Volume 238, Issue 4 p. 1685-1694
Full paper
Open Access

Photographs as an essential biodiversity resource: drivers of gaps in the vascular plant photographic record

Thomas Mesaglio

Corresponding Author

Thomas Mesaglio

Evolution & Ecology Research Centre, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

Centre for Ecosystem Science, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

Author for correspondence:

Thomas Mesaglio

Email: [email protected]

Search for more papers by this author
Hervé Sauquet

Hervé Sauquet

Evolution & Ecology Research Centre, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

National Herbarium of New South Wales, Royal Botanic Gardens and Domain Trust, Sydney, NSW, 2000 Australia

Search for more papers by this author
David Coleman

David Coleman

Evolution & Ecology Research Centre, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

Search for more papers by this author
Elizabeth Wenk

Elizabeth Wenk

Evolution & Ecology Research Centre, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

Search for more papers by this author
William K. Cornwell

William K. Cornwell

Evolution & Ecology Research Centre, School of Biological, Earth and Environmental Sciences, UNSW Sydney, Sydney, NSW, 2052 Australia

Search for more papers by this author
First published: 13 March 2023
Citations: 1

Summary

  • The photographic record is increasingly becoming an important biodiversity resource for primary research and conservation monitoring. However, globally, there are important gaps in this record even in relatively well-researched floras.
  • To quantify the gaps in the Australian native vascular plant photographic record, we systematically surveyed 33 sources of well-curated species photographs, assembling a list of species with accessible and verifiable photographs, as well as the species for which this search failed.
  • Of 21 077 Australian native species, 3715 lack a verifiable photograph across our 33 surveyed resources. There are three major geographic hotspots of unphotographed species in Australia, all far from current population centres. Many unphotographed species are small in stature or uncharismatic, and many are also recently described.
  • The large number of recently described species without accessible photographs was surprising. There are longstanding efforts in Australia to organise the plant photographic record, but in the absence of a global consensus to treat photographs as an essential biodiversity resource, this has not become common practice. Many recently described species are small-range endemics and some have special conservation status. Completing the botanical photographic record across the globe will facilitate a virtuous feedback loop of more efficient identification, monitoring and conservation.

Introduction

Whilst physical vouchers remain the gold standard for botanical specimens (Culley, 2013; Funk et al., 2018), allowing for tissue and DNA sampling and the examination of microscopic features among many benefits, the importance of photographic vouchers is being increasingly recognised (Baskauf & Kirchoff, 2008; Pitman et al., 2021), especially given they can capture information that is often absent or lost in physical vouchers (Gómez-Bellver et al., 2019). Perhaps most obviously, photographs record colour, an important feature for plant identification across many taxa, but one which is typically lost over time in herbarium sheets and other preserved specimens (Heberling & Isaac, 2018). Photographs can also allow for more comprehensive documentation of large individuals (e.g. many tree species), taxa that may be difficult to collect or preserve due to toxicity, spines or stinging hairs (e.g. cacti; Gómez-Bellver et al., 2019), and the habitat and general environmental context in which a collected plant was found (Heberling & Isaac, 2018). Although we emphasise they cannot replace specimens by any means, photographs therefore serve as an invaluable complement to physical vouchers, especially when the two are combined as a ‘fusion voucher’ for the same specimen (Gómez-Bellver et al., 2019). The ‘Extended Specimen’ concept further illustrates and develops the tremendous value of combining physical vouchers with field photographs (and other secondary data) as part of a ‘constellation of specimen preparations and data types’ (Webster, 2017), which serves to enhance existing data, stimulate new discoveries and improve accessibility of these data (Schindel & Cook, 2018; Lendemer et al., 2020).

Concurrently, with the increasing availability and affordability of digital cameras, and the rapid advances in photographic quality seen in smartphone cameras over the past decade (Delbracio et al., 2021), it is easier than ever to take high-quality photographs of the natural world. These advances have had important implications for botanical photography, with fine details and small features such as leaf venation, anther arrangement and hair morphology more easily photographed than before. Whilst many of these images are taken by professional botanists, taxonomists or consultants, and hosted on platforms curated by government agencies or scientific institutions, for example state herbaria websites, the rise in popularity of biodiversity citizen science (also known as community science) platforms such as Pl@ntNet (Joly et al., 2016), Flora Capture (Boho et al., 2020) and iNaturalist (Seltzer, 2019) has seen a tremendous influx of plant photographs online.

With over 7 million photographic observations of vascular plants submitted in the first 6 months of 2022 alone, iNaturalist has become a world leader in photographic documentation of the planet's vascular plant flora. Importantly, Heberling & Isaac (2018) provide a comprehensive protocol for connecting physical vouchers deposited at herbaria with photographs uploaded to iNaturalist, enhancing the research value of both data types. Increasingly, records submitted to iNaturalist represent the only known field photographs of rare species. For example, photographs of the rare rock daisy Perityle grandifolia, endemic to Mexico, were the first images ever taken of this species, and the first record since 1904: www.inaturalist.org/observations/64944539 (iNaturalist Contributors, iNaturalist, 2023). After the 2021 rediscovery of the presumed extinct species Gasteranthus extinctus (Pitman et al., 2022), a search through iNaturalist observations uncovered previously unidentified records of the species from almost 2 yr earlier. As usage of iNaturalist continues to increase exponentially (Mesaglio & Callaghan, 2021), records such as these will only become more common.

Yet, there are still many plant species which are only represented by images of herbarium sheets, have no photographs in any online databases, or which have never been photographed at all. A recent census of the vascular flora of the Americas found that just 53% of c. 125 000 vascular plant species across North America, Central America, South America and the Caribbean were represented by at least one field photograph in one of 25 surveyed online databases (Pitman et al., 2021). Whilst the USA and Canada were well-documented (93% of species photographed), the three most plant-rich regions (> 20 000 vascular plant species), Mexico, Colombia and Brazil, were relatively poorly documented (c. 55–65% of species photographed). Some of these ‘missing’ species have actually been photographed, with the images remaining unidentified or not accessible in a major online database (Pitman et al., 2021), but many have genuinely never been photographed in the field. Whilst the driving factors behind these true absences are not entirely clear, highly endemic species and those with small populations are generally less likely to be photographed (Pitman et al., 2021).

As one of 17 ‘megadiverse’ countries (Mittermeier & Mittermeier, 1997; Steffen et al., 2009), and containing 2 of 20 global centres of vascular plant diversity (Rafiqpoor et al., 2005) and 2 of 35 global biodiversity hotspots (Mittermeier et al., 2011), Australia has one of the richest vascular plant floras in the world, with an estimated diversity of > 21 000 native species (Chapman, 2009). Given that > 91% of these species are estimated to be endemic to Australia (Chapman, 2009), and that c. 1400 Australian plant taxa are listed as threatened or extinct (Threatened Species Recovery Hub, 2020), understanding the photographic record of Australian vascular plants is important for research and conservation; it can be difficult to accurately identify species for which field photographs do not exist (Pitman et al., 2021). To our knowledge, Australia's online vascular plant photographic record has never been surveyed. Australia is an ideal region to survey not only because of its large total number of native species and the broad diversity of biomes in which they occur, but also the valuable resources available for their inventory (Australian Plant Census, biodiversity.org.au/nsl/services/search/taxonomy) and occurrences (Australasian Virtual Herbarium, avh.chah.org.au; Atlas of Living Australia, www.ala.org.au).

Here, we: compile a list of described, native, Australian vascular plant species; conduct a photographic census of well-curated online resources for these species; assess factors determining whether any given species has been photographed; and discuss the importance of field photographs.

Materials and Methods

We downloaded a complete list of all Australian vascular plant taxa on 14 February 2022 from the Australian Plant Census (APC; Supporting Information Table S1), an online database informing which scientific names for Australian vascular plant species are currently accepted, which names are currently regarded as synonyms, and state or territory level distributions (Cowley et al., 2011). This list contains all taxa recognised as occurring in at least one Australian state, territory or external territory, namely Western Australia, the Northern Territory, Queensland, New South Wales, the Australian Capital Territory, Victoria, Tasmania, South Australia, Cocos (Keeling) Islands, Christmas Island, Ashmore Reef, Cartier Island, Coral Sea Islands, Norfolk Island, Lord Howe Island, Macquarie Island, Heard Island and McDonald Islands (Table S2). This list covers all species described in 2018 or earlier, although four species described after 2018 (Thelymitra tubulina R.J.Bates, Pecteilis dockrillii D.L.Jones & M.A.Clem. and Diuris brockmanii D.L.Jones & C.J.French in 2019, and Cadetia maideniana Schltr. ex Kellermann in 2021) were also present. We removed supraspecific taxa, infraspecific taxa, hybrids and taxa with phrase names (formulaic names used to document undescribed, putatively new plant species; Chapman, 2005), so that only formally described taxa at the level of species were retained. We removed all species recognised only as ‘naturalised’ or ‘doubtfully naturalised’, so that only species recognised as ‘native’ or ‘native and naturalised’ in at least one state or territory were retained. We removed 46 species that were erroneously annotated as native in the APC list (Table S3; note that we retained species for which considerable uncertainty exists regarding their native status in Australia) and added two species (Diuris leopardina D.L.Jones & R.J.Bates and Baumea articulata (R.Br.) S.T.Blake) that were accepted by the APC at the time of analysis but did not come across in the download. We also amended location data for nine species (Table S3). The final list contained 21 077 native, vascular plant species across 2190 genera and 259 families (Table S4).

We selected 33 well-curated, easily searched online resources containing images of native Australian plant species (Table S5) and surveyed them for the species on our final list. Although the majority of our surveyed resources are state herbarium portals, or identification resources created by professional botanists and taxonomists, we also surveyed several reputable personal online databases and citizen science platforms, such as iNaturalist. We acknowledge that some taxa are typically not identifiable from photographs and that, in some cases, photographs may be misidentified, especially on the citizen science platforms we surveyed. However, many of these cases are rarer species misidentified as common species (Mesaglio et al., 2023), which are already well-represented within the photographic record, and thus, such misidentifications may actually cause an underestimate of the number of photographed species. Once an image was found for any given species, that species was no longer searched for in subsequent surveys of other resources. To qualify as accepted, an image was required to:
  1. Be a photograph or a digital scan of a photographic slide. Illustrations and paintings were excluded.
  2. Be taken in Australia (although an exact georeferenced location was not required). Photographs of Australian species with extra-Australian distributions taken in other countries (e.g. New Zealand and Papua New Guinea) were excluded.
  3. Depict a live plant. Photographs of herbarium sheets or otherwise pressed or preserved specimens were excluded. Photographs of collected specimens that had clearly been uprooted immediately before photography were accepted.
  4. Depict a significant part of the plant that could be recognised in the field. Photographs of features such as stem cross-sections or macro shots of seeds were excluded.
  5. Be identified without indications of uncertainty, that is photographs with annotations such as ‘aff.’, ‘?’, ‘possibly’ or ‘cf.’ were excluded.

We surveyed iNaturalist first due to its broad taxonomic and geographic scope, the large number of observed Australian species on the platform and its dedicated export tool, which allowed for downloads of large quantities of data. We accepted all verifiable records, that is georeferenced and dated observations. In addition to ‘Research Grade’ observations (Mesaglio & Callaghan, 2021), we also accepted ‘Needs ID’ observations; many such records (especially rare species for which only one photograph or photographic set exists) have been uploaded by taxonomists, botanists and other experts, including photographs associated with physical vouchers, and are reliably labelled despite currently lacking a confirming identification. Because iNaturalist largely follows the taxonomy used by Plants of the World Online (i.e. The World Checklist of Vascular Plants; Govaerts et al., 2021), we standardised all names to APC taxonomy where conflicts existed by matching names to synonym data downloaded as part of the APC list. The Atlas of Living Australia (ALA) was surveyed second, also due to its broad scope and species coverage, and the ability to execute large queries using the R package galah (Stevenson et al., 2022). A list of the remaining unphotographed species present in Florabase (Western Australia herbarium website), FloraNT (Northern Territory herbarium) and PlantNet (New South Wales herbarium) was generated programmatically in R using the package rvest (Wickham, 2022) and filtered to those with images on their webpages. Using the package downloader (Chang, 2015), the images associated with each species were downloaded and manually checked to exclude illustrations and photographs of herbarium sheets. All other resources were surveyed by manually searching for each putatively unphotographed species within the relevant taxonomic and spatial scope (Table S5 for notes on resource type, scope and survey methods used). Images for 74 species were found under synonyms of their accepted names (Table S6). All surveys were conducted between 28 March and 15 April 2022.

Family names and original dates of publication for each species were obtained from the Australian Plant Name Index (biodiversity.org.au/nsl/services/search/names). These data were downloaded on 27 May 2022. Growth habit data for each species were obtained from the development branch of AusTraits (Falster et al., 2021) and recoded into nine growth habits (Table S7). One species, Marsdenia tubulosa F.Muell., was excluded as its growth habit is unknown. We divided Australia into 1° × 1° (c. 12 000 km2) grid cells and, for each cell, counted the number of unphotographed species recorded at least once in that cell based on ALA records. We defined a cell as a hotspot if > 5% of the total number of unphotographed species have been recorded in it. All analyses of geographic distribution, growth habit, publication date and higher taxonomy were conducted in R v.4.1.2 (R Core Team, 2021).

Results

Of 21 077 native Australian vascular plant species (Fig. 1), 3715 (17.6%) species across 832 genera and 158 families were unphotographed (Table S8) across 33 online resources. Conversely, 17 362 (82.4%) species are represented by at least one field photograph in at least one of 33 online resources (Table S9). There are 1358 photographically ‘complete’ genera, that is all species within each of these genera are photographed, and 101 photographically complete families (Table S10). Only two genera containing > 40 species are photographically complete: Banksia (Proteaceae; 176 species) and Hakea (Proteaceae; 151). Broadly, charismatic genera tend to have comprehensive photographic records, such as Grevillea (Proteaceae; 343/367) or Verticordia (Myrtaceae; 98/101), whilst less charismatic taxa tend to be poorly represented in the photographic record relative to their diversity, such as Deyeuxia (Poaceae; 11/34) or Spermacoce (Rubiaceae; 33/63). There are 98 unphotographed genera (Table S10), that is none of the species in these genera are photographed, but zero unphotographed families. In absolute terms, the genera with the most unphotographed species are Acacia (Fabaceae; 98 unphotographed species), Hibbertia (Dillenaceae; 78), Fimbristylis (Cyperaceae; 57), Solanum (Solanaceae; 54) and Leucopogon (Ericaceae; 53), and the families with the most unphotographed species are Poaceae (343), Fabaceae (324), Asteraceae (265), Myrtaceae (260) and Cyperaceae (254). In relative terms, for genera containing 50 or more species, the least complete genera photographically are Fimbristylis (36.7% of 90 species photographed), Triumfetta (Malvaceae; 40.7% of 59 species), Mitrasacme (Loganiaceae; 46% of 50 species) and Heliotropium (Boraginaceae; 47.7% of 88 species).

Details are in the caption following the image
Representative species from Australia's most diverse native plant genera. (a) Acacia pyrifolia, © David Muirhead; (b) Eucalyptus gongylocarpa, © Dean Nicolle; (c) Grevillea venusta, © Graham Corbin; (d) Caladenia filifera, © Stephen Buckle; (e) Stylidium humphreysii, © Robert Davis; (f) Hibbertia riparia, © Tim Hammer; (g) Melaleuca quinquenervia, © Thomas Mesaglio; (h) Pterostylis concinna, © Michael Keogh; (i) Eremophila longifolia, © David H. Fischer. All photographs are from iNaturalist.

Whilst southeastern Australia (New South Wales, the Australian Capital Territory, Victoria and Tasmania) is comprehensively (c. 95+% of species) documented, Western Australia and the Northern Territory each only just exceed 80% completeness (Fig. 2). Western Australia in particular is the great frontier for unphotographed Australian plant species: 51.9% (1927) of all unphotographed species can be found in Western Australia. Of those 1927 species, 80.7% (1555) are endemic to Western Australia. Unsurprisingly, Western Australia also has the highest total vascular plant diversity of any Australian state or territory with 10 269 species, followed by Queensland (8139) and New South Wales (5821). Aside from Cartier Island, for which only a single vascular plant species has been recorded (the seagrass Thalassia hemprichii), there are no Australian states or territories for which 100% of their recorded vascular plant species have been photographed.

Details are in the caption following the image
Geographic summary of Australia's native vascular plant photographic record. (a) Relative positions of each external territory to mainland Australia; (b) mainland Australia and Tasmania; (c) sub-Antarctic islands; (d) tropical islands; (e) subtropical islands. For each state and territory, the first number is the percentage of species represented by at least one field photograph in at least one of 33 online resources, and the number in parentheses is the total number of recorded species. Note the external territories are not depicted to scale in (c–e).

There are 16 hotspots for unphotographed species across Australia (Fig. 3). The most significant hotspot spans Kakadu National Park and West Arnhem Land in the Northern Territory; > 9% of all unphotographed species in Australia have been recorded in this region on the ALA. The second and third greatest hotspots cover Kakadu National Park (further south of the first hotspot) and Stirling Range National Park in southwestern Western Australia respectively, followed by Nitmiluk National Park south of Kakadu, Wooroonooran National Park in northern Queensland and Fitzgerald River National Park in southwestern Western Australia (Fig. 3). There are also seven ‘coldspots’ for which all species recorded there in the ALA have been photographed, distributed across remote regions of Western Australia and the Northern Territory in the Great Sandy, Little Sandy, Gibson, Simpson and Great Victoria Deserts. It is likely that some of these coldspots, and other areas with low numbers of unphotographed species, are influenced by low species occurrence sampling, especially in remote and difficult-to-access regions.

Details are in the caption following the image
Geographic distribution of unphotographed Australian vascular plant species (n = 3714). Each value represents the percentage of all unphotographed species for which there is at least one record from the Atlas of Living Australia in that cell. Each grid cell has a resolution of 1° × 1° (c. 12 000 km2).

Most unphotographed species were described relatively recently: 35% of all unphotographed species were originally described from 2000 onwards (Fig. 4). Across all unphotographed species, 37.8% are shrubs, and just three of nine growth habits (shrubs, herbs and graminoids) constitute 88.6% of all unphotographed species. Herbs and graminoids have disproportionately more unphotographed species relative to their total diversity; the latter constitute 16.1% of unphotographed species despite making up just 8.2% of all species. Trees show the opposite pattern, constituting just 5.7% of unphotographed species vs 16.7% of all species (Fig. 5).

Details are in the caption following the image
Distribution of original years of publication for all unphotographed Australian vascular plant species (n = 3714).
Details are in the caption following the image
Growth habit composition for all Australian vascular plant species (closed; n = 21 076) and unphotographed Australian vascular plant species (open; n = 3714). Original growth habit data were obtained from AusTraits and then recoded (Supporting Information Table S7). Note that Marsdenia tubulosa was omitted from this figure as its growth habit is unknown.

Discussion

We present the first photographic census of Australia's native vascular plant flora and show that c. 18% of Australian native species lack a verifiable photograph in one of 33 online resources. The global number is unknown; however, 18% is more than double the percentage of unphotographed species for the United States and Canada, but less than half of that for Mexico (Pitman et al., 2021). Building on previous work, we also identify the drivers of these photographic gaps for Australia. Most unphotographed species are found in Western Australia, Queensland, or the Northern Territory, are shrubs or herbs, are in the families Poaceae or Fabaceae and were described within the last 30 yr. The large number of recently described unphotographed species suggests that these gaps arise, at least in part, from a failure to recognise the photographic record as an important biodiversity resource. Moreover, citizen scientists, who are rapidly filling many gaps in our biodiversity data, may nonetheless be slow to find and photograph species with certain characteristics.

Photographs of vascular plants are crucial tools in conservation and research, facilitating easier recognition of species in the field, and forming the cornerstone of field guides, identification keys and public engagement with plant science (Pitman et al., 2021; Wäldchen et al., 2022). Although attributes important for field identification are typically captured in field notes associated with vouchered specimens, this is not a universal practice. Field photographs are thus invaluable for complementing physical vouchers and contributing towards an ‘extended specimen’ (Lendemer et al., 2020) by capturing information such as growth habit, orientation of leaves or branchlets and colour, which, in some cases, cannot be retrospectively determined from pressed specimens (Heberling & Isaac, 2018). Field photographs, especially those from citizen science datasets, are also increasingly being utilised in largescale trait-based research, with plant images already used to evaluate potential impacts of climate change on phenology (Puchałka et al., 2022), delineate unusual flowering events (Barve et al., 2020) and map global trait patterns (Wolf et al., 2022). With 136 unphotographed Australian species listed as Critically Endangered, Endangered or Vulnerable under the Environment Protection Biodiversity Conservation Act (Table S11), it is possible that some Australian species may go extinct without ever being photographed in the field.

Although our methods precluded an exact quantification of the number of photographs for each species across all 33 surveyed resources, it is clear that some species are represented by a comprehensive photographic record spanning hundreds, or even thousands, of images, whilst others are represented by just a single image. Efforts to improve the vascular plant photographic record should therefore focus not only on unphotographed species, but also poorly represented species, and should aim to both increase the number of photographs for each species and capture a standardised set of diagnostic features to maximise identifiability (Baskauf & Kirchoff, 2008; Gómez-Bellver et al., 2019; Rzanny et al., 2019, 2022). This verified set of photographs is likely to then facilitate future identification by both professional and citizen scientists, as well as the improvement of automated species identification and phenological data extraction through the use of, for example, deep convolutional neural networks (Mäder et al., 2021; Reeb et al., 2022; Yang et al., 2022), leading to more verified photographs and better data on spatial–temporal distributions and phenology. The development of a centralised photographic repository may improve the discoverability of online plant photographs (Pitman et al., 2021) and better facilitate efforts to complete the photographic record for underrepresented species.

Citizen science, especially iNaturalist, is one of the most important drivers of the photographic documentation of vascular plants, with almost 50 million observations uploaded to iNaturalist. As of 13 April 2022, when we finalised our iNaturalist data download, photographs for 57.1% of all Australian vascular plant species had been uploaded to the platform, an impressive statistic given iNaturalist only started to become well-used in Australia in mid-late 2016 (Mesaglio & Callaghan, 2021). Indeed, in the 9-month period after our iNaturalist data download, photographs for an additional 1230 Australian plant species (including 214 species without photographs in any of our surveyed resources) were identified or uploaded to the platform. However, professionally collected photographs deposited in national and regional online floras, which typically depict vouchered specimens identified by botanists and thus are associated with an especially high-veracity identification, are also a crucial component of the photographic record: a quarter of Australian vascular plant species have photographs only available through these resources. There is also a considerable overlap in the species recorded by iNaturalist and noncitizen science resources.

Why are some plant taxa more likely to be photographed than others? Gaps in the Australian photographic record are mainly driven by three factors: plant charisma (lack thereof); geographic difficulty; and identification difficulty. These biases are especially prevalent in citizen science, where photographic subject choice is often influenced by attributes such as stem height, flower colour or identifiability (Boakes et al., 2016; Seregin et al., 2020; Marcenò et al., 2021).

Many of Australia's most recognised and charismatic genera have comprehensive, or even complete, photographic records, and many species within these taxa are shrubs or trees and have spectacular floral displays. Conversely, small-statured herbs or graminoids with small or dull flowers tend to be more poorly documented (Caley et al., 2020). It is generally easier to notice and photograph a taller plant such as a tree (Boakes et al., 2016) vs a small herb or grass, especially if the latter is not flowering. For graminoids, this becomes an issue of both detectability and charisma, with ‘grass blindness’ a pervasive phenomenon (Thomas, 2019; Marcenò et al., 2021).

The three most notable Australian regions for unphotographed species – northern Australia spanning from the Kimberley to Arnhem Land, Queensland's Wet Tropics World Heritage Area and southwestern Western Australia – are characterised by often large distances from population centres, remote areas and difficult-to-access environments, including those with poor or no road infrastructure. The nature of these hotspots is not unexpected given the strong biases towards main roads and roadside habitats in many citizen science datasets (Geldmann et al., 2016; Croce & Nazzaro, 2017; Hughes et al., 2021).

Taxa that are difficult to identify also tend to be underrepresented in the photographic record. This may be due to the similarity of many species and the difficulty of capturing necessary diagnostic characters in photographs, or the disincentivisation to either photograph, or attempt to identify from photographs, taxa renowned as hard to identify. For example, species identification within Lepidosperma (Cyperaceae, 38/75 species photographed) can be difficult even from specimens if material is incomplete (Barrett & Wilson, 2012), let alone from photographs, and grasses are often difficult to identify from photographs, especially when not flowering (Rzanny et al., 2019, 2022).

These three driving factors are also more broadly applicable to the vascular plant photographic records in other regions. For example there are almost 1500 unphotographed species in the genus Piper (Piperaceae) in the Americas (Pitman et al., 2021); many of these species are relatively plain and lack charismatic flowers, are narrow range endemics in remote regions and are difficult to differentiate from each other.

Despite the rise of digital photography over the past three decades, more recently described species are more likely to be unphotographed across our 33 surveyed resources. Among unphotographed species, almost half were described recently – from 1993 to 2018, with four possible reasons why a species described in an era with cheap and easy photography may still lack a photograph. First, new species may be described solely based on old collections (Bebber et al., 2010), so the taxonomists describing them may have never seen them in the field. Second, new species may be illustrated with line drawings rather than field photographs. Third, some species have photographs on social media platforms such as Facebook, or personal websites such as blogs. It is often complicated to search for and scrape data from these platforms (Marcenò et al., 2021), and sometimes difficult to assess the accuracy of these records. Fourth, there are also many cases where the species on our unphotographed list have been photographed, but the images have not been uploaded to any of our 33 surveyed resources. These photographs may be associated with print field guides, or the original description or revisions published in scientific journals. Whilst many of these papers are open access, some remain behind paywalls, which inhibits access by many potential identifiers (Penev et al., 2008). The trend toward open-access publication may partially solve this problem (Bénichou et al., 2018), but the discoverability of the photographs by potential identifiers is still an issue. We therefore make a four-part recommendation critical to the improved photographic documentation of the vascular flora: the open-access publication of new species descriptions; the greater promotion of uploading photographs, including of newly described species, to publicly accessible, easily searchable online databases, such as iNaturalist, that keep track of metadata and allow use of the photographs for construction of identification resources; the application of copyright licenses to any uploaded photographs that will maximise their use in research, such as CC BY or CC BY-SA (Hagedorn et al., 2011); and the inclusion of an explicit requirement in the International Code of Nomenclature for algae, fungi and plants (Turland et al., 2018) for field photographs of newly described plant species in all new species descriptions, where taking such photographs is feasible.

Conclusions

Despite the impressive recent growth and potential of searchable online databases to build a photographic record of the Australian flora, this study highlights how incomplete this biodiversity resource currently is. We suspect that Australia is not a global outlier in this sense (Pitman et al., 2021) and that upwards of 100 000 vascular plant species globally would similarly lack a findable photograph. This shortfall is crucial because it massively inhibits identification and monitoring for these species. We argue that photographs should be seen and quantified alongside other data resources (Hortal et al., 2015) and, ideally, integrated as part of an ‘extended specimen’ (Heberling & Isaac, 2018; Lendemer et al., 2020). The combined efforts of citizen scientists and botanists are crucial for the continued documentation of vascular flora.

Acknowledgements

We thank Anna Monro for providing invaluable information and data from the Australian Plant Census and the Australian Plant Image Index, and Hayden Schilling and Sophie Yang for providing useful code. Mason Heberling and three anonymous reviewers provided thoughtful comments that substantially improved this manuscript. Open access publishing facilitated by University of New South Wales, as part of the Wiley - University of New South Wales agreement via the Council of Australian University Librarians.

    Competing interests

    None declared.

    Author contributions

    TM, HS and WKC contributed to the research design. TM contributed to the performance of the research and first draft. TM, WKC, DC and EW contributed to the data analysis and collection. HS, WKC, DC and EW contributed to the manuscript comments and edits.

    Data availability

    All data and code are available at https://github.com/tmesaglio/Australian-plant-photos.