You are on page 1of 5

Conservation biology

The dynamics underlying avian extinction trajectories forecast a wave of extinctions
Melanie J. Monroe1,2,3, Stuart H. M. Butchart4,5, Arne O. Mooers2
and Folmer Bokma1,6
Research 1
Center for Ecological and Evolutionary Synthesis (CEES) Department of BioSciences, University of Oslo, PO Box
Cite this article: Monroe MJ, Butchart SHM, 1066, Blindern, 0316 Oslo, Norway
Department of Biological Sciences and the IRMACS Center for Interdisciplinary Research, Simon Fraser
Mooers AO, Bokma F. 2019 The dynamics University, 8888 University Drive, Burnaby, BC, Canada, V5A 1S6
underlying avian extinction trajectories forecast Department of Ecology and Genetics, Evolutionary Biology Center (EBC), Uppsala University, Norbyvägen 18D,
a wave of extinctions. Biol. Lett. 15: 20190633. 75236 Uppsala, Sweden
BirdLife International, David Attenborough Building, Pembroke Street, Cambridge CB2 3QZ, UK 5
Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK
Department of Ecology and Environmental Science and IceLab, Umeå University, 90187 Umeå, Sweden
SHMB, 0000-0002-1140-4049; AOM, 0000-0003-0383-8856; FB, 0000-0002-0049-2127
Received: 26 August 2019
Population decline is a process, yet estimates of current extinction rates often
Accepted: 15 November 2019
consider just the final step of that process by counting numbers of species
lost in historical times. This neglects the increased extinction risk that affects
a large proportion of species, and consequently underestimates the effective
extinction rate. Here, we model observed trajectories through IUCN Red List
Subject Areas: extinction risk categories for all bird species globally over 28 years, and esti-
ecology mate an overall effective extinction rate of 2.17 × 10−4/species/year. This is
six times higher than the rate of outright extinction since 1500, as a conse-
quence of the large number of species whose status is deteriorating. We
very conservatively estimate that global conservation efforts have reduced
biodiversity, birds, mass extinction, endangered the effective extinction rate by 40%, but mostly through preventing critically
species, conservation, population decline endangered species from going extinct rather than by preventing species
at low risk from moving into higher-risk categories. Our findings suggest
that extinction risk in birds is accumulating much more than previously
appreciated, but would be even greater without conservation efforts.
Author for correspondence:
Folmer Bokma
1. Introduction
Recent global biodiversity loss is estimated to be at least one hundred times
pre-human levels [1–3]. However alarming, these estimates may be too optimis-
tic [4]. Estimates of current or recent extinction rates have typically been based
on the numbers of species within a particular group that we know or suspect to
have gone extinct over a set period of time [5,6]. However, this simple calcu-
lation combines species that are currently not at risk with those whose
populations are declining but that have not yet been lost [7]. Given that species
must decline from ‘not at risk’ through various levels of risk before extinction,
including these trajectories in calculations would offer a more comprehensive
measure of ongoing extinction dynamics. Here, we estimate the overall effective
extinction rate from changes in the IUCN Red List [8] categories of extinction
risk for all 11 064 recognized avian species over 28 years (1988–2016), and
assess the impact of conservation efforts on this rate.

Electronic supplementary material is available

online at 2. Material and methods
The IUCN Red List uses seven categories to classify extinction risk: least concern
(LC), near threatened (NT), vulnerable (VU), endangered (EN), critically endangered

© 2019 The Author(s) Published by the Royal Society. All rights reserved.
Table 1. Annual rates at which bird species moved between IUCN Red List categories. The values are based on genuine movements between categories (see §2 2
for abbreviations) from the years 1988–2016. For example, the value in row NT and column VU is the probability (0.0024) that a species currently classified as
NT will next year be classified as VU. Thus, values across each row sum to one (with rounding errors). Shading intensity indicates the magnitude of transition
rates, with reds denoting movement to higher, and greens denoting movement to lower, risk levels. #spp gives the numbers of currently recognized species per
category in 2016, including extinct species, and K is the distribution of species over categories in 2016 excluding species already extinct. CR includes CR(PE) and
CR(PEW). Lifetime T is the expected time to extinction in years for a species currently in a given category (see §2). ‘with conservation’ is the scenario where
rate estimates include both up- and down-listings between extinction risk categories. ‘without conserv.’ is the scenario where estimates exclude those down-
listings that were the result of conservation efforts.

probability (× 10−4 yr−1) of transition to category lifetime T (yr)

from LC NT VU EN CR EW EX #spp K (%) with conservation without conserv.

LC 9993 5 2 0 0 0 0 8417 76.0 5161 (3770–7502) 3347 (2736–4287)

Biol. Lett. 15: 20190633

NT 6 9967 24 1 3 0 0 1017 9.2 3959 (2544–6230) 2161 (1578–3062)
VU 1 11 9950 33 5 0 0 786 7.1 3432 (2090–5664) 1691 (1140–2566)
EN 0 6 18 9937 37 1 0 461 4.2 3054 (1696–5253) 1371 (839–2241)
CR 1 1 6 57 9898 32 5 222 1.8 2503 (1226–4612) 1015 (537–1849)
EW 2 2 4 14 142 9679 156 5 0.1 1366 (482–2997) 598 (252–1249)
EX 0 0 0 0 0 0 104 156 0 0
weighted average: 4780 (3418–7093) 2985 (2400–3893)

(CR), extinct in the wild (EW) and extinct (EX). The classification inverse, the per-species, per-year extinction rate: the global
of a species changes if it becomes more or less threatened over average across all birds in 2016 was 4780 years (95% credible
time, and we assume that this process can be modelled as a interval: 3418–7093). (Henceforth, intervals following esti-
time-homogeneous Markov process with annual transition mates are 95% credible intervals.) This corresponds to an
matrix Q. (We found no evidence of a trend over time that
overall extinction rate of 2.17 × 10−4 (1.41–2.92 × 10−4)/
suggests a more complex model to better describe the data.)
species/year (≈2090 extinctions per million species years
This process has EX as the absorbing state, because once a species
(E/MSY) [13] but see electronic supplementary material,
is extinct, it will not re-appear.
The ongoing rate of extinction is calculated from Q, using Methods). Because times to extinction are exponentially dis-
standard Markov chain theory [9], as follows: let R denote the tributed, the median time to extinction is considerably
6 × 6 transient segment of Q, that is, Q without the row and shorter than the mean: 50% of present-day species would
column EX (table 1). The matrix F = (I − R)−1, where I is the iden- be lost already after loge(2) · 4780 = 3313 (2369–4917) years.
tity matrix, is the fundamental matrix of Q: entry fij of F is the This projection is 1000-fold shorter than the 3 Myr estimate
expected number of times that a species currently in the ith cat- for pre-human avian species durations [14].
egory will be in the jth category before going extinct. The Importantly, our Q-based estimate of the time to extinction,
expected time until extinction for each transient starting state which takes into account transition rates between all categories
is, therefore, T = Fc, where c is a 6 × 1 column vector of ones.
from LC through EX, is much shorter than traditional estimates
Let K be a 1 × 6 vector with fractions describing the current dis-
that only count transitions to EX. To illustrate the underestima-
tribution of species over the transient extinction risk categories
tion of the extinction rate caused by lumping all non-EX
(ΣK = 1; table 1). The average time to extinction is then KT,
and the rate of extinction is the inverse of the average time to categories and considering only transitions between ‘non-EX’
extinction: (KT)−1. Thus, we can calculate the scalar extinction and EX, we can sum the product of the columns labelled ‘EX’
rate from transition matrix Q, which is based on transitions and ‘K’ in table 1 to get the per-year extinction probability of
between all extinction risk categories, not just between the an average species: 1/24 492 (1/42 788–1/15 983)/species/
most threatened categories and EX. year. Thus, lumping all non-EX categories causes a 24 492/
We estimated Q from the extinction risk categories of 11 064 bird 4780 = 5-fold (3.4–8.5) overestimation of the time to extinction,
species in 1988, 1994, 2000, 2004, 2008, 2012 and 2016 (the years in because it neglects the net tendency for species at low risk to
which the status of all species has been assessed) according to the move into higher-risk categories. More directly, consider that
IUCN Red List, which is based on data provided by BirdLife Inter-
during the past 500 years, about 187 of 11 064 avian species
national [10,11]. Improved knowledge about taxonomy or threat
are documented to have gone extinct [15]. If the per species
factors was retroactively applied [10,12] (electronic supplementary
per year probability of extinction is pe, then the fraction of
material, Data). We used a Bayesian algorithm (electronic supplemen-
tary material), which allowed us to take into account uncertainty species expected to be extinct after 500 years is 1 − (1 − pe)500.
about the status of species, including those CR species tagged as Equating that fraction to 187/11 064 yields an expected time
‘possibly extinct’ (PE) or ‘possibly extinct in the wild’ (PEW). to extinction of p−1
e = 29 333 years: six times longer (4.1–8.6)
than our current estimate based on Q.
To illustrate how the tendency for low-risk species to
move to higher-risk categories affects the extinction rate, we
3. Results used the matrix product QK (see §2) to project the classifi-
From our estimate of the full matrix Q (table 1), we estimate cation of the present-day species far into the future
the expected time to extinction per species and therefore its (figure 1). We emphasize that this is an illustration of the
(a) 3 3


per year
with conservation
1 no conservation

10 000 EX EW
10 000 EN CR
7500 7500 NT
no. species

5000 LC

5000 2500

Biol. Lett. 15: 20190633

2500 0 100 200 300 400 500

500 2500 5000 7500 10 000 12 500 15 000
time (years from present)
Figure 1. Illustration of ongoing bird extinction dynamics as implied by the rate matrix Q, based on the rates at which bird species have moved between IUCN Red
List categories from 1988 to 2016. We project these dynamics to illustrate how current bird species diversity would hypothetically decrease in the future given recent
trends, rather than to represent realistic predictions, because Q is unlikely to remain constant over this time period. Solid lines: projected extinctions based on Q
estimated using all observed category transitions. Dashed lines: projected extinctions based on Q estimated without category transitions attributed to conservation
efforts. (a) Projected yearly number of extinctions, which first rise as the extinction wave passes, and then ultimately fall as the pool of species shrinks. (b) Projected
changes in the numbers of species in each extinction risk category, with (c) showing magnified detail over the first 500 years during which many species become
more threatened. (Online version in colour.)

process currently taking place and not a prediction of what time to extinction of the world’s bird species by 1795
will happen in the future, because Q would not remain con- (44–4045) years per species, from 2985 to 4780 years (table 1),
stant over such long time periods. During the next 500 years, resulting in 40% (1.4–60%) reduction of the effective extinc-
this approach suggests that 471 (226–589) species would go tion rate. These estimates (electronic supplementary material,
extinct, about three times as many as we have lost over the table S2) are very conservative because they consider only
past 500 years. About 109 of these are projected to be species conservation efforts that resulted in improvements in status
currently classified as ‘least concern’. The graver problem is that were of sufficient magnitude to down-list species to
that most species become more threatened. This build-up of lower categories of risk, while conservation actions presumably
extinction risk [16] then causes a sharp increase in the more often allow species to remain in their current category
number of extinctions. Using the current Q, this would last or to transition to more threatened categories at a lower
for about 2000 years, after which the wave gradually fades rate [18].
as ever fewer species remain under this illustrative model.
To gain insight into the overall effect of global conserva-
tion efforts [3,17] from 1988 to 2016, we estimated Q
excluding category changes for species whose status, owing 4. Discussion
to conservation action, improved sufficiently to qualify for Red List assessments of extinction risk are based on a broad
down-listing to a lower Red List category, but including cat- literature on population demography [19] so that different
egory changes for six species down-listed owing to natural species may qualify under a particular IUCN Red List cat-
factors (electronic supplementary material, table S1). As egory for very different reasons and may have substantially
expected, conservation has had the largest impact on the different population sizes, range extents and threatening fac-
most threatened categories (table 1). For instance, the fate of tors. Owing to inaccurate estimates of e.g. population size,
species categorized as CR may seem less dire based on rate of decline or extent of occurrence or owing to time-lags
recent trajectories than their category implies, because they in information reaching Red List assessors, some assessments
are twice as likely to improve as to deteriorate (table 1). With- will be erroneous. It is challenging to model such error. We
out conservation, however, these species are twice as likely to assumed that species may be erroneously classified in a cat-
deteriorate as to improve (electronic supplementary material, egory adjacent to the true category, including the distinction
table S2). Conservation efforts resulting in the ‘down-listing’ between CR and EX, although Red List assessors are very
of a single species from CR to EN extend that species’ cautious about assigning taxa to EX. (Most media stories
expected time to extinction by 551 years, from 2503 to 3054 about ‘lazarus’ bird species relate to species still classified
years under our model (table 1). Efforts targeting CR species as critically endangered, rather than extinct.) Compared
increase the expected time to extinction of LC species as well, with assuming that assessments are error-free, our modelling
because these will become CR before going extinct. Thus, of error yields lower estimates of extinction rate (electronic
global conservation efforts have increased the projected supplementary material). Furthermore, deteriorations in
status are more likely to go undetected than improvements, targeted species that are already threatened. According to 4
because species benefiting from conservation action tend to our calculations (table 1), these efforts have considerably
be well monitored, such that our estimates of the rates at extended the projected time to extinction of all avian species,
which species are moving towards extinction may be conser- because all species become threatened before going extinct.
vative. Because some extinction risk categories contain few However, such efforts may not be the most cost-effective
species, estimates of their transition rates are also influenced strategy in the long term [27] because they do not prevent
by the choice of prior. Appropriate choice of priors and mod- least concern species from becoming more threatened, and
elling of assessment error will, therefore, likely be crucial to thus do not prevent increasing numbers of species in immedi-
better estimate current extinction rates. ate need of conservation. Therefore, it is important that any
Importantly, however, transitions of large numbers of post-2020 biodiversity framework negotiated through the
species through broad classes of relative extinction risk [19] Convention includes a renewed target to prevent extinctions,
provide useful information on the dynamics of extinction but, as emphasized by our analyses, also to prevent non-
given the diversity and pattern of human impacts on the threatened species or those at low risk from moving into
natural world. As shown in the electronic supplementary higher-risk categories, i.e. keeping common species

Biol. Lett. 15: 20190633

material, relaxing simplifying assumptions of the model or common and improving the status of currently threatened
using alternative priors on transition rates and classification species. This latter emphasis would help dampen an
error has little effect on estimates of the extinction rate com- ever-building wave of conservation need.
pared with estimates that treat all extant species as secure.
Comparatively few avian species are currently critically Data accessibility. The data and code used in this study are available as
endangered [8] and many of these appear to be benefiting electronic supplementary material.
from conservation efforts [5,12,20,21], so failure to account Authors’ contributions. M.J.M. and F.B. conceived the study. S.H.M.B.
for species becoming more threatened leads to considerable oversaw data collection and compiled the dataset. M.J.M. prepared
data for analysis. F.B. developed algorithms and analysed the data.
underestimation of effective extinction risk. We expect that
A.O.M. and F.B. interpreted results and designed model compari-
analyses for other taxa would show similar patterns. Most sons. M.J.M., F.B. and A.O.M. framed and wrote the manuscript
other taxa are less well studied than birds and contain a with input from S.H.M.B. All authors approved the final version
significant proportion of data deficient species, which com- and agree to be held accountable for the work it presents.
plicate risk assessment [22]. In such cases, information Competing interests. The authors declare no competing interests.
about e.g. ecology, geography [23], traits [24] and phylogeny Funding. Funding was provided by the Swedish Research Council (VR)
[25] could be combined to assign prior probabilities to the (M.J.M., grant no. 637-2013-274) and NSERC Canada (A.O.M.).
categorization of these species. Acknowledgements. We thank the many thousands of individuals and
The build-up of potential future extinction should inform organizations who contribute to BirdLife’s assessments of all the
world’s birds for the IUCN Red List. We thank Karen Magnusson-
assessment of progress towards international conservation Ford for research assistance and Giulio Della Riva and Simon
obligations such as the UN Convention on Biological Diver- Whelan for discussion, as well as three reviewers for many valuable
sity’s Aichi Targets [26]. Conservation efforts have mainly comments.

1. Ceballos G, Ehrlich PR, Barnosky AD, Garcia A, 7. Ceballos G, Ehrlich PR, Dirzo R. 2017 Biological Natl Acad. Sci. USA 103, 10 941–10 946. (doi:10.
Pringle RM, Palmer TM. 2015 Accelerated modern annihilation via the ongoing sixth mass extinction 1073/pnas.0604181103)
human-induced species losses: entering the sixth signaled by vertebrate population losses and 14. Stanley MS. 1998 Macroevolution: pattern and
mass extinction. Sci. Adv. 1, e1400253. (doi:10. declines. Proc. Natl Acad. Sci. USA 114, process. London, UK: Johns Hopkins Univ. Press.
1126/sciadv.1400253) E6089–E6096. (doi:10.1073/pnas.1704949114) 15. Butchart SHM, Lowe S, Martin RW, Symes A,
2. Barnosky AD et al. 2011 Has the Earth’s sixth mass 8. IUCN. 2017 The IUCN Red List of Threatened Westrip JRS, Wheatley H. 2018 Which bird species
extinction already arrived? Nature 471, 51–57. Species. Version 2017-1. have gone extinct? A novel quantitative
(doi:10.1038/nature09678) 9. Kemeny JG, Snell JL. 1976 Finite Markov chains, 2nd classification approach. Biol. Conserv. 227, 9–18.
3. De Vos JM, Joppa LN, Gittleman JL, Stephens PR, edn. New York, NY: Springer. (doi:10.1016/j.biocon.2018.08.014)
Pimm SL. 2015 Estimating the normal background 10. Butchart SHM et al. 2007 Improvements to the Red 16. Cardillo M, Mace GM, Gittleman JL, Purvis A.
rate of species extinction. Conserv. Biol. 29, List Index. PLoS ONE 2, e140. (doi:10.1371/journal. 2006 Latent extinction risk and the future
452–462. (doi:10.1111/cobi.12380) pone.0000140) battlegrounds of mammal conservation. Proc.
4. Ceballos G, Ehrlich PR. 2002 Mammal population 11. Butchart SHM et al. 2010 Global biodiversity: Natl Acad. Sci. USA 103, 4157–4161. (doi:10.
losses and the extinction crisis. Science 296, indicators of recent declines. Science 328, 1073/PNAS.0510541103)
904–907. (doi:10.1126/science.1069349) 1164–1168. (doi:10.1126/science.1187512) 17. Pimm SL, Jenkins CN, Abell R, Brooks TM, Gittleman
5. Butchart SHM, Stattersfield AJ, Brooks TM. 2006 12. Brooke MDL, Butchart SHM, Garnett ST, Crowley GM, JL, Joppa LN, Raven PH, Roberts CM, Sexton JO.
Going or gone: defining ‘Possibly Extinct’ species to Mantilla-Beniers NB, Stattersfield AJ. 2008 Rates of 2014 The biodiversity of species and their rates of
give a truer picture of recent extinctions. Bull. Br. movement of threatened bird species between IUCN extinction, distribution, and protection. Science 344,
Ornithol. Club 126A, 7–24. Red List categories and toward extinction. Conserv. 1246752. (doi:10.1126/science.1246752)
6. Akçakaya HR, Keith DA, Burgman M, Butchart SHM, Biol. 22, 417–427. (doi:10.1111/j.1523-1739.2008. 18. Hoffmann M, Duckworth JW, Holmes K, Mallon DP,
Hoffmann M, Regan HM, Harrison I, Boakes E. 2017 00905.x) Rodrigues ASL, Stuart SN. 2015 The difference
Inferring extinctions III: a cost–benefit framework 13. Pimm S, Raven P, Peterson A, Şekercioğlu ÇH, conservation makes to extinction risk of the world’s
for listing extinct species. Biol. Conserv. 214, Ehrlich PR. 2006 Human impacts on the rates of ungulates. Conserv. Biol. 29, 1303–1313. (doi:10.
336–342. (doi:10.1016/J.BIOCON.2017.07.027) recent, present, and future bird extinctions. Proc. 1111/cobi.12519)
19. Mace GM, Collar NJ, Gaston KJ, Hilton-Taylor C, 22. Burgman MA, Keith DA, Walshe TV. 1999 uncertainties in a time of urgency. Phil. 5
Akcakaya HR, Leader-Williams N, Milner-Gulland EJ, Uncertainty in comparative risk analysis for Trans. R. Soc. B 370, 20140002. (doi:10.1098/rstb.
Stuart SN. 2008 Quantification of extinction risk: threatened Australian plant species. Risk Anal. 19, 2014.0002)
IUCN’s system for classifying threatened species. 585–598. (doi:10.1111/j.1539-6924.1999.tb00430.x) 26. CBD. 2019 Draft summary for policymakers of
Conserv. Biol. 22, 1424–1442. (doi:10.1111/j.1523- 23. Bland LM, Collen B, Orme CDL, Bielby J. 2015 Predicting the 5th Edition of the Global Biodiversity
1739.2008.01044.x) the conservation status of data-deficient species. Outlook, Document CBD/SBSTTA/23/2/Add.3.
20. Hoffmann M et al. 2010 The impact of conservation Conserv. Biol. 29, 250–259. (doi:10.1111/cobi.12372) (
on the status of the world’s vertebrates. Science 24. Foden WB et al. 2013 Identifying the world’s most e02639e37191f353553e513d/sbstta-23-02-
330, 1503–1509. (doi:10.1126/science.1194442) climate change vulnerable species: a systematic trait- add3-en.pdf )
21. Butchart SHM, Stattersfield AJ, Collar NJ. 2006 based assessment of all birds, amphibians and corals. 27. McCarthy MA, Thompson CJ, Garnett ST. 2008
How many bird extinctions have we prevented? PLoS ONE 8, e65427. (doi:10.1371/journal.pone.0065427) Optimal investment in conservation of species.
Oryx 40, 266–278. (doi:10.1017/ 25. Forest F, Crandal KA, Chase MW, Faith DP. 2015 J. Appl. Ecol. 45, 1428–1435. (doi:10.1111/j.1365-
S0030605306000950) Phylogeny, extinction and conservation: embracing 2664.2008.01521.x)

Biol. Lett. 15: 20190633