Recent advances in conceptual frameworks in vegetation classifications, such as the EcoVeg approach that underpins the International Vegetation Classification (IVC) developed by NatureServe staff and colleagues, offer opportunities to enhance national classification initiatives. For example, the United States National Vegetation Classification, after years of struggling with a primarily physiognomic and floristic classification, adopted EcoVeg as its new standard because it incorporated physiognomic, biogeographic, floristic, and ecological elements in a flexible yet consistent hierarchy to serve a broad spectrum of ecological research and management needs. Other countries in the Americas, especially Canada and Bolivia, helped advance the approach. Similarly, Australia has a long history of developing various vegetation typologies at local to regional scales, but ecologists recognize the need for an Australia-wide, plot-based classification system that will need to incorporate similar elements to those of the EcoVeg approach and fit within an international classification system. Here we provide examples of structures and criteria for some Australian classifications along with the IVC, and give illustrations of how Australian classifications of forest, shrublands, grasslands, deserts and wetlands could potentially link into the IVC hierarchy. We then discuss how the IVC has the capacity to summarize the full range of Australian vegetation at a broad formation (biome) scale and potentially inform future work towards an Australian vegetation classification system at lower levels, and, vice versa, the implications of Australian vegetation classification work for IVC development.

The National Vegetation Information System (NVIS) was originally developed to underpin the National Land and Water Resources Audit (NLWRA) assessment of vegetation in Australia. It is a good example of ongoing collaborative initiative between the Australian and State and Territory governments. This initiative aims to manage floristic and structural vegetation data to improve native vegetation planning and management within Australia.

NVIS’s first 15 years of development focused on national vegetation standards, protocols and classifications. The last five years have focused on automation, data exchange and vegetation modelling. Automation has significantly reduced the resources needed to produce the NVIS; for example, a data-driven tool was recently developed to assign vegetation groups based on defined rulesets. A National Information Model and Data Exchange Protocol working group has been established to better standardise vegetation data interchange nationally. Some of the outputs of the group is the establishment of standard national vegetation vocabulary and data exchange framework.

Pilot programs have been run to investigate the integration of modeled data into the NVIS, such as modeled vegetation cover data or height data. Other recent work involves using machine learning models to directly predict vegetation communities from raw State and Territory data, successfully predicting TasVeg data to Major Vegetation Groups. The need for better vegetation information is only increasing, with the largest emerging need being temporal vegetation extent/change information and its associated condition. NVIS is a critical piece of this puzzle, and its ongoing improvement will complement other initiatives in vegetation ecology, modelling and mapping.

Various vegetation classification systems have been applied to plot-based vegetation data in the Northern Territory since the 1990s. About a decade ago, the national vegetation guidelines and National Vegetation Information System (NVIS) were adopted and continue to be used as the standard classification system and field protocol in the Northern Territory. Parallel to implementing the national guidelines, the Northern Territory Vegetation Site Database was developed to store hierarchical plot data. The system has semi-automated tools that can classify plot data equivalent to levels two (structural formation), five (association) and six (sub-association) of the NVIS Information Hierarchy. As a result of inconsistencies between surveys and varying degrees of floristic and structural information collected in the field, only a portion of the data is compliant with the NVIS hierarchical levels.
Classification of plot data in the Northern Territory is more commonly used for vegetation mapping. The only Territory-wide vegetation classification is the 1:1 million vegetation map where 112 broad vegetation types were described from an intuitive appraisal of numerical analytical techniques. This hybrid approach continues to be used, however a standard method needs to be developed in the Northern Territory to classify plot data into meaningful vegetation types. The key challenges ahead are to collect standardised plot data that is transferrable and can comply with a consistent vegetation classification for Australia. In addition to reviewing the compatibility of plot data to NVIS, we also describe how it conforms with the International Vegetation Classification.

The assessment of terrestrial biodiversity in NSW utilises a master typology of native plant assemblages known as Plant Community Types (PCTs). New biodiversity legislation has increased scrutiny of PCTs and emphasised the need for a schema that enables a consistent and objective identification of types. Current PCTs in eastern NSW are an iteratively compiled set of types interpreted from multiple independent sources that vary widely in scale and reliability. This accumulated complexity reduces the reliability of biodiversity assessments.
We applied a bottom-up approach to identify a revised set of PCTs for eastern NSW from analyses of approximately 50,000 standard vegetation survey plots. Multivariate techniques were used to identify floristic and environmental patterns, commencing with a mixture modelling approach to partition the dataset into regions of common probability profile. We evaluated a suite of contemporary clustering algorithms before adopting k-means clustering to explore finer patterns at a consistent classification scale within each defined region. Draft groups were reviewed against multiple factors using a standardised work flow to produce final revised PCTs.
Approximately 1,100 revised PCTs are identified for eastern NSW. Floristic, environmental and spatial attributes of PCTs are defined by member plots and explicit plot membership is now stored in an online database. A new online tool allows quantitative metrics to be used to objectively assess new plots against PCTs. We argue that the revised classification is a major advance, resulting in reduced complexity and uncertainty for users, improved access to primary data, and enhanced functionality for other applications including mapping.

Historic vegetation clearing and land use change of the Swan Coastal Plain (SCP) in south-western Australia has resulted in a significant reduction in the extent and distribution of Banksia woodlands, which are a Threatened Ecological Community (TEC) at a Federal level under the EPBC Act 1999. Floristic community type 20a (FCT20a) is a Western Australian TEC characterised by Banksia attenuata woodlands with species rich understorey. A database was developed to combine flora and vegetation data from a wide range of sources for the SCP from 1991 to 2017. Extensive data cleaning and curation was conducted to standardise nomenclature across the dataset. Ordination was conducted to identify outlier plots, which were removed. A meta-analysis was conducted on approximately 1800 flora plots. Preliminary classification results have identified that threatened Banksia woodlands appear to be more distinct at a local-scale than previously thought. This suggests that units such as FCT20a could be further split into restricted area-specific vegetation/flora units for management and approvals purposes.

The Regional Ecosystem classification system was adopted by the Queensland government in 1999 as a state-wide landscape classification system. It is a tripled-tiered hierarchy with a regional ecosystem being identified by an IBRA bioregion at the first tier, a geomorphological category at the second tier and one, or more, vegetation communities at the third tier. Regional ecosystems are used as a biodiversity surrogate, and a conservation planning instrument and incorporated in to legislation at three levels of government. They have been mapped at 1:100,000 scale or better across the whole state. Regional ecosystems are therefore used by a wide variety of practitioners across government agencies, conservation organisations, consultants and land holders. Vegetation communities making up the regional ecosystems are recognised, using plot-based data, at the plant association level and incorporate structure and floristic composition . To date, identifying communities has been done using fully supervised plot-grouping techniques. In this talk I will evaluate the RE classification system in a global context as well as outline work being done to move the classification approach to one incorporating un-supervised plot-grouping techniques.

There is a significant amount of data collected in ecology to monitor the environment by measuring biodiversity and ecological process at a certain point in time and space. Plot-based monitoring is used to monitor vegetation and ecosystem processes that use repeatable methods and procedures. Repeated measurements of the same observed properties would enable us to decide on on-going environmental and resource management practices. Generally, data collections are project-based, collected for a specific purpose and use the same monitoring methodologies at different plots covering more extensive geographical locations. These datasets are of considerable significance if they are integrated with other similar projects or programs. The handling and integration of ecology data are complex and challenging. Some of the challenges faced during the data integration include uncertainty of source data management and capturing and harmonising different data terminologies and methodologies.
We will discuss an approach of integrating and harmonising plot-based ecology data. The approach uses core domain-specific ontology to describe ecology plots, re-visits, sampling different observation collections associated with plots. The ontology is based on Observation and Measurement (O&M) and W3C Semantic Sensor Network Ontology. All the database-centric terminologies and methodologies are described as controlled vocabularies in a machine-readable format and mapped as properties of core ontology classes. This approach enables each data provider terminologies described in a linked data format (RDF) and enables to build a global vocabulary to describes terminologies and methodologies. We described the complete system to integrate ecology data to enable customised output.

Whole landscape, fine-scale, reliable data on habitat condition across Australia is a long standing need for improved conservation decision making and reporting on the state of our environment. A range of national and state/territory based agencies have been working to address this need for decades, with mixed success. In 2014, the Australian Government Department of Environment and Energy and the CSIRO commenced a collaborative project to develop a Habitat Condition Assessment System (HCAS). The underlying concept is a novel, big-data approach developed earlier by CSIRO alone. The HCAS progresses from previous approaches based on remote sensing, environmental data layers, or site based condition scores, by integrating data from all three sources. The model output is a simple-to-understand 250m national grid with each pixel’s current habitat condition predicted on a scale from zero to one. This presentation will outline how the HCAS works, from its conceptual basis, data inputs and modelling algorithms, to the current national output layer, key limitations, expected schedule for publication and future development work.