Philippa Kaur

Institute of Carbon Neutrality, Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing, ChinaQian Zhao, Yao Zhang & Shilong PiaoSchool of Urban Planning and Design, Shenzhen Graduate School, Peking University, Shenzhen, ChinaZaichun Zhu & Hui ZengKey Laboratory of Earth Surface System and Human—Earth Relations, Ministry of Natural Resources of China, Shenzhen Graduate School, Peking University, Shenzhen, ChinaZaichun Zhu & Hui ZengDepartment of Earth and Environment, Boston University, Boston, MA, USARanga B. MyneniCSIC, Global Ecology Unit CREAF-CSIC-UAB, Barcelona, Catalonia, SpainJosep PeñuelasCREAF, Barcelona, Catalonia, SpainJosep PeñuelasState Key Laboratory of Tibetan Plateau Earth System, Resources and Environment (TPESRE), Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, ChinaShilong Piao More

Chase, J. M. et al. The interaction between predation and competition: A review and synthesis. Ecol. Lett. 5, 302–315. https://doi.org/10.1046/j.1461-0248.2002.00315.x (2002).Article 

Google Scholar 
Droge, E., Creel, S., Becker, M. S. & M’Soka, J. Risky times and risky places interact to affect prey behaviour. Nat. Ecol. Evol. 1, 1123–1128. https://doi.org/10.1038/s41559-017-0220-9 (2017).Article 

Google Scholar 
Allesina, S. & Levine Jonathan, M. A competitive network theory of species diversity. Proc. Natl. Acad. Sci. U.S.A. 108, 5638–5642. https://doi.org/10.1073/pnas.1014428108 (2011).Article 
ADS 

Google Scholar 
Bairey, E., Kelsic, E. D. & Kishony, R. High-order species interactions shape ecosystem diversity. Nat. Commun. 7, 12285. https://doi.org/10.1038/ncomms12285 (2016).Article 
ADS 
CAS 

Google Scholar 
Letten, A. D. & Stouffer, D. B. The mechanistic basis for higher-order interactions and non-additivity in competitive communities. Ecol. Lett. 22, 423–436. https://doi.org/10.1111/ele.13211 (2019).Article 

Google Scholar 
Lotka, A. J. Elements of physical biology. Sci. Prog. Twent. Century (1919–1933) 21, 341–343 (1926).
Google Scholar 
Volterra, V. Variazioni e Fluttuazioni del Numero d’Individui in Specie Animali Conviventi. (Società Anonima Tipografica “Leonardo da Vinci”, 1926).Schmitz, O. J. Top predator control of plant biodiversity and productivity in an old-field ecosystem. Ecol. Lett. 6, 156–163. https://doi.org/10.1046/j.1461-0248.2003.00412.x (2003).Article 

Google Scholar 
Fey, K., Banks, P. B., Oksanen, L. & Korpimäki, E. Does removal of an alien predator from small islands in the Baltic Sea induce a trophic cascade?. Ecography 32, 546–552. https://doi.org/10.1111/j.1600-0587.2008.05637.x (2009).Article 

Google Scholar 
Terborgh John, W. Toward a trophic theory of species diversity. Proc. Natl. Acad. Sci. U.S.A. 112, 11415–11422. https://doi.org/10.1073/pnas.1501070112 (2015).Article 
ADS 
CAS 

Google Scholar 
Pringle, R. M. et al. Predator-induced collapse of niche structure and species coexistence. Nature 570, 58–64. https://doi.org/10.1038/s41586-019-1264-6 (2019).Article 
ADS 
CAS 

Google Scholar 
Sandom, C. et al. Mammal predator and prey species richness are strongly linked at macroscales. Ecology 94, 1112–1122. https://doi.org/10.1890/12-1342.1 (2013).Article 

Google Scholar 
Louette, G. & De Meester, L. Predation and priority effects in experimental zooplankton communities. Oikos 116, 419–426. https://doi.org/10.1111/j.2006.0030-1299.15381.x (2007).Article 

Google Scholar 
Johnston, N. K., Pu, Z. & Jiang, L. Predator identity influences metacommunity assembly. J. Anim. Ecol. 85, 1161–1170. https://doi.org/10.1111/1365-2656.12551 (2016).Article 

Google Scholar 
Karakoc, C., Radchuk, V., Harms, H. & Chatzinotas, A. Interactions between predation and disturbances shape prey communities. Sci. Rep. 8, 2968. https://doi.org/10.1038/s41598-018-21219-x (2018).Article 
ADS 
CAS 

Google Scholar 
Hubbell, S. P. The Unified Neutral Theory of Biodiversity and Biogeography (MPB-32) (Princeton University Press, 2011).Book 

Google Scholar 
MacArthur, R. H. & Wilson, E. O. The Theory of Island Biogeography (Princeton University Press, 2001).Book 

Google Scholar 
Daniel, J., Gleason, J. E., Cottenie, K. & Rooney, R. C. Stochastic and deterministic processes drive wetland community assembly across a gradient of environmental filtering. Oikos 128, 1158–1169. https://doi.org/10.1111/oik.05987 (2019).Article 

Google Scholar 
Lehner, B. & Döll, P. Development and validation of a global database of lakes, reservoirs and wetlands. J. Hydrol. 296, 1–22. https://doi.org/10.1016/j.jhydrol.2004.03.028 (2004).Article 
ADS 

Google Scholar 
Chase, J. M., Biro, E. G., Ryberg, W. A. & Smith, K. G. Predators temper the relative importance of stochastic processes in the assembly of prey metacommunities. Ecol. Lett. 12, 1210–1218. https://doi.org/10.1111/j.1461-0248.2009.01362.x (2009).Article 

Google Scholar 
Werner, E. E. & Peacor, S. D. A review of trait-mediated indirect interactions in ecological communities. Ecology 84, 1083–1100. https://doi.org/10.1890/0012-9658(2003)084[1083:AROTII]2.0.CO;2 (2003).Article 

Google Scholar 
Pearson, D. E., Ortega, Y. K., Eren, Ö. & Hierro, J. L. Community assembly theory as a framework for biological invasions. Trends Ecol. Evol. 33, 313–325. https://doi.org/10.1016/j.tree.2018.03.002 (2018).Article 

Google Scholar 
Duchesne, É. et al. Variable strength of predator-mediated effects on species occurrence in an arctic terrestrial vertebrate community. Ecography 44, 1236–1248. https://doi.org/10.1111/ecog.05760 (2021).Article 

Google Scholar 
Ryberg, W. A., Smith, K. G. & Chase, J. M. Predators alter the scaling of diversity in prey metacommunities. Oikos 121, 1995–2000. https://doi.org/10.1111/j.1600-0706.2012.19620.x (2012).Article 

Google Scholar 
Carrete Vega, G. & Wiens, J. J. Why are there so few fish in the sea?. Proc. R. Soc. B 279, 2323–2329. https://doi.org/10.1098/rspb.2012.0075 (2012).Article 

Google Scholar 
Barrett, M. et al. Living planet report 2018: Aiming higher. (2018).Reid, A. J. et al. Emerging threats and persistent conservation challenges for freshwater biodiversity. Biol. Rev. 94, 849–873. https://doi.org/10.1111/brv.12480 (2019).Article 

Google Scholar 
Di Marco, M. et al. Changing trends and persisting biases in three decades of conservation science. Glob. Ecol. Conserv. 10, 32–42. https://doi.org/10.1016/j.gecco.2017.01.008 (2017).Article 

Google Scholar 
Hammerschlag, N. et al. Ecosystem function and services of aquatic predators in the anthropocene. Trends Ecol. Evol. 34, 369–383. https://doi.org/10.1016/j.tree.2019.01.005 (2019).Article 

Google Scholar 
Wang, T. et al. Amur tigers and leopards returning to China: direct evidence and a landscape conservation plan. Landsc Ecol 31, 491–503. https://doi.org/10.1007/s10980-015-0278-1 (2016).Article 

Google Scholar 
Hong, S. et al. Stream health, topography, and land use influences on the distribution of the Eurasian otter Lutra lutra in the Nakdong River basin, South Korea. Ecol. Indic. 88, 241–249. https://doi.org/10.1016/j.ecolind.2018.01.004 (2018).Article 

Google Scholar 
Guter, A., Dolev, A., Saltz, D. & Kronfeld-Schor, N. Using videotaping to validate the use of spraints as an index of Eurasian otter (Lutra lutra) activity. Ecol. Indic. 8, 462–465. https://doi.org/10.1016/j.ecolind.2007.04.009 (2008).Article 

Google Scholar 
Sittenthaler, M., Bayerl, H., Unfer, G., Kuehn, R. & Parz-Gollner, R. Impact of fish stocking on Eurasian otter (Lutra lutra) densities: A case study on two salmonid streams. Mamm. Biol. 80, 106–113. https://doi.org/10.1016/j.mambio.2015.01.004 (2015).Article 

Google Scholar 
Zheng, B., Huang, H., Zhang, Y. & Dai, D. The Fishes of Tumen River (Jilin People’s Publishing House, 1980).
Google Scholar 
Fleishman, E., Murphy, D. D. & Brussard, P. F. A new method for selection of umbrella species for conservation planning. Ecol Appl 10, 569–579. https://doi.org/10.1890/1051-0761(2000)010[0569:ANMFSO]2.0.CO;2 (2000).Article 

Google Scholar 
Roberge, J.-M. & Angelstam, P. E. R. Usefulness of the umbrella species concept as a conservation tool. Conserv. Biol. 18, 76–85. https://doi.org/10.1111/j.1523-1739.2004.00450.x (2004).Article 

Google Scholar 
McGowan, J. et al. Conservation prioritization can resolve the flagship species conundrum. Nat. Commun. 11, 994. https://doi.org/10.1038/s41467-020-14554-z (2020).Article 
ADS 
CAS 

Google Scholar 
Katano, I., Doi, H., Eriksson, B. K. & Hillebrand, H. A cross-system meta-analysis reveals coupled predation effects on prey biomass and diversity. Oikos 124, 1427–1435. https://doi.org/10.1111/oik.02430 (2015).Article 

Google Scholar 
Leibold, M. A. A graphical model of keystone predators in food webs: Trophic regulation of abundance, incidence, and diversity patterns in communities. Am. Nat. 147, 784–812. https://doi.org/10.1086/285879 (1996).Article 

Google Scholar 
McPeek, M. A. The consequences of changing the top predator in a food web: A comparative experimental approach. Ecol. Monogr. 68, 1–23. https://doi.org/10.1890/0012-9615(1998)068[0001:TCOCTT]2.0.CO;2 (1998).Article 

Google Scholar 
Chase, J. M. & Leibold, M. A. Ecological Niches: Linking Classical and Contemporary Approaches (University of Chicago Press, 2003).Book 

Google Scholar 
Gravel, D., Canham, C. D., Beaudet, M. & Messier, C. Reconciling niche and neutrality: The continuum hypothesis. Ecol. Lett. 9, 399–409. https://doi.org/10.1111/j.1461-0248.2006.00884.x (2006).Article 

Google Scholar 
Yoshida, T., Jones, L. E., Ellner, S. P., Fussmann, G. F. & Hairston, N. G. Rapid evolution drives ecological dynamics in a predator–prey system. Nature 424, 303–306. https://doi.org/10.1038/nature01767 (2003).Article 
ADS 
CAS 

Google Scholar 
Yin, X., Wang, J., Yin, H. & Ruan, Y. Does inducible defense mitigate physiological stress responses of prey to predation risk?. Hydrobiologia 843, 173–181. https://doi.org/10.1007/s10750-019-04046-7 (2019).Article 

Google Scholar 
Chalcraft, D. R. & Resetarits, W. J. Jr. Predator identity and ecological impacts: Functional redundancy or functional diversity?. Ecology 84, 2407–2418. https://doi.org/10.1890/02-0550 (2003).Article 

Google Scholar 
Petchey, O. L. & Gaston, K. J. Functional diversity: Back to basics and looking forward. Ecol. Lett. 9, 741–758. https://doi.org/10.1111/j.1461-0248.2006.00924.x (2006).Article 

Google Scholar 
Burner, R. C. et al. Functional structure of European forest beetle communities is enhanced by rare species. Biol. Conserv. 267, 109491. https://doi.org/10.1016/j.biocon.2022.109491 (2022).Article 

Google Scholar  More

Ethics statementAnimals’ care was in accordance with institutional guidelines. Ethical permit was issued by Malmö-Lund djurförsöksetiska nämnd 5.8.18-00848/2018.Field workWe carried out the field work in Sweden during four breeding seasons (2018–2021). Adult male willow warblers were captured in their breeding territories using mist nets and playback of a song. From each bird, we collected the innermost primary feather from the right wing. From the birds that returned with a logger we also collected ~20 μl of blood from the brachial wing vein. The blood was stored in SET buffer (0.015 M NaCl, 0.05 M Tris, 0.001 M of EDTA, pH 8.0) at room temperature until deposited for permanent storage at −20 °C. We deployed Migrate Technology Ltd geolocators (Intigeo-W30Z11-DIP 12 × 5 × 4 mm, 0.32 g) and used a nylon string to mount them on birds with the “leg-loop” harness method as outlined in our previous work24. The mass of the logger relative to that of the bird was on average 3.3% (range 2.7–3.8%).The tagged birds were ringed with a numbered aluminum ring, and two, colored plastic rings for later identification in the field. In total, we tagged 466 males (349 in 2018 and 117 in 2020) at breeding territories. During the first tagging season (2018), birds were trapped at 17 locations (average 22 birds per site; range 7–30) distributed across Sweden (Fig. S1). Three of the sites were in southern Sweden to document migration routes of allopatric trochilus and three sites were located above the Arctic circle to record migratory routes of allopatric acredula, whereas the remaining (239) loggers were spread over 11 sites located in the migratory divide. Given the observed densities and distribution of hybrids after analyzing returning birds in 2019, we deployed 117 more loggers at one single site (63.439°N, 14.831°E) in 2020. We successfully retrieved tracks from 57 birds tagged in 2019 and 16 from birds tagged in 2021. In search for birds with loggers, we checked circa 3000 willow warbler males and covered an area of at least 0.5 km radius around each site the year after tagging.Geolocator data treatmentThe R package GeoLight (version 2.0)25 was used to extract and analyze locations from raw geolocator data. All twilight events were obtained with light threshold of 3 lux. The most extreme outliers were trimmed with “loessFilter” function and a K value of 3. We used GeoLight’s function “getElevation” for estimating the sun elevation angle for the breeding period: these sets of locations were used to infer the positions for autumn departure direction. In addition, we carried out a “Hill-Ekström” calibration for the longest stationary winter site during the period before the spring equinox. Winter calibration produced location sets that better reflected the winter coordinates of the main winter site in sub-Saharan Africa26. We reduced some of the inherent geolocation “noise” by applying cantered 5-day rolling means to the coordinates. The equinox periods were visually identified by inspecting standard deviations in latitude. Latitudes from equinox periods were omitted (on average autumn equinox obscured data for 45 days (range 25–68). For the main winter site, we used the longest period at which bird stayed stationary and from which in all cases begun the spring migration (mean = 118, SD = 23 days). Timing of autumn departure was estimated by manual inspection of longitudes and latitudes plotted in time series. To estimate at which longitude the birds crossed the Mediterranean, we extracted the longitude when birds crossed latitude 35 N° (Mediterranean crossing longitude). For 29 birds, it was possible to directly extract the longitude at crossing latitude 35 N°. For the rest of the cases, the birds had not reached latitude 35 N° before the latitude was obscured by the equinox, we calculated the mean longitude of 10 days from the onset of fall equinox as a measure of the Mediterranean crossing. This measurement correlated highly with the winter longitude (r = 0.78, p = 2.8 × 10−16). To control for the birds relative breeding site longitude, we extracted the departure direction (1°–360°) relative from the tagging site to the location where the birds crossed the Mediterranean (departure direction). The departure data was of circular type (measured in 360°), however the variance did not span more than 180° degrees (range 151°–224°). Therefore, we proceeded with analyses using linear statistics. Geographic distances and departure direction were calculated using R package “geosphere” (version 1.5-10). Complete set of positions of each individual bird with equinoxes excluded is presented in Supplementary Data 1.Laboratory work and molecular data extractionWe extracted DNA from blood samples following the ammonium acetate protocol16. Genotyping for divergent regions on chromosome 1 (InvP-Ch1) and chromosome 5 (InvP-Ch5) was done using a qPCR SNP assay16, which is based on one informative SNP per region (SNP 65 for chromosome 1 and SNP 285 for chromosome 5). Probes and primers were produced by Thermo Fisher Scientific and were designed using the online Custom TaqMan® Assay Design tool (Table S4). We used Bio-Rad CFX96™ Real-time PCR system (Bio-Rad Laboratories, CA, USA) and the universal Fast-two-steps protocol: 95 °C, 15 min—40*(95 °C, 10 s–60 °C, 30 s, plate read. Both regions contain inversion polymorphisms that restrict recombination between subspecies-specific haplotypes and contain nearly all the SNPs separating the two subspecies13. For each region, we scored genotypes as either “Tro” (homozygous for trochilus haplotypes), “Acr” (homozygous for acredula haplotypes) or “Het” (heterozygous). The method that we used to assess the presence of MARB-a is based on a qPCR assay that quantifies the copy number of a novel TE (previously known as AFLP-WW212) that has expanded in acredula. The quantification of repeats by this method has been shown to be highly repeatable (R2 = 0.88) when comparing estimates obtained from DNA in blood and feathers15. We used the forward (5′-CCTTGCATACTTCTATTTCTCCC-3′) and reverse (5′-CATAGGACAGACATTGTTGAGG-3′) primers developed by Caballero-López et al.15 to amplify the TE motif. For reference of a single copy region we used the primers SFRS3F and SFRS3R27. We diluted DNA to 1 ng/μl−1 and used a Bio-Rad CFX96™ Real-time PCR system (Bio-Rad Laboratories, CA, USA) with SYBR-green-based detection. Total reaction volume was 25 μl of which 4 μl of DNA, 12.5 μl of SuperMix, 0.1 μl ROX, 1 μl of primer (forward and reverse), and 6.4 μl of double distilled H2O. We ran quantifications of the single copy gene and the TE variant found on MARB-a on separate plates with the following settings: 50 °C for 2 min as initial incubation, 95 °C for 2 min X 43 (94 °C for 30 s [55.3 °C SFRS3 and 55.5 °C for TE, 30 s] and 72 °C for 45 s). Each sample was run in duplicate and together with a two-fold serial standard dilution (2.5–7.8 × 10−2 ng). Allopatric trochilus have 0–6 copies whereas allopatric acredula have 8–45 copies15; a bimodal distribution was also confirmed in this new data set (Fig. S2). Accordingly, for the present analyses, we split the data in two groups: birds with ≤6 TE copies and birds with >7, translating into absence or presence of MARB-a, with the former assumed to be homozygous for the absence of MARB-a and the latter heterozygous or homozygous for the presence of MARB-a. Data from two investigated willow warbler families suggest a Mendelian inheritance pattern and provide support for our interpretation of how TE copy numbers reflect the three genotypes (Table S5). Moreover, the TE copy numbers within the hybrid swarm have a distribution similar to a combination of allopatric trochilus and acredula, further supporting that the copies are inherited as intact blocks (haplotypes). However, a precise distinction between heterozygotes and homozygotes on MARB-a is still not possible15.Statistical analysisWe used linear models with departure direction, winter longitude, migration distance and departure timing as response variables and the three genetic markers: MARB-a (a factor with two levels), InvP-Ch1 (a factor with three levels) and InvP-Ch5 (a factor with three levels) as explanatory variables. Models were constructed with R base package “stats”. We reported Type II ANOVA for models with more than one explanatory variable and no interactions and type III ANOVA results for models with interaction term by using R package “Car” (version 3.0-12)28. We initially constructed mixed effect models with timing of departure and tagging year as random factors however, this delivered singular fits due to insufficient sample sizes across categories. Normality of residuals was checked with a Shapiro–Wilk test. For carrying out circular statistics on autumn migration direction we used the R package “circular” (version 0.4-93). Watson’s U2 pairwise comparisons of different groups delivered the same results as linear models (Table S2 and Fig. S5). Circular means were identical to conventional linear means in our data set, which we take as another evidence that linear models are appropriate for the analysis of our data (Table S3 and Fig. S5). Maps in Figs. 1 and 2b and S1, S3 and S4 were created with R package “ggplot2” (version 3.3.6) using continent contours from Natural Earth, naturalearthdata.com/. Heat gradient over the maps in Fig. 1a–d were created with R package “gstat” (version 2.0-8) and the inverse distance weighting power of 3.0. Circular plots were created with ORIANA (version 4.02). All analyses were carried out with R version 4.1.1 (R Core Team 2021).Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article. More

Newmark, W. D. A land-bridge island perspective on mammalian extinctions in western North American parks. Nature 325, 430–432 (1987).Article 
ADS 
CAS 

Google Scholar 
Newmark, W. D. Isolation of African protected areas. Front. Ecol. Environ. 6, 321–328 (2008).Article 

Google Scholar 
Radeloff, V. C. et al. Housing growth in and near United States protected areas limits their conservation value. Proc. Natl. Acad. Sci. U. S. A. 107, 940–945 (2010).Article 
ADS 
CAS 

Google Scholar 
Jones, K. R. et al. One-third of global protected land is under intense human pressure. Science 360, 788–791 (2018).Article 
CAS 

Google Scholar 
Elsen, P. R., Monahan, W. B., Dougherty, E. R. & Merenlender, A. M. Keeping pace with climate change in global terrestrial protected areas. Sci. Adv. https://doi.org/10.1126/sciadv.aay0814 (2020).Article 

Google Scholar 
Wasser, S. K. et al. Genetic assignment of large seizures of elephant ivory reveals Africa’s major poaching hotspots. Science 349, 84–87 (2015).Article 
ADS 
CAS 

Google Scholar 
Davis, C. R. & Hansen, A. J. Trajectories in land use change around U,S. national parks and challenges and opportunities for management. Ecol. Appl. 21, 3299–3316 (2011).Article 

Google Scholar 
Newmark, W. D. Extinction of mammal populations in western North American national parks. Conserv. Biol. 9, 512–526 (1995).Article 

Google Scholar 
Newmark, W. D. Insularization of Tanzanian parks and the local extinction of large mammals. Conserv. Biol. 10, 1549–1556 (1996).Article 

Google Scholar 
Brashares, J. S., Arcese, P. & Sam, M. K. Human demography and reserve size predict wildlife extinction in West Africa. Proc. R. Soc. B Biol. Sci. 268, 2473–2478 (2001).Article 
CAS 

Google Scholar 
Woodroffe, R. & Ginsberg, J. R. Edge effects and the extinction of populations inside protected areas. Science 280, 2126–2128 (1998).Article 
ADS 
CAS 

Google Scholar 
Turner, M. G. & Dale, V. H. Comparing large, infrequent disturbances: What have we learned?. Ecosystems 1, 493–496 (1998).Article 

Google Scholar 
Berger, J. The last mile: How to sustain long-distance migration in mammals. Conserv. Biol. 18, 320–331 (2004).Article 

Google Scholar 
Bolger, D. T., Newmark, W. D., Morrison, T. A. & Doak, D. F. The need for integrative approaches to understand and conserve migratory ungulates. Ecol. Lett. 11, 63–77 (2008).
Google Scholar 
Sawyer, H., Kauffman, M. J., Nielson, R. M. & Horne, J. S. Identifying and prioritizing ungulate migration routes for landscape-level conservation. Ecol. Appl. 19, 2016–2025 (2009).Article 

Google Scholar 
Tucker, M. A. et al. Moving in the anthropocene: Global reductions in terrestrial mammalian movements. Science 469, 466–469 (2018).Article 
ADS 

Google Scholar 
Soulé, M. E. & Terborgh, J. Conserving nature at regional and continental scales-a scientific program for North America. Bioscience 49, 809–817 (1999).Article 

Google Scholar 
Hilty, J. et al. Guidelines for conserving connectivity through ecological networks and corridors. Best Pract. Prot. Area Guidel. Ser. 30, 122 (2020).
Google Scholar 
Haddad, N. & Tewksbury, J. Impacts of corridors on populations and communities. in Connectivity Conservation (eds. Crooks, K. R. & Sanjayan, M.) 390–415 (Cambridge University Press, 2010).
Google Scholar 
Ramiadantsoa, T., Ovaskainen, O., Rybicki, J. & Hanski, I. Large-scale habitat corridors for biodiversity conservation: A forest corridor in Madagascar. PLoS One 10, 1–18 (2015).Article 
CAS 

Google Scholar 
Newmark, W. D., Jenkins, C. N., Pimm, S. L., McNeally, P. B. & Halley, J. M. Targeted habitat restoration can reduce extinction rates in fragmented forests. Proc. Natl. Acad. Sci. USA. 114, 9635–9640 (2017).Article 
ADS 
CAS 

Google Scholar 
Diamond, J. M. Biogeographic kinetics: Estimation of relaxation times for avifaunas of southwest Pacific islands. Proc. Natl. Acad. Sci. 69, 3199–3203 (1972).Article 
ADS 
CAS 

Google Scholar 
Terborgh, J. Preservation of natural diversity: The problem of extinction prone species. Bioscience 24, 715–722 (1974).Article 

Google Scholar 
Tilman, D., May, R. M., Lehman, C. L. & Nowak, M. A. Habitat destruction and the extinction debt revisited. Nature 371, 65–66 (1994).Article 
ADS 

Google Scholar 
Halley, J. M., Monokrousos, N., Mazaris, A. D., Newmark, W. D. & Vokou, D. Dynamics of extinction debt across five taxonomic groups. Nat. Commun. 7, 1–6 (2016).Article 

Google Scholar 
Wearn, O. R., Reuman, D. C. & Ewers, R. M. Extinction debt and windows of conservation opportunity in the Brazilian amazon. Science 337, 228–232 (2012).Article 
ADS 
CAS 

Google Scholar 
Hanski, I. Extinction debt and species credit in boreal forests: Modelling the consequences of different approaches to conservation. Ann. Zool. Fennici 37, 271–280 (2000).
Google Scholar 
LaBarbera, M. Analyzing body size as a factor in ecology and evolution. Annu. Rev. Ecol. Syst. 20, 97–117 (1989).Article 

Google Scholar 
Oakleaf, J. K. et al. Habitat selection by recolonizing wolves in the northern Rocky mountains of the United States. J. Wildl. Manage. 70, 554–563 (2006).Article 

Google Scholar 
Cushman, S. A., McKelvey, K. S. & Schwartz, M. K. Use of empirically derived source-destination models to map regional conservation corridors. Conserv. Biol. 23, 368–376 (2009).Article 

Google Scholar 
Schwartz, M. K. et al. Wolverine gene flow across a narrow climatic niche. Ecology 90, 3222–3232 (2014).Article 

Google Scholar 
McKelvey, K. S. et al. Climate change predicted to shift wolverine distributions, connectivity, and dispersal corridors. Ecol. Appl. 21, 2882–2897 (2011).Article 

Google Scholar 
Carroll, C., Mcrae, B. H. & Brookes, A. Use of linkage mapping and centrality analysis across habitat gradients to conserve connectivity of gray wolf populations in western North America. Conserv. Biol. 26, 78–87 (2012).Article 

Google Scholar 
Parks, S. A., McKelvey, K. S. & Schwartz, M. K. Effects of weighting schemes on the identification of wildlife corridors generated with least-cost methods. Conserv. Biol. 27, 145–154 (2013).Article 

Google Scholar 
Peck, C. P. et al. Potential paths for male-mediated gene flow to and from an isolated grizzly bear population. Ecosphere 8, e01969 (2017).Article 

Google Scholar 
Wild Migrations: Atlas of Wyoming’s Ungulates. (Oregon State University, 2018).Singleton, P. H., Gaines, W. L. & Lehmkuhl, J. F. Landscape permeability for large carnivores in Washington: A geographic information system weighted-distance and least-cost corridor assessment. (2002).Long, R. A. et al. The Cascades carnivore connectivity project: A landscape genetic assessment of connectivity in Washington’s north Cascades ecosystem. Final report for the Seattle City Light Wildlife Research Program (2013).Diamond, J. M. The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. Biol. Conserv. 7, 129–146 (1975).Article 

Google Scholar 
Wilson, E. O. & Willis, E. O. Applied biogeography. In Ecological structure of ecological communities (eds. Cody, M. L, & Diamond, J. M.) 522–534 (Harvard University Press, 1975)
Google Scholar 
Halley, J. M. & Iwasa, Y. Neutral theory as a predictor of avifaunal extinctions after habitat loss. Proc. Natl. Acad. Sci. USA 108, 2316–2321 (2011).Article 
ADS 
CAS 

Google Scholar 
Cushman, S. A., Lewis, J. S. & Landguth, E. L. Evaluating the intersection of a regional wildlife connectivity network with highways. Mov. Ecol. 1, 1–11 (2013).Article 

Google Scholar 
Singleton, P. H. & Lehmkuhl, J. F. I-90 Snoqualmie pass wildlife habitat linkage assessment. Final Report. USDA, Pacific Northwest Research Station. (2000).Craighead, L., Craighead, A., Oeschslia, L. & Kociolek, A. Bozeman pass post-fencing wildlife monitoring. Final Report. FHWA/MT-10-006/8173 (2011).Andis, A. Z., Huijser, M. P. & Broberg, L. Performance of arch-style road crossing structures from relative movement rates of large mammals. Front. Ecol. Evol. 5, 1–13 (2017).Article 

Google Scholar 
Millward, L. Small mammal microhabitat use and species composition at a wildlife crossing structure compared with nearby forest (Central Washington University, 2018).
Google Scholar 
Bischof, R., Steyaert, S. M. J. G. & Kindberg, J. Caught in the mesh: Roads and their network-scale impediment to animal movement. Ecography 40, 1369–1380 (2017).Article 

Google Scholar 
Balkenhol, N. & Waits, L. P. Molecular road ecology: Exploring the potential of genetics for investigating transportation impacts on wildlife. Mol. Ecol. 18, 4151–4164 (2009).Article 

Google Scholar 
Clevenger, A. P. & Wierzchowski, J. Maintaining and restoring connectivity in landscapes fragmented by roads. In Connectivity Conservation, (eds. Crooks, K. R. & Sanjayan, M.) 502–535 (Cambridge University Press, 2010.)
Google Scholar 
Sawaya, M. A., Kalinowski, S. T. & Clevenger, A. P. Genetic connectivity for two bear species at wildlife crossing structures in Banff National Park. Proc. R. Soc. B Biol. Sci. 281, 20131705 (2014).Article 

Google Scholar 
Sawaya, M. A., Clevenger, A. P. & Schwartz, M. K. Demographic fragmentation of a protected wolverine population bisected by a major transportation corridor. Biol. Conserv. 236, 616–625 (2019).Article 

Google Scholar 
Kamal, S., Grodzińska-Jurczak, M. & Brown, G. Conservation on private land: A review of global strategies with a proposed classification system. J. Environ. Plan. Manag. 58, 576–597 (2015).Article 

Google Scholar 
Wasserman, T. N., Cushman, S. A., Littell, J. S., Shirk, A. J. & Landguth, E. L. Population connectivity and genetic diversity of American marten (Martes americana) in the United States northern Rocky Mountains in a climate change context. Conserv. Genet. 14, 529–541 (2013).Article 

Google Scholar 
Wasserman, T. N., Cushman, S. A., Shirk, A. S., Landguth, E. L. & Littell, J. S. Simulating the effects of climate change on population connectivity of American marten (Martes americana) in the northern Rocky Mountains, USA. Landsc. Ecol. 27, 211–225 (2012).Article 

Google Scholar 
Cushman, S. A., Landguth, E. L. & Flather, C. H. Evaluating the sufficiency of protected lands for maintaining wildlife population connectivity in the U.S. northern Rocky Mountains. Divers. Distrib. 18, 873–884 (2012).Article 

Google Scholar 
Beier, P., Spencer, W., Baldwin, R. F. & Mcrae, B. H. Toward best practices for developing regional connectivity maps. Conserv. Biol. 25, 879–892 (2011).Article 

Google Scholar 
R Core Team. R: A language and environment for statistical computing. (2020). More

Identification of broad-host-range rhizoplane colonization genes by Tn-seqThis work was focused on SF2 harboring a typical multipartite genome of Sinorhizobium (chromosome, chromid, and symbiosis plasmid) [59]. To perform genome-wide survey of rhizoplane colonization genes of SF2 (Fig. 1), the input mutant library was inoculated on filter paper of plant culture dish, and output mutant libraries were collected from filter papers at 1 h post inoculation (F1h) and 7 days post inoculation (dpi; F7d), and from rhizoplane of cultivated soybean (CS7d), wild soybean (WS7d), rice (R7d), and maize (Z7d) at 7 dpi. To facilitate Tn-seq library construction, all output mutant libraries were subject to 32 h cultivation in the TY rich medium, with input libraries cultivated at the same condition as control (TY). Tn-seq revealed that transposon insertion density in three input and 21 output samples ranged from 57.03 to 86.99% (Table S3), which are above the threshold of 50% insertion density for a good Tn-seq dataset [49]. A reproducible rhizosphere effect was observed in three independent experiments (Fig. S1), i.e., rhizoplane samples (CS7d, WS7d, R7d, and Z7d) consistently formed distinct clusters compared to those of TY, F1h, and F7d. A considerable signature of three independent input libraries was also identified (Data S1, Data S2, and Fig. S1). These results highlight that stochastic variations among multiple independent input libraries should be considered before making conclusions on gene fitness, which has been largely overlooked in earlier studies based on just one input library [49].Based on gene fitness scores of rhizoplane samples (CS7d, WS7d, R7d and Z7d) compared to corresponding F1h datasets (Fig. S2A; Data S2), 93, 91, 127, and 206 genes were identified as rhizoplane colonization genes for test plants of cultivated soybean, wild soybean, maize, and rice, respectively, accounting for 1.4–3.1% of the SF2 genome (p values  More

Durden, L. Lice (Phthiraptera). In Medical and Veterinary Entomology 3rd edn (eds Mullen, G. & Durden, L.) 79–106 (Academic Press, 2019).Chapter 

Google Scholar 
Stork, N. E. & Lyal, C. H. C. Extinction or ‘co-extinction’ rates?. Nature 366, 307. https://doi.org/10.1038/366307a0 (1993).Article 
ADS 

Google Scholar 
Koh, L. P. et al. Species coextinctions and the biodiversity crisis. Science 305, 1632–1634. https://doi.org/10.1126/science.1101101 (2004).Article 
ADS 
CAS 

Google Scholar 
Gerlach J (2014) Haematopinus oliveri. The IUCN Red List of Threatened Species 2014: e.T9621A21423551. https://doi.org/10.2305/IUCN.UK.2014-1.RLTS.T9621A21423551.en.Mingozzi, T. & Stève, R. Analysis of a historical extirpation of the bearded vulture Gypaetus barbatus (L.) in the Western Alps (France-Italy): former distribution and causes of extirpation. Biol. Conserv. 79, 155–171. https://doi.org/10.1016/S0006-3207(96)00110-3 (1997).Article 

Google Scholar 
Schaub, M., Zink, R., Beissmann, H., Sarrazin, F. & Arlettaz, R. When to end releases in reintroduction programmes: demographic rates and population viability analysis of bearded vultures in the Alps. J. Appl. Ecol. 46, 92–100. https://doi.org/10.1111/j.1365-2664.2008.01585.x (2009).Article 

Google Scholar 
BirdLife International. Gypaetus barbatus (amended version of 2017 assessment). The IUCN Red List of Threatened Species 2017: e.T22695174A118590506. https://doi.org/10.2305/IUCN.UK.2017-3.RLTS.T22695174A118590506.en, accessed 07 Apr 2021 (2017).Price, R. D., Hellenthal, R. A., Palma, R. L., Johnson, K. P., & Clayton, D. H. The chewing lice: World checklist and biological overview. Illinois Natural History Survey Special Publication 24. Illinois (2003).Clay, T. Revisions of mallophaga genera. Degeeriella from the Falconiformes. Bull. Br. Mus. (Nat. Hist.) 7, 123–207 (1958).
Google Scholar 
Martín Mateo, M. P. Fauna Ibérica, Vol. 32. Phthiraptera, Ischnocera. Museo Nacional de Ciencias Naturales (CSIC), Madrid (2009).Hoberg, E. P., Brooks, D. R. & Siegel-Causey, D. Host-parasite co-speciation: history, principles, and prospects. In Host-Parasite Evolution: General Principles and Avian Models (eds Clayton, D. H. & Moore, J.) 212–235 (Oxford University Press, 1997).
Google Scholar 
Johnson, K. P., Weckstein, J. D., Witt, C. C., Faucett, R. C. & Moyle, R. G. The perils of using host relationships in parasite taxonomy: phylogeny of the Degeeriella complex. Mol. Phylogenet. Evol. 23, 150–157. https://doi.org/10.1016/S1055-7903(02)00014-3 (2002).Article 
CAS 

Google Scholar 
Catanach, T. A. & Johnson, K. P. Independent origins of the feather lice (Insecta: Degeeriella) of raptors. Biol. J. Linn. Soc. 114, 837–847. https://doi.org/10.1111/bij.12453 (2015).Article 

Google Scholar 
Pérez, J. M., Ruiz-Martínez, I. & Cooper, J. E. Occurrence of chewing lice on Spanish raptors. Ardeola 43, 129–138 (1996).
Google Scholar 
Ash, J. S. A study of the mallophagan of birds with particular reference to their ecology. Ibis 102, 93–110. https://doi.org/10.1111/j.1474-919X.1960.tb05095.x (1960).Article 

Google Scholar 
Askew, R. R. Parasitic Insects (Heinemann Educational, 1971).
Google Scholar 
Marshall, A. G. The Ecology of Parasitic Insects (Academic Press, 1981).
Google Scholar 
Bartlow, A. W., Villa, S. M., Thompson, M. W. & Bush, S. E. Walk or ride? Phoretic behaviour of amblyceran and ischnoceran lice. Int. J. Parasitol. 46, 221–227. https://doi.org/10.1016/j.ijpara.2016.01.003 (2016).Article 

Google Scholar 
Leonardi, M. S., Crespo, E. A., Raga, J. A. & Fernández, M. Scanning electron microscopy of Antarctophthirus microchir (Phthiraptera: Anoplura: Echinophthiriidae): Studying morphological adaptations to aquatic life. Micron 43, 929–936. https://doi.org/10.1016/j.micron.2012.03.009 (2012).Article 

Google Scholar 
Ortega Insaurralde, I., Minoli, S., Toloza, A. C., Picollo, M. I. & Barrozo, R. B. The sensory machinery of the head louse Pediculus humanus capitis: from the antennae to the brain. Front. Physiol. 10, 434. https://doi.org/10.3389/fphys.2019.00434 (2019).Article 

Google Scholar 
Ortega Insaurralde, I., Picollo, M. I. & Barrozo, R. B. Sensory features of the human louse antenna: New contributions and comparisons between ecotypes. Med. Vet. Entomol. 35, 219–224. https://doi.org/10.1111/mve.12485 (2021).Article 
CAS 

Google Scholar 
Page, R. D. M., Lee, P. L. M., Becher, S. A., Griffiths, R. & Clayton, D. H. A different tempo of mitochondrial DNA evolution in birds and their parasitic lice. Mol. Phylogenet. Evol. 9, 276–293. https://doi.org/10.1006/mpev.1997.0458 (1998).Article 
CAS 

Google Scholar 
Cruickshank, R. H. et al. Phylogenetic analysis of partial sequences of elongation factor 1α identifies major groups of lice (Insecta: Phthiraptera). Mol. Phylogenet. Evol. 19, 202–215. https://doi.org/10.1006/mpev.2001.0928 (2001).Article 
CAS 

Google Scholar 
Murrell, A. & Barker, S. C. Multiple origins of parasitism in lice: phylogenetic analysis of SSU rDNA indicates that the Phthiraptera and Psocoptera are not monophyletic. Parasitol. Res. 97, 274–280. https://doi.org/10.1007/s00436-005-1413-8 (2005).Article 

Google Scholar 
Whiteman, N. K., Kimball, R. T. & Parker, P. G. Co-phylogeography and comparative population genetics of the threatened Galápagos hawk and three ectoparasite species: ecology shapes population histories within parasite communities. Mol. Ecol. 16, 4759–4773. https://doi.org/10.1111/j.1365-294X.2007.03512.x (2007).Article 
CAS 

Google Scholar 
Palma, R. L. Slide-mounting of Lice: a detailed description of the Canada Balsam technique. N. Z. Entomol. 6, 432–436. https://doi.org/10.1080/00779962.1978.9722313 (1978).Article 

Google Scholar 
Soler-Cruz, M. D. & Martín-Mateo, M. P. Scanning electron microscopy of legs of two species of sucking lice (Anoplura: Phthiraptera). Micron 40, 401–408. https://doi.org/10.1016/j.micron.2008.10.001 (2009).Article 
CAS 

Google Scholar 
Hafner, M. S. et al. Disparate rates of molecular evolution in cospeciating hosts and parasites. Science 265, 1087–1090. https://doi.org/10.1126/science.8066445 (1994).Article 
ADS 
CAS 

Google Scholar 
Simon, C. et al. Evolution, weighting and phylogenetic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain reaction primers. Ann. Entomol. Soc. Am. 87, 651–701. https://doi.org/10.1093/aesa/87.6.651 (1994).Article 
CAS 

Google Scholar 
Danforth, B. N. & Ji, S. Elongation factor-1α occurs as two copies in bees: Implications for phylogenetic analysis of EF-1α sequences in insects. Mol. Biol. Evol. 15, 225–235. https://doi.org/10.1093/oxfordjournals.molbev.a025920 (1998).Article 
CAS 

Google Scholar 
Smith, V. S., Page, R. D. M. & Johnson, K. P. Data incongruence and the problem of avian louse phylogeny. Zool. Scr. 33, 239–259. https://doi.org/10.1111/j.0300-3256.2004.00149.x (2004).Article 

Google Scholar 
Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment search tool. J. Mol. Biol. 215, 403–410. https://doi.org/10.1016/S0022-2836(05)80360-2 (1990).Article 
CAS 

Google Scholar 
Darriba, D., Taboada, G. L., Doallo, R. & Posada, D. jModelTest 2: More models, new heuristics and parallel computing. Nat. Methods 9, 772. https://doi.org/10.1038/nmeth.2109 (2012).Article 
CAS 

Google Scholar 
Huelsenbeck, J. P. & Ronquist, F. MRBAYES: Bayesian inference of phylogeny. Bioinformatics 17, 754–755. https://doi.org/10.1093/bioinformatics/17.8.754 (2001).Article 
CAS 

Google Scholar 
Zwickl, D. J. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Ph.D. Thesis Dissertation, The University of Texas at Austin, Texas (2006).Rambaut, A. FigTree v1.4.2. Institute of Evolutionary Biology (University of Edinburgh, 2014).
Google Scholar 
Brown, C. J. Plumages and measurements of the Bearded Vulture in Southern Africa. Ostrich 60, 165–171 (1989).Article 

Google Scholar 
Chatterjee, P., Payra, A. & Sen, S. Insecta: Phthiraptera. In Faunal Diversity of Indian Himalaya (eds Chandra, K. et al.) 297–304 (Zoological Survey of India, 2018).
Google Scholar 
Liébanas, G. et al. The morphology of Colpocephalum pectinatum (Phthiraptera: Amblycera: Menoponidae) under scanning electron microscopy. Arthropod Struct. Dev. 64, 101085. https://doi.org/10.1016/j.asd.2021.101085 (2021).Article 

Google Scholar 
Pérez, J. M. Sobre algunos aspectos de la parasitación por malógafos en aves de presa. Ph.D. thesis Dissertation, Granada University (1990).Arya, G., Ahmad, A., Bansal, N., Saxena, R. & Saxena, A. K. Nature of placodean sensilla of four ischnoceran Phthiraptera. Entomon 35, 199–202 (2010).
Google Scholar 
Khan, V., Bansal, N., Arya, G., Ahmad, A. & Saxena, A. K. Contribution to the morphology of Degeeriella regalis (Insecta, Phthiraptera, Ischnocera). J. Entomol. Res. 35, 93–96 (2011).
Google Scholar 
Agarwal, G. P. et al. Bio-ecology of the louse, Upupicola upupae, infesting the Common Hoopoe, Upupa epops. J. Insect. Sci. 11, 46. https://doi.org/10.1673/031.011.4601 (2011).Article 
CAS 

Google Scholar 
Singh, P., Gupta, N., Khan, G., Kumar, S. & Ahmad, A. Diagnostic characters of three nymphal instars and morphological features of adult Collard-dove louse Columbicola bacillus (Phthiraptera: Insecta). J. Appl. Nat. Sci. 11, 7–11. https://doi.org/10.31018/jans.v11i1.1855 (2019).Article 

Google Scholar 
Clayton, D. H. & Johnson, K. P. Linking coevolutionary history to ecological process: Doves and lice. Evolution 57, 2335–2341. https://doi.org/10.1111/j.0014-3820.2003.tb00245.x (2003).Article 

Google Scholar 
Barker, S. C. Lice, cospeciation and parasitism. Int J Parasitol 26, 219–222 (1996).Article 
CAS 

Google Scholar 
Page, R. D. M., Clayton, D. H. & Paterson, A. A. Lice and cospeciation: A response to Barker. Int. J. Parasitol. 26, 213–218. https://doi.org/10.1016/0020-7519(95)00115-8 (1996).Article 
CAS 

Google Scholar 
Paterson, A. M. & Gray, R. D. Host-parasite cospeciation, host-switching and missing the boat. In Host-Parasite Evolution: General Principles and Avian Models (eds Clayton, D. H. & Moore, J.) 236–250 (Oxford University Press, 1997).
Google Scholar 
Paterson, A. M., Palma, R. L. & Gray, R. D. How frequently do avian lice miss the boat? Implications for coevolutionary studies. Syst. Biol. 48, 214–223. https://doi.org/10.1080/106351599260544 (1999).Article 

Google Scholar 
Frey, H. & Walter, W. The reintroduction of the bearded vulture Gypaetus barbatus into the Alps. In Raptors in the Modern World (eds Meyburg, B. U. & Chancellor, R. D.) 341–344 (WWGBP, 1989).
Google Scholar 
Pérez, J. M., Sánchez, I. & Palma, R. L. The dilemma of conserving parasites: the case of Felicola (Lorisicola) isidoroi (Phthiraptera: Trichodectidae) and its host, the endangered Iberian lynx (Lynx pardinus). Insect. Conserv. Divers. 6, 680–686. https://doi.org/10.1111/icad.12021 (2013).Article 

Google Scholar  More

Geographical and temporal settingPaleogeographic and biogeographic evidence suggests Madagascar has been an isolated insular unit since the splitting from Greater India around 88 million years ago (Mya)1,57. However, a worldwide mass extinction event is known to have taken place around the K-Pg boundary ~66 Mya affecting Madagascar’s biodiversity2, so using this event as an island age for the current biota may also be appropriate. While Madagascar had a rich Mesozoic vertebrate fauna, including many mammals58,59, molecular phylogenetic data suggest that no mammalian lineage that colonized before the K-T event has survived until the present, although the confidence intervals for the colonization time of the Malagasy Afrosoricida extend until before this age (Fig. 1). Furthermore, the existence of short-lived land-bridges connecting the island to continental Africa at various stages has been proposed60, although this hypothesis has been contested61. We accounted for the uncertainty in island age in our sensitivity analyses, using both 88 and 66 million years (Myr) as island age. We do not consider the possible existence of temporary land-bridges, as we estimate average rates (e.g. of colonization) throughout the entire history of the island (see “DAISIE models” section).Malagasy mammalian diversityWe compiled a comprehensive taxonomic/phylogenetic dataset of the entire assemblage of Malagasy non-marine mammals, including information on phylogenetic relationships, timing and causes of extinction and levels of threat (Supplementary Data S1). We first compiled a checklist of all mammalian species known to have been present on Madagascar in the late Holocene before human arrival and expansion, ~2500 years ago (Supplementary Data S1). The checklist includes all native species that are still extant today and all those that are known or presumed to have gone extinct in the late Holocene. Taxa known only from ancient early- or pre-Holocene fossils that are assumed to have gone extinct long before human presence were excluded, as well as all non-native species. We followed the taxonomy and nomenclature of the Mammal Diversity Database of the American Society of Mammalogists62,63, as of May 2022. Our checklist does differ from that database because of ongoing taxonomic and nomenclatural revisions based on recent molecular phylogenetic analyses, and recent discoveries or descriptions of new species on the island. All such cases and cases where our taxonomy or nomenclature differs from that used in the phylogeny of43 are explained in the column “Taxonomy note” in Supplementary Data S1. Molecular-based taxonomy and morphological taxonomy (such as that from the palaeontological record) can be incongruent because cryptic species “detected” in DNA-based analyses may not be identifiable based on fossil data. The number of species in the existing fossil record is therefore likely an underestimate—we address this in the section “Impact of increased knowledge on the ERT”.To compile the checklist, we used a variety of sources from both the neontological and palaeontological literature. The main sources on Malagasy extant and extinct mammalian species are4,8,25,29,52,57,58, but other published studies were used, particularly for bats, lemurs and cryptic and recently discovered species (Supplementary Data S1). We classified species as endemic or non-endemic to Madagascar, as information on endemicity status is one of the types of data that DAISIE uses to estimate rates of speciation. For non-endemic species, only represented in the dataset by certain species of bats, we noted their range outside of Madagascar in the column “Additional range note”. For the ERT analyses, we also compiled the IUCN Red List status for each species in 2010, 2015 and 2021. We used the digital archive “wayback machine” to obtain the 2010 and 2015 IUCN Red List data, as older versions of listings are not kept online by the IUCN8,41,42. We compiled the checklist using Excel v16.63.Extinct speciesIn the DAISIE analyses we treat species that went extinct due to non-anthropogenic causes (before or after human arrival) as if they were species not known to science. This is because the natural extinction rate that is estimated based on the colonization and branching times extracted from phylogenies (without these extinct species), already accounts for such missing species. These include species that went extinct before the late Holocene and may or may not be known from the fossil record, but also species that have gone naturally extinct after human arrival. In contrast, we consider that anthropogenic extinctions do not contribute to the natural extinction rate. Therefore, we treat species for which an anthropogenic cause of extinction is likely as if they had survived into the present and we include them in the phylogenies, following the approach of Valente et al.22. The rates of speciation, colonization and natural extinction that are estimated from such phylogenies are the natural average rates assuming that humans had no impact on the island—these would be the natural average background rates in the periods pre-dating human arrival.Our checklist includes all species that are hypothesized to have gone extinct in the late Holocene, hereafter termed “recently extinct species”. For these species, we compiled information on whether they have gone extinct before or after human arrival (columns “Last date 14C age BP” and “Extinction before/after humans” in Supplementary Data S1). In assessing whether species went extinct before or after human colonization and expansion, we employ the circa 2500 years BP date as time zero3,4,5,64,65. After permanent settlement, anthropogenic pressures on Madagascar’s biodiversity have intensified, with a visible increase in the past few decades4. Species which are considered in the literature to have gone extinct before human arrival were excluded. For some taxa, it is unclear whether they went extinct before or after human colonization, because the fossil record is insufficiently known. These were included in the list and the effect of their inclusion/exclusion was evaluated in sensitivity analyses (see section below). In addition, for all extinct species in our list, we compiled information on the hypothesized causes of extinction cited in the literature, classifying each extinction as anthropogenic, natural or uncertain. References for timing and causes of extinction are provided in Supplementary Data S1.To account for these uncertainties, we re-ran analyses for two datasets, assuming high and low human impact (see sensitivity analyses). For the high human impact scenario, we assumed all recent extinctions (less than 2500 years ago) to have an anthropogenic cause and therefore included them in the phylogenies and in the counts of pre-human species diversity. In this scenario we also include recently extinct species for which no 14C dates are available, but which have been hypothesized to have gone extinct in the last 2500 years, as well as species with a putative natural cause of extinction, because even natural recent extinctions may have had an indirect human influence. For the high human impact scenario, we thus assume all recent species loss is linked to potential human influence, and we include all those species in the phylogenies. In the low human impact scenario we assume a natural cause for all recent extinctions that have previously been hypothesized to have had a natural cause, for all recent extinctions whose cause is unknown, and for the cases for which it is unclear whether extinction pre-dates or post-dates human arrival (e.g. no radiocarbon dating available). For the low human impact scenario, we excluded all such species from the phylogenies and from the counts of pre-human diversity (it is assumed they are unknown). See column “Low human impact scenario” in Supplementary Data S1. For two of the species for which no date currently exists indicating whether they went extinction before or after humans (the lemurs Mesopropithecus dolichobrachion and Palaeopropithecus kelyus), an anthropogenic cause of extinction has been hypothesized despite the lack of a precise last occurrence date, and we thus considered them to be anthropogenic extinctions in both high and low human impact scenarios.Phylogenetic dataThe source of all our phylogenetic data—including divergence times of Malagasy lineages—is the phylogeny by Upham et al.43, the most comprehensive and complete mammalian phylogeny published to date, including 5911 species of mammals. From this tree, we extracted phylogenetic information with reference to Madagascar on the number of colonization events, the estimated dates of colonization (divergence times from the most closely related non-Malagasy relatives), number of species per monophyletic colonist lineage and the timing of within-island speciation events. We created a Madagascar-specific dataset consisting of a series of multiple subtrees drawn from the same Mammalia-wide dating framework, representing all colonization events for most known late Holocene native mammals on the island, including bats and recently extinct species. We visually inspected the trees using Figtree v1.4.4.Upham et al.43 used two approaches to calibrate their phylogeny: node dating and tip dating. Following the recommendations in that publication, we used the trees based on the node-dating approach, in which node-age priors were placed on the tree based on 17 mammalian fossils and one root constraint. Regarding molecular sampling, they produced two types of trees: DNA-only, with 4098 species sampled in the phylogeny based on molecular data; and completed trees, where they placed an additional 1813 species that were unsampled for DNA in the tree using taxonomic constraints (across multiple posterior trees). The DNA-only trees have the advantage that the topology is based on molecular data, and is likely more reliable, but the disadvantage that DNA sequences were not available for many Malagasy species and so these species needed to be added to the phylogeny for the DAISIE analyses. The completed trees have the advantage that they are near-complete, but the disadvantage that some Malagasy species—particularly several bats that are unsampled for DNA—were placed randomly within a given clade constraint, which may lead some trees in the posterior to have some incorrectly inferred colonizations. We ran analyses on data extracted from both types of trees (see sensitivity analyses).An alternative to using this phylogenetic dataset would be to extract data from separate individual trees from publications with phylogenies focusing on specific clades. There are many such studies, and indeed some of them include taxa that are not present in the Upham et al.43 tree—e.g. new recently described cryptic species that were only identified after molecular analyses, including, for example, the nesomyine rodent Eliurus tsingimbato66 and the mouse lemur species Microcebus jonahi67; or extinct species for which no molecular data exists, but which were included in phylogenetic dating analyses based on morphological data, such as members of the lemur genus Mesopropithecus46. However, we favoured using phylogenetic data from a single study to ensure divergence times are comparable (i.e. same models, assumptions and data), even though this is done at the expense of reduced species sampling. Although we use a single tree (or posterior distribution of single trees) for our dataset, DAISIE treats each Malagasy colonizing lineage as its own separate tree, so we deal with a “forest” of phylogenetic trees, each representing a single Malagasy lineage resulting from one colonization event. For example, the lineages that have radiated on the island have a tree that includes the stem age of the lineage (splitting from the closest sampled mainland relative) and all branching events within the radiation. Lineages with a single species on Madagascar (endemic or non-endemic) are essentially a tree with a single tip and with an age equal to the splitting of that species from its closest (sampled) continental relative.Alternative colonization scenariosThe number of colonization events of Malagasy mammals inferred from the phylogenetic data can vary depending on the placement of some missing taxa in the tree or because some clades have poor branch support and could be the result of one or more colonizations in different trees from the posterior. We considered two alternative colonization scenarios (CS), one where we favour fewer colonizations (CS1) and one where we favour more (CS2). The differences between the two scenarios are summarized in Table S1 (all colonizations shown in Supplementary Data S2, S3). We considered lemurs to be the result of a single colonization event in both scenarios. A recent study68 has suggested that Daubentonia is a separate colonization of Madagascar, but we assigned both the extant aye-aye and the extinct giant aye-aye to the single Lemuroidea (lemurs) clade because that is the only scenario supported by the mammal tree.Adding missing speciesA total of 34 out of 249 species in our Madagascar mammal checklist are not present in the mammal phylogenetic tree43. Most of these (23 species) are extinct species (Data S1). The other 11 species are recently described species (two bats, eight lemurs and one nesomyine rodent), these are indicated in column “Taxonomy note” in Supplementary Data S1. An additional 61 species are included in the completed trees, but not in the DNA-only trees, as no molecular data were available for these. We added the 34 species missing from the mammal tree to both DNA-only and completed trees, and the 61 species missing molecular data to the DNA-only trees. Instead of adding species directly to the posterior distribution of trees and then extracting information from the phylogenies, we assign those species to specific clades using the “missing species” option in DAISIE. This tool allows them to be placed anywhere within the Malagasy clade they are believed to belong to, without specifying a specific topological position within the clade—DAISIE does not use topological information for its estimates. For example, a species of lemur that was missing from the tree was added to the species count of the lemur clade. The information on the clade to which each missing species was added to (under either CS1 or CS2) is provided in Supplementary Data S1.Most recently extinct species are not included in the mammal tree because the original study was primarily focused on the extant mammalian taxa43. Three extinct species of lemur, Archaeolemur majori, Megaladapis edwardsi and Palaeopropithecus ingens are included in their tree based on molecular data obtained from subfossil material. One extinct species of lemur (A. edwardsi), one extinct species of carnivoran (Cryptoprocta spelea) and two extinct hippopotamus species (Hippopotamus madagascariensis and H. lemerlei) are included in their completed trees, i.e. not based on molecular data. We added the remaining extinct species to the phylogenies using the approach explained above (Supplementary Data S1). These were: 1 tenrec (assigned to the Malagasy Afrosoricida (CS1) or Tenrecidae (CS2)); the 2 bibymalagasy (assigned to Malagasy Afrosoricida (CS1) or to Bibymalagasia (CS2)); 1 hippopotamus (assigned to the single hippopotamus clade (CS1) or to one of the two hippopotamus clades (CS2); 1 euplerid carnivore (Cryptoprocta sp. nov., assigned to Eupleridae); 2 bats (1 Paratriaenops, assigned to the Paratriaenops clade; 1 Macronycteris assigned to Macronycteris (CS1) or as its own colonization (CS2)); 3 nesomyine rodents (assigned to Nesomyinae); and 13 lemurs (assigned to the Lemuroidea clade).In a few cases, all descendants from a colonization of Madagascar were missing from the mammal tree. These were added as a separate colonization, using the DAISIE_max_age option, which assumes that they could have colonized at any time since the given age and the present. These were: the two species of Bibymalagasia (CS2, using the stem age of Afrosoricida in the mammal tree as the maximum age of colonization); Chaerephon leucogaster (CS1 and CS2, using crown age of Molossidae family as maximum colonization time); Macronycteris cryptovalorona (CS1 and CS2, using crown age of Hipposideridae family as maximum colonization time); Macronycteris besaoka (CS2, using crown age of Hipposideridae family as maximum colonization time); Hippopotamus laloumena (CS2, using stem age of genus Hippopotamus as maximum colonization time); and Pipistrellus raceyi (absent from the DNA-only tree, we used the crown age of Vespertilionidae as the maximum colonization time). For Miniopterus, the phylogenetic resolution for this radiation is poor (including both Malagasy and non-Malagasy taxa), and we therefore used the crown age of the genus as a maximum colonization time of Madagascar (we chose the crown and not stem because Miniopterus of Madagascar do not diverge early in the genus).Colonization and branching timesFor endemic Malagasy clades (radiations (e.g. lemurs) or clades with a single endemic species, e.g. Pteropus rufus), we assumed the time of colonization of Madagascar coincides with the divergence time from its closest non-Malagasy lineage, i.e. the stem age of the clade. These ages are likely overestimates (e.g. if the tree is incompletely sampled, or if the closest continental ancestor has gone extinct, see ref. 18), but are a good approximation, and we repeated analyses over the posterior distribution of trees to account for age uncertainties. Non-endemic species are represented by a single tip in the mammal tree, and we therefore used the age of that tip as a maximum age of colonization, as the actual colonization time of the Madagascar population is most certainly younger than that age. DAISIE integrates through all possible ages between that maximum age and the present. The only exception are three non-endemic species belonging to the Chiroptera genus Miniopterus, which likely resulted from cladogenesis within Madagascar and became non-endemic by colonizing the Comoros. These are treated as part of the Miniopterus clade or clades (Table S1), and thus contribute to the estimates of cladogenesis on Madagascar (rather than being assigned their own colonist lineage). The branching times within Madagascar radiations were taken directly from the trees. When species within a radiation were missing from the phylogeny, they were included using the DAISIE missing species option (see section above), thus contributing to the estimates of cladogenesis rates for the given clade.We wrote an R script to extract colonization and branching times from the maximum clade credibility (MCC) and posterior trees from the Upham et al.43 mammal phylogeny, add missing species and assign maximum colonization times (if relevant), assuming a variety of scenarios (see “Main dataset and sensitivity analysis” section below). Once the data were extracted from the trees, the script creates DAISIE objects, i.e. datasets in DAISIE format that can be read by DAISIE functions. This script uses functions from phytools v1.2-0 R69, ape v5.6-270 and DAISIE47 v4.0.5 R packages. The R script describes all the steps taken to prepare the phylogenetic data for the DAISIE analyses. The script, the precise source trees that we used from the Upham et al. mammal phylogeny43, as well as all DAISIE objects for the main analyses and the sensitivity analyses are provided in an online repository (https://github.com/luislvalente/madagascar). Analyses were run in R v4.2.1 and RStudio v2022.02.3.DAISIEWe used the DAISIE R package47 to estimate rates of speciation, colonization and extinction (CES rates) of Madagascar mammals using maximum likelihood (ML) under a range of different models and to identify the preferred model given the phylogenetic data. The DAISIE likelihood inference approach is based on theory and methods developed for phylogenetic birth-death models71,72. It has been demonstrated that the shape of phylogenies of extant species contains information about natural extinction rates72. While these approaches have many known limitations73, we have shown in different studies that the DAISIE model is able to accurately estimate extinction rate from simulated datasets for which the extinction rate is known18,74. In addition, unlike most phylogenetic birth-death models, which are single-clade approaches and use only information from branching times, DAISIE has the advantage that it uses information from multiple independent clades and from both colonization and branching times, increasing its statistical power to estimate parameters.We fitted a set of 30 DAISIE models to the phylogenetic data, explained in Table S2. Models M1–M4 assume homogeneous CES rates for all Malagasy mammals, while for models M5–M30 we allow one or more of the CES parameters to vary between non-volant mammals and bats. The set of models include both diversity-dependent and diversity-independent models. Models can differ in the number of parameters: for example, M1 has five parameters (colonization, cladogenesis, anagenesis, K and extinction); M3 has four parameters (same as M1, except that anagenesis is fixed to zero); M5 has six parameters (the same five parameters as M1, plus a parameter for colonization rate which differs for bats); and M6 has five parameters (same as M5 but K is fixed to infinite, i.e. there is no diversity dependence).CES rate heterogeneityDAISIE estimates average CES rates for the island and assumes that these rates are constant through time (except for models that include diversity dependence, in which rates decline with increasing diversity). However, from the geology of the island and the fossil record, we can infer that rates have most likely not been constant. For example, periods of large-scale natural extinction may have taken place throughout the history of the island3,75,76. While there may have been important temporal rate changes, when estimating the future island evolutionary return time (the main purpose of our analyses), we seek to estimate the overall average natural background rates, which incorporate periods of both low and high rates (which will certainly also occur in the future). We fitted two models (M31—bats and non-volant mammals share same rates; and M32—bats and non-volant mammals have different rates) in which colonization rates can shift to a lower or higher rate46 at a certain point in time, but these models were not preferred (Table S3). Therefore, when we estimate the ERTs, we use the average rates for the island as a whole over its entire geological history in the absence of humans. Importantly, although the preferred models assume constant rates, the DAISIE model has been shown to perform very well for ancient continental islands (separated from the mainland very deep in geological time, such as Madagascar), in terms of accurately predicting the number of species, and the number of species and colonizations through time77. In addition, although rates may have been lower or higher at some periods, the average rates are nevertheless informative of the unique geographical setting of the island and the ecological characteristics of the target community—this is particularly valuable, for example, when comparing Malagasy mammals with ERTs from other systems, such as in Caribbean bats22 and New Zealand birds21, both in which rates have also most likely varied through time.There is evidence for rate variation among mammalian lineages78. We therefore chose to test for differential rates for two groups: non-volant mammals and bats. In the context of islands, it is likely that bats will have different rates of colonization due to their higher dispersal abilities, and they may also vary in other parameters. We used the two-type DAISIE model approach first applied to the birds of the Galápagos (Darwin’s finches vs other birds45. While there may also be differences in rates between specific non-volant and bat clades, we favour obtaining average ERTs across the whole fauna, rather than specific ERTs for each lineage. First, assigning unique rates to each lineage would lead to over parameterization, and estimating lineage-specific rates would not be reliable for some individual Malagasy clades that are the product of a single colonization and have few species (e.g. many bat lineages, hippopotamuses, euplerid carnivorans). Thus, we restricted the test of idiosyncrasies to the comparison between bats and non-volant species. Second, an advantage of our approach is that the rates we obtained are based entirely on the phylogenies of Malagasy species and therefore our rates are already very specific to the Malagasy context—whereas comparable methods use average rates worldwide and then extrapolate to the focal lineages24. Third, we are interested in whether total diversity will recover, not whether specific types of species will recover. A trait-dependent diversification model for insular communities that would allow us to obtain ERTs based on, for example, certain morphological traits that may promote diversification, does not currently exist.Main dataset and sensitivity analysesWe consider the main dataset (D1) to comprise: colonization scenario 1 (CS1) with high human impact, using the DNA-only mammal tree (MCC and posterior), island age of 88 Myr. The reason for this is that we consider the CS1 (fewer colonizations) and high human impact scenarios to be the most realistic given the level of isolation of the island and because the evidence for anthropogenic mammalian extinction on Madagascar is compelling and growing6. We also consider the DNA-only tree more appropriate, as species were sampled based on molecular data, and all missing species were included in clades using the DAISIE missing species option, i.e. placing them in a clade but without forcing a given topology within that clade.There are currently ~6500 species of mammals62, but this number was certainly different in the past and only a subset of these constitute the potential mainland pool for Madagascar, which would include African species and to a much lesser extent from the Indian subcontinent or other portions of Asia. For the main dataset we considered the number of species on the mainland pool (M) to be 1000 (approximately the current number of African terrestrial mammal species), but we re-ran analyses with 2000 and 5000 species. For the models where bats differ from non-volant mammals (M5–M30), the proportion of bat species in the mainland pool was set to 0.22, equivalent to the proportion of all mammal species that are bats today.To account for uncertainty in island age, mainland pool size, colonization scenarios, human impact, topology, dating (colonization and branching times), and tree sampling completeness, we ran a series of sensitivity analyses. We re-ran analyses for the MCC tree of the main dataset (that is: DNA-only tree, high impact, CS1), assuming an island age of 66 Myr, and varying pool sizes (for both island ages). Then, fixing the mainland pool to 1000 species and the island age to 88 Myr, we ran DAISIE analyses assuming colonization scenarios CS1 and CS2, high and low human impact. We also repeated analyses using the completed and DNA-only trees, using the corresponding MCC trees for each scenario. In total we ran 13 different scenarios (D1–D13) for the sensitivity analysis (Supplementary Data S4).We used the following approach for ML optimizations on the main dataset and the sensitivity analyses, using the DAISIE_ML function implemented in the DAISE R package. For the analyses on a single MCC tree (all 13 scenarios, including the main dataset), we fitted each of the 30 DAISIE models to each dataset 10 times, using different random sets of starting values for the likelihood optimization (30 × 10 = 300 ML optimizations per scenario, total 3900 ML optimizations). For each scenario, we selected the preferred model by comparing Bayesian information criterion (BIC) and Akaike information criterion (AIC) scores between models. For the main dataset, to examine if the same model is preferred across the posterior distribution of trees, we also ran analyses on the posterior, fitting each model 4 times to each of 100 datasets from the posterior (30 × 4 × 100 = 12,000 optimizations). To obtain confidence intervals for the preferred model of the main dataset, we ran analyses on 1000 trees from the posterior, with 2 random sets of starting values (2 × 1000 = 2000 ML optimizations, Table S4). We ran ERT analyses using the parameters of the preferred model for all 13 scenarios. For the main dataset D1 we also fitted two models with a temporal shift in colonization rate (M31 and M32, Table S3). All analyses were run on the Peregrine cluster of the University of Groningen.For the main dataset, we ran simulations of the best overall rate model using the DAISIE_sim function. Under the parameters of the model, we simulated 5000 islands for 88 million years. We then assessed the goodness of fit of the model to the data by comparing diversity metrics in the simulated datasets to those in the empirical data.In the sensitivity analyses, varying mainland pool size, island age, human impact, colonization scenario or phylogenetic dataset (DNA-only vs completed), had a limited impact in the preferred models or parameters values (Supplementary Data S4). Varying island age, human impact or colonization scenario generally led to only minor changes in parameter values. Varying mainland pool sizes affected the colonization rate, which decreases with mainland pool size because colonization rate is measured per mainland species. When using BIC as the criterion for model selection, M26 was the preferred model in 10 out of 13 scenarios, with M22 being the preferred model under one scenario (D8, DNA-only data, island age 88, M = 1000, CS2, high human impact), and M11 preferred under two scenarios (D12 and D13, completed trees, island age 88, M = 1000, C2, for both high and low human impact) (Supplementary Data S4). Using AIC, alternative models to M26 were preferred for two additional scenarios – M10 was preferred for D10 (completed trees, island age 88, M = 1000, CS1, high human impact) and M11 was preferred for D11 (completed trees, island age 88, M = 1000, CS1, low human impact). We consider the M26 model to be the preferred model overall, because we favour the DNA-only trees (for which M26 was consistently selected as the best model under both AIC and BIC) and the BIC criterion for model selection (shown to perform better than AIC when selecting between DAISIE models18). However, like M26, all three alternative models preferred in some of the sensitivity analyses (M10, M11 and M22) are two-rate models under which bats have a higher rate of colonization than non-volant mammals and differ from non-volant mammals in one or more parameters.Preferred modelThe preferred model is the M26 model. Under this model, the background rate of cladogenesis for Malagasy mammals is 0.33 (0.27–0.36) events per lineage Myr−1 and the rate of anagenesis is 1.47 (1.18–2.12) events per lineage Myr−1 (Table S3, S4). The model is diversity independent (for both bats and non-volant species), meaning that there is no carrying capacity per clade (K per clade is infinite). The rate of natural extinction for non-volant mammals is 0.29 (0.22–0.31) events per lineage Myr−1, and for bats it is 0.46 (0.40–0.50) species per lineage Myr−1. The rate of colonization for non-volant mammals is 0.00036 (0.00027–0.00038) events per mainland species Myr−1, equivalent to 0.28 colonizations per Myr (0.21–0.30). The rate of colonization for bats is much higher, at 0.034 (0.030–0.038) events per mainland lineage Myr−1, equivalent to 7.5 successful bat colonization events per Myr (6.6–8.4).Evolutionary return timesThe island evolutionary return time (ERT) metric estimates the time it would take for an insular community to reach a given species diversity level assuming a given model of macroevolution with certain rates of colonization, speciation and natural extinction22. To estimate ERT, we first counted the number of species that were present on Madagascar in the late Holocene (pre-human diversity) (Table S5). We also counted the number of mammal species estimated to remain extant on the island if currently threatened species go extinct. Threatened species we classified as those that fall under the IUCN categories Vulnerable (VU), Endangered (EN) or Critically Endangered (CR). We estimated the ERT of Malagasy mammals for the following scenarios: (1) the return from current diversity to pre-human diversity, and (2) the return from the diversity that will remain if threatened species (VU + EN + CR) go extinct back to current diversity.There were some differences in ERTs between datasets (D1–D13). Most differences were subtle and some appear counterintuitive (Table S6). For example, a lower number of colonization events in the data (CS1 vs CS2) or a larger mainland pool both lead to a lower colonization rate, which evidently should increase the ERT. However, changes in the mainland pool and the number of colonization events can also lead to very small changes in the diversification rate (cladogenesis minus extinction), which has a much higher impact on the ERT than colonization rate. The largest differences in ERT were between the high and low human impact scenarios, because not only do estimated rates differ, but also the start and target diversities.Change in ERT between 2010 and 2021To compare how ERT for threatened species has been changing through time as human impact and working knowledge increases, we repeated analysis (2) using the IUCN threat statuses from 2010, 2015 and 2021. As the threat status of some species increases, they are uplisted by the IUCN, e.g. from Near Threatened (NT) to Vulnerable (VU) (e.g. the Madagascar rousette (Rousettus madagascariensis)), i.e. becoming threatened under IUCN classification of threat (VU, EN, CR). Although there have been many changes in status between the categories that we considered as threatened (e.g. from VU to EN, or from EN to CR), these are not considered in our analyses, as we only consider changes from non-threatened to threatened. Therefore, our analyses can be considered conservative; hence, the increase in ERT from 2010 to 2021 is probably higher than what we estimate. It is important to keep in mind that for different reasons, which include more intensively studied Malagasy mammalian groups and biopolitics, the manner certain data are weighed during assessments and resulting statutes are not necessarily the same across mammal groups.A total of 72 species were uplisted by the IUCN from a non-threat to a threat category between 2010 and 2021 (Supplementary Data S1). Of these, 57 moved from Not Evaluated (NE) or Data Deficient (DD) categories to one of the threat categories (VU, EN, CR), so may simply represent an increase in knowledge rather than a real increase in threat status. In addition, IUCN categories are not comparable across assessments for species that have undergone taxonomic revisions that may have altered their threat status—e.g. splitting of one species into two allopatric species will lead to a range size reduction. We therefore also repeated analyses only for those species that changed from an evaluated non-threat category (LC, NT) to a threat category (VU, EN, CR) and which have not undergone taxonomic changes between 2010 and 2021 or for which a taxonomic change was not the cause the up-listing. For all species that changed from a non-threat to a threat category between 2010 and 2021 (15 species, Supplementary Data S5) we consulted the literature to find out whether changes in taxonomy took place for that taxon, and whether those changes influenced the up-listing. We identified 10 species (Supplementary Data S5) for which there was no taxonomic change between 2010 and 2021 or for which a taxonomic change did not lead to an up-listing between those years. We then calculated ERTs for a scenario where we assume that only those 10 species were uplisted between 2010 and 2021 (Table S8, scenario C).Species diversity lost versus ERTThe number of species lost through anthropogenic extinctions and the ERT are two alternative ways of looking at the impact of humans on island biota. To assess whether the number of species lost is a good proxy for ERT, we ran simulations to measure how ERT varies with the number of extinct species to be recovered. For example, we compared how the time to return to pre-human diversity varies with starting diversity, e.g. assuming an increasing number of species have gone extinct. We ran simulations using the parameters of the best overall model for the main dataset. We first created 10,000 random start diversities, sampling between 0 species and a target species diversity, assuming variable numbers of species have gone extinct. For each of the starting diversities, we randomly specified a proportion of endemic and non-endemic species. Simulations were run in R.The shape of the island species diversity curve (how the total number of species on an island varies through time) under different DAISIE models can vary at different stages. For example, in an equilibrium model, diversity increases rapidly at early stages and low diversities, but at later stages it plateaus and increases slowly. In non-equilibrium models, diversity increases can for example be low at early stages and faster later. This will have implications for how ERT relates to the number of species that need to be recovered. We therefore separated the results into (a) returning to pre-human diversity (capturing a later stage of the diversity curve), (b) return to contemporary diversity (capturing an intermediate stage) and (c) return to half of the contemporary diversity (capturing early stage of the diversity curve). We did this separately for non-volant mammals and bats.Impact of increased knowledge on the ERTNew species discoveries and increasingly complete IUCN Red List assessments are likely to affect ERT estimates in the future. The discovery of new extant species may lead to an increase in ERT because undiscovered species are more likely to already be threatened (e.g. due to small range and population sizes). Taxonomic revisions may lead to species splits, resulting in additional threatened species. The known fossil record may also include cryptic species that cannot be identified using molecular methods if DNA is not available. The discovery of more taxa that have gone extinct since humans arrived will also likely increase the ERT (return to pre-human diversity). However, if new species are discovered, the rates of colonization and speciation estimated in DAISIE will also increase, and therefore ERTs may not rise dramatically.To assess how future species discoveries may affect our results, we performed analyses where we assume 30 new mammal species (15 bats species and 15 non-volant mammal species) will be discovered in the next 10 years on Madagascar. This is likely an overestimate. We simulated datasets by adding these bat and non-volant species at random locations to the main phylogenetic dataset D1, and repeated this procedure 1000 times. We fitted the preferred DAISIE model to these 1000 datasets and estimated the ERT for each of them. We then assumed that the newly discovered species were (a) all threatened, (b) half of them threatened and half already extinct or (c) all already extinct (since human arrival). The results of these analyses are summarized in Fig. S4. We found that the ERTs for non-volant mammals do not change substantially, increasing slightly or even declining under some scenarios. This is because although the number of species to recover increases, the estimated DAISIE rates also increase. On the other hand, under some scenarios, an increase in the number of extinct or extant bat species leads to large increases in ERT, as it takes longer on average to recover bat species according to the preferred DAISIE model.IUCN Red List assessments should become more comprehensive in the future; currently 8% of recognized Malagasy mammal species are Not Evaluated or assessed as Data Deficient, corresponding to 18 species (10 bats and 8 non-volant species). Thus, we also estimated how the completion of IUCN assessments may affect our results. If all species yet to be assessed by the IUCN were evaluated as threatened in the future, the ERT to return to contemporary diversity would rise to 26.2 (20.8–32) Myr for non-volant mammals (~13% increase) and 6.6 (5.5–7.8) Myr for bats (more than double) (Table S8, scenario A). The increase is proportionally higher for bats because it takes longer to recover bat species and because there are more unevaluated bat species.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article. More

Lander, E. A. A colonialidade do saber: eurocentrismo e ciencias sociais. Perspectivas latino-americanas (CLACSO, 2005).Césaire, A. Guaraguao 20, 157–193 (2005).
Google Scholar 
Fanon, F. Black Skin, White Masks (Pluto, 1986).Mbembe, A. Arts Humanit. High. Educ. 15, 29–45 (2016).Article 

Google Scholar 
Rivera Cusicanqui, S. Ch’ixinakax utxiwa: una reflexión sobre prácticas y discursosdescolonizadores (Tinta Limón, 2010).Castro-Gómez, S. & Grosfoguel, R. El Giro Decolonial (Siglo del hombre, 2007).Grosfoguel, R. Rev. (Fernand Braudel Center) 25, 203–224 (2002).
Google Scholar 
Quijano, A. Int. Sociol. 15, 215–232 (2000).Article 

Google Scholar 
Visvanathan, S. in Science and Citizens: Globalization and the Challenge of Engagement (eds Leach, M. et al.) 83–97 (Zed Books, 2005).Spivak, G. C. Can the Subaltern Speak? (Routledge, 1994).Bhargava, R. Glob. Policy 4, 413–417 (2013).Article 

Google Scholar 
Bhambra, G. K. Crit. Times 4, 73–89 (2021).Article 

Google Scholar 
Sousa Santos, B. Rev. (Fernand Braudel Center) 30, 45–89 (2007).
Google Scholar 
Maas, B. et al. Conserv. Lett. 14, e12797 (2021).Article 

Google Scholar 
Nuñez, M. A. et al. J. Appl. Ecol. 56, 4–9 (2019).Article 

Google Scholar 
Mohammed, R. S. et al. Am. Nat. 200, 140–155 (2022).Article 

Google Scholar 
Rau, J. et al. Ecol. Austral 27, 312–496 (2017).Article 

Google Scholar 
Raja, N. B. et al. Nat. Ecol. Evol. 6, 145–154 (2022).Article 

Google Scholar 
Trisos, C. H., Auerbach, J. & Katti, M. Nat. Ecol. Evol. 6, 1205–1212 (2021).Article 

Google Scholar 
Eichhorn, M. P., Baker, K. & Griffiths, M. Front. Biogeogr. 12, 1–7 (2020).
Google Scholar 
Baker, K., Eichhorn, M. P. & Griffiths, M. Biotropica 51, 288–292 (2019).Article 

Google Scholar 
Brandt, S. et al. Ecol. Evol. 10, 12450–12456 (2020).Article 

Google Scholar 
McGill, B. M. et al. Ecol. Evol. 11, 3636–3645 (2021).Article 

Google Scholar 
Melles, S. J. et al. Ecoscience 26, 323–340 (2019).Article 

Google Scholar 
Adebisi, F. Decolonisation is not about ticking a box: it must disrupt. criticallegalthinking.com, https://criticallegalthinking.com/2020/03/12/decolonisation-is-not-about-ticking-a-box/ (12 March 2020).Güttler, N. Ber. Wiss. 42, 235–258 (2019).
Google Scholar 
Worster, D. J. Nature’s Economy: A History of Ecological Ideas, 2nd edn (Cambridge Univ. Press, 1994).Nicolson, M. Hist. Sci. 26, 183–200 (1988).Article 

Google Scholar 
Lewinsohn, T. M. Filos. História Biol. 11, 347–381 (2016).
Google Scholar 
Raby, M. The Study of Ecology in Latin America and the Caribbean (Oxford Univ Press, 2021).Mignolo, W. D. Coloniality at Large: Time and the Colonial Difference (Taylor & Francis, 2020).Simpson, D. Y., Beatty, A. E. & Ballen, C. J. Trends Ecol. Evol. 36, 4–8 (2021).Article 

Google Scholar 
Smith, L. T. Decolonizing Methodologies: Research and Indigenous Peoples (Zed Books, 2013).Grosfoguel, R. Tabula Rasa 24, 123–143 (2016). More