Australian threatened birds for which the risk of extinction declined between 1990 and 2020

ABSTRACT Reducing extinction risk is a common aim of threatened species management. However, over the period 1990 to 2020, extinction risk was recently assessed as having declined in only 25 out of the 199 Australian bird taxa eligible for assessment. Here we analyse patterns that emerge from these taxa. Some of these improvements may be only temporary; the extinction risk of three taxa increased after it had initially declined. Invasive predator control on islands was the conservation intervention with greatest impact, benefitting 13 taxa (with nine of these from Macquarie Island). For four taxa, intensive management was the primary driver of reduced risk. Another four benefited from habitat protection and one from law enforcement. For seven taxa, conservation actions had no discernible effect; for two albatrosses a shift in fishing patterns may have reduced bycatch, for one, losses on the mainland meant that most birds now persist only in a stable island population and, for four taxa, reasons for changes in population trend are unknown. Never was there only one driver of reduced extinction risk with most taxa benefitting from at least five drivers. Macquarie Island was the only geographic cluster of taxa; there was little overlap among other taxa. Although the number of improvements is small, our results demonstrate that reduced extinction risk can be achieved with the right combination of targeted actions and, in some cases, serendipity. However, due to insufficient data, our ability to predict accurately the drivers of, or changes in, extinction risk for most species remains poor. POL ICY HIG HLIG HTS Reductions in extinction risk are rare but achievable. Up to ten different factors contributed to reduced extinction risk in a taxon; there were rarely fewer than five. Extinction risk reduction had no geographical focus apart from the cluster of taxa on Macquarie Island. There is currently no widely accepted approach to classifying the many ways in which extinction risk can be reduced. Over a quarter of risk reduction examples were not the result of conservation interventions.

• Reductions in extinction risk are rare but achievable.
• Up to ten different factors contributed to reduced extinction risk in a taxon; there were rarely fewer than five.• Extinction risk reduction had no geographical focus apart from the cluster of taxa on Macquarie Island.• There is currently no widely accepted approach to classifying the many ways in which extinction risk can be reduced.• Over a quarter of risk reduction examples were not the result of conservation interventions.

Introduction
A core objective of threatened species management is to reduce the likelihood of extinction.A reduction in extinction risk may be the outcome of effective and targeted management of key threats, or a consequence of changed threat environments, or may arise serendipitously.Determining the extent and causes of improvements in species' conservation status is important because it can provide a measure of return on conservation investment, help identify actions that could be transferred effectively for comparable species (Cinner et al. 2016), and help refine management actions that are not working; for those involved in threatened species conservation, highlighting cases of such improvement can also provide hope and motivation in a sea of woe (Knowlton 2021).
Assessing changes in extinction risk is greatly assisted by the categories and criteria used to determine status under the IUCN Red List (IUCN 2001).Although CONTACT Stephen T. Garnett stephen.garnett@cdu.edu.auThis article has been corrected with minor changes.These changes do not impact the academic content of the article.
Supplemental data for this article can be accessed at https://doi.org/10.1080/01584197.2023.2291140.
extinction risk is a continuum, and improvement or deterioration in extinction risk may occur without resulting in changes in IUCN Red List category, reductions in extinction risk category -from Critically Endangered, to Endangered, to Vulnerable, to Near Threatened, to Least Concern -all require that at least one of the factors driving extinction risk has crossed a threshold and, mostly, changed for the better.Such factors include the size or patterns in a population or its distribution; all can affect extinction risk.We note, however, that reductions in extinction risk may or may not equate to recovery or conservation success because both terms have many definitions and varied uses, and changes in extinction risk are only one potential criterion of 'success' (e.g.Redford et al. 2011;Westwood et al. 2014).For example, the Green List for Species (now Green Status: https://www.iucnredlist.org/about/green-status-species) describes success in terms of 'viability, ecological functionality, and representation' (Akçakaya et al. 2018), which is well beyond what we have attempted here.Our focus is simply on those taxa exhibiting some step-change improvement in extinction risk, and some of the taxa for which we describe such improvements have not fully recovered and are still threatened.Bolam et al. (2021) showed that, globally, since 2003, for species whose otherwise likely extinction had been prevented, the principal reasons were invasive species control, ex situ conservation and/or site/area protection.Threatened bird taxa known to occur at a single site in 2005 were more likely to show reductions in extinction risk if they were subject to more intense land, water or direct species management or better law and policy (Luther et al. 2021).Globally, BirdLife International has determined that, of 540 genuine changes in extinction risk over seven assessment periods of four years each since 1988-1992, 89 had been reductions in the extinction risk (BirdLife International 2023).A recent review of all Australian threatened birds (Garnett and Baker 2021) provides an opportunity to identify factors that may have influenced reductions in extinction risk at a continental scale based on retrospective assessments of extinction risk categories.
The review of changes in extinction risk (Garnett and Baker 2021) complements partial analyses over the last 15 years.In one, the Orange-bellied Parrot Neophema chrysogaster was included in a global list of 45 bird species thought to have been saved from extinction by conservation actions since 1993 (Bolam et al. 2021).However, although its extinction has so far been prevented, the parrot has remained Critically Endangered for over 30 years (Menkhorst et al. 2021).The Orangebellied Parrot was also one of four Australian bird taxa, including subspecies, that Garnett and Crowley (2008) considered to have been likely to have become extinct over the period 1950 to 2000, were it not for conservation action (the others were the Norfolk Island Green Parrot Cyanorhamphus novaeselandiae cookii, Noisy Scrub-bird Atrichornis clamosus, and White-chested White-eye Zosterops albogularis, with the white-eye now thought to have gone extinct between 2000 and 2010; Clarke et al. 2021).They also considered that the extinction risk of six taxa was lower in 2000 than 1950, although this assessment was based on older, slightly different, interpretations of the IUCN Red List criteria.A set of chapters about success in Australian threatened species conservation (Garnett et al. 2018b) also included 11 bird taxa as examples of preventing extinction or reducing its risk; however, not all of these examples showed a step change in conservation status: e.g.Helmeted Honeyeater Lichenostomus c. cassidix has also remained Critically Endangered for over three decades, but is likely to have gone Extinct without conservation interventions (Harley et al. 2018;Quin et al. 2021).Most recently, Woinarski et al. (2023b) listed eight bird taxa that had previously been listed as threatened under Australia's Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) but no longer meet the listing criteria (while not necessarily being eligible for delisting; Woinarski et al. 2023a).
Neither globally nor in Australia has there been a review of all the taxa within a taxonomic group that have, in any time period, been shifted to a lower extinction risk category.Nor has there been a detailed analysis of the characteristics of those taxa or the circumstances that led to the change in extinction risk.Here we do so for Australian bird taxa since 1990 based on a recent assessment of extinction risk and of all changes in risk status at decadal intervals, assessed retrospectively in 2020 (Garnett and Baker 2021).Using this data set, we assess the examples of extinction risk reduction to determine which types of conservation action have been important.To do this we classified the actions using a framework that combined approaches recommended by the IUCN (2012a, 2012b), Garnett et al. (2018a) and Ingeman et al. (2022), a synthesis of classifications that may have relevance to similar analyses with other groups of taxa.

Scope
This paper considers changes in extinction risk, using the IUCN Red List criteria, based on retrospective assessments published in Garnett and Baker (2021) with a small number of subsequent corrections (see Berryman et al. 2024).Only genuine changes were considered, so omitting any resulting from changes in knowledge, taxonomy, interpretation of guidelines or error.Here we consider all 1,273 Australian bird taxa occurring naturally in Australia (i.e.not introduced taxa) that were not Extinct in 1990 or considered vagrants, based on the taxonomy of BirdLife Australia (2022).Our analysis focused on the 166 taxa that were (retrospectively) assigned Near Threatened or threatened status in at least one decadal period between 1990 and 2010, as these were the only ones that could show an improvement in extinction risk by 2020.The taxonomy follows BirdLife Australia (2022) and considers all subspecies and monotypic species (ultrataxa).The geographical scope covers all of Australia and the six oceanic island groups under its jurisdiction (Heard, Macquarie, Norfolk, Lord Howe, Christmas and Cocos (Keeling)).For seabird taxa that breed both inside and outside Australian territory, extinction risk was assessed only for the population breeding inside Australia.

Drivers of reduction in extinction risk
For taxa assessed in Garnett and Baker (2021) as having a lower extinction risk than in the decade immediately preceding it, conservation interventions and their impact were identified from the relevant literature (Supplementary Table S2), augmented by the personal experience of five of the authors (STG, GBB, NC, SML, JCZW) with conservation actions of all of the taxa over the 30-year period described in the paper.The effectiveness of each intervention in reducing extinction risk was assessed subjectively against a counterfactual assumption of the probable extinction risk had not the intervention occurred.Interventions were interpreted as having made an active contribution aimed at reducing extinction risk of that taxon, a passive contribution (e.g.habitat protection with no additional active management targeted at the taxon), no difference (while actions had been implemented, they were judged not to have contributed to the change in extinction risk even if there may have been local benefits), or of there having been no action.Note that the actions considered to have contributed to reduced extinction risk need not have been instituted since the taxon became threatened; laws that set aside habitat for protection or reduced hunting frequency were often essential precursors to subsequent improvements but may have been advocated for and enacted many decades before extinction risk met IUCN Red List thresholds.Likely drivers of lowered extinction risk were assessed against the nine categories of conservation action that might need to be taken to conserve a taxon (IUCN 2012a), the ten potential Conservation Actions In Place (IUCN 2012b) which the IUCN (2012a) recommends be consulted to inform the actions required, the 13 categories of action that Ingeman et al. (2022) considered likely to benefit higher-order predators, and the seven types of action that Garnett et al. (2018a) considered most important for successful threatened species management in Australia (Table 1).The combination of factors likely to have had the greatest influence on reducing extinction risk were then described in narrative form.

Characteristics of improving taxa
We used Generalised Linear Models (GLMs) to investigate whether particular characteristics were associated with improvements in conservation status, as these types of model make no assumptions about the error distribution of the data.We measured the influence of 14 predictor variables on reduced extinction risk (measured as a binary variable, where 0 = no change) for all Australian bird taxa that had been listed as threatened by 1990, 2000 or 2010 (n = 184).Predictors included a set of 14 factors considered likely to influence the impact of threats, e.g.land protection, changes in policy (see Garnett et al. 2019Garnett et al. , 2024;;Supplementary Tables S1-S3 for details, data sources and data).Continuous variables for each taxon were: (i) detectability: a measure of the effort needed to find them (ii) generation time (as per IUCN 2001) (iii) monitoring score (see Verdon et al. 2024 We tested for collinearity among predictor variables, finding no correlation coefficients > 0.7.We then modelled all possible combinations of predictor variables (using a logit link function) and used a second-order form of Akaike's Information Criterion (AICc) (Burnham and Anderson 2002) to rank the performance of competing models.We identified a candidate set of models within 2 Akaike units of the best ranked model (those more than 2 units away were considered to have less support for explaining the observed patterns in the data; Burnham and Anderson 2002) and used model averaging to obtain a final output for inference (i.e. where the coefficient for a given predictor was averaged across all models in the candidate set in which it appeared).We considered a predictor to have a significant influence on reduced extinction risk when the 95% confidence intervals of the averaged regression coefficients did not overlap zero.All analyses were undertaken using the 'lme4' package (Bates et al. 2015) in R version 4.2.1.

Taxa with reduced extinction risk
Of the 1,273 taxa, 1,041 remained Least Concern throughout the 30-year period and 33 were newly listed as imperilled in 2020, or were extinct by 2010, so could not have a reduced extinction risk by 2020.Of the remaining 199 taxa, the extinction risk of 25 was judged to have declined sufficiently in at least one decadal interval over the period 1990 to 2020 for the taxon to be placed in a lower extinction risk category as judged by the IUCN Red List criteria (IUCN 2001).Eleven of these 25 were judged to have oscillated such that they experienced both an improvement and a deterioration in extinction risk in different decades.In contrast, a far larger number (100) showed a deterioration in extinction risk, and 74 remained in the same category of extinction risk throughout.Of the 25 taxa judged to have experienced a reduction in extinction risk (Table 2), four taxa were shifted to successively lower extinction risk categories twice and one three times; six shifts occurred after the first decade, five after the second, and 20 after the third.However, for the same set of 25 taxa, there were also 11 increases in extinction risk: two after the first decade, eight after the second and two after the third with one taxon, having reached Vulnerable, retreating to Critically Endangered before returning to Vulnerable.Considering only the most recent (2020) assessment and the baseline assessment (1990), only 17 taxa showed overall improvements in extinction risk.
Allowing for the fact that the extinction risk of Least Concern taxa cannot be reduced between decades, and that of Extinct taxa cannot change, there were 31 inter-decadal reductions in extinction risk out of 488 opportunities for a taxon to be imperilled in one decade and less so in the subsequent decade (6.4%).For the same set of taxa there were 200 increases in extinction risk out of 698 opportunities (28.7%) or 5.2% of 3,818 opportunities if all taxa that were not Extinct at the start of an assessment period (i.e.all 1,273 taxa over three decades, minus one taxon that was assessed as Extinct by 2010) are to be considered.
Of the 25 taxa in which extinction risk was lower, 12 are seabirds including nine from Macquarie Island.The extinction risk of two of the seabirds with lower extinction risk (Amsterdam and Sooty albatrosses) and two shorebirds (Eastern Greater Sand Plover and Great Knot) was likely to have been reduced because of changes that occurred outside Australian territory.The taxa for which extinction risk was lower were widely distributed across Australia, with overlap only for taxa on Macquarie Island, the two visiting populations of albatross species and the two migratory shorebird taxa (Figure 1).Otherwise, the taxa have largely separate distributions, apart for Partridge Pigeons and Gouldian Finches overlapping in the Northern Territory.Flame Robins have a large distribution but differ ecologically from any of the other taxa with which they may occasionally co-occur.

Factors that may have contributed to improvements
Assessment against the proposed pre-requisites for successful conservation revealed a complex range of influences (Table 3; Supplementary Table S4 For the Southern Cassowary, Tasmanian Wedgetailed Eagle and Bulloo Grey Grasswren, conservation for the birds themselves was a major reason for the land protection; Albert's Lyrebird was the serendipitous beneficiary of conversion of an important NSW state forest into a conservation area.The first three taxa were also subject to targeted conservation beyond the prevention of their habitat being cleared.The Southern Cassowary has had some habitat rehabilitated, but not enough to be a major driver of reduced extinction risk. For seven of the taxa, the reduction of extinction risk was not a result of conservation action at all.For the Amsterdam and Sooty albatrosses, a shift in fishing grounds (not a consequence of conservation) is thought the most likely reason for reduced bycatch deaths.For the Eastern Partridge Pigeon, downlisting occurred because the population declined in one area but not another, such that the overall rate of decline slowed as the declining population became an increasingly small percentage of the taxon's total population.Finally, for Greater Sand-Plover, Great Knot, the Flame Robin and the Gouldian Finch, there is no clear explanation for the change in population trend that reduced their extinction risk.For four taxa, protection of their habitat was sufficient to reduce extinction risk.
However, while the principal, proximate drivers of reduced extinction risk can be identified in each case, all have worked in concert with others (Figure 2; see Supplementary Table S4 for detailed arguments for all factors and taxa).For most taxa there were at least five  Evidence that populations on mainland Australia were declining from 1990-2010 suggested that declines in the total populations was approaching 30% in three generations (15 years).However, mainland declines meant that the proportion of the population on the Tiwi Islands, where the population density is at least 10 times that in Kakadu National Park and probably exceeds 80,000 individuals, must have increased.By 2020, it was mathematically unlikely that losses were still approaching 30%.Mainland declines have been attributed to changes in fire regime and cat predation, but the same phenomena appear not to have affected birds on the Tiwi Islands.Drivers: Not related to conservation.Amsterdam Albatross Non-conservation drivers 0 A gradual increase in the breeding population that has occurred since the 1980s is thought to have been a consequence of reductions in longline fishing for tuna in the southern Indian Ocean.No management focused on the albatross is likely to have benefitted the species, although removal of cattle from Île Amsterdam may benefit some birds.Drivers: Not related to conservation.

Sooty Albatross
Non-conservation drivers 0 Rapid population declines in the 20th century as a result of longline fishing have been followed by a period of stabilisation or increases in the breeding population at many sites since 2000.The change is possibly because of reductions in longline fishing for tuna in the southern oceans, although this is uncertain.No management focused specifically on the albatross is likely to have benefitted the species.The change in extinction risk is recognised by Garnett and Baker (2021) but not yet by the IUCN.Drivers: Not related to conservation.Macquarie Island taxa (9 taxa: Light-mantled Sooty, Grey-headed and Black-browed Albatrosses, Blue, Softplumaged, Whiteheaded and Grey Petrels, Antarctic Prion and Subantarctic Diving-Petrel Land/water management 8 Eradication of cats in 2000 benefitted some petrel species but then allowed rabbits to proliferate, overgrazing the island and the resulting habitat degradation threatened the recovery of some petrels and three of the four breeding albatross species.A successful campaign to eradicate rabbits and rodents has not only allowed recovery of existing populations but has also enabled immigration of some taxa whose populations were previously restricted by the availability of safe breeding habitat.The invasive species eradication was inspired by well-documented campaigns in New Zealand, strongly advocated by Australian NGOs and government officers and pushed through, despite some operational difficulties.Monitoring ensured that the campaign was successful and was having the desired impact.Drivers: Land/water protection and management, Law and policy, Education and awareness, Research and Monitoring, Conservation planning, Other: continuity and adequacy of resources, advocacy.Australian Gould's Petrel Species management 10 The subject of adaptive management since the early 1990s with research on threats followed by threat amelioration that included elimination of rabbits that degraded nesting habitat on the only breeding island and establishment of new colonies on other islands (by translocation and self-introduction combined with provision of additional artificial habitat) to spread risk.(Continued) factors contributing; for Australian Gould's Petrel and the Norfolk Island Green Parrot, 10 factors could be identified.The factor that affected fewest taxa, livelihoods, covers the compensation/purchase prices of land for the Southern Cassowary and Bulloo Grey Grasswren and the higher value of land occupied by Kangaroo Island Glossy Black-Cockatoos.External capacity building, which applied to none of the examples, was included only because the IUCN (2012a) identified it as an action that might be required to reduce extinction risk.

Characteristics of taxa with reduced extinction risk
The most important predictors of reduced extinction risk in Australia birds were generation time, range size and number of mature individuals; taxa that had moved  (1977)(1978)(1979)(1980)(1981) and the second (1999)(2000)(2001), but the population subsequently stabilised for unknown reasons, with no major contraction in range.Drivers: Not related to conservation.

Gouldian Finch
Non-conservation drivers 0 A steep decline from east to west across its once extensive range across northern Australia starting before the 1930s and continuing until the early 2000s, followed by a gradual increase in numbers and range from west to east.While there have been improvements in fire management in some areas, the population increase appears to have occurred largely independent of management.Drivers: Not related to conservation.to a lower extinction risk category were more likely to be longer-lived, have smaller distributions, and have a higher number of mature individuals (Figure 3, Supplementary Table S5).Detectability, monitoring score, taxonomic distinctiveness, whether a taxon had a conservation plan, legislative status and ultrataxon class all appeared in the candidate model set but had less support for explaining reductions in extinction risk (evidenced by 95% confidence intervals overlapping zero, Figure 3).Site accessibility, variability, weight, geography, and taxonomic group did not appear at all in the candidate set of models, suggesting they have little support for explaining the observed patterns.

Frequency of extinction risk reduction
We found that a small proportion of the Australian imperilled bird taxa experienced a reduction in extinction risk between 1990 and 2020, notwithstanding considerable conservation investment for many of them.The 31 extinction reductions of 25 taxa over three decades is equivalent to 1.0 per year, and less than that if reversals are included.For the period of study, about five times more taxa experienced an increase in extinction risk rather than a reduction in that risk.This reflects the ratio of threat impact to conservation effort in Australia (Berryman et al. this issue) and is consistent with ongoing declining trajectories reported for threatened Australian birds generally (Bayraktarov et al. 2021).Another reason for the far smaller pool of taxa showing improvements in extinction risk is the time lag between interventions and the outcome of moving taxa to a lower category of risk.This time lag is incorporated in the IUCN Red List guidelines, which state that 'A taxon may be moved from a category of higher threat [extinction risk] to a category of lower threat if none of the criteria of the higher category has been met for five years or more' (IUCN 2001).However, just to reach the lower category often requires decades.Just as it is usual for species and habitats in marine environments to take one to three decades to approach a reference level after threats have been ameliorated (Duarte et al. 2020), the same has been true for many of the bird taxa in the current analysis.Even passive interventions, such as the declaration of protected areas,  S3 and S4).
require a lag period to achieve beneficial results.Sometimes, as in the case of Southern Cassowary, the long generation time has meant that the benefits have taken several decades to be realised.Such caution is greater where listings have legal ramifications, as is the case with Australia's EPBC Act (Woinarski et al. 2023a).Under the Act, removing a taxon from the list cannot be undertaken if it is considered that changing that taxon's status will result in conservation detriment (EPBC Act Section 186 2A).However, the same is not true of the IUCN Red List criteria.Such caution about moving taxa to lower risk categories does not apply to moving them to a category of higher risk, which biases the rate of risk reductions over increases.In the examples discussed in this paper, Southern Eastern Bristlebird and Australia Gould's Petrel were both immediately returned to higher risk categories despite earlier improvements after monitoring suggested that new but still unknown threats had started to reverse population trends.The timing of assessments is also important.One reason why actions taken on Macquarie Island benefited so many taxa over the 1990 to 2020 period is that the earlier eradication of cats Felis catus from the island was followed by a proliferation in rabbit numbers, although cause and effect is not entirely understood (Bergstrom et al. 2009a(Bergstrom et al. , 2009b;;Dowding et al. 2009).The 2010 assessment (Garnett et al. 2011) was completed at the time the damage to the island was at its peak.Eradication of all introduced mammals shortly afterwards (2011; Springer 2018) allowed the five-year cautionary lag to occur before 2020.Similarly, the risk status of the Norfolk Island Green Parrot fluctuated between decades partly because realisation that management was failing occurred as part of the 2010 risk assessment, with rapid action occurring immediately afterwards (Ortiz-Catedral et al. 2018).

Categorising extinction risk reduction
There appears to have been no systematic large-scale analysis of extinction risk reduction apart from the study of extinction avoidance (Bolam et al. 2021) and that of single site species (Luther et al. 2021), both of which are special cases.The first of two reasons for the lack of analysis may be that the categorisation of risk reduction actions seems not to be comprehensive.The IUCN currently has two sets of categories for conservation action but neither was adequate for what we have attempted here.The list of actions in place (IUCN 2012a), i.e. the actions that have been taken in the past, only partially overlaps the list of the actions that should be taken in the future (IUCN 2012b).However, when looking at trends over time, and hence what has been achieved, all assessments need to be retrospective.Therefore, the list of actions in place should include anything that might already have been initiated, such as habitat and species management.They should also include many of the other factors that drive extinction risk, such as those listed by Garnett et al. (2018a) and Ingeman et al. (2022).Factors like the importance of species champions for effecting risk reduction have been known for over 20 years (Carlile et al. 2003) but are ignored by the IUCN categorisation.
The second reason is that shifts to a lower risk category are so infrequent.This was true in our analysis here in which our sample size of taxa exhibiting improvements in extinction risk is too small to allow us to create a comprehensive listing of drivers of reduced extinction risk because we do not have the diversity of social and environmental circumstances in which risk has been reduced.To provide an adequate sample size for what is a rare event, a global analysis of extinction risk reduction in birds, and preferentially of a range of plant and animal groups, should be undertaken that will provide more empirical support for the types of driver that have most impact on extinction risk reduction.
As it is, our analysis has revealed some anomalous reasons for reducing extinction risk category, as well as highlighting the complexity of the issue.This is shown most readily in the seven cases, out of the 26 we examined, where extinction risk reduction apparently occurred for reasons other than conservation interventions.Simply categorising these as serendipitous would lose important information because there were at least three different reasons -one to do with changes in fishing effort, one related to criteria thresholds and the third having no known explanation but which are likely to be different in each case.
For the Sooty and Amsterdam albatrosses, the most parsimonious reason continuing declines have ceased or been avoided is that they no longer encounter fishing boats as often as they did.While this could be seen as an example of effective bycatch mitigation, which is either rare or rarely documented (Reid et al. 2023), the threat could return unless closely monitored, i.e. the extinction reduction can be seen as a temporary pause rather than a long-term solution and should be categorised as such.
Even harder to categorise or interpret is the Eastern Partridge Pigeon for which the extinction risk category improved because of continued population decline.When Near Threatened, the total population was thought to be declining at a rate approaching 30% because of losses on the north Australian mainland.However, as the mainland population shrunk, the proportional contribution of the Tiwi Islands population (relative to total population), which is large and probably stable (Davies et al. 2019), increased, meaning the taxon no longer met any extinction risk criteria (Davies et al. 2021).Such a perverse reduction in extinction risk category can happen even with rare declining species.For example, the Australian population of Australasian Bittern Botaurus poiciloptilus is classified as Endangered under Criterion C2a(i) because > 95% of the population of < 2500 mature individuals occurs in eastern Australia (Herring et al. 2021).However, were the western population to stabilise while the eastern one continues to decline it might exceed 5% of the total population, meaning the bittern would have been categorised as Vulnerable only because of differences in the rates of population loss.
The four taxa for which reasons for reduced extinction risk are unknown -the two migratory shorebird taxa, as well as the Flame Robin and the Gouldian Finch -may all have different reasons for population stabilisation.The shorebird trends may be fragile, given the ongoing threats to the environments on which they rely (Murray et al. 2014), so understanding why their numbers have declined then increased so dramatically over the period 1990 to 2020 will improve the focus of conservation efforts.That so many of the Great Knot stop-over sites remained unknown until recently (Chan et al. 2019) suggests that either the species is more flexible about where it stops than previously appreciated or that a high reproductive rate can compensate rapidly from occasional large losses, such as those that probably occurred when habitat in South Korea was reclaimed (Moores et al. 2008).For both taxa, the factors affecting breeding success, often the main driver of population trends (Newton 1988), are very poorly known.If breeding success is the major driver of population trends, then greater effort may be needed to understand how they can be influenced before numbers fall again.A similar need applies to the Flame Robin which underwent the greatest decline between the first major atlas of Australian birds (1977)(1978)(1979)(1980)(1981) and the second (1998)(1999)(2000)(2001) and has continued to decline locally even if the population as a whole has stabilised (Newman et al. 2021).By contrast, factors that affect the Gouldian Finch are now well known (Legge et al. 2021), so understanding reasons for their return to many sites from which they have been absent for decades could hold important lessons for other species.
Two other special cases that emerge from our analysis are those for which the shift in risk reduction occurred after conservation interventions were first abandoned and then reinstated (Norfolk Island Green Parrot) and where a conservation intervention may have been the cause of the extinction risk which was then remedied by further action (albatrosses breeding on Macquarie Island).Such cases are likely to be rare but understanding why they occur will be important to avoid repetition.Just as there are good arguments for coronial investigations of extinction to determine why they have occurred (Woinarski et al. 2017(Woinarski et al. , 2019)), similar reviews are warranted where conservation interventions produce unintended consequences.

Factors driving extinction risk reduction
We have identified what we consider to be the main proximal reasons for extinction risk reduction but note that other assessors may interpret data differently because the results make apparent the complexity of the process and the large number of factors that applied in almost all cases.While analysis of taxon characteristics suggested that those in our small sample were longer-lived, had smaller distributions and a higher number of mature individuals than did those that moved to a higher extinction risk category or were unchanged, we think these results provide a testable hypothesis for a larger sample rather than necessarily being causative or universal.While threats may be easier to reduce within small ranges, generation time and population size may be artefacts of our sample, and certainly need more detailed analysis across a broader range of examples.Rarely is it possible to tease out which factors were most important, or the sequence in which they influenced the risk reduction process.For example, listing of taxa and the enforcement of laws relating to that listing, was instrumental in achieving improvements of eight taxa as well as the nine Macquarie Island birds (see Supplementary Table S4).However, research was also important to every one of these examples that led to improved management, and adequate funding and advocacy was critical to all but two.For five of these taxa, none of these factors may have been influential had not someone stepped up to take personal responsibility and devoted a substantial proportion of their career to driving the conservation action.The decision that a taxon could be moved to a lower extinction category was often the culmination of decades of effort through many pathways.Understanding the ways in which such pathways interact and their sequencing could help emulate them with other taxa.Often the processes are more about social drivers, reflecting personal values and organisational change, rather than the biological characteristics of the birds or the institutional environment in which they find themselves (Holmes et al. 2017;Ainsworth et al. 2018).
Such an analysis need not be confined to taxa formally moved from a high to a lower extinction risk category.Our selection of taxa for the current analysis has been highly conservative.Across Australian birds there has a been a great deal of activity which has not led to changes in extinction risk category (Garnett et al. 2018b(Garnett et al. , 2019(Garnett et al. , 2024)), and may never do so given the size of the population and the intensity of management required.For some, the aim has been simply to prevent extinction (Harley et al. 2018;Bolam et al. 2021).For others, more time is required before a change in category can be achieved, but there are still valuable lessons to be learned.As with the categories for extinction risk recovery, there could helpfully be a formal assessment of the extent to which conservation actions have rendered benefits to a taxon even without the empirical evidence that comes with extinction risk category change.While the sort of counterfactual assessment needed for such analysis can seem conjectural (Coetzee and Gaston 2021) -e.g. if one champion had not emerged to conserve a species, perhaps another would have -there are clear benefits from applying counterfactual thinking to conservation (Ferraro 2009) and frameworks for its formal evaluation (McMurdo Hamilton et al. 2023).
Alternatively, a broader review of taxa could occur, either beyond Australia or across life-forms, as recently undertaken for vertebrate species no longer meeting any eligibility criteria for listing as threatened in Australia but which remained listed under the EPBC Act (Woinarski et al. 2023a(Woinarski et al. , 2023b)).Analyses of the social and institutional processes combined with data on the biology of the taxa concerned and the threats they face could yield many more lessons than could be achieved studying birds alone.

Predicting and monitoring reductions in extinction risk
Governments, conservation NGOs and others routinely invest in conservation actions (including research, management and monitoring), with many factors contributing to prioritisation of that investment.Our analyses show that it should be possible to predict explicit benefits (e.g.reduction in extinction risk, or prevention of extinction) and retrospectively evaluate benefits in a standardised manner that can be used to estimate return on investment and audit the conservation consequences of such investment.The most compelling example apparent from our analysis is that the AU $20 million investment in eradication of feral mammals on Macquarie Island resulted in the stepped improvement in extinction risk for nine bird taxa, representing 28% of the overall stepped improvements across Australian birds as a whole since 1990.
This analysis also highlights the fragility of reductions in risk.The fluctuations in the extinction risk status of the Norfolk Island Green Parrot are evidence that, at least in some cases, management attention cannot relax.That a taxon has had a higher risk category at some stage in its history is likely to be evidence that there is a sensitivity about the taxon that makes it susceptible to threats in the future, either old threats being resumed or new threats from another source.For the parrot, it was the same threat after management was relaxed.For the Australian Gould's Petrel and the Southern Eastern Bristlebird, the threats that increased extinction risk in 2020 were novel and unrelated to those that had been affecting the taxa before.Most recently there is evidence that the mortality of Tasmanian Wedge-tailed Eagles is high and likely to get higher -from 2010 until 2022, 272 were killed or injured at Tasmania's 197 wind turbines and power infrastructure (distribution and transmission lines) with an additional 996 turbines planned for construction by 2032 (Pullen 2023).An analysis of the latest shorebird count data also suggests that the Great Knot may have negative population trends that would justify relisting, but that declines of several other imperilled shorebird taxa do not now meet IUCN Red List criteria (Rogers et al. 2023).None of these recent examples has been assessed against the IUCN Red List criteria but all examples demonstrate how tenuous improvements may be.
Conversely, some of the examples we describe suggest that conservation actions are likely to have achieved benefits that endure without the need for ongoing management: conservation does not necessarily have to be an endless Sisyphean sentence.The most notable examples are for those bird species that have benefitted from the eradication of feral mammals from islands.Other than ongoing biosecurity, the job has been done, and the outcomes are likely to be robust and persistent.

Conclusion
Despite the small number of Australian bird taxa for which extinction risk has been reduced over the last 30 years, important lessons can be gained from each case where this has occurred.First, shifts to lower risk categories are rare.Second, the types of action that results in such shifts are varied, with some requiring only time to adapt to environmental change or shifts in pressure unrelated to conservation interventions.The third is that, for some taxa, the declaration of habitat as being protected has been sufficient to reduce risk to allow a shift between categories.Fourth, intensive interventions can yield successful reductions in extinction risk, but such hard-won improvements can be fragile and need ongoing monitoring and a willingness to continue adaptive management if it is required.
The variability in drivers of reduced extinction risk suggest a need for a larger data set if patterns are to emerge, either for groups other than birds or for a global set of birds, bearing in mind that the threat context, institutional environments and conservation capacity will have a profound impact on risk reduction.Even at the scale at which the current study could be conducted; however, the results indicate, happily, that extinction risk trends are not necessarily all one way.

Figure 1 .
Figure 1.Distributions of Australian imperilled bird taxa for which extinction risk has been reduced in at least one decade from 1990 to 2020.Distributions and vignettes from accounts in Garnett and Baker (2021).

Figure 2 .
Figure 2. Actions that contributed to the reduced extinction risk of 25 Australian imperilled bird taxa that experienced a reduction in extinction risk in at least one decade from 1990 to 2020.Categories are from IUCN (2012a) augmented with details of those within the category 'other'.The nine Macquarie Island taxa are labelled separately because they all benefited from the same set of interventions.

Figure 3 .
Figure 3. Results of generalised linear modelling of factors (median ± SE) affecting whether imperilled Australian birds have shown a reduction in extinction risk in any decadal period from 1990 to 2020.Scores are from the average of a candidate set of models within 2 Akaike units of the best model, after all possible models were estimated.For model details see Supplementary TablesS3 and S4).

Table 1 .
Classifications of drivers of extinction risk reduction identified in four publications.

Table 2 .
The IUCN Red List status (assessed retrospectively) in each decade from 1990 to 2020 for Australian imperiled bird taxa for which extinction risk was lower than the previous decade on at least one occasion.Green up arrows between decades indicate an improvement and red down arrows a deterioration in extinction risk (CR: Critically Endangered; EN: Endangered; VU: Vulnerable; NT: Near Threatened; LC: Least Concern;IUCN (2001)).

Table 3 .
Drivers of extinction risk reduction among threatened Australian birds assessed as moving to a lower extinction risk category in any decade from 1990 to 2020 (see Supplementary Table S4 for details).World Heritage listing of the Queensland Wet Tropics in 1988 prevented further clearing of rainforest, after which the southern population stabilised.No substantive threats to the northern population have ever been proposed.The gradual reduction in extinction risk reflects the proportion of the three generations (36 years) that occurred before habitat loss ceased.Drivers: Land/water protection, Law and policy, Education and awareness, Livelihood, economic and other incentives, Research and Monitoring, Other: continuity and adequacy of resources;

Table 3 .
(Continued).to have a small range that was being adversely affected by timber harvesting.Once that ceased in a large area of prime habitat, extinction risk declined.Drivers: Land/water protection, Other: continuity and adequacy of resources.Evidence of decline prompted a successful program aiming to spread, and thus lower, extinction risk by reintroducing taxon to previously occupied habitat.Unexplained declines in the two of the largest populations in 2020 resulted in re-assessment as Vulnerable in 2020.Drivers: Land/ water and Species protection and management, Law and policy, Education and awareness, Research and Monitoring, Conservation planning, Other: recovery 'champion', continuity and adequacy of resources.Had the greatest decline in reporting rate of any species between the first national bird Atlas