The results demonstrate how community records can reveal extensive plant invasions while also showing why poorly sampled towns cannot be interpreted confidently from raw species lists. The Western Cape provided a useful setting because iNaturalist has attracted substantial plant-observation activity in the province. By October 2024, South Africa had more than 2.7 million plant records on the platform, including 1.7 million concentrated in the Western Cape, with a disproportionate share originating in urban areas.

Researchers used built-up areas identified in South Africa’s National Land Cover database to define 100 urban centres. They exported more than 300,000 vascular-plant observations made after 2017 that included at least one photograph. Records with stated positional accuracy greater than 1,000 metres were excluded because they were unlikely to fall within the urban boundaries; approximately 2% of records met that condition.

Records without a stated accuracy were retained because they were assumed to fall within the designated areas. Central to the study was the semi-Automated Non-Native Inventory Compilation workflow, or sANNIC, which operates in four stages: acquiring iNaturalist observations, compiling an automated non-native list, creating and curating an iNaturalist project for the resulting taxa, and evaluating the inventory’s completeness. The workflow’s R-based Automated Non-Native Inventory Compilation script standardized names and matched them to the World Checklist of Vascular Plants.

Native status was then evaluated using the checklist’s botanical-country distributions. For this case study, only taxa recorded as native to the Cape Provinces were classified as native; taxa with native ranges elsewhere, or introduced ranges in the selected region, were classified as non-native. Taxa with difficult species-level distinctions, hybrids, unclear names or missing distribution information were flagged for manual assessment rather than accepted automatically or silently discarded.

Automated classification produced both useful matches and evidence of where an entirely hands-off workflow would fail. Of 4,467 vascular-plant taxa considered, 4,405 received a World Checklist match, including 4,364 exact matches, 34 edit-distance fuzzy matches and seven phonetic matches. Multiple possible matches were returned for 496 taxa, while 62 received no match.

Researchers manually checked 126 matches involving 108 taxa, resolving cases that included obscure hybrids, typographical errors, non-standard names and taxa not yet listed in the checklist. Review also exposed errors in the checklist’s geographic information. Seven taxa were incorrectly labelled as non-native and removed, including Pteridium aquilinum, Tarchonanthus littoralis, Tetragonia fruticosa, Scabiosa columbaria, Solanum rubetorum, Senecio pinifolius and Salicornia pilansii.

Cotula australis and Rumex acetosella were added after their non-native status was recognized during local review. The preliminary taxa were then loaded into an iNaturalist project so that individual observations could be reviewed on the platform where corrections would also improve the underlying community dataset. Curation addressed both identification and cultivation status.

Images, observation dates, positional information, observer notes, environmental context and signs of care were examined to determine whether plants were growing wild or remained cultivated. This distinction is central because an ornamental specimen in a garden does not establish the presence of a wild-growing alien population. Taxonomic experts and the authors reviewed unresolved or conflicting identifications, and records that could not be identified reliably from typical photographs were retained at a broader rank, such as genus, subfamily or section.

The curation effort increased the proportion of conventional project records reaching Research Grade from 86.4% to 94%, excluding the separately retained problematic taxa. Completeness was assessed through sample-coverage profiles at orders q = 0, 1 and 2. The q = 0 measure estimates the proportion of all species, including rare ones, that was observed.

The q = 1 measure estimates sample coverage for the detected species, while q = 2 gives greater weight to abundant species. For the 100 urban centres combined, q = 0 sample completeness was 0.7823, with a 95% confidence interval of 0.7314–0.8331; q = 1 sample coverage was 0.9958, with a 95% confidence interval of 0.9953–0.9962; and the q = 2 result was close to one. These estimates indicate that almost all relatively abundant alien plants were represented, whereas some uncommon species remained undetected.

Indeed, 219 taxa were observed only once and 91 only twice, making rare-species records especially influential in the completeness calculation. Observation effort was highly uneven among the 100 centres. Cape Town supplied 31,034 records and 686 species, while Stellenbosch contributed 13,689 records and 410 species.

Only 85 towns had at least two records, the theoretical minimum for species estimation; 57 had 20 or more records and 25 had more than 100. Thirteen centres achieved q = 1 sample coverage above 0.8, whereas 37 had q = 2 values above 0.8. Cape Town, Stellenbosch, Somerset West, George and Mossel Bay had the highest q = 1 coverage.

These figures did not mean that the other towns contained fewer alien plants. Rather, declining completeness profiles, wide uncertainty intervals or very few records indicated that a reliable completeness estimate was unavailable, leaving sampling gaps as a plausible explanation for observed differences. To rank taxa without simply rewarding centres with the largest number of observations, the researchers developed a prevalence index.

It sums each taxon’s inverse squared rank within an urban centre and weights the contribution by that centre’s q = 1 sample coverage. Acacia cyclops ranked first, followed by Acacia saligna, Plantago lanceolata, Echium plantagineum and Acacia mearnsii. Lantana camara, Tropaeolum majus and Pinus were also among the leading taxa.

Fabaceae, Asteraceae, Poaceae, Solanaceae and Myrtaceae contained the greatest numbers of species, while annual plants were the most common life-form category. Many prominent taxa were alien trees associated with natural or semi-natural vegetation adjoining or embedded within cities, whereas others were widely planted garden species that had spread into urban green spaces. Comparison with the Global Naturalized Alien Flora, or GloNAF, exposed important gaps in existing invasion compilations.

Of the 947 inventoried taxa, 404, or 43%, were represented as naturalised or invasive in the Western Cape by GloNAF; 588, or 62%, appeared as naturalised or invasive somewhere in South Africa, and 856, or 90%, appeared globally. Ninety-one taxa were absent from GloNAF altogether. Some differences may reflect classifications of species as casual rather than naturalised or invasive, while others may represent omissions in previous regional assessments.

The open iNaturalist project and accompanying code now provide a continuously updateable inventory rather than a fixed list. The authors caution, however, that the method requires substantial expert curation, and its completeness estimates remain sensitive to geographic and seasonal sampling bias, observer preferences, plant detectability, cultivation-status errors and imperfectly independent community observations.

Subject of Research: An observational workflow for compiling and assessing non-native vascular plant inventories from iNaturalist across 100 urban centres in South Africa’s Western Cape province

Article Title: Deriving inventories of non-native plant species from iNaturalist: Insights from urban centres of the Western Cape, South Africa

Article References: Gildenhuys, C. P., Potgieter, L. J., Hui, C., & Richardson, D. M. (2025). Deriving inventories of non-native plant species from iNaturalist: Insights from urban centres of the Western Cape, South Africa. NeoBiota, 104, 27-58. https://doi.org/10.3897/neobiota.104.155832

Image Credits: AI Generated

DOI: 10.3897/neobiota.104.155832

Keywords: iNaturalist, biological invasions, non-native plants, citizen science, Western Cape, urban biodiversity, plant inventories, sample completeness, taxonomic curation, South Africa, Deriving, inventories