The abrupt climate shifts identified in Greenland ice cores transformed understanding of the climate system. Although primarily studied in the paleoclimate record, abrupt climate change induced by greenhouse gas rise poses a serious threat to modern humans and ecosystems. We present the first ultra-high-resolution view (hundreds of samples per year) of the abrupt onset (within 1 year) of the current interglacial (warm) climate retrieved from the Greenland Ice Sheet Project Two (GISP2) ice core archive. This abrupt onset is manifested by a marked reduction in storm event frequency and increase in the length of the summer season around Greenland. We apply this metric to the current rapid climatic amelioration in the Arctic as a precursor for future abrupt climate change events.
In this paper we focus on the onset of the current interglacial (a warm period relative to glacial–interglacial cycles), the Holocene, because it is during this period of relatively favorable climate that civilization emerged and has since evolved significant sensitivity to local- and regional-scale changes in water availability, storms and heat. Furthermore, the beginning of the Holocene offers an important past analog for the onset of modern warming. Previous studies demonstrate that the inception of Holocene climate is characterized by an abrupt shift in major features of Northern Hemisphere atmospheric circulation, including a northward shift in the jet stream (the edge of the polar vortex) (Mayewski et al., 1997) and the Intertropical Convergence Zone (ITCZ; Steffensen et al., 2008) and an abrupt decrease in North Atlantic sea ice extent (Mayewski et al., 1994; Cabedo-Sanz et al., 2013; Pearce et al., 2013). These changes are in turn associated with increased moisture availability to Greenland (Alley et al., 1993).
Since the discovery of abrupt climate change, advances in our mechanistic understanding of these events have been greatly enhanced by developing proxies for past behavior of the atmosphere and by more detailed sampling resolution. Past changes in atmospheric circulation derived from chemically fingerprinted and instrumentally calibrated Greenland Ice Sheet Project Two (GISP2) ice core calcium, sodium, potassium, chloride and ammonium records provide proxy reconstructions for the relative size and strength of the polar vortex, westerlies, relative strength of zonal compared with meridional airflow, North Atlantic sea ice extent, and the Icelandic Low and Siberian High (Mayewski et al., 1993, 1994, 1997; Meeker et al., 1997; Meeker and Mayewski, 2002). A recent example of the benefit of finer sampling is an annually layered Greenland ice core sequence sectioned using a microtome that reveals changes in atmospheric circulation during an abrupt warming ∼38 000 NGRIP a BP (Thomas et al., 2009). In this study (2–20-mm resolution) changes in calcium and sodium atmospheric circulation proxies preceded changes in snow accumulation, followed by a change in temperature using stable water isotopes. A second example of the advantage gained by high-resolution sampling is the demonstration, based on continuous melt sampling techniques (2.5–5.0-cm resolution), that the onset of Holocene climate occurred over 1–3 years and that the abrupt shift in atmospheric circulation at this transition preceded a change in temperature (Steffensen et al., 2008). Here, we use previously developed ice core climate proxies, expand upon the concept of finer scale sampling of calcium and sodium with the addition of iron, and focus on the abrupt climate change precursor – a change in atmospheric circulation.
Materials and methods
We use our newly established W. M. Keck Laser Ice Facility LA-ICP-MS (laser ablation − inductively coupled plasma − mass spectrometer) trace element analytical capability to present the first ultra-high-resolution (20 µm) view of the abrupt transition from the end of the Younger Dryas (YD, a near return to glacial conditions during the last deglaciation) to Holocene climate (Fig. 1a). Ice core data presented here include both previously published GISP2 soluble ion analyses at 10-cm sampling resolution (Mayewski et al., 1997; Meeker et al., 1997) and new analyses derived from archived GISP2 ice using LA-ICP-MS at 20-µm sampling resolution. For this new technique we: (i) employed a new cryocell and automated laser positioning; (2) tested replicates for variability in multiple passes and laser penetration depth; (3) calibrated with previous GISP2 and other ice core indicators for annual layer identification, including electrical conductivity, particle concentration and visual stratigraphy; (4) calibrated between laser intensity and chemical concentration; and (5) expanded previous LA-ICP-MS analytical results (Reinhardt et al., 2001, 2003; Müller and Rasmussen, 2011) yielding continuous analysis of up to 1-m-long sections of ice. For this study calcium, sodium and iron were each run on separate, but parallel ablation lines to allow use of freshly cleaned, contamination-free surfaces. Examination of the high-magnification video stream accompanying our LA-ICP-MS runs demonstrates that concentration changes are not coincident with the occurrence of individual large particles, ice crystal grain boundaries, ice crystal triple points or cracks in the ice core, and therefore changes in concentration represent deposition events that at the sampling frequency used here are potentially equivalent in timing to storm event depositional features. Sampling pathways were programmed and viewed using New Wave Research™ positioning software with micrometer scale control.
Discussion and Results
We focus on a section of the GISP2 ice core archive covering the depth range 1677.575–1678.580 m (11 643–11 675 GISP2 a BP (Meese et al., 1997)) spanning the well-established (Steffensen et al., 2008) abrupt YD/Holocene transition. Note that more recent work has placed the YD/Holocene transition at 11 700/ ± 50 NGRIP a BP (Rasmussen et al. (2006), but discussion of this is not within the scope of this paper. Figure 1(b) shows a portion of the original GISP2 calcium and sodium data focused on the original 10-cm sampled data (Mayewski et al., 1997) covering the YD/Holocene transition era. Laser sampling of the transition core section (Fig. 2) yielded 25 739 individual sample levels. For comparison, the entire ∼110 000-year (3053.44 m deep) original GISP2 soluble ion record was based on 16 395 individual sample levels. For each lasered sample the following were measured: total calcium (a ubiquitous terrestrial dust indicator associated with zonal (westerlies) atmospheric circulation (Mayewski et al., 1997), largely transported to Greenland as solubilized carbonate and/or gypsum (Laj et al., 1997) and as in association with potassium also with easterly transport through, for example, the Siberian High (Meeker and Mayewski, 2002)); total sodium (a marine source indicator largely transported by the Icelandic Low to Greenland as soluble sodium (Legrand and Mayewski, 1997; Mayewski et al., 1997; Meeker and Mayewski, 2002)); and total iron (a crustal dust indicator formed by dissolution of iron-rich particles within the atmosphere (Dedick and Hoffmann, 1992)). Comparison of the original GISP2 soluble calcium and sodium produced by ion chromatography to total (soluble and insoluble) derived LA-ICP-MS concentrations demonstrates similar values (Fig. 2). Breaks in the record are a result of either missing ice (the GISP2 archive is now 20 years old) or cracked ice that does not afford the air-tight seal necessary to fully capture the argon transport gas plasma generated by the laser sampling technique.
We capture, for the first time, ‘storm-scale’ detail (500 samples cm−1) allowing comparison of chemical concentration, frequency of chemical events used as proxies for atmospheric circulation, relative phasing of proxies for atmospheric circulation, and length of season before and following the YD/Holocene abrupt transition. Figure 2 shows the marked difference in concentration between pre- and post-Holocene onset evident in calcium and sodium noted in previous work (Mayewski et al., 1993; Steffensen et al., 2008). Iron concentrations are more similar pre- and post-Holocene onset, suggesting a difference in source region between calcium and iron crustal source indicators. Figure 2 also shows the marked reduction in variability of chemical concentration previously interpreted as a decrease in multi-annual to decadal-scale averaged storminess from the YD to the Holocene (Mayewski et al., 1993, 1997) that can now be investigated at storm event scale in our study.
Previous work demonstrates that peaks in calcium and sodium occur during winter/spring, consistent with increased atmospheric storminess and increased transport of terrestrial dusts and seasalt to Greenland at this time (Whitlow et al., 1992). Therefore, lower background concentrations of calcium and sodium over Greenland indicate less stormy periods, as expected during summer. Annual layering (gray rectangles Figs 2 and 3) is well preserved and defined by in-phase seasonal maxima in calcium, sodium and iron. Figure 3 (only smoothed data are shown) details two 5-cm core sections noted in Fig. 2. Figure 3(b) contains the last 3 years of the YD (1678.19–1678.22 m depth), the YD/Holocene transition year (1678.18–1678.19 m depth) as inferred by the decrease in concentration in calcium, sodium and iron over this depth, and the first year of the Holocene starting at 1678.18 m depth. Since the GISP2 archive no longer contains the depth range 1677.99–1678.17 m, due to previous sampling of this section, we cannot assess whether there is another transition step in the missing 18 cm. Figure 3(a) spans just <2 years of the early stages of the Holocene (1677.75–1677.80 m depth). The number of annual layers identified in Fig. 3(a) for the Holocene section is close to 3 cm a−1, consistent with the average of ∼30 a m−1 calculated in this depth range from the original GISP2 record (Meese et al., 1997). The annual layer thickness in the YD section of Fig. 3(b) (1678.18–1678.22 m depth) is closer to 1–2 cm a−1, indicating lower precipitation during the YD. The Holocene sections in Figs 2 and 3 have annual layering in chemistry clearly marked by longer periods of low values than those in the YD. Therefore, the decrease in concentration of calcium and sodium noted in previous lower resolution studies (Mayewski et al., 1997; Steffensen et al., 2008) and the accompanying increase in precipitation (Alley et al., 1993; Meese et al., 1997) from the YD to the Holocene is not just a consequence of a decrease in transport strength during winter/spring and an increase in moisture availability, but also a consequence of longer periods of less stormy, milder conditions in and around Greenland, as would be expected if the summer season lengthened with Holocene onset. Comparison of Fig. 3(a) and Fig. 3(b) potentially suggests at least a doubling in the length of the summer season at the onset of Holocene warming. A longer summer season is consistent with the abrupt decrease in sea ice extent surrounding Greenland suggested for the YD/Holocene transition (Mayewski et al., 1994; Steffensen et al., 2008).
Closer examination of Fig. 3(b) shows that the YD/Holocene transition is defined at the depth over which levels of calcium, sodium and iron decline markedly (1678.18–1678.19 m). This depth range spans approximately 1 year and includes a significantly reduced winter/spring peak in westerly winds (calcium) compared with the previous 3 years (1678.19–1678.22 m). During the 3 years leading up to the transition, westerly wind peak values rise progressively. Westerly winds then weaken to Holocene (Fig. 3a) values at the transition. Marine storm intrusions (sodium) are variable in the 3 years before the YD/Holocene transition as westerly winds strengthen, potentially blocking transport of marine air onto Greenland. However, marine air mass intrusion is greatest during the transition year, suggesting intensified circulation over marine areas and closer proximity of marine air to Greenland as sea ice extent is reduced and westerly wind strength declines. Marine air mass incursions are less intense following the transition into the Holocene, as evidenced by lesser concentration peaks in sodium and longer periods of less intense transport (near flat sodium concentrations) during the longer summer seasons characterizing the onset of the Holocene. Iron concentrations drop at the beginning of the transition year and early Holocene (1678.18 m and shallower depth) iron transport to Greenland is characterized by longer periods of weak transport from continental sources. Reduction in continental source calcium and iron and marine source sodium is consistent with northward migration and weakening of the westerlies and coupled easterlies associated with warming into the Holocene as the north–south thermal gradient weakened.
The YD/Holocene Analog
While Holocene age abrupt climate change events are less pronounced than glacial age and YD/Holocene abrupt transitions, they still have significant relevance to humans and ecosystems (Mayewski et al., 2004). Today, with carbon dioxide levels rising faster than at any time over the last several hundred thousand years (EPICA Community, 2004), and with the addition of other human source radiative forcing related to sulfates, soot carbon and ozone-destroying agents, there is increased likelihood for non-linear climate responses at local to regional scales. To investigate this we use our YD/Holocene analog for change in season length as an indicator of current or impending abrupt climate change over the Northern Hemisphere with specific focus on the Arctic, where summer sea ice extent and volume has shrunk at unprecedented rates in recent years. Examination of the period AD 2007–2012 minus AD 1979–2000 for mean annual temperature and summer sea ice extent reveals significant change. For example, changes in mean annual 2-m temperature are as much as 5 °C over the eastern Arctic (Fig. 4a) accompanying massive decreases in summer sea ice extent. This rise in mean annual temperature exceeds proxy-based reconstructions of even the most recent naturally forced change in climate, the transition from the Medieval Warm Period to the Little Ice Age. The latter is marked by an abrupt shift in atmospheric circulation at 1400–1420 CE (O'Brien et al., 1995) followed by a decrease in temperature of approximately 1 °C over the North Atlantic region (Bradley and Jones, 1993; Mayewski and Maasch, 2006).
To test length of season as an indicator for abrupt climate change we examine the percentage change in melting degrees (accumulated degrees >0 °C) using ECMWF ERA-Interim reanalysis 2-m air temperature for AD 2007–2012 minus AD 1979–2000. The length of the melting season increased 60–100% between the analysis periods over significant areas of the Arctic, with the most dramatic impacts in the coastal eastern Arctic, the Canadian Arctic and Greenland (Fig. 4b). Directly upwind from GISP2 on the west coast of Greenland temperature increased close to 3°C and annual melting degree days increased close to 100% for the comparison period. These changes are similar to the approximate doubling in length of the summer season at the YD/Holocene transition seen in our study.
Implications for the future
Recent Arctic summer sea ice impact is the first in a series of regional-scale abrupt climate changes that will probably continue in regions most sensitive to warming and eventually propagate regionally. These abrupt changes need not always be manifested by length of the melt season, but also in the duration of drought/flood and storminess (e.g. as is currently the case in Saharan Africa.) Our ultra-high-resolution ice core examination of the classic abrupt onset of Holocene warming sheds light on the structure of past abrupt climate changes, and also provides unparalleled perspective with which to assess the potential for near-term rapid shifts in atmospheric circulation and seasonality. As an example, our study demonstrates that the recent warming in the Arctic is equivalent to the warming, inferred by the relative length of the summer season, during the YD/Holocene transition, although thus far being more localized in the modern case. Differences in season length before and after abrupt changes in past climate offer an analog for assessing abrupt change in atmospheric temperature, precipitation and atmospheric circulation. Global circulation models failed to predict the rapidity and scale of decay in Arctic summer sea ice currently in progress. Future investigations into the relative length of seasons, as marked by past and modern physical, chemical and biological parameters, will probably yield details of behavior, such as trends, frequency of outliers and changes in magnitude, that could be precursors for abrupt climate shifts. Identification and use of such precursors will fill an essential void in existing climate prediction models by testing for sensitivity in the context of past analogs.
Support was provided by the W. M. Keck Foundation (to P.M.) and the National Science Foundation Office of Polar Programs (grant No. 1203640 to P.M. and A.K.). We also thank Climate Change Institute members Michael Handley, Tom Beers and Skylar Haines and the US National Ice Core Laboratory. Data used in this paper are available at: http://data-portal.ecmwf.int/, http://nsidc.org/data/submit.html and http://www.icereader.org/icereader/. P.A.M. conducted fieldwork and sample collection. S.B.S. obtained laser data. S.D.B., A.V.K. and K.A.M. provided code and modeling. S.B.S. and P.A.M. compiled the data. P.A.M. wrote the paper. All authors contributed to the final manuscript text and figures.
Greenland Ice Sheet Project Two