Extended Megadroughts in the Southwestern United States During Pleistocene Interglacials

LETTER doi:10.1038/nature09839

Extended megadroughts in the southwestern United States during Pleistocene interglacials

Peter J. Fawcett1, Josef P. Werne2,4,5, R. Scott Anderson6,7, Jeffrey M. Heikoop8, Erik T. Brown3, Melissa A. Berke3, Susan J. Smith7, Fraser Goff1, Linda Donohoo-Hurley1, Luz M. Cisneros-Dozal8, Stefan Schouten9, Jaap S. Sinninghe Damste´9, Yongsong Huang10, Jaime Toney8, Julianna Fessenden6, Giday WoldeGabriel6, Viorel Atudorei1, John W. Geissman1 & Craig D. Allen11

The potential for increased drought frequency and severity linked muted precessional cycles within this interglacial. In comparison to anthropogenic climate change in the semi-arid regions of the with MIS 11, MIS 13 experienced higher precessional-cycle ampli- southwestern United States (US) is a serious concern1. Multi-year tudes, larger variations in MAT (4–6 6C) and a longer period of droughts during the instrumental period2 and decadal-length extended warmth, suggesting that local insolation variations were droughts of the past two millennia1,3 were shorter and climatically important to interglacial climatic variability in the southwestern different from the future permanent, ‘dust-bowl-like’ mega- US. Comparison of the early MIS 11 climate record with the drought conditions, lasting decades to a century, that are predicted Holocene record shows many similarities and implies that, in the as a consequence of warming4. So far, it has been unclear whether absence of anthropogenic forcing, the region should be entering a or not such megadroughts occurred in the southwestern US, and, if cooler and wetter phase. so, with what regularity and intensity. Here we show that periods of The hydroclimatology of the southwestern US shows significant aridity lasting centuries to millennia occurred in the southwestern natural variability including major historical droughts1. Models of US during mid-Pleistocene interglacials. Using molecular palaeoclimate response to anthropogenic warming predict future dust- temperature proxies5 to reconstruct the mean annual temperature bowl-like conditions that will last much longer than historical (MAT) in mid-Pleistocene lacustrine sediment from the Valles droughts and have a different underlying cause, a poleward expansion Caldera, New Mexico, we found that the driest conditions occurred of the subtropical dry zones4. At present, no palaeoclimatic analogues during the warmest phases of interglacials, when the MAT was are available to assess the potential duration of aridity under a warmer comparable to or higher than the modern MAT. A collapse of climate or to evaluate its effect on the seasonality of precipitation. drought-tolerant C4 plant communities during these warm, dry Here we present a high-resolution climate record from an 82-m intervals indicates a significant reduction in summer precipitation, lacustrine sediment core (VC-3) from the Valles Caldera (Fig. 1) that possibly in response to a poleward migration of the subtropical dry spans two mid-Pleistocene glacial cycles from MIS 14 to MIS 10 zone. Three MAT cycles 2 6C in amplitude occurred within (552 kyr ago to ,368 kyr ago; see Supplementary Information). Marine Isotope Stage (MIS) 11 and seem to correspond to the MISs 11 and 13 are long interglacials that may have been as warm as

a b Southern Rocky 106.6° W Longitude 106.3° W Mountains 36.0000° N San Juan Mountains Jemez Mountains

(Valle Grande)

UT CO Latitude

Los Alamos

Valle AZ NM de Pajarito GrandeGran Mountain

South Pajarito Rio Grande Mountain Plateau Sangre de Cristo Mountains 35.8055° N (km) 015 0 Valles Caldera area Figure 1 | Location map of the Valles Caldera. a, Location in northern New Mexico. b, Digital elevation model of the Valles Caldera showing the location of South Mountain rhyolite, Valle Grande, the drilling location of core VC-3 (black square) and a photograph of the drilling site.

1Department of Earth & Planetary Sciences, University of New Mexico, Albuquerque, New Mexico 87131, USA. 2Large Lakes Observatory and Department of Chemistry and Biochemistry, University of Minnesota Duluth, Duluth, Minnesota 55812, USA. 3Large Lakes Observatory and Department of Geological Sciences, University of Minnesota Duluth, Duluth, Minnesota 55812, USA. 4Centre for Water Research, University of Western Australia, Crawley, Western Australia 6009, Australia. 5WA-Organic and Isotope Geochemistry Centre, Curtin University of Technology, Bentley, Western Australia 6845, Australia. 6School of Earth Sciences and Environmental Sustainability, Northern Arizona University, Flagstaff, Arizona 86011, USA. 7Laboratory of Paleoecology, Bilby Research Center, Northern Arizona University, Flagstaff, Arizona 86011, USA. 8Earth and Environmental Sciences Division, EES-14, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, USA. 9NIOZ Royal Netherlands Institute for Sea Research, Department of Marine Organic Biogeochemistry, PO Box 59, 1790 AB Den Burg, Netherlands. 10Department of Geological Sciences, Brown University, Providence, Rhode Island 02912, USA. 11USGS Fort Collins Science Center, Jemez Mountains Field Station, Los Alamos, New Mexico 87544, USA.

the Holocene epoch, and MIS 11 is a good analogue for future natural Isotope stage (MIS) a Insolation (W m climate variability with similar, low-amplitude precessional cycles6,7. 1011a 11c 11e 12 13 14 530 We used novel organic geochemical proxies (the cyclization ratio of 510 branched tetraethers (CBT, related to pH) and the methylation index b 490 5,8 12 of branched tetraethers (MBT, related to temperature and pH )) to 470 reconstruct the annual MAT of the Valles Caldera watershed, and 8 450 –2 ) compared these with proxies of hydrologic balance to evaluate the c 4 Juniperus (%) relationship between warmth and aridity. 30 Quercus (%) Interglacial MATs in the VC-3 record range from ,0to7uC, with 0 20 d the highest temperatures occurring in MIS 13 and early in MIS 11 10 9 (Fig. 2). The highest temperatures (5–7 uC) are similar to modern 0 6 Modern MAT , A1 B1 B2 B3 B4 B6 B7 C1 C3 MATs, of 5 uC. The glacial stages have multiple millennial-scale tem- 3 perature oscillations with amplitudes as large as 7 uC; approximately 0 seven oscillations are preserved in MIS 12 (B1–B7), three in late MIS 14 –3 e 30 Abies (%) Picea + (C1–C3) and one in early MIS 10 (A1). The frequency of these oscilla- (°C) Temperature –6 20 tions (2–10 kyr) is similar to those recorded in contemporaneous f 9 10 Atlantic Ocean sediment records . All VC-3 stadials correlate with high 15 percentages of Picea 1 Abies pollen, whereas interstadials have lower 0 10 Picea 1 Abies pollen percentages and many correlate with local maxima in Juniperus and Quercus (Fig. 2). Increased percentages of 5 Cyperaceae (sedge) pollen during several interstadials suggest a 0

Cyperaceae (%) g shallower lake rimmed by a broad marshy zone, which would have Mud cracks 1 Si/Ti ratio 0.8 been minimized during stadials, when the lake was deeper. Interstadial 0.6 shallowing probably resulted from increased evaporation and/or a h 8 0.4 reduction in the winter precipitation that dominates regional glacial- 6 0.2 10 stage precipitation . 4 Glacial terminations VI and V in the VC-3 record show temperature i TOC (%) 2 δ increases of ,7 and ,8 C, respectively. The d13C record of TOC –18 13 u 0 C (‰ VPDB) (Fig. 2) shows negative isotopic shifts of 2.5–3.5% at the terminations –21 that we interpret as biotic responses to global increases in atmospheric –24 13 11 9 CO2, similar to the Termination I d C response in Lake Baikal . j –27 We subdivide MIS 11 into five distinct substages, three warm and 8 –30 two cool, on the basis of MAT estimates, warm (lower-elevation) 7 Soil pH 8 k Ca Count

versus boreal (higher-elevation) pollen taxa, and variation in aquatic (×1,000) productivity proxies (Fig. 2). The warm substages (MISs 11a, c and e) 6 6 4 are separated by intervals in which the temperature is ,2 uC lower 2 (MISs 11b and d). Although these small temperature variations are 0 within the error limits of the MBT/CBT calibration, their timing is 360 380 400 420 440 460 480 500 520 540 560 supported by decreases in warm pollen taxa and increases in boreal Age (kyr) pollen taxa (with the exception of MIS 11a). The warmest substage, Figure 2 | Multi-proxy profiles of VC-3 plotted versus calendar age. Age MIS 11e, occurs early in the interglacial, and has peak MATs of 6–7 uC model age–depth tie points are shown as arrows at the top and possible and the highest percentages of Juniperus pollen. After MIS 11e, the sedimentary hiatuses are indicated by positions of mud cracks (below). Shading warm substages become progressively cooler. indicates interglacial periods including odd-numbered MISs 11 and 13 and The preservation of five MIS 11 substages in VC-3 is unusual. Most substages within MIS 11 (a, c and e). a, June insolation at latitude 30u N (ref. 6). published records recognize only three substages, although a weak MIS b, Quercus (warm) pollen percentages. c, Juniperus (warm) pollen percentages. 11e was noted in the Lake Baikal biogenic silica record12 and there are d, MAT estimates from MBT/CBT, with data size marker equivalent to 2 uC three distinct (warm) peaks in MIS 11 pollen influx from Greenland (blue). Red line shows the modern MAT, of 4.8 uC, in the Valle Grande. MBT/ preserved in ODP Site 646 sediments13. The VC-3 MIS 11a substage is CBT temperature estimates have an absolute uncertainty of 5 uC based on uncertainties in the global calibration (see further discussion in Supplementary cooler than the extended warm phases of MIS 11, similar to other mid- Information). Millennial-scale events within the three glacial periods, defined Pleistocene climate records12, and is defined mainly by elevated lacus- 13 by local maxima in MAT and Cyperaceae and local minima in Picea and Abies, trine productivity (Si/Ti and TOC), more-positive d C values, slightly are indicated (A for MIS 10, B for MIS 12 and C for MIS 14). e, Picea 1 Abies higher temperature estimates and a combination of Quercus, Picea and (boreal) pollen percentages. f, Cyperaceae pollen percentages; mud cracks Abies pollen that may not have a good modern climatic analogue. indicated with orange diamonds. g, Si/Ti ratios from core scanning X-ray Within the limits of the VC-3 age model and the calibration uncer- fluorescence (XRF). Large peaks in MIS 14 correspond to pumiceous gravels. 13 13 12 tainty in the MBT/CBT proxies, the warm substages seem to corre- h, Total organic carbon (TOC). i, d CTOC 5 (( C/ C)sample/ 13 12 spond to the three precessional peaks of MIS 11, suggesting that the ( C/ C)standard 2 1) 3 1,000%, relative to the Vienna PeeDee Belemnite temperature response of this region to low-amplitude precessional (VPBD) standard. j, Watershed soil pH estimate from CBT. k, Calcium concentration in sediments from core scanning XRF. cycles is ,2 uC. On the basis of the MIS 11 orbital forcing similarity with the Holocene, we suggest that in the absence of anthropogenic MIS 1114,15, and a smaller Greenland ice sheet13 and a lack of ice rafting forcing future southwestern US climate should see a cooling of ,2 uC in the North Atlantic9 also indicate Northern Hemisphere warmth relative to the early Holocene. during MIS 13, although not necessarily more than during MIS 11. Large parts of MIS 13 seem to have been warmer than most of MIS In contrast, Southern Hemisphere records uniformly show a cooler 11, as shown by MATs of up to 7 uC, higher Juniperus pollen percen- MIS 1314,16. tages and the absence of Picea 1 Abies pollen (Fig. 2). Only MIS 11e The higher MIS 13 temperatures in the southwestern US occur had temperatures approaching the peak warmth of MIS 13. Other despite lower interglacial values of atmospheric CO2 and CH4 (ref. 17). Northern Hemisphere records suggest that MIS 13 was warmer than However, the amplitude of precessional cycles and, hence, extremes in

Northern Hemisphere insolation were larger in MIS 13 than MIS 116 Following the aridity of MIS 11e, the lake expanded during MIS 11d as and may have led to higher continental temperatures during parts of shown by well-laminated sediments and open-basin conditions (low MIS 13. Temperature variability during MIS 13 was as much as 6 uC, calcium values). Despite this interval being ,2 uC cooler, sufficient which is significantly larger than that during MIS 11 (Fig. 2). In com- summer rainfall early in MIS 11d allowed renewed C4 plant growth. bination with the apparent precessional timing of MIS 11 warm sub- Northern New Mexico at present receives ,40–50% of its annual stages, this suggests that the southwestern US responded more strongly precipitation total during the summer monsoon22. During the warmest to insolation variations than to interglacial trends in greenhouse gases phases of the interglacials, we would expect greater summer precipi- or global ice volume. tation, as the monsoon is primarily driven by land surface heating22. Mud cracks present in the warmest phases, MIS 11e and MIS 13, are Indeed, linkages among interglacial warmth, robust summer precipi- unambiguous indicators of drier conditions. One 70-cm mud crack tation and precessional variations are indicated by the presence of C4 occurs within MIS 11e, and ,3 m of section within the upper portion plants in early MIS 13, early MIS 11e and MISs 11c and 11a, when of MIS 13 sediments contains multiple, centimetre-scale mud cracks, MATs were similar to or slightly less than modern values, but the making this portion of the VC-3 age model less certain (Fig. 2). Also, warmest intervals did not have robust summer precipitation. As possible the presence or absence of calcite in VC-3 sediments provides a con- analogues for interglacial aridity, both historical droughts and pre- tinuous indicator of closed-basin or, respectively, open-basin condi- historical megadroughts were characterized by reductions in winter tions in the lake. No calcite precipitated during freshwater open-basin precipitation as a consequence of more-frequent La Nin˜a events22,3,23, conditions, whereas during drier (closed-basin) conditions, evaporative with summer precipitation reduced also. concentration led to calcite precipitation and preservation. XRF core In contrast, the extended arid episodes (centuries to millennia) of scanning results show two intervals with high calcium concentrations MIS 11e and MIS 13 lasted much longer than pre-historical mega- during MISs 11e and MIS 13 (Fig. 2) that correlate with elevated (1–2%) droughts. An analogous relationship between peak interglacial warmth total inorganic carbon (not shown), whereas core sections with low and extended aridity was also noted in a mid-Holocene bog record from calcium concentrations have essentially no total inorganic carbon. the margin of the Valles Caldera24.Here,,2 kyr of desiccation occurred Significant increases in calcium mark the onsets of closed-basin condi- contemporaneously with the highest temperatures of the Holocene in tions coincident with rapid temperature increases a few thousand years the southwestern US25 and with the northernmost extent of the inter- after Terminations V and VI (Fig. 2). Mud cracks develop later within tropical convergence zone in the Gulf of Mexico26. The timing of this these closed-basin periods. dry episode in the Holocene interglacial following the deglaciation is Long-term changes in watershed hydrology are also reflected in the very similar to that of the arid episode in MIS 11e; subsequent late- CBT-derived soil pH record. Changes in soil pH are assumed to reflect Holocene conditions became wetter in the southwestern US, with changes in total precipitation18; greater soil leaching and acidification increased winter precipitation27 similar in timing to wetter conditions occurs with more precipitation, whereas drier conditions result in during MIS 11d in the VC-3 record. weaker soil acidification. The most alkaline soils occur within the The strong correspondence between the warmest temperatures and interglacial mud-cracked facies (Fig. 2) and are more basic for longer extended aridity during at least three interglacials (MIS 13, MIS 11e periods in MIS 13. In contrast, soil pH shows a progressive acidifica- and the early Holocene) in the southwestern US suggests a stable tion through MIS 12, consistent with progressively wetter conditions climate state fundamentally different from conventional drought con- through that glacial stage, and possibly also caused by increases in ditions. These periods of aridity are related to lower winter precipita- boreal tree vegetation. tion (as mid-latitude westerlies shifted polewards during warmer During MIS 11e, Si/Ti (a proxy for diatom productivity) is initially periods), but reductions in summer precipitation seem to be critical very high and then declines to average interglacial values, whereas to their development. Unlike the temporary summer blocking high TOC increases to ,5% in MIS 11e and is also high during early MIS over the southwestern US thought to partly explain the 1950s 13 28 11d and MISs 11c and 11a. After the glacial termination, d CTOC drought , these longer periods of aridity indicate a more permanent rapidly increases to ,220%, indicating an expansion of C4 plants change in atmospheric circulation. Climate model analysis shows that in the watershed and higher lacustrine productivity levels19. Increases the dust-bowl-like conditions predicted for the southwestern US over 13 d CTOC also occurs in MISs 11c and 11a, and early in MIS 11d. the next century in response to anthropogenic warming arise from a 13 Continued high Si/Ti, TOC and more-positive d CTOC values during poleward shift of the mid-latitude westerlies and the poleward branch 4 early stages of closed-basin conditions in MIS 11e and MIS 13 indicate of the Hadley cell . This response to warming is not transient and periods of robust summer precipitation and productivity related to would result in a more arid southwestern US as long as the underlying insolation forcing of monsoon strength, even as reduced winter pre- conditions (warming) remained in place. Our palaeoclimate record cipitation led to less precipitation overall and closed-basin conditions. shows that extended interglacial aridity is strongly linked to higher- Mud-cracked facies in MIS 11e and MIS 13 are characterized by than-modern temperatures and reduced summer rainfall, and we 13 suggest that a similar expansion of the subtropical dry zone has negative shifts in d CTOC of 6–7% and dramatic decreases in percentage TOC (Fig. 2). Si/Ti ratios, however, remain elevated relative to occurred several times in the past in response to natural warming, glacial values, suggesting that low percentage TOC values are due to even though MIS 11 and MIS 13 had different orbital and atmospheric organic degradation in shallow, oxidized sediments rather than lower CO2 forcings. Our results strongly indicate that interglacial climates in 13 the southwestern US can experience prolonged periods of aridity, aquatic productivity. The large negative shifts in d CTOC are best lasting centuries to millennia, with profound effects on water avail- explained by a collapse of the interglacial C4 plant community. ability and ecosystem composition. The risk of prolonged aridity is Variations in C3 and C4 plant communities are a complex function likely to be heightened by anthropogenic forcing1,4. of temperature, atmospheric CO2, and growing-season precipita- tion20,21. These dry intervals include some of the highest MATs in METHODS SUMMARY the VC-3 record that should favour C4 plants, and the relatively high Measurement of fossil branched glycerol dialkyl glycerol tetraether (GDGT) mem- interglacial levels of atmospheric CO2 during MISs 11 and 13 vary by less than 20–30 p.p.m.v. Thus, the best explanation for the decline of brane lipids from soil bacteria were conducted at the Royal Netherlands Institute for Sea Research (NIOZ) and Brown University following procedures outlined in C4 plants in the watershed is a significant decrease in summer precipi- Supplementary Information. At NIOZ we analysed GDGTs on an Agilent 1100 series tation. In contrast to the early interglacial closed-basin phases where LC-MSD SL, and at Brown University we analysed GDGTs on an HP 1200 series LC- significant C4 plant growth provided evidence for robust summer MS. Both labs used an Alltech Prevail Cyano column (2.1 3 150 mm, 3 mm) with the precipitation, we interpret the extended arid periods later in MIS 11e same solvent elution scheme and instrument operating conditions. GDGTs were and MIS 13 to be the result of greatly reduced summer precipitation. detected using atmospheric-pressure chemical ionization mass spectrometry. All

520 | NATURE | VOL 470 | 24 FEBRUARY 2011 ©2011 Macmillan Publishers Limited. All rights reserved LETTER RESEARCH liquid chromatography/mass spectrometry runs were integrated at NIOZ by the 16. Jouzel, J. et al. Orbital and millennial Antarctic climate variability over the past same technician to ensure consistency. To evaluate the compatibility between the 800,000 years. Science 317, 793–796 (2007). 17. Loulergue, L. et al. Orbital and millennial-scale features of atmospheric CH4 over Brown and NIOZ measurements, representative samples were analysed on both the past 800,000 years. Nature 453, 383–386 (2008). machines and the resulting MBT/CBT indices were found to be identical within 18. Johnson, D. W., Hanson, P. J., Todd, D. E., Susfalk, R. B. & Trettin, C. F. Precipitation analytical uncertainty. change and soil leaching: field results and simulations from Walker Branch Processing for pollen included suspension in KOH, dilute HCL, hydrofluoric Watershed, Tennessee. Wat. Air Soil Pollut. 105, 251–262 (1998). acid and acetolysis solution. The pollen sum included all terrestrial pollen types; 19. Meyers, P. A. Applications of organic geochemistry to paleolimnological Cyperaceae percentages were calculated outside the sum. We identified pollen reconstructions: a summary of examples from the Laurentian Great Lakes. Org. Geochem. 34, 261–289 (2003). grains to the lowest taxonomic level using the modern pollen reference collection 20. Ehleringer, J. R., Cerling, T. E. & Helliker, B. R. C4 photosynthesis, atmospheric CO2 at Northern Arizona University. Analysis for organic carbon elemental concen- and climate. Oecologia 112, 285–299 (1997). 13 trations and d CTOC included samples being dried, ground and pretreated twice 21. Huang, Y. et al. Climate change as the dominant control on glacial-interglacial 13 with 6 N HCL at 60 uC to remove the carbonate fraction. TOC and d CTOC were variations in C3 and C4 plant abundance. Science 293, 1647–1651 (2001). analysed using a Costech Elemental Analyser coupled to a Thermo-Finnigan Delta 22. Douglas, M. W., Maddox, R. A., Howard, K. & Reyes, S. The Mexican monsoon. J. Plus isotope ratio mass spectrometer. The bulk elemental composition of core VC- Clim. 6, 1665–1677 (1993). 3 sediments was determined using an ITRAX X-ray Fluorescence Scanner (Cox 23. Schubert, S. D., Suarez, M. J., Pegion, P. J., Koster, R. D. & Bacmeister, J. T. On the cause of the 1930s dust bowl. Science 303, 1855–1859 (2004). Analytical Instruments). XRF scanning was conducted at 1-cm resolution with 60- 24. Anderson, R. S. et al. Development of the mixed conifer forest in northern New s scans using a molybdenum X-ray source set to 30 kV and 15 mA. Mexico and its relationship to Holocene environmental change. Quat. Res. 69, 263–275 (2008). Received 11 June 2010; accepted 12 January 2011. 25. Jime´nez-Moreno, G., Fawcett, P. J. & Anderson, R. S. Millennial- and centennial- scale vegetation and climate changes during the Late Pleistocene and Holocene 1. Woodhouse, C. A. et al. A 1,200-year perspective of 21st century drought in from northern New Mexico (USA). Quat. Sci. Rev. 27, 1448–1452 (2008). southwestern North America. Proc. Natl Acad. Sci. USA 107, 21283–21288 26. Poore, R. Z., Pavich, M. J. & Grissino-Mayer, H. D. Record of the North American (2010). southwest monsoon from Gulf of Mexico sediment cores. Geology 33, 209–212 2. McCabe, G. J., Palecki, M. A. & Betancourt, J. L. Pacific and Atlantic Ocean (2005). influences on multidecadal drought frequency in the United States. Proc. Natl 27. Enzel, Y., Cayan, D. R., Anderson, R. Y. & Wells, S. G. Atmospheric circulation during Acad. Sci. USA 101, 4136–4141 (2004). Holocene lake stands in the Mojave Desert: evidence of regional climate change. 3. Cook, E. R., Seager, R., Cane, M. A. & Stahle, D. W. North American drought: Nature 341, 44–47 (1989). reconstructions, causes and consequences. Earth Sci. Rev. 81, 93–134 (2007). 28. Namias, J. Some meteorological aspects of drought with special reference to the 4. Seager, R. et al. Model projections of an imminent transition to a more arid climate summers of 1952–54 over the United States. Mon. Weath. Rev. 83, 199–205 in southwestern North America. Science 316, 1181–1184 (2007). (1955). 5. Weijers, J. W. H., Schouten, S., van den Donker, J. C., Hopmans, E. C. & Sinninghe Damste´, J. S. Environmental controls on bacterial tetraether membrane lipid Supplementary Information is linked to the online version of the paper at distribution in soils. Geochim. Cosmochim. Acta 71, 703–713 (2007). www.nature.com/nature. 6. Berger, A. & Loutre, M. F. Insolation values for the climate of the last 10 million Acknowledgements We thank A. Mets for analytical support, W. McIntosh for the Ar–Ar years. Quat. Sci. Rev. 10, 297–317 (1991). age determination, T. Wawrzyniec and A. Ellwein for drilling help, and the Valles Caldera 7. Loutre, M. F. & Berger, A. Marine Isotope Stage 11 as an analog for the present Trust for permission to drill in the Valle Grande. Core assistance was provided by LRC/ interglacial. Global Planet. Change 36, 209–217 (2003). LacCore. This work was supported by the NSF Paleoclimate and P2C2 programs, IGPP 8. Weijers, J. W. H., Schefuß, E., Schouten, S. & Sinninghe Damste´, J. S. Coupled LANL and the USGS Western Mountain Initiative. Support from the Gledden Fellowship thermal and hydrological evolution of tropical Africa over the last deglaciation. is acknowledged. This work forms contribution 2399-JW at the Centre for Water Science 315, 1701–1704 (2007). Research, The University of Western Australia and contribution 131 at the Laboratory of 9. McManus, J. F., Oppo, D. W. & Cullen, J. L. A 0.5-million-year record of millennial- Paleoecology, Northern Arizona University. scale climate variability in the North Atlantic. Science 283, 971–975 (1999). 10. Kutzbach, J. E. et al. Climate and biome simulation for the past 21000 years. Quat. Author Contributions Writing and interpretation was done by P.J.F. with significant Sci. Rev. 17, 473–509 (2000). contributions from J.P.W., R.S.A., J.M.H. and E.T.B. MBT/CBT analyses were conducted 11. Prokopenko, A. A., Williams, D. F., Karabanov, E. B. & Khursevich, G. K. Response of by J.P.W., M.A.B., J.S.S.D., S.S., Y.H. and J.T. Organic carbon/nitrogen analyses were Lake Baikal ecosystem to climate forcing and pCO2 change over the Last Glacial/ conducted by P.J.F., J.M.H., L.M.C.-D., J.F. and V.A. XRF core scanning analyses were Interglacial transition. Earth Planet. Sci. Lett. 172, 239–253 (1999). conducted by E.T.B. Pollen analyses and palaeovegetation analyses were conducted by 12. Prokopenko, A. A. et al. Muted climate variations in continental Siberia during the R.S.A., S.J.S. and C.D.A., and F.G., G.W. and P.J.F. conducted core sediment and mid-Pleistocene epoch. Nature 418, 65–68 (2002). stratigraphic analyses. L.D.-H. and J.W.G. investigated palaeomagnetic and rock 13. de Vernal, A. & Hillaire-Marcel, C.Natural variabilityofGreenland climate, vegetation, magnetic core properties. All authors discussed the results and commented on the and ice volume during the past million years. Science 320, 1622–1625 (2008). manuscript. 14. Guo, Z. T., Berger, A., Yin, Q. Z. & Qin, L. Strong asymmetry of hemispheric climates during MIS-13 inferred from correlating China loess and Antarctic ice records. Author Information Reprints and permissions information is available at Clim. Past 5, 21–31 (2009). www.nature.com/reprints. The authors declare no competing financial interests. 15. Rossignol-Strick, M., Paterne, M., Bassinot, F. C., Emeis, K. C. & De Lange, G. J. An Readers are welcome to comment on the online version of this article at unusual mid-Pleistocene monsoon period over Africa and Asia. Nature 392, www.nature.com/nature. Correspondence and requests for materials should be 269–272 (1998). addressed to P.J.F. ([email protected]).

Supplementary Figures

Supplementary Figure 1 Photograph of mudcracked core section in MIS 13 from

46.4 m to 47.05 m depth. Scale bar in cm with dm marked by alternating yellow and

black bars. Note multiple mudcrack horizons, including prominent horizons ~5 cm from

the top (left) and ~56 cm from the top.

WWW.NATURE.COM/NATURE | 1 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

Supplementary Figure 2: Additional proxy profiles from core VC-3. Rosaceae and

Artemisia pollen counts from core VC-3 plotted vs. age (ka B.P.), bulk organic matter

δ15N (‰),C:N ratios of bulk organic matter, and bulk magnetic susceptibility (SI)

measured by a Geotek multisensor core logging system. MIS shading as in Figure 2.

Supplementary Methods

Measurement of fossil Glycerol Dialkyl Glycerol Tetraether (GDGT) membrane

lipids from soil bacteria was conducted at 20 cm to 1 m intervals in VC-3 sediments, with

most spaced at 40 cm (~1200 yrs). All VC-3 samples were carefully chosen to avoid

WWW.NATURE.COM/NATURE | 2 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

turbidites (based on visual and X-radiograph core inspection) and these are most common

in the lowest section of the core where our sampling density was lowest. GDGT analyses

were conducted at both NIOZ and Brown University following supplementary references

1 and 2. Freeze-dried, homogenized sediment was extracted with dichloromethane

(DCM)/methanol (2:1) using Soxhlet or 9:1 DCM/methanol using Dionex™accelerated

solvent extraction (ASE) techniques to obtain the total lipid extract (TLE). An aliquot of

the TLE was further separated into neutral, free fatty acid, and phospholipid fatty acid

fractions using Alltech Ultra-Clean SPE Aminopropylsilyl bond elute columns. Eight mL

each of 1:1 methylene chloride:2-propanol, 4% glacial acetic acid in ethyl ether, and

methanol were used to elute the neutral lipid, fatty acid, and phospholipid fatty acid

fractions, respectively. The neutral lipid fraction was further separated into apolar and

polar fractions using an activated Al2O3 column and eluting with hexane/DCM 9:1 and

DCM/methanol 1:1 (v/v) at NIOZ. At Brown University, the neutral lipid fraction was

separated into alkane, ketone and polar fractions using a silica gel column with four mL

each of hexane, DCM, and methanol (all GDGTs are found in the polar fractions eluted

with methanol). The polar fraction (DCM/methanol) was dried, dissolved in 99:1

hexane/2-propanol and filtered over a 0.45 µm PTFE filter prior to analysis.

At NIOZ, GDGTs were analyzed on an Agilent 1100 series LC-MSD SL, using

an Alltech Pevail Cyano column (2.1 x 150 mm, 3µm). The compounds were eluted

isocratically with 90% A and 10% B for 5min (flow rate 0.2ml/min), and then with a

linear gradient to 16% B for 34min, where A = hexane and B = hexane:isopropanol (9:1,

v/v). At Brown University GDGTs were analyzed with the same column, solvent elution

scheme and intrument operating conditions as at NIOZ. For both instruments, GDGTs

WWW.NATURE.COM/NATURE | 3 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

were detected using atmospheric pressure chemical ionization mass spectrometry (ACPI-

MS) according to supplementary references 1 and 2. Injection volume was 10 μl and

analyses were performed in selective ion monitoring mode to increase sensitivity and

reproducibility. Relative quantification of the compounds was accomplished by

integrating the [M+H]+ (protonated molecular ion) peaks in the mass chromatograms. All

LC/MS runs were integrated at NIOZ by the same technician using Agilent ChemStation

software. To evaluate the compatibility between the Brown and NIOZ measurements,

representative samples were analyzed on both machines and the resulting MBT/CBT

indices were found to be identical within analytical uncertainty. The analytical

reproducibility of results in this study based on replicate sample processing is 0.5°C for

the MAT estimate and 0.1 for pH.

The Branched and Isoprenoid Tetraether (BIT) index is based on the relative

abundance of terrestrially derived branched GDGT lipids vs. crenarchaeol (an isoprenoid

GDGT lipid derived from aquatic Crenarchaeota) and represents a measure for the

relative fluvial input of soil organic matter in sediments3-5. In VC-3 sediments, the BIT

index for the relative input of soil organic matter ranges from 0.83 to 1.0, with an average

value of 0.96 indicating a large input of soil organic matter1,3,5 as would be expected in

this lacustrine setting. The residence time of these branched tetraether lipids in soils

surrounding the paleolake is likely to be short relative to our sampling resolution as we

see no significant lag between MBT/CBT temperature estimates (lipids delivered

fluvially to the lake) and climatically sensitive pollen types primarily delivered to the

lake via wind (e.g. Quercus, Juniperus or Picea and Abies). There is no systematic

difference in BIT index between glacial and interglacial periods, although the lowest

WWW.NATURE.COM/NATURE | 4 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

values are all located near the bottom of the core in MIS 14. Branched GDGT membrane

lipids are thought to be derived predominantly from soil bacteria1,4,5, though recent

studies suggest that these compounds may also be produced in lacustrine and/or marine

sediments6-8. The relative distribution of branched GDGT lipids, expressed as the

Methylation index of Branched Tetraethers (MBT) and the Cyclisation ratio of Branched

Tetraethers (CBT) is a function of annual mean air temperature (MAT) and soil pH1. The

calibration equations for MBT and CBT are taken from supplementary reference 1, in

which a significant linear correlation with modern MAT ranging from -6 to 27oC was

shown:

CBT = 3.33 - 0.38 × pH [1]

MBT = 0.122 + 0.187 × CBT + 0.020 × MAT [2]

The original three dimensional global soil calibration for MBT and CBT had an r2

of 0.77, corresponding to a total calibration error of ~5°C and 1 pH unit1. However,

recent studies on Mt Kilimanjaro9 as well as in East African lakes10 have demonstrated

that local calibrations, particularly for lacustrine sediment reconstructions, are typically

much better than the global calibration. In the global calibration, both vegetation type and

soil type differences appear to contribute to the data scatter, and therefore to the

calibration error. The East African Lake study10 showed that with consistent soil and

vegetation types, the local calibration had an RMS error of ~2oC in the MBT/CBT

temperature estimate. The Mt Kilimanjaro study9 also showed a smaller error than the

WWW.NATURE.COM/NATURE | 5 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

global calibration, and noted that a dramatic vegetation change at the highest elevations

in the study altered the calibration relative to the lower elevation sites.

We therefore expect that the actual error in our temperature estimates is better

than 5°C, but at present there is no local calibration for lakes in the southwest of North

America. Based on the results of the East African Lake study10, we estimate that our

calibration error will be closer to ~2oC for the Valle Grande sediments once a local

calibration is developed. The documented vegetation changes in the VC-3 record (where

we estimate ~600 m of vegetation zone changes between glacial and interglacials) could

potentially alter this (cf. supplemental reference 9) and raises the possibility that the

absolute error in calibration is larger between glacial and interglacial periods than within

either a glacial or an interglacial. Furthermore, these and other studies 7,8 have also

identified a “cold bias” in the MBT-reconstructed temperatures from lake sediments

compared to soils. Thus, our temperatures should be taken as minimum values. For

these reasons, we believe that our MBT-reconstructed temperature trends are quite

robust, but the absolute values of the temperatures reconstructed are likely to be under-

estimates.

In the VC-3 record, two parts of the section are partially oxidized (described in

VC-3 core chronology and sediment characteristics section) which raises the possibility

of complicating the MBT/CBT proxy. However, recent studies show that the branched

GDGTs used in the MBT/CBT proxy are not significantly affected by oxidation of

organic material in sediments11,12. First, isoprenoid GDGT concentrations used in the

11 TEX86 paleothermometer are not sensitive to oxidizing conditions , and more

importantly, a study of branched GDGTs used in the MBT/CBT proxy in oxidized marine

WWW.NATURE.COM/NATURE | 6 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

sediments show that they are much more resistant to oxidation than the isoprenoid

GDGTs12. Thus, the oxidation in portions of the Valles Caldera sediments (resulting in

TOC values of ~0.5%) is unlikely to affect the distribution of branched GDGTs within

those sediments, and therefore the MBT/CBT based reconstructions of temperature and

soil pH.

VC-3 sediments were sampled continuously for pollen and spore content at

intervals ranging from 10 cm to 100 cm, with most spaced at 40 cm (~1200 yrs).

Processing for pollen followed supplementary reference 13 including suspension in

KOH, dilute HCL, HF and acetolesis solution. The pollen sum included all terrestrial

pollen types; Cyperaceae percentages were calculated outside the sum. Pollen grains were

identified to the lowest taxonomic level using the modern pollen reference collection at

Northern Arizona University.

The bulk elemental composition of core VC-3 sediments was determined using an

ITRAX X-ray Fluorescence Scanner (Cox Analytical Instruments) at the Large Lakes

Observatory, University of Minnesota Duluth. XRF core scanning was conducted at 1-cm

resolution (approximately 30 yrs) with 60 second scans using a Mo x-ray source set to 30

kV and 15 mA. For this work, we focus on Si:Ti, an indicator of the abundance of

biogenic opal in lacustrine sediments14. In core VC-3, diatom abundances estimated from

smear slides taken every 50 cm down the length of the core show good agreement with

the Si:Ti measurements. The high Si:Ti values in the lowest portions of the core

correspond directly with pumiceous gravels which have little to no diatom content. In

general, Si:Ti and TOC trends follow each other with two significant exceptions during

the interglacials – the later stage of MIS 13 and late MIS 11e where desiccation and

WWW.NATURE.COM/NATURE | 7 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

oxidation has removed significant amounts of organic carbon. The ITRAX XRF scanner

also provided 0.2 mm resolution x-radiographs of the core.

Sediments were sampled continuously at a 20 cm resolution (~600 yrs) and

analyzed for organic carbon and nitrogen elemental concentration and δ15N and δ13C

values. Samples were dried, ground and pretreated twice with 60oC 6N HCL in order to

remove the carbonate fraction. The hot acid treatment was applied because some of the

samples contain small rhombs of diagenetic siderite, which may not be entirely removed

by standard HCl treatment15. Total Organic Carbon (TOC), Total Organic Nitrogen

(TON) and bulk δ13C and δ15N were analyzed using a Costech Elemental Analyzer

coupled to a Thermo-Finnigan Delta Plus Isotope Ratio Mass Spectrometer. Laboratory

standards were run at the beginning, at intervals between samples and at the end of

analytical sessions. The analytical precision calculated from duplicate measurements of

the standards is ± 0.1‰ for δ13C and ± 0.15‰ for δ15N (1σ standard deviation).

Physical and Climatic Setting of the Valle Grande

The Valles Caldera complex formed after a large shallow magma chamber erupted, at

1.23 Ma16. Over the next million years, a series of rhyolite domes erupted from ring

fracture vents following the caldera collapse and initial resurgence. Several lakes formed

in the moat valleys after 0.8 Ma when the post-caldera eruptive domes temporarily

blocked drainages out of the caldera17. A particularly long-lived lake formed in the Valle

Grande (35o 52’ N, 106o 28’ W, 2553 m asl) following the eruption of the South

Mountain rhyolite dome17,18 which dammed the drainage to the southwest from the basin

WWW.NATURE.COM/NATURE | 8 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

(Figure 1). The Valle Grande catchment has a total area of ~100 km2 (Figure 1) and the

dominant bedrock within the catchment is almost exclusively rhyolite16. The largest lake

present had a maximum surface area of ~20 km2.

Vegetation communities in the Jemez Mountains are strongly affected by aspect19.

At the highest elevations (ca. 2900 to 3500 m), north-facing slopes contain forests of

Picea engelmannii and Abies lasiocarpa var. arizonica, while local areas of high

elevation grasslands are found on south-facing slopes20. Mixed conifer forests occur

from ca. 2300 to 2900 m, with Pinus ponderosa, Pseudotsuga menziesii, A. concolor,

Populus tremuloides and Pinus flexilis. P. ponderosa with shrub oak (Quercus gambelii)

dominates elevations 2100 to ca. 2600 m, while piñon (Pinus edulis) – juniper (Juniperus

monosperma or J. scopulorum) woodlands occur between ca. 1900 and 2100 m elevation.

Juniper grasslands (J. monosperma, big sagebrush [Artemisia tridentata], Bouteloua sp.)

occur from ca. 1600 to 1900 m, while shrub steppe with members of the goosefoot and

aster families (i.e., A. tridentata, Sarcobatus) and joint-fir (Ephedra) occur at elevations

below this.

Modern pollen studies within southern Colorado and northern New Mexico by

Susan Smith, Scott Anderson and others help to elucidate the fossil pollen record. The

percentages of piñon (Pinus edulis), juniper (Juniperus) and oak (Quercus) from MIS 11

and 13 are very similar to those from the modern piñon – juniper (PJ) woodland of today.

For instance, in modern PJ, piñon, juniper and oak constitutes about 20% or more, 10%

and 10%, respectively, of the pollen sum21. The difference between the modern and

fossil assemblages is that juniper is much higher. Clearly, if these species were not

WWW.NATURE.COM/NATURE | 9 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

growing directly around the lake, they probably grew very close to the site, perhaps

below the rim of the Caldera.

Similarly, the percentages of dominant species during MIS 10, 12 and 14 are very

similar to those from modern spruce (Picea) – fir (Abies) forest in northern New Mexico

(Jicarita Bog, 3207 m elevation22; Stewart Bog, 3100 m23) and southern Colorado (Little

Molas Lake, 3370 m24; Hunters Lake, 3516 m25). This includes percentages of Picea to

20%; Abies to 5%; Pinus 20-40%, and Poaceae 5-10%. On these bases, we ascribe

changes in apparent elevation of the vegetation within the VC-3 core (2553 m elevation)

of at least 600 m between interglacial and glacials.

Modern climate data for the Valle Grande was derived from a USCRN co-op

weather station (NM Los Alamos 13 W) located at the Valles Caldera National Preserve

headquarters, on the flank of the VC-3 paleolake valley. The Mean Annual Temperature

of this site is 4.8oC with summer (JJA) temperatures averaging 14.0oC and winter (DJF)

temperatures averaging -4.5oC. Most of the annual average precipitation (~670 mm total)

occurs during the summer and winter with drier spring and fall seasons. Summer (JAS)

precipitation totals average 260 mm (~40% of the annual total) and is mainly convective

(monsoon), while winter (DJF) frontal precipitation totals average 180 mm for the three

months (~26% of the annual total).

Core VC-3 Chronology and Sediment Characteristics

Previous work on sediments from the Valle Grande included a low-resolution pollen

study on cuttings from a deep USGS water well, where two major cycles of wet to dry

WWW.NATURE.COM/NATURE | 10 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

climates (consistent with two glacial cycles) were interpreted from ratios of Picea

(spruce), Abies (fir), Pinus (pine) and Quercus (oak) pollen26. As numerical age control

was not available at the time of that study, the uppermost lacustrine sediments were

assumed to be Holocene to late Pleistocene in age. Subsequent Ar-Ar dating of the South

Mountain rhyolite27 dam forming the paleolake provided a middle Pleistocene age and an

updated mid-Pleistocene age for VC-3 lacustrine sediment was shown by an Ar-Ar date

of 552 ± 3 kyr from a basal tephra18.

The VC-3 core hole was drilled within the floor of the caldera away from the

steep slopes of the topographic margin, the resurgent dome, and the post-caldera eruptive

center to the east, west, and to the north of the well, respectively. Volcaniclastic

sediments eroded from these topographic highs were directly carried into the ancestral

lake. At least five compositionally discrete tephra stratigraphic markers within the VC-3

sequence represent sporadic pulses of volcaniclastic input directly into the lake without

transient sedimentation traps. The earliest marker tuffs within the core originated to the

south of VC-3 and are probably related to fallout eruptions, whereas subsequent units

were of fluvial origin supplied from the north.

Sediments recovered in core VC-3 consist of lacustrine silts and clays with

occasional gravels in the lower strata, and pumiceous gravels and indurated muds at the

base that probably represent a fluvial/meadow environment prior to the eruption of the

South Mountain rhyolite and damming of the Valle Grande drainage18. Above the basal

gravels and turbidites, two major lithologic cycles of about the same thickness are

defined by transitions from silty diatomaceous clays with well-defined, thin (mm)

laminations to silty diatomaceous clays with thicker and less well defined laminations (to

WWW.NATURE.COM/NATURE | 11 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

massive units lacking lamination) with occasional mm-scale silts and sands. Within this

latter facies, cm scale mudcracks, bioturbation and color mottled, gleyed horizons are

also locally present from 47.5 to 44 m depth (e.g. Supplementary Figure 1) and again at

~23 m depth. The basal transitions from thinly laminated to thickly laminated massive

clays are relatively abrupt (cm-scale) while the transitions back to thinly laminated clays

are more gradational (m-scale). We interpret the well-laminated sediments as forming

under deep lake conditions during the glacial periods and the less well-laminated to

mudcracked sediments forming in shallower to desiccated lake conditions during drier

interglacials. The mid-Pleistocene glacial deep and interglacial shallow pattern is

consistent with late Pleistocene, LGM-Holocene lake level changes in the Southwest28.

Two basal tephra Ar-Ar dates of 552 ± 3 ka from a glassy tephra and a well-sorted

pumiceous gravel at 75.8 and 76.0 m depth respectively constrain the onset of lacustrine

sedimentation in core VC-3 to MIS 14 in the middle Pleistocene18 (Supplementary Table

1). Trace element analyses of the tephra and pumiceous gravel show a close match with

the South Mountain Rhyolite, as well as with a tephra mapped outside the caldera dated

at 551 ± 2 ka29,30, further confirming the basal date for VC-3. This is overlain by ~20 m

of sediment that accumulated rapidly, with deposition of numerous turbidites and gravels

as unstable volcanic landforms eroded. Higher in the core, within the finer grained

lacustrine sequence, there are abrupt and dramatic increases in TOC, Si/Ti ratios, MAT

13 estimates, δ CTOC, C:N and warm taxa pollen types (Quercus, Juniperus, Rosaceae) at

51.5 m depth and again at 27.5 m (Figure 2 and Supplementary Figure 2). We correlate

these dramatic climate shifts with glacial terminations VI and V and their globally

constrained ages of 533 ka and 426 ka respectively31. Paleomagnetic work (progressive

WWW.NATURE.COM/NATURE | 12 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

alternating field demagnetization) has identified two short-lived magnetic field “features”

at 17.25 m and at 52.3 m depth in the core, where negative inclination magnetizations

have been either well-resolved or implied by demagnetization trajectories. We correlate

these negative inclination features with globally recognized events 11α (400 ka) and 14α

(536 ka)32 (Supplementary Table S1). The positions of these dated magnetic field events

are consistent with the age model based on glacial termination correlations.

At 43.1 m depth, dramatic decreases in Si:Ti, warm taxa pollen types, and MAT

estimates occur, indicating the end of MIS 13 warmth. A simple linear interpolation

assuming constant sedimentation rates between Glacial Terminations VI and V would

result in the end of MIS 13 occurring at ~505 ka. However, the end of MIS 13 / onset of

MIS 12 has a very consistent age of 480 ka in several long Pleistocene climate records

from different environments31,33-35. Additionally, MIS 13 is an extraordinarily long

interglacial that spans ~50 kyr (e.g.,; ~50 kyr Dome C Antarctica35; ~50 kyr in Lake

Baikal33,34; ~52 kyr deep sea ice volume record31; but ~65 kyr in the Devils Hole

record36). MIS 11 spans a similar amount of time (~50 ka)31,33-35, but in VC-3 it is

represented by 14 m of sediment whereas MIS 13 is represented by 9 m of sediment. This

5 m sediment preservation difference between these two interglacials is probably due to

sediment hiatuses as shown by the presence of multiple (12 to 15) mudcracks

(Supplementary Figure 1) in the upper half of MIS 13 sediments (from 47.5 m to 44.0 m

depth), although we cannot absolutely rule out lower sedimentation rates during MIS 13.

Abrupt changes in MAT estimates, specific pollen taxa counts and other proxies occur

across several of the more prominent mudcracks suggesting that these do represent

hiatuses. To account for this ~5 m of missing sediment, we assign an age of 480 ka to the

WWW.NATURE.COM/NATURE | 13 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

43.1 m depth corresponding to the end of MIS 13. Because we cannot assign precise ages

to the mudcracked sediments between 47.45 m and 43.1 m, we apply a simple linear

interpolation between these depths as the most parsimonious age model for this portion of

the core.

As the lower half of MIS 13 does not contain any obvious mudcracks/hiatuses, a

linear interpolation from the start to end of MIS 13 would result in anomalously high

sedimentation rates the upper interval with hiatuses and anomalously low sedimentation

rates in the lower interval without hiatuses. To determine an approximate age for the

47.45 m depth tie-point (just below the lowest mudcrack), we extrapolated the upper core

average sedimentation rate of 0.32 mm/yr from glacial termination VI (51.5 m depth) to

this point. This gives an age of 520 ka at 47.45 m depth.

The VC-3 lacustrine record terminates in the late middle Pleistocene, likely

following a breach of the South Mountain rhyolite dam as the available accommodation

space in the caldera moat filled. Because there are no available age-depth tie-points near

the top of the core, we apply the upper core average sedimentation rate to the sediments

between 17.25 m depth (400 Ma) and the lacustrine core top at 5.5 m depth. This

extrapolation results in an age estimate for the top of the core of ~363 ka.

WWW.NATURE.COM/NATURE | 14 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

Table S1: Age-depth tie-points for core VC-3 age model.

Depth (m) Age (ka) Nature of age tie-point

17.25 400 Geomagnetic Field Event 11α32

27.50 426 Termination V31

43.1 480 End of MIS 13 (see text)

47.45 520 Onset of mudcracked interval in MIS 13 (see text)

51.50 533 Termination VI31

52.30 536 Geomagnetic Field Event 14α32

75.60 552 Ar-Ar age date18

Additional VC-3 Core Properties

Artemisia (sage) pollen tends to be higher during glacial periods, and increases

through the complete MIS 12 glacial as conditions build toward the glacial maximum

(Supplementary Figure 2). It is almost absent during MIS 13, consistent with the warm

temperatures occurring during this interglacial, and it is highly variable during MIS 11

with low values during the warm substages and very high values during the cool

substages (11b and d). Rosaceae pollen (probably mostly Cercocarpus, but also Cowania

or Amelanchier) is highest during the interglacials, particularly evident during the first

two warm substages of MIS 11 (c and e). Rosaceae pollen is low during MIS 11a,

WWW.NATURE.COM/NATURE | 15 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

consistent with this being the coolest of the three warm substages (Supplementary Figure

2).

C:N ratios in bulk organic matter parallel other indicators of lacustrine aquatic

productivity. Average glacial values range between 5 and 10 while interglacial values

range between 10 and 15 (Supplementary Figure 2). Decreases in C:N ratios in during the

mudcracked, low TOC intervals during MIS 11e and MIS 13 are consistent with intense

decomposition of organic matter37. The degradation of organic matter leads to enrichment

in 15N, explaining the high δ15N values during these intervals (Supplementary Figure 2),

but does not have a large effect on δ13C values38.

The rock magnetic signature of the upper dry MIS 13 interval is exceptional with

high magnetic susceptibility (MS) values (Supplementary Figure 2). The normal

interglacial MS values are much lower (MIS 11a,c,e; early MIS 13) with biogenic silica

and organic carbon dilution of the clastic sediments. This dilution effect is also evident in

the bulk sediment density (not shown). The high MS values, which coincide with the

mudcracked portion of MIS 13, are probably caused by diagenetic formation of

metastable iron oxides and minor amounts of greigite in a shallow lake with fluctuating

water tables39. The presence of multiple gleyed horizons over this interval indicates rapid

changes in sediment redox conditions, consistent with shallow fluctuation water tables.

WWW.NATURE.COM/NATURE | 16 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

Supplemental References

1. Weijers, J.W.H., Schouten, S., van den Donker, J.C., Hopmans, E.C., Sinninghe Damsté, J.S. Environmental controls on bacterial tetraether membrane lipid distribution in soils. Geochim. Cosmochim. Acta 71, 703-713 (2007)

2. Schouten, S., Huguet, C., Hopmans, E.C., Kienhuis, M.V.M., Sinninghe Damsté, J.S., Analytical methodology for TEX86 paleothermometry by high-performance liquid chromatography/atmospheric pressure chemical ionization-mass spectrometry. Anal. Chem. 97, 2940-2944 (2007).

3. Hopmans, E.C., Weijers, J.W.H., Schefuß, E., Herfort, L., Sinninghe Damsté, J.S., Schouten, S. A novel proxy for terrestrial organic matter in sediments based on branched and isoprenoid tetraether lipids, Earth Planet Sc. Lett. 224, 107-116 (2004).

4. Weijers, J.W.H., Schouten, S., Spaargaren, O.C., Sinninghe Damsté, J.S. Occurrence and distribution of tetraether membrane lipids in soils: Implications for the use of the TEX86 proxy and the BIT index. Org. Geochem. 37, 1680-1693 (2006).

5. Sinninghe Damsté, J.S., Ossebaar, J., Abbas, B., Schouten, S. & Verschuren, D. Fluxes and distribution of tetraether lipids in an equatorial African lake: constraints on the application of the TEX86 palaeothermometer and BIT index in lacustrine settings. Geochim. Cosmochim. Acta 73, 4243-4249 (2009).

6. Peterse, F., Kim, J. H., Schouten, S., Kristensen, D. K., Koc, N., and Damste, J. S. S. Constraints on the application of the MBT/CBT palaeothermometer at high latitude environments (Svalbard, Norway). Org. Geochem. 40, 692-699 (2009).

7. Tierney, J. E., and Russell, J. M. Distributions of branched GDGTs in a tropical lake system: Implications for lacustrine application of the MBT/CBT paleoproxy. Org. Geochem. 40, 1032-1036, (2009).

8. Blaga, Cornelia I., Reichart, G.-J., Schouten, S., Lotter, A.F., Werne, J.P., Sinninghe Damsté, J.S. Branched glycerol dialkyl glycerol tetraethers in lake sediments: Can they be used as temperature and pH proxies? Org. Geochem. (2010), doi: 10.1016/j.orggeochem.2010.07.002

9. Sinninghe Damsté, J. S., Ossebaar, J., Schouten, S. and Verschuren, D. Altitudinal shifts in the branched tetraether lipid distribution in soil from Mt. Kilimanjaro (Tanzania): Implications for the MBT/CBT continental palaeothermometer. Org. Geochem. 39, 1072-1076 (2008).

WWW.NATURE.COM/NATURE | 17 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

10. Tierney, J.E., J.M. Russell, H. Eggermont, E.C. Hopmans, D. Verschuren, J.S. Sinninghe Damsté (in press) Environmental controls on branched tetraether lipid distributions in tropical East African lake sediments. Geochimica et Cosmochimica Acta (2010).

11. Kim, J-H., Hugeut, C., Zonneveld, K.A., Versteegh, G.J.M., Roeder, W., Sinninghe Damsté, J.S., and Schouten, S. An experimental field study to test the K stability of lipids used for the TEX86 and U37 palaeothermometers, Geochim. Cosmochim. Acta 73, 2888-2898 (2008).

12. Huguet, C., de Lange, G.J., Gustafsson, Ö, Middelburg, J.J., Sinninghe Damsté, J.S., and Schouten, S. Selective preservation of soil organic matter in oxidized marine sediments (Madeira Abyssal Plain), Geochim. Cosmochim. Acta 72, 6061- 6068 (2008).

13. Fægri, K., and J. Iversen, 1989, Textbook of Pollen Analysis. 4th ed., John Wiley, Chichester.

14. Brown, E.T., Johnson, T.C., Scholz, C.A., Cohen, A.S., and King, J.W. Abrupt change in tropical Africa climate linked to the bipolar seesaw over the past 55,000 years. Geophys. Res. Lett. 34, doi:10.1029/2007GL031240 (2007).

15. Larson, T.E., Heikoop, J.M., Perkins, G., Chipera, S.J. and Hess, M.A. Pretreatment technique for siderite removal in geological samples. Rapid Commun. Mass Sp. 22, 865-872 (2008).

16. Goff, F., and Gardner, J.N. Evolution of a mineralized geothermal system, Valles Caldera, New Mexico. Econ. Geol. 89, 1803-1832, (1994).

17. Reneau, S.L., Drakos, P.G., and Katzman, D., Post-resurgence lakes in the Valles Caldera, New Mexico, New Mexico Geological Society Guidebook, 58th Field Conference, Geology of the Jemez Mountains Region II, 398-408 (2007).

18. Fawcett, P.J., Heikoop, J., Goff, F., Anderson, R.S., Donohoo-Hurley, L., Geissman, J.W., WoldeGabriel, G., Allen, C.D., Johnson, C.M., Smith, S.J., and Fessenden-Rahn, J. Two Middle Pleistocene Glacial-Interglacial Cycles from the Valle Grande, Jemez Mountains, northern New Mexico; New Mexico Geological Society Guidebook, 58th Field Conference, Geology of the Jemez Mountains Region II, 409-417 (2007).

19. Anderson, R.S., Jass, R.B., Toney, J.L., Allen, C.D., Ciseros-Dozal, L.M., Hess, M., Heikoop, J., Fessenden, J., Development of mixed conifer forest in northern New Mexico, and its relationship to Holocene environmental change. Quat. Res. 69, 263-275 (2008).

WWW.NATURE.COM/NATURE | 18 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

20. Reneau, S.L., McDonald, E.V., Gardner, J.N., Kolbe, T.R., Carney, J.S., Watt P.M., and Longmire, P.A., Erosion and deposition on the Pajarito Plateau, New Mexico, and implications for geomorphic responses to late Quaternary climate changes, New Mexico Geological Society Guidebook, 58th Field Conference, Geology of the Jemez Mountains Region, 391-398 (1996).

21. Murray, S.S., Adams, K.R., Smith, S.J., Anderson, R.S., Anderson, K.C., The Ridges Basin modern plant environment. In Potter, JM (ed.), Animas-La Plata Project: Volume X – Environmental Studies. SWCA Anthropological Research Paper 10, SWCA Environmental Consultants, Phoenix (2008).

22. Bair, A.N., A 15,000 year vegetation and fire history record, southern Sangre de Cristo Mountains, New Mexico. MS thesis, Northern Arizona University (2004).

23. Jiménez-Moreno, G., Fawcett, P.J., Anderson, R.S., Millennial- and century-scale vegetation and climate changes during the Late Pleistocene and Holocene from northern New Mexico (USA). Quat. Sci. Rev. 27, 1442-1452 (2008).

24. Toney, J.L., Anderson, R.S., A postglacial paleoecological record from the San Juan Mountains of Colorado: fire, climate and vegetation history. The Holocene 16, 505-517, (2006).

25. Anderson, R.S., Allen, C.D., Toney, J.L., Jass,R.B., and Bair, A.N., Holocene vegetation and fire regimes in subalpine and mixed conifer forests, southern Rocky Mountains, USA., International Journal of Wildland Fire 17, 96-114 2008.

26. Sears, P.B. & Clisby, K.H. Two long climate records. Science 116, 176-178 (1952).

27. Spell, T.L. & Harrison, T.M., 40Ar/39Ar geochronology of post-Valles Caldera rhyolites, Jemez Mountains volcanic field, New Mexico, J. Geophys. Res. 98, 8031-8051 (1993).

28. Smith, G.I., and Street-Perrott, F.A., Pluvial lakes of the western United States, in Wright, H.E., Jr., and S.C. Porter, eds., Late Quaternary environments of the United States, Volume 1: The Late Pleistocene, Minneapolis, University of Minnesota Press, 190-214 (1983).

29. Smith, G.A., McIntosh, W., and Kuhle, A.J. Sedimentologic and geomorphic evidence for seesaw subsidence of the Santo Domingo accommodation-zone basin, Rio Grande rift, New Mexico, Geol. Soc. America Bull. 113, 561-574. (2001).

30. WoldeGabriel, G., Heikoop, J., Goff, F., Counce, D., Fawcett, P.J., and Fessenden-Rahn, J. Appraisal of post-South Mountain volcanism lacustrine

WWW.NATURE.COM/NATURE | 19 doi:10.1038/nature09839 RESEARCH SUPPLEMENTARY INFORMATION

sedimentation in the Valles Caldera using tephra units. New Mexico Geological Society Guidebook, 58th Field Conference, Geology of the Jemez Mountains Region II, 83-85. (2007).

31. Lisiecki, L.E., and Raymo, M.E. A Pliocene-Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography 20, PA1003, doi:10.1029/2004PA001071 (2005).

32. Lund, S., Stoner, J.S., Channell, J.E.T., and Acton, G. A summary of Bruhnes paleomagnetic field variability recorded in Ocean Drilling Program cores. Phy. Earth Planet. In. 156, 194-204 (2006).

33. Prokopenko. A.A., Williams, D.F., Kuzmin, M.I., Karabanov, E.B., Khursevich, G.K., and Peck, J.A., Muted climate variations in continental Siberia during the mid-Pleistocene epoch. Nature 418, 65-68 (2002).

34. Prokopenko, A.A., Hinov, L.A., Williams, D.F., and Kuzmin, M.I. Orbital forcing of continental climate during the Pleistocene: a complete astronomically tuned climatic record from Lake Baikal, SE Siberia. Quat. Sci. Rev. 25, p. 3431-3457 (2006).

35. Jouzel, J.et al. Orbital and millennial Antarctic climate variability over the past 800,000 years. Science 317, 793–796 (2007).

36. Winograd, I.J., Landwehr, J.M., Ludwig, K.R., Coplen, T.Y., and Riggs, A.C. Duration and structure of the past four interglaciations. Quat. Res. 48, 141-154 (1997).

37. Chen, X., Cabrera, M.L., Zhang, L., Shi, Y., Shen, S.M. Long-term decomposition of organic materials with different carbon/nitrogen ratios. Comm. Soil Sci. Plan. 34, 41-54 (2003).

38. Kramer, M.G., Sollins, P., Sletten, R.S. & Swart, P.K. N isotope fractionation and measures of organic matter alteration during decomposition. Ecology 84, 2021- 2025 (2003).

39. Donohoo-Hurley, L., Geissman, J.W., Fawcett, P.J., Wawrzyniec, T., and Goff, F. A 200 kyr lacustrine record from the Valles Caldera: Insight from environmental magnetism and paleomagnetism, New Mexico Geological Society Guidebook, 58th Field Conference, Geology of the Jemez Mountains Region II, 424-432 (2007).

WWW.NATURE.COM/NATURE | 20 NEWS & VIEWS RESEARCH different geometry from that described by Wat- concerned is that the constraints of the double computational studies on both naked double son and Crick (Fig. 1a, b). Similarly, an alterna- helix don’t preclude this possibility. The pres- helices and protein–DNA complexes. ■ tive base-pairing geometry can occur for G•C ence of Hoogsteen base pairs in detectable pairs. Hoogsteen pointed out that if the alterna- amounts — even in free DNA — therefore Barry Honig is at the Howard Hughes tive hydrogen-bonding patterns were present provides a notable example of the remark- Medical Institute, Center for Computational in DNA, then the double helix would have to able plasticity of the canonical double helix. It Biology and Bioinformatics, Department assume a quite different shape. Hoogsteen base also implies that, if DNA bases are regarded as of Biochemistry and Molecular Biophysics, pairs are, however, rarely observed. letters, each letter potentially has two mean- Columbia University, New York, New York Nikolova and colleagues’ key finding1 is ings that determine both hydrogen-bonding 10032, USA. Remo Rohs is in the Molecular that, in some DNA sequences, especially CA patterns and structural variations in the double and Computational Biology Program, and TA dinucleotides, Hoogsteen base pairs helix. Department of Biological Sciences, University exist as transient entities that are present in Structural biologists have long recognized of Southern California, Los Angeles, California thermal equilibrium with standard Watson– that there is no second code in which certain 90089, USA. Crick base pairs. The detection of the transient amino acids recognize complementary DNA e-mails: [email protected]; [email protected] species required the use of NMR techniques bases in protein–DNA interactions. Never- 1. Nikolova, E. N. et al. Nature 470, 498–502 (2011). that have only recently been applied to theless, protein–DNA binding is still com- 2. Parker, S. C. J., Hansen, L., Abaan, H. O., Tullius, T. D. macromolecules6. monly thought of purely in terms of codes and & Margulies, E. H. Science 324, 389–392 (2009). Why is this finding important? Hoog- sequence motifs, rather than as the binding 3. Rohs, R. et al. Nature 461, 1248–1253 (2009). 4. Watson, J. D. & Crick, F. H. C. Nature 171, 737–738 steen base pairs have, after all, previously of two large macromolecules that have com- (1953). been observed in protein–DNA complexes7–9 plex shapes and considerable conformational 5. Hoogsteen, K. Acta Crystallogr. 16, 907–916 (1963). (Fig. 1c). But it has not been possible to deter- flexibility10. Nikolova and colleagues’ discov- 6. Palmer, A. G. & Massi, F. Chem. Rev. 106, 1700–1719 (2006). mine whether Hoogsteen base pairs are pre- ery reminds us that DNA offers proteins not 7. Aishima, J. et al. Nucleic Acids Res. 30, 5244–5252 sent in free DNA. Nikolova and colleagues’ only an enticing linear alphabet, but also a set (2002). study1 reveals that the ability to flip between of conformations that can be recognized in a 8. Nair, D. T., Johnson, R. E., Prakash, S., Prakash, L. & Aggarwal, A. K. Nature 430, 377–380 (2004). Watson–Crick and Hoogsteen base pairing sequence-dependent way. Understanding how 9. Kitayner, M. et al. Nature Struct. Mol. Biol. 17, in free DNA is an intrinsic property of indi- the linear sequence of bases in DNA is recog- 423–429 (2010). vidual sequences. This implies that some nized by proteins is therefore a problem that 10. Rohs, R. et al. Annu. Rev. Biochem. 79, 233–269 (2010). proteins have evolved to recognize only one must be solved in three dimensions. This will 11. Chen, Y., Dey, R. & Chen, L. Structure 18, 246–256 base-pair type, and use intermolecular interac- require structural, biochemical, genomic and (2010). tions to shift the equilibrium between the two geometries9. DNA has many features that allow its CLIMATE CHANGE sequence-specific recognition by proteins. This recognition was originally thought to primarily involve specific hydrogen-bonding interactions between amino-acid side chains Old droughts in and bases. But it soon became clear that there was no identifiable one-to-one correspondence — that is, there was no simple code to New Mexico be read. Part of the problem is that DNA can undergo conformational changes that distort A long climate record reveals abrupt hydrological variations during past the classical double helix. The resulting varia- interglacials in southwestern North America. These data set a natural benchmark tions in the way that DNA bases are presented for detecting human effects on regional climates. See Letter p.518 to proteins can thus affect the recognition mechanism. More significantly, it has become evident JOHN WILLIAMS Dust Bowl4. These droughts are linked to that distortions in the double helix are them- yearly-to-decadal variations in sea surface selves dependent on base sequence. This outhwestern North America is a pretty temperatures in the tropical Pacific, enhanced enables proteins to recognize DNA shape in dry place, and is likely to get drier this by local soil-moisture feedbacks5. Further a manner reminiscent of the way that they century because of anthropogenic understanding of the mechanisms of hydro- recognize other proteins and small ligand Sclimate warming. On page 518 of this issue1, logical variability, together with efforts to molecules. For example, stretches of A and Fawcett et al. provide a climate record from limit societal vulnerability to climate change, T bases can narrow the minor groove of deep in the past that will help in assessing the are priorities in global-change research4. DNA (the narrower of the two grooves in the future hydrological regime for the area. Palaeoclimatic studies have made an essen- double helix), thus enhancing local negative Climate models consistently project tial, if not particularly reassuring, contribu- electrostatic potentials and creating bind- declines in winter precipitation for the tion to this effort, by providing insight into ing sites for appropriately placed, positively southwest, in response to rising greenhouse the natural behaviour of hydrological sys- charged arginine amino-acid residues3. gases as the subtropical dry zones expand tems and a longer-term context for histori- Nikolova and colleagues’ discovery1 that polewards2,3. This precipitation decline, cal and projected changes. Tree-ring records DNA base pairs can so easily leave their combined with expected increases in evapo- offer sobering evidence of widespread and favoured Watson–Crick conformation comes ration rates and reduced snowpack, would decades-long ‘megadroughts’ in the western as a surprise. But viewed in another way, the severely strain the region’s capacity to adapt United States over the past several millennia phenomenon is not so different from protein to climate change. Moreover, the southwest is that dwarfed recorded historical droughts6. side chains undergoing a conformational historically prone to droughts, with six multi- Records spanning the past 11,000 years change so as to optimize binding with another year droughts in the nineteenth and twenti- (the Holocene interglacial) demonstrate protein. The real surprise where DNA is eth centuries, including the infamous 1930s long-term shifts in southwestern monsoon

thousand years, and apparently began and ended abruptly. At least two other multi- D. USNER millennial dry periods occurred during MIS 11, and higher-frequency variations in total organic carbon and δ13C hint at sub- millennial hydrological variability during the warm stages of MIS 11. Mud cracks and low total organic carbon also characterize much of MIS 13, implying that this interglacial was similarly marked by recurring episodes of high temperatures and aridity. Interestingly, MIS 13 seems to have been warmer and drier than MIS 11, despite lower concentrations of atmospheric carbon dioxide and methane during MIS 13 (refs 11, 12). This suggests that the larger variations in insolation during MIS 13 strongly regulated southwestern aridity. What lessons can we draw from the Valles Caldera record? First, in southwestern North America, hydrological variability seems to be the rule rather than the exception during Figure 1 | An aerial view of Valles Caldera, site of the lake sediments from which Fawcett et al.1 interglacial periods. Second, on timescales of 3 4 extracted their climate record. 10 to 10 years, orbital precession strongly influences southwestern monsoon intensity intensity7. Fawcett et al.1 now present a lake hydrology (total organic carbon and and seasonal precipitation. Third, insofar as beautifully detailed record of climatic vari- calcium concentrations). The age model is well MIS 11 is a good analogue for the Holocene, we ability between 370,000 and 550,000 years ago constrained by a basal, radiometrically dated should now be at a time roughly equivalent to from lake sediments at Valles Caldera, New ash layer and by palaeomagnetic reversals the transition between warm MIS 11e and cool Mexico (Fig. 1). This time interval spans two of known age, and the temporal resolution MIS 11d. If so, southwestern climates might earlier interglacial periods, known as Marine is fine enough to resolve millennial-scale naturally be trending towards a somewhat Isotope Stage (MIS) 11 and 13. climate variations. cooler and wetter stage — except that other Among palaeoclimatic aficionados, MIS 11 The Valles Caldera record reveals three factors (us) are affecting climate. Knowing the is of great interest, because Earth’s orbital con- warm and two cool stages within MIS 11, likely natural trends thus helps us to diagnose figuration, and hence the insolation regime with a 2 °C difference between stages. Warm the causes of twenty-first-century hydro- (amount of solar radiation reaching Earth), peaks probably correspond to peaks in logical trends. And perhaps, just perhaps, closely resembles that during the Holocene8. summer insolation. these natural trends will partially mitigate Usually, interglacial periods are short, lasting The first warm stage (MIS 11e) is the the projected drying in the southwest. ■ half of one precessional cycle (about 10,000 counterpart to the past 11,000 years, and its years), and end when summer cooling in the trajectory is documented in exquisite detail. John (Jack) Williams is in the Department of Northern Hemisphere triggers a new round MIS 11e began with a rapid 8 °C warming, Geography and Center for Climatic Research, of continental glaciations. However, during increased abundances of vegetation such University of Wisconsin-Madison, Madison, MIS 11, Earth’s orbit was nearly circular for as oak and juniper that prefer warm condi- Wisconsin 53706-1695, USA. an unusually long time, dampening the effect tions, and increased lake productivity. A few e-mail: [email protected] of precessional variations on seasonal insola- thousand years later, abundances of grasses 1. Fawcett, P. J. et al. Nature 470, 518–521 (2011). tion and causing MIS 11 to last 50,000 years. with the C4 photosynthetic system increased, 2. Intergovernmental Panel on Climate Change (MIS 13 is comparably long and hence also suggesting warm and at least seasonally wet Climate Change 2007: The Physical Science Basis of interest.) The similar orbital configura- conditions. Simultaneously, however, the (Cambridge Univ. Press, 2007). 3. Seager, R. & Vecchi, G. A. Proc. Natl Acad. Sci. USA tion today raises the fascinating possibility Valles Caldera lake became hydrologically 107, 21277–21282 (2010). that, even without rising greenhouse gases, closed (there were no outflowing streams), a 4. Cook, E. R. et al. in Abrupt Climate Change. A Report the current interglacial might have another signal of negative water balance. This some- by the U.S. Climate Change Science Program and 40,000 years to go9. Thus, palaeoclimatic what paradoxical result can be explained the Subcommittee on Global Change Research (eds Clark, P. U. et al.) 143–257 (US Geol. Surv., 2008). records from MIS 11 offer unique insights by differing seasonal precipitation signals, 5. Schubert, S. D., Suarez, M. J., Pegion, P. J., Koster, into the potential trajectory for the con- with the C4 grasses responding to continued R. D. & Bacmeister, J. T. Science 303, 1855–1859 temporary climate system, in the absence of summer precipitation and lake hydrology (2004). humanity’s effects. responding to reduced winter precipita- 6. Cook, E. R., Seager, R., Cane, M. A. & Stahle, D. W. Earth Sci. Rev. 81, 93–134 (2007). Unfortunately, most MIS 11 records are tion. Summer and winter precipitation 7. Thompson, R. S., Whitlock, C., Bartlein, P. J., from Antarctic ice cores or marine sediments; also diverged during the early Holocene, Harrison, S. P. & Spaulding, W. G. in Global Climates terrestrial records are scarce. The Valles Cal- probably due to the effects of insolation on Since the Last Glacial Maximum (eds Wright, H. E. Jr 10 et al.) 468–513 (Univ. Minnesota Press, 1993). dera record is remarkable for the wealth of southwestern monsoon intensity . 8. Loutre, M. F. & Berger, A. Global Planet. Change 36, palaeoclimatic proxy data collected by Fawcett The next phase within MIS 11e is the most 209–217 (2003). 1 13 and colleagues . The proxies include both new striking: abrupt shifts in δ C and lowered 9. Berger, A. & Loutre, M. F. Science 297, 1287–1288 geochemical indicators of past temperatures total organic carbon suggest a rapid collapse (2002). 10. Harrison, S. P. et al. Clim. Dyn. 20, 663–688 (2003). and tried-and-true proxies for terrestrial of C4 grasses and oxidation of lake sediments; 13 13 12 11. Siegenthaler, U. et al. Science 310, 1313–1317 vegetation (pollen and δ C [ C/ C]), lake mud cracks in the sediments show that the (2005). productivity (silica/titanium ratios) and lake dried out. This arid episode lasted several 12. Spahni, R. et al. Science 310, 1317–1321 (2005).