New open-source workflow turns biodiversity data into alien species impact indicators

Invasive alien species are among the leading drivers of biodiversity loss worldwide, yet conservation managers have long lacked a standardised, evidence-based way to measure how much harm these species actually cause across landscapes and through time. A team of researchers writing in the journal NeoBiota has now unveiled an open-source workflow, packaged in an R library called impIndicator, that converts raw biodiversity records and standardised impact assessments into measurable impact indicators for alien species. The tool was demonstrated using Acacia species in three Mediterranean-climate regions: the Iberian Peninsula, South Africa and California.

The negative impacts of alien species are recognised as a major threat to biodiversity and to the ecosystem services that underpin human well-being. Governments and organisations worldwide are developing policies to prevent new introductions, control existing invaders and restore affected ecosystems, but these policies require standardised decision-support tools to monitor changes in impacts and evaluate whether control measures are working. Impact indicators, defined as measurable proxies that quantify the impacts of alien species on native ecosystems across space and time, are essential for informing biological invasion policies and management strategies. Until now, however, most available indicators relied purely on occurrence data, such as the cumulative number of established alien species or their relative abundance. While occurrence-based indicators provide a general measure of invasion trends, they do not necessarily reflect changes in ecological impacts, nor do they account for variation in impact magnitude amongst alien species.

The new workflow addresses this gap by integrating two complementary data streams. The first is occurrence data from the Global Biodiversity Information Facility (GBIF), a large biodiversity data infrastructure that supports the Findable, Accessible, Interoperable and Reusable (FAIR) Data Principles. The workflow uses an occurrence cube format, which aggregates records along spatial, temporal and taxonomic dimensions, making the data suitable for modelling and computational analysis. To minimise the influence of sampling bias, multiple records of a species from the same grid cell within a given year are converted to binary presence or absence records for all computations. The second data stream consists of assessments from the Environmental Impact Classification for Alien Taxa (EICAT), a framework recently adopted by the IUCN that classifies alien species impacts from Minimal Concern to Massive, depending on the reported magnitude of their negative effects on native species. A species is assigned Minimal Concern if there is no reduction in the performance of individuals of a native species, Minor if individual performance is reduced without a population decline, Moderate if populations are reduced, Major if impacts have led to reversible local extinction, and Massive if they have caused naturally irreversible extirpation or extinction. Impacts are also classified according to twelve mechanisms described by the IUCN, including competition, predation, hybridisation, disease transmission, parasitism, toxicity, biofouling, grazing, chemical, physical and structural impacts on ecosystems, and indirect impacts through interactions with other species.

For computational purposes, the semi-quantitative EICAT categories must first be transformed into numerical scores. The researchers considered three encoding schemes: a linear scheme treating Minimal Concern as zero impact, a linear scheme treating it as a non-zero category reflecting some degree of concern, and an exponential scheme giving disproportionate weight to higher-impact categories. For all analyses in the study, they used the default encoding scheme in impIndicator, which assigns scores of zero to four across the five categories, consistent with the original EICAT classification. The choice of encoding matters because it shapes how strongly severe impacts weigh on the final indicator values.

The workflow then applies a hierarchical aggregation framework with three levels: species, site and region. At the species level, because species are often assigned multiple impact categories across study locations and impact mechanisms, the records are aggregated to obtain a single species-level impact score. Three methods are available: Maximum, which assigns a species the highest recorded impact across all available scores; Mean, which computes the average impact across all scores; and Sum of Maximum across Mechanisms, which sums the maximum impact for each mechanism through which the species exerts its impacts. Each method embodies different assumptions. The Maximum method is consistent with a precautionary framework, assuming that the highest observed impact best reflects management concern and that rare but severe impacts matter more than average ones, but it can substantially overestimate impacts and is sensitive to exceptional cases. The Mean method provides an expected impact value appropriate when numerous records exist, but can underestimate species capable of causing severe impacts. The mechanism-based sum favours species affecting ecosystems through many independent pathways, though it depends heavily on the completeness of assessments and may double-count overlapping ecological effects.

At the site level, the impacts of all alien species recorded at each grid cell are aggregated using one of five methods: Precautionary, which computes the maximum impact across species present at a site; Precautionary Cumulative, which sums the maximum scores of all co-occurring species; Mean of Means, which averages species-level means; Mean Cumulative, which sums the mean scores of all species at a site; and Cumulative, which sums mechanism-based scores across species. The regional impact indicator is finally computed as the spatial sum of site-level impacts for each year, using the same five aggregation approaches. Because each method employs a different aggregation strategy, the resulting indicators can produce substantially different estimates. For example, when all species considered have been classified as causing Major or Massive impacts at least once, the Precautionary site-level indicator may show little variation amongst sites, whereas the Mean Cumulative method incorporates the contributions of all species and their average impacts, allowing greater variation and providing a representation of cumulative invasion pressure.

To estimate the uncertainty associated with the indicators, the researchers used bootstrapping. For each year in the occurrence cube, the procedure re-samples individual records with replacement and re-computes the indicator for each re-sampled dataset, yielding a distribution of possible values from which standard errors and confidence intervals are derived. To identify species that disproportionately influence the regional indicator, they applied a leave-one-species-out cross-validation technique, which systematically recalculates the indicator after excluding one species at a time and compares the resulting values to those from the full dataset. The percentage deviation provides a measure of each species’ contribution to the regional impact indicator.

The demonstration with Acacia species revealed striking regional differences. Eleven unique Acacia species with impact evidence occur across the three regions, including Acacia dealbata, Acacia mearnsii, Acacia baileyana, Acacia cyclops, Acacia longifolia and Acacia saligna. Acacia dealbata had the highest impact value in the Iberian Peninsula, Acacia mearnsii in South Africa and Acacia baileyana in California. The leave-one-species-out analysis showed that Acacia dealbata alone contributed approximately 55% to the regional impact indicator in the Iberian Peninsula in 2010, underscoring how strongly regional indicators can depend on particular species. Regional impact values were higher in the Iberian Peninsula and South Africa than in California, aligning with independent assessments by the California Invasive Plant Council, which classifies most Acacia species there as Watch species with potential future rather than current significant impact. The Iberian results also revealed substantial spatial variability in impact across the Peninsula, showing that a widespread distribution does not necessarily translate into uniformly high impact, a heterogeneity especially relevant for management planning in Portugal, where all Acacia species are subject to legal restrictions.

The researchers caution that the indicators are sensitive to variation in sampling effort and the completeness of the underlying occurrence data. Since EICAT impact scores assigned to each species remain constant through time, increases in impact values mainly reflect additional occurrence records, which may indicate either genuine spread to new sites or increased recording effort. The apparent rate of spread inferred from the indicators differs from aerial survey estimates in South Africa, most likely because of differences in sampling effort and data sources, including the incorporation of citizen-science observations into GBIF. Observed trends may also reflect delays between field observations and publication, the exclusion of datasets that are not openly available, and non-systematic sampling. The workflow additionally assumes that impact magnitudes recorded in one location are transferable to other locations, an assumption that may not hold in all ecological contexts, and the indicators therefore represent potential rather than realised impacts. Another limitation is the current lack of EICAT assessments for most alien taxa, and the assessments that do exist are not yet fully open or FAIR-compliant.

Despite these caveats, the workflow offers a robust framework for comparative assessments. Absolute indicator values should not be interpreted in isolation, but comparisons across species, sites, regions or time using the same methodological settings provide meaningful insights into where alien species pose the greatest threats and whether interventions are effectively reducing them. The information is directly relevant to policy frameworks, including Target 6 of the Convention on Biological Diversity’s Kunming-Montreal Global Biodiversity Framework, which aims to eliminate, reduce and mitigate the negative impacts of alien species on biodiversity and ecosystem services. Future versions of the package will extend uncertainty calculation to species-level and site-level indicators, and recent advances in deep learning species distribution models could help project potential impacts across space and time, reducing biases associated with uneven sampling. By highlighting species and regions where impacts may be greatest, the approach can help guide management actions and prioritise efforts to mitigate biological invasions, contributing to the protection of biodiversity and the maintenance of healthy ecosystems.

Subject of Research: A workflow for impact indicators of alien species: A demonstration with Acacia species

Article Title: A workflow for impact indicators of alien species: A demonstration with Acacia species

Article References: A workflow for impact indicators of alien species: A demonstration with Acacia species. (n.d.). https://doi.org/10.3897/neobiota.109.193861

Image Credits: AI Generated

DOI: 10.3897/neobiota.109.193861

Keywords: invasive species, biodiversity, impact indicators, EICAT, GBIF, Acacia, biological invasions, conservation policy, open-source software, Kunming-Montreal Global Biodiversity Framework, workflow, impact