Correlates of avian extinction timing around the world since 1500 CE

1.   Introduction

Humans have brought about an alarming biodiversity crisis, considered by many to be part of a sixth mass extinction, as evidenced by current extinction rates greatly exceeding natural background ones (Ceballos et al., 2015, 2020; Ceballos and Ehrlich, 2018; Cowie et al., 2022). Consequently, we may be heading towards a future bird extinction rate higher than any since the Late Pleistocene (Şekercioğlu et al., 2004, 2008; Monroe et al., 2019; Andermann et al., 2021; Cooke et al., 2023). Avian extinctions have been relatively well documented in modern history, and in the past millennia, more bird species are known to have gone extinct than species in any other vertebrate class (Szabo et al., 2012; Ceballos et al., 2015; Cooke et al., 2023). The situation has been particularly dire for island endemic bird species, which have faced unique threats compared to those on the mainland. Among island endemic bird species living in 1500 CE, nearly half have either gone extinct or face the threat of extinction, with 39 endemic species that are considered Critically Endangered estimated to have fewer than 50 individuals remaining worldwide (Matthews et al., 2022). Most avian extinctions over the past five centuries were caused by people, either directly or indirectly (Szabo et al., 2012; Cooke et al., 2023). Significant drivers of extinction include habitats being lost, degraded, or fragmented by agriculture; hunting; and invasive species (Szabo et al., 2012; Cooke et al., 2023), all of which have been particularly prevalent in island systems. Furthermore, climate change and severe weather systems as well as residential and commercial development have become increasingly important drivers of extinction in the last century or so (Şekercioğlu et al., 2008; Wormworth and Şekercioǧ; lu, 2011; Szabo et al., 2012).

Prior research has identified several key correlates associated with heightened extinction risk in birds. One such trait is island endemism. Islands have served as fertile evolutionary grounds for speciation, such as through adaptive radiation, but they have also functioned as extinction hotspots due to the nature of their isolation (Curnutt and Pimm, 2001; Blackburn et al., 2004; Szabo et al., 2012; Cooke et al., 2023; Whittaker et al., 2023). Seabirds and ground-nesting birds on islands have been particularly vulnerable to introduced mammalian predators (Blackburn et al., 2004; Bellard et al., 2016; Spatz et al., 2017; Richards et al., 2021; Marino et al., 2022). Likewise, the absence of native mammalian predators on islands has supported the evolutionary loss of flight in certain island taxa (Sayol et al., 2020; Matthews et al., 2022; Cooke et al., 2023), which can render a species more susceptible to predation, limiting its ability to escape novel threats like predation by people and other mammalian predators (Sayol et al., 2020; Matthews et al., 2022), as well as impede movements to new, safer habitats and locations. Dispersal ability for both island and flightless species is further limited by their separation from the mainland. Similarly, limited dispersal ability due to shorter, rounder wings can also restrict a species' capacity to move to new areas and to escape deteriorating environments and emerging anthropogenic pressures (Moore et al., 2008; Sheard et al., 2020).

Body mass is another well-known trait associated with extinction (Fromm and Meiri, 2021; Matthews et al., 2022) and heightened vulnerability (Wang et al., 2018; Kittelberger et al., 2021; Rivas-Salvador and Reif, 2023). Bigger size tends to be associated with a number of potentially inhibitive species' life history traits including slower life histories and larger home ranges (Gaston and Blackburn, 1995; Bennett and Owens, 1997; Boyer, 2008; Matthews et al., 2022), as well as making those individuals more targeted by people for food. Generation length has been highlighted as an important determinant of extinction risk (Bird et al., 2020; Andermann et al., 2021; Lévêque et al., 2021), as species with longer generation lengths typically have slower population turnover, which can hinder their ability to adapt to environmental changes and other emerging threats quickly. Species exhibiting a higher degree of trophic specialization, in comparison to generalists in terms of both habitat and diet, also face challenges when their specific ecological niche is altered, resulting in them being less able to respond and adapt to disturbances (Şekercioğlu, 2011; Callaghan et al., 2020; Neate-Clegg et al., 2023). As such, forest specialists tend to be among the most threatened birds today, especially among islands in Southeast Asia (Kittelberger et al., 2021) and in fragmented tropical forests (Şekercioğlu and Sodhi, 2007). Likewise, with regards to diets, longer bill lengths in birds are often associated with specialized feeding behaviors (Matthews et al., 2022). While advantageous in stable environments, specific adaptations and characteristics for species may result in them being more vulnerable to extinction when individuals are faced with alterations in food availability, habitat structure and quality, or disturbance and predation.

A majority of studies examining correlates of extinction risk among avian species have primarily focused on a small number of traits, assessing the effects of these traits individually (e.g., Bennett and Owens, 1997; Norris and Harper, 2004; Şekercioğlu et al., 2004; Bird et al., 2020). Likewise, similar approaches have been used in studies looking at correlates of extinction among species that have gone extinct globally (Fromm and Meiri, 2021), including those from islands (Matthews et al., 2022). However, analyzing traits in isolation can limit our understanding of the power of identified extinction correlates while also making it more difficult to conclude the importance of traits to potential species' extinction (Henle et al., 2004; Kittelberger et al., 2021), especially since species are often impacted by more than one threat or trait (Davies et al., 2004; Wang et al., 2015, 2018; Kittelberger et al., 2021). While several recent studies have investigated the drivers of avian extinctions (Szabo et al., 2012; Fromm and Meiri, 2021; Matthews et al., 2022), studies typically lump extinctions together rather than separating out when species went extinct (but see Szabo et al., 2012). There is consequently a need for additional analyses to examine both the effects of multiple traits simultaneously on species' extinctions and how those traits correspond to when species went extinct. Moreover, global avian taxonomies are not yet unified (Neate-Clegg et al., 2021), which can result in some extinct species not being widely recognized. Furthermore, there are dozens of species that have not yet been officially classified as extinct even though they have not been detected in decades (Hume, 2017). As a result, including extinct species that may not be accepted by all main taxonomies as well those birds without recent confirmed sightings can lead to the inclusion of more taxa in extinction analyses, improve the strength of results, and provide a more complete and robust picture of drivers of global bird loss.

In this study, we examined the biological correlates of extinction timing among all bird species that have experienced modern-day, observed extinctions or disappearances since 1500 CE. We analyzed a broad range of traits and characteristics that have previously been associated with extinction and extinction risk in birds and compared the effects of these simultaneously against one another on how long since present day a species has been extinct or last detected. We predicted that species more likely to go extinct earlier in time would be island inhabitants (Fromm and Meiri, 2021; Matthews et al., 2022), flightless (Sayol et al., 2020; Fromm and Meiri, 2021; Matthews et al., 2022), having longer generation lengths (Bird et al., 2020; Andermann et al., 2021), large-bodied (Wang et al., 2018; Fromm and Meiri, 2021; Kittelberger et al., 2021; Matthews et al., 2022; Rivas-Salvador and Reif, 2023), more trophically specialized and likely inhabiting aquatic environments and feeding on vertebrates (van Weerd et al., 2003; Norris and Harper, 2004; Şekercioğlu, 2011; Sayol et al., 2021; Matthews et al., 2022), having limited dispersal ability (Moore et al., 2008; Sheard et al., 2020), and having longer bill lengths (Matthews et al., 2022). Through a comprehensive analysis of these correlates, we aimed to unravel the complex interplay between ecological, physiological, and behavioral factors that underlie species' susceptibility to extinction. However, we take this a step further and, by examining the time component of when species went extinct and how certain correlates varied in importance over time in relation to bird loss, we provide a novel perspective to understanding bird extinctions over the past five centuries. We also aimed to identify key locations and time periods in recent centuries where birds have gone extinct, and highlight specific taxonomic groups that have experienced heightened levels of extinction, building on prior work illuminating the hotspots and taxonomic patterns of extinction (Szabo et al., 2012; Matthews et al., 2022; Soares et al., 2024).

2.   Materials and methods

2.1   Dataset

We compiled a dataset of 216 bird species that have gone extinct or are potentially extinct since 1500 CE (Appendix Table S1), a time period considered "recent" or "modern" in which extinctions were typically observed and documented (Szabo et al., 2012; Sayol et al., 2021; Cooke et al., 2023). We restricted our analyses to recent extinctions since there tends to be information available for at least some traits, while there are more unknowns for species that disappeared before 1500 CE; additionally, only birds extinct after this date are included in the IUCN Red List (Matthews et al., 2022).

Avian taxonomic classifications were based primarily on BirdLife International and its list of the world's bird species, which are reviewed and adopted by the BirdLife Taxonomic Working Group (BirdLife International, 2024) and utilized by the IUCN Red List (IUCN, 2023). However, due to the taxonomic discord between the main avian taxonomic checklists (Neate-Clegg et al., 2021; Thomson et al., 2021), there were five species included in our dataset that were not recognized by BirdLife International and instead only accepted by either the IOC World Bird List (v. 14.2) and/or the Clements/eBird (v2023b) checklist (Appendix Table S1): Norfolk Ground Dove (Pampusana norfolkensis), Ascension Night-heron (Nycticorax olsoni), Puerto Rican Parakeet (Psittacara maugei), Lord Howe Parakeet (Cyanoramphus subflavescens), and Macquarie Parakeet (Cyanoramphus erythrotis). We also follow the IOC's treatment for two species: Tuamotu Kingfisher (Todiramphus gambieri) solely refers to the extinct population that inhabited Mangareva Island and does not include the extant population inhabiting Niau Island (Todiramphus gertrudae); and Turdus poliocephalus refers to the Tasman Sea Island Thrush which inhabited Norfolk and Lord Howe Islands and has now been split from other populations that were once considered part of the Island Thrush species. Consequently, we provide the scientific name for species as recognized by each of these three avian taxonomic checklists (Appendix Table S1).

The threat statuses for all species in our dataset consisted of Extinct (EX), Extinct in the Wild (EW), Critically Endangered but Possibly Extinct (CR (PE)), or Critically Endangered (CR) (Fig. 1); while one species, Coppery Thorntail (Discosura letitiae), is currently listed as Data Deficient, we considered it to be CR for our analyses. Most species with a status of CR (PE) or CR on our list have either not been detected in decades (some for more than a century) and/or lack any recent confirmed or credible sightings (Hume, 2017) and are believed to be extinct (e.g., Bahama Nuthatch Sitta insularis), except for the 'Akikiki (Oreomystis bairdi), a CR Hawaiian honeycreeper predicted to be on the verge of extinction in the wild (Paxton et al., 2022; Ong, 2023). For the purposes of this study, we analyzed all species that have potentially gone extinct and can currently be considered as lost (45: 22 CR (PE) and 23 CR) together with those that have been confirmed to be extinct (171: 166 EX and 5 EW). We chose to exclude from our dataset Rio de Janeiro Antwren (Myrmotherula fluminensis; collected in 1982), Tana River Cisticola (Cisticola restrictus; last recorded in 1972), and Blue-wattled Bulbul (Pycnonotus nieuwenhuisii; last recorded in 1992), due to speculation of these not being valid species; Red Sea Swallow (Petrochelidon perdita), an enigmatic species known only from the type specimen in 1984 that is suspected to potentially still exist; and two species currently listed as Critically Endangered – Kinglet Calyptura (Calyptura cristata), a poorly known and infrequently encountered Neotropical Tyrannid last recorded in 1996 after a hundred-year absence of records, and Sangihe White-eye (Zosterops nehrkorni), a species that has been sporadically encountered over the past few decades (last possible sighting in 2015) (BirdLife International, 2024). Finally, we followed avian taxonomic family designations as recognized by BirdLife International.

Figure  1.  The number of bird extinctions by decade since 1500 CE, representing 216 species that have disappeared or have likely disappeared over the past five centuries. IUCN Status rankings consist of EX (Extinct), EW (Extinct in the Wild), CR (PE) (Critically Endangered but Possibly Extinct), and CR (Critically Endangered). While most of the species with the latter two designations could still exist, they have not had any recent confirmed sightings in decades.

DownLoad: Full-Size Img PowerPoint

We collected data on 12 biogeographical, ecological, and life-history traits: insularity (i.e., restricted to an island(s)), endemism (i.e., restricted to a single country and/or island [chain]; for seabirds, breeding restricted to a single geographic area), flightlessness, generation length, body mass, primary habitat, habitat breadth, primary diet, dietary breadth, hand-wing index, and beak-culmen length (hereafter, "bill length"). A majority of these data were extracted from BirdBase, a global dataset of avian ecological traits for all of the world's bird species that is managed by the Şekercioğlu Lab at the University of Utah (Şekercioğlu et al., 2004, 2019; Kittelberger et al., 2021). Data for the following variables were supplemented or calculated by datasets from other sources: generation length (Bird et al., 2020), flightlessness (Fromm and Meiri, 2021; Sayol et al., 2021), hand-wing index (Sheard et al., 2020; Sayol et al., 2021; Tobias et al., 2022), and bill length (Sayol et al., 2021; Tobias et al., 2022). Gaps in body mass were also filled using some of these other datasets. If data for any trait in our dataset were not available from the literature for a species, we used congeneric values to fill in gaps. This resulted in the following number of species (out of 216) in our dataset having congeneric values for certain traits: generation length (6), body mass (15), hand-wing index (63), primary habitat (9), habitat breadth (10), primary diet (36), dietary breadth (38), and bill length (72); these values are highlighted in Appendix A.

Information on where species had occurred (particularly, specific islands) and the year of their extinction (or, for those not currently declared extinct or those that are now EW, the last confirmed/widely accepted year the species was sighted in the wild) was taken from Birds of the World (Billerman et al., 2023), Hume (2017), and/or Szabo et al. (2012). We then took the year of extinction and subtracted it from 2024 to calculate the years since present day that a species has been extinct (hereafter, "number of years extinct").

We divided our flightlessness category into three groups: "yes" (completely flightless), "no" (volant, able to fly), and "partial" (weak fliers, species that could fly but did so poorly and were therefore more akin to being flightless) (Sayol et al., 2020, 2021). We used primary diet and primary habitat as indicators of a species trophic level and habitat preference, respectively (Wang et al., 2018; Kittelberger et al., 2021). To simplify our diet categories, we grouped: piscivores, carnivores, and ovivores as "vertivores"; frugivores, nectarivores, and granivores as "herbivores"; kept "invertivores" separate; and used context to specify if the eight omnivorous species were primarily vertivores, invertivores, or herbivores (Kittelberger et al., 2021). To simplify our habitat categories, we grouped coastal, riparian, sea, and wetland habitats as "aquatic"; desert, plains, rocky, savanna, grassland, shrub, and shrubland categories as "non-forest"; and woodland with "forest", a similar approach used by Matthews et al. (2022). We then used habitat breadth and dietary breadth to calculate the ecological specialization index, quantified as Log10[100/(dietary breadth × habitat breadth)], which has a maximum of 2 for specialized species that live in one habitat and eat one type of food (Şekercioğlu, 2011; Kittelberger et al., 2021).

We chose to exclude endemism from our analysis since most of the species were endemic to a location (Fig. 2). There were only eight non-endemic species (4%) in our dataset, which could have skewed the relationship between this trait and the response variable. Our final dataset therefore consisted of nine traits for 216 species: insularity (yes/no), flightlessness (yes/partial/no), generation length, body mass, primary habitat (forest/non-forest/water), primary diet (herbivore/invertivore/vertivore), ecological specialization index, hand-wing index, and bill length.

Figure  2.  The proportion of 216 bird species globally extinct or likely lost since 1500 CE that were endemic to a region, found solely on islands, and were either flightless or partially flightless, as well as the breakdown of species by primary habitat or diet.

DownLoad: Full-Size Img PowerPoint

2.2   Statistical analyses

We log-transformed body mass and scaled all of our numerical variables to have a mean of 0 and a standard deviation of 1. We then created a generalized linear mixed-effects model (GLMM) to analyze the relationship between number of years extinct and our nine traits (Kittelberger et al., 2021). We also checked for variance inflation factors for the nine traits in our GLMM using the "vif" function from the R package car (Fox and Weisberg, 2019), with all below three (Zuur et al., 2010). Since our response variable consisted of integer values and our model was overdispersed, we used a Quasi-Poisson error structure with an 'nbinom1' family for our GLMM through the R package glmmTMB (Brooks et al., 2017). In order to control for taxonomy, we included avian taxonomic family as a random effect in our model. Finally, we ran submodels testing for potential ecologically significant interactions between our significant traits.

We then used the function "dredge" from the R package MuMIn (Bartoń, 2020) on our GLMM to run models for every possible subset of variables, rank these models according to AIC_c, and provide weights (w_i) for each model (Burnham and Anderson, 2002). Next, we subsetted all these possible models to those with ΔAIC_c < 6 (Harrison et al., 2018), considering the one with the lowest AIC_c to be our top model best supported by the data. If there were any competing models within a ΔAIC_c value of 2 from our top model, we noted these along with their Akaike values, weights, and coefficients (Burnham and Anderson, 2002; Cade, 2015) using the R package AICcmodavg (Mazerolle, 2020). For any categorical variables with more than two categories in our top model, we used a likelihood ratio test to determine significance and then used the "tapply" function in R to determine the mean year of extinction for each category.

All statistical analyses and graphing were conducted in R (version 4.3.1, 2023-06-16; R Core Team, 2023).

3.   Results

3.1   Modelling and extinction correlates

There were 512 total models of our GLMM after dredging (Appendix Table S2). After multi-model comparison, there were 20 competing models within ΔAIC_c < 6 and 4 models within ΔAIC_c < 2. Our top-ranked model contained five covariates (AIC_c = 2510.493, AIC_cw_i = 0.397); our second-ranked model included the same significant covariates as well as bill length, which was not significant (ΔAIC_c = 0.510, AIC_c = 2512.443, AIC_cw_i = 0.307); our third-ranked model contained four significant covariates (ΔAIC_c = 1.950, AIC_c = 2512.485, AIC_cw_i = 0.150); and the fourth-ranked model contained the original five significant covariates as well as diet, which was not significant (ΔAIC_c = 1.992, AIC_c = 8387.894, AIC_cw_i = 0.146). Since these top models were nested, we conducted a likelihood ratio test to determine which model fit our data better (Jones et al., 2022). We found that the top-ranked model was a significantly better fit (p = 0.042), and we therefore treat the top model as our best model. As a result, our top model contained five correlates of extinction timing, all of which were present across a majority of competing models: insularity (in 100% of competing models), flightlessness (85%), body mass (100%), ecological specialization (65%), and hand-wing index (100%).

Insularity was significantly, positively associated with the number of years extinct in our top model (0.926 ± 0.148, z = 6.241, p < 0.001), such that island species (189) were more likely to go extinct earlier than non-island species (27; Fig. 3A). The average number of years extinct for non-insular species was 71.1 and 172.0 for island species, which corresponds with the years 1953 and 1852. For flightlessness (χ² = 9.196, p = 0.010), species that were flightless or partially flightless were more likely to go extinct earlier than volant species (Fig. 3C). The average number of years extinct for partially flightless species was 304.6, 213.4 for flightless species, and 140.2 for volant species; this corresponds with the years 1719, 1811, and 1884, respectively.

Figure  3.  Generalized linear model predictions, with a Quasi-Poisson error structure, between how many years a species has been extinct or potentially extinct since the present day and (A) a species' body mass and occurrence on islands, (B) a species' body mass and degree of flightlessness, (C) a species' hand-wing index and degree of flightlessness, and (D) how specialized a species was. We hold other non-categorical variables in our top model to be equal to their mean values (0) for each figure. All of these variables had a significant effect with number of years extinct (p < 0.05; see Results). The lines represent trendlines from GLMs with 95% confidence intervals. The x-axes for A and B are on a logarithmic scale. These figures represent 216 species that have gone extinct or have potentially gone extinct since 1500 CE.

DownLoad: Full-Size Img PowerPoint

Body mass was significantly, positively associated with number of years extinct (0.256 ± 0.043, z = 6.005, p < 0.001), such that heavier birds have been extinct on average longer than smaller-bodied species (Fig. 3A). There was a significant and positive relationship with ecological specialization index (0.080 ± 0.039, z = 2.033, p = 0.042), with more specialized species having gone extinct before more generalist species (Fig. 3D). Finally, hand-wing index was also significantly, positively associated with the number of years extinct (0.200 ± 0.042, z = 4.809, p < 0.001), with birds that had pointier wings going extinct before those with more rounded ones (Fig. 3C).

Among our un-dredged submodels with all traits, we found significant interactions between two sets of significant traits. There was a significant interaction between body mass and insularity (0.276 ± 0.130, z = 2.133, p = 0.033), such that large-bodied, island-endemic birds were predisposed to earlier extinction (Fig. 3A) than smaller-bodied, continental birds. There was also a significant interaction between body mass and flightlessness (χ² = 29.425, p < 0.001), such that large-bodied, flightless birds were predisposed to earlier extinction (Fig. 3B); specifically, heavier partially flightless birds went extinct before heavier fully-flightless species, though this is likely due to there being far fewer partial than fully flightless species.

3.2   Taxonomic extinction proneness

Four avian families have had the highest number of extinct species over the past five centuries (Fig. 4A): Rallidae (rails, 26 species), Fringillidae (true finches, 24 species), Psittacidae (holotropical parrots, 23 species), and Columbidae (pigeons and doves, 20 species). All of the 26 extinct rail species were restricted to islands, and most were flightless. All but one of the 24 extinct finch species were Hawaiian honeycreepers, with a majority disappearing at the end of the 19th or beginning of the 20th centuries. For Psittacids, 18 were island inhabitants and all but three were forest-dependent. Note that Psittacidae in our dataset includes species recognized by other taxonomic authorities as belonging to Psittaculidae. Likewise, among the Columbids, all but one species were island inhabitants and all were forest-dependent.

Figure  4.  (A) The number of extinctions by avian family, and (B) the proportion of extinctions within a bird family, since 1500 CE. IUCN Status rankings consist of EX (Extinct), EW (Extinct in the Wild), CR (PE) (Critically Endangered but Possibly Extinct), and CR (Critically Endangered). In (B), the number of species that have gone extinct in a family is noted on top of each bar.

DownLoad: Full-Size Img PowerPoint

However, while these families listed above may have the highest number of extinct species, they do not represent the families with the highest proportion of extinctions (Fig. 4B). The family Mohoidae (Hawaiian Honeyeaters) has the highest proportion of extinct birds at 100%, with all five species extant after 1500 CE now extinct, followed by Acanthisittidae (New Zealand Wrens: 50% extinct, 2 species) and Callaeidae (New Zealand wattlebirds: 40% extinct, 2 species).

3.3   Geographic and temporal extinctions

Among the decades that 216 species have gone extinct or are potentially extinct since 1500 CE, there are two notable peaks of avian disappearance: the 1890s with 21 species, and the 1980s with 20 species (Fig. 1, Appendix Table S1). All but one of the extinctions in the 1890s were of island-inhabiting species, with 10 (48%) species residing in the Hawaiian Islands and 5 (24%) on New Zealand islands; 9 (43%) of these species come from the family Fringillidae (all of which were Hawaiian Honeycreepers). The trait composition of the species from the 1980s is more varied, in both location, island-inhabitance (14 of the species were insular), and taxonomic family. The location with the highest number of extinctions this decade was still the Hawaiian Islands, with 6 (30%) species; however, 11 (55%) of these species occurred at least in part within the United States or in a territory controlled by the United States (i.e., Guam; Appendix Table S1). Since 1500 CE, the region with the highest loss of bird species has been the Hawaiian Islands (34 species: 32 extinct, ~2 extinct in the wild), followed by the Mascarene Islands (consisting of Mauritius, Réunion, and Rodrigues Islands; 32 species, of which 12 were on Mauritius and 11 on Réunion), New Zealand and its islands (20 species, including the Chatham Islands), and French Polynesia (17 species) (Figs. 5 and 6).

Figure  5.  Global hotspots of avifaunal extinction since 1500 CE, showing top regions with 4 or more lost bird species. The Mascarene Islands consist of Mauritius, Réunion, and Rodrigues Islands; New Zealand includes the Chatham Islands; Australia includes Lord Howe Island (but Norfolk Island is treated separately due to the proximity to the mainland); the United States refers to the continental region; Mexico includes Guadalupe, Socorro, and Cozumel Islands; and the Lesser Antilles consists of Guadeloupe, Martinique, St. Lucia, and St. Kitts. Losses of four species are noted under multiple countries (see Appendix Table S1) to account for species that had cross-territorial ranges (i.e., Labrador Duck, Passenger Pigeon, Eskimo Curlew, and Great Auk in both the United States and Canada).

DownLoad: Full-Size Img PowerPoint

Figure  6.  A map of global bird loss since 1500 CE, including species confirmed extinct and those likely to have disappeared. The size of the points corresponds with the density of bird species for a specific site. Where possible or notable, singular island-occurrences within a prominent archipelago or island group (i.e., Hawaiian or Mascarene Islands) have been specified to a particular island (e.g., Kauai or Mauritius, respectively) to better reflect hotspots of extinction; distributions in continental countries are centered within the country itself.

DownLoad: Full-Size Img PowerPoint

4.   Discussion

In modern times (since 1500 CE), over 200 bird species are observed to have gone or are likely to have gone extinct. This number of extinctions, and the rate of species loss itself, is forecast to further accelerate in the future (Monroe et al., 2019; Andermann et al., 2021; Cooke et al., 2023), including at the subspecies level (Szabo et al., 2012). The situation is further compounded by the fact that many bird extinctions have likely gone undiscovered (Pimm et al., 2006; Duncan et al., 2013; Cooke et al., 2023), particularly on islands and other remote areas. While threats to birds are expanding and multiplying, the conservation efforts to prevent extinctions have grown and become more focused, and therefore there is a need to better understand the traits that may drive some species towards extinction sooner than others. In this study, we conducted a comprehensive analysis of multiple traits to investigate the biological correlates of avian extinctions, providing a novel approach that incorporated the timing of when species went extinct or disappeared across the world's recently lost avian taxa. We found strong support across our competing models for insular, flightless, larger-bodied, ecologically specialized, and pointier-winged species having been more predisposed to go extinct earlier in time over the past five centuries.

4.1   Biological correlates of extinction

In the past five centuries, 88% of known avian extirpations have consisted of island species (Fig. 2), with around half of all island endemic birds (known to be extant in 1500 CE) today either extinct or threatened with extinction (Matthews et al., 2022). The restricted size of islands results in species inhabiting these locations naturally having small populations as well as having access to limited refugia from threats (Steadman, 2006; Whittaker et al., 2023), which would have contributed to these birds disappearing on average earlier than those on the mainland (Fig. 3A). Additionally, there is a strong relationship between the evolution of flightlessness in avian taxa and insularity (Steadman, 2006; Whittaker et al., 2023), and this lack of flight ability makes these birds especially vulnerable to predation from both introduced predators and humans (Sayol et al., 2020; Matthews et al., 2022). It comes as no surprise that 95% (41 species) of the flightless birds that have disappeared since 1500 CE have occurred on islands, with flightless birds comprising 22% of all insular species that have disappeared (Appendix Table S1). We found that species that were partially flightless were more likely to disappear first (perhaps this was influenced though by the smaller sample size of 10 partially flightless species), followed by birds that were entirely flightless (Fig. 3B and C). While fewer than 1% of living bird species are flightless (Fromm and Meiri, 2021), and only about 25% of modern bird orders contain a flightless species, prehistorically this trait was much more abundant and present in around 60% of bird orders (Sayol et al., 2020). In fact, the true scope of extinction proneness of both insular and/or flightless birds is concealed in the modern day since hundreds to thousands of these particular species had already been driven extinct prior to 1500 CE as humans dispersed through and settled across islands (Steadman, 2006; Sayol et al., 2020), especially in the Pacific (Curnutt and Pimm, 2001).

For body mass, we found that larger birds were more likely to go extinct closer to 1500 CE (Fig. 3A), of which the four heaviest birds in our dataset were two Columbids and two Casuariids: Rodrigues Solitaire (Pezophaps solitaria), Dodo (Raphus cucullatus), Kangaroo Island Emu (Dromaius baudinianus), and King Island Emu (Dromaius minor) (Appendix Table S1). Heavier birds have been more likely to be targeted for hunting, with several well-known examples of birds being hunted to extinction in part for food, including the Dodo, Great Auk (Pinguinus impennis), and Spectacled Cormorant (Urile perspicillatus). Body mass has also been linked with a species' inhabitance of islands (Fromm and Meiri, 2021; Matthews et al., 2022), as island birds tend to be larger than mainland species (Olson et al., 2009). Larger size could have aided birds in more successfully colonizing islands, and islands would have also functioned as nesting safe-havens from land predators for large seabirds (Fromm and Meiri, 2021). Of the 43 extinct species with average body masses above 500 g (Appendix Table S1), 32 species (74%) were endemic to an island(s). Large body size was likely even more notable among prehistoric extinctions in the Quaternary (Boyer, 2008; Fromm and Meiri, 2021), with many of the largest bird species, such as the moas (Dinornithiformes), Haast's Eagle (Hieraaetus moorei) of New Zealand, and the elephant birds (Aepyornithiformes) of Madagascar, already completely extirpated by humans before 1500 CE. However, larger size would have also resulted in species not only being more likely to be hunted for food but also to recover more slowly from population declines due to having traits such as slower life histories, smaller population sizes, and larger home ranges (Gaston and Blackburn, 1995; Bennett and Owens, 1997; Boyer, 2008; Matthews et al., 2022). We however did not find a significant influence of generation length on when species went extinct.

We found that birds with larger hand-wing index were more likely to have gone extinct earlier (Fig. 3C). This result was contrary to our expectations. Because hand-wing index is positively associated with dispersal ability and therefore should aid a species in responding to environmental or anthropogenic pressures (Sheard et al., 2020), we predicted that birds with less-pointy and more rounded wings would have been more predisposed to extinction. However, our hand-wing index finding does mirror the relationship found globally among extant threatened island endemic birds (Matthews et al., 2022). There could be an ancestral component to this hand-wing index relationship though, as species that colonized islands would have likely needed to have had a high dispersal ability and therefore high aspect ratio wings to reach these locations (Matthews et al., 2022; Whittaker et al., 2023).

Finally, looking at the trophic niches of species, we found that more ecologically specialized species were more likely to go extinct earlier in time (Fig. 3D). This means that species which existed in only one dietary and habitat niche (ecological specialization index = 2) tended to go extinct first. Highly specialized species are more vulnerable since they face more dietary and habitat constraints compared to more generalist species that can thrive under a wider variety of conditions (Şekercioğlu, 2011; Callaghan et al., 2020; Neate-Clegg et al., 2023; Rivas-Salvador and Reif, 2023) and therefore better respond to environmental or anthropogenic threats and disturbances. While we did not find that a particular habitat or diet had a significant influence on when species disappeared, we found that among the 48 highly specialized species that have become lost, 28 (58%) were invertivores while 34 (71%) inhabited forests.

Additionally, for the most part (Appendix Table S1), the most highly specialized species to have gone extinct since 1500 CE occurred on islands (89% of highly specialized birds; but see Crested Shelduck (Tadorna cristata), Alagoas Curassow (Mitu mitu), Colombian Grebe (Podiceps andinus), Slender-billed Grackle (Quiscalus palustris), and Cryptic Treehunter (Cichlocolaptes mazarbarnetti)), and most of these island species were found outside of the Nearctic or Palearctic (95% of species, except Bermuda Towhee (Pipilo naufragus) and Canarian Oystercatcher (Haematopus meadewaldoi)). These latter findings about geographic realm mirror those among threatened species today (Rivas-Salvador and Reif, 2023).

4.2   Taxonomic and geographic hotspots of extinction

We found that the taxonomic families with the most extinct birds over the past five centuries have been Rallidae, followed by Fringillidae, Psittacidae, and Columbidae (Fig. 4A), which mirrors what has been noted in the literature for these families at both the species and ultrataxon level (Szabo et al., 2012). Rails have been widely known as the most extinction-prone bird family (Curnutt and Pimm, 2001; Steadman, 2006; Lévêque et al., 2021), and today one-third of extant rail species are globally threatened (Lévêque et al., 2021). While 26 rail species were extirpated since 1500 CE (Fig. 4A), the situation is even more dire when we consider extinctions that pre-date 1500 CE: it is estimated that hundreds to possibly over one thousand flightless rail species, particularly in the Pacific, were driven to extinction after the human settlement of islands (Curnutt and Pimm, 2001; Steadman, 2006; Kirchman, 2012; Sayol et al., 2020; Lévêque et al., 2021). Columbiformes and Psittaciformes are also widely represented in the extinction debt among islands, both before and after 1500 CE, with a number of islands historically supporting multiple endemic parrots and pigeons and/or doves (Curnutt and Pimm, 2001; Hume, 2017; Matthews et al., 2022). Similar to rails, it is likely that hundreds of parrot as well as pigeon and dove species went extinct over the past several thousand years across islands (Curnutt and Pimm, 2001). Besides the Dodo, one of the most famous birds to go extinct recently was the mainland Passenger Pigeon of North America, also a Columbid, which at one point was likely the most numerous bird species on the planet (Hume, 2017).

In modern times though, more finches have gone extinct than Columbids or Psittacids as a result of the strong anthropogenic drivers of extinction across the Hawaiian Archipelago; in fact, all but one of the 24 Fringillid species to disappear since 1500 CE have been Hawaiian honeycreepers. The tragedy of this group of endemic, specialized finches is emblematic of the general plight of Hawaiian birds, and this has unfortunately led to these islands becoming the primary hotspot (Figs. 5 and 6) of recent global bird extinctions (i.e., Boyer, 2008; Szabo et al., 2012; Kaushik et al., 2018; Matthews et al., 2022; Cooke et al., 2023). Habitat destruction, introduced predators, and avian malaria from accidently introduced mosquitoes initially wiped out lowland Hawaiian endemics (Boyer, 2008). However, while the spread of the malaria on the islands was originally elevationally restricted due to the thermal thresholds for the reproduction of the parasite, rising temperatures from climate change have now caused the mosquito and the malaria it carries to encroach into remaining high elevational forest refugia and decimate the remaining honeycreepers (particularly on Kauai and Maui) (Berio Fortini et al., 2020; Judge et al., 2021; Paxton et al., 2022; Gallerani et al., 2023; Neddermeyer et al., 2023). Just this year, another Hawaiian species, the 'Akikiki, is likely to become Extinct in the wild due to the expansion of avian malaria on Kauai (Paxton et al., 2022; Ong, 2023), and several other Hawaiian birds (Puaohi (Myadestes palmeri), Palila (Loxioides bailleui), 'Ākohekohe (Palmeria dolei), Maui Parrotbill (Pseudonestor xanthophrys), and 'Akeke'e (Loxops caeruleirostris)) that number in the low hundreds or less are on the precipice of extinction due to factors such as avian malaria and declining habitat quality (Judge et al., 2021; Paxton et al., 2022; Gallerani et al., 2023; BirdLife International, 2024). The most recent extinction wave across the Hawaiian Islands also saw the only complete extirpation of an entire bird family in modern times (Lovette, 2008), Mohoidae (Hawaiian Honeyeaters; Fig. 4B), with the remaining individual of the final species, Kauaʻi ʻōʻō (Moho braccatus), last detected in 1987. While there are 34 total bird species that have gone extinct or likely disappeared from the Hawaiian Islands since 1500 CE (including 'Akikiki), the true extent of loss of endemic birds from the archipelago is much higher when including prehistoric extinctions of taxa such as the moa-nalo (Thambetochenini) that disappeared after the arrival of Polynesians around 1600 years ago (Boyer, 2008; Sayol et al., 2021; Cooke et al., 2023).

However, the Hawaiian Archipelago is not the only set of islands that is a hotspot of extinction since 1500 CE (Fig. 6). The Mascarene Islands, New Zealand islands, and French Polynesia (Fig. 5) also experienced a high amount of avian loss in modern times (Szabo et al., 2012; Matthews et al., 2022; Cooke et al., 2023; Soares et al., 2024). Once again, the situation is even more dire when considering prehistoric extinctions (Duncan et al., 2013; Sayol et al., 2021; Cooke et al., 2023). Across these four island archipelagos, there are currently 38 Endangered and 30 Critically Endangered species (excluding the CR species that we included in our study): Hawaiian Islands, 8 EN/9 CR; Mascarenes, 4/3; New Zealand, 18/8; and French Polynesia, 8/10 (BirdLife International, 2024). As known hotspots of avian loss, it is imperative that conservation efforts be amplified across these islands to prevent these globally threatened species from disappearing and halt the loss of functional diversity from these islands (Sayol et al., 2021). On the other hand, it is worth noting that continental Africa largely has a dearth of recorded avian loss, especially compared to the other [non-Antarctica] continents (Fig. 6); loss of African species has instead been largely restricted to off-shore islands. The one extinction for continental Africa is of Slender-billed Curlew (Numenius tenuirostris), which wintered in parts of northern Africa such as Algeria.

5.   Conclusion

In this study, we used a broad range of biogeographical, ecological, and life history traits to identify the important biological correlates of when species went extinct among the world's avifauna that have disappeared since 1500 CE. We build upon what has been reported previously as characteristics associated with extinction and extinction risk (i.e., Fromm and Meiri, 2021; Kittelberger et al., 2021; Matthews et al., 2022; Rivas-Salvador and Reif, 2023) that can help aid conservation efforts targeting species most susceptible to extinction. Compared to prior studies, we examine a more comprehensive set of traits across a larger, more complete list of avian taxa that includes species currently recognized by at least one of the three main avian taxonomies as well as those species that are not officially listed as extinct but have likely disappeared. Importantly, we examine biological correlates of bird extinctions through the lens of when birds went extinct, providing a novel extinction timing element that helps better inform why birds with certain traits disappeared when they did. Additionally, we provide a complete, unified, and robust dataset of the traits we compiled for extinct and likely extinct bird species (Appendix Table S1). There is a lack of extensive trait data on modern extinct species (Matthews et al., 2022), and prior public datasets tended to include fewer traits with less comprehensive data for extinct species. This dataset will therefore be useful to future researchers examining avifaunal extinctions. For examining trends among extinct species, we employed a novel approach, modelling traits against the number of years since present day that a species has been extinct and providing the average year birds with specific traits went extinct since 1500 CE. We noted that the 1980s represents one of the two highest decades of avian loss since 1500 CE (Fig. 1), which is noteworthy since focused and targeted conservation efforts were already underway globally before and during this period. Though the rate of extinction declined in the subsequent decades (Fig. 1), the number of globally threatened species has only increased (IUCN, 2023), which is compounded by the fact that some of the main predictors of global threat have remained the same over the past few decades (Rivas-Salvador and Reif, 2023)

While we focused on observed avian losses in modern history, we did not take into account known extinctions prior to 1500 CE, both those recorded and those only known from fossils. Our study therefore does not account for the full scope of known bird extirpations, many of which occurred due to humans, and the traits these birds may have possessed. Likewise, we are not able to account for undiscovered extinct birds, of which there may have been close to 1000 before recent history (Curnutt and Pimm, 2001; Duncan et al., 2013; Cooke et al., 2023). It is also possible that the effects of some of our correlates, like hand-wing index and diet, on when species went extinct are affected by the loss of species before the study period, leading to the removal of those trait values from the modern extinction pool (Boyer, 2008). Over the past roughly 7000 years, there have been three primary bird extinction waves, with the most devastating wave occurring across the Eastern Pacific around the 14th century and representing the biggest human-driven vertebrate extinction event to date (Cooke et al., 2023). However, we are likely heading towards an extinction rate that is even larger than any since the Late Pleistocene (Monroe et al., 2019; Andermann et al., 2021; Cooke et al., 2023), underscoring the need to understand correlates that already drove species extinct so that current conservation measures can be better informed on how to minimize and prevent additional extinctions going forward.

CRediT authorship contribution statement

Kyle D. Kittelberger: Writing – review & editing, Writing – original draft, Visualization, Supervision, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Colby J. Tanner: Writing – review & editing, Validation, Methodology. Amy N. Buxton: Writing – review & editing, Writing – original draft, Data curation. Amira Prewett: Writing – original draft, Data curation. Çağan Hakkı Şekercioğlu: Writing – review & editing, Resources, Conceptualization.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could serve as a potential conflict of interest.

Acknowledgements

We thank Montague H. C. Neate-Clegg for his feedback on some of the modeling and visualization approaches utilized in this study. We also thank Natalie Zorn and Nick Putz for their assistance with collecting and proofing data for some of the species utilized in this project. We are grateful to H. Batubay Özkan and Barbara Watkins for their support of the Biodiversity and Conservation Ecology Lab at the University of Utah, School of Biological Sciences.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.avrs.2024.100213.