Trends and patterns in the extinction risk of Australia’s birds over three decades

ABSTRACT Australia recently committed through the Kunming-Montreal Global Biodiversity Framework (GBF) to halt human-induced extinction of known threatened species and to reduce extinction risk of threatened species significantly by 2030. We review recent trends in extinction risk of Australian birds to provide context for current and future conservation efforts. We calculate the Red List Index (RLI) for all Australian birds as well as subsets based on geography, habitat and taxonomy. Over the period 2010 to 2020, the number of taxa reassigned to lower categories of extinction risk (n = 20; 1.5% of all taxa included) was greatly outweighed by the number moved to higher categories owing to deteriorating status (n = 93; 7%). This resulted in the steepest decadal decline in the RLI since data were first compiled in 1990. It was chiefly driven by rapid population declines in migratory shorebirds, loss of suitable habitat for species affected by wildfire in 2019–2020 and, to a lesser extent, declines in the abundance of upland rainforest birds. To a small extent, these losses were counterbalanced by improvements in status of some bird species resulting from local eradication of invasive mammals, primarily from Macquarie Island. For Australia to meet the commitments recently adopted through the GBF, conservation interventions (and hence funding) will need to be scaled up substantially. The RLI is well placed for monitoring progress towards the GBF targets and for communicating trends in the extinction risk to national avifaunas. POLICY HIGHLIGHTS Australia is committed to reducing extinction risk through its adoption of the Kunming-Montreal Global Biodiversity Framework. A Red List Index (RLI) for all Australian birds from 1990 to 2020 shows their extinction risk increased by 3.34%. More than 50% of the increase in extinction risk between 2010 and 2020 was caused by the 2019–2020 wildfires. The greatest increases in overall avian extinction risk were in Queensland, South Australia and New South Wales, where drought and wildfire effects were most pronounced. Red List Index trends can reflect the impact of individual threats or conservation interventions, and represent an important tool for monitoring national and global progress towards international agreements.

• Australia is committed to reducing extinction risk through its adoption of the Kunming-Montreal Global Biodiversity Framework.• A Red List Index (RLI) for all Australian birds from 1990 to 2020 shows their extinction risk increased by 3.34%.• More than 50% of the increase in extinction risk between 2010 and 2020 was caused by the 2019-2020 wildfires.• The greatest increases in overall avian extinction risk were in Queensland, South Australia and New South Wales, where drought and wildfire effects were most pronounced.
• Red List Index trends can reflect the impact of individual threats or conservation interventions, and represent an important tool for monitoring national and global progress towards international agreements.

Introduction
Australia hosts one of the most phylogenetically distinct vertebrate faunas in the world (Holt et al. 2013).Extinction risk to this group is largely posed by threats that are familiar globally, particularly invasive alien species, and habitat loss and degradation driven by unsustainable agriculture (Szabo et al. 2012a;Geyle et al. 2018;Tingley et al. 2019;Lees et al. 2022).Yet while the recent extinction rate of well-documented faunal groups in Australia remains in-step with global trends (e.g.Szabo et al. 2012b;Woinarski et al. 2019), there is concern that the extent of biodiversity decline may accelerate, especially in response to climate change, to which the country is considered especially vulnerable (Urban 2015).
In response to these growing threats to biodiversity, Parties to the Convention on Biological Diversity (CBD), including Australia, adopted the landmark Kunming-Montreal Global Biodiversity Framework (GBF) in December 2022 (CBD 2022a).Goal A in the framework stated that 'human-induced extinction of known threatened species is halted, and, by 2050, the extinction rate and risk of all species are reduced tenfold'.Among 23 action targets to achieve this and other goals was a commitment to 'Ensure urgent management actions to halt human-induced extinction of known threatened species and for the recovery and conservation of species, in particular threatened species, to significantly reduce extinction risk . . .' by 2030 (Target 4).The Australian government has also adopted comparable goals, with a recent Threatened Species Action Plan including the objectives 'the risk of extinction is reduced for all priority species' and 'new extinctions of plants and animals are prevented' (Commonwealth of Australia 2022).
The GBF recommended the Red List Index (RLI) as a headline indicator to be reported by all Parties to measure progress towards this target (CBD 2022b).This builds on its use to measure progress to the CBD's '2010 biodiversity target' (Secretariat of the CBD 2010) and the Aichi targets for 2020 (Secretariat of the CBD 2020), as well as by other multilateral environmental agreements (e.g. the Convention on Migratory Species, and UN Convention to Combat Desertification), the UN Sustainable Development Goals (UN 2022b), and the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES 2019).
The RLI communicates trends over time in the aggregate extinction risk of sets of species (Butchart et al. 2004(Butchart et al. , 2005(Butchart et al. , 2007) ) using data from the IUCN Red List of Threatened Species TM (hereafter 'IUCN Red List'), widely recognised as the most authoritative source of standardised information on the global extinction risk of species (Rodrigues et al. 2006;Brooks et al. 2015).The IUCN Red List places species in categories by consistently applying the best available data to thresholds, with species of lowest extinction risk listed as 'Least Concern', and those of highest as 'Critically Endangered'.The RLI is then calculated from the number of taxa in each IUCN Red List category, and changes in the RLI across time periods is assessed using the number of taxa moving between categories in each reassessment period owing to genuine improvement or deterioration in extinction risk.It excludes (through retrospective re-categorisation) all changes in Red List statuses resulting from improved knowledge or corrections to estimates of parameters applied to the Red List criteria (e.g. a population size calculated to be lower than previously thought), changes in taxonomy or revisions to the Red List Criteria or Guidelines.It is therefore an index of overall survival probability (the inverse of extinction risk) for sets of species, with a value of 1 if all taxa are Least Concern, and a value of 0 if all taxa (from the original complement) are extinct (Butchart et al. 2007).
Although originally developed at the global scale to track international progress against biodiversity goals (Butchart et al. 2004(Butchart et al. , 2005(Butchart et al. , 2006(Butchart et al. , 2010)), Red List Indices are increasingly being employed at national scales to track individual countries' progress towards national and international commitments (e.g.Zamin et al. 2010;Han et al. 2014;Rodrigues et al. 2014;Raimondo et al. 2023) made through the CBD and United Nations, or through national legislation or policy.There exist two means by which national RLI values can be generated (Bubb et al. 2009): (1) a 'disaggregated global RLI' (UN 2022a; Raimondo et al. 2023), in which each species is weighted by the fraction of its range that occurs in each country; and/or (2) using Red List assessments of national extinction risk produced in a country following the Guidelines for Application of IUCN Red List Criteria at Regional and National Levels (IUCN 2012).The latter can be challenging in biodiverse countries with limited resources, but complements the global approach so that national extinction-risk of non-endemic taxa can be adequately tracked (Raimondo et al. 2023).Szabo et al. (2012a) applied the national approach to calculate RLIs from three assessments of all Australian bird species undertaken at 10-year intervals between 1990 and 2010 (Garnett 1992;Garnett and Crowley 2000;Garnett et al. 2011).In 2020, a fourth national-scale iteration was carried out (Garnett and Baker 2021), offering the unique opportunity to generate a national RLI spanning 30 years.This is among the longest for any national RLIs, with only mammals and sharks in Italy (dating back to 1985: Rondinini et al. 2014), and birds in Denmark (dating back to 1990: Pihl and Flensted 2011) being comparable (Hoffmann et al. 2017), and globally is only exceeded by the RLI calculated for all birds (BirdLife International 2022).Importantly, the most recent decade saw several salient events for Australian bird conservation.For example, invasive mammals on Macquarie Island had long been a major contributor of extinction risk for a suite of seabirds, but these mammals were eradicated between 1998 and 2011 (Springer 2016;Garnett et al. 2018).Conversely, devastating 2019-2020 wildfires across extensive areas of temperate Australia caused major population losses of many species and were shown to have caused abrupt deterioration in RLI for a range of taxonomic groups, including birds (Legge et al. 2022;Rumpff et al. 2023).Building on the work of Szabo et al. (2012a), here we evaluate Australia's RLI over the past three decades and attempt to explain the main causes of change.

Scope and taxonomy
The geographic scope considered by Garnett and Baker (2021), and herein, was consistent with previous iterations, and includes mainland Australia, its islands and overseas territories (Christmas, Cocos (Keeling), Norfolk, Lord Howe, Macquarie and Heard Islands) and the 370-km buffer around these that make up the Australian Fishing Zone.
The taxonomic list used includes all regularly occurring Australian birds (i.e.core taxa), totalling 1,302 'ultrataxa' (subspecies and monotypic species), excluding alien invasive species and those considered vagrants to Australia.Analyses were conducted on this ultrataxon dataset (thus giving equal weight to species and subspecies) because this is the unit most often used for Australian conservation management (e.g.Department of Climate Change, Energy, the Environment and Water (2023)), and Szabo et al. (2012a) found similarity between the indices for species and ultrataxon datasets in Australia.The taxonomy used to produce this ultrataxon list is principally derived from del Hoyo andCollar (2014, 2016), and subsequent revisions thereto (HBW and BirdLife International 2019), with discrepancies explained in Garnett and Baker (2021).

Red list data
Assessments were made in accordance with the Guidelines for Using the IUCN Red List Categories and Criteria (IUCN Standards and Petitions Committee 2019) and were applied nationally to the region described above.As recommended by Butchart et al. (2007Butchart et al. ( , 2010)), Red List categories assigned in 1990 (Garnett 1992), 2000 (Garnett and Crowley 2000) and 2010 (Garnett et al. 2011) were reviewed and revised retrospectively if data available in 2020 suggested this was appropriate, thereby ensuring as best as was possible that changes in Red List categories between decades reflect genuine changes in extinction risk.The categories for each taxon chiefly follow Garnett and Baker (2021), but in some instances (n = 6 taxa; see Appendix), categories were revised further after the latter's publication to reflect the most recent data, some of which became available too late to be incorporated therein.

RLI calculations
RLI values were calculated following the method of Butchart et al. (2007).Each Red List category was scored using 'equal steps', where Least Concern (LC), Near Threatened (NT), Vulnerable (VU), Endangered (EN), Critically Endangered (CR) and Extinct (EX) are scored from 0 to 5, respectively.The number of taxa in each category in each year (1990, 2000, 2010 and 2020) is then multiplied by these scores and expressed as a fraction of the maximum sum (where all taxa have gone extinct).This final value is subtracted from 1 to give the Red List Index value.All ultrataxa (including threatened and non-threatened) were included, except for those listed as Extinct or Critically Endangered (Possibly Extinct) in 1990.
Confidence intervals (95%) for RLI values were derived following Butchart et al. (2010).No taxa in the dataset were assessed as Data Deficient (Garnett and Baker 2021), hence the method used to randomly allocate categories to Data Deficient taxa was not used.However, methods to quantify extrapolation uncertainty and temporal variability were used to derive confidence intervals as follows: (1) Extrapolation uncertainty: although RLIs are extrapolated linearly based on the slope of the closest two assessed points, there is uncertainty about how accurate this slope may be, so rather than extrapolating deterministically, the slope is selected from a normal distribution with a probability equal to the slope of the closest two assessed points, and standard deviation equal to 60% of this slope.(2) Temporal variability: because assessments were repeated only at 10-year intervals, the precise value for any particular year is uncertain, so the RLI value each year is assigned from a moving window of 5 years, centred on the focal year (with the window set as 3-4 years for the first two and last 2 years in the series).Hence, to incorporate these uncertainties into the aggregated RLI, an RLI for each taxonomic group is calculated interpolating and extrapolating, and a final RLI value is assigned to each taxonomic group for each year from a window of years as described above.This is repeated 10,000 times and the mean is calculated (Butchart et al. 2010).

RLI disaggregations
To determine the principal drivers of changes in RLI, ultrataxa were disaggregated (sensu Butchart et al. 2004Butchart et al. , 2005) ) into subsets of geography, habitat and taxonomy: (i) Geography.Using the approach of Szabo et al. (2012a), taxa were characterised as occurring on the oceanic islands (Christmas, Cocos (Keeling), Heard, Macquarie, Norfolk and Lord Howe), continental islands (including Tasmania) that were connected to the Australian mainland during the last glacial period, and/or the Australian mainland.Mainland taxa were disaggregated further by state.The extinction risk was assessed at the national scale for all analyses because there has been no analysis of extinction risk for all taxa at a subnational level and global scale analyses do not include subspecies.Taxa that occur in combinations of these geographies were listed under all that were appropriate.(ii) Habitat.Classes of habitat occurrence for each taxon are from Garnett et al. (2015) using definitions based on the Department of the Environment and Water Resources (2007) and Hutchinson et al. (2005).Habitat classes for taxa were derived de novo for the current publication from the same sources as were used in Garnett et al. (2015).Taxa that occur in multiple habitat classes were listed under all considered appropriate.(iii) Taxonomy.The RLI of each family (sensu del Hoyo andCollar (2014, 2016)) with more than 15 ultrataxa.
We also sought to investigate the impact of the 2019-2020 wildfires on Australia's RLI.To do this, we extracted all ultrataxa (n = 25) for which the wildfires were identified as the cause of being moved to a higher extinction risk category in 2020.For none of these 25 were the other threats identified by Garnett and Baker (2021) considered capable of operating at an acuity and spatial scale that would have driven an increase in extinction risk category between 2010 and 2020 without the 2019-2020 wildfires.Consequently, in developing the counterfactual scenario, we assumed that these 25 ultrataxa would have remained in the 2010 Red List category if the fires had not taken place.

Geography
Oceanic island ultrataxa (n = 79 taxa) remain those with the greatest aggregated extinction risk but are the only geographic subset for which the RLI value increased between 2010 (0.555 [0.552-0.560])and 2020 (0.592 [0.584-0.603])(Figure 4).Disaggregating oceanic islands further, however, reveals heterogeneity.While the RLIs for species on Macquarie and Heard Islands increased by 21.2% (n = 28) and 11.3% (n = 19), respectively, the RLIs for species on Norfolk (−1.2%, n = 40), Lord Howe (−2.5%, n = 44), Christmas (−1.5%, n = 30) and Cocos (−2.2%, n = 21) islands (Figure 4) all declined.The steepest decline was observed on Australia's continental islands, where the RLI score decreased by 7% in the same timeframe.Much of this loss was a result of the 2019-2020 fires on Kangaroo Island, which directly led to 16 of the island's 17 endemic/near-endemic taxa being moved to a higher extinction risk category; the aggregated mean RLI value for these 17 taxa consequently halved between 2010 (0.930 [0.842-0.930])and 2020 (0.463 [0.381-0.581]).While the RLI for mainland taxa remained higher than that of insular subsets, and declined less steeply (Figure 4), there was considerable variation among the nation's eight jurisdictions.The RLIs for all states except Western Australia declined between 2010 and 2020 (Figure 4), with the greatest declines observed in Queensland (−1.5%, n = 748), South Australia (−1.6%, n = 474) and New South Wales (−1.3%, n = 506), where the effects of drought, fire and other climate change-related processes were most pronounced.

Habitat
The RLI for all terrestrial habitats declined (Figure 5) with declines observed for the period 2010-2020 in dry sclerophyll, heath, mallee, rainforest and wet sclerophyll faster than in any decade since 1990.Over the past four decades, the aggregated RLI of sandy shore species has declined most rapidly (−12.7%,n = 64) while the RLI scores for arid shrubland, mangroves and rivers and streams have remained both very high and almost unchanged.

Discussion
The change in RLI for Australia's birds between 2010 and 2020 reveals that the level of extinction risk was far worse than during the previous three decades.For the first time, Australia's mean national RLI score for birds in 2020 (0.891) fell below the global RLI score for birds (0.902; BirdLife International 2022).While the two indices are not directly comparable given the different taxonomic units used (species vs ultrataxa), Szabo et al. (2012a) found high congruence between the two.Although derived from relatively broad categories of extinction risk, our data show that the Red List Index can be sensitive to individual events, threats or conservation responses that affect a large suite of taxa, and the net effect of these impacts can be quantified.Looking to the future, the discernible increase for Macquarie Island birds after invasive mammal eradications is likely to foreshadow comparable recoveries in bird populations that should benefit from recent eradications of invasive rodents on Lord Howe Island, and programs aiming to eradicate cats from Christmas and Bruny Islands; by 2030, the beneficial impact of these events on Australia's RLI should be quantifiable.
The decline in the RLI of rainforest species was principally driven by gradational deteriorations in the Red List status of Wet Tropics taxa, which in turn were directly attributed to climate change based on 17 years of intensive surveying showing major range contractions (Williams et al. 2021;de la Fuente et al. 2023).More abruptly, the climate change-driven 2019-2020 wildfires caused c.50% of the reduction in RLI between 2010 and 2020.Taken together, these two analyses indicate that the threats associated with climate change analysed here were a considerable driver of increased extinction risk in Australia's avian RLI (Urban 2015;Garnett and Baker 2021;Legge et al. 2022;de la Fuente et al. 2023).These data also support the findings of Legge et al. (2022), who documented a marked impact of the 2019-2020 wildfires on RLI for each vertebrate class (albeit based on EPBC Act status rather than, as here, Red List status) and Nolan et al. (2021a), who wrote that 'the rate of change in wildfire risk delivered by climate change is outstripping the capacity [of ecological systems to adapt]'.Together, these data suggest that Australia's ability to meet its CBD obligations to halt extinctions and significantly reduce extinction risk is contingent on international climate change cooperation and fundamental shifts in fire management (Nolan et al. 2021a); some of these losses, however, may ultimately be unrecoverable (Nolan et al. 2021b).
Shorebird declines were also a substantial contributor to the worsening in RLI between 1990 and 2020, with Scolopacidae (sandpipers) emerging as the family with the greatest fall in aggregated RLI in the past 30 years.This mirrors widespread concern in shorebird trends across the East Asian-Australasian Flyway (Szabo et al. 2016a(Szabo et al. , 2016b;;Yong et al. 2018), in which, for most species, Australia is the southern end-point of a migration that spans over 10,000 km.It is notable that declines in migratory shorebird abundance (and therefore the associated impact on RLI) are largely thought to be driven by threats outside of Australia, especially habitat loss (particularly of tidal flats) and hunting in East Asia (Studds et al. 2017;Gallo-Cajiao et al. 2020;IUCN 2023) and, to an unknown extent, climate change-induced reductions in breeding productivity (e.g.Kubelka et al. 2018).In 2020, 83% of Australian bird taxa remained Least Concern, and the absence of extinctions in the period 2010-2020 suggests that conservation management in Australia is having some success at preventing extinctions, with most changes to the RLI between 2010 and 2020 driven by rapid declines in species that remain (for now) numerous.Moreover, the completion of alien invasive species eradications on Macquarie Island in 2014 (Springer 2016;Garnett et al. 2018) produced a rapid, detectible and measurable impact on Australia's RLI in 2020, providing more evidence that the metric can be sensitive to successful conservation interventions (Szabo et al. 2012a;Young et al. 2014).The near-parallel rebound in RLI for Heard Island (see Figure 5) is most likely artefactual.Since both islands share a similar avifauna, a reduction in the national extinction risk of species occurring on Macquarie Island has axiomatically impacted that of populations of the same taxa breeding on Heard Island, despite no monitoring or conservation action on the latter to have driven genuine local improvement.Such sensitivities to the way in which species are aggregated indicate a need to disaggregate RLIs cautiously and across multiple traits to determine the true causes of improvement or deterioration.
Globally, in some taxonomic groups, uncertainty in RLI values is principally due to Data Deficient taxa (Butchart et al. 2010).However, no Australian birds are listed as Data Deficient and most uncertainty  (Verdon et al. 2024).There is consequently a critical need to ensure that the collection of data relevant to IUCN Red List assessments (e.g.population size, trend and range metrics) continues, so that the information base for future Red List assessments is not constrained by shortfalls in research and monitoring.Future research may consider using a causal model to explore in more detail the underlying trends in the transition process between conservation categories.The biological traits that predispose Australian bird taxa to extinction risk were analysed by Olah et al. (2024) but this could also be extended to consider more precisely the factors common to species moving between certain categories.

RLI and other threatened species trend measures
Australia supports two other national-scale sources of information on trends over time in the extinction risk of Australian birds: (1) the Threatened Species Index (TSX); and (2) changes to the listings under the Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act; Department of Climate Change, Energy, the Environment and Water 2023).The TSX was developed in 2020 to monitor annual changes in population abundance for threatened taxa for which there are suitable data, starting with birds (Bayraktarov et al. 2021), based on methodologies developed for the Living Planet Index approach of assessing populations against those in a baseline year (Loh et al. 2005).
Consistent with the RLI results herein, the TSX of aggregated monitoring data on threatened birds from 1985 to 2019 showed that migratory shorebirds had a much larger average population decline of 65% compared with 48% for all birds combined (TSX 2023).Although both the RLI and TSX have as their baseline the extinction risk assessments for individual taxa, they are distinct and complementary.The TSX provides a high-frequency measure of performance in reducing threats to taxa that are already considered to have an elevated extinction risk, even if the timeseries data analysis for the TSX means that it may contain taxonomic and geographic biases resulting from historical variations in monitoring effort.In contrast, the RLI assesses the extinction risk trends in entire avifauna, including Least Concern taxa, but at a longer time interval because it considers the impact of all components of extinction risk (e.g.changes in extent and structure of distributions), which typically change more gradually.
The EPBC list includes taxa that have been assessed as threatened under a set of criteria almost identical to those of the IUCN Red List, and it has legislative implications for the species listed.However, listing and delisting are constrained by resources and potentially are influenced by political considerations.The EPBC, for example, can be slow in listing extinctions (Woinarski et al. 2019) and very rarely delists taxa even if they do not meet the IUCN Red List criteria, in part because there are explicit safeguards used to constrain delistings (Woinarski et al. 2023).Also, while there has been exceptionally rapid uptake of recommendations on listing resulting from the latest IUCN Red List assessment of birds relative to previous assessments (STG pers.obs.), the EPBC listing cannot be used to assess status retrospectively.For that, some statistical modelling approaches can be used even with phylogeny control (Olah et al. 2024); however, the complexity of such models might hinder their implementation as standard measures.The RLI is therefore uniquely placed to monitor the ongoing extinction risk of Australia's birds over time, and it will be critical in monitoring the country's progress towards international agreements such as the CBD Global Biodiversity Framework.
Our assessment indicates that over the last three decades, the extinction risk for the Australian bird fauna has increased, with this trajectory deteriorating even further over the most recent decade.On these trends, the goal of 'human-induced extinction of known threatened species is halted, and, by 2050, the extinction rate and risk of all species are reduced tenfold' expressed in the Kunming-Montreal Global Biodiversity Framework (GBF) in December 2022 (CBD 2022a) is slipping further away, and this deterioration indicates that much more needs to be done effectively and strategically if the goal is to be achieved.Our results show some signs that can help.For example, significant conservation management programs (notably eradication of invasive mammal species from Macquarie Island) have clearly delivered benefits.Our assessment also indicates the sets of bird species undergoing most rapid deterioration, and hence where more effective conservation responses may be most needed.

Figure 1 .
Figure 1.Number of ultrataxa of Australian birds in each IUCN Red List category (retrospectively assessed) for each year in which their status was assessed.The number of Least Concern taxa (excluded from plot) in each assessment year was: 1,138 (1990), 1,119 (2000), 1,091 (2010) and 1,054 (2020).
Figure 1.Number of ultrataxa of Australian birds in each IUCN Red List category (retrospectively assessed) for each year in which their status was assessed.The number of Least Concern taxa (excluded from plot) in each assessment year was: 1,138 (1990), 1,119 (2000), 1,091 (2010) and 1,054 (2020).

Figure 4 .
Figure 4. Red List Index of species survival for Australian birds disaggregated by geography between 1990 and 2020.Shaded area represents 95% confidence intervals.

Figure 5 .
Figure 5. Red List Index of species survival for Australian birds disaggregated by habitat between 1990 and 2020.Taxa that occur in multiple habitats are included in all that were appropriate.Shaded area represents 95% confidence intervals.

Table 1 .
Disaggregated mean Red List Index values for all Australian bird families with more than 15 ultrataxa.Presented in the descending order of % change 2010-2020.