
JTH collected all NSW plot data, entered all of NSW data, analysed the data and co-wrote the manuscript. EA contributed equally to writing of the manuscript and in particular the incorporation of the IVC hierarchy to the results presented.