A near-complete dataset of plant growth form, life history, and woodiness for all Australian plants

Introduction

To provide large-scale descriptions of plant biodiversity, make best-practice management decisions, undertake phylogenetic analyses and understand ecological strategies, information on trait values must exist alongside taxon lists, occurrence data and phylogenies. Combined, this infrastructure informs us about which species exist, where they are found (occurrences) and their form and function. Incomplete trait-by-taxon matrices hinder our ability to rapidly assess functional diversity, include traits in descriptions of biodiversity or as covariates in analyses (Maitner et al. 2023). Data on species occurrences are now readily available through GBIF (at a global scale) or the Atlas of Living Australia (ALA) for Australia-specific analyses, but the limited knowledge of the functional traits associated with each taxon introduces biases into large-scale evolutionary or biogeographic analyses (Andrew et al. 2021; Stephens et al. 2023). Meanwhile, managers are unable to rapidly predict or assess species’ responses to climate change or major disturbance events if their functional traits are not readily available in a tabular format. For instance, growth form is a core trait often documented to distinguish between broad types of plants, and plants with these different core life histories may exhibit contrasting responses to climatic shifts or disturbance events (Butt and Gallagher 2018; Gallagher et al. 2022). Ultimately, despite a strong interest in plant traits and morphological and functional strategies, the patchy coverage of tabular trait data across national (or global) floras has limited the scope and uptake of trait-based approaches, especially at large spatial scales.

Recent years have seen ongoing efforts to expand coverage of a few core traits. Although standardised global and regional Australian compilations are beginning to emerge (Oliver et al. 2019; Weigelt et al. 2020; Díaz et al. 2022; Taylor et al. 2023), data on Australian plants are under-represented in global trait compilations (Maitner et al. 2023), furthering the importance of developing and maintaining compilations dedicated to the Australian flora. AusTraits is a large compilation of trait data specifically for Australian plants, including data on plant growth form for 23,338 taxa, on life history for 22,031 taxa and on woodiness of 9026 taxa in its first public release (v3.0.2, 14 July 2021; Falster et al. 2021). When released in 2021, it offered a substantial improvement on other databases for Australia, but coverage still fell far short of providing trait values for all 30,490 species and infraspecific taxa in the Australian flora.

Although tabulated data do not yet exist for most traits across most taxa, taxonomic descriptions in their raw form, or aggregated into floras, contain a wealth of largely untapped trait data and span most taxa. Basic morphological trait data are available for most described species and some of this content links directly to ecological function, such as plant growth form and woodiness (Coleman et al. 2023). In addition, nearly all taxon descriptions include the taxon’s life history. For the Australian flora, taxon descriptions have been compiled into six mostly complete state and territory floras and a partially complete national flora. There also exist regional compilations, such as the Flora of Southeast Queensland (Stanley and Ross 1983). Combined with the original taxonomic treatments, these contain near-complete coverage of these fundamental ecological traits, just not in a structured tabular format.

Recently, Coleman et al. (2023) used human-guided text processing to extract trait values from the text paragraphs given as taxon descriptions in floras. This effort greatly increased coverage for plant growth form, woodiness and life history, as well as a diversity of morphological traits such as leaf dimensions, leaf shape, fruit type, and plant storage organs for Australian plants. These data are available through the AusTraits trait database (v4.2, 18 September 2023). However, the outputs of the automated workflow remained incomplete and required manual reconciling.

In this paper, we aim to further extend Coleman et al. (2023)’s compilation by the following:

Together, these provide a near-complete single resource for 30,279 taxa or 99.3% of accepted taxa in the Australian Plant Census, for plant growth form, life history, and woodiness.

Methods and results

We focused on extending and refining tabular data compilations for three traits, namely plant growth form, life history and woodiness, for all taxon concepts accepted by Australia’s National Species List for vascular plants, the Australian Plant Census (APC). The APC (https://biodiversity.org.au/nsl/services/search/taxonomy) offers a list of the accepted taxon concepts for all Australian vascular plants, including both native and naturalised taxa. As of November 2022, the APC included 30,494 accepted taxa of rank species, subspecies, variety, and form. Trait definitions were sourced from the AusTraits Plant Dictionary (APD; Tables 1, 2, 3). The APD is a recent effort to formalise and publish definitions for all trait concepts used by AusTraits and includes allowed values for each categorical trait (APD; Wenk et al. 2023). The curated values include both broad, near-universal descriptions of plant growth form, life history and woodiness, as well as some more nuanced terms describing the unique forms and functions of Australia’s plants. These definitions were the basis for trait values extracted both here and by Coleman et al. (2023).

Automated trait extractions from Australia’s state and national floras by Coleman et al. (2023) formed the starting point for this work. They used natural language processing (NLP; innovative text-string matching) to extract plant-trait data for plant growth form, life history and woodiness from Australia’s national and state floras (Flora of Australia 2024; National Herbarium of New South Wales 2024; Northern Territory Herbarium 2024; Royal Botanic Gardens Victoria 2024; State Herbarium of South Australia 2024) and map them to the allowable trait values. In addition, trait values were inferred from higher-level taxa if, and only if, the higher-level taxon documented a single trait value. For instance, if the Flora of Australia indicated that a genus was comprised entirely of herbs, all taxa listed as species within the flora were scored as being herbs. For the state floras, only the taxa within that flora were scored by inference because the taxon descriptions may be written to just reflect the regional flora, not all Australian species; that is, a genus described within the Northern Territory flora as exclusively comprised of herbs might include shrubs in Victoria (Coleman et al. 2023). Across all floras, Coleman et al. (2023) ultimately extracted plant growth-form data for 25,396 taxa, life-history data for 23,840 taxa and woodiness for 25,127 taxa (Table 4). As Tasmania does not have a comprehensive online flora that could be processed using NLP, we added trait data on Tasmanian plants manually transcribed by contributors to the Tasmanian print flora and submitted to AusTraits (Jordan 2001; McGlone et al. 2015). This combined dataset was filtered to exclude values that were either inconsistent across floras, had multiple values within a flora, or were designated as climbers. This left the following number of taxa with unique trait scores: plant growth form, 20,784; life history, 22,314; woodiness, 24,084. Hereafter, all listed methods refer to our refinement and expansion of this dataset.

To construct a dataset with a single harmonised trait value for all of Australia’s taxa, the following three additional steps beyond Coleman et al.’s (2023) efforts were required: (1) manually interpreting ambiguous trait values; (2) merging disparate trait values offered by different floras; and (3) manually gap-filling trait data for all remaining taxa from the literature. The processes for completing each step are outlined below.

For each trait-by-taxon combination, if the starting compilation had unique, identical and unambiguous trait values across the floras, this value was accepted without review. This included all taxa with growth form consistently scored as either herbs, shrubs, trees, or graminoids across all floras that included the taxon. For life history, this included all taxa that were scored consistently as annual, biennial or perennial. For woodiness, all taxa scored consistently as either woody or herbaceous were accepted. Taxa that were reported as possessing multiple trait values, either within the same source or between multiple sources, were not automatically accepted. This included taxa that were scored as being either annual or perennial within a single flora, scored in AusTraits as ‘annual perennial’, and taxa with inconsistent scorings across floras. For these taxa, the original descriptions in each flora were read to ensure that it was appropriate to include multiple trait values for that species. The individual taxon descriptions were also read for all climbing plants, as the floras use a diverse vocabulary to describe climbers and automated extractions were inconsistently able to identify a climber as woody versus herbaceous.

For taxa without information in the floras, additional resources were relied on, including the taxonomic literature, print floras, online floras for regions outside Australia, and websites that included full taxon descriptions and/or references to floras. Because the state of Queensland lacks a comprehensive flora and the Northern Territory online flora is incomplete, most plants requiring manual scoring had Australian ranges restricted to these states. Additionally, taxa described in the past decade have generally not yet been added to the online state and national floras. Descriptions for naturalised species were, when available, sourced from an online flora for their native range, including the Flora of North America, the Flora of China and other floras available at www.efloras.org. Additionally, some online compilations were deemed to be reliable sources of information because they included complete taxonomic descriptions, were referenced, and/or were compilations by universities, government agencies or NGOs, such as lists of local weed species. For instance, Wikipedia has an increasing number of taxon descriptions because its taxon descriptions are those displayed on iNaturalist (iNaturalist 2023). Useful Tropical Plants (Fern and Fern 2023a) and Useful Temperate Plants (Fern and Fern 2023b) offer well-curated descriptions of both tropical and temperate plants with known human uses. In addition, for invasive naturalised species, there are many Australian and global weed databases. For all these sources, trait values were manually scored. Plant growth form and woodiness were scored first, with life history-trait values scored during a second data-collection campaign. The same references and process were used each time, but slightly different numbers of trait values were extracted between the two initiatives. For many taxa, life history could be scored on the basis of the plant growth-form and woodiness scores, because any shrub or tree was mapped as perennial and most woody plants could be mapped as perennial.

Among the taxa with trait values that were not automatically accepted from Coleman et al. (2023) or that lacked scores in Coleman et al. (2023), most were scored by manually reading the taxon descriptions in either the various Australian online floras (Flora of Australia 2024; National Herbarium of New South Wales 2024; Northern Territory Herbarium 2024; Royal Botanic Gardens Victoria 2024; State Herbarium of South Australia 2024; Western Australian Herbarium, Biodiversity and Conservation Science 2024), Flora of Australia print volumes (https://www.dcceew.gov.au/science-research/abrs/publications/flora-of-australia), or the Flora of Southeast Queensland (Stanley and Ross 1983) (Table 4). In many cases this simply required looking at the extracted phrase and confirming that multiple trait values within a single flora or disparate scorings across floras were accurate and appropriate. The taxonomic literature added a further 1584 plant growth-form values, 1585 woodiness values, and 1288 life-history values (Table 4). Websites, including online floras for regions outside Australia, were the source of information for 704 plant growth-form and woodiness scores and 728 life-history scores (Table 4).

In total, we report plant growth-form values for 29,993 taxa, woodiness values for 30,279 taxa, and life-history scores for 30,056 taxa; all but a few hundred of the 30,494 taxa in the initial list of APC-accepted taxa were scored (Table 5). Although most taxa display a unique value for each of the three traits, a significant component of Australia’s flora is polymorphic with respect to plant growth form, woodiness, or life history, with multiple trait values appropriate for 4871 taxa for one or more traits.

Data for plant growth form and woodiness are included in AusTraits under the dataset_id Wenk_2022 (from v4.1.0; https://zenodo.org/doi/10.5281/zenodo.3568417) and those for life history under the dataset_id Wenk_2023 (from v4.2.0) and are most easily extracted using the {austraits} R package (Falster et al. 2021; https://github.com/traitecoevo/austraits).

The following R code can be used to access the complete dataset for plant growth form, woodiness, and life history:

remotes::install_github("traitecoevo/austraits", dependencies = TRUE, upgrade = "ask")

austraits <- austraits::load_austraits(version = "6.0.0")

growth_form_woodiness <-

 austraits$traits %>%

  dplyr::filter(dataset_id %in% c("Wenk_2022")) %>%

  dplyr::rename(references = measurement_remarks) %>%

  dplyr::select(taxon_name, trait_name, value, references) %>%

  dplyr::distinct(taxon_name, trait_name,.keep_all = TRUE) %>%

  tidyr::pivot_wider(names_from = trait_name, values_from = value)

complete_traits <-

 austraits$traits %>%

 dplyr::filter(dataset_id %in% c("Wenk_2023")) %>%

 dplyr::rename(life_history = value) %>%

 dplyr::select(taxon_name, life_history) %>%

 dplyr::distinct(taxon_name,.keep_all = TRUE) %>%

 dplyr::left_join(growth_form_woodiness) %>%

 dplyr::select(taxon_name, plant_growth_form, woodiness_detailed, life_history, references)

For further information of accessing and manipulating AusTraits data, see: https://traitecoevo.github.io/traits.build-book/AusTraits_tutorial.html.

Discussion

In recognition of the need for more complete tabular trait data for Australian plants, we present the first near-complete continental-scale scoring of a trio of categorical traits. This dataset is a resource ready for use for diverse research and management programs. The data can be used to explore trait distributions across climatic gradients, as a covariate in ecological analyses, and offers a mechanism to rapidly assess species’ responses to natural disasters or to inform long-term management decisions. Analyses have indicated that plants with different values for the traits in question (life history, growth form, and stem woodiness) exhibit different functional trade-offs (Vico et al. 2016; Šímová et al. 2018; Towers et al. 2024). Similarly, plants with different values for these three categorical traits display different numeric trait–climate associations (Flores-Moreno et al. 2019), require different conservation strategies (Mayfield et al. 2006), have different responses to key disturbances (Clarke et al. 2013; Gallagher et al. 2021), and even differ in the degree to which they have a public photographic record (Mesaglio et al. 2023). Having complete tabular datasets for these traits across the Australian flora, therefore, has direct implications for the management and conservation of these species, in addition to facilitating more broad-scale biogeographic analyses.

Increasing awareness of the existence of this resource should accelerate its use in research and eliminate the need for other researchers to unnecessarily expend time and resources creating their own datasets for these traits. For example, 102 of the 363 datasets within the most recent AusTraits release included the trait plant growth form. This occasionally represents values scored in the field, but mostly these are data transcribed from floras and added into an analysis as a descriptive or statistical covariate. This information and data on life history and woodiness can now be quickly sourced from the compilation presented here. Many additional studies will have chosen study taxa on the basis of their scores for plant growth form, life history, and woodiness; this dataset represents a rapid way to scan a list of potential study taxa for appropriate study taxa. Within our research group, the plant growth-form dataset has already been used twice (Mesaglio et al. 2023; Towers et al. 2024).

The need to manually score or check the scores of these three traits for more than 25% of Australia’s plant taxa highlights the need for comprehensive, open-access, tabular resources for Australia’s flora. Most taxa missing trait values were those restricted (within their Australian range) to Queensland, the Northern Territory or offshore islands, because these regions lack comprehensive online floras. For instance, of the 162 described taxa lacking life-history scores, only 13 have ranges that include New South Wales, only six have ranges that include Victoria, whereas 82 have ranges that include Queensland and 55 have ranges that include the Northern Territory. The eventual completion of the Flora of Australia will alleviate this problem, but as long as this resource is incomplete, trait data about taxa restricted to Australia’s northern regions remain difficult to compile.

We suggest that new taxon descriptions and taxon revisions should explicitly include data for these traits in a standardised format, even if it is only to acknowledge that a taxon displays a range of trait values. A taxon may be rare and known from only a few localities, but its conservation is impeded if its annual or perennial status is unknown (Butt and Gallagher 2018; Gallagher et al. 2021). Another concern is that new taxon descriptions may be presented in books or hard-to-access taxonomic journals, instead of the generally open-access journals associated with national or state herbaria. For instance, many species of Drosera and Eremophila are described only in books (Chinnock 2007; Lowrie 2013). Until the Flora of Australia adds profiles for all taxa, portals such as Wikipedia and World Flora Online are the main online resources (e.g. the profiles for Eremophila species have all been transcribed from Chinnock (2007) into Wikipedia). Although Plants of the World Online includes plant growth-form and life-history data for most taxa, its descriptions are not always correct and were sometimes rejected as references. For instance, several instances of the non-sensical phrase ‘climbing herbaceous trees’ were noted (e.g. for Merremia gemelli), suggesting an automatic extraction of words suggesting a vine that climbed into the treetops.

Whereas it is fantastic to be able to offer a resource with trait values for 99% of Australia’s flora, it is simultaneously disappointing that these core traits were not scorable for all described taxa. For instance, there does not exist a reference that unequivocally states whether Zornia oligantha is annual, perennial, or displays a polymorphic life history. Taxa that could not be scored either had not had a taxon revision since 1950 (and usually since 1920) or the taxon description from the taxonomic literature did not include information on the three traits being scored. References from before 1950 could only haphazardly be relied on because most of these older descriptions, if accessible, rarely included information on life history for herbaceous species and did not consistently distinguish between herbs and subshrubs (and hence, taxon woodiness). There also exists a small number of genera where the most recent taxon revisions incompletely documented these traits, generally because species’ phenotypes span the discrete trait values or display an intermediate, difficult-to-express trait value. Three genera within which individual taxa were rarely scored for life history, namely Eriocaulon, Nicotiana and Spermacoce, can likely grow and reproduce as annuals or perennials, depending on the environmental conditions. Many species in the genera Sida and Solanum are difficult to score for plant growth form and woodiness, because many taxa are short-lived, post-disturbance species that grow rapidly, assuming a tall, well-branched form, but have only partially lignified stems. The phrases ‘annual shrub’ (Roepera reticulata in Western Australian Florabase) and ‘soft shrub’ (Grevillea fililoba in Flora of Australia) exemplify these ‘intermediate’ forms. The woodiness trait value ‘semi_woody’ is intended for such instances, but can be used only if the taxon description includes terminology to suggest a partially lignified stem.

We hope the existence of these datasets encourages researchers to use the data to explore evolutionary and ecological questions about Australia’s diverse flora as well as to use these data as co-variates in other trait-based analyses. We additionally hope that the compilation and release of this dataset will not just offer new research opportunities, but also inspire other researchers to work toward gap-filling other essential plant traits and sharing their results on open-source, harmonised data platforms, such as AusTraits.