Ecology

1.
Tomitani A, Knoll AH, Cavanaugh CM, Ohno T. The evolutionary diversification of Cyanobacteria: molecular-phylogenetic and paleontological perspectives. Proc Natl Acad Sci USA. 2006;103:5442–7.
CAS  PubMed  Google Scholar 
2.
Schirrmeister BE, Gugger M, Donoghue PCJ. Cyanobacteria and the Great Oxidation Event: evidence from genes and fossils. Palaeontology. 2015;58:769–85.
PubMed  PubMed Central  Google Scholar 

3.
Fischer WW, Hemp J, Johnson JE. Evolution of oxygenic photosynthesis. Annu Rev Earth Planet Sci. 2016;44:647–83.
CAS  Google Scholar 

4.
Soo RM, Hemp J, Parks DH, Fischer WW, Hugenholtz P. On the origins of oxygenic photosynthesis and aerobic respiration in Cyanobacteria. Science. 2017;355:1436–40.
CAS  PubMed  Google Scholar 

5.
Biller SJ, Berube PM, Lindell D, Chisholm SW. Prochlorococcus: the structure and function of collective diversity. Nat Rev Microbiol. 2015;13:13–27.
CAS  PubMed  Google Scholar 

6.
Sánchez-Baracaldo P. Origin of marine planktonic Cyanobacteria. Sci Rep. 2015;5:14–17.
Google Scholar 

7.
Shang JL, Chen M, Hou S, Li T, Yang YW, Li Q, et al. Genomic and transcriptomic insights into the survival of the subaerial cyanobacterium Nostoc flagelliforme in arid and exposed habitats. Environ Microbiol. 2019;21:845–63.
CAS  PubMed  Google Scholar 

8.
Chrismas NAM, Anesio AM, Śanchez-Baracaldo P. The future of genomics in polar and alpine Cyanobacteria. FEMS Microbiol Ecol. 2018;94:fiy032.
PubMed Central  Google Scholar 

9.
Kashtan N, Roggensack SE, Rodrigue S, Thompson JW, Biller SJ, Coe A, et al. Single-cell genomics reveals hundreds of coexisting subpopulations in wild Prochlorococcus. Science. 2014;344:416–20.
CAS  PubMed  Google Scholar 

10.
Larsson J, Celepli N, Ininbergs K, Dupont CL, Yooseph S, Bergman B, et al. Picocyanobacteria containing a novel pigment gene cluster dominate the brackish water Baltic Sea. ISME J. 2014;8:1892–903.
CAS  PubMed  PubMed Central  Google Scholar 

11.
Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 2012;19:455–77.
CAS  PubMed  PubMed Central  Google Scholar 

12.
Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes. Genome Res. 2015;25:1043–55.
CAS  PubMed  PubMed Central  Google Scholar 

13.
Gurevich A, Saveliev V, Vyahhi N, Tesler G. QUAST: quality assessment tool for genome assemblies. Bioinformatics. 2013;29:1072–5.
CAS  PubMed  PubMed Central  Google Scholar 

14.
Seemann T. Prokka: rapid prokaryotic genome annotation. Bioinformatics. 2014;30:2068–9.
CAS  Google Scholar 

15.
Mistry J, Finn RD, Eddy SR, Bateman A, Punta M. Challenges in homology search: HMMER3 and convergent evolution of coiled-coil regions. Nucleic Acids Res. 2013;41:e121.
CAS  PubMed  PubMed Central  Google Scholar 

16.
Wu M, Scott AJ. Phylogenomic analysis of bacterial and archaeal sequences with AMPHORA2. Bioinformatics. 2012;28:1033–4.
CAS  PubMed  Google Scholar 

17.
Capella-Gutierrez S, Silla-Martinez JM, Gabaldon T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics. 2009;25:1972–3.
CAS  PubMed  PubMed Central  Google Scholar 

18.
Waterhouse RM, Seppey M, Simão FA, Manni M, Ioannidis P, Klioutchnikov G, et al. BUSCO applications from quality assessments to gene prediction and phylogenomics. Mol Biol Evol. 2018;35:543–8.
CAS  PubMed  Google Scholar 

19.
Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30:772–80.
CAS  PubMed  PubMed Central  Google Scholar 

20.
Miller MA, Pfeiffer W, Schwartz T. Creating the CIPRES Science Gateway for inference of large phylogenetic trees. In: Proceedings of the Gateway Computing Environments Workshop (GCE). New Orleans (LA): IEEE; 2010. pp 1–8.

21.
Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015;32:268–74.
CAS  Google Scholar 

22.
Di Rienzi SC, Sharon I, Wrighton KC, Koren O, Hug LA, Thomas BC, et al. The human gut and groundwater harbor non-photosynthetic bacteria belonging to a new candidate phylum sibling to Cyanobacteria. Elife. 2013;2:e01102.
PubMed  PubMed Central  Google Scholar 

23.
Matheus Carnevali PB, Schulz F, Castelle CJ, Kantor RS, Shih PM, Sharon I, et al. Hydrogen-based metabolism as an ancestral trait in lineages sibling to the Cyanobacteria. Nat Commun. 2019;10:1–16.
CAS  Google Scholar 

24.
Letunic I, Bork P. Interactive Tree Of Life (iTOL) v4: recent updates and new developments. Nucleic Acids Res. 2019;47:W256–9.
CAS  PubMed  PubMed Central  Google Scholar 

25.
Tung HoLS, Ané C. A linear-time algorithm for Gaussian and non-Gaussian trait evolution models. Syst Biol. 2014;63:397–408.
Google Scholar 

26.
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B. 1995;57:289–300.
Google Scholar 

27.
Gan F, Bryant DA. Adaptive and acclimative responses of Cyanobacteria to far-red light. Environ Microbiol. 2015;17:3450–65.
CAS  PubMed  Google Scholar 

28.
Revell LJ. phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol Evol. 2012;3:217–23.
Google Scholar 

29.
Levy A, Salas Gonzalez I, Mittelviefhaus M, Clingenpeel S, Herrera Paredes S, Miao J, et al. Genomic features of bacterial adaptation to plants. Nat Genet. 2018;50:138–50.
CAS  Google Scholar 

30.
Zhu Q, Kosoy M, Dittmar K. HGTector: an automated method facilitating genome-wide discovery of putative horizontal gene transfers. BMC Genom. 2014;15:717.
Google Scholar 

31.
Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIAMOND. Nat Methods. 2014;12:59–60.
PubMed  Google Scholar 

32.
Csurös M. Count: evolutionary analysis of phylogenetic profiles with parsimony and likelihood. Bioinformatics. 2010;26:1910–2.
PubMed  Google Scholar 

33.
Enright AJ. An efficient algorithm for large-scale detection of protein families. Nucleic Acids Res. 2002;30:1575–84.
CAS  PubMed  PubMed Central  Google Scholar 

34.
Komárek J. A polyphasic approach for the taxonomy of Cyanobacteria: principles and applications. Eur J Phycol. 2016;51:346–53.
Google Scholar 

35.
Komárek J, Kaštovský J, Mareš J, Johansen JR. Taxonomic classification of cyanoprokaryotes (Cyanobacterial genera) 2014, using a polyphasic approach. Preslia. 2014;86:295–335.
Google Scholar 

36.
Ponce-Toledo RI, Deschamps P, López-García P, Zivanovic Y, Benzerara K, Moreira D. An early-branching freshwater Cyanobacterium at the origin of plastids. Curr Biol. 2017;27:386–91.
CAS  PubMed  PubMed Central  Google Scholar 

37.
de Vries J, Archibald JM. Endosymbiosis: did plastids evolve from a freshwater Cyanobacterium? Curr Biol. 2017;27:R103–5.
PubMed  Google Scholar 

38.
Dagan T, Roettger M, Stucken K, Landan G, Koch R, Major P, et al. Genomes of stigonematalean Cyanobacteria (subsection V) and the evolution of oxygenic photosynthesis from prokaryotes to plastids. Genome Biol Evol. 2013;5:31–44.
PubMed  Google Scholar 

39.
Shih PM, Wu D, Latifi A, Axen SD, Fewer DP, Talla E, et al. Improving the coverage of the cyanobacterial phylum using diversity-driven genome sequencing. Proc Natl Acad Sci USA. 2013;110:1053–8.
CAS  PubMed  Google Scholar 

40.
Sánchez-Baracaldo P, Raven JA, Pisani D, Knoll AH. Early photosynthetic eukaryotes inhabited low-salinity habitats. Proc Natl Acad Sci USA. 2017;114:E7737–45.
PubMed  Google Scholar 

41.
FitzJohn RG, Maddison WP, Otto SP. Estimating trait-dependent speciation and extinction rates from incompletely resolved phylogenies. Syst Biol. 2009;58:595–611.
PubMed  Google Scholar 

42.
Monk JM, Charusanti P, Aziz RK, Lerman JA, Premyodhin N, Orth JD, et al. Genome-scale metabolic reconstructions of multiple Escherichia coli strains highlight strain-specific adaptations to nutritional environments. Proc Natl Acad Sci USA. 2013;110:20338–43.
CAS  PubMed  Google Scholar 

43.
Tripp HJ, Bench SR, Turk KA, Foster RA, Desany BA, Niazi F, et al. Metabolic streamlining in an open-ocean nitrogen-fixing cyanobacterium. Nature. 2010;464:90–4.
CAS  PubMed  Google Scholar 

44.
Scanlan DJ, Ostrowski M, Mazard S, Dufresne A, Garczarek L, Hess WR, et al. Ecological genomics of marine Picocyanobacteria. Microbiol Mol Biol Rev. 2009;73:249–99.
CAS  PubMed  PubMed Central  Google Scholar 

45.
Poulton NJ, Acinas SG, Lauro FM, Cavicchioli R, Swan BK, Hanson NW, et al. Prevalent genome streamlining and latitudinal divergence of planktonic bacteria in the surface ocean. Proc Natl Acad Sci USA. 2013;110:11463–8.
PubMed  Google Scholar 

46.
Bentkowski P, Van Oosterhout C, Ashby B, Mock T. The effect of extrinsic mortality on genome size evolution in prokaryotes. ISME J. 2017;11:1011–8.
CAS  PubMed  Google Scholar 

47.
Steele JH, Brink KH, Scott BE. Comparison of marine and terrestrial ecosystems: suggestions of an evolutionary perspective influenced by environmental variation. ICES J Mar Sci. 2019;76:50–9.
Google Scholar 

48.
Philippot L, Andersson SGE, Battin TJ, Prosser JI, Schimel JP, Whitman WB, et al. The ecological coherence of high bacterial taxonomic ranks. Nat Rev Microbiol. 2010;8:523–9.
CAS  PubMed  Google Scholar 

49.
Luo H, Csűros M, Hughes AL, Moran MA. Evolution of divergent life history strategies in marine Alphaproteobacteria. MBio. 2013;4:1–8.
Google Scholar 

50.
Whitton BA (editor). Ecology of Cyanobacteria II. Dordrecht, Netherlands: Springer; 2012.

51.
Yoshihara S, Katayama M, Geng X, Ikeuchi M. Cyanobacterial phytochrome-like PixJ1 holoprotein shows novel reversible photoconversion between blue- and green-absorbing forms. Plant Cell Physiol. 2004;45:1729–37.
CAS  PubMed  Google Scholar 

52.
Bhaya D, Takahashi A, Grossman AR. Light regulation of type IV pilus-dependent motility by chemosensor-like elements in Synechocystis PCC6803. Proc Natl Acad Sci USA. 2001;98:7540–5.
CAS  PubMed  Google Scholar 

53.
Yang Y, Lam V, Adomako M, Simkovsky R, Jakob A, Rockwell NC, et al. Phototaxis in a wild isolate of the cyanobacterium Synechococcus elongatus. Proc Natl Acad Sci USA. 2018;115:E12378–87.
CAS  PubMed  Google Scholar 

54.
Kehoe DM, Gutu A. Responding to color: the regulation of complementary chromatic adaptation. Annu Rev Plant Biol. 2006;57:127–50.
CAS  PubMed  Google Scholar 

55.
Sánchez-Baracaldo P, Bianchini G, Di Cesare A, Callieri C, Chrismas NAM. Insights Into the evolution of Picocyanobacteria and Phycoerythrin Genes (mpeBA and cpeBA). Front Microbiol. 2019;10:45.
PubMed  PubMed Central  Google Scholar 

56.
Ting CS, Rocap G, King J, Chisholm SW. Cyanobacterial photosynthesis in the oceans: the origins and significance of divergent light-harvesting strategies. Trends Microbiol. 2002;10:134–42.
CAS  PubMed  Google Scholar 

57.
Gan F, Zhang S, Rockwell NC, Martin SS, Lagarias JC, Bryant DA. Extensive remodeling of a cyanobacterial photosynthetic apparatus in far-red light. Science. 2014;345:1312–7.
CAS  PubMed  Google Scholar 

58.
Thiel V, Tank M, Bryant DA. Diversity of chlorophototrophic bacteria revealed in the Omics Era. Annu Rev Plant Biol. 2018;69:21–49.
CAS  PubMed  Google Scholar 

59.
Kühl M, Trampe E, Mosshammer M, Johnson M, Larkum AWD, Frigaard N-U, et al. Substantial near-infrared radiation-driven photosynthesis of chlorophyll f-containing Cyanobacteria in a natural habitat. Elife. 2020;9:e50871.
PubMed  PubMed Central  Google Scholar 

60.
Oren A. Microbial life at high salt concentrations: phylogenetic and metabolic diversity. Saline Syst. 2008;4:1–13.
Google Scholar 

61.
Sääf A, Baars L, von Heijne G. The internal repeats in the Na+/Ca 2+ exchanger-related Escherichia coli protein YrbG have opposite membrane topologies. J Biol Chem. 2001;276:18905–7.
PubMed  Google Scholar 

62.
Price GD, Woodger FJ, Badger MR, Howitt SM, Tucker L. Identification of a SulP-type bicarbonate transporter in marine Cyanobacteria. Proc Natl Acad Sci USA. 2004;101:18228–33.
CAS  PubMed  Google Scholar 

63.
Sakamoto T, Inoue-Sakamoto K, Bryant DA. A novel nitrate/nitrite permease in the marine cyanobacterium Synechococcus sp. strain PCC 7002. J Bacteriol. 1999;181:7363–72.
CAS  PubMed  PubMed Central  Google Scholar 

64.
Carrieri D, Wawrousek K, Eckert C, Yu J, Maness PC. The role of the bidirectional hydrogenase in Cyanobacteria. Bioresour Technol. 2011;102:8368–77.
CAS  PubMed  Google Scholar 

65.
Tamagnini P, Axelsson R, Lindberg P, Oxelfelt F, Wunschiers R, Lindblad P. Hydrogenases and hydrogen metabolism of Cyanobacteria. Microbiol Mol Biol Rev. 2002;66:1–20.
CAS  PubMed  PubMed Central  Google Scholar 

66.
Huisman J, Codd GA, Paerl HW, Ibelings BW, Verspagen JMH, Visser PM. Cyanobacterial blooms. Nat Rev Microbiol. 2018;16:471–83.
CAS  PubMed  Google Scholar 

67.
Ben Fekih I, Zhang C, Li YP, Zhao Y, Alwathnani HA, Saquib Q, et al. Distribution of arsenic resistance genes in prokaryotes. Front Microbiol. 2018;9:2473.
PubMed  PubMed Central  Google Scholar 

68.
Fürst-Jansen JMR, de Vries S, de Vries J. Evo-physio: on stress responses and the earliest land plants. J Exp Bot. 2020;71:3254–69.
PubMed  PubMed Central  Google Scholar 

69.
Murik O, Oren N, Shotland Y, Raanan H, Treves H, Kedem I, et al. What distinguishes Cyanobacteria able to revive after desiccation from those that cannot: the genome aspect. Environ Microbiol. 2017;19:535–50.
CAS  PubMed  Google Scholar 

70.
Gul N, Poolman B. Functional reconstitution and osmoregulatory properties of the ProU ABC transporter from Escherichia coli. Mol Membr Biol. 2013;30:138–48.
PubMed  Google Scholar 

71.
Pathak J, Ahmed H, Singh PR, Singh SP, Häder D-P, Sinha RP. Mechanisms of photoprotection in Cyanobacteria. In: Mishra AK, Tiwari DN, Rai AN. editors. Cyanobacteria. Cambridge: Academic Press; 2019. pp. 145–171.

72.
Meulenbroek EM, Peron Cane C, Jala I, Iwai S, Moolenaar GF, Goosen N, et al. UV damage endonuclease employs a novel dual-dinucleotide flipping mechanism to recognize different DNA lesions. Nucleic Acids Res. 2013;41:1363–71.
CAS  PubMed  Google Scholar 

73.
Richardson EJ, Bacigalupe R, Harrison EM, Weinert LA, Lycett S, Vrieling M, et al. Gene exchange drives the ecological success of a multi-host bacterial pathogen. Nat Ecol Evol. 2018;2:1468–78.
PubMed  Google Scholar 

74.
Wiedenbeck J, Cohan FM. Origins of bacterial diversity through horizontal genetic transfer and adaptation to new ecological niches. FEMS Microbiol Rev. 2011;35:957–76.
CAS  PubMed  Google Scholar 

75.
Smillie CS, Smith MB, Friedman J, Cordero OX, David LA, Alm EJ. Ecology drives a global network of gene exchange connecting the human microbiome. Nature. 2011;480:241–4.
CAS  PubMed  Google Scholar 

76.
Sheppard SK, Guttman DS, Fitzgerald JR. Population genomics of bacterial host adaptation. Nat Rev Genet. 2018;19:1–17.
Google Scholar 

77.
Oliveira PH, Touchon M, Rocha EPC. Regulation of genetic flux between bacteria by restriction-modification systems. Proc Natl Acad Sci USA. 2016;113:5658–63.
CAS  PubMed  Google Scholar 

78.
Jain R, Rivera MC, Lake JA. Horizontal gene transfer among genomes: the complexity hypothesis. Proc Natl Acad Sci USA. 1999;96:3801–6.
CAS  PubMed  Google Scholar 

79.
Pál C, Papp B, Lercher MJ. Adaptive evolution of bacterial metabolic networks by horizontal gene transfer. Nat Genet. 2005;37:1372–5.
PubMed  Google Scholar  More

Compared to the pre-war baseline, our results show an overall numerical population depletion of 77% across all mammal species during the war period, with some species experiencing a decline of up to 80% of their pre-war baseline abundance. Moreover, this degree of wildlife decline was not reversed by the end of the post-war period. This overall pattern of marked large mammal declines has not been previously documented at sites exposed to intense armed conflicts, which in Angola and other combatant countries profoundly dismantle the socio-political structure, natural resource management activities and enforcement practices such as bushmeat market inspection22,23. We emphasize that even during post-war peace times, wild mammal populations in Angola will fail to recover as long as rural people living in war-torn countries remain armed and wildlife management regulations cannot be enforced.
In Angola, there has been a process of slow disarmament of citizens by the government, which has disrupted hunting practices and reduced hunting pressure on local wildlife populations. However, meaningful recovery of institutional policy on protected areas and wildlife populations have not yet been implemented in all the Quiçama region, which is now largely occupied by a mix of native peoples, war refugees, and former combatants. As a consequence, post-war mammal population rebounds have been entirely restricted to some small-bodied species, likely due to their higher fecundity, in contrast with the low reproductive rate of medium- to large-bodied species, which continue to be slaughtered by fire weapons and other hunting techniques. Automatic rifle confiscation from citizens is an important factor in reducing hunting pressure, thereby favouring the recovery of local game biomass13,24. However, without the critical intervention of well-designed government policies, the baseline structure of large terrestrial vertebrate assemblages is unlikely to recover. For example, in the post war-zone Gorongosa National Park, Mozambique, the total biomass density of nine focal large mammal species had recovered in 2018 by ~ 80% of the pre-war baseline density, but the community composition had shifted dramatically compared to the pre-war baseline due to asymmetric recovery rates across species, with smaller antelope species exceeding the abundance of formerly dominant megaherbivores25. In particular, waterbuck abundance had increased by an order of magnitude, with more than 55,000 individuals accounting for over 74% of large-herbivore biomass by 2018. By contrast, elephant, hippo, and buffalo, which accounted for 89% of the pre-war biomass, now comprised only 23%25.
Considering carnivores, only lion populations in Mozambique’s Gorongosa National Park persisted throughout the war26, whereas leopards also persisted at intermediate abundance in forest environments in our study area. Both of these studies also recorded hyenas and jackals. At Quiçama, however, only two local informants had seen or killed hyenas or jackals over the last 5 years. The collapse of these carnivores has important ecological implications on their roles in key ecosystem linkages, such as necromass scavengers and energy and nutrient transfer27.
Defaunation can have important impacts not only in terms of severe depletion of vulnerable species but also on general ecosystem functions, including predation, herbivory, carrion removal and disease control28,29. For example, the Mozambican Civil War (1977–1992) induced to a catastrophic large‐herbivore die-off in Gorongosa National Park, which was followed by 35 years of woodland expansion, most severely in areas where pre‐war herbivore biomass was greatest7. This expansion included the invasive Mimosa pigra shrub—considered one of the world’s 100 worst invasive plant species30. Tree cover increased in four of the park’s five major habitat zones by 51% to 134%. Local informants in our study explained that in many areas of Quiçama the landscape have become more wooded since the collapsed of large herbivores, although this remains anecdotal. The most parsimonious explanation in both Mozambique and Quiçama is that a severe reduction in browsing pressure enhanced tree growth, survival and/or recruitment7.
Before the Angolan civil war, the protected areas of the Quiçama region once safeguarded one of the largest world populations of Red Buffalos (around 8,000 individuals) across both savannah and forest landscapes31. However, we found that poaching had severely reduced Red Buffalos to small populations restricted to some forest fragments in the southern Quiçama area. Landscape structure and vegetation cover clearly interfere with the degree of hunting efficiency because they affect hunter velocity, understorey visibility, size-selective prey detectability, and hunting techniques. In open savannah areas, larger animals can be easily detected, resulting in far more efficient use of long-range projectiles fired by automatic rifles and other weapons carried by distant hunters17. Also, compared to forest environments, motor vehicles gain much more feasible access into savannah landscapes when both pursuing prey and transporting carcasses to markets, which further explains the higher depletion rates of the savannah megafauna32. Mammals inhabiting more accessible open areas are therefore more vulnerable. For example, a study on Europe’s largest terrestrial mammal (Bison bonasus) showed that stronger pre-historic hunting pressure in open landscapes forced these animals into closed-canopy forest as a refuge habitat since the Pleistocene, leaving the legacy of the last native bison populations being restricted to forest areas15. However, habitat quality in forest refugia is not necessarily suitable. For instance, eland and roan antelope at Quiçama were unable to seek refugia in forest remnants, unlike other large-bodied species such as elephant and red buffalo. This likely explains why over 90% of our interviewees reported the conspicuous absence of those two ungulate species in the entire area.
Our model shows that commercially valuable target species in both savannah and forest habitats were not necessarily the most abundant during the early stages of the war. This is likely because the abundance of large-bodied species was then not low enough to discourage hunters from pursuing them. However, during the late and post-war periods, depletion rates of large-bodied prey in savannahs habitats were so high that pursuing them had become less worthwhile than pursuing midsized species. Because of the elevated time/energy costs of capturing large-bodied prey species in savannah areas, hunters become more selective in this habitat compared to the forest. On the other hand, given that levels of depletion of large-bodied species in forest areas were lower, most of these species continued to be killed in this habitat type, but resulted in smaller offtakes. Hunters also selected midsized species to compensate for any losses in the overall biomass of prey profiles. In the aftermath of the war, the gradual shift in prey size structure towards smaller-bodied species progressed and midsized species were most frequently selected by hunters in both savannah and forest habitats. In a study in Ghana, commercial trophy hunting for ivory, as opposed to subsistence hunting, was more sensitive to the density of elephants and enforcement efforts to inhibit poaching, supporting the notion that commercial hunting often depends mainly on overall prey abundance33.
Hunter preference for large- and medium-bodied species is higher because they yield higher catch-per-unit-effort in terms of meat biomass and other products (e.g. ivory and skin). As such, most species smaller than 12 kg were not a target game species and their relative abundance remained unchanged over the assessed periods. The fact of whether or not any given species had been reported as a hunting target during the war did not affect its pre- to post-war change in perceived abundance (see Fig. 4A) was influenced by the depletion of some small-bodied species which were not commercially harvested during the war, but were still hunted—because they were crop-raiders or depredated livestock—at a time when plenty of ammunition was readily available. That subsistence and/or commercial game hunting can have a profound detrimental effect on the biomass of large-bodied species has been widely documented34,35. However, we note that the abundance of medium-sized species at Quiçama continues to decline. In contemporary Africa, mammal populations have shown a ‘U-shaped’ abundance trend. Perhaps because small-bodied species are higher-fecundity and/or bypassed by hunters, large-bodied species have been targeted by wildlife management and conservation programs, whereas intermediate-sized species have experienced the steepest declines as they are usually hunted, but lack active management and can exhibit slow reproductive rates36. Therefore, there is a need to also directly manage midsized species, rather than assume that management actions targeting the most iconic ‘umbrella’ taxa will lead to effective conservation of all species. In our study area, for example, the greatest conservation focus should be allocated to bushbuck (Tragelaphus scriptus), currently the most hunted species at Quiçama (mainly for trade). This ungulate species has received no attention from regional to national scale conservation programs37.
We found little or no change in the relative abundance of small mammals, perhaps because these small-bodied species were neither commercially valuable nor harvested for local subsistence. However, comparing our results with other studies using combined sampling techniques such as camera traps, net, and microphones16, we recognize that some small mammals could have been undersampled, despite the enormous usefulness of LEK approaches in meeting the aims of this study. Regarding the primates, cultural influences such as food taboos may have important roles in mediating population declines of overexploited species. However, primates elsewhere in Africa and the Neotropics comprise the largest number of species threatened by hunting across the world’s mammals38. We therefore caution that the future bushmeat trade in Angola could, in fact, begin to target primates as other more desirable large-bodied species become gradually depleted and economically extinct. In addition, we highlight the increased risk of zoonotic diseases, given that our close phylogenetic relationship with nonhuman primates increases the likelihood of animal-to-human pathogen spillover39 and because the risk of disease emergence among mammalian orders is highest in bats (risk rate = 2.64), followed by primates (2.23), ungulates (2.09), rodents (1.81) and carnivores (1.39)40.
Modern armed conflicts affect terrestrial wildlife through a range of interactions, including tactical military operations. However, the consequences of socio-economic upheaval and livelihood disruption associated with a civil war can outweigh the direct effects of military activity9. Among the 24 mechanisms through which armed conflicts are known to affect wildlife, eight (86% of all existing case studies by 2016) were “non-tactical” pathways involving institutional decay, displacement of people and economic upheaval13. Accordingly, our results show that the main consequences of the war in the Quiçama region were non-tactical, such as much greater access to powerful fire-weapons, which were widely used by hunters and the military, even though their initial distribution purpose was to arm the population to fight against rival militias. The widespread use of automatic weapons intensified the overkill of large mammals, increasing hunting efficiency and the number of hunted species. In addition, wildlife culls were intensified during all brief periods of cease-fire because once the probability of encountering guerrilla groups was reduced, armed hunters felt safer and increased the amount of time allocated to hunting activities as well as the size of their catchment areas.
Ivory tusks from elephants killed at Quiçama were removed by the natural resource sector of each political party responsible for the catch, probably in exchange for automatic weapons1,41. Consequently, Angola’s elephants during the 1980s drew international alarm with reports of up to 100,000 elephants exterminated within rebel-controlled territories42. Park rangers were also victims of the threat from rebel groups, which was exacerbated by hundreds of outside hunters gaining access to the Quiçama area. Similarly, in the Okapi Reserve in the Democratic Republic of Congo, park guards were forced to abandon their posts following guerrilla attacks and were unable to prevent elephant poaching and bushmeat extraction13,43.
Strategic installation of both fixed and mobile military bases throughout protected areas is a tactical manoeuvre that greatly facilitates access to rifles and ammunition by all residents. However, in some situations this can potentially benefit wildlife populations elsewhere by effectively creating a “no human’s land”. This was the case in the Demilitarized Zone separating North and South Korea, which has been uninhabited by humans, thereby becoming a unique nature reserve containing the last refugia of Korean natural heritage23. Therefore, some pathways can show both positive and negative consequences for wildlife, depending on the spatial extent and timescale considered. In fact, if on one hand, exclusion zones often create protected areas for wild nature, on the other hand, sites overrun by war refugees will succumb to much greater hunting pressure. Where the civil war was most intensive in Eastern Angola, many populations of endangered wild species have been identified44, whereas in Western Angola, where the armed conflict was patchy or episodic, we found that wild populations of a similar set of species spiralled down into steep declines or were driven to local extinction. Despite intensive post-war efforts in clearing and deactivating landmines, millions of hectares of these explosive weapons zones remain under interdiction in Europe, Africa, and Asia45. This unpredictable distribution of landmines is also a double-edged sword because many refugees did not return to their original households after the war terminated because of risks associated with landmines. Some of the most intact ecosystems of Central America, for example, have not been threatened by habitat conversion by agrarian peasants because they were seeded with landmines during the civil wars46. Nevertheless, landmines also pose threats to wildlife, killing for example at least 30 elephants in Angola’s southern provinces42. Also, when landmines explode, they shatter soil systems, rip up plant life and disrupt water flows, all of which accelerate widespread ecosystem disruption46.
The main impacts of the Angolan civil war on terrestrial mammals of Quiçama occurred indirectly from military tactics or from “non-tactical” pathways and resulted from wholesale institutional and socioeconomic changes, rather than directly from military tactics. In view of all our findings and related literature, we present a summary flow diagram showing how modern armed conflicts can impact wildlife in modern war zones (Fig. 6). We divide the impact of wars into (A) tactical pathways, which are directly or indirectly derived from military unrest, associated military tactics or supporting military activities; and (B) “non-tactical” pathways, which stem from broad socio-political and economic changes associated with armed conflicts, including major institutional or policy failure, movement of refugees, and severely altered economies, local livelihoods and ecosystems.
Figure 6

Pathways through which modern armed conflicts can affect wildlife populations within war zones. Distinct pathways linking armed conflict to wildlife outcomes organized thematically in “tactical” pathways (which arise directly from the conflict itself and are associated with military tactics or supporting military activities) and “non- tactical” pathways (which stem from broad socio-political and economic changes associated with the armed conflict, including changing institutional dynamics, movement of people, and altered economies and livelihoods). Blue and red boxes represent either positive or negative effects, respectively.

Full size image

Finally, we highlight that 36 countries worldwide are currently experiencing civil wars and most of these conflicts are fuelled or funded by international interests or started after an external intervention. These internationalized conflicts are more prolonged and less likely to find a political solution47. Mirroring our study area, protected areas confronting military conflicts elsewhere become surrounded by armed citizens and can rely on little, if any, national and international support to combat poaching by armed people48. Therefore, considering measures can reduce the impact of warfare on wildlife, we emphasize the intentional or inadvertent complicity of foreign powers, which should also promote policies to mitigate the detrimental environmental impacts of armed conflicts.
We conclude that armed conflicts remain a poorly understood driver of wildlife population collapses and our results indicate that although individual conflicts can have either positive or negative impacts, the overarching trend is clearly negative and the mere propagation of warzones, regardless of their intensity, is sufficient to heavily deplete wildlife populations. In the interest of preventing wildlife collapses in other parts of the world, we highlight that civil wars can vastly increase the availability of automatic weapons/ammunition which are typically used to deplete wildlife; this consequently leads to intense slaughter and major wildlife declines, especially in more accessible open habitats. This may be easier stated than done, but we conclude that policy strategies that can prevent the consequences of warfare, as shown here, remains a key conservation priority. We realize, however, that this rests on recalcitrant political will to promote robust public policies, which are rare priorities in rebuilding nation-states. It is critical to restore vertebrate community structure, but this may take many decades and require active intervention efforts. A multifaceted strategy to prevent previous war-zones from becoming “empty forests” or “empty savannas’’—severely degrading patterns of diversity, ecosystems functioning and ultimately human welfare—is therefore quintessential. More

Model components
The results of this paper are derived from a model that is built around the economic sectors outlined as the most important drivers of planetary pressures in Supplementary Table 1. This includes production sectors that have an important direct effect on the ESPs or that have important links to such sectors. They may be linked by using output from such sectors as inputs, providing inputs to such sectors, competing for inputs with such sectors or providing outputs that serve as substitutes for the output from those sectors. The resulting set of included production sectors are: agriculture (producing food and biofuel), energy services, fossil-fuel extraction, renewable energy (other than biofuel), fertilizer production, phosphate extraction, water supply, fisheries, and industrial manufacturing. The demand for final consumption goods is derived from the maximization of households’ utility. Since we have economic policies in the model, we are implicitly assuming some government entity that imposes these policies, but since we consider the policies exogenous (not, e.g., determined to optimize some objective) we do not explicitly model the government.
We solve the model as a competitive equilibrium where we assume that all agents maximize their respective objectives while taking prices as given (prices are given from the perspective of the individual agent, but are endogenously determined by aggregate supply and demand). We then analyze changes in the endogenously determined model variables in response to an assumed exogenous change in economic policy.
In the model, competition for resources thus leads to a number of important trade-offs. These arise from three main sources including, alternative uses of the output of a sector (e.g., output from the agricultural sector can be used as food or biofuels), sectors competing for the use of inputs (e.g., land can be used for agriculture, forestry or maintained as undisturbed natural land) or from inputs being substitutes or complements in production or consumption (e.g., nitrogen and phosphorus preferably being used in fixed proportions).
The production sectors are modeled either by using an explicit production function or by a production cost function. A production function is specified for agriculture, energy services, fertilizer production, fisheries, timber production and industrial manufacturing sectors since their factor inputs are directly connected to one or more ESPs (see the previous section on “Economic drivers of planetary pressures”), thus making their input substitutability important. For all sectors except agriculture, we use one level constant elasticity of substitution (CES) functions. For agriculture, we use a nested CES function (see below). Sectors whose production processes are of less importance, are represented by a production cost function. These sectors include phosphate, water, fossil fuel, and renewable energy. Also, in many sectors, certain inputs e.g., labor and capital, are economically important but their explicit modeling is not directly relevant for our analysis (i.e., of negligible importance to the ESPs). To account for these inputs, we include an aggregate input, which we refer to as other inputs, in all production sectors except energy services and assume that these are supplied with a given sector-specific price elasticity of supply. The possibility of adjusting these other inputs leads to decreased use in sectors where their marginal value decreases and increased use in sectors where their marginal value increases, and thus to some extent captures the possibility to move inputs between sectors in response to changing economic conditions.
We will now present the model sectors in more detail. A list of model quantities, their prices and uses can be found in Table 1 (different uses of a quantity are denoted by subscripts).
Table 1 Model quantities, prices and uses.
Full size table

The agricultural sector uses inputs land (LA), fertilizers (P), water (W), energy services (({{mathcal{E}}}_{A})) and other inputs (MA) as inputs to produce output that can be used for food or biofuels. Producers maximize their profit, taking prices as given. Their profit maximization problem is

$$mathop{max }limits_{{L}_{A},P,W,{{mathcal{E}}}_{A},{M}_{A}}{p}_{A}Aleft({L}_{A},P,W,{{mathcal{E}}}_{A},{M}_{A}right)-{p}_{L}{c}_{A}({L}_{A}){L}_{A}\ -, {p}_{P}P-{p}_{W}W-{p}_{{mathcal{E}}}{{mathcal{E}}}_{A}-{p}_{{M}_{A}}{M}_{A},$$
(1)

where cA(LA) captures the cost of converting land to agricultural land. The agricultural production function is a CES function between land and non-land inputs, where non-land inputs are aggregated using a CES function.
The energy-services sector combines energy from different sources into a bundle of energy services (({mathcal{E}})). The different sources are biofuels (AB), fossil fuels (({E}_{{mathcal{E}}})) and renewables (R). The producers in this sector solve the profit maximization problem

$$mathop{max }limits_{{A}_{B},{E}_{{mathcal{E}}},R}{p}_{{mathcal{E}}}{mathcal{E}}({A}_{B},{E}_{{mathcal{E}}},R)-{p}_{A}{A}_{B}-{p}_{E}{E}_{{mathcal{E}}}-{p}_{R}R.$$
(2)

We model production of fertilizers (P) as using fossil fuel (EP), phosphate (({mathcal{P}})) and other inputs (MP). The use of fossil fuel is intended to capture the fossil-fuel (more specifically natural-gas) intensive production of the nitrogen component of fertilizers. We thus treat fossil fuel use in fertilizer production as a proxy for nitrogen. The profit maximization problem of fertilizer producers is

$$mathop{max }limits_{{E}_{P},{mathcal{P}},{M}_{P}}{p}_{P}Pleft({E}_{P},{mathcal{P}},{M}_{P}right)-{p}_{E}{E}_{P}-{p}_{{mathcal{P}}}{mathcal{P}}-{p}_{{M}_{P}}{M}_{P}.$$
(3)

For timber production (T) we only consider the input land (LT) and other inputs (MT). The producers then solve the maximization problem

$$mathop{max }limits_{{L}_{T},{M}_{T}}{p}_{T}T({L}_{T},{M}_{T})-{p}_{L}{c}_{T}({L}_{T}){L}_{T}-{p}_{{M}_{T}}{M}_{T},$$
(4)

where cT is a cost of converting (e.g., clearing) land for forestry.
Industrial manufacturing (Y) requires energy (({{mathcal{E}}}_{Y})) and other inputs (MY). While we refer to this sector as manufacturing, the substitutability between energy and other inputs is chosen to match that of the economy as a whole. The substitutability thus reflects not only the manufacturing sector but also the service sector that has a significantly lower energy intensity but is economically important. The maximization problem of the representative producer is

$$max {p}_{Y}Yleft({{mathcal{E}}}_{Y},{M}_{Y}right)-{p}_{{mathcal{E}}}{{mathcal{E}}}_{Y}-{p}_{{M}_{Y}}{M}_{Y}.$$
(5)

The fisheries sector uses inputs fossil fuel (EF) and other inputs (MF). The producers solve the maximization problem

$$mathop{max }limits_{{E}_{F},{M}_{F}}{p}_{F}F({E}_{F},{M}_{F})-{p}_{E}{E}_{F}-{p}_{{M}_{F}}{M}_{F}.$$
(6)

Extraction of fossil fuel (E) is modeled by assuming a gross extraction cost (gE) that increases with increased extraction (gE(E) thus gives the total cost of extracting quantity E). We assume that the tax on fossil fuels (a percentage tax τE) is paid by the firms that extract and sell it. Extraction firms solve the profit maximization problem

$$mathop{max }limits_{E}frac{{p}_{E}}{1+{tau }_{E}}E-{g}_{E}(E).$$
(7)

The sectors phosphate (({mathcal{P}})), water (W), renewable energy (other than biofuels) (R) and the other inputs (MA, MF, MP, MT, and MY) are similarly represented by a production or extraction cost and the profit-maximization problem of the producers are given by

$$mathop{max }limits_{X}{p}_{X}X-{g}_{X}(X) ,, {rm{for}} ,, Xin {{mathcal{P}},W,R,{M}_{A},{M}_{F},{M}_{P},{M}_{T},{M}_{Y}}.$$
(8)

We have now described the maximization problems underlying decisions made by all producers. The representative household also solves a maximization problem, maximizing the utility derived from consumption. The households’ preferences are represented by utility function U and the utility-maximization problem, subject to the income being I, is given by

$$mathop{max }limits_{{A}_{{mathcal{F}}},F,Y,{L}_{U},T} Uleft({mathcal{F}}left({A}_{{mathcal{F}}},Fright),tilde{{mathcal{F}}}left(Y,{L}_{U},Tright)right)\ {rm{s}}.{rm{t}}. ,, {p}_{A}{A}_{{mathcal{F}}}+{p}_{F}F+{p}_{Y}Y+{p}_{L}{L}_{U}+{p}_{T}Tle I.$$
(9)

This specification has divided consumption into two levels. While this division is not necessary at this level of generality, it clarifies the assumed substitutabilities between goods. We assume greater substitutability within than between categories. The upper level consists of food (({mathcal{F}})) and non-food ((tilde{{mathcal{F}}})) goods, with the former category consisting of food from agriculture and from fisheries, and the latter of manufactured goods, natural land and timber. The inclusion of natural land is intended to capture various ways in which households’ demand for natural lands lead to land being kept from other uses, e.g., preservation of land as national parks. We assume that timber is consumed directly by the households.
This completes the description of the modeling of all decision-making agents in the model. In addition to conditions derived from these maximization problems, we must also specify market-clearing conditions that make sure that supplied and demanded quantities add up.
For land (L), the total supply is assumed to be fixed:

$$L={L}_{A}+{L}_{T}+{L}_{U}.$$
(10)

The remaining market-clearing conditions are for agricultural production

$$A={A}_{{mathcal{F}}}+{A}_{B},$$
(11)

fossil fuel

$$E={E}_{{mathcal{E}}}+{E}_{F}+{E}_{P}$$
(12)

and energy services

$${mathcal{E}}={{mathcal{E}}}_{A}+{{mathcal{E}}}_{Y}.$$
(13)

In summary, production functions, market-clearing conditions, budget constraints and first-order conditions from the maximization problems of representative agents provide us with 41 equilibrium conditions pinning down the 41 endogenous prices and quantities. The full set of equilibrium conditions are available in the Supplementary Methods.
Solution Approach
We note a few features of our model, some of which have already been mentioned: there are no explicit externalities; policies are applied exogenously; all sectors are assumed to be competitive; market clearing determines the equilibrium. In this context, we can work with the decentralized equilibrium, which may be analyzed by considering the first order conditions. In our model, there are 41 unknown prices and quantities in the model, determined by 41 equilibrium conditions. Being exogenous, policies represent parameters that are known in advance; denote a generic “policy” pertaining to any one ESP by τ. Let Xi denote the generic ith variable, an endogenous price or quantity. The jth equilibrium condition can then generally be written as:

$${G}_{j}left({X}_{1},ldots ,{X}_{41};tau right)=0.$$
(14)

This system of equations implicitly define all resulting equilibrium quantities and prices as functions of the policy i.e. ({X}_{i}={X}_{i}left(tau right)).
There are now two solution approaches: the first is to solve the set of resulting non-linear equations (and thereby obtain all the equilibrium values); the second is to trace out marginal changes in the equilibrium values in response to a change in the policy, τ. The latter approach can be illustrated by considering the total derivative of the equilibrium conditions with respect to the policy. This leads to a system of equations, with the jth equation being

$$mathop{sum }limits_{i}^{41}left[frac{{X}_{i}}{{G}_{j}}frac{partial {G}_{j}}{partial {X}_{i}}hat{{X}_{i}}right]=-frac{1}{{G}_{j}}frac{partial {G}_{j}}{partial tau },$$
(15)

where

$$hat{{X}_{i}}equiv frac{1}{{X}_{i}}frac{d{X}_{i}}{dtau }$$
(16)

is the relative change in variable Xi. These can be interpreted as a linear approximation of the percentage change in the variable induced by a one percentage point increase in the fossil fuel tax. Assume, for instance, that we get ({hat{X}}_{i}=2) and consider a one percentage point increase in the tax rate, ΔτE = 0.01. We would then get (frac{1}{{X}_{i}}Delta {X}_{i}approx {hat{X}}_{i}Delta {tau }_{E}=0.02). Hence, a one percentage point increase in the tax induces a two percent increase in the quantity. The result is a system of 41 equations in 41 unknowns, the (hat{{X}_{i}}), and is most useful because of linearity in the unknowns. Indeed this approach can be viewed as linear approximation of the equilibrium response to a change in the policy parameter. The required empirical parameter values needed for numerical computations are fewer, easier to find, and easier to interpret. Furthermore, if considering changes in other parameters of the model (e.g., changes in other policies) only the right-hand side of (15) needs to be changed.
Data and parametrization
We parameterize the model based partly on data extracted directly from the widely-used GTAP database, described below, and partly on empirical estimates from various sources in the literature. As described above, we mainly need three types of values: quantity shares, expenditure shares and elasticities of various kinds. In total, we need 39 empirical estimates to run the model. In our computations, we set the initial carbon price equal to zero. In reality there are various forms of carbon prices. It is difficult to get a precise measure of all these, but the global average is likely a relatively small negative price. For our analysis, this makes little difference. Assuming a different initial price would scale all results somewhat since the effect of a one percentage point increase in the price would, relatively speaking, be smaller or larger depending on the initial price. All other parameter values that we use are empirically derived based on the current effective carbon price. In the following section, we provide tables with parameter values and their sources.
The first type of parameter that occurs are quantity shares. By quantity share ({Q}_{X,{X}_{Z}}) we mean the share of total quantity X used in a specific sector Z. The full set of values, including their sources are given in Table 2. The exceptions are the quantity shares of fossil fuel going to different sectors and the share of agricultural production going to food or biofuel. These were derived as follows.
Table 2 Parameters—quantity shares.
Full size table

Total energy consumption in 2011 was 12,225 Mtoe30. Out of this, 10624 Mtoe came from fossil fuel related sources. Fertilizer production uses about 1.2% of total energy supply and almost all of this comes from fossil fuels31. Hence we assume that the share of fossil fuels going to fertilizer production is ({Q}_{E,{E}_{P}}=frac{12,225}{10,624}times 1.2 % approx 1.4 %). For fisheries production, we assume a global fuel consumption of 40 billion litre’s of fuel32. Assuming that this is mostly diesel, this corresponds to 40 Mtoe of fossil fuel or ({Q}_{E,{E}_{F}}=frac{40}{10,624}approx 0.4 %) of total fossil fuel use. Finally we assume the remaining fossil fuels are used in energy production i.e., 98.2%.
In order to compute the share of agricultural production going to bioufuels we used data underlying the FAO Agricultural Outlook report 2016–202533. For each major agricultural commodity (e.g., wheat, maize, rice, etc.) we computed the share of agricultural production used for biofuels and then computed a weighted sum using the fraction of land used to harvest a specific commodity as weight. This resulted in a quantity share ({Q}_{A,{A}_{B}}approx 3.8 %).
Agriculture accounts for only a relatively small proportion of total final energy demand in both industrialized and developing countries. In OECD countries, for example, around 3–5% of total final energy consumption is used directly in the agriculture sector, while for developing countries, the equivalent figure is likely slightly higher in the a range of 4–8% of total final commercial energy use34. Based on these estimates, we concluded that ({Q}_{{mathcal{E}},{{mathcal{E}}}_{A}}) = 5% constitutes a reasonable baseline.
The second type of that occurs in our equilibrium conditions are expenditure shares. The expenditure share ({Gamma }_{X}^{Z}) of input X in sector Z is the share of total spending on inputs in sector Z that goes to X. To pin down these at the global level, we employed the GTAP database15. More specifically, we used the GTAP data set corresponding to the year 2014, for 141 countries and 57 sectors. The GTAP database is a unique global economic data set constructed by collating and reconciling data on national input-output tables, international trade, production, consumption, and macro-economic data sets from various international data sources. This has further been extended by ref. 35 to include renewable energy commodities, based on several energy data sources, including the International Energy Agency (IEA) data set and the World Bank data set. Furthermore, ref. 36 has extended this even further to include water as an endowment, used in both agricultural and other sectors. Finally, we have a data set in which we can derive the shares of labor, capital, land, water, and several other inputs in producing all commodities. Some inputs, such as fertilizers are not separately identified in this data set, but they are subsumed in broader GTAP sectors such as chemicals, rubber, and plastics. Therefore, we make broad reasonable assumptions to derive the shares of such granular-level inputs; for example, we assume that most of agricultural consumption of output from the GTAP sectors chemicals, rubber, and plastics are fertilizers and pesticides. For all production sectors except energy services, we assign the residual expenditure share, remaining when all inputs of direct interest have been accounted for, to other inputs M. The details are given below and summarized in Table 3.
Table 3 Parameters: expenditure shares (source: GTAP).
Full size table

Agriculture. Our agricultural production function distinguishes between land and non-land inputs (with “other inputs” in the non-land category). The expenditure share of land is 19.2%. The expenditure shares of fertilizers, water, energy, and other inputs are 6.43%, 1.93%, 3.33%, and 71.1%, respectively. Their respective shares out of non-land inputs are their total shares divided by the total non-land share. This means that ({Gamma }_{{L}_{A}}^{A}=0.192), ({Gamma }_{{tilde{L}}_{A}}^{A}=0.808), ({Gamma }_{P}^{{tilde{L}}_{A}}=frac{0.0643}{0.808}=0.0796), ({Gamma }_{W}^{{tilde{L}}_{A}}=frac{0.0193}{0.808}=0.0239), ({Gamma }_{{{mathcal{E}}}_{A}}^{{tilde{L}}_{A}}=frac{0.0643}{0.808}=0.0412), and ({Gamma }_{{M}_{A}}^{{tilde{L}}_{A}}=frac{0.711}{0.808}=0.880).
Energy services. The expenditure shares of biofuels, fossil fuels and renewables are 0.37%, 94.33%, and 5.30% respectively. That is ({Gamma }_{{A}_{B}}^{{mathcal{E}}}=0.0037), ({Gamma }_{{E}_{{mathcal{E}}}}^{{mathcal{E}}}=0.9433), and ({Gamma }_{R}^{{mathcal{E}}}=0.0530).
Utility. The expenditure shares of food from agriculture, fish, manufactured goods, recreational land use, and timber are 11.93%, 0.42%, 86.86%, 0.15%, and 0.65%. This gives expenditure share of food ({Gamma }_{{mathcal{F}}}^{U}=0.1235) and expenditure share of non-food goods ({Gamma }_{tilde{{mathcal{F}}}}^{U}=0.8765). The within-category expenditure shares are ({Gamma }_{{A}_{{mathcal{F}}}}^{{mathcal{F}}}=frac{11.93}{12.35}=0.9660), ({Gamma }_{F}^{{mathcal{F}}}=frac{0.42}{12.35}=0.0340), ({Gamma }_{Y}^{tilde{{mathcal{F}}}}=frac{86.86}{87.65}=0.9910), ({Gamma }_{{L}_{U}}^{tilde{{mathcal{F}}}}=frac{0.15}{87.65}=0.001711), and ({Gamma }_{T}^{tilde{{mathcal{F}}}}=frac{0.65}{87.65}=0.007416).
Timber. The expenditure shares of land and other inputs are 37.48% and 62.52%, respectively. That is ({Gamma }_{{L}_{T}}^{T}=0.3748) and ({Gamma }_{{M}_{T}}^{T}=0.6252).
Composite goods. The expenditure shares of energy services and other inputs are 6.38% and 93.62%, respectively. That is, ({Gamma }_{{{mathcal{E}}}_{Y}}^{Y}=0.0638) and ({Gamma }_{{M}_{Y}}^{Y}=0.9362).
Fertilizers. The expenditure share of energy is 10.95%. The factor share of phosphate is assumed to be a share ({xi }_{{mathcal{P}}}=0.5) out of the factor share of non energy intermediates 62.53%. That is ({Gamma }_{{E}_{P}}^{P}=0.1095) and ({Gamma }_{{mathcal{P}}}^{P}=0.5* 0.6253=0.3127). this leaves the expenditure share or other inputs as ({Gamma }_{{M}_{P}}^{P}=0.5778).
Finally, we need several estimates of elasticities, including the elasticity of substitution, price elasticity of supply and elasticities of conversion costs. For the majority of parameters, we were able to track down estimates from the literature which are presented together with their corresponding reference in Table 4. Where the uncertainty in the estimates were high we employed a wide band for the sensitivity analysis. The parameters that are varied in the sensitivity analysis are indicated as [min, max, and mean] with mean being the baseline values.
Table 4 Parameters—elasticities and quantities.
Full size table

Numerical results
The full sets of changes in our model quantities and prices resulting from the two policies are presented in Table 5.
Table 5 Baseline results.
Full size table

We now describe the mapping from changes in model variables to effects on ESPs. For the model variables freshwater (W), natural land-use (LU), phosphate (({mathcal{P}})), and nitrogen (assumed to be proportional to fossil fuel use in fertilizer production EP), there is a simple one-to-one mapping with model variables. For climate change, ocean acidification, biodiversity loss and aerosol loading, however, the mapping is more complicated. For climate change and ocean acidification, we measure the change in pressure on both ESPs as the net change in CO2 emissions. For biosphere integrity, we measure changes in pressure as a change in threats to endangered species (more details on this are given below). We measure aerosol loading as changes in aerosol optical depth. For chemical pollution and ozone depletion, we map pressures to contributing sectors, but do not make any quantitative analysis of the net effects.
Climate change and ocean acidification—are both driven by carbon emissions and we use these emissions as our proxy for the pressures inflicted on these boundaries. To translate changes in model variables into changes in emissions, we use data from refs. 37,38. From the figure on page 2 of ref. 37 we get the percentage contribution of carbon dioxide emissions per sector outlined in the report. Using these percentages we can thus recover the amount of actual carbon emissions in gigaton carbondioxide (GtCO2) per year connected to a specific variable in our model.
Using this approach, we start by looking at the energy-related emissions that, according to ref. 37, account for a total of 66.5%. Multiplying by the aggregate total emissions in 2005 (44.15 GtCO2) we get 29,36 GtCO2. Next, we allocate these energy-related emissions to the energy service production sector, fossil fuel extraction, and emissions from fertilizer production. From ref. 37 we have that 6.4% (2.826 GtCO2 eq) of the total energy-related emissions is due to extraction processes. Based on ref. 38, fertilizer production is estimated to cause emissions of 0.575 GtCO2 eq. Hence we can split the total energy-related emission of 29.36 GtCO2 based on these percentages. This implies that 25,960 GtCO2 will be connected to the energy services output in our model, 2.826 GtCO2 is attributed to the fossil fuel extraction process and 0.575 GtCO2 is connected to fertilizer production.
The other emission-related variables in our model are more straightforward. Emissions from industrial processes in ref. 37 are assigned to manufacturing in our model (In total 4.6% = 2.031 GtCO2). Emission from land-use change are assigned to the change in natural land in our model (12.2% = 5.387 GtCO2). Emissions from agriculture are assigned to the total agricultural production variable (13.8% = 6.093 GtCO2). For the fisheries sector,39 estimate carbon dioxide emissions to be ~0.14 GtCO2.
Using these assignments as a status quo, we can calculate the total policy impact by simply multiplying the percentage change in our model variables resulting from the policy by the status quo emission levels. In total, our model variables cover ~97.4% of the emissions outlined in ref. 37. The results of this exercise, in terms of percentage changes to each planetary pressure, is outlined in Supplementary Table 2 for the carbon tax policy and Supplementary Table 3 for the combined carbon tax and biofuel tax policy.
To summarize, we find that a 1% increase in the carbon tax leads to a reduction in carbon dioxide emissions by −0.25% or −0.11 GtCO2 yr−1, which is what we use as an indicator of the change in pressure accrued to the climate change and ocean acidification boundary. For the combination of carbon and biofuel tax, the change is −0.26% or −0.12 GtCO2 yr−1.
Biodiversity loss—is a notoriously difficult task to assess at a global scale. Studies that quantify terrestrial biodiversity losses resulting from the environmental pressures of human activities typically focus on land-related impacts40,41. There are, however, multiple other environmental pressures causing loss of biodiversity that are not related to land-use42. In ref. 3 the global extinction rate is used as one way of quantifying this boundary (defined as extinctions per million species-years). Here, we will make use of the IUCN Red List of Threatened Species to derive a measure of biodiversity loss. The Red List identifies not only the species that have been confirmed to have gone extinct but also the species that are currently threatened and, if pressures remain, may become extinct in the future.43 identify the drivers behind the prevalent threats to the species on the Red List in a comprehensive assessment of more than 8000 species. These drivers can be directly identified as variables in our model. In ref. 43 there is overlap between threats in the sense that multiple activities can pose threats to a given species. We refer to a decrease in an activity posing a threat to a certain number of species as a decrease in threats. Without knowing the overlap between threats, we can not translate this into changes in number of threatened species. Therefore, we use the change in threats as our measure. Agricultural activity poses threats to 5295 species, which is the largest number of threats. The second-largest threat comes from logging, which threatens 4049 species, and we assign this to timber production in our model. Apart from those, we make the following assignments. Pollution from agriculture threatens 1523 of the species and this is assigned to fertilizer production. Over exploitation (fishing), threatening 1118 species, is assigned to fisheries production. Energy production (oil and gas) and renewable energy production account for threats to 56 species, which we assign to fossil fuel extraction and renewables. Finally, threats from urban development (industrial), pollution (except agriculture), human disturbance (work), transport, energy production (mining) summed to 3573 which we assign to manufacturing. There are also significant biodiversity effects of climate change, which threatens 1688 of the species. In this analysis, we abstract from the effects of changes in one ESP on other ESPs (unless the ESP is directly captured by a model variable). We can note, however, that including the effects of climate change would lead to larger decreases of biodiversity loss.
Hence, having connected the categories of threats to species by driver in ref. 43 to our model variables, we can measure the biodiversity impact of a policy by assessing whether the number of threats increases or declines as a result of the policy. For example, if the agricultural production increases by 1% as a result of a policy in our model then this would increase the number of threats from agricultural activity by 52.95 (0.01 × 5296).
The results, in terms of percentage change to the number of threatened species, are outlined in the column labeled Biodiv. in Supplementary Table 2 for the carbon tax policy and Supplementary Table 3 for the combined carbon tax and biofuel subsidy removal. To summarize, this implies that the total number of threats decline by 0.018% for the carbon tax and by 0.011% for the combined carbon and biofuel tax.
Finally, it should be noted that there are indeed several caveats to our approach for assessing biodiversity loss. First, it should be noted that this measure of biodiversity loss is just a proxy for true biodiversity loss. Future work would benefit from assessing the drivers of the actual rate of species loss as defined in e.g., ref. 1. Furthermore, we have taken the description of threats in43 and mapped them to our model variables. For instance, all threats assigned to agriculture in ref. 43 are assigned to agricultural production in our model. Perhaps some part of these threats come from land use change associated with agriculture rather than agriculture as such. In that case they should be mapped to our land use variables. We do not have a proper basis for such reassignment and, therefore, stick close to their assignment. Qualitatively, this distinction could matter for the carbon tax in isolation, but will not be important for the carbon tax combined with biofuel policy.
Aerosol loading—is proxied following3 which use aerosol optical depth (AOD) as an indicative measure of planetary pressure. To determine how AOD changes as a result of policy, we use data from three sources44,45,46. The impact is calculated as follows. First, we calculate a global average estimate of AOD from the main regional anthropogenic sources (sulfur (0.0392), black carbon (0.0003) and organic carbon (0.0011)) provided by ref. 44. Second, we use data from ref. 45 to calculate the share of global aerosol contributing emissions for each of these respective sources (sulfur (3.6%), black carbon (32%) and organic carbon (63%)) that stem specifically from biomass burning (assuming that approximately 90% of biomass burning emissions result from land-use change46). Third, using these estimates, we can calculate the amount of global AOD which ought to be attributed to emissions from fossil fuel and biofuels (0.038) and biomass burning (0.0022). These estimates are then connected to the model variables fossil fuel consumption (in energy services, fertilizer production and fisheries), biofuel production and change in natural land. In total, a 1% carbon tax leads to a 0.0136% (−5.5 × 10−6) decline in AOD and the combined carbon tax and biofuel policy leads to a decline of 0.014% (−5.7 × 10−6) (further details can be found in Supplementary Tables 2 and Table 3).
Stratospheric ozone depletion and chemical pollution—are not directly quantified in terms of their effects on the boundaries. Stratospheric ozone depletion increases with N2O emissions from agricultural production, fossil fuel use, manufacturing, and biofuels. For an increase in the carbon tax all these activities except for biofuel use decreases. The net effect is thus potentially ambiguous. If the carbon tax increase is complemented with a decrease in biofuel subsidies, all relevant variables decrease and we conclude that the net effect is a decrease in the pressure.
Chemical pollution. Chemical pollution increases in manufacturing, extracted fossil fuels, total agricultural production, agricultural production for food, fossil fuel use in fertilizer production, and fossil fuel use in energy services production. All these activities decrease with a carbon tax, with or without a biofuel policy. Hence, we conclude that chemical pollution will decrease in both cases.
For the remaining boundaries—the impacts are easier to assess since they are directly tied to specific model variables. First, the impact on the biogeochemical flows is assigned to the model variables phosphate and fossil-fuel use in fertilizer production. While the former is self-explanatory, the latter is used as a proxy for nitrogen, which relies almost entirely on fossil fuels in its production. For phosphorus, we translate the change into Gg P yr−1 using the value for current flows (mined and applied to erodible soils) from ref. 3:  −0.000068 × 14,000 ≈ −0l9 Gg P yr−1 for the carbon tax and  −0.0005 × 14,000 ≈ −7 Gg P yr−1 for the combined carbon and biofuel policy. For nitrogen, we translate the change into TgN yr−1 using ref. 3: −0.0013 × 150 ≈ −0.2 TgN yr−1 for the carbon tax and   −0.0018 × 150 ≈ −0.28 TgN yr−1 for the combined carbon and biofuel policy. Second, for the land-system boundary, we rely on the model variable natural land use as an indicator of the direction this boundary is moving in. This is translated into MHa using the average (between high and low value) for “natural forests” in ref. 47:  −0.00014 × 3507 ≈ −0.5 MHa for the carbon tax and 0.00043 × 3507 ≈ 1.5 MHa for the combined carbon and biofuel policy. Third, the freshwater boundary is directly tied to the water variable in our model. We translate this into km3 yr−1 using the value for current use in ref. 3, we the reduction is given by 2600 × 0.00009 ≈ 0.24 km3 yr−1 for the carbon tax policy and  −2600 × 0.00036 ≈ −0.93 km3 yr−1 combined carbon tax and biofuel policy.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article. More

Cranial skeleton
The right premaxilla, the left maxilla, some teeth, the palatine and the palpebral are the best-preserved cranial remains, although a fragmentary right prefrontal could be also present. Complete descriptions for these bones were possible after a micro CT-scanning (Fig. 2).
Figure 2

3D reconstruction of the skull of Ogresuchus furatus (MCD-7149) in (a) lateral, (b) medial, (c) dorsal, (d) palatal, and (e) cranial view. (f) Volume rendering of the segmented neurovascular network of the trigeminal nerve overlaid on the articulated premaxilla and maxilla. app anterior palpebral, ch choana, dn dentary notch, en external naris, f neuro-vascular foramen, if inferior foramen, l-mx lacrimal-maxilla contact, m1-5 maxillary tooth, mes medial shelf, mx maxilla, paf palatal foramen, pd paramedian depressions, pfr prefrontal, plt palatine, pltf palatine foramen, pm1-4 premaxillary tooth, pmx premaxilla, pmx-mx premaxilla-maxilla contact, poas posantral strut, s apicobasal sulcus, snv-tgn V supranarial vessels and the trigeminal nerve V (ophthalmic branch), mv-tgn V maxillary vessels and the trigeminal nerve V (maxillary branch). Scale bar = 2 cm.

Full size image

The right premaxilla is exposed on the rock in lateral view. It is a medio-laterally thin bone, and dorso-ventrally higher than rostro-caudally wide. The caudal margin is sinuous, making a dorso-caudal projection of the premaxilla for the contact with the nasals, and articulating with the lost right maxilla in a sigmoid suture (see specular image of Fig. 2). This margin is larger than the rostral premaxillary margin, making a sharp snout. The premaxilla makes the ventral, lateral and part of the dorsal margins of the external naris (Fig. 2a,b,e), which opens directly rostrally in the lower part of the snout. Except for the sloping wall of the naris and the lateral side of the tooth row, the lateral surface of the premaxilla is ornamented by a shallow pit-and-bulge pattern. Four premaxillary tooth positions are present (Fig. 2a,b,d). All the alveoli are of similar length and elliptical shape, although the third is slightly larger than the others. There is also a large foramen between the first alveolus and the naris. The third premaxillary tooth is preserved. It is conical with a very sharp crown and very labiolingually compressed. The crown is curved lingually and mesially. The mesial and distal margins of the crown are rounded and do not bear carinae. The enamel is ornamented with few apico-basal ridges that cross the crown continuously (Fig. 2a). In palatal view, the premaxilla makes a large incisive foramen separated from the tooth row. The premaxilla-maxilla suture is oriented anteromedially. At least two paramedial depressions are visible mesially and distally to the second tooth position of the premaxilla.
The left maxilla is also exposed in its lateral side, but on the opposite side of the rock respect to the premaxilla (Fig. 1C). It is latero-medially compressed, dorso-ventrally large and rostro-caudally short. The lateral surface is gently rugose, ornamented with the same pit-and-bulge pattern as the premaxilla. The maxilla is subpentagonal in outline. The anterior margin of the maxilla is oblique, because the premaxilla-maxilla suture is located into a notch for the reception of the dentary caniniform (Fig. 2a). The dorsal surface for the contact with the nasals is straight and reduced, and then, the dorsal border of the maxilla slopes ventro-caudally for contacting the lacrimal and, ventrally, the jugal. There is no evidence for an anteorbital fenestra. The maxilla projects medially from its ventral margin, making the secondary palate. In medial view, a septum appears on the lateral wall of the maxilla and turns caudally over the palatal portion of the bone, covering the internal breathing chamber. Caudally to the origin of this septum in the lateral wall, a big foramen opens to trigeminal passage. In palatal view, the maxillae branches meet completely anteriorly to the palatines (Fig. 2d). The maxilla makes the anterior border of the suborbital fenestra, precluding ectopterygoid-palatine contact in this margin.
Only five maxillary tooth positions are present (Fig. 2a,b,d). The maxilla preserves the second, third and fourth erupted teeth. The fist is partially preserved unerupted within the alveolus. The third maxillary tooth is the largest, whereas the fourth is the smallest of the three, although the fifth might be even smaller. The alveolar margin is ventrally arched, reaching the greatest depth at the third maxillary position. After the third maxillary tooth, the alveolar margin turns dorsally making a small notch, where the fourth alveolus is located. A row of eight foramina is present in the lateral side over the alveoli. The crowns are curved lingually and distally. The cross section is labio-lingually compressed. The mesial margins of the crowns are rounded, but the distal margins bear unserrated carinae. The enamel is ornamented with several conspicuous ridges that cross the crowns continuously from the base to the apex (Fig. 2a).
The palatine is an elongated bone rostro-caudally oriented, forming part of the narial passage. It is almost straight, though the caudal end is slightly wider than its rostral one. The anteriormost edge is not preserved, but the maxillary outline reveals a sharp anterior margin of the palatine, exceeding the anterior end of the suborbital fenestra and extending between the maxillae (Fig. 2d). The palatine forms the medial margin of the suborbital fenestra. The posterior ends of the palatines define the anterior and lateral margins of a large choanal opening. The anterior margin of the choana is situated between the suborbital fenestrae. Another D-shaped fenestra opens in the middle of the palatal shaft, anteriorly to the choana.
The anterior palpebral is large, and it is not sutured to the adjacent bones. The bone is subtriangular with a wider anterior end, and its major axis oriented antero-posteriorly (Fig. 2c). The anteromedial border is projected medially, forming a sharp crest for the articulation with the prefrontal. The bone is elongate posteriorly and forms the lateral margin of the supraorbital fenestra. The contact of palpebrals is not preserved, but the preserved portion suggests an oval supraorbital fenestra with an antero-posterior major axis.
Axial skeleton
Most of the dorsal series and few caudal vertebrae are identified. Preserved dorsal series includes seven complete and three fragmentary vertebrae, almost in articulation. These vertebrae are tentatively identified as 5th to 14th dorsal vertebrae. They are exposed in dorsal view, except 6th and 14th vertebrae that show their caudal view. Vertebral centra are amphicoelous. Prezygapophyses and postzygapophyses are well developed, with rounded margins, and laterally oriented. However, no variation in their orientation is observed along the dorsal series. The matrix partially hides the vertebrae, therefore some additional characters (i.e., orientation of articular facets; presence and morphology of a suprapostzygapophyseal lamina) cannot be assessed. The prezygapophyses seem to fuse with the transversal processes from the 7th dorsal vertebra on, as described in other related taxa as Notosuchus terrestris18, Baurusuchus albertoi19, Pissarrachampsa sera10 and Campinasuchus dinizi20. However this condition must be taken with caution, because it is only based on the 7th and 11th vertebrae. Transversal processes are hidden by the matrix in the rest of the series. Neural spines are broken in all the vertebrae except in the 14th dorsal. This spine is well developed and high, corresponding to half of the total height of the vertebra. However, based on the broken basis of neural spines along the series, the spine is medio-posteriorly located on the neural arch, as in B. albertoi19 and Campinasuchus20. A few distal caudal vertebral centra are also preserved, without association with neural spines and transverse processes.
In addition, some dorsal ribs are also identified. These elements are flattened. The proximal end shows the capitulum and the tuberculum for articulating with the associated vertebrae. Capitulum and tuberculum are separated by a well-marked U-shaped depression. The shaft is ventrally curved and shows a median longitudinal depression, unlike Campinasuchus20. At middle length the shaft makes torsion, being antero-posteriorly flattened at proximal half and medio-laterally fattened at distal half.
Forelimb
Only the right ulna, and the metacarpals I, II, III and IV are well identified. The proximal epiphysis of the right radius is also probably preserved (Fig. S9).
The ulna is an elongated and latero-medially flattened bone, as in other sebecids, baurusuchids, and notosuchians10,19,20. It is exposed in lateral side. In lateral view, the bone is arquated, displaying a concave anterior margin and a concave posterior one. The bone becomes shaper on its distal portion. The distal condyles are lost. The proximal end is cranio-caudally expanded. The proximal articular surface is concave, with the caudal olecranon process more developed than the cranio-lateral one. The lateral face bears a shallow longitudinal groove for the insertion of M. extensor carpi radialis brevis pars ulnaris, delimited caudally by a ridge for the insertion of M. flexor ulnaris10,21.
The proximal epiphysis of the radius is not well preserved. In proximal view it is a sub-squared bone with wide condyles, but it is strongly damaged hampering the assessment of detailed morphology.
Metacarpals were identified based on it general outline. The metecarpals I and II are almost complete, but the II, IV and the probable V are distally broken. Metacarpals decrease in width and robustness from the I to the V, being the first the largest. Each of them has an expanded proximal portion for articulating with the next metacarpal. The width of this expansion also decreases in size accordingly. In MI and III, the distal condyles bear a circular central depression for the attachment of M. interossei is observed21. These bones are similar to those referred to other baurusuchids10,19.
Hindlimb
A partial left femur, both tibiae and an indeterminate metatarsal were identified.
The femur is broken in two parts. The shaft seems almost cylindrical, but both proximal and distal ends are lost hindering any accurate morphological description.
Both tibiae are exposed in posterior view. They are long and medially curved bones, as in B. albertoi16, Sebecus22, Stratiotosuchus23 and Mariliasuchus8, differing from the straight condition in Crocodylia. Left tibia is preserved only in its distal portion, but the right tibia is almost complete. The tibial shaft is bowed posteriorly and medially, as in Sebecus22. This tibia is expanded at both ends, although the proximal articular surface is not well preserved. The distal end of the tibia is divided into lateral and medial portions. The medial portion is mesio-distally projected, forming an oblique distal margin. The lateral portion is well developed. This condition is present in other notosuchians as Stratiotosuchus, Notosuchus, Araripesuchus, Yacarerani, Pissarrachampsa and Sebecus8,10,22,24,25.
A left metatarsal is well preserved. It is a long and slender bone, compressed cranio-caudally. The shaft is almost straight with expanded proximal and distal ends. The proximal end shows well-marked lateral and medial condyles separated by a shallow concavity. The distal condyles are rounded, making a squared epiphysis. A lateral circular concavity is observed in both sides for the attachment of the M. interossei, as in the metacarpals. Based on the moderate expansion of the proximal end, this bone is tentatively considered as the metatarsal I.
Remarks
Based on the reconstructed 3D model, the general outline of the skull (Fig. 2), especially the lateromedially compressed and dorsoventrally high premaxillae and maxillae and the reduced dental formula (four or five maxillary teeth), resemble the typical doggy-shaped baurusuchid skulls7,26. The maxilla of Ogresuchus specially resembles that of Gondwanasuchus27, but although both taxa show apicobasal sulci on their teeth, Gondwanasuchus bears serrated dentition. The ornamentation pattern of Ogresuchus is also similar to Caipirasuchus teeth, although Caipirasuchus shows a highly specialized dentition composed by three serrated morphotypes in a continuous tooth row, not separated by a premaxillary-maxillary notch12. The anterior palpebral of Ogresuchus is unusually elongated. This bone differs from the morphology observed in most basal notosuchians, baurusuchids and sebecids; although comparisons are hindered because of the palpebral is not preserved in several species. On the other hand, the shape of the anterior palpebral is reminiscent to Gondwanasuchus27 and Araripesuchus tsangtsangana24. Finally, The absence of antorbital fenestra in Ogresuchus differs from many basal notosuchians and some baurusuchids12,26,27,28,29,30,31,32. This condition is similar to those observed in some basal notosuchians, a few baurusichids, and sebecids7,31,33. More

1.
Pabi, S., Van Dijken, G. L. & Arrigo, K. R. Primary production in the Arctic Ocean, 1998–2006. J. Geophys. Res. 113, 1998–2006 (2008).
Google Scholar 
2.
Uitz, J., Claustre, H., Gentili, B. & Stramski, D. Phytoplankton class-specific primary production in the world’s oceans: Seasonal and interannual variability from satellite observations. Global Biogeochemical Cycles 24, GB3016 (2010).
ADS  Google Scholar 

3.
Arrigo, K. R. & van Dijken, G. L. Continued increases in Arctic Ocean primary production. Prog. Oceanogr. 136, 60–70 (2015).
ADS  Google Scholar 

4.
Moore, C. M. et al. Processes and patterns of oceanic nutrient limitation. Nat. Geosci. 6, 701–710 (2013).
ADS  CAS  Google Scholar 

5.
de Baar, H. J. W. et al. Importance of iron for plankton blooms and carbon dioxide drawdown in the Southern Ocea. Nature 373, 412–415 (1995).
ADS  Google Scholar 

6.
Tremblay, J. -É & Gagnon, J. The effects of irradiance and nutrient supply on the productivity of Arctic waters: a perspective on climate change. In Influence of Climate Change on the Changing Arctic and Sub-Arctic Conditions (eds Nihoul, J. C. J. & Kostianoy, A. G.) 73–89 (Springer, Berlin, 2009).
Google Scholar 

7.
De Jong, J. T. M. et al. Sources and fluxes of dissolved iron in the Bellingshausen Sea (West Antarctica): the importance of sea ice, icebergs and the continental margin. Mar. Chem. 177, 518–535 (2015).
Google Scholar 

8.
Popova, E. E. et al. What controls primary production in the Arctic Ocean? Results from an intercomparison of five general circulation models with biogeochemistry. J. Geophys. Res. Ocean. 117, (2012).

9.
Nielsdóttir, M. C., Moore, C. M., Sanders, R., Hinz, D. J. & Achterberg, E. P. Iron limitation of the postbloom phytoplankton communities in the Iceland Basin. Glob. Biogeochem. Cycles 23, 1–13 (2009).
Google Scholar 

10.
Ryan-Keogh, T. J. et al. Spatial and temporal development of phytoplankton iron stress in relation to bloom dynamics in the high-latitude North Atlantic Ocean. Limnol. Oceanogr. 58, 533–545 (2013).
ADS  Google Scholar 

11.
Taylor, R. L. et al. Colimitation by light, nitrate, and iron in the Beaufort Sea in late summer. J. Geophys. Res. Ocean. 118, 3260–3277 (2013).
ADS  Google Scholar 

12.
Findlay, H. S. et al. Late winter biogeochemical conditions under sea ice in the Canadian High Arctic. Polar Res. 34, 24170 (2015).
Google Scholar 

13.
Mills, M. M. et al. Nitrogen limitation of the summer phytoplankton and heterotrophic prokaryote communities in the Chukchi Sea. Front. Mar. Sci. 5, 1–22 (2018).
ADS  Google Scholar 

14.
Hopwood, M. J. et al. Non-linear response of summertime marine productivity to increased meltwater discharge around Greenland. Nat. Commun. 9, 1–9 (2018).
CAS  Google Scholar 

15.
Codispoti, L. A. et al. Synthesis of primary production in the Arctic Ocean: III. Nitrate and phosphate based estimates of net community production. Prog. Oceanogr. 110, 126–150 (2013).
ADS  Google Scholar 

16.
Schauer, U. et al. Variation of measured heat flow through the Fram Strait between 1997 and 2006. In Arctic-Subarctic Ocean Fluxes: Defining the Role of the Northern Seas in Climate (eds Dickson, R. R. et al.) 65–85 (Springer, Berlin, 2008).
Google Scholar 

17.
Mouginot, J. et al. Fast retreat of Zachariæ Isstrøm, northeast Greenland. Science (80-, ) 350, 1357–1361 (2015).
ADS  CAS  Google Scholar 

18.
Smedsrud, L. H., Halvorsen, M. H., Stroeve, J. C., Zhang, R. & Kloster, K. Fram Strait sea ice export variability and September Arctic sea ice extent over the last 80 years. Cryosphere 11, 65–79 (2017).
ADS  Google Scholar 

19.
Rijkenberg, M. J. A., Slagter, H. A., Rutgers Van Der Loeff, M., Ooijen, J. Van. & Gerringa, L. J. A. Dissolved Fe in the deep and upper Arctic Ocean with a focus on Fe limitation in the Nansen Basin. Front. Mar. Sci. 5, 88 (2018).
Google Scholar 

20.
de Steur, L. et al. Freshwater fluxes in the East Greenland Current: a decade of observations. Geophys. Res. Lett. 36, 1–5 (2009).
Google Scholar 

21.
Beszczynska-Möller, A., Woodgate, R. A., Lee, C., Melling, H. & Karcher, M. A synthesis of exchanges through the main oceanic gateways to the Arctic Ocean. Oceanography 24, 82–99 (2011).
Google Scholar 

22.
Rudels, B. Arctic ocean circulation. in Encyclopedia of Ocean Sciences 262–277 (Elsevier Inc., 2019). https://doi.org/10.1016/B978-0-12-409548-9.11209-6

23.
Laukert, G. et al. Ocean circulation and freshwater pathways in the Arctic Mediterranean based on a combined Nd isotope, REE and oxygen isotope section across Fram Strait. Geochim. Cosmochim. Acta 202, 285–309 (2017).
ADS  CAS  Google Scholar 

24.
Michel, C. et al. Arctic Ocean outflow shelves in the changing Arctic: a review and perspectives. Prog. Oceanogr. 139, 66–88 (2015).
ADS  Google Scholar 

25.
Rudels, B. et al. The interaction between waters from the Arctic Ocean and the Nordic Seas north of Fram Strait and along the East Greenland Current: Results from the Arctic Ocean-02 Oden expedition. J. Mar. Syst. 55, 1–30 (2005).
Google Scholar 

26.
Lalande, C. et al. Lateral supply and downward export of particulate matter from upper waters to the seafloor in the deep eastern Fram Strait. Deep. Res. Part I Oceanogr. Res. Pap. 114, 78–89 (2016).
ADS  Google Scholar 

27.
Norwegian Polar Institute. Sea ice extent in the Fram Strait in September, 1979–2018. Environmental Monitoring of Svalbard and Jan Mayen (MOSJ) (2020). https://www.mosj.no/en/climate/ocean/sea-ice-extent-barents-sea-fram-strait.html. Accessed 10th January 2020.

28.
de Steur, L., Peralta-Ferriz, C. & Pavlova, O. Freshwater export in the East Greenland current freshens the North Atlantic. Geophys. Res. Lett. 45, 13359–13366 (2018).
ADS  Google Scholar 

29.
Marnela, M., Rudels, B., Houssais, M.-N., Beszczynska-Möller, A. & Eriksson, P. B. Recirculation in the Fram Strait and transports of water in and north of the Fram Strait derived from CTD data. Ocean Sci. 9, 499–519 (2013).
ADS  Google Scholar 

30.
Beszczynska-Möller, A., Fahrbach, E., Schauer, U. & Hansen, E. Variability in Atlantic water temperature and transport at the entrance to the Arctic Ocean, 1997–2010. ICES J. Mar. Sci. 69, 852–863 (2012).
Google Scholar 

31.
de Steur, L., Hansen, E., Mauritzen, C., Beszczynska-Möller, A. & Fahrbach, E. Impact of recirculation on the East Greenland Current in Fram Strait: Results from moored current meter measurements between 1997 and 2009. Deep. Res. Part I Oceanogr. Res. Pap. 92, 26–40 (2014).
ADS  Google Scholar 

32.
Grasshoff, K., Kremlingl, K. & Ehrhardt, M. Methods of Seawater Analysis (Wiley, Hoboken, 1999). https://doi.org/10.1002/9783527613984.
Google Scholar 

33.
Cutter, G. et al. Sampling and Sample-handling Protocols for GEOTRACES Cruises. (2014).

34.
Rapp, I., Schlosser, C., Rusiecka, D., Gledhill, M. & Achterberg, E. P. Automated preconcentration of Fe, Zn, Cu, Ni, Cd, Pb Co, and Mn in seawater with analysis using high-resolution sector field inductively-coupled plasma mass spectrometry. Anal. Chim. Acta 976, 1–13 (2017).
CAS  PubMed  Google Scholar 

35.
Moore, C. M. Diagnosing oceanic nutrient deficiency. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 374, 20150290 (2016).
ADS  Google Scholar 

36.
Browning, T. J. et al. Nutrient co-limitation at the boundary of an oceanic gyre. Nature 551, 242–246 (2017).
ADS  CAS  PubMed  Google Scholar 

37.
Achterberg, E. P. et al. Natural iron fertilization by the Eyjafjallajökull volcanic eruption. Geophys. Res. Lett. 40, 921–926 (2013).
ADS  CAS  Google Scholar 

38.
Welschmeyer, N. A. Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and pheopigments. Limnol. Oceanogr. 39, 1985–1992 (1994).
ADS  CAS  Google Scholar 

39.
Van Heukelem, L. & Thomas, C. S. Computer-assisted high-performance liquid chromatography method development with applications to the isolation and analysis of phytoplankton pigments. J. Chromatogr. A 910, 31–49 (2001).
PubMed  Google Scholar 

40.
Mackey, M. D., Mackey, D. J., Higgins, H. W. & Wright, S. W. CHEMTAX – a program for estimating class abundances from chemical markers: application to HPLC measurements of phytoplankton. Mar. Ecol. Prog. Ser. 144, 265–283 (1996).
ADS  CAS  Google Scholar 

41.
Coupel, P. et al. Pigment signatures of phytoplankton communities in the Beaufort Sea. Biogeosciences 12, 991–1006 (2015).
ADS  Google Scholar 

42.
Hong, C. N. et al. Sediment efflux of silicon on the Greenland margin and implications for the marine silicon cycle. Earth Planet. Sci. Lett. 529, 115877 (2020).
Google Scholar 

43.
Tonnard, M. et al. Dissolved iron in the North Atlantic Ocean and Labrador Sea along the GEOVIDE section (GEOTRACES section GA01). Biogeosciences 17, 917–943 (2020).
ADS  Google Scholar 

44.
Colombo, M., Jackson, S. L., Cullen, J. T. & Orians, K. J. Dissolved iron and manganese in the Canadian Arctic Ocean: On the biogeochemical processes controlling their distributions. Geochim. Cosmochim. Acta (2020). (in press).

45.
Ardiningsih, I. et al. Natural Fe-binding organic ligands in Fram Strait and over the Northeast Greenland shelf. Mar. Chem. 224, (2020).

46.
Le Moigne, F. A. C. et al. Sequestration efficiency in the iron-limited North Atlantic: implications for iron supply mode to fertilized blooms. Geophys. Res. Lett. 41, 4619–4627 (2014).
ADS  Google Scholar 

47.
Beszczynska-Möller, A. & Wisotzki, A. Physical oceanography during POLARSTERN cruise ARK-XXIII/2. Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, PANGAEA (2010). https://doi.org/10.1594/PANGAEA.733424.

48.
Kattner, G. & Ludwichowski, K.-U. Inorganic nutrients measured on water bottle samples during POLARSTERN cruise ARK-XXIII/2. Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, PANGAEA (2014). https://doi.org/10.1594/PANGAEA.832402.

49.
Randelhoff, A., Fer, I., Sundfjord, A., Tremblay, J. -É & Reigstad, M. Vertical fluxes of nitrate in the seasonal nitracline of the Atlantic sector of the Arctic Ocean. J. Geophys. Res. Ocean. 121, 3010–3028 (2016).
Google Scholar 

50.
Achterberg, E. P. et al. Iron biogeochemistry in the high latitude North Atlantic ocean. Sci. Rep. 8, 1–15 (2018).
ADS  CAS  Google Scholar 

51.
Torres-Valdés, S. et al. Export of nutrients from the Arctic Ocean. J. Geophys. Res. 118, 1625–1644 (2013).
ADS  Google Scholar 

52.
Ooijen, J. Van, Rijkenberg, M. J. A., Gerringa, L. J. A., Rabe, B. & van der Loeff, M. R. Inorganic nutrients measured on water bottle samples during POLARSTERN cruise PS94 (ARK-XXIX/3). (2016). https://doi.org/10.1594/PANGAEA.868396

53.
Slagter, H. A. et al. Organic Fe speciation in the Eurasian Basins of the Arctic Ocean and its relation to terrestrial DOM. Mar. Chem. 197, 11–25 (2017).
CAS  Google Scholar 

54.
Gerringa, L. J. A., Rijkenberg, M. J. A. & Slagter, H. A. Dissolved iron measured on board with Flow injection analysis and iron-binding dissolved organic ligands from Ultra Clean CTD collected depth profiles during GEOTRACES PS94 Arctic cruise on Polarstern. (2018). https://doi.org/10.1594/PANGAEA.890975

55.
Krawczyk, D. W. et al. Seasonal succession, distribution, and diversity of planktonic protists in relation to hydrography of the Godthåbsfjord system (SW Greenland). Polar Biol. 41, 2033–2052 (2018).
Google Scholar 

56.
Meire, L. et al. High export of dissolved silica from the Greenland Ice Sheet. Geophys. Res. Lett. 43, 9173–9182 (2016).
ADS  CAS  Google Scholar 

57.
Nöthig, E.-M. et al. Summertime plankton ecology in Fram Strait: a compilation of long- and short-term observations. Polar Res. 34, 1–18 (2015).
Google Scholar 

58.
Moore, C. M. et al. Relative influence of nitrogen and phosphorus availability on phytoplankton physiology and productivity in the oligotrophic sub-tropical North Atlantic Ocean. Limnol. Oceanogr. 53, 291–305 (2008).
ADS  CAS  Google Scholar 

59.
Browning, T. J. et al. Nutrient co-limitation at the boundary of an oceanic gyre_Supplementary Material. Nature 551, 242–246 (2017).
ADS  CAS  PubMed  Google Scholar 

60.
Kattner, G. & Budéus, G. Nutrient status of the Northeast Water Polynya. J. Mar. Syst. 10, 185–197 (1997).
Google Scholar 

61.
Smith, R. E. H., Gosselin, M. & Taguchi, S. The influence of major inorganic nutrients on the growth and physiology of high arctic ice algae. J. Mar. Syst. 11, 63–70 (1997).
Google Scholar 

62.
Maestrini, S. Y., Rochet, M., Legendre, L. & Demers, S. Nutrient limitation of the bottom-ice microalgal biomass (southeastern Hudson Bay, Canadian Arctic). Lim 31, 969–982 (1986).
ADS  CAS  Google Scholar 

63.
Ortega-Retuerta, E., Jeffrey, W. H., Ghiglione, J. F. & Joux, F. Evidence of heterotrophic prokaryotic activity limitation by nitrogen in the Western Arctic Ocean during summer. Polar Biol. 35, 785–794 (2012).
Google Scholar 

64.
Mann, E. L. & Chisholm, S. W. Iron limits the cell division rate of Prochlorococcus in the eastern equatorial Pacific. Limnol. Oceanogr. 45, 1067–1076 (2000).
ADS  CAS  Google Scholar 

65.
Vernet, M. et al. Influence of phytoplankton advection on the productivity along the Atlantic water inflow to the Arctic Ocean. Front. Mar. Sci. 6(583), 1–18 (2019).
Google Scholar 

66.
Moore, C. M. et al. Iron limits primary productivity during spring bloom development in the central North Atlantic. Glob. Chang. Biol. 12, 626–634 (2006).
ADS  Google Scholar 

67.
Browning, T. J. et al. Nutrient regulation of late spring phytoplankton blooms in the midlatitude North Atlantic. Limnol. Oceanogr. 9999, 1–13 (2019).
Google Scholar 

68.
Blain, S. et al. Availability of iron and major nutrients for phytoplankton in the northeast Atlantic Ocean. Limnol. Oceanogr. 49, 2095–2104 (2004).
ADS  CAS  Google Scholar 

69.
Boyd, P. W. & Ellwood, M. J. The biogeochemical cycle of iron in the ocean. Nat. Geosci. 3, 675–682 (2010).
ADS  CAS  Google Scholar 

70.
Twining, B. S. & Baines, S. B. The trace metal composition of marine phytoplankton. Ann. Rev. Mar. Sci. 5, 191–215 (2013).
PubMed  Google Scholar 

71.
Saito, M. A. et al. Multiple nutrient stresses at intersecting Pacific Ocean biomes detected by protein biomarkers. Science (80-. ) 345, 1173–1177 (2014).
ADS  CAS  Google Scholar 

72.
Ward, B. A., Dutkiewicz, S., Moore, C. M. & Follows, M. J. Iron, phosphorus, and nitrogen supply ratios define the biogeography of nitrogen fixation. Limnol. Oceanogr. 58, 2059–2075 (2013).
ADS  CAS  Google Scholar 

73.
Hattermann, T., Isachsen, P. E., Von Appen, W. J., Albretsen, J. & Sundfjord, A. Eddy-driven recirculation of Atlantic Water in Fram Strait. Geophys. Res. Lett. 43, 3406–3414 (2016).
ADS  Google Scholar 

74.
Klunder, M. B. et al. Dissolved iron in the Arctic shelf seas and surface waters of the central Arctic Ocean: impact of Arctic river water and ice-melt. J. Geophys. Res. Ocean. 117, 1–18 (2012).
Google Scholar 

75.
Charette, M. A. et al. The Transpolar Drift as a Source of Riverine and Shelf-Derived Trace Elements to the Central Arctic Ocean. J. Geophys. Res. Ocean. 125, e2019JC015920 (2020).
ADS  Google Scholar 

76.
Yamamoto-Kawai, M., Carmack, E. & McLaughlin, F. Nitrogen balance and Arctic throughflow. Nature 443, 43 (2006).
ADS  CAS  PubMed  Google Scholar 

77.
Rijkenberg, M. J. A. et al. The distribution of dissolved iron in the West Atlantic Ocean. PLoS ONE 9, 1–14 (2014).
Google Scholar 

78.
de Boyer Montégut, C., Madec, G., Fischer, A. S., Lazar, A. & Iudicone, D. Mixed layer depth over the global ocean: an examination of profile data and a profile-based climatology. J. Geophys. Res. Ocean. 109, 1–20 (2004).
Google Scholar 

79.
Randelhoff, A., Sundfjord, A. & Reigstad, M. Seasonal variability and fluxes of nitrate in the surface waters over the Arctic shelf slope. Geophys. Res. Lett. 42, 3442–3449 (2015).
ADS  CAS  Google Scholar 

80.
Randelhoff, A. et al. Pan-Arctic ocean primary production constrained by turbulent nitrate fluxes. Front. Mar. Sci. 7, 1–15 (2020).
Google Scholar 

81.
Rafter, P. A., Sigman, D. M. & Mackey, K. R. M. Recycled iron fuels new production in the eastern equatorial Pacific Ocean. Nat. Commun. 8, 1100 (2017).
ADS  PubMed  PubMed Central  Google Scholar 

82.
Stohl, A. et al. Arctic smoke: record high air pollution levels in the European Arctic due to agricultural fires in Eastern Europe in spring 2006. Atmos. Chem. Phys. 7, 511–534 (2007).
ADS  CAS  Google Scholar 

83.
Marsay, C. M. et al. Concentrations, provenance and flux of aerosol trace elements during US GEOTRACES Western Arctic cruise GN01. Chem. Geol. 502, 1–14 (2018).
ADS  CAS  Google Scholar 

84.
Conca, E. et al. Source identification and temporal evolution of trace elements in PM10 collected near to Ny-Ålesund (Norwegian Arctic). Atmos. Environ. 203, 153–165 (2019).
ADS  CAS  Google Scholar 

85.
Kadko, D. et al. The residence times of trace elements determined in the surface Arctic Ocean during the 2015 US Arctic GEOTRACES expedition. Mar. Chem. 208, 56–69 (2019).
CAS  Google Scholar 

86.
Wehrmann, L. M. et al. Iron and manganese speciation and cycling in glacially influenced high-latitude fjord sediments (West Spitsbergen, Svalbard): evidence for a benthic recycling-transport mechanism. Geochim. Cosmochim. Acta 141, 628–655 (2014).
ADS  CAS  Google Scholar 

87.
Bown, J. et al. Evidences of strong sources of DFe and DMn in Ryder Bay, Western Antarctic Peninsula. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 376, 20170172 (2018).
ADS  Google Scholar 

88.
Bucciarelli, E., Blain, S. & Tréguer, P. Iron and manganese in the wake of the Kerguelen Islands (Southern Ocean). Mar. Chem. 73, 21–36 (2001).
CAS  Google Scholar 

89.
Nishino, S. et al. Enhancement/Reduction of biological pump depends on ocean circulation in the sea-ice reduction regions of the Arctic Ocean. J. Oceanogr. 67, 305–314 (2011).
CAS  Google Scholar 

90.
Kipp, L. E., Charette, M. A., Moore, W. S., Henderson, P. B. & Rigor, I. G. Increased fluxes of shelf-derived materials to the central arctic ocean. Sci. Adv. 4, 1–10 (2018).
Google Scholar 

91.
Mayot, N. et al. Springtime export of Arctic Sea ice influences phytoplankton production in the Greenland Sea. J. Geophys. Res. Ocean. 125, 1–16 (2020).
Google Scholar 

92.
Randelhoff, A. et al. The evolution of light and vertical mixing across a phytoplankton ice-edge bloom. Elem. Sci. Anthr. 7, 1–19 (2019).
Google Scholar 

93.
Kahru, M., Brotas, V., Manzano-Sarabia, M. & Mitchell, B. G. Are phytoplankton blooms occurring earlier in the Arctic?. Glob. Chang. Biol. 17, 1733–1739 (2011).
ADS  Google Scholar 

94.
Ardyna, M. et al. Environmental drivers of under-ice phytoplankton bloom dynamics in the Arctic Ocean. Elem. Sci. Anthr. 8, 30 (2020).
Google Scholar 

95.
Tremblay, J. É et al. Vertical stability and the annual dynamics of nutrients and chlorophyll fluorescence in the coastal, southeast Beaufort Sea. J. Geophys. Res. Ocean. 113, 1–14 (2008).
Google Scholar 

96.
Castro de la Guardia, L. et al. Assessing the role of high-frequency winds and sea ice loss on arctic phytoplankton blooms in an ice-ocean-biogeochemical model. J. Geophys. Res. Biogeosci. 124, 2728–2750 (2019).
ADS  Google Scholar 

97.
Ardyna, M. et al. Recent Arctic Ocean sea ice loss triggers novel fall phytoplankton blooms. Geophys. Res. Lett. 41, 6207–6212 (2014).
ADS  Google Scholar 

98.
Wang, Q. et al. Intensification of the Atlantic Water supply to the Arctic Ocean through Fram Strait induced by Arctic sea ice decline. Geophys. Res. Lett. 47, e2019GL086682 (2020).
ADS  Google Scholar 

99.
Arrigo, K. R., van Dijken, G. & Pabi, S. Impact of a shrinking Arctic ice cover on marine primary production. Geophys. Res. Lett. 35, 1–6 (2008).
Google Scholar 

100.
Carmack, E. C., Macdonald, R. W. & Steve, J. Phytoplankton productivity on the Canadian Shelf of the Beaufort Sea. Mar. Ecol. Prog. Ser. 277, 37–50 (2004).
ADS  Google Scholar 

101.
Lasternas, S. & Agustí, S. Phytoplankton community structure during the record Arctic ice-melting of summer 2007. Polar Biol. 33, 1709–1717 (2010).
Google Scholar 

102.
Harding, K. et al. Symbiotic unicellular cyanobacteria fix nitrogen in the Arctic Ocean. Proc. Natl. Acad. Sci. 115, 13371–13375 (2018).
CAS  PubMed  Google Scholar 

103.
Sipler, R. E. et al. Preliminary estimates of the contribution of Arctic nitrogen fixation to the global nitrogen budget. Limnol. Oceanogr. 2, 159–166 (2017).
Google Scholar 

104.
Zehr, J. P. & Kudela, R. M. Nitrogen cycle of the open ocean: from genes to ecosystems. Annu. Rev. Mar. Sci. 3, 197–225 (2011).
ADS  Google Scholar 

105.
Acker, J. G. & Leptoukh, G. Online analysis enhances use of NASA Earth Science Data. Eos (Washington, DC) 88, 14–17 (2007).
ADS  Google Scholar 

106.
Schaffer, J. et al. A global, high-resolution data set of ice sheet topography, cavity geometry, and ocean bathymetry. Earth Syst. Sci. Data 8, 543–557 (2016).
ADS  Google Scholar  More

Field experiment
Estimation of turtle catch per unit effort
We conducted our field experiment in February–April 2018 at two wetland complexes near Murray Bridge, South Australia, selecting two study sites at each complex (Supplementary Fig. S3). At each site, we estimated turtle population density using catch-per-unit-effort (CPUE; Supplementary Table S1). We conducted three 3-day rounds of turtle trapping using a combination of fyke and cathedral traps, baited with offal. Up to eight traps were deployed at a time. We calculated turtle CPUE by dividing the total number of turtles caught (regardless of species) by the total trap-hours. The number of trap-hours was similar across all four sites (average 1685 ± 7.6 SE total trap-hours).
Carp carcass decomposition
After the first and the second trapping rounds, we deployed whole carp carcasses at each site to measure carp decomposition rates depending on turtle accessibility. We placed each carp in a pre-weighed plastic box (340 × 230 × 120 mm), securing it with cable ties. Carp were made non-accessible to turtles in half of the deployments by covering the plastic boxes with 25 × 25 mm mesh (Supplementary Fig. S4). The mesh prevented turtle access to the carp, but was large enough to allow scavenging by crayfish (Cherax destructor) and other freshwater invertebrates. We tied each box to a brick and submerged the boxes around the four study sites ≥ 30 m away from each other, sunk at an average depth of 436 mm (± 13 SE). We used a total of 38 accessible and 40 non-accessible carp, split between our four study sites over two rounds (Supplementary Table S9). Every day, starting from day 2, the box and carp were weighed together with a digital scale. In all measurements, we calculated the wet mass of the carp by subtracting the box weight from the total weight. Carp carcasses were left in the wetlands for up to 10 days, or until they were fully consumed. All work was performed in accordance with DEWNR Permit M26663-1, PIRSA permits MP0085 and ME9902980, and The University of Sydney Animal Ethics Committee approval (project number 2017/1208), observing all relevant guidelines and regulations.
Statistical analysis
We analysed our data using RStudio 1.1.45633 (packages: “lme4” 1.1-2134, “MuMIn” 1.42.135). To assess whether turtles were important scavengers of our carcasses, we computed a linear mixed model testing whether turtle CPUE and carp access (yes/no) affected the rate of mass loss of the carp carcasses. The turtle CPUE values used were the average CPUE in the trapping round before and after each carp was deployed. We used the rate of mass loss per day as a dependent variable. We included the carp mass before deployment as an independent variable to account for initial mass variation, and we included study site as a random variable. We log-transformed all data before analysis. We assessed model fit by examining predicted versus residual and Q–Q plots, and testing the normality of residuals.
Mesocosm experiment
Turtle trapping and experiment procedure
We caught 20 adult male E. macquarii with fyke nets baited with offal at Hawkesview Lagoon, Albury, NSW, in November 2018. The E. macquarii captured at this site belong to the same genetic population as the E. macquarii trapped in South Australia36, therefore we expect behaviours to be similar between the two populations. We focussed on E. macquarii as this is the most common species in the Murray–Darling Basin, and fish carrion is an important part of its diet19,37. The turtles were transported by car to the Experimental Wetlands facility at Western Sydney University, in Richmond, NSW (Supplementary Fig. S5). This facility is comprised of 10 circular mesocosms (0.42 m depth × 2.1 m diameter) filled with 1,450 L of tap water. Each is an independent flow-through system where the water flow is regulated, and was maintained at 1998.6 ml/min (± 149.5 SE) throughout the experiment. Each mesocosm had two cement blocks for the turtles to bask on, and two plastic tunnels for shelter. The experiment was conducted for 40 days, therefore it is a short-term study (Supplementary Fig. S6). Upon arrival at the facility, we placed four adult male E. macquarii turtles in each of five random mesocosms, which means the experimental replication was 5. The remaining five mesocosms were controls and had no turtles. The four turtles comprised an average 5,376.6 g total biomass per pond, each being 3.46 m2. This would result in a biomass of 11,560 individuals/ha or 15,537 kg/ha on average. Kinosternon integrum has been estimated reaching densities of 20,000 individuals/ha in Sonora, Mexico, while Podocnemis vogli may reach 10,300 individuals/ha or 15,450 kg/ha in Venezuela, likely in temporary aggregations38. Emydura macquarii tend to congregate around food sources, therefore we considered four turtles per carp carrion as a realistic density. After 7 days of acclimation, we introduced one carp carcass to all mesocosms, and a second 6 days later. We used one ~ 1 kg carp at a time to simulate a density close to 3,144 kg carp/ha31. The turtles had continual access to the carp, which was their main food source throughout the experiment. The day all carp carcasses were fully eaten in all turtle mesocosms, we removed turtles from their mesocosms and released them at the point of capture. On the same day (day 10), we ended the data collection in their mesocosms, because any further change in water quality here would not have been related to carp decomposition. We continued the daily water quality measurements in the five control mesocosms until all carp were fully decomposed (day 32). This experimental design allowed us to collect water quality data without the need to add turtle food to the mesocosms, which would have biased our measurements once carp were removed from the turtle mesocosms.
We measured water temperature, dissolved oxygen, conductivity, turbidity, phosphate, and ammonia concentration in all mesocosms every morning from the day before the first carp introduction (see Supplementary Materials for equipment used). We also photographed the carcasses daily to estimate their decomposition rate based on a scale (Supplementary Table S10) designed after the decomposition stages described by Benninger et al.39 Due to the short transit time of fish matter in E. macquarii’s gut37, the effects of the turtles’ metabolic wastes on water quality are included in our experiment for carp 1. All work was performed in accordance with OEH Permit SL100401, DPI permit P09/0070-3.0, and Western Sydney University Animal Ethics Committee Animal Research Authority approval A12390, observing all relevant guidelines and regulations.
Statistical analysis
To assess whether the presence of turtles affected the decomposition of carp we computed a mixed linear model using the repeated measures PROC MIXED procedure using SAS (3.8 University Edition, SAS Institute Inc., Cary, NC, USA). For this model, days to total decomposition/removal was a dependent variable, turtle presence/absence was a fixed effect, and carp number (first or second) was a repeated fixed effect.
We used DO, conductivity, turbidity, phosphate, and ammonia to carry out a principal component analysis (PCA) using PROC PRINCOMP. We conducted a PCA because the parameters are a multivariate response and have potential to covary with each other, which would not be detected in univariate analyses. We considered a parameter loaded onto a PC when the absolute value of its eigenvector was > 0.300. If the same parameter loaded onto more than one PC, we considered it only on the PC where its eigenvector had a higher absolute value.
To test the effect of turtles scavenging on water quality, we computed general linear mixed models (GLMMs), using PCs as response variables, in PROC MIXED. We used a PC as response variable if its eigenvalue was greater than one (Kaiser criterion40). For each of these GLMMs, we included turtle presence (yes/no) and day number after the first carp introduction as fixed effects, water temperature and flow as covariates, and mesocosm ID as a random effect. We computed a model with full interactions first, and then, in absence of four- or three-way interactions, simplified the model to focus on main effects.
Finally, to test the effect of turtle scavenging on each water quality parameter, a GLMM was computed for each (logged) parameter that loaded onto a PC with eigenvalue > 1, i.e. dissolved oxygen, ammonia, turbidity, conductivity, phosphate. For these GLMMs, turtle presence and day number were fixed effects, water temperature and flow were covariates, and mesocosm ID was a random effect. More

Animals and housing conditions
The present study was conducted with 24 male C57BL/6J mice, purchased from a professional breeder (Charles River Laboratories, Research Models and Services, Germany GmbH, Sulzfeld, Germany) at the age of five weeks. Upon arrival, mice were housed in same-sex groups of 3 individuals per cage (Makrolon cages type III, 38 × 23 × 15 cm3), since in sub-adult male mice, the occurrence of escalated aggression is very unlikely. However, with the males becoming adult, the probability of escalated agonistic encounters increases. Therefore, at the age of nine weeks, mice were transferred to single housing conditions to avoid any escalated aggressive interactions. Please note that the question whether to house male laboratory mice singly or in groups is under ongoing discussion and there is still no “gold standard” regarding its solution. For current discussions about recommendations for male mouse housing see37,38. Cages were equipped with wood chips as bedding material (TierWohl Super, J. Rettenmaier & Söhne GmbH + Co.KG, Rosenberg, Germany), a wooden stick, a semi-transparent red plastic house (11.1 × 11.1 × 5.5 cm3, Tecniplast Deutschland GmbH, Hohenpeißenberg, Germany), and a paper tissue. Housing rooms were maintained at a reversed 12 h dark/light cycle with lights off at 8 a.m., a temperature of approximately 23 °C, and a relative humidity of about 50%. The animals had ad libitum access to water and food (Altromin 1324, Altromin Spezialfutter GmbH & Co. KG, Lage, Germany) until the beginning of the touchscreen training phase. From then on they were mildly food restricted to 90–95% of their ad libitum feeding weights in order to enhance their motivation to work for food rewards. As neither distinct negative effects of such a restricted feeding protocol39, nor an interference with judgement bias assessment17,18 could be detected in previous studies, we considered this method to not affect the emotional state of the mice itself. Weights were monitored on a daily basis using a digital scale (weighing capacity: 150 g, resolution: 0.1 g; CM 150-1 N, Kern, Ballingen, Germany).
In addition to the experimental animals, 16 group-housed adult female C57BL/6J mice and 5 single-housed adult male NMRI mice, purchased from Charles River Laboratories, were used to provide the test animals with social experiences.
Ethics statement
All procedures complied with the regulations covering animal experimentation within Germany (Animal Welfare Act), the EU (European Communities Council DIRECTIVE 2010/63/EU), and the fundamental principles of the Basel Declaration, and were approved by the local (Gesundheits- und Veterinäramt Münster, Nordrhein-Westfalen) and federal authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen “LANUV NRW”, reference number 84-02.04.2015.A441).
Experimental design
In this study, the effects of different social experiences on important correlates of animal emotions, comprising cognitive (judgement bias), behavioural (anxiety-like and exploratory behaviour) as well as physiological (stress hormone levels) measures, were investigated. The experiment comprised six phases: a handling phase, a training phase, a first cognitive judgement bias (CJB) test phase, an experience phase, a second CJB test phase, and a behavioural test phase (Fig. 1).
Figure 1

Experimental design. Mice were habituated to cup handling before they underwent daily training sessions until successful acquisition of the discrimination task. Afterwards, they were tested in the cognitive judgement bias (CJB) test. During the following phase, mice repeatedly made one out of three different experiences: mildly “adverse”, “beneficial”, or “neutral”. They were then tested for their CJB again. During this second test phase, a so-called reminder was presented before each test session with the aim to re-evoke the affective state the animals experienced during the treatment phase. On the last day of each CJB test phase, faecal corticosterone metabolite concentrations (FCMs) were assessed. Subsequently, animals were tested for anxiety-like behaviour. Again, they were presented with reminders before each behavioural test.

Full size image

During the handling phase starting at PND 69, mice were first habituated to cup handling for 5 days and thereafter underwent daily training sessions to learn the discrimination task required for CJB testing, starting at PND 76. Afterwards, the animals’ initial CJB was assessed (start test phase 1: PND 223 ± 77; for details on CJB training and testing see following section).
During a subsequent experience phase starting at PND 230 ± 77, mice were exposed to one of three different experiences, each comprising three group-specific encounters, classified as either mildly “adverse”, “beneficial”, or “neutral”. Encounters took place under red light between 2:45 p.m. and 4:35 p.m. on 3 different days, always separated by a gap day. The mildly “adverse experience” group (AE group, n = 8) repeatedly encountered a dominant opponent of the aggressive NMRI strain40, with each confrontation lasting maximally 10 minutes30,41. Confrontations were terminated in cases of high aggression. The “beneficial experience” group (BE group, n = 8) was repeatedly presented with freshly collected urine of an unfamiliar C57BL/6J female for 10 minutes31. To provide all subjects with comparable experiences, we controlled for the females’ oestrus state. Since the time of oestrus in mice is relatively short42, urine from non-oestrous females was used in order to keep the total number of involved females low. The “neutral experience” group (NE group, n = 8) served as a control group and was repeatedly placed into a novel cage containing clean bedding material for 10 min.
Following the experience phase, CJB was assessed again to investigate the influence of the respective experience on the animals’ judgement bias (start test phase 2: PND 237 ± 77). In this second test phase, a so-called reminder was presented immediately before each test session. These reminders were introduced to acutely re-evoke the affective state the mice experienced during the encounters of the treatment phase. Reminders took place immediately before each test session of the second CJB test phase. For this purpose, mice were placed into a cage (Makrolon type II cage; 22 × 16 × 14 cm3) filled with bedding for 3 min. For AE mice, another 25 ml of soiled bedding from the home cage of the last NMRI male encountered were added. For BE mice, the same was done with soiled bedding from the home cage of the last female of which urine had been presented.
On the last day of each CJB test phase, faeces samples were obtained to assess corticosterone metabolite (FCM) concentrations. Finally, animals underwent a battery of standard behavioural tests for anxiety-like behaviour and exploratory locomotion (elevated plus maze test (EPM), dark-light test (DL), and open field test (OF); start: PND 245 ± 77). Before each test session, a reminder was presented again.
The allocation of mice to the treatment groups was pseudo-randomised, so that balanced numbers of mice with different learning speeds were present in each group. The testing order of mice was randomised once before the first CJB test and subsequently maintained for the following CJB and behavioural test sessions. As reminders were provided immediately before CJB testing as well as before the subsequent behavioural tests, blinding of the experimenter was not possible.
The touchscreen-based cognitive judgement bias test
Procedure
The same apparatus as described previously was used28,36 (Bussey-Saksida Mouse Touch Screen Chambers, Model 80614, Campden Instruments Ltd., Loughborough, Leics., UK). Mice underwent daily touchscreen sessions at intervals of approximately 24 h on maximally 6 consecutive days. Before each session, each mouse was taken out of its home cage and weighed. In a red semi-transparent box (21 × 21 × 15 cm3) the animal was then transported to a separate room where it was placed into the touchscreen chamber. The session was started and ended after a maximum number of trials had been performed or after a training step-specific duration. All touchscreen sessions were conducted during the dark phase between 8.15 a.m. and 1 p.m.
Paradigm
The paradigm applied here was the same as described previously with minor modifications36. Briefly, mice were trained to distinguish between a positive and a negative condition (Fig. 2). The positive condition was signalled by a bar at the bottom (5 cm below upper edge) of the cue-presentation field, the negative condition by a bar at the top (1 cm below upper edge). Mice had to touch either the left or right touch field in response to the cues. A correct touch in the positive condition led to the delivery of a large reward (12 μl of sweet condensed milk, diluted 1:4 in tap water, in the following “SCM”). An incorrect touch resulted in the delivery of a small reward (4 μl of SCM). In the negative condition, correct touches led to the delivery of a small reward (4 μl of SCM), while incorrect touches resulted in a mild “punishment” (5 s time out and houselight on). Mice had to learn to touch the high-rewarded side in the positive condition and the small-rewarded side in the negative condition. The small-rewarded touch field was the same in both conditions. The association between condition and correct touch side was the same for each individual but counterbalanced between mice. For a detailed description of the training procedure please see the supplementary material. After successful training, animals underwent CJB testing. The two cognitive bias test phases took place on five consecutive days each. During each CJB test session, three types of ambiguous cues, interspersed with the learned reference cues, were presented. These were bars at three intermediate positions: near positive (NP, 4 cm below upper edge), middle (M, 3 cm below upper edge) and near negative (NN, 2 cm below upper edge). Touches in response to these ambiguous cues resulted in a neutral outcome (neither a reward nor a “punishment”). The animals’ judgements made in response to these cues indicated whether they interpreted them according to the positive (“optimistic” response) or negative (“pessimistic” response) reference cue, serving as a measure of CJB.
Figure 2

Touchscreen-based cognitive bias paradigm. Mice were trained to distinguish between bars displayed at the top (negative condition) or bottom (positive condition) of a central field of a touchscreen. In this example, mice learned to touch right for a large reward during the positive condition and to touch left for a small reward during the negative condition (the association between positive/negative cue and the correct touch side was counterbalanced across mice). During the test, mice were presented with cues displayed at three intermediate positions: near positive, middle and near negative. The relative number of “optimistic”-like responses to these ambiguous conditions served as outcome measures of the animals’ cognitive judgement bias. Figure adopted from Krakenberg et al. (2019) with permission from Elsevier36.

Full size image

Each test session comprised 54 trials. Per session, each type of ambiguous cue was presented twice, interspersed with 48 training trials. Per test phase, each mouse was presented with each ambiguous cue ten times and each trained cue 120 times.
Behavioural measures
Responses to ambiguous cues served as a measure of the animals’ CJB. Touches according to the positive condition were defined as “optimistic” choices, touches according to the negative condition were defined as “pessimistic” choices. Out of all responses per condition, a “choice score” was calculated as previously28,36 according to the following formula:

$$ Choice,Score = frac{{N,choices ( {text{“}}optimistic{text{”}} ) – N,choices ({text{“}}pessimistic{text{”}})}}{ N,choices ({text{“}}optimistic{text{”}} + {text{“}}pessimistic{text{”}})} $$

The choice score could range between − 1 to + 1. Higher scores indicated a higher proportion of “optimistic” choices and consequently a relatively positive CJB compared to lower scores. Please note that choice scores are not an absolute, but a relative measure of CJB and that the term was chosen for the sake of intuitiveness.
Anxiety-like behaviour and exploratory locomotion
Mice were tested in three tests on anxiety-like behaviour and exploratory locomotion in the following order: the elevated plus-maze test (EPM), the dark-light test (DL) and the open field test (OF). The sequence of tests followed recommendations to schedule tests that are more sensitive to previous experience at the beginning of such a battery, and to conduct potentially more stressful tests towards the end43,44. Tests were carried out at intervals of at least 48 h and were performed in a room different from the housing room between 12:45 p.m. and 3:35 p.m. Test equipment was cleaned with 70% ethanol between subjects. Behaviour was recorded with a webcam (Logitech Webcam Pro 9000) and the animals’ movements during the EPM and OF were automatically analysed by the video tracking system ANY-maze (ANY-maze version 4.99, Stoelting Co., Wood Dale, IL, USA). Videos of the DL were analysed manually by an experienced observer (Sophie Siestrup). For apparatus descriptions and details about testing procedures see supplementary material.
Faecal corticosterone metabolites
The basal levels of adrenocortical activity of the subjects were monitored non-invasively by measuring faecal corticosterone metabolites45,46,47 (FCMs). Faeces samples of each individual were collected on the last day of the first CJB test week (= before the experience phase) and on the last day of the second CJB test week (= after the experience phase). During the dark phase, a peak of FCMs can be found in the faeces 4–6 h after the exposure to a stressor45. For this reason, faeces samples were collected 5.5–8.5 h after an individual finished CJB testing to ensure that faeces collection could be finished in the dark phase. For sample collection, mice were placed in Makrolon cages type III with a thin layer of bedding material and clean enrichment items as present in the home cage. Water was available ad libitum. After the sampling period of 3 h, mice were transferred to novel clean cages together with the enrichment items. All faeces produced during this time were collected and frozen at − 20 °C. Faecal samples were dried and homogenised and aliquots of 0.05 g were extracted with 1 ml of 80% methanol. Samples were then analysed using a 5α-pregnane-3β, 11β, 21-triol-20-one enzyme immunoassay (for details see45,46). Intra- and inter-assay coefficients of variation were More

1.
Steeman, M. E. et al. Radiation of extant cetaceans driven by restructuring of the oceans. Syst. Biol. 58, 573–585 (2009).
PubMed  PubMed Central  Google Scholar 
2.
Tolley, K. A. & Rosel, P. E. Population structure and historical demography of eastern North Atlantic harbour porpoises inferred through mtDNA sequences. Mar. Ecol. Prog. Ser. 327, 297–308 (2006).
ADS  CAS  Google Scholar 

3.
Banguera-Hinestroza, E., Bjørge, A., Reid, R. J., Jepson, P. & Hoelzel, A. R. The influence of glacial epochs and habitat dependence on the diversity and phylogeography of a coastal dolphin species: Lagenorhynchus albirostris. Conserv. Genet. 11, 1823–1836 (2010).
Google Scholar 

4.
Taguchi, M., Chivers, S. J., Rosel, P. E., Matsuishi, T. & Abe, S. Mitochondrial DNA phylogeography of the harbour porpoise Phocoena phocoena in the North Pacific. Mar. Biol. 157, 1489–1498 (2010).
CAS  Google Scholar 

5.
Amaral, A. R. et al. Influences of past climatic changes on historical population structure and demography of a cosmopolitan marine predator, the common dolphin (genus Delphinus). Mol. Ecol. 21, 4854–4871 (2012).
PubMed  Google Scholar 

6.
Moura, A. E. et al. Recent diversification of a Marine Genus (Tursiops spp.) tracks habitat preference and environmental change. Syst. Biol. 62, 865–877 (2013).
PubMed  Google Scholar 

7.
Whitehead, H. Cultural selection and genetic diversity in matrilineal whales. Science 282, 1708–1711 (1998).
ADS  CAS  PubMed  Google Scholar 

8.
Fontaine, M. C. et al. Postglacial climate changes and rise of three ecotypes of harbour porpoises, Phocoena phocoena, in western Palearctic waters. Mol. Ecol. 23, 3306–3321 (2014).
CAS  PubMed  Google Scholar 

9.
Louis, M. et al. Ecological opportunities and specializations shaped genetic divergence in a highly mobile marine top predator. Proc. Biol. Sci. 281, 20141558–20141558 (2014).
PubMed  PubMed Central  Google Scholar 

10.
Foote, A. D. et al. Genome-culture coevolution promotes rapid divergence of killer whale ecotypes. Nat. Commun. 7, 11693 (2016).
ADS  CAS  PubMed  PubMed Central  Google Scholar 

11.
Hare, M. P., Cipriano, F. & Palumbi, S. R. Genetic evidence on the demography of speciation in allopatric dolphin species. Evolution 56, 804–816 (2002).
PubMed  Google Scholar 

12.
Pastene, L. A. et al. Radiation and speciation of pelagic organisms during periods of global warming: The case of the common minke whale, Balaenoptera acutorostrata. Mol. Ecol. 16, 1481–1495 (2007).
CAS  PubMed  Google Scholar 

13.
Barnes, L. G. Evolution, taxonomy and antitropical distributions of the porpoises (Phocoenidae, Mammalia). Mar. Mammal Sci. 1, 149–165 (1985).
Google Scholar 

14.
Burridge, C. P. Antitropicality of Pacific fishes: Molecular insights. Environ. Biol. Fishes 65, 151–164 (2002).

15.
Banguera-Hinestroza, E., Hayano, A., Crespo, E. & Hoelzel, A. R. Delphinid systematics and biogeography with a focus on the current genus Lagenorhynchus: Multiple pathways for antitropical and trans-oceanic radiation. Mol. Phylogenet. Evol. 80, 217–230 (2014).
PubMed  Google Scholar 

16.
Marx, F. G. & Uhen, M. D. Climate, critters, and cetaceans: Cenozoic drivers of the evolution of modern whales. Science 327, 993–996 (2010).
ADS  CAS  PubMed  Google Scholar 

17.
McGowen, M. R., Spaulding, M. & Gatesy, J. Divergence date estimation and a comprehensive molecular tree of extant cetaceans. Mol. Phylogenet. Evol. 53, 891–906 (2009).
CAS  PubMed  Google Scholar 

18.
Gaskin, D. E. The ecology of whales and dolphins (Heinemann, London, 1982).
Google Scholar 

19.
Zhou, X. et al. Population genomics of finless porpoises reveal an incipient cetacean species adapted to freshwater. Nat. Commun. 9, 1276 (2018).
ADS  PubMed  PubMed Central  Google Scholar 

20.
Teilmann, J. & Sveegaard, S. Porpoises the World over: Diversity in behavior and ecology. in Ethology and Behavioral Ecology of Odontocetes (ed. Würsig, B). Vol. 54, 449–464 (Springer International Publishing, New York, 2019).

21.
Ridgway, S. H. & Johnston, D. G. Blood oxygen and ecology of porpoises of three genera. Science 151, 456–458 (1966).
ADS  CAS  PubMed  Google Scholar 

22.
Morell, V. World’s most endangered marine mammal down to 30. Science 355, 558–559 (2017).
ADS  CAS  PubMed  Google Scholar 

23.
Amante, C. & Eatkins, B. W. ETOPO1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC. https://doi.org/10.7289/V5C8276M.

24.
Berta, A., Sumich, J. L. & Kovacs, K. M. Chapter 6 – Evolution and geography. in Marine Mammals: Evolutionary Biology 131–166 (Elsevier, Amsterdam, 2015). https://doi.org/10.1016/B978-0-12-397002-2.00006-5.

25.
Chen, M. et al. Genetic footprint of population fragmentation and contemporary collapse in a freshwater cetacean. Sci. Rep. 7, 14449 (2017).
ADS  PubMed  PubMed Central  Google Scholar 

26.
Hayano, A., Amano, M. & Miyazaki, N. Phylogeography and population structure of the Dall’s porpoise, Phocoenoides dalli, in Japanese waters revealed by mitochondrial DNA. Genes Genet. Syst. 78, 81–91 (2003).
CAS  PubMed  Google Scholar 

27.
Rosa, S. et al. Population structure of nuclear and mitochondrial DNA variation among South American Burmeister’s porpoises (Phocoena spinipinnis). Conserv. Genet. 6, 431–443 (2005).
CAS  Google Scholar 

28.
Méndez-Fernandez, P. et al. Ecological niche segregation among five toothed whale species off the NW Iberian Peninsula using ecological tracers as multi-approach. Mar. Biol. 160, 2825–2840 (2013).
Google Scholar 

29.
Galatius, A., Kinze, C. C. & Teilmann, J. Population structure of harbour porpoises in the Baltic region: Evidence of separation based on geometric morphometric comparisons. J. Mar. Biol. Ass. 92, 1669–1676 (2012).
Google Scholar 

30.
Fontaine, M. C. Harbour porpoises, Phocoena phocoena, in the Mediterranean Sea and adjacent regions: Biogeographic relicts of the Last Glacial Period. Adv. Mar. Biol. 75, 333–358 (2016).
CAS  PubMed  Google Scholar 

31.
Tezanos-Pinto, G. et al. A worldwide perspective on the population structure and genetic diversity of bottlenose dolphins (Tursiops truncatus) in New Zealand. J. Hered. 100, 11–24 (2009).
CAS  PubMed  Google Scholar 

32.
Thomas, L. et al. Last call: Passive acoustic monitoring shows continued rapid decline of critically endangered vaquita. J. Acoust. Society Am. 142, EL512–EL517 (2017).
Google Scholar 

33.
Jaramillo Legorreta, A. M. et al. Decline towards extinction of Mexico’s vaquita porpoise (Phocoena sinus). R. Soc. Open Sci. 6, 190598 (2019).
ADS  PubMed  PubMed Central  Google Scholar 

34.
Wang, J. Y. & Reeves, R. R. Neophocaena phocaenoides. The IUCN Red List of Threatened Species. e.T198920A50386795. https://doi.org/10.2305/IUCN.UK.2017-3.RLTS.T198920A50386795.en. Downloaded on 04 April 2019 (2017).

35.
Wang, D., Turvey, S. T., Zhao, X. & Mei, Z. Neophocaena asiaeorientalis ssp. asiaeorientalis. The IUCN Red List of Threatened Species. e.T43205774A45893487. https://doi.org/10.2305/IUCN.UK.2013-1.RLTS.T43205774A45893487.en. Downloaded on 04 April 2019. (2013).

36.
Birkun, A. A., Jr & Frantzis, A. Phocoena phocoena ssp. relicta. The IUCN Red List of Threatened Species. e.T17030A6737111. https://doi.org/10.2305/IUCN.UK.2008.RLTS.T17030A6737111.en. Downloaded on 04 April 2019 (2008).

37.
Read, F. L., Santos, M. B. & González, A. F. Understanding Harbour Porpoise (Phocoena phocoena) and Fishery Interactions in the North-West Iberian Peninsula. (Final report to ASCOBANS, 2012).

38.
Dufresnes, C. et al. Conservation phylogeography: Does historical diversity contribute to regional vulnerability in European tree frogs (Hyla arborea)?. Mol. Ecol. 22, 5669–5684 (2013).
PubMed  Google Scholar 

39.
Malaney, J. L. & Cook, J. A. Using biogeographical history to inform conservation: The case of Preble’s meadow jumping mouse. Mol. Ecol. 22, 6000–6017 (2013).
PubMed  Google Scholar 

40.
Moritz, C. C. & Potter, S. The importance of an evolutionary perspective in conservation policy planning. Mol. Ecol. 22, 5969–5971 (2013).
PubMed  Google Scholar 

41.
Fajardo-Mellor, L. et al. The phylogenetic relationships and biogeography of true porpoises (Mammalia: Phocoenidae) based on morphological data. Mar. Mammal Sci. 22, 910–932 (2006).
Google Scholar 

42.
Rosel, P. E., Haygood, M. G. & Perrin, W. F. Phylogenetic relationships among the true porpoises (Cetacea: Phocoenidae). Mol. Phylogenet. Evol. 4, 463–474 (1995).
CAS  PubMed  Google Scholar 

43.
Torroni, A., Achilli, A., Macaulay, V., Richards, M. & Bandelt, H.-J. Harvesting the fruit of the human mtDNA tree. Trends Genet. 22, 339–345 (2006).
CAS  PubMed  Google Scholar 

44.
Viricel, A. & Rosel, P. E. Evaluating the utility of cox1 for cetacean species identification. Mar. Mammal Sci. 28, 37–62 (2011).
Google Scholar 

45.
Andrews, S. FastQC: A quality control tool for high throughput sequence data. https://www.bioinformatics.babraham.ac.uk/projects/fastqc (2010).

46.
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
CAS  PubMed  PubMed Central  Google Scholar 

47.
Kearse, M. et al. Geneious Basic: An integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics 28, 1647–1649 (2012).
PubMed  PubMed Central  Google Scholar 

48.
Arnason, U., Gullberg, A. & Janke, A. Mitogenomic analyses provide new insights into cetacean origin and evolution. Gene 333, 27–34 (2004).
CAS  PubMed  Google Scholar 

49.
Hahn, C., Bachmann, L. & Chevreux, B. Reconstructing mitochondrial genomes directly from genomic next-generation sequencing reads—A baiting and iterative mapping approach. Nucleic Acids Res. 41, e129–e129 (2013).
CAS  PubMed  PubMed Central  Google Scholar 

50.
Morin, P. A. et al. Complete mitochondrial genome phylogeographic analysis of killer whales (Orcinus orca) indicates multiple species. Genome Res. 20, 908–916 (2010).
CAS  PubMed  PubMed Central  Google Scholar 

51.
Edgar, R. C. MUSCLE: Multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32, 1792–1797 (2004).
CAS  PubMed  PubMed Central  Google Scholar 

52.
Clayton, D. A. Transcription and replication of mitochondrial DNA. Hum. Reprod. 15(Suppl 2), 11–17 (2000).
PubMed  Google Scholar 

53.
Guindon, S. et al. New algorithms and methods to estimate maximum-likelihood phylogenies: Assessing the performance of PhyML 3.0. Syst. Biol. 59, 307–321 (2010).
CAS  Google Scholar 

54.
Ronquist, F. et al. MrBayes 3.2: Efficient Bayesian phylogenetic inference and model choice across a large model space. Syst. Biol. 61, 539–542 (2012).
PubMed  PubMed Central  Google Scholar 

55.
Darriba, D., Taboada, G. L., Doallo, R. & Posada, D. jModelTest 2: More models, new heuristics and parallel computing. Nat. Methods 9, 772–772 (2012).
CAS  PubMed  PubMed Central  Google Scholar 

56.
Rambaut, A., Suchard, M. A., Xie, D. & Drummond, A. J. Tracer v.1.6. (2014). https://tree.bio.ed.ac.uk/software/tracer/. Accessed 26 Feb 2017.

57.
Yu, G., Lam, T.T.-Y., Zhu, H. & Guan, Y. Two methods for mapping and visualizing associated data on phylogeny using Ggtree. Mol. Biol. Evol. 35, 3041–3043 (2018).
CAS  PubMed  PubMed Central  Google Scholar 

58.
Bouckaert, R. et al. BEAST 2: A software platform for bayesian evolutionary analysis. PLoS Comput. Biol. 10, e1003537–e1003546 (2014).
PubMed  PubMed Central  Google Scholar 

59.
Nabholz, B., Glemin, S. & Galtier, N. Strong variations of mitochondrial mutation rate across mammals—The longevity hypothesis. Mol. Biol. Evol. 25, 120–130 (2007).
PubMed  Google Scholar 

60.
Fontaine, M. C. et al. Genetic and historic evidence for climate-driven population fragmentation in a top cetacean predator: The harbour porpoises in European water. Proc. Biol. Sci. 277, 2829–2837 (2010).
PubMed  PubMed Central  Google Scholar 

61.
Rambaut, A. & Drummond, A. J. FigTree version 1.4.3. (tree.bio.ed.ac.uk/software/figtree, 2012).

62.
Librado, P. & Rozas, J. DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25, 1451–1452 (2009).
CAS  PubMed  Google Scholar 

63.
Sanders, H. L. Marine benthic diversity: A comparative study. Am. Nat. 102, 243–282 (1968).
Google Scholar 

64.
McDonald, J. H. & Kreitman, M. Adaptive protein evolution at the Adh locus in Drosophila. Nature 351, 652–654 (1991).
ADS  CAS  PubMed  Google Scholar 

65.
Hervé, M. RVAideMemoire: Testing and plotting procedures for biostatistics. https://cran.r-project.org/web/packages/RVAideMemoire/index.html (2019).

66.
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J. Roy. Stat. Soc. Ser. B (Methodol.) 57, 289–300 (1995).
MathSciNet  MATH  Google Scholar 

67.
Kimura, M. The Neutral Theory of Molecular Evolution (Cambridge University Press, Cambridge, 1983). https://doi.org/10.1017/CBO9780511623486.
Google Scholar 

68.
Hughes, A. L. Near neutrality: Leading edge of the neutral theory of molecular evolution. Ann. N. Y. Acad. Sci. 1133, 162–179 (2008).
ADS  PubMed  PubMed Central  Google Scholar 

69.
Phifer-Rixey, M. et al. Adaptive evolution and effective population size in wild house mice. Mol. Biol. Evol. 29, 2949–2955 (2012).
CAS  PubMed  PubMed Central  Google Scholar 

70.
Eyre-Walker, A. Changing effective population size and the McDonald–Kreitman test. Genetics 162, 2017–2024 (2002).
PubMed  PubMed Central  Google Scholar 

71.
Parsch, J., Zhang, Z. & Baines, J. F. The influence of demography and weak selection on the McDonald-Kreitman test: An empirical study in Drosophila. Mol. Biol. Evol. 26, 691–698 (2009).
CAS  PubMed  Google Scholar 

72.
Romiguier, J. et al. Fast and robust characterization of time-heterogeneous sequence evolutionary processes using substitution mapping. PLoS ONE 7, e33852 (2012).
ADS  CAS  PubMed  PubMed Central  Google Scholar 

73.
Dutheil, J. & Boussau, B. Non-homogeneous models of sequence evolution in the Bio++ suite of libraries and programs. BMC Evol. Biol. 8, 255 (2008).
PubMed  PubMed Central  Google Scholar 

74.
Dutheil, J. Y. et al. Efficient selection of branch-specific models of sequence evolution. Mol. Biol. Evol. 29, 1861–1874 (2012).
CAS  PubMed  Google Scholar 

75.
R Core Team. R: A language and environment for statistical computing. Vienna, Austria. https://www.R-project.org/ (2018).

76.
Figuet, E., Romiguier, J., Dutheil, J. Y. & Galtier, N. Mitochondrial DNA as a tool for reconstructing past life-history traits in mammals. J. Evol. Biol. 27, 899–910 (2014).
CAS  PubMed  Google Scholar 

77.
Tajima, F. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123, 585–595 (1989).
CAS  PubMed  PubMed Central  Google Scholar 

78.
Fu, Y. X. & Li, W. H. Statistical tests of neutrality of mutations. Genetics 133, 693–709 (1993).
CAS  PubMed  PubMed Central  Google Scholar 

79.
Excoffier, L. & Lischer, H. E. L. Arlequin suite ver 3.5: A new series of programs to perform population genetics analyses under Linux and Windows. Mol. Ecol. Res. 10, 564–567 (2010).
Google Scholar 

80.
Schneider, S. & Excoffier, L. Estimation of past demographic parameters from the distribution of pairwise differences when the mutation rates vary among sites: application to human mitochondrial DNA. Genetics 152, 1079–1089 (1999).
CAS  PubMed  PubMed Central  Google Scholar 

81.
Drummond, A. J., Rambaut, A., Shapiro, B. & Pybus, O. G. Bayesian coalescent inference of past population dynamics from molecular sequences. Mol. Biol. Evol. 22, 1185–1192 (2005).
CAS  PubMed  Google Scholar 

82.
Kingman, J. F. C. The coalescent. Stochastic Process. Appl. 13, 235–248 (1982).
MathSciNet  MATH  Google Scholar 

83.
Kass, R. E. & Raftery, A. E. Bayes factors. J. Am. Stat. Assoc. 90, 773–795 (1995).
MathSciNet  MATH  Google Scholar 

84.
Moura, A. E. et al. Phylogenomics of the genus Tursiops and closely related Delphininae reveals extensive reticulation among lineages and provides inference about eco-evolutionary drivers. Mol. Phylogenet. Evol. 146, 106756 (2020).
PubMed  Google Scholar 

85.
Slater, G. J., Price, S. A., Santini, F. & Alfaro, M. E. Diversity versus disparity and the radiation of modern cetaceans. Proc. Biol. Sci. 277, 3097–3104 (2010).
PubMed  PubMed Central  Google Scholar 

86.
McGowen, M. R. et al. Phylogenomic resolution of the cetacean tree of life using target sequence capture. Syst. Biol. 31, 2553 (2019).
Google Scholar 

87.
Ho, S. Y. W., Saarma, U., Barnett, R., Haile, J. & Shapiro, B. The effect of inappropriate calibration: Three case studies in molecular ecology. PLoS ONE 3, e1615 (2008).
ADS  PubMed  PubMed Central  Google Scholar 

88.
Zheng, Y. & Wiens, J. J. Do missing data influence the accuracy of divergence-time estimation with BEAST? Mol. Phylogenet. Evol. 85, 41–49 (2015).
PubMed  Google Scholar 

89.
Lindberg, D. R. Marine biotic interchange between the northern and southern hemispheres. Paleobiology 17, 308–324 (1991).
Google Scholar 

90.
Perrin, W. F. Coloration. in Encyclopedia of Marine Mammals (eds. Würsig, B., Perrin, W. & Thewissen, J. G. M.) 243–249 (Elsevier, 2009). https://doi.org/10.1016/B978-0-12-373553-9.00061-4.

91.
Koopman, H. N., Pabst, D. A., McLellan, W. A., Dillaman, R. M. & Read, A. J. Changes in blubber distribution and morphology associated with starvation in the harbor porpoise (Phocoena phocoena): Evidence for regional differences in blubber structure and function. Physiol. Biochem. Zool. 75, 498–512 (2002).
CAS  PubMed  Google Scholar 

92.
Hoekendijk, J. P. A., Spitz, J., Read, A. J., Leopold, M. F. & Fontaine, M. C. Resilience of harbor porpoises to anthropogenic disturbance: Must they really feed continuously? Mar. Mammal Sci. 34, 258–264 (2018).
Google Scholar 

93.
Escorza-Treviño, S. & Dizon, A. E. Phylogeography, intraspecific structure and sex-biased dispersal of Dall’s porpoise, Phocoenoides dalli, revealed by mitochondrial and microsatellite DNA analyses. Mol. Ecol. 9, 1049–1060 (2000).
PubMed  Google Scholar 

94.
Wang, J. Y., Frasier, T. R., Yang, S. C. & White, B. N. Detecting recent speciation events: The case of the finless porpoise (genus Neophocaena). Heredity (Edinb) 101, 145–155 (2008).
CAS  Google Scholar 

95.
Lin, W. et al. Phylogeography of the finless porpoise (genus Neophocaena): Testing the stepwise divergence hypothesis in the northwestern Pacific. Sci. Rep. 4, 6572 (2014).
CAS  PubMed  PubMed Central  Google Scholar 

96.
Rosel, P. E., Dizon, A. E. & Haygood, M. G. Variability of the mitochondrial control region in populations of the harbour porpoise, Phocoena, on interoceanic and regional scales. Can. J. Fish. Aquat. Sci. 52, 1210–1219 (1995).
CAS  Google Scholar 

97.
Harris, S. A. Thermal history of the Arctic Ocean environs adjacent to North America during the last 3.5 Ma and a possible mechanism for the cause of the cold events (major glaciations and permafrost events). Progress Phys. Geogr. Earth Environ. 29, 218–237 (2005).
Google Scholar 

98.
Chivers, S. J., Dizon, A. E. & Gearin, P. J. Small-scale population structure of eastern North Pacific harbour porpoises (Phocoena phocoena) indicated by molecular genetic analyses. J. Cetacean Res. Manag. 4, 111–122 (2002).
Google Scholar 

99.
Pimper, L. E., Goodall, R. N. P. & Remis, M. I. First mitochondrial DNA analysis of the spectacled porpoise (Phocoena dioptrica) from Tierra del Fuego, Argentina. Mamm. Biol. Zeitschrift für Säugetierkunde 77, 459–462 (2012).
Google Scholar 

100.
Lundmark, C. Science sings the blues: Other words for Nothin’ left to lose. Bioscience 57, 208–208 (2007).
Google Scholar 

101.
Ehlers, J. R. & Gibbard, P. Quaternary glaciation. in Encyclopedia of Snow, Ice and Glaciers 873–882 (Springer, Dordrecht, 2014). https://doi.org/10.1007/978-90-481-2642-2_423

102.
Norris, K. S. & McFarland, W. N. A new harbor porpoise of the genus Phocoena from the Gulf of California. J. Mammal. 39, 22 (1958).
Google Scholar 

103.
Rosel, P. E. & Rojas-Bracho, L. Mitochondrial DNA variation in the critically endangered Vaquita Phocoena Sinus Norris and Macfarland, 1958. Mar. Mammal Sci. 15, 990–1003 (1999).
Google Scholar 

104.
Allendorf, F. W., Luikart, G. H. & Aitken, S. N. Conservation and the Genetics of Populations (Wiley, New York, 2012).
Google Scholar 

105.
Moritz, C. Defining ‘Evolutionarily Significant Units’ for conservation. Trends Ecol. Evol. 9, 373–375 (1994).
CAS  PubMed  Google Scholar 

106.
Nabholz, B., Mauffrey, J.-F., Bazin, E., Galtier, N. & Glemin, S. Determination of mitochondrial genetic diversity in mammals. Genetics 178, 351–361 (2008).
PubMed  PubMed Central  Google Scholar 

107.
Bazin, E., Glemin, S. & Galtier, N. Population size does not influence mitochondrial genetic diversity in animals. Science 312, 570–572 (2006).
ADS  CAS  PubMed  Google Scholar 

108.
Degnan, J. H. & Rosenberg, N. A. Gene tree discordance, phylogenetic inference and the multispecies coalescent. Trends Ecol. Evol. 24, 332–340 (2009).
PubMed  Google Scholar 

109.
Fontaine, M. C. et al. Mosquito genomics. Extensive introgression in a malaria vector species complex revealed by phylogenomics. Science 347, 1258524 (2015).
PubMed  Google Scholar 

110.
Heliconius Genome Consortium. Butterfly genome reveals promiscuous exchange of mimicry adaptations among species. Nature 487, 94–98 (2012).
ADS  Google Scholar 

111.
Miles, A. et al. Genetic diversity of the African malaria vector Anopheles gambiae. Nature 552, 96–100 (2017).
ADS  Google Scholar  More