Ecology

1.Barnosky, A. D. et al. Has the Earth’s sixth mass extinction already arrived?. Nature 471, 51–57 (2011).CAS 
PubMed 
Article 
ADS 
PubMed Central 

Google Scholar 
2.Zaffos, A., Finnegan, S. & Peters, S. E. Plate tectonic regulation of global marine animal diversity. Proc. Natl. Acad. Sci. 114, 5653–5658 (2017).CAS 
PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
3.Claramunt, S. & Cracraft, J. A new time tree reveals Earth historys imprint on the evolution of modern birds. Sci. Adv. 1, e1501005 (2015).PubMed 
PubMed Central 
Article 
ADS 
CAS 

Google Scholar 
4.Leprieur, F., Descombes, P., Gaboriau, T., Cowman, P. F. & Parravicini, V. Plate tectonics drive tropical reef biodiversity dynamics. Nat. Commun. 7, 11461 (2016).CAS 
PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
5.Ficetola, G. F., Mazel, F. & Thuiller, W. Global determinants of zoogeographical boundaries. Nat. Ecol. Evol. 1, 0089 (2017).Article 

Google Scholar 
6.Mazel, F. et al. Global patterns of β-diversity along the phylogenetic time-scale: The role of climate and plate tectonics. Glob. Ecol. Biogeogr. 26, 1211–1221 (2017).Article 

Google Scholar 
7.Hofreiter, M. & Stewart, J. Ecological change, range fluctuations and population dynamics during the pleistocene. Curr. Biol. 19, R584–R594 (2009).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
8.Pellissier, L. et al. Quaternary coral reef refugia preserved fish diversity. Science 344, 1016–1019 (2014).CAS 
PubMed 
Article 
ADS 
PubMed Central 

Google Scholar 
9.Jaramillo, C. et al. Effects of rapid global warming at the paleocene-eocene boundary on neotropical vegetation. Science 330, 957–961 (2010).CAS 
PubMed 
Article 
ADS 
PubMed Central 

Google Scholar 
10.Svenning, J.-C., Eiserhardt, W. L., Normand, S., Ordonez, A. & Sandel, B. The influence of paleoclimate on present-day patterns in biodiversity and ecosystems. Annu. Rev. Ecol. Evol. Syst. 46, 551–572 (2015).Article 

Google Scholar 
11.Steeman, M. E. et al. Radiation of extant cetaceans driven by restructuring of the oceans. Syst. Biol. 58, 573–585 (2009).PubMed 
PubMed Central 
Article 

Google Scholar 
12.Antonelli, A. & Sanmartín, I. Mass Extinction, gradual cooling, or rapid radiation? reconstructing the spatiotemporal evolution of the ancient angiosperm genus hedyosmum (Chloranthaceae) using empirical and simulated approaches. Syst. Biol. 60, 596–615 (2011).PubMed 
Article 
PubMed Central 

Google Scholar 
13.Nee, S., May, R. M. & Harvey, P. H. The reconstructed evolutionary process. Philos. Trans. R. Soc. Lond. B 344, 305–311 (1994).CAS 
Article 
ADS 

Google Scholar 
14.Morlon, H., Parsons, T. L. & Plotkin, J. B. From the cover: Reconciling molecular phylogenies with the fossil record. Proc. Natl. Acad. Sci. 108, 16327–16332 (2011).CAS 
PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
15.Condamine, F. L., Rolland, J. & Morlon, H. Macroevolutionary perspectives to environmental change. Ecol. Lett. 16, 72–85 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
16.Condamine, F. L. et al. Deciphering the evolution of birdwing butterflies 150 years after Alfred Russel Wallace Deciphering the evolution of birdwing butterflies 150 years after. Sci. Rep. 5, 11860 (2015).PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
17.Lagomarsino, L. P., Condamine, F. L., Antonelli, A., Mulch, A. & Davis, C. C. The abiotic and biotic drivers of rapid diversification in Andean bellflowers (Campanulaceae). New Phytol. 210, 1430–1442 (2016).PubMed 
PubMed Central 
Article 

Google Scholar 
18.Rolland, J. & Condamine, F. L. The contribution of temperature and continental fragmentation to amphibian diversification. J. Biogeogr. 46, 1857–1873 (2019).Article 

Google Scholar 
19.Gaboriau, T. et al. Ecological constraints coupled with deep-time habitat dynamics predict the latitudinal diversity gradient in reef fishes. Proc. R. Soc. B Biol. Sci. 286, 20191506 (2019).Article 

Google Scholar 
20.Descombes, P. et al. Linking species diversification to palaeo-environmental changes: A process-based modelling approach. Glob. Ecol. Biogeogr. 00, 1–12 (2017).
Google Scholar 
21.Rangel, T. F. et al. Modeling the ecology and evolution of biodiversity: Biogeographical cradles, museums, and graves. Science 361, 5452 (2018).Article 
CAS 

Google Scholar 
22.Pontarp, M. et al. The latitudinal diversity gradient: Novel understanding through mechanistic eco-evolutionary models. Trends Ecol. Evol. 34, 211–223 (2019).PubMed 
Article 
PubMed Central 

Google Scholar 
23.Cowman, P. F. Historical factors that have shaped the evolution of tropical reef fishes: A review of phylogenies, biogeography, and remaining questions. Front. Genet. 5, 1–15 (2014).Article 

Google Scholar 
24.Bellwood, D. R. et al. Evolutionary history of the butterflyfishes (f: Chaetodontidae) and the rise of coral feeding fishes. J. Evol. Biol. 23, 335–349 (2010).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
25.Cowman, P. F. & Bellwood, D. R. Coral reefs as drivers of cladogenesis: Expanding coral reefs, cryptic extinction events, and the development of biodiversity hotspots. J. Evol. Biol. 24, 2543–2562 (2011).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
26.Sorenson, L., Santini, F., Carnevale, G. & Alfaro, M. E. A multi-locus timetree of surgeonfishes (Acanthuridae, Percomorpha), with revised family taxonomy. Mol. Phylogenet. Evol. 68, 150–160 (2013).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
27.Dornburg, A., Moore, J., Beaulieu, J. M., Eytan, R. I. & Near, T. J. The impact of shifts in marine biodiversity hotspots on patterns of range evolution: Evidence from the Holocentridae (squirrelfishes and soldierfishes). Evolution 69, 146–161 (2015).PubMed 
Article 
PubMed Central 

Google Scholar 
28.Cowman, P. F. & Bellwood, D. R. The historical biogeography of coral reef fishes: Global patterns of origination and dispersal. J. Biogeogr. 40, 209–224 (2013).Article 

Google Scholar 
29.Lohman, D. J. et al. Biogeography of the Indo-Australian archipelago. Annu. Rev. Ecol. Evol. Syst. 42, 205–226 (2011).Article 

Google Scholar 
30.Gaboriau, T., Leprieur, F., Mouillot, D. & Hubert, N. Influence of the geography of speciation on current patterns of coral reef fish biodiversity across the Indo-Pacific. Ecography 41, 1295–1306 (2017).Article 

Google Scholar 
31.McManus, J. W. Marine speciation, tectonics and sea- level changes in Southeast Asia. Proc. Fifth Int. Coral Reef 4, 133–138 (1985).
Google Scholar 
32.Potts, D. C. Sea-level fluctuations and speciation in Scleractinia. Proc. Fifth Int. Coral Reef 4, 51–62 (1985).
Google Scholar 
33.Hou, Z. & Li, S. Tethyan changes shaped aquatic diversification. Biol. Rev. https://doi.org/10.1111/brv.12376 (2017).Article 
PubMed 
PubMed Central 

Google Scholar 
34.Stadler, T. Mammalian phylogeny reveals recent diversification rate shifts. Proc. Natl. Acad. Sci. USA 108, 6187–6192 (2011).CAS 
PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
35.Bellwood, D. R. & Wainwright, P. C. The history and biogeography of Fishes on Coral Reefs. in Coral Reef Fishes, Dynamics and Diversity in a Complex Ecosystem, 5–32 (2002).36.Williams, S. T. & Duda, T. F. Did tectonic activity stimulate Oligo-Miocene speciation in the Indo-West Pacific?. Evolution 62, 1618–1634 (2008).PubMed 
Article 
PubMed Central 

Google Scholar 
37.Renema, W. et al. Hopping hotspots: Global shifts in marine biodiversity. Science 321, 654–657 (2008).CAS 
PubMed 
Article 
ADS 
PubMed Central 

Google Scholar 
38.Tea, Y.-K. et al. Phylogenomic analysis of concatenated ultraconserved elements reveals the recent evolutionary radiation of the fairy wrasses (teleostei: labridae: cirrhilabrus). Syst. Biol. 1, 1–12 (2021).
Google Scholar 
39.Hall, R. Southeast Asia’s changing palaeogeography. Blumea J. Plant Taxon. Plant Geogr. 54, 148–161 (2009).Article 

Google Scholar 
40.Keith, S. A., Baird, A. H., Hughes, T. P., Madin, J. S. & Connolly, S. R. Faunal breaks and species composition of Indo-Pacific corals: The role of plate tectonics, environment and habitat distribution. Proc. Biol. Sci. 280, 20130818 (2013).CAS 
PubMed 
PubMed Central 

Google Scholar 
41.Cowman, P. F. & Bellwood, D. R. Vicariance across major marine biogeographic barriers: Temporal concordance and the relative intensity of hard versus soft barriers. Proc. Biol. Sci. 280, 20131541 (2013).PubMed 
PubMed Central 

Google Scholar 
42.Price, S. A., Claverie, T., Near, T. J. & Wainwright, P. C. Phylogenetic insights into the history and diversification of fishes on reefs. Coral Reefs 34, 997–1009 (2015).Article 
ADS 

Google Scholar 
43.Bellwood, D. R., Goatley, C. H. R. & Bellwood, O. The evolution of fishes and corals on reefs: Form, function and interdependence. Biol. Rev. 92, 878–901 (2017).PubMed 
Article 
PubMed Central 

Google Scholar 
44.Bowen, B. W., Rocha, L. A., Toonen, R. J. & Karl, S. A. The origins of tropical marine biodiversity. Trends Ecol. Evol. 28, 359–366 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
45.Price, S. A., Tavera, J. J., Near, T. J. & Wainwright, P. C. Elevated rates of morphological and functional diversification in reef-dwelling haemulid fishes. Evolution 67, 417–428 (2012).PubMed 
Article 
PubMed Central 

Google Scholar 
46.Kiessling, W., Simpson, C., Beck, B., Mewis, H. & Pandolfi, J. M. Equatorial decline of reef corals during the last Pleistocene interglacial. Proc. Natl. Acad. Sci. USA. 109, 21378–21383 (2012).CAS 
PubMed 
PubMed Central 
Article 
ADS 

Google Scholar 
47.Floeter, S. R. et al. Atlantic reef fish biogeography and evolution. J. Biogeogr. 35, 22–45 (2007).
Google Scholar 
48.Riginos, C., Buckley, Y. M., Blomberg, S. P. & Treml, E. A. Dispersal capacity predicts both population genetic structure and species richness in reef fishes. Am. Nat. 184, 52–64 (2014).PubMed 
Article 
PubMed Central 

Google Scholar 
49.Rocha, L. A. & Bowen, B. W. Speciation in coral-reef fishes. J. Fish Biol. 72, 1101–1121 (2008).Article 

Google Scholar 
50.Tedesco, P. A., Paradis, E., EvEque, C. L. & Hugueny, B. Explaining global-scale diversification patterns in actinopterygian fishes. J. Biogeogr. 44, 773–783 (2016).Article 

Google Scholar 
51.Rosenzweig, M. L. Species Diversity in Space and Time (Springer, 1995).Book 

Google Scholar 
52.Kisel, Y. & Barraclough, T. G. Speciation has a spatial scale that depends on levels of gene flow. Am. Nat. 175, 316–334 (2010).PubMed 
Article 
PubMed Central 

Google Scholar 
53.Fine, P. V. A. & Ree, R. H. Evidence for a time-integrated species-area effect on the latitudinal gradient in tree diversity. Am. Nat. 168, 796–804 (2006).PubMed 
Article 
PubMed Central 

Google Scholar 
54.Jetz, W. & Fine, P. V. A. Global gradients in vertebrate diversity predicted by historical area-productivity dynamics and contemporary environment. PLoS Biol. 10, e1001292 (2012).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
55.Konow, N., Price, S., Abom, R., Bellwood, D. & Wainwright, P. Decoupled diversification dynamics of feeding morphology following a major functional innovation in marine butterflyfishes. Proc. Biol. Sci. 284, 20170906 (2017).PubMed 
PubMed Central 

Google Scholar 
56.Clements, K. D., German, D. P., Piché, J., Tribollet, A. & Choat, J. H. Integrating ecological roles and trophic diversification on coral reefs: Multiple lines of evidence identify parrotfishes as microphages. Biol. J. Linn. Soc. 120, 729–751 (2017).
Google Scholar 
57.Lobato, F. L. et al. Diet and diversification in the evolution of coral reef fishes. PLoS ONE 9, e102094 (2014).PubMed 
PubMed Central 
Article 
ADS 
CAS 

Google Scholar 
58.Siqueira, A. C., Morais, R. A., Bellwood, D. R. & Cowman, P. F. Trophic innovations fuel reef fish diversification. Nat. Commun. 11, 1–11 (2020).Article 
CAS 

Google Scholar 
59.Louca, S. & Pennell, M. W. Extant timetrees are consistent with a myriad of diversification histories. Nature 580, 502–505 (2020).CAS 
PubMed 
Article 
ADS 
PubMed Central 

Google Scholar 
60.Morlon, H., Hartig, F. & Robin, S. Prior hypotheses or regularization allow inference of diversification histories from extant timetrees. bioRxiv (2020).61.McCord, C. L. & Westneat, M. W. Phylogenetic relationships and the evolution of BMP4 in triggerfishes and filefishes (Balistoidea). Mol. Phylogenet. Evol. 94, 397–409 (2016).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
62.Santini, F. & Carnevale, G. First multilocus and densely sampled timetree of trevallies, pompanos and allies (Carangoidei, Percomorpha) suggests a Cretaceous origin and Eocene radiation of a major clade of piscivores. Mol. Phylogenet. Evol. 83, 33–39 (2015).PubMed 
Article 
PubMed Central 

Google Scholar 
63.Santini, F., Carnevale, G. & Sorenson, L. First multi-locus timetree of seabreams and porgies (Percomorpha: Sparidae). Ital. J. Zool. 81, 55–71 (2014).Article 

Google Scholar 
64.Müller, R. D., Sdrolias, M., Gaina, C., Steinberger, B. & Heine, C. Long-term sea-level fluctuations driven by ocean basin dynamics. Science 319, 1357–1362 (2008).PubMed 
Article 
ADS 
CAS 
PubMed Central 

Google Scholar 
65.Heine, C., Yeo, L. G. & Müller, R. D. Evaluating global paleoshoreline models for the Cretaceous and Cenozoic. Aust. J. Earth Sci. 62, 275–287 (2015).CAS 

Google Scholar 
66.Kleypas, J. A. & Mcmanus, J. W. Environmental Limits to Coral Reef Development : Where Do We Draw the Line ?. Am. Zool. 39, 146–159 (1999).Article 

Google Scholar 
67.Bugayevskiy, L. M. & Snyder, J. P. Map Projections: A Reference Manual (Taylor & Francis, London, 1995).
Google Scholar 
68.Chang, J., Rabosky, D. L. & Alfaro, M. E. Estimating diversification rates on incompletely sampled phylogenies: Theoretical concerns and practical solutions. Syst. Biol. 69, 602–611 (2020).PubMed 
Article 
PubMed Central 

Google Scholar 
69.Morlon, H. et al. RPANDA: An R package for macroevolutionary analyses on phylogenetic trees. Methods Ecol. Evol. 7, 589–597 (2016).Article 

Google Scholar  More

Detailed population structure of L1–4 and a hierarchical sub-lineage naming systemWe assembled a high-quality data set of whole genomes, antibiotic resistance phenotypes, and geographic sites of isolation for 9584 clinical Mtb samples (“Methods” section and Supplementary Data 1). Of the total, 4939 (52%) were pan-susceptible, i.e., susceptible to at least isoniazid and rifampicin (and all other antibiotics when additional phenotypic data were available), and 4645 (48%) were resistant to one or more antibiotics (Supplementary Fig. 1a). Using the 62 SNS lineage barcode6, 738 isolates were classified as L1 (8%), 2193 as L2 (22%), 1104 as L3 (12%) and 5549 as L4 (58%, Supplementary Fig. 1b). Among the 4939 pan-susceptible isolates, we identified high-quality genome-wide SNSs (83,735 for L1, 56,736 for L2, 76,817 for L3, and 185,622 for L4) that we used in building maximum-likelihood phylogenies for each major lineage (L1–4, “Methods” section). We computed an index of genetic divergence (FST) between groups defined by each bifurcation in each phylogeny. Sub-lineages were defined as monophyletic groups that had high FST ( >0.33) and were also clearly separated from other groups in principal component analysis (PCA, see “Methods” section). We also defined internal groups to sub-lineages (see “Methods” section): an internal group is a monophyletic group genetically divergent (by FST and PCA) from its neighboring groups, but has one or more ancestral branches that show a low degree of divergence or low support (bootstrap values). Internal groups do not represent true sub-lineages in a hierarchical fashion, but defining them allows us to further characterize the Mtb population structure. We provide code to automate all the steps described above. Our approach is scalable and can be used on other organisms (see “Methods” section).To better classify Mtb isolates in the context of the global Mtb population structure, we developed a hierarchical sub-lineage naming scheme (Supplementary Data 2) where each subdivision in the classification corresponds to a split in the phylogenetic tree of each major Mtb lineage. Starting with the global Mtb lineage numbers (e.g., L1), we recursively introduced a subdivision (e.g., from 1.2 to 1.2.1 and 1.2.2) at each bifurcation of the phylogenetic tree whenever both subclades sufficiently diverged. Formally, we defined these splits using bootstrap criteria, and independent validations by FST and PCA (see “Methods” section). Internal groups were denoted with the letter “i” (e.g., 4.1.i1). This proposed system overcomes two major shortcomings of the existing schemas: same-level sub-lineages are never overlapping (unlike the system of Stucki et al.8 sub-lineage 4.10 includes sub-lineages 4.7–4.9), and the names reflect both phylogenetic relationships and genetic similarity (unlike semantic naming such as the “Asia ancestral” lineage in the system of Shitikov et al.7). Further, this naming system can be standardized to automate the process of lineage definition. These advantages come at the price of long sub-lineage names in the case of complex phylogenies (e.g., for L4, sub-lineage 4.10 gets the lineage designation 4.2.1.1.1.1.1.1). For compatibility with naming conventions already in use and to keep names as short as possible, we designed a second, shorthand, naming system that expands the Coll et al. lineage schema by adding new subdivisions and differentiating between sub-lineages and internal groups. For instance, sub-lineage 4.3.1 is designated as 4.3.i1, informing the user that this is an internal group of sub-lineage 4.3. To simplify the use of the hierarchical naming schema and the updated shorthand schema, we provide a table that compares them side by side along with naming systems currently in use (Supplementary Data 2).Using the sub-lineage definition rules and the sub-lineage naming scheme described above, we characterized six previously undescribed sub-lineages of L1 (Fig. 1 and Supplementary Fig. 2); five of which expand the current description of 1.2. We also detected an internal group of 91 isolates (1.1.3.i1) characterized by a long defining branch in the phylogeny (corresponding to 82 SNSs), a high FST (0.48), and geographically restricted to Malawi (85/91, 93% isolates, Fig. 1 and Supplementary Fig. 3). We estimated the date of the emergence of the MRCA of such a group (see “Methods” section) and we found it to be between 1497 and 1754. We found four previously undescribed sub-lineages of L3 (Fig. 2 and Supplementary Fig. 4), revising L3 into four main groups, whereas previously only two partitions of one sub-lineage were characterized (3.1). We found that the latter two partitions are in fact internal groups of the largest sub-lineage (3.1.1) in our revised classification.Fig. 1: Phylogenetic tree reconstruction of lineage 1 (binary tree).Gray circles define splits where the FST (fixation index) calculated using the descendants of the two children nodes is greater than 0.33. The sub-lineages are defined by colored areas (blue: sub-lineages already described in the literature; green: sub-lineages described here; purple: internal sub-lineages). Source data are provided as a Source Data file.Full size imageFig. 2: Phylogenetic tree reconstruction of lineage 3 (binary tree).Gray circles define splits where the FST (fixation index) calculated using the descendants of the two children nodes is greater than 0.33. The sub-lineages are defined by colored areas (green: sub-lineages described here; purple: internal sub-lineages). Source data are provided as a Source Data file.Full size imageL2 is divided into two groups: proto-Beijing and Beijing with the latter in turn partitioned into two groups: ancient- and modern-Beijing7. Each one of these groups is characterized by further subdivisions (three for the ancient-Beijing group and seven for the modern-Beijing group; see Supplementary Fig. 4). We found a new sub-lineage (2.2.1.2, Fig. 3, and Supplementary Fig. 5) within the previously characterized ancient-Beijing group. However, genetic diversity within the modern-Beijing group (2.2.1.1.1) was lower than in the other L2 sub-lineages and the tree topology and FST calculations did not support further hierarchical subdivisions. Although we did find three internal groups of modern-Beijing: two undescribed and one that corresponds to the Central Asia group7. For L4, our results support a complex population structure with 21 sub-lineages and 15 internal groups. In particular, we found 11 previously undescribed sub-lineages and 5 internal groups that expand our understanding of previously characterized sub-lineages (e.g., 4.2.2; 4.2 in the Coll et al. classification) or that were not characterized since these isolates were simply classified as L4 (e.g., 4.11, Fig. 4, and Supplementary Fig. 6) using the other barcodes.Fig. 3: Phylogenetic tree reconstruction of lineage 2 (binary tree).Gray circles define splits where the FST (fixation index) calculated using the descendants of the two children nodes is greater than 0.33. The sub-lineages are defined by colored areas (blue: sub-lineages already described in the literature; green: sub-lineages described here; purple: internal sub-lineages). Source data are provided as a Source Data file.Full size imageFig. 4: Phylogenetic tree reconstruction of lineage 4 (binary tree).Gray circles define splits where the FST (fixation index) calculated using the descendants of the two children nodes is greater than 0.33. The sub-lineages are defined by colored areas (blue: sub-lineages already described in the literature; green: sub-lineages described here; purple: internal sub-lineages). Source data are provided as a Source Data file.Full size imageA new barcode to define L1–4 Mtb sub-lineages and a software package to type Mtb strains from WGS dataWe defined a SNS barcode for distinguishing the obtained sub-lineages (Supplementary Data 3). We characterized new synonymous SNSs found in 100% of isolates from a given sub-lineage, but not in other isolates from the same major lineage, compiling 95 SNSs into an expanded barcode (Supplementary Data 3). We validated the barcode by using it to call sub-lineages in the hold-out set of 4645 resistant isolates and comparing the resulting sub-lineage designations with maximum-likelihood phylogenies inferred from the full SNS data (Supplementary Figs. 7–10). A sub-lineage was validated if it was found in the hold-out data and formed a monophyletic group in the phylogeny. Considering the “recent” sub-lineages, i.e., the most detailed level of classification in our system, we were able to validate eight out of nine L1 sub-lineages including five out of six of the new sub-lineages described here, with the exception of 1.1.1.2. We validated all four new L3 sub-lineages, all five L2 sub-lineages including the one previously undescribed, and 16 of the 21 L4 sub-lineages including two described here. The sub-lineages we could not confirm were not represented by any isolate in the validation phylogenies. We did not observe any paraphyletic sub-lineages in the revised classification system.We developed fast-lineage-caller, a software tool that classifies Mtb genomes using the SNS barcode proposed above. For a given genome, it returns the corresponding sub-lineage as output using our hierarchical naming system in addition to four other existing numerical/semantic naming systems, when applicable (see “Methods” section). The tool also informs the user on how many SNSs support a given lineage call and allows for filtering of low-quality variants. The tool is generalizable and can manage additional barcodes defined by the user to type the core genome of potentially any bacterial species.Geographic distribution of the Mtb sub-lineagesNext, we examined whether certain sub-lineages were geographically restricted, which would support the Mtb-human co-evolution hypothesis, or whether they constituted prevalent circulating sub-lineages in several different countries (i.e., geographically unrestricted)8. We used our SNS barcode to determine the sub-lineages of 17,432 isolates (see “Methods” section) sampled from 74 countries (Supplementary Fig. 11 and Supplementary Data 4, 5). We computed the Simpson diversity index (Sdi) as a measure of geographic diversity that controls for variable sub-lineage frequency (see “Methods” section) for each well-represented sub-lineage or internal group (n  > 20). We hypothesized that geographically unrestricted lineages would have a higher Sdi. We found Sdi to correlate highly (⍴ = 0.68; p-value = 5.7 × 10−7) with the number of continents from which a given sub-lineage was isolated (Supplementary Fig. 12). The Sdi ranged between a minimum of 0.05 and a maximum of 0.72, with a median value of 0.46 (Fig. 5). The known geographically restricted sub-lineages8 had an Sdi between 0.28 and 0.5 (Fig. 5 and Supplementary Table 1), while the known geographically unrestricted sub-lineages8,9 had an Sdi between 0.55 and 0.61 (Fig. 5 and Supplementary Table 2). We found 11 sub-lineages/internal groups with Sdi 0.61 (Supplementary Table 4), i.e., more extreme than previously reported geographically restricted or unrestricted sub-lineages, respectively.Fig. 5: Histogram of the Simpson diversity index calculated for sub-lineages of lineages 1–4.A data set of 17,432 isolates from 74 countries was used to perform this analysis. Yellow triangles designate the Simpson diversity index values of sub-lineages designated as geographically restricted by Stucki et al. Light gray circles designate the Simpson diversity index values of sub-lineages designated as geographically unrestricted by Stucki et al. Source data are provided as a Source Data file.Full size imageWhile the currently known geographically restricted sub-lineages are all in L4, we found evidence of geographic restriction for two sub-lineages/internal groups of L1. The first, the L1 internal group 1.1.3.i1, showed a very low Sdi (0.06) and was only found at high frequency among the circulating L1 isolates in Malawi (Fig. 6). This finding is also in agreement with the L1 phylogeny (Fig. 1) that shows a relatively long (82 SNS) branch defining this group. The second geographically restricted L1 sub-lineage is 1.1.1.1 (Sdi = 0.12) that was only found at high frequency among circulating L1 isolates in South-East Asia (Vietnam and Thailand, Fig. 7). To exclude the possibility that these two groups appeared geographically restricted as a result of oversampling transmission outbreaks, we calculated the distribution of the pairwise SNS distance for each of these two sub-lineages. We measured a median SNS distance of 204 and 401, respectively, refuting this kind of sampling error for these groups (typical pairwise SNS distance in outbreaks 0.67 and results on L4 transmissibility below.Differences in transmissibility between the Mtb global lineagesThe observation that some lineages/sub-lineages are more geographically widespread than others raises the question of whether this results from differences in marginal transmissibility across human populations. On a topological level, we observed L2 and L3 phylogenies to be qualitatively different from those of L1 and L4 (Figs. 1–4): displaying a star-like pattern with shorter internal branches and longer branches near the termini. We confirmed this quantitatively by generating a single phylogenetic tree for all 9584 L1–4 isolates and plotting cumulative branch lengths from root to tip for each main lineage (Supplementary Fig. 20). Star-like topologies have been postulated to associate with rapid or effective viral or bacterial transmission e.g., a “super-spreading” event in outbreak contexts25. To compare transmissibility between the four lineages, we compared the distributions of terminal branch lengths expecting a skew toward shorter terminal branch lengths supporting the idea of higher transmissibility. We found L4 to have the shortest median terminal branch length, followed in order by L2, L3, and L1 (medians: 6.2 × 10−5, 8.2 × 10−5, 10.2 × 10−5, 17.5 × 10−5, respectively; all pairwise two-sided Wilcoxon rank-sum tests significant p-value < 0.001; Fig. 9). Shorter internal node-to-tip distance is a second phylogenetic correlate of transmissibility; the distribution of this measure across the four lineages revealed a similar hierarchy to the terminal branch length distribution (Supplementary Fig. 21). We also computed the cumulative distribution of isolates separated by increasing total pairwise SNS distance (Supplementary Fig. 22). The proportion of L4 isolates separated by More

Review of C–R stability theoryTo set the context for how the R–N module will be used to understand the dynamics of nutrient-limited ecosystem models, we first briefly review stability results from modular food web theory. We do this by laying out a set of examples that serve to illustrate that in general, strong C–R interactions promote oscillatory dynamics while carefully placed weak C–R interactions dampen them5. We begin with the Rosenzweig–MacArthur C–R system as our base C–R module (Fig. 1a). It is biologically supported and produces a range of biologically plausible dynamics5, making it an appropriate system for this analysis. It exhibits three different dynamical phases over a gradient of interaction strengths (energetically defined sensu Nilsson et al. 2018) such that increasing the attack rate (({a}_{{CR}})) increases interaction strength15 (Fig. 2). We use the return time after a small perturbation (i.e., eigenvalues) to highlight the natural stability trade-off that occurs as interaction strength is changed, (i.e., the “checkmark” stability pattern)5,6. Equations and parameters can be found in Supplementary Results 1A. We draw your attention to three notable dynamical phases of the C–R module. At low interaction strengths the dominant eigenvalue (({lambda }_{{max }})) is negative and real and the C–R module follows a monotonic return to a stable equilibrium (Fig. 2a). During this phase ({lambda }_{{max }}) decreases from 0 (i.e., where ({a}_{{{CR}}}) allows the consumer to persist) to more negative values and thus stronger interactions tend to increase stability (Fig. 2d, i). At moderate interaction strengths, there is a sudden shift to eigenvalues with a non-zero complex part and population dynamics overshoot the equilibrium (Fig. 2b). Increases in interaction strength then further excite population dynamics and we observe less stable dynamics across this phase (Fig. 2d, ii). Last, the system reaches a Hopf bifurcation where the dominant eigenvalue becomes positive, yielding sustained cycles or oscillations (Fig. 2c, d, iii). As interaction strength increases across this phase, it is difficult to determine stability from the magnitude of a positive dominant eigenvalue; however, destabilization with increased interaction strength is readily observed in that the cycles become increasingly larger oscillations with a high coefficient of variation (CV)5. Note that while the Rosenzweig–MacArthur C–R system is shown here under a single set of parameters, analysis of the Jacobian shows the qualitative results to be general5. Moreover, the qualitative stability pattern remains for a type I and type III functional response5.Fig. 1: C–R and R–N base modules.a Rosenzweig–MacArthur C–R module modelled with Holling type II functional response and logistic resource growth, where (R) is resource biomass and (C) is consumer biomass. Parameters: (r) is the intrinsic growth rate of (R), (K) is the carrying capacity of (R), ({a}_{{mathrm {CR}}}) is the attack rate of (C) on (R), (e) is the assimilation rate of (C), ({R}_{0}) is the half-saturation density of (C), ({m}_{R}) and ({m}_{C}) are the mortality rates of (R) and (C), respectively. b R–N module modelled with a Monod nutrient uptake equation and external nutrient input, where (N) is a limiting-nutrient pool and (R) is the resource biomass. Parameters: ({I}_{N}) is external nutrient input to (N), ({a}_{{RN}}) is nutrient uptake rate by (R), (k) is the half-saturation density of (R), ({l}_{N}) and ({l}_{R}) are nutrient loss rates from (N) and (R), respectively.Full size imageFig. 2: C–R checkmark stability response.d Local stability (real and complex parts of the dominant eigenvalue; ({lambda }_{{max }})) as a function of interaction strength (({a}_{{{mathrm {CR}}}})) for the Rosenzweig–MacArthur C–R module. Time series reflect dynamics associated with region i, ii, and iii, respectively, following a perturbation that removes 50% of consumer biomass: a Stable equilibrium; monotonic dynamics. b Stable equilibrium; overshoot dynamics. c Unstable equilibrium; limit cycle. Boldness of arrows indicates the strength of interaction (({a}_{{CR}})).Full size imageWe now couple C–R modules into higher order food web modules to demonstrate how the addition of weak and/or strong interactions to a system can be used to predict dynamics at steady state (Fig. 3), constituting the “algebra” of C–R modules. Equations and parameters can be found in Supplementary Results 1B–D. We start with the three trophic level food chain (Fig. 3a), consisting of two coupled C–R modules (i.e., C1-R and P–C1). Theory has tended to find two weakly interacting C–R modules to generally produce locally stable equilibria16 (Fig. 3a). Increasing the strength of the C1–R interaction causes it to act like an oscillator (see Fig. 2c, above), and with enough increase this underlying oscillation is reflected in the limit cycles of the entire food chain (Fig. 3b). If the P–C1 interaction is strengthened as well, we end up with two coupled oscillators—the recipe for chaos17,18 (Fig. 3c). As such, coupled strong interactions are not surprisingly the recipe for complex and highly unstable dynamics.Fig. 3: Algebra of C–R modules.Time series showing the general dynamical outcomes for the food chain and diamond module at steady state with varied combinations of C–R interaction strengths. a Weak–weak interaction; point attractor. b Strong–weak interaction; limit cycle. c Strong–strong interaction; chaos. d Strong–strong, weak interaction; limit cycle. e Strong–strong, weak–weak interaction; point attractor.Full size imageFollowing McCann et al.19, we now add a weakly coupled consumer C2 to the food chain system of Fig. 3c. This weak consumer essentially draws energy away from the strong P–C1–R pathway and in doing so partially mutes the coupled oscillators, bringing the dynamics back to a more even limit cycle (Fig. 3d) and under certain conditions can drive equilibrium dynamics19. Last, the predator is weakly coupled to C2, creating a strong and weak pathway. The second weak interaction further draws energy away from the strong pathway, muting the oscillators entirely and bringing the system in this example to a point attractor (Fig. 3e). These examples show that well placed weak interactions (i.e., non-oscillatory phases, Fig. 2a, b) can be used to draw energy away from strong pathways and act as potent stabilizers of potentially oscillatory pathways. Note that weak interactions play a similarly stabilizing role in the omnivory module20 and further, weak interactions have been shown to stabilize large food web networks4,6 suggesting the principles derived from modular theory scale up to whole systems. Taken altogether, the oscillatory nature of strong C–R interactions generally promotes oscillatory dynamics in higher order systems, while the careful placement of weak C–R interactions—which are monotonic in nature—act to dampen oscillations. Although not discussed to our knowledge, we conjecture that if a subsystem exists such that strong interactions lead to monotonic dynamics (i.e., without oscillatory decay), strong interactions in this case would serve as a potent stabilizer. Below, we show the R–N module appears to be such a case.R–N module and stabilityTowards understanding how the R–N subsystem may interact in a higher order system, we first briefly consider the stability of the R–N module alone (akin to what we discussed for the C–R module above). The R–N module consists of a resource that takes up nutrients according to a Monod-like growth term, is open to flows from the external environment as a result of geochemical processes, and nutrients are lost to the external environment according to a linear term11 (Fig. 1b). Performing a local stability analysis about the interior equilibrium reveals the R–N module to be locally stable for all biologically feasible parameterizations, as determined by the signs of the trace and determinant of the Jacobian matrix (see Supplementary Results 2B). We now perform further numerical and analytical analyses to understand how stability is influenced by interaction strength.As the maximum rate of nutrient uptake (({a}_{{RN}})) is increased (i.e., R–N interaction strength), stability is generally increased (Fig. 4d), with the real part of the dominant eigenvalue (({lambda }_{{max }})) tending from 0 (i.e., where ({a}_{{RN}}) allows the resource to persist) towards an asymptote of ({-l}_{R}) (see Supplementary Results 2C). Numerical analysis reveals that the asymptote at ({-l}_{R}) can be approached from above or below depending on the relative leakiness of the R and N compartments (i.e., the rate at which nutrients are lost to the external environment from compartment R (({l}_{R})) and N (({l}_{N}))). For ({l}_{N} , > , {l}_{R}) (Fig. 4d), the R–N module only follows a monotonic return to equilibrium as interaction strength is increased, with increased interaction strength only tending to increased stability (i.e., reduce return time). For ({l}_{N} < {l}_{R}) (Fig. 4d), the R–N module follows a monotonic return to equilibrium for weak (Fig. 4a) and strong (Fig. 4c) interaction strength, but modest overshoot dynamics are observed for intermediate interaction strength (Fig. 4b). Stability tended to increase with interaction strength for weak to intermediate interaction strength (i.e., dominant eigenvalue becomes more negative), then slightly decrease as interaction strength became strong. A special case exists when ({l}_{R}={l}_{N}) (Fig. 4d), where stability increases with interaction strength until ({lambda }_{{max }}) becomes locked in at ({-l}_{R}), indicating stability does not change regardless of any further increase in interaction strength. Overall, the R–N interaction tends to generally stabilize in all cases (dominant eigenvalue goes from zero to a more negative saturating value with monotonic dynamics), although there are some intermediate cases that produce complex eigenvalues that suggest population dynamic overshoot potential (Fig. 4b). Note that we obtain qualitatively similar results when implicitly strengthening the R–N interaction by increasing nutrient loading (see Supplementary Results 2D and Supplementary Fig. 1). Now, given the above framework for coupled C–R modules—where weak C–R interactions with underlying monotonic dynamics dampen the oscillatory potential of strong C–R interactions—the underlying monotonic dynamics of the R–N module suggest that R–N interactions ought to be stabilizing when coupled to strong C–R interactions. Further, the underlying increase in stability (i.e., more rapid return to equilibrium) as R–N interaction strength is increased suggests the stabilizing potential of the R–N module ought to increase as the interaction becomes stronger.Fig. 4: R–N stability response to increasing interaction strength.Time series showing R density following a perturbation that lowered R density to 50% of equilibrium density for a low (({a}_{{RN}}=0.8)), b intermediate (({a}_{{RN}}=1)), and c high maximum rate of nutrient uptake (({a}_{{RN}}=2.8)). d Local stability (dominant eigenvalue; ({lambda }_{{max }})) of the R–N subsystem as ({a}_{{RN}}) is increased for ({l}_{N} , > , {l}_{R}), ({l}_{N}={l}_{R}), and ({l}_{N} < {l}_{R}), where ({l}_{R}) and ({l}_{N}) are the rate at which nutrients are lost to the external environment from compartment R and N, respectively. Solid lines are real parts and dashed lines are complex parts of ({lambda }_{{max }}).Full size imageTo look into this conjecture, we first coupled R–N to multiple configurations of strong and expectantly oscillatory C–R interactions and increased R–N interaction strength (({a}_{{RN}})). Following this, we added nutrient cycling and repeated the experiment to demonstrate that our results can be generalized to nutrient-limited ecosystem models. The full equations and parameter values for each model are listed in Supplementary Results 3A–D and 4A, B. We begin with the C–R–N system, where C–R and R–N are coupled through R (Fig. 5a). The initial increase in ({a}_{{RN}}) implicitly strengthens the C–R interaction and fuels the oscillatory potential of C–R and cycles emerge almost immediately after C is able to persist. As ({a}_{{RN}}) is increased further the cycles disappear and we obverse a steep stabilization phase, followed by a modest period of destabilization. Adding a weakly coupled predator gives a similar outcome, although the system continually stabilizes as ({a}_{{RN}}) is increased (Fig. 5b). If the P–C interaction is strengthened (i.e., both C1–R and P–C1 are strong, the recipe for chaos), R–N is unable to dampen oscillations even with a strong interaction strength, although a strong interaction gives tighter bound cycles than a weak interaction (Fig. 5c). We next add a weakly coupled consumer to the nutrient-limited food chain with strong P–C1 and C1–R interactions (Fig. 5d). As seen previously, this interaction draws energy out of the strong pathway, partially muting oscillatory potential. Thus, the ability for a strong R–N interaction to once again return the system to a stable equilibrium is not surprising. Finally, we add a detrital compartment to show that strong R–N interactions remain potent stabilizers in the context of nutrient cycling (Fig. 6b) when compared to a nutrient-limited food chain without nutrient cycling (Fig. 6a).Fig. 5: Nutrient-limited food chain stability.a–d Non-equilibrium dynamics (log10(C1,max/C1,min)) and equilibrium stability (real part of the dominant eigenvalue; ({lambda }_{{max }})) of the C–R–N, P–C–R–N with a single oscillator, P–C–R–N with coupled oscillators, and P–C1–C2–R–N modules, respectively, as ({a}_{{RN}}) is varied.Full size imageFig. 6: Nutrient-limited ecosystem module stability.a, b Non-equilibrium dynamics (log10(Cmax/Cmin)) and equilibrium stability (real part of the dominant eigenvalue; ({lambda }_{{max }})) of the C–R–N nutrient-limited food chain model and the C–R–N–D nutrient-limited ecosystem model, respectively, as ({a}_{{RN}}) is varied.Full size imageNote that we repeat our analysis of higher order modules by implicitly increasing R–N interaction strength through nutrient loading (see Supplementary Results 3E and 4C and Supplementary Figs. 2 and 3). In all cases, increased nutrient loading led to less stable dynamics, consistent with DeAngelis’ (1992) paradox of enrichment finding where increased nutrient loading lead to destabilizing autotroph–herbivore oscillations. More

1.Binney, H. et al. Vegetation of Eurasia from the last glacial maximum to present: key biogeographic patterns. Quat. Sci. Rev. 157, 80–97 (2017).ADS 
Article 

Google Scholar 
2.Clark, P. U. et al. The Last Glacial Maximum. Science 325, 710–714 (2009).ADS 
CAS 
PubMed 
Article 

Google Scholar 
3.Bigelow, N. H. Climate change and Arctic ecosystems: 1. Vegetation changes north of 55°N between the last glacial maximum, mid-Holocene, and present. J. Geophys. Res. 108, https://doi.org/10.1029/2002jd002558 (2003).4.Graham, R. W. et al. Timing and causes of mid-Holocene mammoth extinction on St. Paul Island, Alaska. Proc. Natl Acad. Sci. USA 113, 9310–9314 (2016).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
5.Stuart, A. J. Late Quaternary megafaunal extinctions on the continents: a short review. Geol. J. 50, 338–363 (2015).Article 

Google Scholar 
6.Koch, P. L. & Barnosky, A. D. Late Quaternary extinctions: state of the debate. Ann. Rev. Ecol. Evol. Syst. 37, 215–250 (2006).Article 

Google Scholar 
7.Rabanus-Wallace, M. T. et al. Megafaunal isotopes reveal role of increased moisture on rangeland during late Pleistocene extinctions. Nat. Ecol. Evol. 1, 0125 (2017).Article 

Google Scholar 
8.Mann, D. H., Groves, P., Kunz, M. L., Reanier, R. E. & Gaglioti, B. V. Ice-age megafauna in Arctic Alaska: extinction, invasion, survival. Quat. Sci. Rev. 70, 91–108 (2013).ADS 
Article 

Google Scholar 
9.Capo, E. et al. Lake sedimentary DNA research on past terrestrial and aquatic biodiversity: overview and recommendations. Quaternary 4, https://doi.org/10.3390/quat4010006 (2021).10.Edwards, M. E. et al. Metabarcoding of modern soil DNA gives a highly local vegetation signal in Svalbard tundra. Holocene 28, 2006–2016 (2018).ADS 
Article 

Google Scholar 
11.Hughes, P. D., Gibbard, P. L. & Ehlers, J. Timing of glaciation during the last glacial cycle: evaluating the concept of a global ‘Last Glacial Maximum’ (LGM). Earth Sci. Rev. 125, 171–198 (2013).ADS 
Article 

Google Scholar 
12.Willerslev, E. et al. Fifty thousand years of Arctic vegetation and megafaunal diet. Nature 506, 47–51 (2014).ADS 
CAS 
PubMed 
Article 

Google Scholar 
13.Rasmussen, S. O. et al. A new Greenland ice core chronology for the last glacial termination. J. Geophys. Res. 111, https://doi.org/10.1029/2005jd006079 (2006).14.Mangerud, J. The discovery of the Younger Dryas, and comments on the current meaning and usage of the term. Boreas 50, 1–5 (2020).Article 

Google Scholar 
15.Bauska, T. K. et al. Carbon isotopes characterize rapid changes in atmospheric carbon dioxide during the last deglaciation. Proc. Natl Acad. Sci. USA 113, 3465–3470 (2016).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
16.Wesser, S. D. & Armbruster, W. S. Species distribution controls across a forest‐steppe transition: a causal model and experimental test. Ecol. Monogr. 61, 323–342 (1991).Article 

Google Scholar 
17.Rijal, D. P. et al. Sedimentary ancient DNA shows terrestrial plant richness continuously increased over the Holocene in northern Fennoscandia. Sci. Adv. 7, eabf9557 (2021).18.Birks, H. H. Aquatic macrophyte vegetation development in Kråkenes Lake, western Norway, during the late-glacial and early-Holocene. J. Paleolimnol. 23, 7–19 (2000).ADS 
Article 

Google Scholar 
19.Guthrie, R. D. Origin and causes of the mammoth steppe: a story of cloud cover, woolly mammal tooth pits, buckles, and inside-out Beringia. Quat. Sci. Rev. 20, 549–574 (2001).ADS 
Article 

Google Scholar 
20.Mann, D. H., Peteet, D. M., Reanier, R. E. & Kunz, M. L. Responses of an Arctic landscape to Lateglacial and early Holocene climatic changes: the importance of moisture. Quat. Sci. Rev. 21, 997–1021 (2002).ADS 
Article 

Google Scholar 
21.Ritchie, M. in Competition and Coexistence (eds Sommer, U. & Worm, B.) 109–131 (Springer, 2002).22.Signor, P. W., Lipps, J. H., Silver, L. & Schultz, P. in Geological Implications of Impacts of Large Asteroids and Comets on the Earth vol. 190 (eds Silver, L. T. & Schultz, P. H.) 291–296 (1982).23.Haile, J. et al. Ancient DNA reveals late survival of mammoth and horse in interior Alaska. Proc. Natl Acad. Sci. USA 106, 22352–22357 (2009).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
24.Librado, P. et al. Tracking the origins of Yakutian horses and the genetic basis for their fast adaptation to subarctic environments. Proc. Natl Acad. Sci. USA 112, E6889–E6897 (2015).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
25.Nikolskiy, P. A., Sulerzhitsky, L. D. & Pitulko, V. V. Last straw versus Blitzkrieg overkill: climate-driven changes in the Arctic Siberian mammoth population and the Late Pleistocene extinction problem. Quat. Sci. Rev. 30, 2309–2328 (2011).ADS 
Article 

Google Scholar 
26.Pavlov, P., Svendsen, J. I. & Indrelid, S. Human presence in the European Arctic nearly 40,000 years ago. Nature 413, 64–67 (2001).ADS 
CAS 
PubMed 
Article 

Google Scholar 
27.Kuzmin, Y. V. & Keates, S. G. Siberia and neighboring regions in the Last Glacial Maximum: did people occupy northern Eurasia at that time? Archaeol. Anthropol. Sci. 10, 111–124 (2016).Article 

Google Scholar 
28.Stuart, A. J. & Lister, A. M. Extinction chronology of the woolly rhinoceros Coelodonta antiquitatis in the context of late Quaternary megafaunal extinctions in northern Eurasia. Quat. Sci. Rev. 51, 1–17 (2012).ADS 
Article 

Google Scholar 
29.Chang, D. et al. The evolutionary and phylogeographic history of woolly mammoths: a comprehensive mitogenomic analysis. Sci. Rep. 7, 44585 (2017).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
30.Vartanyan, S. L., Arslanov, K. A., Karhu, J. A., Possnert, G. & Sulerzhitsky, L. D. Collection of radiocarbon dates on the mammoths (Mammuthus primigenius) and other genera of Wrangel Island, northeast Siberia, Russia. Quat. Res. 70, 51–59 (2017).Article 
CAS 

Google Scholar 
31.Rogers, R. L. & Slatkin, M. Excess of genomic defects in a woolly mammoth on Wrangel island. PLoS Genet. 13, e1006601 (2017).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
32.Zimov, S. A., Zimov, N. S., Tikhonov, A. N. & Chapin, F. S. Mammoth steppe: a high-productivity phenomenon. Quat. Sci. Rev. 57, 26–45 (2012).ADS 
Article 

Google Scholar 
33.Yurtsev, B. A. The Pleistocene “Tundra-Steppe” and the productivity paradox: the landscape approach. Quat. Sci. Rev. 20, 165–174 (2001).ADS 
Article 

Google Scholar 
34.Rybczynski, N. et al. Mid-Pliocene warm-period deposits in the High Arctic yield insight into camel evolution. Nat. Commun. 4, 1550 (2013).ADS 
PubMed 
Article 
CAS 
PubMed Central 

Google Scholar 
35.Reimer, P. J. et al. The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0–55 cal kBP). Radiocarbon 62, 725–757 (2020).CAS 
Article 

Google Scholar 
36.Pedersen, M. W. et al. Postglacial viability and colonization in North America’s ice-free corridor. Nature 537, 45–49 (2016).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
37.Slon, V. et al. Neandertal and Denisovan DNA from Pleistocene sediments. Science 356, 605–608 (2017).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
38.Lorenz, M. G. & Wackernagel, W. Adsorption of DNA to sand and variable degradation rates of adsorbed DNA. Appl. Environ. Microb. 53, 2948–2952 (1987).ADS 
CAS 
Article 

Google Scholar 
39.Meyer, M. & Kircher, M. Illumina sequencing library preparation for highly multiplexed target capture and sequencing. Cold Spring Harb. Protoc. 2010, pdb.prot5448 (2010).PubMed 
Article 
PubMed Central 

Google Scholar 
40.Willerslev, E., Hansen, A. J. & Poinar, H. N. Isolation of nucleic acids and cultures from fossil ice and permafrost. Trends Ecol. Evol. 19, 141–147 (2004).PubMed 
Article 

Google Scholar 
41.Alsos, I. G. et al. The treasure vault can be opened: large-scale genome skimming works well using herbarium and silica gel dried material. Plants 9, https://doi.org/10.3390/plants9040432 (2020).42.Hill, M. O. Diversity and evenness: a unifying notation and its consequences. Ecology 54, 427–432 (1973).Article 

Google Scholar 
43.Koleff, P., Gaston, K. J. & Lennon, J. J. Measuring beta diversity for presence-absence data. J. Anim. Ecol. 72, 367–382 (2003).Article 

Google Scholar 
44.Dixon, P. VEGAN, a package of R functions for community ecology. J. Veg. Sci. 14, 927–930 (2003).Article 

Google Scholar 
45.Grootes, P. M. & Stuiver, M. Oxygen 18/16 variability in Greenland snow and ice with 10−3- to 105-year time resolution. J. Geophys. Res. Oceans 102, 26455–26470 (1997).ADS 
CAS 
Article 

Google Scholar 
46.Andersen, K. K. et al. High-resolution record of Northern Hemisphere climate extending into the last interglacial period. Nature 431, 147–151 (2004).ADS 
CAS 
PubMed 
Article 

Google Scholar 
47.Stuiver, M. & Grootes, P. M. GISP2 oxygen isotope ratios. Quat. Res. 53, 277–284 (2017).Article 
CAS 

Google Scholar 
48.Johnsen, S. J. et al. The δ18O record along the Greenland Ice Core Project deep ice core and the problem of possible Eemian climatic instability. J. Geophys. Res. Oceans 102, 26397–26410 (1997).ADS 
CAS 
Article 

Google Scholar 
49.Fuhrer, K., Neftel, A., Anklin, M. & Maggi, V. Continuous measurements of hydrogen peroxide, formaldehyde, calcium and ammonium concentrations along the new grip ice core from summit, Central Greenland. Atmos. Environ. A 27, 1873–1880 (1993).ADS 
Article 

Google Scholar 
50.Mayewski, P. A. et al. Major features and forcing of high-latitude northern hemisphere atmospheric circulation using a 110,000-year-long glaciochemical series. J. Geophys. Res. Oceans 102, 26345–26366 (1997).ADS 
CAS 
Article 

Google Scholar 
51.Alley, R. B. et al. Abrupt increase in Greenland snow accumulation at the end of the Younger Dryas event. Nature 362, 527–529 (1993).ADS 
Article 

Google Scholar 
52.Holden, P. B. et al. PALEO-PGEM v1.0: a statistical emulator of Pliocene–Pleistocene climate. Geosci. Model Dev. 12, 5137–5155 (2019).ADS 
CAS 
Article 

Google Scholar 
53.Karger, D. N. et al. Climatologies at high resolution for the earth’s land surface areas. Sci. Data 4, 170122 (2017).PubMed 
PubMed Central 
Article 

Google Scholar 
54.Martindale, A. et al. Canadian Archaeological Radiocarbon Database (CARD 2.1) (Laboratory of Archaeology at the University of British Columbia, and the Canadian Museum of History, accessed 6 February 2020).55.Vermeersch, P. M. Radiocarbon Palaeolithic Europe database: a regularly updated dataset of the radiometric data regarding the Palaeolithic of Europe, Siberia included. Data Brief 31, 105793 (2020).PubMed 
PubMed Central 
Article 

Google Scholar 
56.Rue, H., Martino, S. & Chopin, N. Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. J. R. Stat. Soc. B 71, 319–392 (2009).MathSciNet 
MATH 
Article 

Google Scholar 
57.Lindgren, F. & Rue, H. Bayesian spatial modelling with R-INLA. J. Stat. Softw. 63, 1–25 (2015).Article 

Google Scholar 
58.Martiniano, R., De Sanctis, B., Hallast, P. & Durbin, R. Placing ancient DNA sequences into reference phylogenies. Preprint at https://doi.org/10.1101/2020.12.19.423614 (2020).59.Wickham, H. ggplot2: Elegant Graphics for Data Analysis (Springer, 2016).60.Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32, 2847–2849 (2016).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
61.Wang, Y. et al. Supporting Data for: Late Quaternary Dynamics of Arctic Biota from Ancient Environmental Metagenomics https://dataverse.no/privateurl.xhtml?token=86979109-5605-43b5-b3fb-f470d85b114c (2021).62.Theodoridis, S. et al. Climate and genetic diversity change in mammals during the Late Quaternary. Preprint at https://doi.org/10.1101/2021.03.05.433883 (2021). More

COMMENT
20 October 2021

Congo Basin rainforest — invest US$150 million in science

The world’s second-largest rainforest is key to limiting climate change — it needs urgent study and protection.

Lee J. T. White

0
,

Eve Bazaiba Masudi

1
,

Jules Doret Ndongo

2
,

Rosalie Matondo

3
,

Arlette Soudan-Nonault

4
,

Alfred Ngomanda

5
,

Ifo Suspense Averti

6
,

Corneille E. N. Ewango

7
,

Bonaventure Sonké

8
&

Simon L. Lewis

9

Lee J. T. White

Lee J. T. White is Minister of Water, Forests, Oceans, Environment, Climate Change and Land-use Planning, Gabonese Republic.

View author publications

You can also search for this author in PubMed
 Google Scholar

Eve Bazaiba Masudi

Eve Bazaiba Masudi is Deputy Prime Minister and Minister of Environment and Sustainable Development, Democratic Republic of the Congo.

View author publications

You can also search for this author in PubMed
 Google Scholar

Jules Doret Ndongo

Jules Doret Ndongo is Minister of Forestry and Wildlife, Republic of Cameroon.

View author publications

You can also search for this author in PubMed
 Google Scholar

Rosalie Matondo

Rosalie Matondo is Minister of Forest Economy, Republic of the Congo.

View author publications

You can also search for this author in PubMed
 Google Scholar

Arlette Soudan-Nonault

Arlette Soudan-Nonault is Minister of Environment, Sustainable Development and the Congo Basin, Republic of the Congo.

View author publications

You can also search for this author in PubMed
 Google Scholar

Alfred Ngomanda

Alfred Ngomanda is director of the National Centre for Scientific Research and Technology (CENAREST), Gabonese Republic.

View author publications

You can also search for this author in PubMed
 Google Scholar

Ifo Suspense Averti

Ifo Suspense Averti is an associate professor in tropical forest ecology at Marian Ngouabi University, Republic of the Congo.

View author publications

You can also search for this author in PubMed
 Google Scholar

Corneille E. N. Ewango

Corneille E. N. Ewango is a professor of tropical forest ecology and management at the University of Kisangani, Democratic Republic of the Congo.

View author publications

You can also search for this author in PubMed
 Google Scholar

Bonaventure Sonké

Bonaventure Sonké is a professor of plant systematics and ecology at the University of Yaounde I, Republic of Cameroon.

View author publications

You can also search for this author in PubMed
 Google Scholar

Simon L. Lewis

Simon L. Lewis is professor of global change science at University College London and the University of Leeds, UK.

View author publications

You can also search for this author in PubMed
 Google Scholar

Share on Twitter
Share on Twitter

Share on Facebook
Share on Facebook

Share via E-Mail
Share via E-Mail

Download PDF

A warden with an orphaned mountain gorilla in the Virunga National Park sanctuary in the Democratic Republic of the Congo.Credit: Phil Moore/AFP/Getty

Earth’s second-largest expanse of tropical forest lies in central Africa, in the Congo Basin. The region supports the livelihoods of 80 million people. The rainfall that the forest generates as far away as the Sahel and the Ethiopian highlands supports a further 300 million rural Africans. These forests are crucial to regulating Earth’s climate, and are home to forest elephants, gorillas and humans’ closest relatives, chimpanzees and bonobos.Such services to people and the planet are not guaranteed, given rapid climate change and ongoing development in the region. The forest’s ability to absorb carbon dioxide is slowing as temperatures rise1. Deforestation, although lower than elsewhere in the tropics over recent decades, has led to the loss of more than 500,000 hectares of forest in 2019 alone (see go.nature.com/3dnxm9e). Without new policies, this is expected to increase.Yet, too often, central Africa’s rainforests are ignored or downplayed. The Congo Basin forests receive much less academic and public attention than do those in the Amazon and southeast Asia. Between 2008 and 2017, the Congo Basin received just 11.5% of international financial flows for forest protection and sustainable management in tropical areas, compared with 55% for southeast Asia and 34% for the Amazon region2.The area is neglected even by comparison with the rest of Africa. For example, a key UK-funded programme of climate research, called Future Climate for Africa, invested £20 million (US$27 million) in modelling and four projects focused on eastern, western and southern Africa. None focused on the Congo Basin or central Africa.The result of this neglect is clear in high-level climate assessments. Central Africa was one of only two regions worldwide without enough data for the Intergovernmental Panel on Climate Change to assess past trends in extreme heat in its 2021 Working Group I report (the other was the southern tip of South America).
A collaborative look at the Congo Basin
We are a group of ministers who have responsibility for forests in the region, and scientists who work on the ground and advise governments. Together we call for a Congo Basin Climate Science Initiative. This should comprise a $100-million, decade-long programme of research, tied to a separate $50-million fund to train Congo Basin nationals to become PhD-level scientists. Such funding would transform our understanding of these majestic forests, providing crucial input for policymakers to help them enact policies to avoid the region’s looming environmental crises.There is precedent for such a transformation. In the mid-1990s, rainforest science in the Amazon region was limited and was largely conducted by overseas scientists. Formally beginning in 1998 and led by Brazilians, the Large-Scale Biosphere-Atmosphere Experiment in Amazonia programme, known as the LBA, was a 10-year, $100-million effort. It revolutionized understanding of the Amazon rainforest and its role in the Earth system.The LBA involved 6 years of intensive measurements and covered climatology, hydrology, ecology and biogeochemistry across an area of 550 million hectares. It comprised 120 projects and 1,700 participants, 990 of whom were Brazilians3. One of its greatest legacies was the creation of a new cadre of Brazilian researchers. Two decades on, Brazil is now widely acknowledged as the world’s leading nation for tropical forest monitoring, and is at the forefront of rainforest science.We should — we must — do the same for central Africa.Known unknownsThe Greater Congo Basin covers some 240 million hectares of contiguous forests, straddling 8 nations (see ‘Earth’s second green lung’). Merely sampling this vast area is daunting. Access often requires days of travel in dugout canoes and long treks through the humid jungle, punctuated by wading through swamps. There is also a pervasive prejudice: too many people think working in the Congo Basin region is perilous, whether the hazards are political instability, unfamiliar diseases or dangerous animals. In reality, for the vast majority of central Africa, the risks are similar to working in the Amazon rainforest or east African savannah ecosystems.

Source: Ref. 1

These various challenges can be surmounted. Papers from the past few years, co-authored by many of us, highlight how important and understudied the region is. In 2017, the world’s largest tropical peatland complex was mapped for the first time — an area spanning 14.6 million hectares in the heart of the Congo basin4. This work radically shifted our understanding of carbon stores in the region. In March 2020, an international consortium showed that Africa’s rainforests annually absorb the same amount of carbon1 as was emitted each year by fossil-fuel use across the entire African continent in the 2010s5.In December 2020, a striking 81% decline in fruit production over 3 decades in an area of forest in Gabon was shown to coincide with climate warming and an 11% decline in the body condition of forest elephants (they rely on fruit for part of their diet)6. And in April, the first region-wide assessment of tree community composition in central Africa was published7, mapping areas that are vulnerable to climate change and human pressures.
Biodiversity needs every tool in the box: use OECMs
Overall, the strikingly recent (although somewhat limited) data suggest that the tropical forests of the Congo Basin are more carbon-dense8, more efficient at slowing climate change1 and more resistant to our changing climate9 than are Amazon tropical forests. But we do not know how increasing droughts, higher temperatures, selective logging and deforestation might interact — including the possibility of reduced rainfall in the Sahel10 and Ethiopian highlands11. Some 2,500 years ago, vast swathes of the Congo Basin forests were lost during a period of climate stress, but researchers do not understand the historic context of that event, nor the likelihood of a repeat12.Little is known about the region because not enough science is done in central Africa. Remarkably, researchers still do not understand the basic principles of why different types of forest occur where they do in the Congo Basin. Climate models for this region are poor, both because of the complex interplay of Atlantic, Indian and Southern ocean influences and because of a lack of local climate data. Without more data and more specialists, it is impossible to make reliable predictions of these forests’ responses to changes in climate and land use.Next stepsInvestment in basic science is urgently needed to fill these gaps. A Congo Basin Climate Science Initiative should focus on three important overarching questions: how does the Congo Basin currently operate as an integrated system? How will changes in land use and climate affect its function? And how sustainable are different options for development?Within these broad topics are more specific questions that politicians will need answers to if nations are to achieve net-zero CO2 emissions by 2050. One such question is how much carbon is stored in vegetation and soils. These and other quantities must be reported as part of countries’ commitments to the 2015 Paris climate agreement. At present, most central African countries rely on default values, which could be way off the mark. A recent paper13 on African montane forests largely near the edges of the basin, for example, showed that measured carbon storage values were 67% higher than the default values.

A child on the Mongala River in the dense forest of the Democratic Republic of the Congo.Credit: Pascal Maitre/Panos

A science initiative will work only if there is enthusiasm and leadership from researchers and active support from key Congo Basin countries, alongside buy-in from funders. We envision three steps to achieve these aims.First, scientists from the Congo region should hold a workshop with the LBA architects and participants to assess lessons from the Amazon region. This south–south cooperation would build a scientist-led framework to address the crucial research questions.Second, a meeting of politicians and advisers from the region would facilitate discussions of the policy-relevant questions that scientists should investigate. This would be led by Cameroon, the Democratic Republic of the Congo, Gabon and the Republic of the Congo — the four nations conducting the most research in the region. The meeting will help to lock in political support across ministries responsible for forests, environment, water, climate, science and universities.
Nature-based solutions can help cool the planet — if we act now
Third, partners will need to develop an overarching science programme that is acceptable to funders. Such a programme will probably include scaling up many efforts that are already under way, but which are currently insufficient in scope or unreliably funded. This would speed up scientific progress.For example, a handful of established, long-term field sites already exist in the Greater Congo Basin, including in Lopé National Park in Gabon and in the Yangambi Biosphere Reserve in the Democratic Republic of the Congo. These ‘supersites’ are sophisticated field stations with full-time staff who collect reliable, long-term data sets on vegetation, animals and the physical environment, including greenhouse-gas fluxes at Yangambi. But the sites are too few in number, and they rely on the heroic efforts of local champions. There should be a dozen or so locations across the region, with consistent funding to support complex research projects.Similarly, the African Tropical Rainforest Observation Network (AfriTRON), established in 2009, tracks every tree in permanent sample plots to estimate the carbon balance of undisturbed forests. Although this observatory has ramped up from its original 40 sites in central Africa to more than 200 today, these cover just 250 hectares of the roughly 240-million-hectare total. That is very sparse sampling from which to draw regional conclusions.Meanwhile, the Forest Global Earth Observatory (ForestGEO), established in 1990 to understand how tropical forests maintain such a diverse number of tree species, has established just 4 sites in central Africa in 30 years, with none in the centre of the basin. There is an obvious need for expansion.

African forest elephants in Ivindo National Park, Gabon.Credit: Amaury Hauchard/AFP/Getty

Finally, the 2016 AfriSAR airborne field campaign, a collaboration between NASA, the European Space Agency and the Gabonese Agency for Space Studies and Observation, showed how to combine different data sets to carefully map forest types and their carbon stocks in Lopé National Park in Gabon. This model could be replicated elsewhere in the basin.All of this work will require linking theory, observations, experiments and modelling. It should attract a diversity of leading international experts to focus on Africa and provide training to Congo Basin nationals. A $100-million research programme would provide new opportunities and much-needed career options for African scientists. The tied investment of $50 million, focused on building talent, could produce approximately 200 PhDs awarded by leading universities worldwide. This would create a new generation of scientists, including future leaders, from central Africa. The training programme would ensure the necessary step-change in science capacity, and provide opportunities for young African researchers who currently find it hard to compete for international scholarships, which are often won by students from Asia or South America.Agreeing on open access for all the data collected, as in the LBA programme, will significantly boost the initiative’s science impact.Money well spentThis $150-million science programme over 10 years needs investors. One option would be to combine funds from governments that have made large forest- and science-related investments in the Congo Basin in the past, notably Belgium, France, Germany, Norway, the United Kingdom, the United States and the European Union. Alternatives include United Nations agencies, international climate funds and private philanthropy organizations. Such a programme should be high on funders’ agendas, given the UN Sustainable Development Goals (SDGs). These include raising capacity for effective climate-change-related planning and management (SDG13), increasing financial resources to conserve and sustainably use biodiversity and ecosystems (SDG15), boosting the number of researchers in lower-income countries, and increasing research and development (R&D) funding (SDG9), all before 2030.Global R&D funding was $2.2 trillion in 201914. Thus, investing $150 million over a decade to better understand and protect the world’s second-largest extent of tropical forest is modest. To put this sum in context, the US government’s total projected cost for the Human Genome Project was $2.7 billion, and the European Space Agency spends approximately $500 million on its larger, long-lasting scientific satellites. The $100 million that the LBA brought to the Amazon in the 1990s is equivalent to about $160 million in today’s terms.
Ethiopia, Somalia and Kenya face devastating drought
The investment in science will pay for itself many times over. Consider just the role of forests as reservoirs of zoonotic diseases. Better forest management lowers the risk of disease outbreaks, let alone a pandemic15.Critics might argue that direct interventions in development aid are more urgent than investing in climate and ecological science. However, these funds are usually independent and do not compete. Furthermore, the old division between ending poverty and protecting the environment no longer applies: Africans will suffer disproportionately if temperatures are not limited as per the Paris agreement. That must include protection of the forests of the Congo Basin.Further efforts could help to support the goals of the Congo Basin science programme. For example, there is a lack of economic models that show how standing forests can become more valuable than converted landscapes. Developing these would support policy decisions to maintain forest cover.There are also several efforts under way to improve forest management that aim to empower local people, increase income and protect the environment. These include the transfer of land-management decisions to local populations, such as through community forestry, and creating high-value end products from selective logging rather than relying on the export of raw, unprocessed timber. A new science initiative could assess various approaches to understand what works best.We know so little about the majestic forests of central Africa. A Congo Basin Climate Science Initiative would curb our collective ignorance. A lack of investment is the barrier to safeguarding these precious ecosystems. Surmount this, and the future of Earth’s second ‘great green lung’ will be brighter.

Nature 598, 411-414 (2021)
doi: https://doi.org/10.1038/d41586-021-02818-7

References1.Hubau, W. et al. Nature 579, 80–87 (2020).PubMed 
Article 

Google Scholar 
2.Atyi, R. E. et al. International Financial Flows to Support Nature Protection and Sustainable Forest Management in Central Africa (Central Africa Forest Observatory, 2019).
Google Scholar 
3.Lahsen, M. & Nobre, C. A. Environ. Sci. Policy 10, 62–74 (2007).Article 

Google Scholar 
4.Dargie, G. C. et al. Nature 542, 86–90 (2017).PubMed 
Article 

Google Scholar 
5.Ayompe, L. M., Davis, S. J. & Egoh, B. N. Environ. Res. Lett. 15, 124039 (2020).Article 

Google Scholar 
6.Bush, E. R. et al. Science 370, 1219–1222 (2020).PubMed 
Article 

Google Scholar 
7.Réjou-Méchain, M. et al. Nature 593, 90–94 (2021).PubMed 
Article 

Google Scholar 
8.Lewis, S. L. et al. Phil. Trans. R. Soc. B 368, 20120295 (2013).PubMed 
Article 

Google Scholar 
9.Bennett, A. C. et al. Proc. Natl Acad. Sci. USA 118, e2003169118 (2021).PubMed 
Article 

Google Scholar 
10.Salih, A. A. M., Zhang, Q. & Tjernström, M. J. Geophys. Res. Atmospheres 120, 6793–6808 (2015).Article 

Google Scholar 
11.Gebrehiwot, S. G. et al. WIREs Water 6, e1317 (2019).Article 

Google Scholar 
12.Malhi, Y. Proc. Natl Acad. Sci. USA 115, 3202–3204 (2018).PubMed 
Article 

Google Scholar 
13.Cuni-Zanchez, A. et al. Nature 596, 536–542 (2021).PubMed 
Article 

Google Scholar 
14.Sargent, J. F. Global Research and Development Expenditures: Fact Sheet R44283 (Congressional Research Service, 2021).15.Everard, M., Johnston, P., Santillo, D. & Staddon, C. Environ. Sci. Policy 111, 7–17 (2020).PubMed 
Article 

Google Scholar 
Download references

Competing Interests
L.J.T.W. (Gabon), E.B.M. (Democratic Republic of the Congo), J.D.N. (Cameroon), R.M. (Republic of the Congo) and A.S-N. (Republic of the Congo) are ministers of forests and/or the environment. Their countries stand to benefit if international donors take on board the recommendations in this Comment article.

Related Articles

Nature-based solutions can help cool the planet — if we act now

Biodiversity needs every tool in the box: use OECMs

Ethiopia, Somalia and Kenya face devastating drought

A collaborative look at the Congo Basin

Subjects

Conservation biology

Climate change

Ecology

Latest on:

Climate change

Before making a mammoth, ask the public
World View 20 OCT 21

Why fossil fuel subsidies are so hard to kill
News Feature 20 OCT 21

The broken $100-billion promise of climate finance — and how to fix it
News Feature 20 OCT 21

Ecology

Rhinoceros genomes uncover family secrets
News & Views 19 OCT 21

A toxic ‘tide’ is creeping over bountiful Arctic waters
Research Highlight 06 OCT 21

Spatiotemporal origin of soil water taken up by vegetation
Article 06 OCT 21

Jobs

Community Practice Pathologist (AP/CP)

University of Vermont (UVM)
Plattsburgh, NY, United States

Postdoctoral Fellowship in Vascular Biology at Boston Children’s Hospital and Harvard Medical School

Boston Children’s Hospital (BCH)
Boston, MA, United States

Assistant Professor/Associate Professor of Physiology and Cellular Biophysics (Tenure-Track)

Columbia University in the City of New York (CU)
New York, United States

Associate or Senior Editor, Nature Biomedical Engineering

Springer Nature
New York, NY, United States More

Trends in global biomass C stocksThe CRAFT model reliably reproduces the observed trends in primary and managed forest biomass C stocks (including both above-ground and belowground biomass) in 1990–2020 with a relative root mean square error (RMSE) of 0.57% between simulated and observed biomass C stocks by the FRA2 at the global level. These low divergences between stock estimates result, however, in global C emissions c. 2 times lower according to the CRAFT simulations than the estimates derived from the FRA (Supplementary Table 2). Still, the CRAFT simulations corroborated the FRA observations while adding information on annual estimations of forest C stocks, rather than the 5-years interval data provided by the FRA (Fig. 1a), and dynamic annual net C emissions (Fig. 1c, d) from managed and primary forests. The five sensitivity analyses carried out on the most uncertain model inputs and assumptions (see ‘Methods’ descriptions and Supplementary Figs. 5–10) confirmed the results presented in Fig.1: The largest deviation derived from the sensitivity analysis considering forest gross instead of net area changes results in a relative RMSE of 1.94% with global C emissions c.2.5 times higher than the FRA estimates (Supplementary Table 2). The simulations from the reference model assumptions yield the best RMSE and closest agreement with the C budgets derived from the FRA, indicating that they are the most optimal.Fig. 1: Global trends in total, primary, and managed forests.a Forest biomass C stocks (GtC); b cumulated change in forest area (Mha; negative values indicate area loss); c cumulated C net emissions (GtC; positive values indicate a C source while negative values indicate a C sink); and d cumulated net change in C-stock densities (tC/ha). See Supplementary Fig. 1 for annual fluxes.Full size imageIn line with the FRA data, we find here that the main trend is a loss of total biomass C stocks following three phases: increase in annual emissions, stagnation and slight recovery of C stocks, resulting in net C emissions from forest biomass (Fig. 1c) by 0.74 GtC or 0.03 GtC/yr between 1990 and 2020, contrasted by an opposite trend of increasing biomass density from 70 to 73 tC/ha in total forest (Fig. 1b, d). These figures are within the range of the estimated sink in forest soil and biomass of 0.1 ± 7.3 GtC/yr in 2001–2019 found by Harris et al.17. Our estimation is also consistent with that of Tubiello et al.3 of 0.11 GtC/yr net C emissions from forest ecosystems. A comparison3 of FRA-derived global forest C emissions with other independent estimates reported in 1990–2015 by National Greenhouse Gas Inventories (NGHGIs)—including the Russian Federation, the USA, China, Indonesia, and India—and by the United Nations Framework Convention on Climate Change for other countries (UNFCCC, 202018) yields a slight difference of c. 18%, although the UNFCCC and NGHGI’s account, by definition, only for emissions from managed land3. Further independent comparisons at the national and macro-regional levels are compiled in Supplementary Table 1 and reveal that C emissions estimated in the present study are in good agreement with other research.Here we find that the net C emissions mostly arise from primary forests, which undergo area loss, but also biomass thickening (Fig. 1b, d). By contrast, in spite of area loss, managed forests act as C-sinks following biomass thickening (Fig. 1b, d). Increasing biomass density is therefore key to counteract net C emissions from forest biomass in 1990–2020. While both harvest rate and burnt area increase globally over the period of observation, the increased forest growth rate that we calculate with CRAFT for both primary and managed forests over 1990–2020 emerges here as the only factor explaining increased biomass density at the global level. This is in line with other research pointing to the relevance of biomass thickening for forest C sequestration19. In addition, our finding that the forest growth rate increased annually by 0.19%, 0.21%, and 0.21% from 1990 to 2020, respectively, for primary, managed and total forests of the world is consistent with Kolby Smith et al.20 who find that also net primary production (NPP) increased annually between 0.10 and 0.25% in the period 1982–2011, as well as with other modeling and remote-sensing studies documenting a global greening trend, i.e., vegetation thickening following increased vegetation growth rate21,22. Note that estimates of annual growth rate increase in 1990–2020 by the sensitivity analyses provide narrow ranges of 0.17–0.19, 0.21–0.23, and 0.20–0.22%, respectively, for primary, managed, and total forests of the world (Supplementary Table 2).Proximate drivers of net C emissionsWe develop six counterfactual scenarios23,24,25 in order to investigate how forest biomass density and forest biomass C stocks would evolve in the hypothetical absence of (i) changes in harvest (CF1); (ii) changes in forest growth rates (CF2); (iii) change in burnt area (CF3); (iv) change in forest area (CF4); (v) harvest (CF5); (vi) burnt area (CF6) (see “Methods” section). The comparison of observed and simulated counterfactual trends allows us to isolate and quantify the influence of these four main drivers on global forest C-stock changes at national resolution (CF1 to 4) as well as to quantify the overall effects of total wood extraction and burnt area (CF5 and 6).At the global level, we find that loss of forest area (CF4) is the main driver of the net C emissions from forest biomass (Fig. 2a). In the absence of changes in area, global forest biomass would act as a cumulative net C sink of c. 26.9 GtC in the study period, creating a difference of 27.6 GtC between the actual and the CF4 C budget. This effect in the absence of area change, however, is a composite of an additional C sink of 30.7 in deforesting countries and an additional C source of 3.8 GtC in reforesting countries. Changes in harvest and burnt area from 1990 to 2020 also drove net C emissions from global forest biomass as emissions drop by c. 5.7 and 1.4 GtC in the respective counterfactual scenarios, thus generating net C-sinks of c. 4.9 and 0.63 GtC (Fig. 2a). These figures are in stark contrast with the estimated total sink of c. 49.1 and 5.4 GtC that would emerge in the hypothetical absence of harvest (CF5) and burnt area (CF6; Fig. 2a), respectively. Only changes in forest growth rates counteract the net C emissions from global forest biomass (CF2; Fig. 2a). In the absence of changes in forest growth rates, global forests would act as net C source of c. 7.4 GtC in 1990–2020, i.e., c. 10 times the actually observed source. This net effect in the absence of growth rate change results from an additional C source of 30.4 in countries experiencing growth rate increase and an additional C sink of 23.0 GtC in countries experiencing growth rate decline.Fig. 2: Counterfactual scenarios (1990–2020) assessing the cumulative impact of: changes in harvest (CF1); changes in forest growth rate (CF2); changes in burnt area (CF3); changes in forest area (CF4); total harvest (CF5); and total fire (CF6) on C-dynamics.Panels (a) and (b) show the global country-level gross and net CF C budgets (GtC) and changes in biomass density (tC/ha), respectively, with negative (red) and positive values (blue) indicating net emissions and sinks, respectively, error bars indicate the range of C budgets estimated across the five sensitivity analyses performed to test the model robustness (see Supplementary Fig. 5 for additional figures showing the net difference between CF and actual C budgets and changes in biomass density, Supplementary Table 3 and Supplementary Fig. 5 for results from sensitivity analyses). Maps show the effects of c CF1; d CF2; e CF3; f CF4; g CF5; h CF6, and are represented as the % of actual biomass C stocks that would be reached in each CF in 2020. Values above 100% (red) indicate that actual change result in net C emissions while values below 100% (blue) indicate that actual change result in a net C sink.Full size imageA sensitivity analysis on the potential underestimation of C-dynamics resulting from the use of net area change data at country level (see “Methods” section and Supplementary Fig. 5) reveals that accounting for gross area changes26 instead of net area change would result in higher global C emissions estimates (4.19 GtC in the sensitivity test versus 0.74 GtC in the reference simulation) but would reveal the same patterns of forest C-dynamic drivers (Supplementary Fig. 5). However, the magnitude of the main drivers would be slightly changed with a lower effect of changes in area (C sink in the hypothetical absence of area changes reaching 20.8 GtC in the sensitivity tests versus 26.9 GtC in the reference assessment) and a higher effect of growth rate changes (C source in the hypothetical absence of growth rate changes reaching 13.1 GtC in the sensitivity tests versus 7.4 GtC in the reference assessment). Generally, the range of results derived from the five sensitivity analyses does not change the relative importance of the individual drivers in any of the scenarios (Fig. 2a, Supplementary Table 3, and Supplementary Fig. 5). However, the sensitivity analyses highlight that the uncertainty is large enough to reverse the cumulated C signal in the absence of changes in harvest (CF1), changes in burnt area (CF3), and the complete absence of burnt areas (CF6). By contrast, the signals of CF2 (no growth rate change), CF4 (no area change), and CF5 (no harvest) are larger than the uncertainty across sensitivity analyses, signaling that our findings on these drivers are most robust.The global trends displayed in Fig. 2a, b are the combined results of diverging national forest dynamics (Fig. 2c, h). In particular, shifts in forest area (CF4) contribute to global net C emissions only in the Global South, excluding Vietnam, India, and Chile (Fig. 2f). The impacts of changes in burnt area and harvest are similarly heterogenous, with considerable effects only in some regions (e.g., Vietnam, Mozambique, Fig. 2c, e). In contrast, changes in forest growth rates are more ubiquitous, mainly positive (leading to C-sinks) for most countries, with a few notable exceptions, mainly in arid or boreal regions (e.g., India, Spain, Argentina, Canada; Fig. 2d). Possible reasons explaining the negative effect of change in forest growth rate are forest degradation, increasing drought, cloudiness, or insect outbreaks15,16,17,18,19. Over the period 1990–2020, the strongest harvest impacts are observed in countries with large area of managed forest and high harvest pressure, mostly located in temperate and subtropical areas (CF5; Fig. 2g), while fire impacts are strong in only a few countries (CF6; Fig. 2h).The fact that we use here country-level data comes both with limitations and advantages. The main limitation associated with national data is that it conceals gross C fluxes in forest biomass dynamics and blurs heterogeneity in growth conditions and anthropogenic management within countries. The country-level resolution aggregates the effects of manifold, partly counteracting processes at the local level—including photosynthesis, maintenance respiration, growth respiration, as well as forest area loss and expansion—on the annual dynamic of primary and managed forest biomass. As a consequence, our optimization of the growth function actually reflects apparent national growth rates resulting from the aggregate of these processes. However, this simplification of forest ecosystem functioning is also an advantage. Our approach reproduces forest biomass dynamics very accurately, which is complementary to most process-based models aimed at depicting biological processes and their abiotic controls27 but providing a wide range of C flux estimations1 and hardly reproducing observation from inventory data1,28,29. By contrast, the strength of the modeling approach implemented here is that it can be run with parsimonious data availability and allows to disentangle the major drivers behind forest C-stock and flux trajectories.Typology of forest biomass changeIn order to identify spatial and temporal patterns of drivers in forest biomass trends, we establish a typology of the main drivers over the period 1990–2020 (Fig. 3b). The typology we established is based on the positive versus negative shift in biomass C stocks, and highlights the most important driver of this shift as assessed through the counterfactual assessment, irrespective of the relative importance of the other drivers shown in Fig. 2. However, as the early separation between increasing and decreasing biomass C stocks in the decision tree (Fig. 3b) may conceal the effect of a major driver counteracting the observed C dynamic, the typology also accounts for possible antagonistic effects by identifying cases in which the main driver of observed C-dynamics is not, in absolute terms, the most important driver (e.g., C stocks increase but the driver with the strongest absolute effect counteracts this positive budget, see also Supplementary Fig. 3). By pinpointing the major drivers of forest change at national levels, such an approach enables to identify major levers for forest conservation.Fig. 3: Main drivers of the net C emissions from forest biomass.a Applied at the national level to the 1990–2020 period; b established according to a Boolean typology using the results from the counterfactual scenario assessment as criteria; c enabling to calculate the sum of net C-sinks and net C sources in each type of forest C-dynamics trajectory identified through the typology, error bars indicating the range of C-sinks and sources by main driver estimated across the five sensitivity analyses, with black and gray bars standing, respectively, for solid and hatched countries (see Supplementary Figs. 6–7 for results from sensitivity analyses). The hatches on the countries (a), typology (b), and bar chart (c) stand for cases in which the driver with the strongest effect actually counteracts the observed carbon budget. The color of the hatches corresponds to the main factor identified by the decision tree algorithm. Abbreviation on the typology: E: C sink driven by forest area Expansion; LH: C sink driven by Lower Harvest; FR: C sink driven by Fire Reduction; EG: C sink driven by Enhanced Growth rate; DG: C source driven by Declining Growth rate; FI: C source driven by Fire Increase; HH: C source driven by Higher Harvest; D: C source driven by Deforestation; NS: non-significant change.Full size imageDeforestation was the dominant driver of net C emissions from forest biomass in most countries of South America and Sub-Saharan Africa, corroborating findings from the literature11,30,31 (Fig. 3a, c). The net C emissions by countries where deforestation is the most significant driver reach c. 21.3 GtC, with only 0.3 GtC of these emissions being counteracted by another major driver (either increased growth rate or lower harvest pressure). These emissions represent c. 92.7% of the 21.9 GtC net emissions arising from all countries acting as net C sources (Fig. 3c). Changes in forest growth rates act as the primary drivers in most countries experiencing a net C sink over the period (Fig. 3a, c). The net C-sinks by countries where changes in forest growth rates are the main driver reach c. 16.4 GtC, with 0.9 GtC of these sinks being counteracted by another major driver (increased harvest pressure in all cases except for Sudan where area loss was the major driver counteracting the C sink). These C-sinks mainly driven by increased growth rate represented c. 77.5% of the 21.1 GtC net sink created by all countries acting as net C-sinks (Fig. 3c).Forest area expansion from 1990 to 2020 is the main driver of forest biomass net C sink in only a few Northern countries but also some Southern countries, namely Vietnam, India, and Chile, in line with findings reported for these countries32,33,34, all together accounting for a net C sink of 3.9 GtC. However, more than half of the C-sinks mainly driven by reforestation are counteracted by another major driver (either declining forest biomass growth rate or increased harvest pressure). Similarly, changes in harvest as well as changes in burnt areas are the main drivers of net C sink or source for a handful of countries in 1990–2020 (Fig. 3a). Finally, declining forest biomass growth rate is the primary driver of net C emissions only in Mongolia and Canada, which is consistent with other studies highlighting slower growth, higher mortality, and insect outbreak events in Canadian forests35,36,37.These highlights derived from the typology remained the same in all sensitivity analyses (Supplementary Figs. 6–7), despite some possible changes in country type identification (Fig. 3a and Supplementary Fig. S6) and amplitude shifts in the attribution of main drivers globally (Fig. 3c and Supplementary Fig. 7). The ranges of values in the attribution of main drivers result from the previously reported differences between the counterfactual and actual C budget estimates across sensitivity analyses (see also Supplementary Tables 2–3) combined with some changes in the type of forest C-dynamics trajectory identified through the typology in countries with large forest biomass stocks: China, India, and Australia (Supplementary Note 1 and Supplementary Fig. 6). However, these shifts do not affect the main conclusions derived from Fig. 3c: in all sensitivity analyses, growth rate changes remain the main driver of global forest biomass C sink with total net C-sinks in countries where increasing growth rate is the main driver (including both solid and hatched countries) ranging from 12.1 to 21.1 GtC, while afforestation always holds the second place of global C sink driver (total net C-sinks in countries where afforestation is the main driver ranging from 2.4 to 7.7 GtC). Similarly, total net C sources by countries where deforestation is the main driver range from −21.9 to −14.0 GtC, thus highlighting that deforestation would by far remain the main driver of forest biomass C emissions across all sensitivity analyses.Implications for forest-based solutionsOur results allow to identify major mechanisms behind observed forest biomass C changes that are immediately relevant for forest-based climate-change-mitigation strategies. We show that deforestation, increasing harvest, and burnt area have driven the net C emissions from forest biomass over the last three decades. Deforestation is the dominant driver, corroborating that protection from deforestation is indispensable1,11,38. On the other hand, forest growth rate is identified as the major driver counteracting net C emissions (Fig. 2a, d). In fact, most of the temperate and boreal countries, with the noteworthy exception of Canada, fall under a type in which enhanced forest growth rate is the major driver of a net C sink (Fig. 3b). Besides, even countries dominated by deforestation in the tropics show significant increases in growth rate (Figs. 2d and 4). These results highlight that enhanced growth rate, rather than reforestation, is the main driver counteracting biomass C emissions in 1990–2020.Fig. 4: Change in forest growth rate and its effects on global carbon stocks.The diagrams show national forest growth rate changes (y-axis) scaled along the cumulated size of the carbon stock in 1990 (x-axis). The area between the graph and the x-axis indicates the C-stock change due to growth rate for total (a), primary (b), and managed forests (c) (see Supplementary Figs. 8–10 for results from sensitivity analyses).Full size imageThese increases in forest growth rate may arise from diverse processes, including climatic and land-use drivers. On the one hand, several studies highlight the effects of environmental drivers—such as warming, atmospheric carbon dioxide (CO2), and nitrogen (N) fertilization1,6,8,11,21,39—on the terrestrial C sink. On the other hand, changes in forest growth rate can also be driven by shifts in forest management practices, such as tree species selection, forest recovery from past degradation and lesser litter grazing12,40,41. Advancing the understanding of the underlying processes of forest growth rate change is key for forging climate-change-mitigation strategies, but it is not straightforward to isolate climatic (e.g., altered CO2 concentration or temperature) from land-use drivers (e.g., non-timber forest uses such as grazing)42. Still, a comparison of trajectories in primary and managed forest growth rate change based on our results allows to derive insights into the interplay of these different drivers (Fig. 4 and Supplementary Fig. 3). From the fact that only 11% of primary forest carbon stocks show declining growth rate trends (Fig. 4c) while a relatively larger carbon stock in managed forest (22%) is affected by declining growth rate trends (Fig. 4b), we can infer that in overall terms—and assuming primary and managed forests of a given country to be similarly affected by climatic drivers —land use is likely to exert a degrading effect on growth rate dynamics. Nevertheless, some countries reveal declining growth rate in primary forest but increasing growth rate in managed forest, thus suggesting that forest management may have an improving effect on forest growth rate in those countries (e.g., USA, Fig. 4b, c, see also Supplementary Fig. 4). In overall terms, this result suggests that globally a reduction of forest use may have the potential to enhance growth rate, thus corroborating previous findings by Quesada et al.14. However, these interpretations warrant a caveat that primary versus managed forest growth rate changes are derived from the FRA data and a state-of-the-art of the literature on changes in primary forest density (see “Methods” section and Supplementary Note 2), the latter being associated with higher uncertainties although the corresponding sensitivity analysis testing suggests these uncertainties to have little impact on the figures displayed here (see Supplementary Tables 2-3 and Supplementary Figs. 5–10).Independent of their origin (management or climate driven), the future trajectories of this driver, forest growth rate, is subject to large uncertainties43,44,45. Research suggests that increasing forest growth rate is a transient phenomenon and might be discontinued in the future46. For instance, several recent studies have pointed toward the saturating effect of CO2 fertilization, which is suspected to be a key process underlying vegetation greening and ensuing thickening21, the risk of increasing mortality and slower growth rate following increasing drought6,47,48, temperature49, and natural disturbances such as insect outbreaks50,51. Even more recently, Duffy et al.52 showed that, in the near-future, temperature increases from business-as-usual trajectories of climate change shall result in a severe reduction, and possibly a reversal, of the terrestrial C sink, despite the remaining unknowns.Therefore, we conclude that, while increasing forest growth rate is the dominant driver counteracting the global net C emissions from forest biomass in the past three decades, it is against a precautionary principle to forge climate strategies that rely on a continuous net C sink effect from the same processes in the future. By contrast, our results suggest that reducing wood harvest (Fig. 2g) and halting deforestation (Fig. 2c) are key strategies to address the challenge of climate-change mitigation. In this context, increasing forest harvest volumes—a strategy often promoted in the course of climate-change-mitigation efforts embraced as the “bioeconomy”—appears to have critical unintended side-effects, despite the potential of wood for substituting some emissions-intensive products and processes53,54,55: by not only reducing the carbon sink function in forests, but also accelerating the overall C turnover rates through rejuvenation of forests and transfer to harvested wood products of lifetimes shorter than those of old-growth forests56,57,58, such strategies result in a critical loss of C sink capacity. Overall, our results plead for a double strategy to enable future forest-based solutions for climate-change mitigation: in the Global South, ending deforestation is the main priority to reverse the net C source toward a net C sink, while in the Global North, lowering wood harvest has the strongest potential to immediately enhance the C sink in forest biomass. More

1.BirdLife International. State of the World’s Birds: Taking the Pulse of the Planet (BirdLife International, 2018).
Google Scholar 
2.Dias, M. P. et al. Threats to seabirds: A global assessment. Biol. Conserv. 237, 525–537 (2019).Article 

Google Scholar 
3.Gales, R. in Albatross Biology and Conservation (eds Robertson, G. & Gales, R.) 20–45 (Surrey Beatty and Sons, Chipping Norton, 1998).4.Gales, R., Brothers, N. & Reid, T. Seabird mortality in the Japanese tuna longline fishery around Australia, 1988–1995. Biol. Conserv. 86, 37–56 (1998).Article 

Google Scholar 
5.Anderson, O. R. et al. Global seabird bycatch in longline fisheries. Endanger. Species Res. 14, 91–106 (2011).Article 

Google Scholar 
6.Warham, J. The Petrels: Their Ecology and Breeding Systems (Academic Press, 1990).
Google Scholar 
7.Warham, J. The Behaviour, Population Biology and Physiology of the Petrels (Academic Press, 1996).
Google Scholar 
8.Dietrich, K. S., Parrish, J. K. & Melvin, E. F. Understanding and addressing seabird bycatch in Alaska demersal longline fisheries. Biol. Conserv. 142, 2642–2656 (2009).Article 

Google Scholar 
9.Zhou, C., Jiao, Y. & Browder, J. A. Seabird bycatch vulnerability to pelagic longline fisheries: Ecological traits matter. Aquat. Conserv.: Mar. Freshw. Ecosyst. https://doi.org/10.1002/aqc.3066 (2019).Article 

Google Scholar 
10.Brothers, N. The incidental catch of seabirds by longline fisheries: Worldwide review and technical guidelines for mitigation. FAO Fish. Circ. 937, 1–100 (1999).ADS 

Google Scholar 
11.Gilman, E. Integrated management to address the incidental mortality of seabirds in longline fisheries. Aquat. Conserv.: Mar. Freshw. Ecosyst. 11, 391–414 (2001).Article 

Google Scholar 
12.Li, Y. & Jiao, Y. Modeling spatial patterns of rare species using eigenfunction-based spatial filters: An example of modified delta model for zero-inflated data. Ecol. Model. 299, 51–63 (2015).Article 

Google Scholar 
13.Li, Y., Jiao, Y. & Browder, J. A. Assessment of seabird bycatch in the US Atlantic pelagic longline fishery, with an extra exploration on modeling spatial variation. ICES J. Mar. Sci. 73, 2687–2694 (2016).Article 

Google Scholar 
14.Brothers, N. Albatross mortality and associated bait loss in the Japanese longline fishery in the Southern Ocean. Biol. Conserv. 55, 255–268 (1991).Article 

Google Scholar 
15.Croxall, J. P. & Nicol, S. Management of Southern Ocean fisheries: Global forces and future sustainability. Antarct. Sci. 16, 569–584 (2004).ADS 
Article 

Google Scholar 
16.Løkkeborg, S. Best practices to mitigate seabird bycatch in longline, trawl and gillnet fisheries – efficiency and practical applicability. Mar. Ecol. Prog. Ser. 435, 285–303 (2011).ADS 
Article 

Google Scholar 
17.Beerkircher, L. R., Brown, C. J., Abercrombie, D. L. & Lee, D. W. Overview of the SEFSC pelagic observer program in the Northwest Atlantic from 1992–2002. ICCAT CVSP 58, 1729–1748 (2005).
Google Scholar 
18.Diaz, G. A., Beerkircher, L. R. & Restrepo, V. R. Description of the US pelagic observer program (POP). ICCAT CVSP 64, 2415–2426 (2009).
Google Scholar 
19.Lee, D. W. & Brown, C. J. SEFSC pelagic observer program data summary for 1992–1996. US Department of Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Southeast Fisheries Science Center (1998).20.Lo, N. C., Jacobson, L. D. & Squire, J. L. Indices of relative abundance from fish spotter data based on delta-lognornial models. Can. J. Fish. Aquat. Sci. 49, 2515–2526 (1992).Article 

Google Scholar 
21.Martin, T. G. et al. Zero tolerance ecology: improving ecological inference by modelling the source of zero observations. Ecol. Lett. 8, 1235–1246 (2005).PubMed 
Article 
PubMed Central 

Google Scholar 
22.Winter, A., Jiao, Y. & Browder, J. A. Modeling low rates of seabird bycatch in the US Atlantic longline fishery. Waterbirds 34, 289–303 (2011).Article 

Google Scholar 
23.Cortés, V., Arcos, J. M. & González-Solís, J. Seabirds and demersal longliners in the northwestern Mediterranean: Factors driving their interactions and bycatch rates. Mar. Ecol. Prog. Ser. 565, 1–16 (2017).ADS 
Article 

Google Scholar 
24.Bi, R., Jiao, Y., Zhou, C. & Hallerman, E. M. A Bayesian spatiotemporal approach to inform management unit appropriateness. Can. J. Fish. Aquat. Sci. 76, 217–237 (2018).Article 

Google Scholar 
25.Wikle, C. K. Hierarchical models in environmental science. Int. Stat. Rev. 71, 181–199 (2003).MATH 
Article 

Google Scholar 
26.Cressie, N., Calder, C. A., Clark, J. S., Hoef, J. M. V. & Wikle, C. K. Accounting for uncertainty in ecological analysis: The strengths and limitations of hierarchical statistical modeling. Ecol. Appl. 19, 553–570 (2009).PubMed 
Article 
PubMed Central 

Google Scholar 
27.Banerjee, S., Carlin, B. P. & Gelfand, A. E. Hierarchical Modeling and Analysis for Spatial Data (CRC Press, 2014).MATH 
Book 

Google Scholar 
28.Besag, J., York, J. & Mollié, A. Bayesian image restoration, with two applications in spatial statistics. Ann. Inst. Stat. Math. 43, 1–20 (1991).MathSciNet 
MATH 
Article 

Google Scholar 
29.Rue, H. & Held, L. Gaussian Markov Random Fields: Theory and Applications (CRC Press, 2005).MATH 
Book 

Google Scholar 
30.Rue, H., Martino, S. & Chopin, N. Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. J. R. Stat. Soc. Ser. B Stat. Methodol. 71, 319–392 (2009).MathSciNet 
MATH 
Article 

Google Scholar 
31.Held, L., Schrödle, B. & Rue, H. in Statistical Modelling and Regression Structures (eds Kneib, T. & Tutz, G.) 91–110 (Springer, Berlin Heidelberg, 2010).32.Lindgren, F., Rue, H. & Lindström, J. An explicit link between Gaussian fields and Gaussian Markov random fields: The stochastic partial differential equation approach. J. R. Stat. Soc. Ser. B Stat. Methodol. 73, 423–498 (2011).MathSciNet 
MATH 
Article 

Google Scholar 
33.Bakka, H., Vanhatalo, J., Illian, J. B., Simpson, D. & Rue, H. Non-stationary Gaussian models with physical barriers. Spat. Stat. 29, 268–288 (2019).MathSciNet 
Article 

Google Scholar 
34.NCAR (National Center for Atmospheric Research). The Climate Data Guide: Hurrell North Atlantic Oscillation (NAO) Index (PC-based). Available from: https://climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao-index-pc-based. Retrieved: March 1, 2019.35.ESRL (Earth Science Research Laboratory, NOAA). Climate timeseries: AMO (Atlantic Multidecadal Oscillation) Index. Available from: http://www.esrl.noaa.gov/psd/data/timeseries/AMO/. Retrieved: March 1, 2019.36.Spiegelhalter, D. J., Best, N. G., Carlin, B. P. & Van Der Linde, A. Bayesian measures of model complexity and fit. J. R. Stat. Soc. Ser. B Stat. Methodol. 64, 583–639 (2002).MathSciNet 
MATH 
Article 

Google Scholar 
37.Watanabe, S. Asymptotic equivalence of Bayes cross validation and widely applicable information criterion in singular learning theory. J. Mach. Learn. Res. 11, 3571–3594 (2010).MathSciNet 
MATH 

Google Scholar 
38.Shumway, R. H. & Stoffer, D. S. Time Series Analysis and Its Applications (Springer, 2011).MATH 
Book 

Google Scholar 
39.Bi, R., Jiao, Y., Bakka, H. & Browder, J. A. Long-term climate ocean oscillations inform seabird bycatch from pelagic longline fishery. ICES J. Mar. Sci. 77, 668–679 (2020).Article 

Google Scholar 
40.Lear, W. H. History of fisheries in the Northwest Atlantic: The 500 year perspective. J. Northwest Atl. Fish. Sci. 23, 41–73 (1998).Article 

Google Scholar 
41.Veit, R. R., Goyert, H. F., White, T. P., Martin, M. C., Manne, L. L. & Gilbert, A. Pelagic Seabirds off the East Coast of the United States 2008–2013. US Dept. of the Interior, Bureau of Ocean Energy Management, Office of Renewable Energy Programs, Sterling, VA. OCS Study BOEM, 24, 186 (2015).42.Harrison, P. Seabirds, an identification guide (Houghton Mifflin, 1983).
Google Scholar 
43.Onley, D. & Scofield, P. Albatrosses, petrels and shearwaters of the world (Princeton University Press, 2013).
Google Scholar 
44.Gladics, A. J. et al. Fishery-specific solutions to seabird bycatch in the U.S. West Coast sablefish fishery. Fish. Res. 196, 85–95 (2017).Article 

Google Scholar 
45.Grieve, B. D., Hare, J. A. & Saba, V. S. Projecting the effects of climate change on Calanus finmarchicus distribution within the U.S. Northeast Continental Shelf. Sci. Rep. 7, 6264 (2017).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
46.Petersen, S. L., Honig, M. B., Ryan, P. G. & Underhill, L. G. Seabird bycatch in the pelagic longline fishery off southern Africa. Afr. J. Mar. Sci. 31, 191–204 (2009).Article 

Google Scholar 
47.Arcos, J. M. & Oro, D. Significance of fisheries discards for a threatened Mediterranean seabird, the Balearic shearwater Puffinus mauretanicus. Mar. Ecol. Prog. Ser. 239, 209–220 (2002).ADS 
Article 

Google Scholar 
48.Furness, R., Edwards, A. & Oro, D. Influence of management practices and of scavenging seabirds on availability of fisheries discards to benthic scavengers. Mar. Ecol. Prog. Ser. 350, 235–244 (2007).ADS 
Article 

Google Scholar 
49.Grémillet, D. et al. A junk-food hypothesis for gannets feeding on fishery waste. Proc. Biol. Sci. 275, 1149–1156 (2008).PubMed 
PubMed Central 

Google Scholar 
50.Skov, H. & Durinck, J. Seabird attraction to fishing vessels is a local process. Mar. Ecol. Prog. Ser. 214, 289–298 (2001).ADS 
Article 

Google Scholar 
51.Chapman, D. C., Barth, J. A., Beardsley, R. C. & Fairbanks, R. G. On the continuity of mean flow between the Scotian Shelf and the Middle Atlantic Bight. J. Phys. Oceanogr. 16, 758–772 (1986).ADS 
Article 

Google Scholar 
52.Steimle, F. W. & Zetlin, C. Reef habitats in the middle Atlantic bight: Abundance, distribution, associated biological communities, and fishery resource use. Mar. Fish. Rev. 62, 24–42 (2000).
Google Scholar 
53.Lee, D. S. Pelagic seabirds and the proposed exploration for fossil fuels off North Carolina: A test for conservation efforts of a vulnerable international resource. J. Elisha Mitchell Sci. Soc. 115, 294–315 (1999).
Google Scholar 
54.Kai, E. T. et al. Top marine predators track Lagrangian coherent structures. Proc. Natl. Acad. Sci. 106, 8245–8250 (2009).ADS 
CAS 
Article 

Google Scholar 
55.Li, Y., Browder, J. A. & Jiao, Y. Hook effects on seabird bycatch in the United States Atlantic pelagic longline fishery. Bull. Mar. Sci. 88, 559–569 (2012).Article 

Google Scholar 
56.Taylor, A. H. & Stephens, J. A. The North Atlantic Oscillation and the latitude of the Gulf Stream. Tellus 50, 134–142 (1998).Article 

Google Scholar 
57.Hobday, A. J., Hartog, J. R., Spillman, C. M. & Alves, O. Seasonal forecasting of tuna habitat for dynamic spatial management. Can. J. Fish. Aquat. Sci. 68, 898–911 (2011).Article 

Google Scholar 
58.FAO (Food and Agriculture Organization of the United Nations). Guidelines to reduce sea turtle mortality in fishing operations. FAO Technical Guidelines for Responsible Fisheries Prepared by Gilman, E., Bianchi, G. FAO: Rome. ISBN 978-92-106226-5 (2009).59.Bethoney, N. D., Schondelmeier, B. P., Kneebone, J. & Hoffman, W. S. Bridges to best management: Effects of a voluntary bycatch avoidance program in a mid-water trawl fishery. Mar. Policy 83, 172–178 (2017).Article 

Google Scholar 
60.Lindgren, F. & Rue, H. Bayesian spatial modelling with R-INLA. J. Stat. Softw. 63, 19 (2015).Article 

Google Scholar 
61.R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2019).62.Rue, H., Martino, S. & Lindgren, F. INLA: Functions which allow to perform a full Bayesian analysis of structured (geo-)additive models using integrated nested Laplace approximation. R package version 0.0., GNU General Public License, version 3 (2009).63.Simpson, D., Rue, H., Riebler, A., Martins, T. G. & Sørbye, S. H. Penalising model component complexity: A principled, practical approach to constructing priors (with discussion). Stat. Sci. 32, 1–28 (2017).MATH 

Google Scholar 
64.Fuglstad, G.-A., Simpson, D., Lindgren, F. & Rue, H. Constructing priors that penalize the complexity of Gaussian random fields. J. Am. Stat. Assoc. 114, 445–452 (2018).MathSciNet 
MATH 
Article 
CAS 

Google Scholar 
65.Plummer, M. Penalized loss functions for Bayesian model comparison. Biostat. 9, 523–539 (2008).MATH 
Article 

Google Scholar 
66.Gelman, A., Hwang, J. & Vehtari, A. Understanding predictive information criteria for Bayesian models. Stat. Comput. 24, 997–1016 (2014).MathSciNet 
MATH 
Article 

Google Scholar  More

1.Krebs, C. J. Two paradigms of population regulation. Wildl. Res. 22, 1–10 (1995).MathSciNet 
Article 

Google Scholar 
2.McLaren, I. A. Natural Regulation of Animal Populations (Routledge, 2017).Book 

Google Scholar 
3.Deffner, D. & McElreath, R. The importance of life history and population regulation for the evolution of social learning. Philos. Trans. R. Soc. B 375, 20190492 (2020).Article 

Google Scholar 
4.Leão, S. M., Pianka, E. R. & Pelegrin, N. Is there evidence for population regulation in amphibians and reptiles? J. Herpetol. 52, 28–33 (2018).Article 

Google Scholar 
5.Hanski, I. A. Density dependence, regulation and variability in animal populations. Philos. Trans. R. Soc. B 330, 141–150 (1990).ADS 
Article 

Google Scholar 
6.Reznick, D., Bryant, M. J. & Bashey, F. r-and K-selection revisited: The role of population regulation in life-history evolution. Ecology 83, 1509–1520 (2002).Article 

Google Scholar 
7.Stenseth, N. C., Falck, W., Bjørnstad, O. N. & Krebs, C. J. Population regulation in snowshoe hare and Canadian lynx: Asymmetric food web configurations between hare and lynx. Proc. Natl Acad. Sci. USA 94, 5147–5152 (1997).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
8.Lande, R. et al. Estimating density dependence from population time series using demographic theory and life-history data. Am. Nat. 159, 321–337 (2002).CAS 
PubMed 
Article 

Google Scholar 
9.Ferguson, G. W., Bohlen, C. H. & Woolley, H. P. Sceloporus undulatus: Comparative life history and regulation of a Kansas population. Ecology 61, 313–322 (1980).Article 

Google Scholar 
10.Andrews, R. M. Population stability of a tropical lizard. Ecology 72, 1204–1217 (1991).Article 

Google Scholar 
11.Tinkle, D. W., Dunham, A. E. & Congdon, J. D. Life history and demographic variation in the lizard Sceloporus graciosus: A long-term study. Ecology 74, 2413–2429 (1993).Article 

Google Scholar 
12.Altwegg, R., Dummermuth, S., Anholt, B. R. & Flatt, T. Winter weather affects asp viper Vipera aspis population dynamics through susceptible juveniles. Oikos 110, 55–66 (2005).Article 

Google Scholar 
13.Madsen, T. & Shine, R. Rain, fish and snakes: Climatically driven population dynamics of Arafura filesnakes in tropical Australia. Oecologia 124, 208–215 (2000).ADS 
CAS 
PubMed 
Article 

Google Scholar 
14.Madsen, T., Ujvari, B., Shine, R. & Olsson, M. Rain, rats and pythons: Climate-driven population dynamics of predators and prey in tropical Australia. Austral Ecol. 31, 30–37 (2006).Article 

Google Scholar 
15.Brown, G. P. & Shine, R. Rain, prey and predators: Climatically driven shifts in frog abundance modify reproductive allometry in a tropical snake. Oecologia 154, 361–368 (2007).ADS 
PubMed 
Article 

Google Scholar 
16.Brown, G. P., Ujvari, B., Madsen, T. & Shine, R. Invader impact clarifies the roles of top-down and bottom-up effects on tropical snake populations. Funct. Ecol. 27, 351–361 (2013).Article 

Google Scholar 
17.Massot, M., Clobert, J., Pilorge, T., Lecomte, J. & Barbault, R. Density dependence in the common lizard: Demographic consequences of a density manipulation. Ecology 73, 1742–1756 (1992).Article 

Google Scholar 
18.Fordham, D. A., Georges, A. & Brook, B. W. Experimental evidence for density-dependent responses to mortality of snake-necked turtles. Oecologia 159, 271–281 (2009).ADS 
PubMed 
Article 

Google Scholar 
19.Burns, G. & Heatwole, H. Home range and habitat use of the olive sea snake, Aipysurus laevis, on the Great Barrier Reef, Australia. J. Herpetol. 32, 350–358 (1998).Article 

Google Scholar 
20.Ward, T. M. Age structures and reproductive patterns of two species of sea snake, Lapemis hardwickii Grey (1836) and Hydrophis elegans (Grey 1842), incidentally captured by prawn trawlers in northern Australia. Mar. Freshw. Res. 52, 193–203 (2001).Article 

Google Scholar 
21.Dennis, B. & Ponciano, J. M. Density-dependent state-space model for population-abundance data with unequal time intervals. Ecology 95, 2069–2076 (2014).PubMed 
Article 

Google Scholar 
22.Bonnet, X., Naulleau, G. & Shine, R. The dangers of leaving home: Dispersal and mortality in snakes. Biol. Conserv. 89, 39–50 (1999).Article 

Google Scholar 
23.Yacelga, M., Cayot, L. J. & Jaramillo, A. Dispersal of neonatal Galápagos marine iguanas Amblyrhynchus cristatus from their nesting zone: Natural history and conservation implications. Herpetol. Conserv. Biol. 7, 470–480 (2012).
Google Scholar 
24.Lukoschek, V. & Shine, R. Sea snakes rarely venture far from home. Ecol. Evol. 2, 1113–1121 (2012).PubMed 
PubMed Central 
Article 

Google Scholar 
25.Shine, R., Cogger, H. G., Reed, R. R., Shetty, S. & Bonnet, X. Aquatic and terrestrial locomotor speeds of amphibious sea-snakes (Serpentes, Laticaudidae). J. Zool. 259, 261–268 (2003).Article 

Google Scholar 
26.Forsman, A. Body size and net energy gain in gape-limited predators: A model. J. Herpetol. 30, 307–319 (1996).Article 

Google Scholar 
27.Shine, R., LeMaster, M. P., Moore, I. T., Olsson, M. M. & Mason, R. T. Bumpus in the snake den: Effects of sex, size and body condition on mortality in red-sided garter snakes. Evolution 55, 598–604 (2001).CAS 
PubMed 
Article 

Google Scholar 
28.Sherratt, E., Rasmussen, A. R. & Sanders, K. L. Trophic specialization drives morphological evolution in sea snakes. R. Soc. Open Sci. 5, 172141 (2018).ADS 
PubMed 
PubMed Central 
Article 

Google Scholar 
29.Lemen, C. A. & Voris, H. K. A comparison of reproductive strategies among marine snakes. J. Anim. Ecol. 50, 89–101 (1981).Article 

Google Scholar 
30.Shine, R., Shine, T. & Shine, B. Intraspecific habitat partitioning by the sea snake Emydocephalus annulatus (Serpentes, Hydrophiidae): The effects of sex, body size, and colour pattern. Biol. J. Linn. Soc. 80, 1–10 (2003).Article 

Google Scholar 
31.Heatwole, H., Grech, A., Monahan, J. F., King, S. & Marsh, H. Thermal biology of sea snakes and sea kraits. Integr. Comp. Biol. 52, 257–273 (2012).PubMed 
Article 

Google Scholar 
32.Van Dyke, J. U., Beaupre, S. J. & Kreider, D. L. Snakes allocate amino acids acquired during vitellogenesis to offspring: Are capital and income breeding consequences of variable foraging success? Biol. J. Linn. Soc. 106, 390–404 (2012).Article 

Google Scholar 
33.Masunaga, G., Matsuura, R., Yoshino, T. & Ota, H. Reproductive biology of the viviparous sea snake Emydocephalus ijimae (Reptilia: Elapidae: Hydrophiinae) under a seasonal environment in the Northern Hemisphere. Herpetol. J. 13, 113–119 (2003).
Google Scholar 
34.Goiran, C., Dubey, S. & Shine, R. Effects of season, sex and body size on the feeding ecology of turtle-headed sea snakes (Emydocephalus annulatus) on IndoPacific inshore coral reefs. Coral Reefs 32, 527–538 (2013).ADS 
Article 

Google Scholar 
35.Phillips, B. L. The evolution of growth rates on an expanding range edge. Biol. Lett. 5, 802–804 (2009).PubMed 
PubMed Central 
Article 

Google Scholar 
36.Shine, R., Shine, T. G., Brown, G. P. & Goiran, C. Life history traits of the sea snake Emydocephalus annulatus, based on a 17-yr study. Coral Reefs 39, 1407–1414 (2020).Article 

Google Scholar 
37.Goiran, C. & Shine, R. Decline in sea snake abundance on a protected coral reef system in the New Caledonian Lagoon. Coral Reefs 32, 281–284 (2013).ADS 
Article 

Google Scholar 
38.Somaweera, R. et al. Pinpointing drivers of extirpation in sea snakes: A synthesis of evidence from Ashmore Reef. Front. Mar. Sci. 8, 658756 (2021).Article 

Google Scholar 
39.Udyawer, V. et al. Future directions in the research and management of marine snakes. Front. Mar. Sci. 5, 399 (2018).Article 

Google Scholar 
40.Udyawer, V., Cappo, M., Simpfendorfer, C. A., Heupel, M. R. & Lukoschek, V. Distribution of sea snakes in the Great Barrier Reef Marine Park: Observations from 10 yrs of baited remote underwater video station (BRUVS) sampling. Coral Reefs 33, 777–791 (2014).ADS 
Article 

Google Scholar 
41.Udyawer, V., Goiran, C. & Shine, R. Peaceful coexistence between people and deadly wildlife: Why are recreational users of the ocean so rarely bitten by sea snakes? People Nat. 3, 335–346 (2021).Article 

Google Scholar 
42.Shine, R., Goiran, C., Shine, T., Fauvel, T. & Brischoux, F. Phenotypic divergence between seasnake (Emydocephalus annulatus) populations from adjacent bays of the New Caledonian Lagoon. Biol. J. Linn. Soc. 107, 824–832 (2012).Article 

Google Scholar 
43.Goiran, C., Brown, G. P. & Shine, R. Niche partitioning within a population of sea snakes is constrained by ambient thermal homogeneity and small prey size. Biol. J. Linn. Soc. 129, 644–651 (2020).Article 

Google Scholar 
44.Li, M., Fry, B. G. & Kini, R. M. Eggs-only diet: Its implications for the toxin profile changes and ecology of the marbled sea snake (Aipysurus eydouxii). J. Mol. Evol. 60, 81–89 (2005).ADS 
CAS 
PubMed 
Article 

Google Scholar 
45.Heatwole, H. Predation on sea snakes. In The Biology of Sea Snakes (ed. Dunson, W. A.) 233–250 (University Park Press, 1975).
Google Scholar 
46.Rancurel, P. & Intes, A. L. requin tigre, Galeocerdo cuvieri Lacepede, des eaux neocaledoniennes examen des contenus stomacaux. Tethys 10, 195–199 (1982).
Google Scholar 
47.Ineich, I. & Laboute, P. Les Serpents Marins de Nouvelle-Calédonie (IRD éditions, 2002).
Google Scholar 
48.Masunaga, G., Kosuge, T., Asai, N. & Ota, H. Shark predation of sea snakes (Reptilia: Elapidae) in the shallow waters around the Yaeyama Islands of the southern Ryukyus, Japan. Mar. Biodivers. Rec. 1, e96 (2008).Article 

Google Scholar 
49.Wirsing, A. J. & Heithaus, M. R. Olive-headed sea snakes Disteria major shift seagrass microhabitats to avoid shark predation. Mar. Ecol. Progr. Ser. 387, 287–293 (2009).ADS 
Article 

Google Scholar 
50.White, G. C. & Burnham, K. P. Program MARK: Survival estimation from populations of marked animals. Bird Study 46, S120–S139 (1999).Article 

Google Scholar 
51.Brischoux, F., Rolland, V., Bonnet, X., Caillaud, M. & Shine, R. Effects of oceanic salinity on body condition in sea snakes. Integr. Comp. Biol. 52, 235–244 (2012).PubMed 
Article 

Google Scholar 
52.Lovich, J. E. & Gibbons, J. W. A review of techniques for quantifying sexual size dimorphism. Growth Dev. Aging 56, 269–269 (1992).CAS 
PubMed 

Google Scholar 
53.Dennis, B., Ponciano, J. M., Lele, S. R., Taper, M. L. & Staples, D. F. Estimating density dependence, process noise, and observation error. Ecol. Monogr. 76, 323–341 (2006).Article 

Google Scholar  More