Introduction

Vegetation management is an important mechanism for sustaining and securing natural resource benefits, including water, food and fibre, carbon sequestration (Maltby and others 1999; Abel and others 2003; Millennium Assessment 2005), soil health (McKenzie and others 2004), and biodiversity (Hobbs and Saunders 1993). The history of European settlement in Australia, since first contact with indigenous peoples in the late 1700s, has seen large areas of native vegetation either modified, replaced, or removed to meet changing social-ecological needs and aspirations (Walker and others 2006). National reporting requirements for native vegetation oblige stakeholders to monitor and report information on extent (i.e., its coverage across the landscape), type (i.e., associations), and condition (i.e., anthropogenic effects on vegetation extent and type) (NLWRA 2007).

Increasing recognition of the strong linkage between vegetation management, vegetation modification and natural resource outcomes has generated new demands for vegetation information—including information that adequately describes anthropogenic effects on vegetation. Information is required in a form that meaningfully translates a wide variety of vegetation, land management, and ecological data into terms that can be applied to natural resource management policy, program development, and reporting.

Our review of the literature on the uses and values of native vegetation shows a consistent need to describe and map the relative degree of anthropogenic modification of native vegetation against either an explicit or implicit benchmark or reference condition state. This is set to be its condition at the time of European settlement. We discuss the rationale for using a pre-European benchmark for measuring vegetation condition below. At the national level, such reference condition states have been, and continue to be, widely recognized in national policies and programs including the National Wilderness Inventory (Lesslie and Maslen 1995), the National Forest Policy Strategy (Commonwealth of Australia 1992), the Native Vegetation Management Framework (Commonwealth of Australia 2004), the National Monitoring and Evaluation Framework (NLWRA 2005), and for listing of nationally threatened ecological communities (DEWHA 2006). In addition to these national requirements, numerous state and territory vegetation programs have also defined similar condition state benchmarks (e.g., Oliver and others 2002; Parkes and others 2003).

While there have been considerable advances in methods for describing, classifying, and mapping native condition states based on survey, inventory, and modeling (EMR 2006), there is no established national framework that can guide the collection and/or compilation of information on condition states. Having said that, Australian researchers are nevertheless leaders in this area of applied ecology (EMR 2006).

This article describes the application of the Vegetation Assets, States and Transitions (VAST) framework as a consistent national framework to translate and compile existing mapped information on the modification of native vegetation. We discuss the correspondence between these compiled VAST datasets at national and regional scales and describe their relevance for natural resource policy and planning.

VAST classifies vegetation associations by degree of anthropogenic modification as a series of condition states, from a natural condition state through to total removal. For the purposes of this framework each dataset that describes vegetation communities must have either an explicit or implicit benchmark or reference state for native vegetation, which is set to be its condition at the time of European settlement of Australia. Condition states in the VAST framework are defined by breakpoints in vegetation composition, structure, and regenerative capacity resulting from land use and land management practices in relation to the identified benchmark condition state.

Approaches to Describing Vegetation Extent and Type

Vegetation classification entails the grouping of vegetation characteristics according to defined criteria. Vegetation classification systems are usually framed around extent, structure, taxonomic composition, and functional attributes and are tailored to address particular sectoral interests. For example, forest industry vegetation classifications generally focus on structural and functional attributes relevant to timber production (e.g., tree height, stem density, or age structure). Agriculture and pastoral interests often focus on structural, taxonomic composition, and functional attributes associated with marketing selected crop types (e.g., tonnes per hectare, cultivars) and pasture productivity (e.g., palatability of species and tonnes per hectare).

Vegetation information requirements for natural resource management (including water, salinity, carbon, and biodiversity management) are much broader. Additional requirements include support for assessments of multiple outcomes, trade-offs, and development scenarios and the monitoring of change and performance reporting (McIntyre and others 2002; Thackway and others 2006). In this context, vegetation is conventionally described and classified in terms of its extent, structural arrangement (height and spacing), and floristics (taxonomic grouping) (AUSLIG 1990; ESCAVI 2003). The National Vegetation Information System (NVIS) is the agreed national protocol for surveying, classifying, and mapping vegetation extent and type (ESCAVI 2003; Hnatiuk and others 2008).

Approaches to Describing Vegetation Modification

In addition to extent, structure, taxonomic composition, and functional attributes, information is also required about vegetation modification, or the degree of vegetation change measured against a putative base-line or reference condition (Hobbs 1994; Hobbs and Norton 1996; McIntyre and Hobbs 1999; Hobbs and Hopkins 1990; Hnatiuk and others 2008; Thackway and others 2006). This analytical approach to assessing vegetation condition states is required in order to understand and address patterns, levels and, effects of management intervention on vegetation, assess natural resource condition, and to maintain and/or restore habitat (McIntyre and Hobbs 1999; Hobbs and Hopkins 1990; Parkes and others 2003). It is also needed to support the development of sustainable production systems, and inform policy and program discussions and debate on options for trade-offs and land management practices.

Several broad methods for assessing vegetation modification have been developed in Australia and elsewhere as part of various research projects, but they are applied at different scales and have not been consistently implemented across jurisdictions:

  1. 1.

    Analysis of vegetation resilience and regenerative capacity in applications such as ecological restoration—(e.g., Hobbs and Hopkins 1990; Tongway and Hindley 1995; Walker and others 2006),

  2. 2.

    Analysis of vegetation landscape processes such as succession and fragmentation, (e.g., Forman and Godron 1986; Forman 1977; McIntyre and Hobbs 1999, 2000; Lesslie 1997, 2001),

  3. 3.

    Scoring, ranking, and indices derived from key vegetation structural, floristic and/or functional attributes associated with native vegetation condition states (Keighery 1994; Parkes and others 2003; Newell and others 2003; Oliver and others 2002), and

  4. 4.

    State and transition models (e.g., Westoby and others 1989; Phelps and Bosch 2002; Macleod and others 1993; Hobbs and Norton 1996; Yates and Hobbs 1997).

While these methods contribute important insights into the condition states of native vegetation communities, there is value in a conceptual framework that can integrate the ecological principles from these approaches into an ordinal classification.

The classical approach to the analysis of modification of natural vegetation generally distinguishes natural cover (native species dominant) from nonnatural (built-up areas and agricultural cover) (Forman 1977). McIntyre and Hobbs (2000) argue that such a one-dimensional binary approach is too limiting, contending rather that landscapes vary on two dimensions, i.e., destruction and modification. While McIntyre and Hobbs (2000) suggest four condition or habitat modification states (i.e., unmodified, modified, highly modified, and destroyed), they do not provide diagnostic criteria to enable the assessment and classification of these condition states or their transitions.

Methods for describing anthropogenic modification of vegetation should address the identification of a benchmark or reference states and transitions from that reference states (Hnatiuk and others 2008). Relevant issues in this context include choice of (1) an appropriate conceptual perspective for describing vegetation, (2) ambiguity in determining whether effects represent a fundamental transition in state or a change within the normal limits of persistence and development, and (3) uncertainty regarding the place and effects of humans in the environment (Lesslie 1997).

Values and Perspectives on Vegetation

Description of characteristics such as “condition,” “modification,” or “integrity” for complex natural systems such as vegetation are often ambiguous or contradictory because important differences in interpretation can arise as a result of differences in conceptual perspective and the use of limited observation sets (Levin 1992; O’Neill and others 1986).

A process-functional view of vegetation, for instance, may focus on the effect of the removal of dominant or key species in terms of the alteration of material and energy flows and process rates but ignore essential features of taxonomic composition. Alternatively, a structural-compositional focus may account for changes in taxonomic composition, species abundance and community structure, but fail to account for functional substitution and compensation effects. Thus, the replacement of native forest vegetation with a plantation forest of species alien to a locality may be regarded as a major modification from a population-community standpoint while from an ecological function perspective, using measures of nutrient cycling or standing biomass, such a change may be not be regarded as a major modification.

Alternative perspectives on vegetation condition states involving aesthetic, ethical or economic criteria may similarly lead to differing conclusions concerning vegetation condition (King 1993). For example, the condition states of forests disturbed by fire, disease, or logging may be unchanged from an economic and process standpoint, but be highly changed from a biotic, aesthetic, and ethical perspective (Lesslie 1997; Thackway and others 2006). In view of these issues, Thackway and others (2006) argue that it is vital that any definition of vegetation condition include:

  1. 1.

    a statement of perspective and values to which the condition applies,

  2. 2.

    consideration of the long-term stability of the vegetation under current management conditions,

  3. 3.

    key attributes of the vegetation, such as structure (e.g., open forest, grassland) and species of plants present,

  4. 4.

    key attributes of the environment (e.g., soil, water)

  5. 5.

    key attributes that may form the basis of particular perspectives (e.g., social, cultural or spiritual perspectives), and

  6. 6.

    a clearly defined or documented method of assessing the attributes, including benchmarks and reference sites.

On this basis we propose that vegetation modification should be described in structural-compositional terms on the basis that, regardless of conceptual standpoint or interpretation, vegetation change does not occur without a change in structure and composition.

Vegetation Dynamics and Regenerative Capacity

Highly modified vegetation, such as crops and plantation forests may be obvious and readily identifiable using available land use and land cover information. However, describing the modification of natural vegetation raises difficulties akin to the problem of describing “integrity” or “resilience”—concepts which involve the description of adaptive capacity, characteristic composition, and functional organization (Angermeier and Karr 1994; Frey 1975; Holling 1973, 1996; Walker and others 2006).

For native vegetation, the notion of a “static” reference condition, based on the Clementsian (1936) concept that communities develop in a predictable fashion toward a specified end point or climax, is unrealistic. Even if there is unambiguous evidence of a change in vegetation (e.g., change in species or change in structure), there may be doubt as to whether this new condition represents a fundamental transition, or simply a shift within natural limits of persistence or development of communities (Pickett and White 1985; Trudgill 1977).

Consequently, we propose that a pragmatic approach to vegetation modification should regard native vegetation structure and composition as dynamic and that a change in state (i.e., a sustained change in the structure and composition of native vegetation) is linked to a change in regenerative capacity.

To assist in the process of classifying regenerative condition states we propose four native condition states based on change in regenerative capacity: (1) no apparent perturbation, (2) a shift within the normal limits of persistence, (3) regenerative capacity is limited or at risk, and (4) regenerative capacity is suppressed or lost.

Human Intervention: “Natural” Versus Anthropogenic Change

There are philosophical issues regarding the role of humans in shaping vegetation characteristics and how these influences should be regarded in terms of condition states and transitions. At one extreme is the view that any anthropogenic effect on vegetation is equivalent to the effect of any species and thus should be considered “natural.” An opposing view would hold all vegetation to be a human artifact, and therefore disturbed, because of the pervasive influence of humans in the biosphere, both in pre-technological and post-technological societies (Lesslie 1997). A useful approach to resolving the issue of the role and effect of humans in shaping vegetation is to identify anthropogenic change as change that results from the application of artificial technology and energy subsidies (Odum 1975; Lesslie 1997; McIntyre and others 2002). Unless such an approach is adopted, profoundly different interpretations of vegetation modification can occur. It follows from this approach that vegetation conditions prevailing at the time of European settlement of Australia should be accepted as a notional (and where necessary assumed) reference or benchmark state.

There remains the problem of distinguishing anthropogenic conditions from those occurring naturally. Even if an effect on vegetation is clearly anthropogenic, it may not differ significantly in degree, extent or effect to natural disturbances such as landslides, storm damage, wildfire, or disease. Moreover, the pervasive effects of human activity in the biosphere and complex linkages with natural systems and processes make it increasingly difficult to separate anthropogenic from naturally occurring condition states. Of global significance in this context is the impact on vegetation of changing atmospheric composition and enhanced climate change (Climate Change Science Program and the Subcommittee on Global Change Research 2004). In the Australian context fire has played a major role in shaping the structure and composition of vegetation in most Australian landscapes since the Tertiary Period (Gill and others 1981). This includes the profound effects of Aboriginal burning for up to 50,000 years (Jones 1969; Bowman 1998) and additional post-settlement change through altered fire intensity and frequency (Leigh and Noble 1981). The deliberate or inadvertent introduction of invasive species in concert with fire has also been responsible for major vegetation changes, the full effects of which are complex and yet to be fully realized (McIntyre and others 2002).

Despite the desirability of untangling natural from human-induced effects, in some circumstances it may not be feasible or realistic (Hobbs and Hopkins 1990). For native vegetation it may not be possible to distinguish anthropogenic from natural states in terms of structure, composition, or regenerative capacity alone.

Faced with the complexity and uncertainty of untangling natural from human-induced effects, we propose a pragmatic solution, the key distinguishing feature for classifying the condition states of vegetation must be evidence of land use, land management practices, or other active intervention, such as the deliberate or inadvertent introduction of organisms into natural vegetation, e.g., promoting exotic pasture species in rangelands to increase productivity.

The VAST Framework

A state and transition model is used as the framework for describing the modification states of vegetation types because it provides (1) an appropriate gradient of change for describing vegetation modification, (2) assistance in classifying the effects of land management practices in changing the state of vegetation, (3) assistance in identifying what data and information are necessary to define normal limits of change and succession, and (4) assistance in identifying gaps and uncertainty where more data and information are needed. State and transition models are useful tools to assist decision-makers assess and communicate the actions and effort needed to plan realistic goals for vegetation management, and to monitor and report progress toward desired management outcomes.

The VAST framework explicitly distinguishes the extent of native and non-native vegetation from the VAST condition states (mapping criteria), which are defined by diagnostic criteria. Diagnostic criteria include objective criteria (composition and structure) with more ‘interpretative’ criteria (current regenerative capacity) (Table 1). This clarification is a refinement of the VAST framework presented by Thackway and Lesslie (2006).

Table 1 The vegetation assets, states and transitions (VAST) classification framework. Increasing vegetation modification from left to right

Input datasets comprising information that informs the requisite diagnostic criteria, including explicit or implicit benchmarks for each vegetation type (i.e., association), can be reclassified into VAST condition states. To assist prospective users in applying the VAST framework we provide the following seven guiding principles:

  1. 1.

    Contemporary patterns of vegetation in highly modified Australian landscapes comprise native, non-native, and non-vegetated areas (states 0–VI in Table 1). The seven broad condition states encompass all vegetation types (native, non-native, and non-vegetated areas) found across the landscape. Depending on requirements, additional condition substates can be added within each of the seven main states.

  2. 2.

    Natural non-vegetated states and substates are bare areas. In the context of the NVIS framework, naturally non-vegetated “definitive vegetation types” (Hnatiuk and others 2008) could be included in state 0 (e.g., salt lakes, sand, mud flats, and rock).

  3. 3.

    Condition assessments can be reported at different points in time for the same area using structural, compositional, and functional attributes. To enable such comparisons to be made, it is necessary to collect and compare the same diagnostic attributes to assess changes in the condition state of particular vegetation associations and their extent in different parts of the landscape.

  4. 4.

    Native vegetation refers to those condition states and substates that can be defined and mapped where the regeneration of species/communities and ecosystems is not predominately prevented or excluded by land management practices. Because native vegetation can be identified by characteristics of its structure and composition (Hnatiuk and others 2008), it provides a distinctive, but not exclusive, set of attributes that can be surveyed and mapped or monitored.

  5. 5.

    Non-native vegetation includes those condition states and substates where the vegetative cover is predominately non-native and regeneration of the native vegetation is repeatedly suppressed or prevented by land management practices. Such areas include VAST V (e.g., crops, plantations, and improved pasture) and VAST VI (areas where the vegetation has been removed, e.g., water reservoirs, urban areas, salt crusted areas, and tilled bare soil).

  6. 6.

    In the context of point 3 above, where condition states can be defined and mapped across the whole landscape, management actions can be used to facilitate transitions between condition states: (a) management actions can “transition” a condition state from VAST state I to a state III or even a state VI; (b) depending on the value system and perspective of the land manager, a manager with sufficient resources and knowledge about ecological restoration can “transition” a condition state from VAST state III to a state I. As noted by McIntyre and Hobbs (2000), land managers should be strategic and aim for least cost solutions when planning the restoration of vegetation associations, e.g., differentiate those sites where the regenerative capacity can be reinstated from those areas where the regenerative capacity has been lost; and (c) noting in the short to medium term (e.g., 10–50 years and longer for more complex vegetation communities) it is not possible to “transition” a non-native condition state (i.e., states IV–VI) back to a native condition state. Where stakeholders plan to restore areas that were formally non-native vegetation types with native species, the structure, composition and function and the regenerative capacity of the “reconstructed native vegetation” will, in the short to medium term, be discernable as a revegetated type. For the purposes of reporting, such revegetated areas should be denoted as VAST state V.

  7. 7.

    Datasets that are eligible for translation and /or interpretation into the VAST framework must have implicit or explicit benchmarks (Thackway and Lesslie 2006) for each vegetation association.

In addition to the above seven guiding principles, we provide four assumptions which underlie any application of the VAST framework that aims to create a condition state dataset. These assumptions have been developed through extensive consultation with conservation biologists and field ecologists:

  1. 1.

    Under natural environmental conditions (i.e., absence of anthropogenic disturbances), the structure, composition, and function (including the regenerative capacity) of vegetation is a response to environmental gradients (Whittaker 1967).

  2. 2.

    In managed native vegetation types, the regenerative capacity of native vegetation, can to a large extent be measured /observed and interpreted to be the result of previous and current land use and land management practices.

  3. 3.

    The effects of managing vegetation can be observed and interpreted as condition states at a range of scales. Condition state datasets can be derived using a range of methods, including inventory, mapping, and modeling. For example, appropriate input datasets can be reclassified and /or remapped into VAST condition states provided the diagnostic attributes are inherent or can be inferred or interpreted in a vegetation condition dataset. The reliability of condition state datasets can be demonstrated and documented to assist prospective users of this information.

  4. 4.

    Within a condition state, management interventions that aim to restore ecological processes must be based on sound ecological research and what is practical and feasible in the field (Society for Ecological Restoration International Science and Policy Working Group 2004). Restoration projects require information on the extent and duration of past disturbances and their effect on biophysical processes, cultural conditions that have shaped the landscape, species availability, and species performance and assembly rules (Hobbs and Norton 1996; Lockwood 1997; Thackway and Lesslie 2006).

National and Regional Case Studies

To illustrate application of the VAST framework at national and regional scales we describe the development of a national scale VAST dataset and discuss relationships between it and three different regional scale VAST datasets; Shoalhaven region New South Wales, north west Victoria, and Northern Territory. These case studies show that the VAST national and regional scale datasets are different in the three regional case studies because of different input datasets. These differences highlight the need to understand issues of accuracy and precision as well as levels of detail associated with the scale of mapping or modeling and the need for consistency between the attributes used to derive the mapped condition state datasets.

National Case Study

An interim national dataset on vegetation condition states was derived from several readily accessible sources of mapped information (Lesslie and others 2008). An expert model involving an implicit pre-European benchmark vegetation condition for each vegetation association along with knowledge of the effects of land use and land management practices upon the integrity of the native vegetation associations was used to classify each 1-km grid cell. The national dataset comprises VAST condition states 0, I-III, V, and VI (Fig. 1 and Table 2). The dataset covers the land area of Australia (768 million hectares) and comprises information collected between 1995 and 2003.

Fig. 1
figure 1

Interim VAST classification for Australia (2005)

Table 2 A comparison between the area and relative proportion of VAST condition states for Australia’s states and territories derived from the national scale VAST dataset

Key inputs used to derive the national dataset include the Biophysical Naturalness layer within the Australian Land Disturbance Database (ALDD) held by the Australian Government Department of the Environment, Water, Heritage, and the Arts (cf Lesslie and Maslen 1995), a national land use dataset prepared for the National Land and Water Resources Audit (Stewart and others 2001), a variety of catchment scale land use datasets produced through the Australian Collaborative Land Use Mapping Program (Lesslie and others 2006), and MODIS satellite imagery (Loveland and Belward 1997). GIS methods were used to overlay input datasets and the VAST states in each dataset were averaged to derive a synthetic VAST condition state for each grid cell.

This view of vegetation condition states highlights the characteristic regional patterns within each of Australia’s states and territories (Table 2). The national pattern is characterized by:

  1. 1.

    Very large areas of residual and modified (VAST I and II) vegetation in central and northern Australia’s rangelands (refer to Table 2; Western Australia, Northern Territory, South Australia, Queensland, and New South Wales),

  2. 2.

    Large areas of residual and modified (VAST I and II) vegetation in temperate areas less suitable for agricultural production, mainly mountainous forested locations (refer to Table 2; Australian Capital Territory, New South Wales, Tasmania, Victoria, and Western Australia),

  3. 3.

    Replaced (VAST V) vegetation, mainly cropping and improved pasture, with remnant VAST II and III (modified and transformed) vegetation in fertile, better watered regions (refer to Table 2; Queensland, New South Wales, Victoria, Western Australia, South Australia, and Tasmania),

  4. 4.

    Extensive modified and transformed (VAST II and III) vegetation from livestock grazing in arid and semi-arid rangelands (the key controls being the presence of palatable vegetation and proximity to water). Refer to Table 2; Queensland, New South Wales, South Australia, Western Australia, and Northern Territory), and

  5. 5.

    At a national level relatively small areas of removal (VAST VI) located on the coastal margin associated with human settlement (urban areas and water reservoirs). Refer to Table 2; New South Wales, Victoria, Western Australia, Queensland, and South Australia.

Regional Case Studies

In this section the application of the VAST framework is illustrated using three regional case studies using regional scale datasets: Shoalhaven region, New South Wales, north western Victoria, and the Northern Territory. In addition, we also compare these results with the national scale VAST dataset for the same study areas.

Shoalhaven Region, New South Wales

Figure 2a shows condition states for Shoalhaven region, New South Wales at the national scale (Fig. 2a(i)) and the regional scale (Fig. 2a(ii)). The Shoalhaven region area covers 983,000 hectares and comprises information compiled from 2000–2004 from vegetation site surveys and mapped at 1:100,000 scale (Tindall and others 2004). The dataset, also known as NSW Native Vegetation Mapping Program Series 4 or P5MA (Priority 5 Mapping Area), is part of the Native Vegetation Mapping Program (NVMP) (Sivertsen and Smith 2001).

Fig. 2
figure 2

Thumbnail images comparing VAST condition states in three regional study areas of Shoalhaven region: New South Wales (a), north western Victoria (b), and Northern Territory (c) for national and regional scale datasets, respectively

The New South Wales Department of Environment and Climate Change (DECC) (formerly NSW Department of Natural Resources and NSW Department of Environment and Conservation) used an expert model combining site-based information, an implicit pre-European benchmark vegetation condition for each vegetation association along with the relative position of each mapping unit in the catchment and land use and land management practices to derive VAST states 0, I–III, V and VI (Thackway and Lesslie 2006). DECC also worked with the original vegetation surveyors, the mapping team, and regional experts to validate and improve the reliability of the final vegetation condition states.

A comparison between the national and regional datasets shows obvious differences in the distribution of types of VAST states found in the study area. Figure 2a(i) and (ii) show differences in the areas and the relative proportions of VAST condition states are shown in Fig. 3a. The national level information has the same number of VAST condition states as the regional dataset. The national scale dataset depicts the vegetation of the study area as comprising VAST I (353,000 hectares or 36.0%), VAST II (266,000 hectares or 27.0%), VAST III (47,000 hectares or 4.8%) VAST V (310,000 hectares or 31.6%), and VAST VI. (5000 hectares or 0.5%) of the study area. In contrast, the regional scale dataset depicts the same area as having 42.9% or 422,000 hectares mapped as VAST I, VAST II (35,000 hectares or 3.6%), VAST III (234,000 hectares or 23.8%), VAST V (236,000 hectares or 24.0%), and VAST VI (39,000 hectares or 4.0%) of the study area.

Fig. 3
figure 3

A histogram of the area of each VAST condition states in three regional study areas of Shoalhaven region: New South Wales (a), north western Victoria (b), and Northern Territory (c) for regional and national scale datasets, respectively

North Western Victoria

Figure 2b shows condition states for north western Victoria at the national scale (Fig. 2b(i)) and the regional scale (Fig. 2b(ii)). The dataset covers an area of approximately 9 million hectares and comprises information collected between 1999 and 2003 (Thackway and Lesslie 2005). Information on vegetation condition states was collected at sites from a number of sources (Newell and others 2006) using the “habitat hectares” approach (Parkes and others 2003). The “habitat hectare” score comprises 10 separate components (7 “site condition” and 3 “landscape context” components), which are scored in relation to an explicit pre-European benchmark for each vegetation type present at a site. The Department of Sustainability and Environment (DSE) used a Neural Network model to spatially extend the site-based condition data to whole region (Newell and others 2006). The model used 13 mapped (i.e., GIS and remote-sensed data) variables including land use, vegetation type, geology, climate, and tree density to map the final regional “habitat hectare” scores. The north western Victoria VAST dataset uses a 30-meter square grid cell. The scale of the corresponding dataset on vegetation types (i.e., ecological vegetation classes) is 1:25,000 scale (NLWRA 2001).

The five condition states for the regional scale native vegetation dataset from DSE were reclassified into three classes (i.e., VAST states I, II, and III), which comprises 75 units of the “habitat hectare” 100-unit score range. Information required to complete the remaining condition states, i.e., naturally bare areas (VAST 0) and non-native vegetation (VAST IV, V, and VI) were obtained from the DSE’s land use and vegetation type datasets.

A comparison between the national and regional datasets show a similar distribution of VAST condition states found in the study area Fig. 2b(i and ii), as well as areas and relative proportions for each VAST condition states Fig. 3b. The regional dataset defines 3,051,000 hectares, or 33.4%, native vegetation (comprising VAST I, II, and III) compared to 2,630,000 hectares, or 28.8%, (comprising VAST I, II, and III). The regional dataset shows majority of the vegetation is VAST V with 6,076,000 hectares, or 66.6%, and the national dataset shows VAST V as having 6,439,000 hectares, or 70.5%, of the study area (Fig. 3b).

Northern Territory

Figure 2c shows condition states for the Northern Territory at the national scale (Fig. 2c(i) and the regional scale (Fig. 2c(ii)). The Northern Territory covers an area of approximately 134,620,000 million hectares and the database comprises information collected between 1995 and 2005 (Thackway and others 2005). The map of vegetation types for the Northern Territory is compiled from various scales of mapping; however, the dominant scale is 1;1,000,000 (NLWRA 2002). The Northern Territory VAST dataset uses a 1000-meter square grid cell.

The developers of the regional scale VAST dataset used a heuristic model involving an implicit pre-European benchmark vegetation condition for each vegetation association along with land use and land management practices information to allocate VAST states 0, I–III, V, and VI. The primary drivers of vegetation condition state changes were land use (scale 1:1,000,000), fire frequency 1997–2003 (scale 1:250,000), and a surrogate of grazing intensity (distribution of domestic stock based on distance from water points). Each dataset was classified separately involving reclassifying attribute class intervals into appropriate VAST states, e.g., a grid cell of native vegetation with a fire frequency of 7 burns in 7 years was classified as VAST III. All three input datasets were overlain and the VAST states in each dataset were averaged to derive a synthetic VAST condition state for each grid cell (Thackway and others 2005).

A comparison of the national and regional datasets showed obvious differences in the distribution (Fig. 2c(i and ii)), as well as differences in the areas of VAST condition states (Fig. 3c). The national level information shows a greater area of naturally bare areas (VAST 0) 948,000 hectares, or 0.7%, compared to only 135,000 hectares or 0.1% of the regional dataset. There are marked differences between the area and proportions of all three VAST condition states for native vegetation (VAST I, II, and III) in the national and regional scale datasets (Fig. 3c). For example, VAST I at the national scale recorded 84,741,000 hectares, or 63.0%, and the regional scale 43,899,000 hectares, or 32.7%; VAST II at the national scale recorded 35,648,000 hectares, or 26.5%, and the regional scale 66,594,000 hectares, or 49.5%; and VAST III at the national scale recorded 12,886,000 hectares, or 9.6%, and the regional scale 22,982,000 hectares, or 17.1%. There was no obvious difference between the area of VAST V in the national and regional scale datasets. However, there was an obvious difference between the area of VAST VI between the national and regional scale datasets (i.e., 12,000 hectares, or 0.0%, and 407,000 hectares, or 0.3%, respectively [Fig. 3c]).

Discussion

Publicly funded programs associated with natural resource management and the sustainable use and management of native vegetation require information on the status and condition of native vegetation. This information is needed at both national and regional scales to identify priorities, to set targets for investment, and to monitor and evaluate progress to agreed outcomes.

In this article we have demonstrated the flexibility of the VAST framework in translating and compiling disparate vegetation condition state datasets derived from different scales (national and regional) and in comparing the resultant datasets. In all four VAST datasets, that is the national scale Australia-wide dataset and the three regional scale case studies (Shoalhaven region New South Wales, north western Victoria, and the Northern Territory) we show that, provided the input datasets for VAST condition states I–III contain information on how each map unit has been modified by land use and land management practices relative to a VAST I benchmark, native vegetation condition state datasets can be derived. Depending on the requirements for assessing, monitoring, and reporting, non-vegetated (VAST state 0, i.e., naturally bare areas, and VAST VI, i.e., removed, as described in Table 1) and non-native cover types (VAST states IV and V, such as crops and plantation forests as described in Table 1), many of these attributes can be inferred and translated from land cover and land use mapping datasets. Given that many non-vegetated and non-native cover types have obvious spectral signatures, such land cover types can readily be detected and mapped using remotely sensed satellite imagery with adequate ground-based sampling.

Accuracy and Precision

The process of translating input datasets into condition states using the VAST framework has been evaluated in each case by stakeholders with expert knowledge of vegetation condition in the relevant study areas. The qualitative review suggested that VAST mapping broadly corresponds with condition state/s expected in the field, given the accuracy and currency of input data used to derive the condition states.

Comparisons between the national and the three regional case studies in Shoalhaven region, New South Wales, north western Victoria, and the Northern Territory highlighted differences between the national and regional scale datasets. One such example was the obvious difference observed in the national and regional scale datasets in the Northern Territory for VAST VI (i.e., water impoundments, urban and industrial landscapes, quarries and mines, transport infrastructure, and bare areas caused by land management practices). The Northern Territory national and regional scale datasets recorded 12,000 hectares, or 0.0%, and 407,000 hectares, or 0.3% (Fig. 2c(i and ii)) and Fig. 3c) respectively. These differences can be attributed to the currency of land cover datasets used to define VAST VI, i.e., removed, where the currency of the national dataset was 1995 and that of the regional scale dataset was 2005. Based on this comparison it can be surmised that, over the 10 years from 1995 to 2005, approximately 396,000 hectares has been converted to VAST VI where the vegetation had been removed.

Level of Detail

The VAST datasets presented and described in this article have been developed from different sources and at different scales using different approaches. The Shoalhaven region study area in New South Wales is a good example. The national dataset showed the predominance of Residual, Replaced-managed, and Modified vegetation (i.e., VAST I, II, and V, 36.0%, 27.0%, and 31.6%, respectively) and correspondingly few grid cells with Transformed native vegetation (4.8%). The largest areas of least modified native vegetation (i.e., VAST I and II) are confined to upper slopes of steeper hills and shallow stony and rocky sandstone plateaus. In addition, there is also some association with land use, e.g., National Parks and protected water catchments or sheds tend to be classified as VAST I. The national dataset used relatively coarse 1-km grid cells and each grid was scored using the best available heuristic knowledge of historic and current forestry, grazing, and cropping land uses and land management practices. Condition states in this dataset were derived by inferring the effect that land uses and land management practices have on modifying the structure, composition, and regenerative capacity of the native vegetation.

In contrast, the regional scale dataset recorded a similar total area of native vegetation (i.e., VAST I, II, and III, 70.3% compared to the national 67.8%), however, the regional level dataset depicted much finer scale patterns of native vegetation. The regional dataset mapped the spatial extent and the types of native vegetation using a combination of detailed color aerial photography (approximately 1:25,000 scale) and spatial modeling of floristic assemblages using abiotic environmental variables using a grid of 25-m cell resolution. Many of the patches defined as native vegetation were ground-truthed, where assessments were made of their condition states by scoring observed disturbances at sites and at the level of the patch. The primary difference between the national and regional scale datasets relates to the method used to map the extent and types of the native vegetation and to ascribe condition scores to map units. The national dataset underrepresented several VAST states including VAST 0, III, and VI in the Shoalhaven region. These differences can be explained by the currency of the input datasets and the scale of mapping.

The national representation of condition states using the VAST framework is of value strategically; however, at this scale, the national VAST dataset is generally of limited worth and should be limited to broad continental assessments of condition states. This is because of the coarseness of the spatial data and temporal variability in the accuracy and precision of input data.

Access to VAST condition datasets coupled with spatial analysis makes it possible to routinely describe and map the four landscape alteration levels of McIntyre and Hobbs (1999), i.e., Intact, Fragmented, Variegated, and Relictual. The McIntyre and Hobbs model is widely known and understood and can assist conservation biologists and natural resource managers address the full spectrum of human impacts observed across agricultural and fragmented landscapes. The habitat modification states of McIntyre and Hobbs (1999) correspond to the VAST condition states as follows: Unmodified (VAST 0 and I), Highly modified (VAST II), Modified and retained (VAST III), and Destroyed (VAST IV, V, and VI). We use VAST condition state datasets, such as the Shoalhaven region (Fig. 2a(ii)) as the input VAST dataset (100-m × 100-m raster) to derive landscape alteration levels (Fig. 4). The distribution and percent area of each landscape alteration level is shown in Figs. 4 and 5, respectively. These were derived using the proportions (i.e., % area) of each condition state measured in a 500-m moving window radius using FRAGSTATS software (Mutendeudzi and Thackway 2008).

Fig. 4
figure 4

A map showing the landscape alteration levels for the Shoalhaven region, New South Wales. The % levels are based on the McIntyre and Hobbs (1999) conceptual model and correspond to (or are estimates of) the remaining amount of native vegetation cover (i.e., the sum of VAST I, II, and III)

Fig. 5
figure 5

A histogram of the landscape alteration levels for the Shoalhaven region, New South Wales. The % levels are based on the conceptual model of McIntyre and Hobbs (1999) and correspond to (or are estimates of) the remaining amount of native vegetation cover (i.e., the sum of VAST I, II, and III)

Together the two datasets, namely the VAST condition states—regional scale (Fig. 2a(ii)) and landscape alteration levels (Fig. 4) help conservation biologists and natural resource managers assess vegetation management, investment options, and set work priorities at both the regional and national levels.

Maps and statistics of condition states and landscape alteration levels provide insights into the spatial patterns and processes that facilitate transitions between condition states. McIntyre and Hobbs (1999) discuss how the context of a patch of native vegetation (i.e., its type, extent, and condition) will influence its likely trajectory in a spatial context. For example, targeting those remnant patches of native vegetation which are large in size, relatively well connected to other large patches, and are least modified can reduce the cost of ongoing restoration and vegetation management. Such an approach has been proposed by Terry and others (2006) in planning the restoration and management of the European Green Belt project. Equally, this information can be used to highlight those areas where there could be resource conflict in regard to sustainable production (i.e., resource access and security).

At the national level VAST condition states and landscape alteration levels could be used to better target land managers to achieve change in land management practices and bring about improvements in vegetation condition state/s within the context of nominated landscape alteration levels. For example, there is an increasing reliance in the National Natural Resource Management Monitoring and Evaluation Framework (NRMMC 2002) on demonstrating how program investments in improving vegetation condition states are meeting natural resource condition objectives.

At the local and regional scale VAST condition states and landscape alteration level datasets could be used to select local projects that aim to restore native vegetation associations. However it should be noted that the condition of the herbaceous layer is harder to describe and map into particular condition states than the overstorey. The key to restoring the functionality of the ground layer vegetation requires on-ground evidence of local land use histories within particular vegetation association is e.g. the application of fertilisers and the loss of top soil. This local level information may, in turn, give clues as to what land management practices, within condition states, might be used to facilitate transitions between condition states. This approach could be used, for example, to improve the selection of landholders to be targeted in developing the Box-Gum Stewardship Program (Australian Government 2007) and in reporting the performance of restoration projects relative to regional environmental targets.

Where decision makers need access to information about the modification states of native vegetation, greatest versatility in decision making can be achieved using the VAST framework where users of VAST datasets have sound working knowledge of how land management practices can be used to transition or change one condition state to another for a vegetation type over time by influencing the vegetation type’s structure, composition and regenerative capacity. Depending on the values and ecosystem goods and services desired from an area, decision-makers can model expected changes in the structure of a vegetation type (e.g., cover/density, basal area, number of layers or strata, growth forms), its composition (e.g., dominant structuring species), and its regenerative capacity (e.g., growth stages/age classes, weeds, viability of propagules, and vegetative reproductive material) by changing land management practices, i.e., treatments that modify the diagnostic attributes which in turn effect the mapping of vegetation condition states. We suggest that three parameters may be used to describe and map the effect of a land management practice/s on the VAST diagnostic attributes for a vegetation type (i.e., an association); its frequency (e.g., monthly/yearly), its seasonality (e.g., spring), and its intensity of effect on the life form (e.g., hand cutting trees versus mechanized tree harvesting, hand pruning versus bulldozer chaining of trees). As an example, Thackway and others (2007) illustrated the application of VAST as a tool for reporting changes and trends in vegetation over a period 1980 to 2004 for a 250 hectares property in southern New South Wales. That example utilized detailed records of on-ground land management practices and corresponding patterns of vegetation structure observed in a range of geographic positions and in large format aerial photography. Observed states and transitions in the vegetation over the 24 years were interpreted relative to a VAST I benchmark.

In terms of broad-scale mapping of condition states, the challenges of making the framework operational are considerable. While alterations to the structural nature of the overstorey can be detected using remotely sensed images, the condition of the herbaceous layer is harder to perceive, yet it holds the key to vegetation functionality in many cases. While such vegetation associations present particular difficulties, we consider that these issues can be largely overcome by using specialized methods for field surveying supported by spatial modeling and accuracy assessments. McIntyre and Lavoral (2007) give specific examples and general principles in relation to states and transitions resulting from land use and effects on nutrients, soil disturbance, and leaf traits in relation to regeneration.

Conclusion

The VAST framework is a classification that may be used to guide the development of condition state datasets using a range of methods and data types. The framework enables inferences to be made about vegetation composition, structure, and regenerative capacity, relative to an undisturbed benchmark.

The VAST framework provides a simple communication and reporting tool designed to assist in describing and accounting for human-induced modification of vegetation and arresting or reversing it to achieve desired outcomes. The scheme presented here provides a broad classification of vegetation modification as a basis on which strategic planning and management decisions can be made. McIntyre and others (2002) and DEWHA (2006) identified the need for more informed decisions about the level of intervention that is needed to ensure long-term maintenance of a vegetation type. In this article we have argued that more informed decisions can be made by using information derived from condition state mapping for native vegetation. Such information can be combined with the extent of fragmentation of Australia’s native vegetation, along with who owns it and how it is managed, where in the landscape a patch is located, the size and connectivity of patches, and the modification states within the patches.

The VAST framework can help describe and account for changes in the status and condition of vegetation, make explicit the links between land management and vegetation modification, provide a mechanism for describing the consequences of land management on vegetation, and contribute to the analysis of ecosystems services (including trade-offs) provided by vegetation.

While attributes have been developed for describing and mapping Australia’s vegetation extent and types through the National Vegetation Information System (NLWRA 2007), there is a need for an agreed national framework to make explicit the links between land management practices and vegetation modification. VAST has the potential to provide a consistent national framework for monitoring and reporting vegetation condition states at a range of scales. VAST datasets, like the national and regional scale datasets discussed here, have the potential to describe the response of vegetation to changes in land use and land management practices, to be used for describing and mapping vegetation changes, and for monitoring progress toward vegetation targets. Of course, the VAST framework could be scaled-up and adapted internationally, at the country level, for reporting on the performance of progress toward targets such as those described in the Millennium Ecosystem Assessment. In addition, given the flexibility of the VAST framework, there is also potential to use it as a state and transition model for predicting the likely future vegetation condition states. Such information can be used by policy and program investors to influence or procure desired changes in land management practices as a basis for achieving multiple ecosystem service outcomes.