First Plants Cooled the Ordovician

FEATURE | FOCUS

First plants cooled the Ordovician

Timothy M. Lenton, Michael Crouch, Martin Johnson, Nuno Pires and Liam Dolan

The Late Ordovician period, ending 444 million years ago, was marked by the onset of glaciations. The expansion of non-vascular land plants accelerated chemical weathering and may have drawn down enough atmospheric carbon dioxide to trigger the growth of ice sheets.

etween 488 and 444 million years ago, the Late Ordovician, but it is unclear which rates are exceptionally high, could also have during the Ordovician period, the process (or processes) could have lowered enhanced weathering12. Together with basalt Bclimate cooled gradually1, culminating these levels temporarily at the time. production, this process could have lowered in the abrupt onset of periods of temporary The slow decline in CO2 levels can be CO2 to approximately 12 PAL from the glaciation. These Late Ordovician glaciations explained by invoking the drawdown of CO2 Middle Ordovician to Early Silurian period. are puzzling, because they occurred when by increased silicate weathering. The Taconic However, these geological mechanisms 10 atmospheric CO2 concentrations, as estimated Orogeny led to extensive mountain building seem insufficient to lower 2CO to 8 PAL and by geochemical models2–4 and proxy data5–7, — and weathering — along what is now the they do not explain the temporary abrupt were roughly 14–22 times present-day mid-Atlantic and northeastern coast of the glaciations1, or other striking global changes, atmospheric levels (PAL). Yet complex climate United States. The eruption rate of basalt, in the Late Ordovician (Fig. 1). In particular, 8,9 13 models suggest atmospheric CO2 levels had a relatively weatherable rock type, was also the δ C record of carbonates contains two to drop to about 8 PAL to trigger glaciations elevated during the Ordovician, according global positive excursions13 that indicate at this time. The uncertainty in the proxy to seawater 87Sr/86Sr records11. Finally, the pulses of high organic carbon burial rates in estimates5–7 allows for such a temporary movement of the continents through the the ocean. Extensive shallow-water phosphate 14 reduction in atmospheric CO2 levels during intertropical convergence zone, where rainfall deposits from the Late Ordovician suggest that the input of phosphorus to the ocean, 15 Cryptospores Trilete spores where it is the ultimate limiting nutrient, was 16 Non-vascular plants First vascular plants high. The resultant increase in productivity 87 86 seems to have triggered regional ocean ) Drop in seawater Sr/ Sr 17 ‰ anoxia, as well as high rates of organic carbon ( e 18 burial, as evidenced by the deposition of

apatit 15

O black shales . Hirnantian

18 19 δ Here we suggest that this suite of Late

20 Climate cooling Ordovician global changes was caused, Glacial deposits at least in part, by the origination and 21 8 24 expansion of the first land plants. All plants 6 20 require rock-derived minerals — including CO 4 δ 16 phosphorus, potassium, calcium, magnesium 13 C ( 2

(P and iron — for growth. Land plants have

2 ‰ 12 AL evolved a variety of systems and symbiotic ) 0 8 ) relationships that accelerate the release of GICE HICE these essential nutrients from rocks16. This –2 4 Phosphate deposits acceleration has been well characterized 0 in vascular plants17, which enhance the 480 470 460 450 440 breakdown of nutrient-containing silicate 1d 2a 2b 2c 3a 3b 4a 4b 4c 5a 5b 5c 5d 6a 6b 6c minerals (silicate weathering) by factors of T. Floian Dapingian Darriwillian Sandbian Katian H. 2 to 10 (as is the case for calcium). The step Early Ordovician Mid Ordovician Late Ordovician Silurian change in silicate weathering rates caused 472 460.5443 by the radiation of vascular plants is widely Age (Ma) thought to have caused the drawdown of

atmospheric CO2 between 400 and 360 Ma in Figure 1 | Global changes during the Ordovician period. Appearances of non-vascular18 and vascular23 the Devonian period that led to global cooling plants dot the timeline of climate change as recorded by conodont δ18O values1 (diamonds, left- and polar glaciations2. hand scale). The appearance of non-vascular plants is followed by an increase in the abundance and Yet it is likely that land plants affected the weathering of volcanic rock, as recorded by seawater 87Sr/86Sr (ref. 11). The weathering seems to coincide global carbon cycle before the Devonian. with a drop in atmospheric CO2 concentrations (solid squares, outer right-hand scale), though proxy- Fossilized plant spores found in ~470 Ma based estimates are scarce and highly uncertain5–7. Glacial deposits appear early in the Late Ordovician sediments18,19 indicate that non-vascular land and increase in frequency as it progresses13, and phosphate deposits appear then in marine settings14. Two plants were growing in damp environments carbon isotope excursions also mark the Late Ordovician: the Guttenberg and Hirnantian isotopic carbon on continental surfaces by the middle of the excursions (GICE and HICE)13. T., Tremadogian age; H., Hirnantian age. Ordovician (Fig. 1). Furthermore, phylogenies

86 NATURE GEOSCIENCE | VOL 5 | FEBRUARY 2012 | www.nature.com/naturegeoscience

constructed using DNA sequences support respectively. Similarly, mosses increased uncertain parameters, which altered initial the contention that the earliest plants were magnesium weathering by a factor of model conditions such that the atmospheric 16,20 non-vascular . Because non-vascular land 1.5 ± 0.2 and 5.4 ± 0.9 from granite and CO2 concentrations ranged from 14 to plants lack deep rooting structures, existing andesite, respectively. Given the relatively 21 PAL 460 Myr ago. Application of the models3,4,17 assume that the early expansion small amounts of biomass present in these same colonization scenario for non-vascular of land plants had negligible effects on the microcosms, these data demonstrate that non- plants then lowered CO2 to about 7 to 10 PAL existing weathering regime. But given their vascular plants considerably enhance silicate (Supplementary Fig. S2, Supplementary need to access rock-bound minerals, we weathering over background abiotic levels, Table S1), broadly consistent with values suggest that these first colonizers actually and do so by a factor comparable to that of necessary to trigger glaciation. Alternatively, caused a substantial change in chemical vascular plants17. varying the extent of colonization by weathering, with a significant impact on the non-vascular plants between 5–45 %, or global carbon cycle and climate. Effects of plant colonization their amplification of silicate weathering To determine the potential effects of over a range of ~1.6–6.7 (global average Weathering experiments enhanced weathering by the first plants on the amplification factor of 1.085–1.85), lowered

We set out to determine if non-vascular Ordovician Earth system, we used an adapted CO2 to ~5–13 PAL (Supplementary Fig. S3, plants enhance chemical weathering using version of the COPSE model4 (Supplementary Supplementary Table S2). These sensitivity the moss Physcomitrella patens (also known Methods). In our baseline version, which does analyses all resulted in a significant decrease as Aphanoregma patens). This modern moss not take non-vascular-plant weathering into in global mean temperature, in the range has evolved considerably compared with its account, the CO2 concentration was ~16 PAL 1.4–7.2 °C. In our simulations, atmospheric Ordovician ancestors, but is a representative and global temperature was 21.4 °C 460 Myr CO2 concentrations and temperature non-vascular plant that can be used for ago, the start of the Late Ordovician (Fig. 3, remained low during the Late Ordovician experimentation. red dotted line). These are consistent with (Fig. 3), consistent with the reported sustained To measure the effect of non-vascular- proxy estimates5 for the background levels cold interval of over 10 Myr1,13,14 (Fig. 1). plant growth on silicate weathering, we of atmospheric CO2 of ~14–22 PAL, and the However, if we used an earlier date for compared the release of elements from lack of evidence for glaciations up to that the colonization of the land by plants19, in granite and andesite in microcosms that were point. When we simulated non-vascular-plant this case, about 490 Myr ago, the model incubated with or without moss. Microcosms colonization of 15% of the currently vegetated produced Early to Middle Ordovician cooling containing either granite or andesite were land surface between 475 and 460 Myr ago (Supplementary Fig. S4a). This simulation incubated with either a suspension containing — amplifying silicate weathering by a factor better matches the temperature trend inferred macerated moss cultures or with suspensions of ~5 (a global average amplification factor of from oxygen isotopes of conodonts1 (Fig. 1). from which the moss had been removed by 1.75) — the model predicted a CO2 drop to However, evidence for early colonization by filtration (control). Moss-containing and 8.4 PAL and global cooling of 4.4 to 17.0 °C land plants is contested18. Alternatively, if control microcosms were incubated in by 460 Myr ago (Fig. 3, blue dashed line). colonization started later or took longer, CO2 16-hour light, 8-hour dark cycles at 25 °C for This is consistent with the requirement of drawdown and cooling were correspondingly

~130 days. During this time 18.4 ± 14.9 mg CO2 concentrations ~8 PAL for the start of delayed (Supplementary Fig. S4b,c). Taking and 13.4 ± 6.2 mg of biomass accumulated glaciations, as suggested by more complex all the sensitivity analyses into account, the in the granite- and andesite-containing climate models8,9. model demonstrates that the enhancement microcosms respectively; extensive mats In our sensitivity analyses (see of silicate weathering by the first land plants of haploid protonema, gametophores and Supplementary Information) we varied is likely to have reduced atmospheric CO2 rhizoids formed (Supplementary Fig. S1). ab Sporophytes did not develop because of the 7 high temperature, and hence the mosses did Granite Andesite not complete their life cycle. 6 Moss Moss We determined the extent of calcium- Water Water

mol) Abiotic control Abiotic control and magnesium-silicate weathering in each µ 5 treatment by measuring the total moles of elements released from rock during the 4 incubation period using inductively coupled ed amount ( plasma-atomic emission spectroscopy. To 3

measure the elements released from rock eather in the moss-containing microcosms, we 2 determined the abundance of elements Mean w in the liquid (consisting of water and 1 inoculation media) and the biomass. We used the abundance of elements present in the 0 incubating media of the control (moss-free) Al (61) Ca (1.4) Fe (170) K (2.5) Mg (1.5) Al (40) Ca (3.6) Fe (360) K (1.5) Mg (5.4) microcosms to assess the amount of abiotic Element (ampliﬁcation factor) Element (ampliﬁcation factor) background weathering. The presence of moss enhanced the Figure 2 | Moss enhances the weathering of Al, Ca, Fe, K and Mg from silicates. Summary of results chemical weathering of all measured from all weathering experiments on different substrates:a , granite (control n = 77; moss n = 72) and elements from both granite and andesite b, andesite (control n = 37; moss n = 41). The amount weathered in micromoles (mean of all microcosms (Fig. 2). Specifically, moss increased of each substrate) are shown, with abiotic and total biotic (moss + water) data in separate columns. calcium weathering by factors of 1.4 ± 0.2 The weathering ‘amplification factor’ due to the presence of moss = total biotic weathering/abiotic and 3.6 ± 0.9 from granite and andesite, weathering. Error bars represent ~95% confidence intervals applied using the Student’st -test.

NATURE GEOSCIENCE | VOL 5 | FEBRUARY 2012 | www.nature.com/naturegeoscience 87

a concentrations and global temperature 18 could have caused the GICE. Our model towards or below the threshold for glaciations. 16 predicts the GICE was accompanied by 14 However, increasing silicate weathering alone a temporary CO2 drop to 6.2 PAL and a ) 12 could not generate large, positive spikes in AL 10 2.0 °C cooling to average temperatures of 13 (P the δ C record, or create short-lived cooling 2 8 15.0 °C, comparable to today. Reproduction

CO 6 events, notably the Hirnantian glaciation 4 of the HICE requires a transient tripling of around 445 Myr ago (Fig. 1). 2 phosphorus weathering, causing CO2 to drop 0 480 470 460 450 440 to 4.5 PAL and a 3.4 °C cooling to 13.6 °C. Phosphorus weathering Age (Ma) This might have been due to the evolution To explain these positive spikes in the δ13C b of the first vascular plants23, colonizing new 25 C) record and short-lived cooling events, we ° habitats and thus causing a second pulse of propose that the colonization of land by plants e ( phosphorus weathering. 20 enhanced the weathering of phosphorus atur Sensitivity analyses produce atmospheric from the land surface. This would have CO2 concentrations in a range of 3.5–9.8 PAL transiently increased the flux of phosphorus 15 for the GICE and 2.5–7.2 PAL for the to the oceans, where it would have fuelled HICE (Supplementary Tables S1 and S2),

Global temper 10 new production and organic carbon burial. 480 470 460 450 440 and corresponding temperature ranges of Phosphorus is essential for plant growth Age (Ma) 12.0–18.0 °C and 10.7–16.2 °C, respectively. and reproduction, and constitutes ~0.1% of c These predicted atmospheric CO and 7 2 the dry weight of mosses (Supplementary 6 temperature ranges are consistent with Methods). Crucially, phosphorus has no glaciations starting during the GICE and

) 5 1 significant gaseous form. Hence there must ‰ 4 become most pronounced in the Hirnantian . have been strong selection pressure for C ( 3 After each event, our simulations suggest 13 δ effective phosphorus weathering, which 2 atmospheric CO2 concentrations and vascular plants achieve by secreting organic 1 temperature rebounded temporarily above 0 acids, using their rooting systems and 480 470 460 450 440 the steady-state level determined by silicate associated mycorrhizal fungi. Age (Ma) weathering. The rebound could help explain We therefore determined if our non- the Boda warming event between the δ13C vascular-plant analogue secreted organic acids Figure 3 | Model results. Predicted Ordovician excursions, as well as the observed warming into the growth medium (Supplementary after the Hirnantian glaciation. Our enhanced variations in a, atmospheric CO2, b, global Methods). We identified malic, citric, glyceric temperature, and c, δ13C of marine carbonates, for: phosphorus weathering scenario is also and succinic acids in purified exudates baseline model with Ordovician geological forcing consistent with phosphate-rich deposits 14 (Supplementary Fig. S5), indicating that and changing solar luminosity only (red dotted formed during the Late Ordovician : Our non-vascular and vascular plants secrete the line), including enhancement of silicate weathering simulations show an increase in phosphorus same acids into the rhizosphere. We also by non-vascular plants (blue dashed line), and burial that closely tracks the phosphorus input 4 5 determined whether our moss analogue adding transient enhancement of phosphorus because phosphate only has a 10 –10 year enhanced the release of phosphorus weathering by early plants (green solid line). residence time in the ocean. from granite, and found that phosphorus Finally, as iron can limit productivity of the weathering was amplified by a factor of ~60 remote oceans, we also determined the effects compared with the controls that lack moss million years the ecosystem exhausted its of moss on iron release from granite and (see Supplementary Methods). This value substrate and shifted to internal recycling of andesite. Moss increased iron weathering by is likely to be a large underestimate because phosphorus with little leakage21. Each time factors of 170 ± 20 for granite and 360 ± 100 the moss used in these experiments was plants evolved to colonize new regions of the for andesite (Fig. 2). Today, approximately grown without fungal symbionts, whereas land surface, or innovations allowed plants to 0.5% of the total weathered iron (~800 Tg yr–1) earliest land plants are likely to have evolved further exploit previously colonized areas, a that enters the global river system reaches the symbiotic associations with fungi similar to pulse of phosphorus weathering and supply to deep ocean in reactive (that is, bioavailable) those that form mycorrhizae today. the ocean would have occurred, as has been form24. Thus, despite potential changes in Enhancement of phosphorus weathering argued for the Late Devonian22. redox and chelation states with a transition by a factor of ~60, even if restricted to a To simulate transient selective from abiotic to biotic chemical weathering, few per cent of the land surface, could enhancement of phosphorus weathering, we land colonization would have significantly have more than doubled the global flux of introduced a new forcing parameter into the enhanced the iron flux to the open ocean. This phosphorus to the ocean. However, once COPSE model (Supplementary Methods), in turn would have increased new production the first plants began to exhaust the supply and varied it to reproduce the two Late and organic carbon burial, and contributed of easily accessible phosphorus in available Ordovician peaks in the carbonate carbon further to the positive δ13C excursions. rocks, and they became effective at recycling isotope (δ13C) record (Fig. 3, green solid line): phosphorus from organic sources as soils the Guttenberg (GICE; δ13C +4‰, ~455 Myr Biological and geological drivers formed, phosphorus flux to the oceans would ago) and Hirnantian (HICE; δ13C +6‰, We demonstrate that the evolution of the have decreased. This would have resulted in ~445 Myr ago) isotope excursions13 (Fig. 1). first land plants, through their effects on a transient spike of increased phosphorus Both events involved pulses of organic carbon silicate weathering and fluxes to the ocean, flux to the ocean, analogous to what happens burial that further lowered atmospheric CO2 could explain Ordovician global changes, during the colonization of modern lava and temperature (Fig. 3, green solid line). We culminating in one of only two glacial flows (for example, on the Hawaiian islands). found that a temporary doubling of global episodes in the Paleozoic era. Geological

Initially, phosphorus weathering and leakage phosphorus weathering by the first land plants factors may also have contributed to CO2 to the ocean were high, but after several (well within the range of our experiments) drawdown and long-term cooling, but cannot

88 NATURE GEOSCIENCE | VOL 5 | FEBRUARY 2012 | www.nature.com/naturegeoscience

at present explain the short-lived Hirnantian of Exeter, Hatherly Laboratories, Prince of Wales Road, 15. Berry, W. B. N. Geol. Soc. Am. S. 466, 141–147 (2010). glaciation or other global biogeochemical Exeter, EX4 4PS, UK, 2Earth and Life Systems Alliance, 16. Bateman, R. M. et al. Annu. Rev. Ecol. Syst. 29, 263–292 (1998). 17. Moulton, K. & Berner, R. A. Geology 26, 895–898 (1998). changes in the Late Ordovician. These School of Environmental Sciences, University of East 18. Wellman, C. H. & Gray, J. Phil. Trans. R. Soc. Lond. B environmental changes are associated with Anglia, Norwich, NR4 7TJ, UK, 3Earth and Life Systems 355, 717–732 (2000). a marine mass extinction25 that defines the Alliance, Department of Cell and Developmental 19. Rubinstein, C. V., Gerrienne, P., de la Puente, G. S., Astini, R. A. & Steemans, P. New Phytol. 188, 365–369 (2010). end of the Ordovician Period. Thus, the Biology, John Innes Centre, Norwich, NR4 7UH, UK, 20. Qiu, Y.-L. et al. Proc. Natl Acad. Sci.USA 103, 15511–15516 (2006). 4 evolution of the first land plants could have Department of Plant Sciences, University of Oxford, 21. Chadwick, O. A., Derry, L. A., Vitousek, P. M., Huebert, B. J. & indirectly contributed to killing off many of South Parks Road, Oxford, OX1 3RB, UK. Hedin, L. O. Nature 397, 491–497 (1999). e-mail: [email protected]; 22. Algeo, T. J. & Scheckler, S. E. Phil. Trans. R. Soc. Lond. B their compatriots in the ocean. More precise 353, 113–130 (1998). dates on the timing of the evolution of the first [email protected] 23. Steemans, P. et al. Science 324, 353 (2009). land plants and better estimates of the extent 24. Poulton, S. W. & Raiswell, R. Am. J. Sci. 302, 774–805 (2002). of these early floras, as well as more detailed References 25. Sheehan, P. M. Annu. Rev. Earth Planet. Sci. 29, 331–364 (2003). quantification of silicate weathering on a 1. Trotter, J. A., Williams, I. S., Barnes, C. R., Lecuyer, C. & Nicoll, R. S. Science 321, 550–554 (2008). Acknowledgements range of rock types, will increase the accuracy 2. Berner, R. A. Science 276, 544–546 (1997). We thank the Earth and Life Systems Alliance (ELSA) for of our models. 3. Berner, R. A. Geochim. Cosmochim. Acta 70, 5653–5664 (2006). funding this research. T.M.L. is also supported by NERC 4. Bergman, N. M., Lenton, T. M. & Watson, A. J. Am. J. Sci. An ongoing puzzle is why CO2 levels (NE/I005978/1). L.D. and N.P. are also supported by an and temperature increased later in the 304, 397–437 (2004). EVO500 grant from the ER AdG program. P. Bota provided 5. Tobin, K. J. & Bergstrom, S. M. Palaeogeog. Palaeoclim. assistance with GC-MS. Silurian. Here we must appeal to a geological Palaeoecol.181, 399–417 (2002). mechanism12: the movement of continents 6. Tobin, K. J., Bergström, S. M. & De La Garza, P. Palaeogeog. Palaeoclim. Palaeoecol. 226, 187–204 (2005). Author Contributions away from the intertropical convergence 7. Yapp, C. J. & Poths, H. Earth Planet. Sci. Lett. 137, 71–82 (1996). T.M.L. and L.D. designed the study. M.C. conducted the zone. We suggest that future work consider 8. Gibbs, M. T., Barron, E. J. & Kump, L. R. Geology 25, 447–450 (1997). microcosm experiments with input from N.P. and L.D. both biological and geological influences on 9. Herrmann, A. D., Patzkowsky, M. E. & Pollard, D. Geology M.C., M.J., L.D. and T.M.L. conducted geochemical analyses. weathering through this pivotal interval of 31, 485–488 (2003). N.P. and L.D. identified acids in moss exudates. T.M.L. did ❐ 10. Kump, L. R. et al. Palaeogeog. Palaeoclim. Palaeoecol. the modelling and sensitivity analyses. T.M.L. and L.D. wrote early land colonization. 152, 173–187 (1999). the paper with input from M.J., M.C. and N.P. 11. Young, S. A., Saltzman, M. R., Foland, K. A., Linder, J. S. & Timothy M. Lenton1,2, Michael Crouch2,3, Martin Kump, L. R. Geology 37, 951–954 (2009). 12. Nardin, E. et al. Geol. Soc. Am. Bull. 123, 1181–1192 (2011). Additional information 2 3,4 3,4 Johnson , Nuno Pires and Liam Dolan are at 13. Saltzman, M. R. & Young, S. A. Geology 33, 109–112 (2005). Supplementary information accompanies this paper on 1College of Life and Environmental Sciences, University 14. Pope, M. C. & Steffen, J. B. Geology 31, 63–66 (2003). www.nature.com/naturegeoscience.

NATURE GEOSCIENCE | VOL 5 | FEBRUARY 2012 | www.nature.com/naturegeoscience 89