Report to the Australian Department of the Environment 2015
Review of Australia’s Major Vegetation classification and descriptions
Prepared by the Centre for Ecosystem Science (UNSW)
in association with Australian Museum Consulting (AMC)
for the Department of the Environment
David A. Keith
(Professor of Botany, UNSW)
Belinda J. Pellow
(Senior Botanist, AMC)
30 June 2015
Citation: Keith, D. A. & Pellow, B. J. (2015). Review of Australia’s Major Vegetation classification and descriptions. Centre for Ecosystem Science, UNSW, Sydney.
Cover photos: D. Keith
Introduction The Australian Government, in collaboration with state and territory jurisdictions, maintains the National Vegetation Information System (NVIS), a repository for spatial data and associated thematic information on Australian native vegetation (NLWRA 2001; ESCVI 2003). NVIS includes a hierarchical database with some 18000 records of mapped vegetation types from all states and territories within Australia. The NVIS information hierarchy was augmented with an additional attribution of ‘Major Vegetation Groups’ (MVGs), enabling a high-level continental-scale synthesis to support the Australian Government’s reporting responsibilities under international agreements such as the Montreal and Kyoto Protocols (Montreal Process Implementation Group for Australia and National Forest Inventory Steering Committee 2013), and to inform the first national audit of native vegetation (NLWRA 2001).
Major Vegetation Groups are composed of a large number of vegetation types with similar structural and/or floristic characteristics (NLWRA 2001). Subsequent revisions of the MVG classification have introduced Major Vegetation Subgroups (MVSs), enabling a finer level of characterisation in a continental context. Due to the complex relationships between structural, floristic and ecological attributes of vegetation, MVSs are not strictly nested within MVGs (i.e. individual MVSs may be represented in one or more MVGs). The current classification (NVIS 4.1) includes 27 MVGs (with five additional land cover categories based on FAO classification) and 85 MVSs (Department of the Environment and Water Resources 2007).
To support improved delivery of the Australian Government's Natural Resource Management programmes, the Environmental Resources Information Network (ERIN), the custodian of the NVIS, engaged the Centre for Ecosystem Science (CES) at the University of NSW to update and improve the descriptive fact sheets for Major Vegetation Groups and to undertake a review of the MVG/MVS classification. The aims of this project were to:
Review and update of MVG Fact Sheets, including new fact sheets for MVGs 31 and 32; and
Review all MVSs to determine their current applicability.
This report describes the outcomes of that work. It first describes the strategic review of MVSs (aim 2) in the context of their relationships to MVGs within the Major Vegetation classification framework. It then describes the detailed revisions to MVG fact sheets (aim 1).
Methods Strategic Review of Major Vegetation Classification A brief email survey was undertaken to obtain an overview of the current usage, demands and expectations of the Major Vegetation classification. The survey included: i) all state and territory members of the NVIS network; ii) a sample of vegetation scientists; iii) a sample of non-government users. Participants were asked for responses to three questions:
What are you primary needs for a classification and national-scale map of Major Vegetation Groups (MVGs) and Major Vegetation Subgroups (MVS)?
Do you think the MVG & MVS data currently served through NVIS would meet those needs more effectively by:
a. Revising the descriptions of the units?
b. Adjusting the circumscriptions of some units?
c. Revising interpretation of the map data for some units?
If you answered yes to 2a-2c, please write a couple of sentences to explain the kind of revisions most needed to improve MVGs and MVSs.
Responses were collated and summarised to interpret the primary needs for a spatially explicit national-scale vegetation classification. Relevant scientific literature was reviewed to identify the features of a classification required to meet these needs. This, together with improvements elicited from NVIS users in the email survey, was used to shape recommendations on future development of the Major Vegetation classification framework to meet current and future needs.
Revision of Major Vegetation Subgroup (MVS) units In consultation with ERIN data managers, each of the 85 MVS units currently within NVIS 4.1 were reviewed to evaluate the degree to which they represented major ecological variation within and between MVGs. Where representation of ecological relationships could be improved, alternative configurations of MVSs within MVGs were prepared for discussion with ERIN data managers. When classifications were agreed, brief descriptions were compiled using information from regional literature (Beadle 1981; Groves 1994; Keith 2004; Harris & Kitchener 2005; Victorian Department of Sustainability and Environment 2004; Beard et al. 2013; Neldner et al. 2014) in combination with records from the NVIS database. Alternative names were proposed for MVSs, where these could transparently reflect their key characteristics.
Preparation of diagnostic dichotomous keys to Major Vegetation Subgroups (MVSs) Dichotomous keys were prepared to assist the diagnosis of MVSs within each MVG. These were developed by expanding and adjusting draft keys prepared by ERIN. Couplets within the keys were based on features that permit the clearest possible discrimination between MVS units within an MVG. In most cases, the key enumerates multiple features to strengthen diagnostic power. It was not possible to test the keys in the time available for the project.
Update of descriptions for Major Vegetation Groups Existing fact sheets for 23 MVGs (and other cover types) were revised and two new fact sheets were prepared for MVGs 31 and 32, respectively. This was based on a comprehensive review and synthesis of relevant descriptive data including:
Available descriptive data for NVIS levels 5 and 6 relating to the relevant MVGs;
Current map data served by ERIN for the relevant MVGs;
Relevant literature pertaining to the relevant MVGs throughout their distribution.
Photographs illustrating key features of each MVG.
During synthesis, all relevant descriptive data were critically evaluated, and examined for consistency between sources and with mapped distributions. This enabled identification and interpretation of key descriptive features of MVGs and MVSs. As far as possible, variability within MVGs was documented by incorporating brief descriptions of MVSs. Information on vegetation change, key values and management issues was also extracted from the literature for inclusion in the updated fact sheets. To improve transparency, sources of information were attributed with in-text citations. Additional photographs to illustrate key features of MVGs were sought to supplement or replace existing photographs to improve illustration of key features. These were licenced for creative commons release.
In consultation with ERIN data managers, the format of the fact sheets was revised to enable explicit reference to key features of MVGs. The fact sheets were formatted in the following sections.
Results and Discussion Identifying needs: applications of the Major Vegetation Classification
Seven of the nine government agencies responded to the survey, as well as six scientists and three non-government organisations. Three main groups of uses were identified from the responses (Table 1). The documented uses are unlikely to be exhaustive, but provide an informative overview of applications by many essentially independent user groups.
Table 1. Synopsis of uses of Major Vegetation classification (MVGs and MVSs) in Australia. Numbers in parentheses represent the numbers of users that identified each application. Bottom row shows total number of users for the three broad groups of application. The numbers exclude two of the nine government jurisdictions that did not respond to the survey and two that stated that they did not use the national classification.
National Forest Inventory time series for Montreal Protocol (1)
conservation investment planning (3)
survey stratification (2)
vegetation component of carbon accounting for Kyoto Protocol (2)
representation in protected areas (3)
cross-border comparison and national context (2)
status evaluation for specific vegetation types (6)
grouping of fine-scale vegetation units (3)
mapping of threatened ecological communities (1)
global vegetation mapping (1)
distribution modelling (2)
fire management (1)
threatened species habitat characterisation (3)
grant application reporting (1)
heritage area assessments (1)
state of environment reporting and related audits for biodiversity (1)
Australia has reporting obligations under a number of international agreements (e.g. Montreal Process; Kyoto Protocol), which are met by Australian Government agencies. Two of these involve reporting on the national extent of forest at regular intervals (Montreal Process Implementation Group for Australia and National Forest Inventory Steering Committee 2013; Commonwealth of Australia 2015), for which the map base for Major Vegetation Groups forms the primary data source (Table 1).
There is a diverse and growing range of applications of the Major Vegetation classification for biodiversity assessments (Table 1). These include tracking the remaining area of native vegetation, design of protected area networks, mapping of potential habitat for threatened species and communities, and decision making on conservation investments. Conservation assessment was the application of the Major Vegetation classification that had the largest and most diverse group of users, including the Commonwealth, states and territories, research scientists and non-government conservation and land management organisations.
The third group of applications involves use of the Major Vegetation classification as a comparative framework to stratify surveys and research sites, scaling up of fine-resolution vegetation units, comparing vegetation across jurisdictional borders and contributions to global vegetation mapping (Table 1). These applications are mainly implemented by government agencies, but other users include the Terrestrial Ecosystem Research Network.
The earliest applications of the Major Vegetation classification were to meet Commonwealth reporting responsibilities, particularly for the Montreal Process and national land and water resources audit (NLWRA 2001; Montreal Process Implementation Group for Australia and National Forest Inventory Steering Committee 2013). However, the current range of applications reflects a growth and diversification of needs over the past decade.
Fit for purpose: classification features to meet needs The diversity of needs for a national vegetation classification and map base creates difficulties in the design of a general system to meet all needs. Some applications require the classification to predict particular vegetation properties, while other applications require quite different features to be represented by the units of a classification. For example, effective carbon accounting over large geographic domains is predicated on map units that are informative about vegetation biomass. Use of vegetation maps in fire management requires the units to be informative about fuel type and accumulation rates. Forest reporting requires the units to conform with the nationally agreed definition of ‘forest’ (viz. dominated by trees having usually a single stem and a mature or potentially mature stand height exceeding 2 metres and with existing or potential crown cover of overstorey strata about equal to or greater than 20 per cent) (Montreal Process Implementation Group for Australia and National Forest Inventory Steering Committee 2013). At a broad level, these applications are best served by a classification founded on structural features reflecting the height and density of plant biomass across different vertical strata.
In contrast, the wide range of conservation applications and most of the comparative applications (Table 1) require the units to represent different groups of biota – they must perform well as ‘biodiversity surrogates’ (Pressey 2004; Hermoso et al. 2013). This is essential to draw credible inferences about representation of biodiversity in protected areas, losses of biodiversity related to change in vegetation extent; the distribution of threated species and ecological communities; and is therefore key to informing sound decisions that seek effective returns on conservation investments. To perform well on these tasks, a vegetation classification needs to be based on ecological features. These include factors that influence the distribution of biota (environmental gradients and ecological processes that define species niches), as well as the composition of the biota itself (Keith 2009; Hermoso et al. 2013; Keith et al. 2015).
Thus for one group of applications, the structural form (specifically biomass) of vegetation is key to outcomes, irrespective of how that structure is produced, while for another group of uses the ecological identity of biotic elements is key to outcomes. In nature, structural features of vegetation sometimes coincide with ecological features, sometimes they do not. Consequently, there may be conflicting goals in the design of a unifying classification system. The emphasis on characteristics of vegetation structure and taxonomy as a basis for defining Major Vegetation Groups and Subgroups reflects early priority needs to support reporting on forests. Consequently, this appears to limit the performance on some conservation applications, the demand for which has increased in more recent years.
One of the key issues for biodiversity applications, such as state of environment reporting, is averaging effects. This applies to cases where two or more ecologically contrasting units with different biota have similar vegetation structure, and are thus lumped into a single structurally defined category. If the ecological units are exposed to different types of resource use and threatening processes, they may have very different status in terms of conservation metrics. However, the true status is masked when the ecological units are lumped into a single structural category because the extreme metric values for each ecological unit are averaged to moderate values for the combined structural category. Box 1 illustrates an example of the effect.
Box 1. Averaging effects: sub optional fit-for-purpose performance for contrasting applications.
In Australia, eucalypt woodlands are distributed extensively in a belt around the arid core of the continent. These vary in structure throughout their range, particularly with regard to canopy cover. In the NVIS, this structural variation is recognised by classifying eucalypt woodlands into two Major Vegetation Groups: MVG 5 (Eucalypt Woodlands) and MVG 11 (Eucalypt Open Woodlands). These categories are useful for reporting on forest inventories and related applications because MVG 5 meets the agree definition of ‘forest’, whereas MVG 11 does not Montreal Process Implementation Group for Australia and National Forest Inventory Steering Committee 2013).
As well as varying structurally, Australian eucalypt woodlands vary greatly in their biodiversity. For example there is a major ecological contrast between tropical eucalypt woodlands, which share a number of ecological features with other savannah biomes in Africa, southeast Asia and South America, and temperate eucalypt woodlands, which have no close ecological analogues on other land masses. Moreover, tropical and temperate woodlands share very few species of plants and animals in common. Consequently, any effective conservation strategy should aim to conserve representative and viable samples of both tropical and temperate eucalypt woodlands and track their status independently (e.g. in state-of-environment reporting).
The tropical eucalypt woodlands are one of the last largely intact tracts of woodland on earth, having so-far avoided effects of high-density human population growth and exploitation for broad-scale agriculture. In contrast, the status of temperate eucalypt woodlands is rather dire due to their exposure to broad-scale agricultural development over more than two centuries. When classified ecologically into two broad groups that represent major contrasts in species composition, the status of tropical and temperate woodlands is highly contrasting: more than 95% of tropical woodlands remaining intact (data from Fox et al. 2001); while less than 15% remains of temperate woodlands within the wheatbelts of southeastern and southwestern Australia (data from Lunt and Bennett 2000; Keith 2004; Beard et al. 2013). When classified into broad structural categories, the equivalent metrics are 66% remaining for eucalypt woodland (MVG 5) and 92% remaining for open eucalypt woodland (MVG 11) (data from NVIS v4.1; see also NLWRA 2001). The % remaining statistics derived from the structural classification and map therefore mask the greatly contrasting conservation status of biodiversity in tropical and temperate eucalypt woodlands (see figure). In essence, the large initial and remaining area of tropical woodlands masks the precipitous trends in the extent of temperate woodlands.
Similar artefacts derived from averaging effects are evident when these types of analyses are applied to other vegetation types such as eucalypt forests, acacia forests and woodlands, grasslands and various shrublands. Conversely, averaging effects would also reduce the performance of forest reporting if a structural classification was replaced by an inappropriate ecological classification that did not align with the agreed definition of forest.
A number of users identified difficulty in extracting distributional data for ecological units such as savanna, temperate woodlands, grasslands, alpine vegetation and wetlands that correspond poorly with structurally defined groups or where it is practically difficult to interpret their structural features in a consistent way. In many cases, this can be resolved by reference to Level 5 or 6 of the NVIS information hierarchy, as many of the source vegetation units that are aggregated into MVGs and MVSs are ecologically homogeneous to the degree required to support inferences about biodiversity. However, interpretation at this level for national (or subnational) synthesis is extremely onerous given the large number of records (>18000) within the NVIS and the scope of accessible descriptive data, which is limited to between three and five dominant taxa in each vertical stratum.
The need to improve predictive performance of broad classification units is reflected in a range of suggested improvements to the classification elicited from NVIS users. In summary these include:
Restructuring the classification towards an exclusive hierarchical relationship between MVGs and MVSs. Ideally any given MVS should be uniquely nested within a single MVG.
Evaluate and improve alignment between MVGs/MVSs and broad units currently adopted by states/territories where these exist (e.g. Keith 2004; Beard et al. 2013; Neldner et al. 2014).
Increasing the number of MVSs to c. 100 by splitting the most heterogeneous units, and providing more detailed descriptions. While the number of MVGs (25) provide a useful level of resolution for national overview, some users felt more resolution at the MVS level would benefit their applications.
Adjusting MVSs and adding more detailed descriptions to capture major structural differences, major compositional differences and major landform differences.
Rationalising ‘Other’ groups, which are currently circumscribed as leftovers from more clearly defined groups (MVGs 10, 17, 21, 31 and numerous MVSs). The structural and ecological heterogeneity of these groupings was seen as problematic. Some could be resolved by merging elements into other groups.
Improving consistency or better-accommodating inconsistencies between data contributors in attribution of NVIS levels V and VI through the interpretation of concepts such as dominance, regrowth, grassland with scattered trees cf. open woodland, heathland and mallee cf. mallee heath, etc.
Disentangling ‘regrowth’ from the concept of ‘MVG’ by revising the NVIS attribution such that any MVG can exist in a range of regrowth states.
Recent improvements to edge matching across state borders were widely acknowledged. Some further work on this is needed to resolve remaining anomalies.
Adjusting MVGs/MVSs where necessary to enable simple cross-walk relationships between Australia’s national vegetation classification and the international vegetation classification system currently under development (Faber-Langendoen et al. 2014).
Adjusting circumscriptions of particular MVGs and MVSs to make them more homogeneous and consistent (see details below).
Maintaining and strengthening close working relationships with data contributors in state agencies to ensure currency and consistency.
Linking the database entries and descriptions of MVGs and NVSs to the Australian Plant Name Index to ensure nomenclature is up to date.
Streamlining the process of assigning NVIS Level 5/6 records to MVGs and MVSs.
Proposed framework for a national Major Vegetation classification A key challenge for future development of the national vegetation classification system is how to improve its ability to meet requirements for biodiversity applications without compromising its performance or consistency for reporting changes in forest inventory or carbon accounting mandated by international agreements (Box 1). Temporal consistency is important, for example in forest mapping, to ensure that recorded trends in forest extent reflect genuine change, rather than changes in methods of forest classification and mapping. This places significant constraints on modifications to existing Major Vegetation Groups to improve their utility for biodiversity applications.
One strategic solution to this trade-off is to adopt an ecological concept for Major Vegetation Subgroups that is at a resolution fine enough for them to display a greater degree of homogeneity in both compositional and structural properties. Such an approach is consistent with many of the suggestions elicited from NVIS users and would facilitate the establishment of major ecological groupings that could sit alongside the structurally based MVGs (Figure 1).